ABSTRACT
The segment-polarity class of segmentation genes in Drosophila are primarily involved in the specification of sub-segmental units. In addition, some of the segmentpolarity genes have been shown to specify cell fates within the central nervous system. One of these loci, gooseberry, consists of two divergently transcribed genes, gooseberry and gooseberry neuro, which share a paired box as well as a paired-type homeobox. Here, the expression patterns of the two gooseberry gene products are described in detail. The gooseberry protein appears in a characteristic segment-polarity pattern of stripes at gastrulation and persists until head involution. It is initially restricted to the ectodermal and neuroectodermal germ layer, but is later detected in mesodermal and neuronal cells as well. The gooseberry neuro protein first appears during germ band extension in cells of the central nervous system and also, much later, in epidermal stripes and in a small number of muscle cells. P-elementmediated transformation with the gooseberry gene has been used to demonstrate that gooseberry transactivates gooseberry neuro and is sufficient to rescue the gooseberry cuticular phenotype in the absence of gooseberry neuro.
INTRODUCTION
During early Drosophila embryogenesis, a number of developmental programs unfold, including segmentation, the generation of the germ layers, and neurogenesis. While segmentation may be viewed as a process beginning with the specification of position along the anteroposterior axis of the embryo (reviewed by Akam, 1987; Ingham, 1988), the germ layers are established by the division of the embryo, along the dorsoventral axis, into longitudinal regions (Hartenstein et al., 1985; Mayer and Nüsslein-Volhard, 1988; reviewed by Govind and Steward, 1991). Later, during germ band extension, the process of neurogenesis begins. The founder cells of the central nervous system (CNS), the neuroblasts, begin to delaminate from the neuroectoderm and migrate inwards (Hartenstein and Campos-Ortega, 1984; Hartenstein et al., 1987). Most neuroblasts divide asymmetrically several times to generate a string of progeny known as ganglion mother cells. Each ganglion mother cell then divides once symmetrically to generate a pair of sibling neurons. Thus, the approximately twenty neuroblasts per hemisegment that leave the ectoderm give rise to about 250 neurons. During germ band retraction, the CNS continues to differentiate and neurons send out their axons (Goodman et al., 1984).
From a genetic and molecular analysis of these early developmental events, it has become increasingly clear that they are directed by relatively small groups of genes, which interact with each other in complex hierarchical regulatory networks. For example, the segmentation genes direct the proper establishment of the metameric organization of the embryonic body plan. The hierarchical activation of three classes of segmentation genes the gap, pair-rule, and segment-polarity genes defines position along the anteroposterior axis in progressively smaller units (Nüsslein-Volhard and Wieschaus, 1980). Similarly, a set of hierarchically acting genes has been described, which controls dorsoventral patterning and thus determines the anlagen of the germ band. Subsequently, the proneural and neurogenic gene sets specify which cells become neuroblasts and which remain on the surface of the embryo and become epidermal cells (reviewed by Campos-Ortega and Knust, 1990).
Interestingly, some of the genes involved in segmentation are redeployed in other developmental processes such as neurogenesis (reviewed by Doe and Scott, 1988). For example, many of the segment-polarity genes are expressed in the neuroectodermal region at the onset of neurogenesis and are also expressed later by subsets of neurons. Detailed analysis of neural development in segment-polarity mutants suggests that certain segment-polarity genes indeed play a specific role in neurogenesis (Patel et al., 1989a).
One of the segment-polarity loci, gooseberry, is unique for several reasons. It encodes two transcripts that share extensive sequence homology with each other and with the pair-rule gene paired (prd). The homologous regions comprise two domains, the paired-domain and the prd-type homeodomain (Bopp et al., 1986). Moreover, the 5′ ends of the two gooseberry transcription units face each other, being separated by about 10 kb (Baumgartner et al., 1987; Li et al., 1993), raising the intriguing possibility that both transcripts share common cis-regulatory elements. Finally, two independent mutagenesis screens failed to produce point mutations for either of the two gooseberry genes (Nüsslein-Volhard et al., 1984; Côté et al., 1987). All known gooseberry mutants are the result of deficiencies. If indeed point mutations of the gooseberry locus cannot be obtained, an explanation might be that either of the two gooseberry products functionally substitutes for the other. Here we report, in detail, the developmental expression of the two gooseberry genes, gooseberry (gsb; previously called gsb-BSH9 or gsb-d) and gooseberry neuro (gsbn; the former gsb-BSH4 or gsb-p). In addition, we demonstrate that gsb is sufficient to rescue fully the gooseberry cuticular phenotype and that gsb activates gsbn in trans.
MATERIALS AND METHODS
Construction of expression and rescue plasmids
Plasmids expressing gsb or gsbn protein, pAR-gsb.fl and pARgsb-neuro.fl, in bacteria were constructed as follows. To obtain pAR-gsb.fl, an EcoRV-EcoRI gsb-cDNA fragment of BSH9c2 (Baumgartner et al., 1987) was subcloned with blunt ends into the BamHI site of the bacterial expression vector pAR3039 (Studier and Moffat, 1986). Since the EcoRV site of BSH9c2 is 40 bp downstream of the gsb start codon, the bacterially expressed gsb protein lacks the 15 N-terminal amino acids of the full-length gsb protein (427 amino acids). To obtain pAR-gsb-neuro.fl, a NcoINsiI fragment of the gsbn-cDNA BSH4c4 (Baumgartner et al., 1987) was subcloned with blunt ends into the BamHI site of the bacterial expression vector pAR3040. As the NcoI site in BSH4c4 contains the start codon of the gsbn protein, the bacterially expressed protein contains the full-length gsbn protein (452 amino acids).
The P-element plasmid containing the gsb gene, gsb-pKSpL2, was constructed by subcloning a 20 kb genomic fragment of the gsb region (Fig. 6B), obtained from a partial EcoRI digest of the genomic clone P920 (in EMBL 4), into pKSpL2. The vector pKSpL2 was constructed as follows. The NotI site of Bluescript pKS+ was destroyed by filling in the cleaved ends with Klenow enzyme and subsequent religation, a short stretch of the polylinker between HindIII and XhoI was removed (ligation of the filled up sites restores the HindIII site), and a NotI site was introduced into the cleaved EcoRV site of the polylinker by blunt end ligation of (GCGGCCGC). The newly created polylinker was confirmed by sequencing. The final gsb rescue plasmid, BSH9-16.18, was constructed in two steps. First, a 17 kb XbaI-NotI fragment of gsbpKSpL2 was subcloned into cp20.2, which had been constructed by removing the KpnI-SalI lacZ fragment from HZ50pL (Hiromi et al., 1985), and second, the 3.1 kb XbaI fragment of gsb-pKSpL2 was inserted to generate BSH9-16.18.
Preparation of purified antisera and immunocytochemistry on whole-mount embryos
Rabbit antisera were generated and purified essentially as described previously for the anti-prd antiserum (Gutjahr et al., 1993) with the following modifications. Both antisera were directed against the full-length proteins and cross-reacted on western blots with bacterially expressed gsb and prd proteins. The antigsb (anti-gsbn) antiserum was depleted of such cross-reactive antibodies by passing it over a column to which a crude bacterial extract containing gsbn (gsb) protein had been bound. Subsequently, the antisera were further affinity-purified (positive adsorption) as described previously for the anti-prd antiserum (Gutjahr et al., 1993). The specificity of both antisera was confirmed by staining embryos homozygous for the deficiency Df(2R)IIX62, which removes both gsb genes. In these embryos, no staining was observed using either antiserum (not shown). The specificity of the anti-gsb antiserum was further corroborated by staining embryos homozygous for the deficiency Df(2R)KrSB1, which removes only the gsb gene (Bopp et al., 1986; Côté et al., 1987). In such embryos, gsbn was expressed at high levels in the head region but only at extremely low levels in very few cells of the CNS whereas no staining was detected using the gsb antiserum. Finally, the specificity of the gsbn antiserum is inferred from the fact that during gastrulation, when gsb is expressed at high levels, no staining was seen with the anti-gsbn antiserum. Staining of fixed embryos with 100-fold diluted anti-gsb and antigsbn antisera and photography on a Zeiss Axiophot with Nomarski optics (unless otherwise indicated) were as described (Patel et al., 1989b; Gutjahr et al., 1993).
Embryos of a wingless lacZ-enhancer trap line, 17en40/CyO (kindly provided by Norbert Perrimon), were stained with mouse anti-β-galactosidase (Promega) and either anti-gsb or anti-en antiserum (mAb 4D9; Patel et al., 1989b) to determine the relative positions of the wingless (wg), engrailed (en), and gsb domains in the ectoderm and in neuroblasts. The relationship of wg and en domains in the neuroblast map (Fig. 4) was also checked by examining embryos that had been both hybridized in situ with a digoxigenin-labeled wg probe and immunostained for en expression and were kindly provided by Armen Manoukian. The relative positions of en and gsb or gsbn protein were also determined by double labeling embryos with mAb 4D9 and anti-gsb or anti-gsbn antiserum.
Neuroblast patterns were initially sketched by hand and are schematically illustrated in Fig. 4. The particular stages illustrated were chosen because they are easily recognizable. The stage in Fig. 4A is characterized by the appearance of the first row 6 neuroblast, and the stage shown in Fig. 4B by the appearance of the extremely medially located row 5 neuroblast. We note that the general neuroblast patterns shown in Fig. 4 closely match those drawn by Doe (1992). In some cases, however, we are not sure of a one-to-one correspondence between specific neuroblasts in our map and the numbered neuroblasts of Doe’s map. Therefore, we have not attempted to use Doe’s numbering system at present. Future double-labeling experiments using the additional neuroblast markers described by Doe (1992) should allow a precise integration of the two maps.
Transgenic fly stocks
To generate transgenic flies, the gsb rescue plasmid was injected into ry506/ry506 embryos. Four independent transformed lines were obtained from which the following genotypes were generated: Df(2R)KrSB1/CyO; P[ry+; gsb+] and Df(2R)IIX62/CyO; P[ry+; gsb+]. Eggs of these lines or from crosses of these lines with each other were collected and either fixed for antibody staining or allowed to age for 48 hours for cuticle preparation.
RESULTS
Both gsb and gsbn proteins are localized in the nucleus, as is the case for other paired- and homeodomain containing proteins, and are expressed in a typical, segmentally reiterated, segment-polarity pattern (Figs 1, 2). Although gsb and gsbn expression overlap, the two proteins predominate at different stages of development and in different tissues.
Expression of gooseberry
The gsb protein is initially expressed in a segmentally reiterated pattern of stripes with a pair-rule modulation of intensity. The first set of stripes is detectable at the end of cellularization and includes the odd-numbered stripes 1-13 plus an anterior stripe 0 (Fig. 1A) that probably corresponds to the antennal segment of the head (Jürgens et al., 1986). The antennal stripe and stripe 1 are the first to appear, followed after a short delay by stripes 3, 13, 7, 11, and finally 5 and 9. At mid gastrulation, the even-numbered stripes 212 emerge simultaneously (Fig. 1B). Stripe 14 appears at the onset of germ band extension (Fig. 1C,D). All the stripes quickly reach equal levels during the rapid phase of germ band extension (Fig. 1D). At the same time, the shape of the stripes changes, acquiring a distinct triangular appearance.
Towards the end of germ band extension, stripes 4-14 become laterally restricted to the neuroectodermal portion of the ectoderm (Fig. 1E,F). At this stage, gsb protein reaches its highest levels and is detectable in the maximum number of segments, including 14 body stripes, 4 regions anterior to the mandibular segment, and 3 regions posterior to the eighth abdominal segment (Fig. 1F,G). The gsb stripes assume a barbell shape as the more medial areas of expression narrow. This exposes a pair of distinct gsb-expressing cells which are located close to the mesectodermal region (Fig. 3G) and are probably the most medial neuroblasts of row 5 (Fig. 4B). Subsequently, gsb protein levels begin to decrease in the head and later in the trunk segments (Figs 1G,H, 3D), and gsb stripes broaden at the end of the extended germ band stage and during germ band retraction (Fig. 1H,I). Towards the end of germ band retraction, gsb expression increases again in the ectoderm (Figs 1L, 3E) albeit to levels lower than the preceding peak of ectodermal gsb expression. This low ectodermal gsb expression (stage 13) is no longer detectable by the time of head involution (stage 14).
During germ band extension (stage 9), neuroblasts begin to delaminate and the gsb ectodermal stripes narrow (compare widths of stripes in Fig. 1D,E). Those gsb-expressing ectodermal cells that become neuroblasts maintain gsb expression. Eventually, all neuroblasts of row 5 and 6 express gsb, and transient gsb expression is also seen in the most medial neuroblast of row 7 (Fig. 4B). In addition, gsb appears to be weakly and transiently expressed by three midline cells directly anterior to the median neuroblast (Fig. 4B). Based on their position, they may be the precursors to the VUM neurons (Klämbt et al., 1991). Expression of gsb persists at low levels in a few neuroblasts and ganglion mother cells until germ band retraction (Fig. 3D,E). Very low levels of gsb protein also remain detectable until head involution in large cells at the extreme ventral surface of the CNS, which may be the remnants of the embryonic neuroblasts (Fig. 1M,N).
It should be noted that during its initial expression, gsb protein is mostly excluded from the mesoderm, the mesectodermal cells and the region generating the amnioserosa (Fig. 1A,B). At mid germ band extension, gsb protein appears in the mesoderm (Fig. 3A,B) where it seems to persist until the end of germ band retraction (Fig. 3C-E) when it is most prominent in three patches of mesodermal cells in the thoracic segments (Fig. 1K). Later these patches appear to merge into a single patch, before gsb expression disappears from the thoracic mesoderm during head involution.
The patterns of gsb expression in the tail and, particularly, in the head region are more complex. At mid germ band extension, the antennal stripe divides into two independent regions, and a novel stripe, intercalated between the antennal stripe 0 and the mandibular stripe 1 and corresponding to the anlagen of the intercalary segment (Jürgens et al., 1986), begins to express gsb protein (Fig. 1D,E). During the slow phase of germ band extension, a bilaterally symmetric pair of patches expressing gsb emerges in the dorsal region of the clypeolabrum, and a small number of cells express gsb in the non-segmented pre-antennal region of the head (Fig. 1E,F). Expression in the posteriormost abdominal region starts with the appearance of stripe 15 in A9 at the end of the rapid phase of germ band extension (Fig. 1E). Subsequently, gsb expression begins in ‘stripe’ 17 (Fig. 1F) as a pair of bilateral patches of cells in the central region of the anal pads (A11; Jürgens, 1987). Finally, ‘stripe’ 16 emerges as a very narrow string of cells that initially abuts stripe 15, but separates from it during the extended germ band stage (Fig. 1F,G). The late appearance and reduced size of stripes 15–17 reflect the rudimentary nature of the terminal abdominal segment anlagen, A9A11, in Drosophila (Baumgartner et al., 1987; Jürgens, 1987). During head involution, gsb is transiently expressed at high levels in a subset of cells of the pharynx and anal pads (Fig. 1M).
Expression of gooseberry neuro
As shown in Fig. 2A, gsbn protein first appears at stage 10 in a small number of neuroblasts, ganglion mother cells, and neurons. As neurogenesis proceeds, gsbn protein levels rise, an increasing number of ganglion mother cells and neurons express gsbn, and a low level of gsbn protein persists in some neuroblasts (Figs 2A-E, 3H,I). The ganglion mother cells and neurons that express gsbn are predominantly, though perhaps not exclusively, the progeny of the gsb-expressing neuroblasts. By the end of stage 11, gsbn is clearly expressed in a segmentally reiterated neural pattern from the mandibular to the tenth abdominal neuromere. In the trunk segments, gsbn expression in the CNS forms a typical L-shaped pattern in each hemisegment (Fig. 2C,D,I). Similarly to gsb, gsbn is expressed in the terminal regions. In the head, gsbn protein is detected in neurons of the brain (Fig. 2F,G) while, in the tail region, it appears in neurons of A9 and in cells of the anal pad (Fig. 2C,D). Finally, gsbn is also expressed in a small number of neurons lying between A9 and the anal pads. These neurons are derived from ‘stripe’ 16 of gsb and may be evidence for a rudimentary tenth abdominal neuromere (Fig. 2C,D). During subsequent stages of development, gsbn protein persists in a subset of neurons until nerve cord retraction during stage 17 (Fig. 2F,G,I,L). Furthermore, after germ band retraction gsbn becomes expressed in a few lateral cells per hemisegment that might belong to the muscle founder cells or the PNS (arrows in Fig. 2I), as well as in a striking stripe of mesodermal cells of T2 (Fig. 2F,H). Similarly to gsb, after germ band retraction gsbn is also expressed in ventral ectodermal stripes in the posterior region of each segment (Fig. 2K). However, the gsbn stripes persist until much later in development (up to stage 17; Fig. 2O) than the gsb stripes. In addition, gsbn protein is detectable in a number of patches of epidermal cells, or derivatives thereof, in the head region, including the pharynx (Fig. 2L,M). Finally, gsbn protein appears in the nuclei of one of the ventral superficial oblique muscles (Fig. 2N,O).
Coexpression of gooseberry and gooseberry neuro with engrailed
To determine the relative positions of gsb and gsbn-expressing cells with respect to the parasegmental and the segmental boundaries, double-labeling experiments were performed with an anti-gsb or anti-gsbn antiserum and an anti-en monoclonal antibody (Patel et al., 1989b). As is apparent from Fig. 5A-C, the anterior border of a gsb stripe is anterior to that of en-expressing cells by one to two rows of cells at the extended germ band stage. Similarly, the posterior boundary of en is one to three cells posterior to that of gsb expression. The greatest extent of overlap between gsb and en protein is seen in the widest and most lateral regions of the gsb stripes. Furthermore, gsb protein is expressed by all neuroblasts of rows 5 and 6, and transiently by the most medial neuroblast of row 7, while en is expressed by all neuroblasts of rows 6 and 7 (Figs 4B, 5C). Extensive overlap of gsbn and en is seen in the CNS. During the early extended germ band stage, en is expressed by a large number of ganglion mother cells and neurons derived from the neuroblasts of rows 6 and 7 as well as from the median neuroblast (Fig. 4B). At this stage, gsbn is expressed in ganglion mother cells and neurons derived from neuroblasts of row 5 and 6, and from the most medial neuroblast of row 7 as evident from the overlap between the en and gsbn expression patterns (Fig. 5D). Further details of gsb and gsbn expression in the CNS will be discussed in the context of the specific role of the gooseberry locus in neural development (Patel, Li, Gutjahr, Ferres-Marco, Noll and Goodman, unpublished data).
We further examined the overlap of gsb with wg (see Materials and methods). This analysis revealed that the wg domain coincides with the anterior gsb domain during the extended germ band stage (not shown). In the CNS, wg is expressed by the neuroblasts of row 5 (Fig. 4). Thus, in both the ventral ectoderm and the underlying neuroblasts, gsb expression includes most, if not all, of the wg domain, plus part of the anterior portion of the en domain. Since the boundary between the ectodermal en and wg domains demarcates adjacent parasegments, it follows that gsb expression spans the parasegmental boundary at the extended germ band stage.
Rescue of the gooseberry cuticular phenotype
In order to test the contributions of gsb and gsbn to the cuticular (Nüsslein-Volhard and Wieschaus, 1980) and CNS gsb-phenotypes (Patel et al., 1989a) and to detect a potential transregulation of gsbn by gsb, we generated transgenic flies carrying a 20 kb genomic DNA fragment harboring the intact gsb gene, the region separating the two gsb transcripts, and the 5′ portion of the gsbn gene comprising the paired-domain and the first two introns of the gsbn gene (Fig. 6B). Conceivably, this construct permits the expression of a functional gsb protein (see below), yet only of a truncated gsbn protein (consisting of the paireddomain fused to vector sequences). Four independent transgenic lines were crossed into gsb mutant backgrounds to test the ability of the gsb gene to rescue the gsb cuticular phenotype.
We first analyzed the cuticles of embryos transheterozygous for the deficiencies Df(2R)KrSB1 and Df(2R)IIX62 that carried one copy of the exogenous gsb gene (Df(2R)KrSB1/Df(2R)IIX62; P[ry+, gsb+]/ry506). The deficiency Df(2R)IIX62 removes both the gsb and the gsbn gene while Df(2R)KrSB1 removes gsb, but not gsbn (Fig. 6B; Bopp et al., 1986; Baumgartner et al., 1987; Côté et al., 1987). Therefore, Df(2R)KrSB1/Df(2R)IIX62; P[ry+, gsb+]/ry506 embryos carry one copy each of the endogenous gsbn and of the exogenous gsb gene. All four transgenic gsb lines tested were able to reverse the gsb cuticular phenotype (similar to the embryo shown in the central panel of Fig. 6A).
To exclude the possibility that gsbn contributes to the cuticular rescue, we also tested embryos homozygous for Df(2R)IIX62. These embryos exhibit both the zipper phenotype, characterized by defects in the head skeleton, and the gsb cuticular phenotype (right panel of Fig. 6A; Nüsslein-Volhard et al., 1984). Again, all four transgenic gsb lines were able to completely rescue the gsb cuticular phenotype of Df(2R)IIX62 homozygotes but still displayed the zipper phenotype (middle panel of Fig. 6A). Since several loci are deleted in addition to gsb in both Df(2R)KrSB1 and Df(2R)IIX62 deficiencies (Côté et al., 1987), embryonic lethality was not rescued. The rescue of the gsb cuticular phenotype does not depend on the gsbn sequences that are also present in the rescue construct and encode a truncated gsbn protein because complete rescue is also achieved by a shorter construct carrying no gsbn sequences (not shown). We conclude that the gsb gene is able to rescue fully the gsb cuticular phenotype in the absence of gsbn. Moreover, as shown below, relatively low levels of gsb protein appear to be sufficient to rescue completely the gsb cuticular phenotype.
Transactivation of gooseberry neuro by gooseberry
The observed general overlap of gsbn and gsb expression in the CNS suggests a possible activation of gsbn by gsb. To test this possibility, we examined whether gsbn protein, which is undetectable in the trunk of Df(2R)KrSB1/Df(2R)IIX62 embryos, is expressed in transgenic Df(2R)KrSB1/Df(2R)IIX62; P[ry+, gsb+]/P[ry+, gsb+] embryos carrying two exogenous gsb genes and one copy of gsbn. In such embryos, gsbn is clearly expressed in ganglion mother cells and neurons (Fig. 7A), and later in the epidermis (not shown), of the same regions as in wild-type embryos although at much lower than wild-type levels. Since no gsbn protein was observed in Df(2R)IIX62/Df(2R)IIX62; P[ry+, gsb+]/P[ry+/gsb+] embryos (not shown), expression of the transgenic truncated gsbn gene (paired domain) is undetectable with the cross-absorbed anti-gsbn antiserum (see Materials and methods), demonstrating that the gsbn protein detected in Df(2R)KrSB1/Df(2R)IIX62; P[ry+, gsb+]/P[ry+, gsb+] embryos must be attributed to the activation of the intact endogenous gsbn gene. Hence, these results demonstrate that the exogenous gsb gene is indeed able to activate gsbn expression in the transgenic transheterozygous gsb mutants. The relatively low expression of gsbn in these embryos probably reflects two different effects. First, these embryos contain only one copy of the gsbn gene. In fact, higher levels of gsbn expression were observed in homozygous Df(2R)KrSB1/Df(2R)KrSB1; P[ry+, gsb+]/P[ry+, gsb+] embryos, which carry two copies of gsbn, both in the CNS and the epidermis (not shown). Second, expression of the exogenous gsb gene is weaker than that of the endogenous gsb gene, especially in the CNS (compare Fig. 7B with Figs 7C and 1F). It should be noted that the low, but easily detectable, expression of gsbn in the pharynx and anal pads in Df(2R)KrSB1/Df(2R)IIX62 embryos (not shown; Ouellette et al., 1992) indicates that not all cell types require gsb to activate gsbn expression.
DISCUSSION
Both gsb genes, gsb and gsbn, encode transcriptional regulators whose N-terminal halves consist of a paired-domain and a prd-type homeodomain (Bopp et al., 1986). Their extreme structural conservation suggests that the function of the gsb and gsbn proteins are probably very similar at the molecular level. The difference in function between the two genes might then consist of a difference in their expression patterns rather than in their specificity of molecular action. We have shown here that gsb protein is continuously expressed in a typical segment-polarity pattern in the epidermis until head involution, transiently in the developing CNS and mesoderm, and finally in specific structures of the head and tail region. The epidermal and CNS expression of gsb in segmentally repeated stripes strikingly parallels the delayed expression of gsbn in these tissues, which suggests a possible dependence of gsbn expression on gsb.
Gooseberry functions in the specification of the cuticular pattern
Since in all presently known gsb alleles the gsb gene is deleted and expression of the neighboring gsbn gene is entirely eliminated or at least reduced to undetectable levels in most parts of the embryo, it was not clear which of the two genes is responsible for the cuticular phenotype. Their patterns of transcripts, however, suggested that gsb rather than gsbn specifies the cuticular pattern (Bopp et al., 1986; Baumgartner et al., 1987). Due to the late expression of gsbn in the epidermis (Fig. 2K), however, the possibility remained that gsbn is also involved in the specification of the cuticle. The rescue of the gsb cuticular phenotype by a gsb transgene in the absence of both gsb and gsbn demonstrates that gsb is sufficient while gsbn is dispensible for proper development of the cuticle (Fig. 6).
Similarly to other segment-polarity genes, gsb is first activated by pair-rule gene products (Baumgartner, 1988). For example, prd and odd-paired (opa) are required for the activation of gsb in oddand even-numbered stripes, respectively (Bopp et al., 1989; Li et al., 1993). Activation by prd is further reflected in the initial pair-rule pattern of gsb (Fig. 1A) which precisely parallels that of the prd protein (Gutjahr et al., 1993). In other words, the prd bands appear in the same order as and immediately precede the corresponding gsb bands, suggesting that the prd protein probably activates the gsb gene directly by binding to the corresponding gsb cis-regulatory elements (Li et al., 1993). The later ectodermal expression of gsb, accompanied most notably by the lateral restriction of the gsb stripes to the neuroectodermal region of the extended germ band, is activated and maintained in response to the wg signal (Li et al., 1993). The cuticular pattern only clearly depends on the wg product before germ band retraction (Bejsovec and Martinez-Arias, 1991), exhibiting a wg-dependent mutant phenotype very similar to that of gsb. Since only gsb but not gsbn is expressed during the temperature-sensitive period of the temperature-sensitive wg allele, it is not surprising that gsb rather than gsbn is responsible for the determination of the cuticular pattern (Fig. 6). Moreover, by the same argument the late epidermal expression of gsbn does not influence the cuticular pattern. Since this late epidermal expression of gsbn depends on gsb, which is activated by wg, the wg signal is also required for the late epidermal expression of gsbn. Hence, specification of the cuticular pattern by gsbn would also be in conflict with the observed temperature-sensitive period of the temperaturesensitive wg allele (Bejsovec and Martinez-Arias, 1991). The function of the late epidermal gsbn expression remains to be elucidated.
Gooseberry activates gooseberry neuro in trans
The observation that gsbn is not expressed in transheterozygous Df(2R)KrSB1/Df(2R)IIX62 embryos, in which one copy of the gsbn gene is retained but both copies of the gsb gene are deleted, could be explained by inactivation of the remaining gsbn gene in cis or trans. Since we could show that gsbn is expressed in transgenic embryos into which an exogenous gsb gene had been introduced, we conclude that the inactivation occurs in trans and that gsb protein is required for the activation of gsbn.
The expression patterns of both gsb and gsbn are altered in pair-rule mutants in the same manner (Bopp et al., 1989;X. Li, unpublished observations). A possible explanation would be that both genes are regulated by the same combinations of pair-rule gene products that interact with the cis-regulatory region of each gene to activate its transcription. Alternatively, one of the two gsb gene products could activate the other gene in trans. Our finding that the expression of gsbn depends on the expression of gsb favors the second alternative. In all cells and tissues expressing gsbn, expression of gsb immediately precedes that of gsbn, indicating that the transactivation of gsbn by gsb might be direct. In the CNS, for example, gsb protein appears in those neuroblasts and ganglion mother cells that subsequently express gsbn and apparently give rise to gsbn-expressing neurons. Also in the epidermis, where gsbn expression is initiated during stage 13, it is preceded by and dependent on gsb expression. However, gsbn expression does not always completely depend on gsb activity as suggested by the expression of gsbn in the pharynx and anal pads of the transheterozygous gsb embryos.
Expression of gsb does not persist in cells and tissues that continue to express gsbn, as for example in the CNS or epidermis. Therefore, gsbn expression is maintained by (a) protein(s) different from gsb. The simplest mechanism for gsbn to maintain its expression would be by autoregulation.
Role of gooseberry genes in neurogenesis
The expression of gsb and gsbn in the CNS suggests that both genes play a role in the development of the CNS. In fact, the known gsb deficiencies also exhibit a CNS phenotype in which even-skipped-expressing cell lineages are altered and the posterior commissures are missing (Patel et al., 1989a). The redeployment of segmentation genes in neurogenesis seems to be a general phenomenon as most of them are reexpressed in the developing CNS at various stages. This expression in the CNS is crucial for the proper specification of neuronal fates as demonstrated for the pairrule genes fushi tarazu, even-skipped, and runt (Doe et al., 1988a,b; Duffy et al., 1991). Our studies shown here suggest that one evident function of gsb is the activation of gsbn expression in the CNS. In addition, we have found that an exogenous copy of gsb rescues the neural defects seen in Df(2R)KrSB1/Df(2R)IIX62 embryos and that both gsb and gsbn are required for a complete rescue of all neural phenotypes (Patel, Li, Gutjahr, Ferres-Marco, Noll, and Goodman, unpublished data).
Are there no point mutants of gooseberry?
Two independent screens for gsb mutations failed to produce point mutants but generated only deletions (Nüsslein-Volhard et al., 1984; Côté et al., 1987). Hence, the question arose whether point mutations have not been obtained because both gsb genes need to be inactivated to observe the gsb cuticular phenotype. Our results argue against such an assumption for two reasons. Our demonstration that gsb is sufficient to specify the cuticle renders the gsbn gene dispensible with respect to cuticular patterning. Moreover, since we have shown that gsbn expression depends on a functional gsb protein, inactivation of the gsb product by point mutations is expected to inactivate both genes. Therefore, we expect that it should be possible to generate point mutations in the gsb gene that result in a cuticular phenotype.
Is there an ancestral gooseberry gene?
The organization of the gsb locus and the sequence homology between the two genes suggest that the two gsb genes have originated from a common ancestral gene through gene duplication. If this interpretation is correct, the question arises whether the two genes of the gsb locus exert specialized and separate functions which were previously the task of a single gene. It may thus be possible to isolate the gsb gene from a more distantly related insect or arthropod in which only one gsb gene exists which performs both functions in segmentation and neurogenesis.
ACKNOWLEDGEMENTS
We would like to thank Fritz Ochsenbein for his patience and help in the photographic artwork. We are grateful to Leslie Pick for injecting the rescue construct and critically reading the manuscript. We would also like to thank Leslie Pick for her continued interest in this project as well as the many invaluable discussions and suggestions. We thank Norbert Perrimon for the wg lacZ-enhancer trap line and Armen Manoukian for embryos double-labeled for wg transcripts and en protein. We also thank Chris Doe for communicating results concerning the neuroblast patterns prior to publication. This work has been supported by the Swiss National Science Foundation grant 31-26652.89 (to M. N.), by a McKnight Neuroscience Scholars Award (to N. H. P.) and by the Kanton Zürich. C. S. G. is an Investigator with the Howard Hughes Medical Institute.