A GATA family transcription factor is expressed along the embryonic dorsoventral axis in Drosophila melanogaster

In vertebrates, the GATA family of zinc finger DNAbinding proteins has been implicated in the transcriptional regulation of a variety of genes in several cell types including erythroid cells and T lymphocytes. This family of proteins, which includes four members (GATA-1 through GATA-4), has been extensively characterized in chickens, frogs, mice and humans (see Orkin, 1992, for review). These proteins all contain two C-X₂-C-X₁₇-C-X₂-C (C₄) zinc fingers that mediate binding to a WGATAR consensus DNA sequence and each of these proteins has been shown to activate transcription.

The first identified member of this multigene family, GATA-1 (Evans and Felsenfeld, 1989; Tsai et al., 1989), was originally defined as an erythroid-specific DNAbinding protein that recognized GATA sites in the regulatory regions of globin (Evans et al., 1988; Martin et al., 1989; Wall et al., 1988) and other erythroid-specific genes (Mignotte et al., 1989). GATA-1 was later shown to be expressed in other myeloerythroid lineage cells (Martin et al., 1990; Romeo et al., 1990) and in a distinct subset of testis cells (Ito et al., 1993). Targeted disruption of the GATA-1 gene blocks erythroid differentiation in vitro and in vivo, thus underscoring the critical role of GATA-1 in the genesis of mature erythroid cells (Pevny et al., 1991; Simon et al., 1992).

The GATA family in vertebrates was originally identified by homology to GATA-1 (Yamamoto et al., 1990). GATA2, although found in erythroid cells, is expressed in many non-erythroid tissues (Yamamoto et al., 1990). Ectopic expression of GATA-2 inhibits the ability of erythroid cells to differentiate (Breigel et al., 1993). GATA-3, like GATA1, has a highly defined tissue distribution, with highest levels in T lymphocytes, placental trophoblasts and the developing nervous system (Joulin et al., 1991; Ko et al., 1991; Yamamoto et al., 1990; P. Mellon, personal communication). GATA-4 is expressed in a limited set of mesodermally and ectodermally derived tissues (Arceci et al., 1993; L. Zon, personal communication). Studies in lower eukaryotes have demonstrated the existence of GATA family members in Caenorhabditis elegans (Spieth et al., 1991), Aspergillus (Kudla et al., 1990) and Neurospora (Fu and Marzluf, 1990a). The two fungal GATA factors have a single C₄ zinc finger that binds to the consensus GATA site (Fu and Marzluf, 1990b). This single finger is most homologous to the carboxyl finger of the two-finger vertebrate GATA factors, an observation supporting the important role of the carboxyl finger of the vertebrate GATA factors in DNA binding (Martin and Orkin, 1990; Yang and Evans, 1992).

The observation that murine GATA-1 plays a crucial role in determining vertebrate cell fate prompted us to explore the possible role of GATA family members in a well-characterized genetic developmental system, the Drosophila embryo. Functional analysis of transcriptional regulatory proteins (e.g., the homeodomain factors) in a variety of organisms has suggested that these gene products are involved in cell fate decisions across species lines (Häcker et al., 1992; McGinnis and Krumlauf, 1992). We have isolated a Drosophila member of the GATA family, dGATAa. The predicted protein sequence of dGATAa contains two zinc fingers which are highly homologous to their vertebrate counterpart. In a companion paper (Abel et al., 1993), the cloning of ABF (also called dGATAb), a single-finger GATA-encoding gene from Drosophila, is reported.

The dGATAa gene is transcribed at high levels in embryos from 2 to 12 hours postfertilization. The transcripts are initially localized to a region of the cellular blastoderm that will give rise to the amnioserosa and the dorsal epidermis. As embryonic development continues, expression becomes restricted to the dorsal epidermis and continues until dorsal closure is complete. An analysis of the distribution of dGATAa transcripts in mutants of maternal and zygotic dorsoventral patterning genes reveals that the maternal gene dorsal (dl) is required to restrict dGATAa expression to the dorsal ectoderm, whereas the zygotic gene decapentaplegic (dpp) is required for dGATAa expression in the dorsal ectoderm. Thus, dGATAa is a presumptive transcription factor that lies downsteam of dpp in the cascade of gene expression that specifies the dorsal epidermis, and the dGATAa expression pattern suggests that this gene product may play a role in cell fate determination along the dorsoventral axis during Drosophila development.

DNAs homologous to mGATA-3 were isolated from a Drosophila genomic λEMBL3 library. Plaque hybridization (Benton and Davis, 1977) using randomly labeled (Feinberg and Vogelstein, 1983) mGATA-3 cDNA clone mc5b8 (Ko et al., 1991) as probe was performed on 3.5×10⁴ pfu of the Drosophila genomic library (Paro et al., 1983) plated on E. coli 803 supF. The filters were UV crosslinked, hybridized (42°C for 24 hours in 50% formamide, 1 M NaCl, 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 10× Denhardts, 50 μg/ml denatured sonicated calf thymus DNA, 0.1% SDS), washed (42°C in 50 mM Tris-HCl (pH 8.0), 1 mM EDTA and 0.1% SDS) at low stringency and exposed for autoradiography (Yamamoto et al., 1990). After plaque purification, bacteriophage DNA was prepared by standard methods (Chisholm, 1989). A 900 bp BamHI DNA fragment within the single genomic λ recombinant (λ2a) recovered from the first screen was identified by Southern blot hybridization with the mGATA-3 probe and was later shown to contain the dGATAa zinc finger domain. This dGATAa fragment was used to rescreen 3.0×10⁵ pfu of the same genomic library under the same conditions, yielding two new overlapping clones. A final screen using only the most 5′ cDNA sequences as probe resulted in the isolation of the final recombinant, λ3a. This set of four non-identical overlapping recombinants thus represents approximately 30 kb of the dGATAa genomic locus. Using Southern blot analysis and restriction mapping, the recombinants were oriented with respect to one another. Appropriate restriction fragments were isolated and subcloned into pGEM vectors (Promega).

Two different Drosophila cDNA libraries were used to isolate dGATAa cDNAs. A 0to 20-hour Drosophila λgt11 embryo cDNA library (see Ryseck et al., 1985) was probed with the 900 bp genomic dGATAa zinc finger BamHI fragment, resulting in the identification of three overlapping clones. The initial clones did not encode a full-length cDNA. A 4to 8-hour Drosophila embryo plasmid cDNA library in vector pNB40 (Brown and Kafatos, 1988) was screened next, using the same 900 bp genomic DNA fragment, to isolate longer dGATAa cDNAs. After plaque purification and phage DNA preparation, the λgt11 cDNA clones were subcloned into pGEM3, whereas the pNB40 plasmid cDNA clones were subcloned into pGEM7 for further analysis.

Utilizing primer binding sites within the various plasmid vectors, all independent cDNA and genomic DNA clones were sequenced from both ends. cDNA subclones and genomic subclones were subjected to directional serial deletion by exonuclease III plus S1 nuclease (Henikoff, 1987). The genomic sequence of dGATAa was determined from various genomic subclones, while the cDNA sequence was determined from both subclones Dmc2-4a and pDmc2b. DNA sequencing was performed by the dideoxy chain termination method using T7 DNA polymerase and alkalidenatured double-stranded plasmid DNA as template (Choi and Engel, 1986; Sanger et al., 1977). 7-deaza-dGTP was used in place of dGTP (Mizusawa et al., 1986). Samples were electrophoresed on Long Ranger (AT Biochemistry) denaturing polyacrylamide gels. Sequences were analyzed using the Gene Works 2.0 software (Intelligenetics).

dGATAa cDNA clone pDmc2b (Fig. 1A) was linearized with XhoI or BamHI and transcribed into RNA in vitro using Sp6 RNA polymerase (Krieg and Melton, 1987). The RNAs representing the in vitro transcription products of XhoIor BamHI-digested DNA were then translated in vitro in the presence of [³⁵S]methionine using rabbit reticulocyte lysate (Promega) and run on 10% denaturing polyacrylamide/SDS gels as described (Laemmli, 1970).

In situ hybridization to Drosophila polytene chromosomes was carried out using ³H-labeled nick-translated genomic DNA (clone λ2a) as probe. Salivary gland polytene chromosomes were prepared from the third instar larva of Oregon R flies grown at 18°C. The following modifications were made to the standard method (Pardue and Gall, 1975) of in situhybridization to polytene chromosomes: slides were treated with 2 μg/ml RNase A followed by boiling in 200 mM NaOAc (pH 5.2) for 5 minutes; cover slips were sealed with rubber cement prior to hybridization at 37°C in a solution consisting of 50% formamide, 10% dextran sulfate, 10 mM Pipes (pH 7.0), 1 mM EDTA and 300 mM NaCl; slides were developed for 2.5 minutes in Kodak D19 and fixed for 2.5 minutes at room temperature.

Primer extension analysis was used to determine the amount of missing cDNA from the 5 ′ end of clone pDmc2b by analysis of 4to 8-hour Drosophila embryo RNA. A synthetic oligonucleotide (JWDGPE21: 5′ GCACAACGATCCGACTTC 3′), which matched the cDNA sequence, located 39 bases from the 5′ end of the cloned cDNA (nt 236 to 219 of Fig. 1B) was labeled with [γ³²P]ATP. Reactions containing 20 μg of 4to 8-hour RNA and 50 fmol of labeled oligonucleotide were heated at 70°C for 3 minutes and annealed at 45°C for 4 hours. Primer extension reactions contained the oligonucleotide annealed to the RNA in a solution containing 40 units of RNasin, 10–20 units of AMV reverse transcriptase and extension buffer (55 mM Tris-HCl (pH 8.3), 11 mM DTT, 6.7 mM MgCl₂, 27.8 μg/ml actinomycin D and 0.55 mM of each dNTP) and were carried out for one hour at 42°C. Reaction products were fractionated on denaturing 6% polyacrylamide sequencing gels.

Quantitative PCR analysis was performed essentially as described (Camp et al., 1991; Foley and Engel, 1992). Approximately 1 μg of total RNA, isolated from each embryonic developmental stage as previously described (Corces et al., 1980) was denatured (65°C) prior to use as a template in a 20 μl cDNA synthesis reaction containing: 1× RT-PCR buffer (50 mM KCl, 20 mM Tris-HCl (pH 8.4), 2.5 mM MgCl₂, 100 mg/ml BSA, 2.5 mM DTT, 1 mM each dNTP), 17.5 units RNasin (Promega), 100 pmol random d(N)₆ primers (Pharmacia) and 8 units of AMV reverse transcriptase (Promega). The reaction was incubated for 10 minutes at 22°C followed by 90 minutes at 42°C. Aliquots of the cDNA reaction were analyzed for dGATAa and ABF (dGATAb) in a 100 μl PCR reaction containing: 1× Taq DNA polymerase buffer (Promega), 0.2 mM each dNTP, 25 pmol of gene-specific primer (see Figure legends), 0.2 μl [³²P]dCTP (3,000 Ci/mmol; ICN) and 2.5 units Taq DNA polymerase (Promega). PCR conditions were 94°C, 2.5 minutes followed by 24 cycles of 94°C, 1 minute; 66°C, 1 minute; 72°C, 1 minute. Control reactions were performed to ensure that the conditions used were within the linear range of PCR amplification (not shown). All primers were chosen to be in separate exons in order to distinguish correctly spliced mRNA from any unspliced RNA or contaminating genomic DNA. Samples were analyzed on 6% polyacrylamide gels (loaded for equivalent rp49 internal control) and exposed for autoradiography. Bands were quantified using a Molecular Dynamics PhosphorImager.

Flies were grown at 25°C on standard yeast-cornmeal-agar media. The dl¹ allele was obtained from R. Lehmann (MIT) and is described by Santamaria and Nüsslein-Volhard (1983). The dpp^H46 mutation was isolated by Irish and Gelbart (1987). zen^w36 is an EMS-induced allele generated by (Wakimoto et al., 1984). The twi^ID96 mutation was isolated in the laboratory of Nüsslein-Volhard (Tearle and Nüsslein-Volhard, 1987). The dpp^H46, zen^w36 and twi^ID96 stocks were provided by the Gelbart laboratory (Harvard University). The Tl^10b allele is a dominant ventralizing mutation described by Erdelyi and Szabad (1989) and provided by K. Anderson (University of California, Berkeley).

In situ hybridizations to whole Oregon R or mutant embryos were carried out using the Genius kit (Boehringer Mannheim) as described by Tautz and Pfeifle (1989) and modified by Michelson et al. (1990). The probe used in these experiments was a 1.4 kb fragment from the dGATAa cDNA clone extending from the 5′ end to the BamHI site. For microscopy and photography, stained embryos were mounted in 90% glycerol so that they could be rotated between a coverslip and microscope slide for visualization from all perspectives.

Southern blot hybridization of genomic Drosophila DNA with a probe corresponding to the whole murine GATA-3 cDNA (Ko et al., 1991) revealed only a single hybridizing band when the blot was washed at low stringency (Fig. 2A). This probe was used to screen a λEMBL3 genomic Drosophila DNA library (Paro et al., 1983), resulting in the identification of one recombinant. Rescreening of the same library with additional GATA probes led to the identification of three additional overlapping genomic recombinants representing the single genomic locus, which we designated dGATAa (λ3a, λ4a, λ2a/3 and λ2c; Fig. 1A). A 900 bp dGATAa zinc finger genomic DNA fragment was also used to screen two cDNA libraries (Brown and Kafatos, 1988; Ryseck et al., 1985); in this manner, we were able to isolate two dGATAa cDNAs (pDm2-4a and pDmc2b; Fig. 1A).

Genomic and cDNA clones were sequenced by exonuclease III-generated serial deletions and dideoxy chain termination of plasmid DNA (Choi and Engel, 1986; Henikoff, 1987; Sanger et al., 1977). The two overlapping dGATAa cDNA clones corresponded to a sequence of 2683 nucleotides of mRNA (not shown). Primer extension analysis of 4-to 8-hour embryonic RNA using a synthetic oligonucleotide corresponding to the antisense strand of nucleotides 39 to 22 of the longest dGATAa cDNA sequence (pDmc2b; not shown) yielded a 237 (±3) nucleotide extension product (Fig. 2B), indicating that the longest cDNA clone was missing 198 bp of sequence information from its 5′ end, and that dGATAa mRNA is therefore 2881 (±3) nucleotides in length.

We learned, during proof correction, that the dGATAa gene corresponds to the pannier gene cloned by Ramain et al. (1993). Comparison of the dGATAa sequence with that of the pannier gene as determined by Ramain et al. (1993) revealed multiple discrepancies between the two. The sequence shown in Fig. 1B was revised based on the knowledge of the pannier gene sequence and careful reexamination of our own sequencing data.

The sequence of the dGATAa gene was determined by analysis of overlapping genomic subclones. The end points of the gene were determined by primer extension (5′ end) and by the position of the poly(A) tail in the cDNA clones (3′ end). The combined and overlapping sequences encoding the dGATAa gene total 5740 bp (Fig. 1B). When the cDNA and genomic sequences were aligned, we found three introns in the dGATAa locus. Each of the discontinuities of the dGATAa gene in comparison to the cDNA clones contained consensus splice donor and acceptor sequences. There were five nucleotide sequence discrepancies between the cDNA sequence and the genomic DNA (not shown); all five differences were in the third coding position of the reading frame and did not affect amino acid identity.

The first exon (nt 1 to 716) contains a long 5′ untranslated region (UTR) and the translational initiation codon followed by only an additional 8 codons prior to the first intron. The second exon (621 bp) contains a single C₄ zinc finger homologous to vertebrate GATA factor zinc fingers (from nucleotides 3923-3997). The third exon (277 bp) contains a second C₄ zinc finger which shows significant homology to vertebrate GATA factor carboxyl zinc fingers (from nucleotides 4156-4230). The fourth exon (1266 bp) contains a series of repetitive triplets (4581 to 4640) encoding a predicted domain of 20 consecutive glutamine residues followed later by a stop codon (at nucleotide 5169) and the beginning of the 3′ UTR sequences. These CAX triplets, which encode the glutamine residues, correspond to the Drosophila opa repetitive element (Wharton et al., 1985).

Homology between dGATAa and the vertebrate GATA factors lies exclusively within the zinc finger region and is quite divergent outside of this domain (see Abel et al., 1993, for a comparison of the zinc finger regions of the GATA family members). Comparison of the dGATAa cDNA and genomic sequences shows that dGATAa introns 2 and 3 are quite small (62 and 68 bp, respectively), whereas intron 1 is of moderate size (2,729 bp).

The dGATAa cDNA clones contain a single long open reading frame (ORF) encoding 540 amino acid residues with a predicted relative molecular mass (M_r) of 57×10³ (Fig. 1B). We examined the capacity of the larger cDNA clone (pDmc2b) to encode the predicted protein by in vitro transcription/translation. A single major protein product of this reaction was found to migrate on a SDS-polyacrylamide gel with an apparent M_r of 57×10³, in agreement with the predicted molecular weight (Fig. 2C). After cleaving pDmc2b with BamHI (position 4352, Fig. 1B), the single translation product was found to migrate at approximately 32×10³, also in agreement with predictions based on cDNA sequence.

Drosophila salivary gland polytene in situ localization, using dGATAa genomic clone λ2a (Fig. 1A) as a probe, mapped the dGATAa locus to chromosome 3R 89A (Fig. 2D), a region near the embryonic lethal mutations serpent and pannier (Lindsley and Zimm, 1992).

The vertebrate GATA factors, as discussed in the Introduction, are expressed with sometimes overlapping, but distinct, tissue-specific profiles. In addition to the two-finger Drosophila GATA factor identified in this study, a related but distinct single-finger GATA family member (ABF, also called dGATAb; Abel et al., 1993) has been identified, thus raising the possibility that, as in vertebrates, multiple GATA factors might be differentially expressed and carry out distinct developmental programs during Drosophila development. The GATA factor cDNA clones described here were isolated from libraries prepared from 0to 16and 4to 8-hour Drosophila embryo RNA. We therefore wished to determine whether or not these two Drosophila GATA family members (dGATAa and ABF) are differentially expressed during embryogenesis.

Preliminary RNA blot analysis suggested that dGATAa mRNA accumulated to its highest level at 4 to 6 hours of embryogenesis (data not shown). These observations were confirmed and extended using quantitative RT-PCR. Total RNA was isolated from staged Drosophila embryos at the indicated times and from adults. This RNA was used as a template for cDNA synthesis and finally assayed for the presence of dGATAa and ABF mRNAs by PCR. The Drosophila ribosomal protein 49 gene has been shown to be abundantly and uniformly expressed throughout development (O′Connell and Rosbash, 1984). rp49 was therefore used as an internal control for cDNA synthesis and a measure of the total amount of RNA in a given PCR reaction. All PCR amplification primers were designed to be in separate exons to enable correctly spliced mRNAs to be distinguished from any unspliced RNA or contaminating genomic DNA that might be present. Control reactions were performed to ensure that the various primers could be used in combination in a single PCR reaction without adverse effects and that the conditions used were within the geometric range of PCR amplification.

The results of these experiments (Fig. 2E) revealed that both dGATAa and ABF were expressed at various times throughout Drosophila embryonic development; furthermore, the profile of accumulation of the two mRNAs was strikingly similar (both factors were maximally expressed in 4to 6-hour embryos). However, quantitative analysis revealed that the levels of ABF mRNA appeared to increase more rapidly up to this time and decreased more rapidly from this point to 22 hours of development when compared to dGATAa (Fig. 2F), a result consistent with the ABF expression profile presented in the accompanying paper (Abel et al., 1993). Furthermore, whereas dGATAa appears to be expressed at similarly low levels in 0to 2-hour and 16to 22-hour embryos as in adult flies, ABF mRNA was in twofold greater abundance in adults than in 0to 2-hour embryos. Analysis of the same RNA samples for expression of the engrailed gene (Poole et al., 1985) and the -tubulin gene (Theurkauf et al., 1986) revealed that the former is maximally expressed in the 6to 8-hour embryo sample whilst the latter, like rp49, was expressed at essentially constant level throughout embryogenesis (not shown). Thus the similarity in profile of dGATAa and ABF expression is not a consequence of the methodology employed here, but appears to reflect the genuine temporal accumulation of these two mRNAs during Drosophila embryonic development.

In order to determine the spatial pattern of dGATAa expression during embryogenesis, we used a whole-mount in situ hybridization protocol (Tautz and Pfeifle, 1989). Distinct localization of dGATAa transcripts is first detected in late stage 5 embryos following the completion of cellularization (Fig. 3A,B). dGATAa mRNA is localized to the dorsal 30% of the blastoderm embryo in a region that will give rise to the amnioserosa and dorsal epidermis. The lateral extent of dGATAa expression, therefore, includes most, but not all, of the dorsal epidermis. The zygotic patterning genes zerknüllt (zen) and decapentaplegic (dpp) are also localized to the dorsal epidermis in a broader domain of the embryo (Ray et al., 1991; Rushlow et al., 1987; St. Johnson and Gelbart, 1987). Unlike the initial zen and dpp domains, dGATAa expression is restricted along the anterior-posterior axis, with expression extending from approximately 20% to 60% (anterior to posterior) of the egg length. Within this region, expression is seen in a series of 5 stripes, the most posterior of which is separated from the others by a wider nonexpressing interstripe (Fig. 3B). The spacing between these stripes suggests that they may reflect parasegmental units.

As gastrulation proceeds and germ band extension begins, dGATAa expression remains localized to the dorsal side of the embryo in a domain having distinct anteroposterior limits (Fig. 3C,D). The anterior boundary remains fixed at the cephalic furrow while the posterior boundary expands to the proctodeum during the early phase of germ band extension. Midway through germ band extension, dGATAa transcripts begin to concentrate in the dorsal ectoderm although expression continues in the amnioserosa (Fig. 3E,F). In the late germ band extended embryo, dGATAa expression is localized exclusively in the epidermis just dorsal to the tracheal pits (Fig. 3G,H). An additional change in the dGATAa expression pattern that occurs during late germ band extension is the exclusion of dGATAa transcripts from the proctodeum (compare Fig. 3F with Fig. 3H).

By the completion of germ band shortening, dGATAa staining persists in the dorsalmost epidermis, but transcripts also appear in the posterior spiracles (Fig. 3I,J). During dorsal closure, the dorsal epidermis stretches over the amnioserosa until the two sides meet at the dorsal midline. dGATAa expression continues in the dorsal epidermis throughout the process of dorsal closure, although the staining intensity reproducibly declines at this stage and becomes confined to a single cell on either side of the midline (Fig. 3K). Anteriorly, the dGATAa-expressing epidermal cells form an expanded cluster.

The dorsoventral axis of the Drosophila embryo is specified by one set of maternal genes which includes the eleven genes of the dorsal group and cactus (Anderson, 1987; Rushlow and Arora, 1990; St. Johnson and NüssleinVolhard, 1992). The axis is organized by the differential nuclear localization of maternal dorsal (dl) protein, displaying highest levels in the nuclei of ventral cells. This dorsal protein gradient appears to specify cell fate along the dorsoventral axis by dividing the embryo into three domains based on the expression pattern of the zygotic dorsoventral patterning genes: a ventral domain of twist (twi) and snail (sna) expression, a dorsal domain of dpp and zen expression and a lateral region in which none of these four genes is expressed (Ray et al., 1991). These regions give rise to the mesoderm, dorsal ectoderm and neuroectoderm, respectively. The dorsal gene product activates twi and sna in the mesoderm and represses dpp and zen in the ventral and lateral domains of the embryo (Ray et al., 1991). In the case of zen, dl appears to be a direct transcriptional repressor, preventing zen expression in the ventral portion of the embryo (Ip et al., 1991; Jiang et al., 1992). Dorsal binding sites within the twi promoter mediate the dorsal-dependent activation of twi in the ventral portion of the embryo (Jiang et al., 1991; Pan et al., 1991; Thisse et al., 1991).

The spatial and temporal patterns of dGATAa expression suggest that it may play a role in patterning along the dorsoventral axis. In order to determine where dGATAa fits within the hierarchy of determinative events in the dorsal epidermis, we determined the pattern of dGATAa expression in various dorsoventral polarity mutants (Fig. 4). In embryos produced from dl¹ homozygous mutant mothers, the epidermis is completely dorsalized (Santamaria and Nüsslein-Volhard, 1983). Consistent with its presumptive role in promoting dorsal cell fate, dGATAa transcripts extend across the entire dorsoventral axis of these mutant embryos (compare Fig. 4A with Fig. 3A). This could be a direct effect of dorsal on the transcription of the dGATAa gene. Alternatively, it could be the result of the effect of dorsal on dpp expression (Ray et al., 1991). The dGATAa expression pattern is largely unchanged in its anteriorposterior extent, although the stripes seen in wild-type embryos are not distinct. Toll^10b (Tl^10b) is a dominant gain-of-function maternal effect mutation which results in ventralized embryos (Anderson et al., 1985; Erdelyi and Szabad, 1989). In embryos produced by Tl^10b mutant mothers, dGATAa is not expressed (data not shown), consistent with the loss of dorsal derivatives.

The zygotic genes dpp and zen are involved in specifying dorsal cell fate (Ferguson and Anderson, 1991). dpp, a member of the TGF-β family, appears to be the primary patterning gene for the dorsal ectoderm (Ray et al., 1991) and recent experiments have revealed that dpp acts as a morphogen to organize dorsoventral patterning within the embryonic ectoderm (Ferguson and Anderson, 1992). zen, a homeobox-containing transcriptional activator protein, acts downstream of dpp and may control the development of the amnioserosa (Rushlow and Levine, 1990). In embryos homozygous for the dpp^H46 mutation, no dGATAa expression is observed (Fig. 4C), consistent with the ventralized phenotype of these embryos and the critical role that dpp plays in the differentiation of the dorsal epidermis (Ray et al., 1991). In homozygous zen^W36 mutant embryos, the lateral border of dGATAa expression is restricted to a narrower, more dorsal portion of the embryo (compare Fig. 4B with Fig. 3D), consistent with the absence of amnioserosa in these embryos and the concomitant dorsal shift of position values (Arora and Nüsslein-Volhard, 1992). The initial domain of dGATAa expression includes the amnioserosa and this factor could also play a role in the development of this tissue. Interestingly, ABF is also expressed in the amnioserosa at this stage, although ABF is absent from the dorsal epidermis (Abel et al., 1993). In the amnioserosa, therefore, these two GATA family members may perform redundant functions. In addition to its transient expression in the amnioserosa, ABF is expressed in a dynamic pattern that begins in the presumptive endoderm and cephalic mesoderm in the cellular blastoderm and ends in the fat body of the later embryo (see Abel et al., 1993).

The zygotic dorsoventral patterning genes, whose localized expression is established by the maternal gene dl, appear to act independently in the dorsal and ventral regions of the embryo. Thus mutations in dpp and twi do not affect the expression of each other (Ray et al., 1991). Consistent with this view, the domain of dGATAa expression in embryos mutant for the ventrally expressed zygotic gene twi (Thisse et al., 1988) is similar to that seen in wild-type embryos except for the obvious distortions imposed by the twist phenotype (compare Fig. 4D with Fig. 3G).

Studies in vertebrates have revealed the important role that GATA family members play in determining cell fate (Orkin, 1992). Indeed, GATA-1 is required for the formation of mature erythroid cells in the mouse (Pevny et al., 1991; Simon et al., 1992). The analysis of dGATAa expression in wild-type and mutant embryos, together with its potential role as a transcriptional regulator, suggests that this gene may be involved in crucial developmental decisions along the dorsoventral axis of the Drosophila embryo. The dGATAa gene (or pannier, according to Ramain et al., 1993) is associated with incomplete anteriordorsal closure (Lindsley and Zimm, 1992). It will be of interest to determine the exact role that this putative transcription factor plays in the zygotic process of dorsoventral patterning in the Drosophila embryo.

The λEMBL3 Drosophila genomic library was kindly provided by Mike Goldberg and the pNB40 plasmid cDNA library was generously provided by Nick Brown (Harvard University). We are grateful to our colleague K. George for her patience in instructing one of us (J. D. W.) in the art of library screening. We thank R. Lehmann, N. Perrimon, K. Anderson, V. Hartenstein, H. Skaer and the Gelbart laboratory for supplying fly stocks and providing insight in interpreting the expression pattern. I. C.-L. took part in this work while on sabbatical leave from the Faculty of Pharmaceutical and Biological Sciences, University of Paris V. This work was supported by NIH research grants GM 28896 (J. D. E.) and GM 29379 (T. M.), an NIGMS Training Fellowship (GM 08061; J. D. W.), a National Science Foundation Graduate Fellowship (T. A.) and postdoctoral fellowships from the Leukemia Society of America (M. W. L.) and the Medical Foundation (A. M. M.). A. M. M. is an Assistant Investigator of the Howard Hughes Medical Institute.