ABSTRACT
The Drosophila ovo gene, which encodes a putative transcription factor (Ovo) with TFIIIA-like zinc fingers, is required for female germline survival and proper oogenesis. Three dominant female-sterile ovoD mutations cause ovarian abnormalities that define an allelic series, with ovoD1 displaying the stronger phenotype and ovoD3 the weaker. We report here that all three ovoD mutations are point mutations that create new in-frame methionine codons in the 5′ part of ovo. There are two types of over-lapping ovo transcription units, ovoα and ovoβ. By using various ovo-lacZ reporter genes, we determined that the long Ovo isoforms starting at methionine M1, present in transcripts ovoα, are expressed at low levels only in mature oocytes. Short Ovo isoforms are translated from methionine M373, the first in-frame start codon present in tran-script ovoβ, and correspond to the activity defined by recessive loss of function ovo mutations. The new AUGs created in ovoD mutations all are located upstream of the M373 initiation site. Our results support the hypothesis that they can substitute for M373 as translation starts and initiate the synthesis of Ovo proteins that have extra amino acids at their N termini. We propose that premature expression of long Ovo protein isoforms occurs in ovoD mutants and interferes with wild-type Ovo function in con-trolling female germline differentiation.
INTRODUCTION
Two systems act in parallel to determine the sex differentiation pathway of Drosophila germ cells. An autonomous signal causes XY germ cells to develop in the male mode, irrespective of the somatic environment, while XX germ cells require both an autonomous signal determined by the X/A ratio and an inductive signal from the soma (Steinmann-Zwicky et al., 1989; Steinmann-Zwicky, 1994). Systematic screens for female sterile mutations have identified a large number of genes specifically required for proper oogenesis (Gans et al., 1975; Mohler, 1977; Perrimon et al., 1986; Schüpbach and Wieschaus, 1989, 1991). Mutants belonging to the so called ‘ovarian tumor class’ present small ovaries with egg chambers filled with an excess of undifferentiated germ cells. The initial observation that female germline cells defective in Sxl also form tumorous cysts (Schüpbach, 1985) led to the idea that this phenotype identifies loci involved in germline sex determination. This was supported by the fact that, among tumorous mutants (bam, McKearin and Spradling, 1990; otu, King et al., 1986; snf, Oliver et al., 1988; Steinmann-Zwicky, 1988; Sxl, Schüpbach, 1985; and ovo, Busson et al., 1983), only bam has a function in the male germline. Identification of the genes required for germline cells to respond to the somatic feminization signal, and determination of the biochemical pathway that is involved, are of primary interest. Because it is specifically required in females and acts upstream of Sxl in the germline, it has been proposed that ovo is a key gene in this pathway (Pauli and Mahowald, 1990; Oliver et al., 1993, 1994; Horabin et al., 1995).
Recessive, null, ovo alleles produce rudimentary ovaries in which germ cells have degenerated, leaving only somatic tissue. Hypomorphic ovo alleles permit more extensive, although abnormal, development: germ cells survive but fail to differentiate, and egg chambers are filled with a large number of mitotically active, undifferentiated cells. Three dominant ovoD mutations, ovoD1, ovoD2 and ovoD3, behaving as antimorphic alleles, have also been isolated and these, like recessive alleles, are fully penetrant for female sterility and have no effect on the male germline. In heterozygous ovoD3/+ females, egg chambers complete vitellogenesis and eggs are laid, but they do not develop. In ovoD1/+ females, oogenesis is arrested prior to or at stage 4. ovoD2 presents an intermediate phenotype with most egg chambers degenerating around stage 10 (Fig. 4D). The complete lack of vitellogenesis in ovoD1/+ females has made this mutation a widely used tool for germline clonal analysis (Chou et al., 1993; Mével-Ninio et al., 1994).
Consistent with their ovarian phenotype, the antimorphic effect of ovoD mutations seems to irreversibly affect germ cell differentiation only late during development, since elimination of the ovoD allele by mitotic crossing over in germ cells of adult females allows normal oogenesis to proceed (Perrimon, 1984). This suggests that the ovo early requirement for germ cell survival (Staab and Steinmann-Zwicky, 1995) and later function during oogenesis may involve the regulation of separate sets of genes, with only the late function being sensitive to ovoD mutations. This led us to investigate the molecular nature of these mutations.
ovo is part of the complex locus ovo-svb, as it shares most of its coding sequences with svb, a gene involved in embryo patterning. Several predicted Ovo protein isoforms are generated from transcripts initiated at two separate sites (Mével-Ninio et al., 1995, and this report) and subject to alter-native exon splicing (Garfinkel et al., 1994; Mével-Ninio et al., 1995). All isoforms are, however, predicted to include the same TFIIIA-like zinc fingers. Together the genetic and molecular data suggest that ovo is a transcription factor gene, controlling the expression of functions specifically required for female germline differentiation.
ovoD mutations constitute, to our knowledge, the only existing allelic series of dominant antimorphic mutations in a transcription factor. The antimorphic effect of ovoD mutations could be predicted to result from either ovo mis-expression during early stages of oogenesis or the production of proteins with altered, dominant, properties (see Little et al., 1995). We report here that all three ovoD mutations are point mutations that create in-frame methionine codons, at sites upstream of the normal initiation site. The observed expression patterns of wild-type Ovo and OvoD1 fusion proteins support the hypothesis that the neo-AUG created in ovoD1 is a functional translation initiator codon. Thus, the dominant-negative effects of the ovoD mutations are likely to arise from the expression of novel protein isoforms, starting early during germline cell differentiation. The nature of revertants of ovoD1 provides possible mechanisms for the dominance of the ovoD mutations.
MATERIALS AND METHODS
Drosophila stocks
Flies were raised on standard Drosophila medium at 23°C (Gans et al., 1975). The three X-linked dominant female-sterile ovoD mutations (Busson et al., 1983) were kept as attached-X stocks: C(1)DX,y f/Y females crossed to ovoD, v24/Y males. As wild-type homologs of ovoD1 and ovoD2 on the one hand and ovoD3 on the other, we used the strains fs(1)K1075 and fs(1)K1540, respectively, which carry iso-genized X chromosomes (Busson et al., 1983). Description of balancers and mutations that are not described in the text can be found in Lindsley and Zimm (1992).
Plasmid construction and Drosophila transformation
Plasmids containing the 7 kb SalI-HindIII ovo genomic fragment (previously shown to be able to rescue ovo− but not svb− mutations), or the homologous fragment derived from ovoD1 DNA, have been previously described (Mével-Ninio et al., 1991, 1994). Hybrid [ovo+-ovoD1] transgenes, depicted in Fig. 1, were cloned into the pW6 transformation vector (Klemenz et al., 1987). ovo-lacZ fusion genes were constructed, starting from the 7 kb SalI-HindIII wild-type or ovoD1 genomic fragments. In-frame fusion of ovo and lacZ coding sequences was verified by DNA sequencing. The nucleotide positions (nt) refer to the EMBL database ovo sequence DMOVO with accession number X59772. For the pP[ovoM1] and pP[ovoM-D1] constructs, the 1.65 kb SalI-BstYI ovo fragments (nt 0-1678), issued from ovo+ and ovoD1 DNA, respectively, were introduced into the pCaSpeR β-gal vector (Thummel et al., 1988) upstream of, and in-frame with, the bacterial β-galactosidase (lacZ) gene. The first wild-type in-frame codon for methionine (M1) is in exon 1a (nt 463-465), while the new methionine present in OvoD1 is at position 79 (nt 1281-1283). In both pP[ovoM1] and pP[ovoM-D1], fusion between ovo and lacZ is at the Ovo alanine residue 212 (nt 1677-1679). In pP[ovoM373], the 2.26 kb long ovo+ DNA fragment (nt 0-2268), was fused to lacZ. In this construct, the Ovo-lacZ fusion proteins initiate at either methionine M1 or M373 (see Figs 1, 2).
The pW6 and the pCaSpeR β-gal vectors contain a mini-w gene (Klemenz et al., 1987; Thummel et al., 1988). The hybrid pP[ovo+-ovoD1] constructs and the pP[ovo-lacZ] reporter constructs were injected together with the helper plasmid pUChsDelta2-3 (Flybase ID:FBmc0002087) into the w1118 host line, and transformants were selected as described by Spradling and Rubin (1982). Several trans-formed lines were obtained (17 with pP[ovoM1], 10 tested; 4 with pP[ovoM373], 4 tested; 11 with pP[ovoM-D1], 6 tested). X-gal staining reactions were carried out as described in Mével-Ninio et al. (1995).
Isolation and sequencing of the 800 bp PstI-BstXI DNA fragment from wild type and ovoD mutant strains
Adult genomic DNA was isolated from the ovoD2 and ovoD3 mutants and from their parental strains, according to Bingham et al. (1981).
The 800 bp PstI-BstXI DNA fragment was amplified using specific primers: Primer 1, 5′-AGTTGCTGCAGCGTTTGACACCAA-3′ (nt 1087-1110); Primer 2, 5′-GAGCAGAATTCGTGCGGCCAAAATG-3′, (nt 1934-1909). A single nucleotide change was introduced into primer 2 at nt 1926 to create an EcoRI restriction site. PCR reactions were performed as described in Sambrook et al. (1989) in 50 μl solution containing 300 ng genomic DNA and 10 pmoles oligomers. The amplified DNA was recovered after phenol-chloroform treatment, digested by EcoRI and PstI and cloned into pBluescript KS+. Sequencing was performed using the dideoxy-chain termination method (Sanger et al., 1977). Two independent PCR amplifications from ovoD2 and ovoD3 DNAs were sequenced.
RESULTS
Mapping the ovoD1 mutation by using chimaeric transgenes causing female sterility
ovoD mutations behave genetically as typical dominant anti-morphic mutations (Busson et al., 1983), suggesting that they encode proteins that interfere with the activity of wild-type Ovo in a dose-dependent manner. Preliminary molecular analyses suggested that ovoD are point mutations, since no dif-ferences were detected between the restriction patterns of ovoD and wild-type genomic DNAs (Mével-Ninio et al., 1989). The ovo+ function was restricted to a 7 kb SalI-HindIII fragment by transformation experiments (Mével-Ninio et al., 1991). The corresponding fragment from ovoD1 conferred a dominant-sterility phenotype (Chou et al., 1993; Mével-Ninio et al., 1994), showing that the ovoD1 mutation lies within this fragment.
We first mapped the ovoD1 mutation by constructing hybrid transgenes in which different regions of the ovo+ gene have been replaced by the homologous regions taken from ovoD1 (Fig. 1). Two properties of these transgenes were examined when injected into flies. First, since the ovoD1 mutation resulted in an arrest of oocyte development, we expected that transformed germ cells of injected female embryos (G0) would degenerate, giving no transformant. Second, we expected that these transgenes would mimic the ovoD1 phenotype in the female progeny of transformed males.
ovo-specific (as opposed to ovo-svb) mutations have been mapped in two separate regions of the ovo-svb locus (Mével-Ninio et al., 1991, 1995; Garfinkel et al., 1992).
P[ovo+-ovoD1] hybrid transgenes I to IV were designed to assay separately each of these two regions (Fig. 1). Table 1 shows that these transgenes fall into two classes. For transgenes III and IV, trans-formed progeny were obtained from G0 females, and transformed females in the next generation were fully fertile. We further verified that these two trans-genes carry an ovo+ function by complementation tests using ovoD2 (Busson et al., 1983). In contrast, transgenes I and II led to defective oogenesis, as pre-viously observed for P[ovoD1], a transgene that contains the whole 7 kb ovoD1 fragment (Mével-Ninio et al., 1994). These results mapped the ovoD1 mutation to within 1.8 kb in the 5′ region of the gene. Two additional transgenes P[ovoV] and P[ovoVI] were then tested (Table 1), allowing us to map the ovoD1 mutation within a 800 bp PstI-BstXI genomic fragment (nt 1092-1869) (Fig. 1).
Each of the three ovoD mutations creates an in-frame AUG codon with the potential to generate new Ovo isoforms
To determine the nucleotide position of the ovoD1 mutation, we compared the sequences of the PstI-BstXI fragment from both the ovoD1 and its parental ovo+ chromosome (from strain fs(1)K1075). While the sequence of fs(1)K1075 proved to be identical to the published Oregon R sequence (Mével-Ninio et al., 1991), a single difference was found between ovoD1 and wild-type DNA: the A at nucleotide position 1282 in wild type is replaced by a T in ovoD1. This substitutes a methionine for a lysine at position 79 in the open reading frame in exon 2 (Fig. 2).
The fact that all three dominant ovoD mutations display similar sterility phenotypes, albeit of different severity, led us to predict that the position in the Ovo protein is similar. Starting from this hypothesis, we sequenced the PstI-BstXI fragment from the ovoD2 and ovoD3 mutant genes, and the ovoD3 parental gene (fs(1)K1540 strain). As in ovoD1, a single nucleotide change was detected in either ovoD2 or ovoD3. In ovoD2, the G at nucleotide position 1572 in wild type is replaced by an A. This substitutes a methionine for a valine at position 176. In ovoD3, the A at nucleotide position 1336 is replaced by a T. This creates a methionine in place of a lysine at position 97 (Fig. 2). Therefore, in all three ovoD mutations, a new in-frame AUG codon is created in the open reading frame in exon 2.
Two separate AUGs are differentially used for wild-type Ovo protein synthesis
The unexpected finding that all three dominant ovoD mutations create new in-frame AUG codons raised the possibility that these codons are used as bona fide translation initiators, for the synthesis of novel Ovo isoforms with anti-morphic activity. We therefore first determined which of two possible methionine codons is used for initiation of the wild-type Ovo protein(s). Indeed, there is evidence for two types of ovo transcripts, differing in their 5′ ends and referred to below as ovoα and ovoβ, respectively (Mével-Ninio et al., 1995; Figs 1, 2). The first in-frame methionine codon present in transcript ovoα is methionine M1, while the first in-frame AUG in transcript ovoβ codes for methionine M373. M373 is located in exon 2a, an exon shared by transcripts ovoα and ovoβ. Initiation of translation at M1 and M373 would therefore result in two different proteins, designated below as OvoM1 and OvoM373, differing at their N termini by the 372 amino acids present solely in OvoM1. To determine at what stage during development M1 and M373 are used for initiating Ovo protein synthesis, we made translational fusions between different parts of ovo and lacZ coding sequences, to serve as reporter genes in transgenic lines. Transgenes ovoM1 and ovoM373 are lacZ fusions upstream and downstream of M373, respectively (see Fig. 2 and Materials and Methods). Developmental expression of ovoM1 and ovoM373 was examined and compared to expression of ovoB, a lacZ fusion gene that contains almost the entire ovo coding region, with only the two carboxy-terminal codons missing (Mével-Ninio et al., 1995).
Similar to ovoB, ovoM373 expression is detected in the nuclei of germline cells throughout oogenesis before accumulating in the mature oocyte (Fig. 3B,F). Expression of ovoM373 is also detected in germ cells of female larval gonads (not shown). The only difference with ovoB is in the intensity of staining, which is weaker in case of ovoM373, making its detection in the germarium variable from line to line. This difference may, at least partly, reflect different stability of the OvoM373 and OvoB proteins, possibly due to the difference in their intracellular localization; OvoB is strictly nuclear while Ovo M373 is both nuclear and cytoplasmic. Possibly for the same reason, ovoM373 is not detected in male larval gonads or testes. In contrast to ovoM373, ovoM1 expression is detected only late during oogenesis, in the ooplasm of mature oocytes (Fig. 3D). Furthermore, this late expression is weak in every transformant line tested.
Notwithstanding minor quantitative differences between different lines, the main conclusion that could be drawn from comparing the expression of ovoM1, ovoM373 and ovoB is that the Ovo protein(s) present during early germline cell differentiation is initiated at methionine M373. The in-frame methionine codons created by ovoD mutations therefore represent potential start codons for the synthesis of proteins larger than wild-type OvoM373 during these early stages.
The AUG codon created by the ovoD1 mutation is most probably a new translation start
To test for translation initiator activity of the neo-AUG codon at position 79 in ovoD1, we made an ovoD1-lacZ reporter construct, P[ovoM-D1], representing the mutant equivalent of P[ovoM1] (Fig. 2). The presence of the ovoD1 mutation in the ovoM1 fusion transgene resulted in a spectacular change in the transgene expression pattern. While ovoM1 is expressed at very low levels, exclusively in mature oocytes, ovoM-D1 is expressed in germline cells throughout oogenesis, starting in the germarium (Fig. 3C,G). In mature oocytes, strong staining is observed, uniformly distributed in the ooplasm. As a whole, the expression patterns of ovoM-D1 and ovoM373 are indistinguishable. A complete transition from the ovoM1 to the ovoM373 expression pattern therefore results from the presence of the OvoD1 methionine at position 79. This allowed us to conclude that the AUG created by the ovoD1 mutation is most likely to be a functional start codon. Because ovoD2 and ovoD3 also create new in-frame AUGs upstream of M373, it seems very likely that the same mechanism, i.e. initiation of translation at the newly created AUG, is used in all three ovoD mutations. The antimorphic effect of the OvoD mutations is therefore probably due to the synthesis of novel Ovo protein isoforms with NH2-terminal extensions of 294, 197 and 276 amino acids in OvoD1, OvoD2 and OvoD3, respectively.
Expression of the ovo-lacZ reporter genes in ovoDfemales
In order to test whether OvoD interferes with expression of wild-type Ovo, we compared the expression of the ovoM1, ovoM373, ovoB and ovoM-D1 reporter genes in wild-type and ovoD ovaries. Whereas no detectable change of ovoM1 expression was observed, (i.e. it remained very weak and only detectable in mature oocytes), expression of ovoM373 (or ovoB) was significantly weaker in ovoD2 (Fig. 4,A-D) and ovoD3 (not shown), compared to wild-type ovaries. In contrast, expression of ovoM-D1 appeared stronger in ovoD2 than in wild-type ovaries (Fig. 4E,F). In ovoD1/+ females, oogenesis is arrested prior to or at stage 4 and reporter gene activity was either not detected (ovoM373) or only detected in a few nuclei (ovoB and ovoM-D1; data not shown), consistent with a previous report suggesting that OvoD may negatively regulate wild-type ovo expression (Oliver et al., 1994). We noted, however, that the presence of either ovoB or ovoM-D1 appeared to improve significantly the fertility of the dominant-sterile ovoD2 females. The average number of eggs laid per day per ovoD2 female rose from 2 to 30 and 24, in the presence of ovoB and ovoM-D1, respectively. We noted that ovoB could allow the development up to the adult stage of more than 10% of the eggs. This rescue is somewhat more efficient than what we observed in a previous study (Mével-Ninio et al., 1995) and is due to the use of a different ovoB strain producing higher levels of OvoB protein. In the case of ovoM-D1, a very small fraction of laid eggs (fewer than 1/1000) could develop. By contrast, the presence of ovoM373 did not detectably change the ovoD2/+ phenotype. These results indicate that the relative expression of the Ovo+ and OvoD proteins, which is critical for the strength of the ovoD phenotype, is very sensitive to the number and nature of the ovo genes or transgenes present in ovoD mutant cells, even though these transgenes lack most of the Ovo protein-coding information.
DISCUSSION
ovo is required in XX germline cells for both viability and proper differentiation. Because it is specifically required in females and acts upstream of Sxl in the germline, ovo has been proposed to be a germline target of the somatic femi-nization signal (Pauli and Mahowald, 1990; Oliver et al., 1993). The exact role of ovo has proved difficult to assess, due in part to the loss of germ cells in amorphic alleles (Oliver et al., 1987; Staab and Steinmann-Zwicky, 1995) and the genetic and molecular complexity of the ovo-svb locus (Mével-Ninio et al., 1991, 1995; Garfinkel et al., 1994). We report here a molecular characterization of the dominant ovoD mutations and show that they correspond to the creation of new in-frame AUG codons, which most probably initiate the synthesis of novel Ovo isoforms. Expression of these novel isoforms with antimorphic activity would antagonize the function of wild-type Ovo, leading to abnormal differentiation of the oocyte.
Different functions for different Ovo protein isoforms?
ovo is required during larval stages for survival of the female germline (Staab and Steinmann-Zwicky, 1995) and later for proper oogenesis. Furthermore, the existence of a large pool of maternal ovo transcripts inherited by the embryo, a fraction of which is incorporated into pole cells (Mével-Ninio et al., 1991, 1995), suggests a possible maternal function of ovo, a role which cannot easily be assessed because all ovo mutations lead to defective oogenesis. Distinct ovo functions could possibly be fulfilled by different ovo isoforms, since a diversity of Ovo protein products results from the use of two separate transcription start sites, generating two types of tran-scripts, ovoα and ovoβ, and the alternative splicing of one protein coding exon, exon 2b (Mével-Ninio et al., 1991, 1995; Garfinkel et al., 1994). Our results, using various ovo-lacZ reporter genes, support the hypothesis that the develop-mental expression of only one class of Ovo isoforms, starting at methionine M373, correlates with the ovo functional requirement, based on the phenotypes of ovo mutations. A longer isoform, starting at the upstream methionine M1, is only detected in mature oocytes and early embryos. This could indicate that Ovo isoforms initiating at methionine M1 have no role in oogenesis per se, but may provide maternal information to the embryo. The ovoD ovarian phenotype that is likely to result from misexpression of long Ovo isoforms (although different from those initiated at M1) predicts distinct functions for OvoM1 and OvoM373. A somewhat related situation, where two protein isoforms are synthesized by alternative start codons, has recently been described for the gene oskar; only the short isoform has full oskar activity (Markussen et al., 1995). A perhaps more closely related situation (although in that case the use of two different AUGs involves a leaky ribosome scanning mechanism) is the trans-lation of the liver-enriched LAP transcriptional activator and the LIP transcriptional inhibitor from two separate AUGs on the same mRNA (Descombes and Schibler, 1991). Because long Ovo isoforms may be synthesised late in oogenesis, from transcripts ovoα, we propose that the ovoD phenotype is due to the premature expression of such long isoforms due to the creation of upstream AUGs in transcripts ovoβ.
The OvoD mutations create new initiator AUGs, which are likely to result in the production of novel Ovo protein isoforms
The dominant female-sterility phenotype of ovoD mutations is strictly cell autonomous and can be reversed up to the adult stage by induction of +/+ germline clones (Perrimon et al., 1984). This indicates that mitotically active germline stem cells are present in ovoD ovaries and remain competent for under-going oogenesis and suggests that the OvoD proteins antagonize ovo function during oogenesis. ovoD mutations are point mutations that create in-frame methionine codons upstream of the wild-type translation start M373 and the expression pattern of wild-type Ovo and OvoD fusion proteins supports the hypothesis that the ovoD1 neo-AUG is a functional initiator codon. By analogy, we postulate that the AUGs present in ovoD2 and ovoD3 are also new initiator codons. Based on these results, we favor the interpretation that the dominant effect of ovoD mutations is due to the synthesis of novel forms of the Ovo protein, rather than to the consequence of single amino acid changes.
The observation that the relative dominance and severity of ovoD ovarian abnormalities do not correlate with the relative lengths of the N-terminal extensions potentially present in the different OvoD proteins, is rather intriguing. It is possible that the respective strengths of the ovoD1, ovoD2 and ovoD3 phenotypes reflect different rates of protein synthesis initiated at the different neo-AUGs present in each mutant.
Possible mechanisms for the antimorphic properties of the OvoD proteins
Because ovoD mutations create new AUGs, their antimorphic effects are likely to arise from the expression of Ovo proteins that differ from wild-type Ovo by N-terminal extensions. Several observations indicate that these extensions have no antimorphic activity by themselves, but only within the context of an otherwise normal Ovo protein. First, the 133-amino-acid long N-terminal OvoD1 region (from methionine 79 to proline 211), expressed as part of the OvoD1-lacZ fusion protein, is not antimorphic since ovoM-D1 transgenic lines are fully fertile. Second, insertions of transposable elements in ovoD, down-stream of the M373 codon (but either upstream of or within the zinc-finger coding region) completely reverse the ovoD mutations (Mével-Ninio et al., 1989; Garfinkel et al., 1992). These data suggest that the zinc-finger DNA binding domain is required in OvoD mutant proteins for their dominant-negative activity. This contrasts with the situation observed for the WT1 protein encoded by the Wilm’s tumour suppressor gene, in which either point mutations or deletions in the zinc finger region are dominant-negative and lead to the Denish-Drash syndrome (Coppes et al., 1993; Little et al., 1995). Since OvoD1 requires the presence of the zinc-finger domain, the most straightforward interpretation to explain its antimorphic activity is a direct competition between the mutant and Ovo+ proteins for occupying target sites on the DNA (assuming, as predicted from its zinc-finger structure, that Ovo is a sequence-specific transcription factor), with the OvoD proteins being inactive. An equally possible mechanism would be interference by squelching (Ptashne and Gann, 1990). That is, the OvoD proteins may be titrating a limiting factor, different from Ovo, but necessary for Ovo function. A third possibility is that Ovo forms homodimers. Dimer formation between one Ovo+ and one OvoD molecule could result in an inactive complex sequestering the functional Ovo protein in a dose-dependent manner. In either case, competition or squelching, the resulting phenotype should be close to that of ovo hypomorph alleles. This appears to be the case, since among the diverse ovarian abnormalities observed in ovoD1 and ovoD2 females, tumorous cysts are frequently found as they are in recessive hypomorphic alleles.
A strong inhibition of the expression of transgenes ovoB, ovoM373 and ovoM-D1 is observed in the germline of ovoD1 females. A similar observation was previously made by Oliver et al. (1994), using transcriptional rather than translational ovo-lacZ fusion genes, and leading to the interpretation that the OvoD1 product is a negative trans-regulator of ovo+, (even though the presence of ovo+ is not absolutely required for ovo-lacZ expression; Oliver et al., 1994). We have now extended this observation to ovoD2 females. Expression of the transgenes ovoM373 and ovoB is down-regulated in ovoD2 compared to wild-type ovaries. Yet, although unable to sub-stitute for the wild-type ovo gene in rescuing the lack of function ovo phenotype, the ovoB transgene significantly improves fertility of ovoD2 females (Mével-Ninio et al., 1995 and this report). This observation supports the existence of a direct interaction between OvoB and OvoD2 (or Ovo+), shifting the equilibrium between inactive (Ovo+/OvoD2) and active (Ovo+/Ovo+ and, possibly Ovo+/OvoB) forms towards more of the active forms.
In contrast to ovoM373 and ovoB, ovoM-D1 leads consistently to more intense lacZ staining, although mostly cyto-plasmic, in ovoD2 compared to wild-type females. However, as is the case for ovoB, the presence of ovoM-D1 significantly improves the ovoD2 phenotype. One possible interpretation of these data is that a direct interaction occurs between the OvoM-D1 and OvoD2 proteins to form a stable complex sequestering OvoD2 in the cytoplasm. This interaction would, however, have to be different from the interaction between OvoD2 and either OvoB or Ovo+, since the 132 amino acid region of OvoD1 expressed from the ovoM-D1 transgene is absent from the short Ovo isoform while it overlaps by 35 amino acids with OvoD2. Another possibility is that a bias in the relative translation of the native and different ovoD proteins (or lacZ fusion proteins) due to the different AUG contexts is exacerbated in competition conditions, suggesting the existence of a limiting factor for ovo translation (that remains to be identified). Such a bias in translation efficiency would account for both the high sen-sitivity of the ovoD phenotype to the gene copy number and the differential expression of Ovo-lacZ and OvoD1-lacZ in different ovoD mutant contexts.
The developmentally regulated production of distinct transcription factor isoforms by alternative use of promoters and/or splicing is a widespread phenomenon (review by Lopez, 1995). The alternative use of different AUGs of unequal strength from the same mRNA has recently emerged as an addition to this potential regulatory complexity. Whereas dominant-negative mutations of WT1 have previously been shown to map in zinc-finger domains and affect DNA binding (Little et al., 1995), the molecular basis of ovoD dominance is unprecedented. A functional dissection of the mechanism of this dominance will certainly provide deeper insight into the role of ovo in the female germline differentiation pathway and the control of this pathway at the level of transcription in Drosophila.
ACKNOWLEDGEMENT
We thank Girish Deshpande, Roger Karess and François Payre for their comments on the manuscript, A. Bucheton and M. Gans for constant encouragement. This work was supported by CNRS, INSERM (Grant number 930112), ARC (grant number 1132), CEE (grant number SC1-CT920811) and a grant from Groupement de Recherches et Etudes du Genome (GREG).