Function of Drosophila ovo⁺ in germ-line sex determination depends on X-chromosome number

An important step in germ-line development, is the bifurcation leading to spermatogenesis or oogenesis. One of the principal inputs in both the germ-line and somatic sex determination hierarchies (reviewed by Burtis, 1993) involves counting the number of X chromosomes (Schüpbach, 1985; Cline, 1986, 1988; Steinmann-Zwicky et al., 1989). Nothing is known about how X chromosomes are counted by germ cells. In the soma the balance between positively acting, X-chromosome-encoded transcription factors, and negatively acting transcription factors encoded by the autosomes or supplied maternally (Caudy et al., 1988; Torres and Sánchez, 1989; Parkhurst et al., 1990; Erickson and Cline, 1991, 1993; Van Doren et al., 1991; Younger-Shepard et al., 1992; reviewed by Cline, 1993) translates into an all-or-nothing response through the Sxl⁺ gene (Cline, 1984, 1986, 1988; Keyes et al., 1992). One might imagine, a priori, that the same transcription factors would be involved in counting X chromosomes in the germ line. This is not true. Mosaic analysis suggests that the genes involved in X-chromosome counting (and/or Sxl⁺ activation) in the soma are not required for germ-line sex determination (Cline, 1983; Schüpbach, 1985; Cronmiller and Cline, 1987; Steinmann-Zwicky, 1993; Granadino et al., 1993).

In addition to the number of X chromosomes and somatic sex determination signals (Nöthiger et al., 1989; Steinmann-Zwicky et al., 1989; Oliver et al., 1993; Steinmann-Zwicky, 1994), several X-chromosome genes (ovo⁺, otu⁺, snf⁺ and Sxl⁺) have been implicated in germ-line sex determination (Steinmann-Zwicky et al., 1989; Oliver et al., 1990, 1993; Wei et al., 1991, 1994; Pauli et al., 1993). Germ-line Sxl⁺ function or female-specific pre-mRNA splicing requires an XX karyotype, a female soma (Schüpbach, 1985; Steinmann-Zwicky et al., 1989; Oliver et al., 1993) and the action of the ovo⁺, otu⁺ and snf⁺ genes (Bopp et al., 1993; Oliver et al., 1993; Pauli et al., 1993), which would be expected if these genes are involved in counting the number of X chromosomes or processing somatic sex determination signals. While snf⁺ is an RNA binding protein (Flickinger and Salz, 1994) and may therefore act directly on Sxl⁺ pre-mRNA (cf. Albrecht and Salz, 1993; Bopp et al., 1993; Oliver et al., 1993), the mechanism of otu⁺ and ovo⁺ function is unknown. We have analyzed the ovo⁺ gene in order to find its place within this regulatory scheme.

We have identified a germ-line-specific promoter mapping to the ovo⁺ part of the ovo-svb locus (Oliver et al., 1987; Mével-Ninio et al., 1989; Garfinkel et al., 1992) and show that this promoter activity is much higher in females than in males. The germ-line-specific activity is then used as a molecular phenotype in the analysis of flies mutant for various sex determination genes. Our data indicate that ovo⁺ transcription is higher in females than in males because of the number of X chromosomes in female flies. We also find that ovo⁺ is required in all cases where the karyotype is XX, and is never required in cases where the karyotype is X0 or XY. These data suggest that ovo⁺ expression responds to the number of X chromosomes and is required in XX germ cells independently of germ-line or somatic sexual identity.

The ovo:lacZ constructs were derived from a 10 kb genomic fragment previously shown to be sufficient to rescue ovo⁻mutations, but not svb⁻mutations (Garfinkel et al., 1992). The 5.2 kb EcoRI fragment was placed directly upstream of the promoterless bacterial β-galac-tosidase gene (lacZ) in pCaSpeR AUG β-gal (Thummel et al., 1988). There is an in-frame stop codon upstream of the AUG used to drive expression of the reporter, which eliminates the possibility of trans-lational readthrough. The resulting plasmid is pCOW 5.2. Unidirectional deletions of the ovo genomic DNA were made from the 3′ end of the 5.2 kb ovo fragment with Exonuclease III (Sambrook et al., 1989). Three of these pCOW 5.2 derivatives with 5′ ovo DNA termini at −1.2 and 3′ termini at approximately +2.1 kb, +1.9 kb and 0.0 kb respectively were used in this study. Transformation of flies was performed essentially as described by Spradling (1986).

All flies were grown at 25°C, except for snf⁻flies, which were grown at 25°C or 31°C, and ovo⁻flies, which were grown at 18°C or 25°C (see text and figure legends). See Lindsley and Zimm (1992) or Flybase (1994) for descriptions of mutant alleles and rearranged chromosomes. See Fuller (1993) and Spradling (1993) for gonad morphology.

The X-gal reactions for gonads were carried out according to the method of Pauli et al. (1993), except that gonads were fixed in 2.2% formaldehyde and X-gal was used at a concentration of 0.04%. For long staining reactions fresh solutions were added at 12-hour intervals. In all cases, positive and negative controls were incubated in the same tubes as the experimental gonads. The vasa antigen was recognized by rat anti-vasa serum (a gift of P. Macdonald) and per-oxidase linked goat anti-rat IgG.

Analysis of the ovo⁺ sex determination function has been hindered because the ovo⁺ locus is required for a number of different developmental events only some of which are likely to be related to sex determination (Oliver et al., 1987; Mével-Ninio et al., 1989, 1991; Garfinkel et al., 1992). Most prominently, mutations disrupting one of the ovo⁺ locus functions results in male-specific instead of female-specific Sxl⁺ pre-mRNA splicing, while a second function of the ovo⁺ locus, shavenbaby⁺ (svb⁺), is required non-sex-specifically in somatic cells for the elaboration of denticle belts (Oliver et al., 1987, 1993). These functions can be separated genetically. The ovo⁺ locus can be divided into five domains (Fig. 1A): Domain I is required only for svb⁺ function; domains II and IV are required only for ovo⁺ function; and domains III and V are required for both ovo⁺ and svb⁺ functions. The region required for wild-type ovo⁺ function is further delimited by two genomic fragments that rescue ovo⁻mutations (Mével-Nino et al., 1991; Garfinkel et al., 1992). svb⁻mutations are not rescued by these fragments, indicating that sequences required for svb⁺ function lie elsewhere. These data along with the knowledge that ovo⁺ is transcribed from left to right (Mével-Ninio et al., 1991) suggest the presence of an ovo promoter (P_ovo) in Domain II.

To examine the putative P_ovo region further, we have made fusions between ovo locus sequences and lacZ to serve as reporters of promoter activity (Fig. 1A). Two slightly different constructs containing domain II sequences fused upstream of lacZ were made (pCOW +2.1 and pCOW +1.9). A third construct containing DNA upstream of the predicted promoter location was also used (pCOW 0.0). All three constructs were introduced into flies by P-element mediated transformation. Flies bearing pCOW +2.1 or pCOW +1.9 have detectable enzymatic activity in germ cells, flies bearing pCOW 0.0 do not (Fig. 1B). Thus, sequences required for germ-line-specific expression of the reporter gene are localized in domain II, a region required for germ-line-specific ovo⁺, but not somaspecific svb⁺, genetic function.

Chromogenic staining in wild-type female germ cells (Fig. 1B,C; also see controls in later figures) resembles the expression pattern for ovo-svb RNA as visualized by in situ hybridization (Mével-Ninio et al., 1991; Garfinkel et al., 1994; A. P. Mahowald, personal communication; A. Vincent, personal communication). In the germarium, chromogenic staining was seen in the stem cells (region 1), cystoblasts (region 1) and young cysts (region 2), with region 2 being more strongly stained. Staining intensity decreases markedly in region 3 of the germarium, but shortly thereafter (stage 3 or 4 egg chambers) the staining signal is as strong as in region 2 of the germarium. By stage 8 the signal is very strong. The enzymatic activity is ultimately dumped from the nurse cells into the maturing egg. In situ hybridization shows high steady state levels of ovo-svb RNA in all germarial regions, and very strong expression in stage 8+ egg chambers (Mével-Ninio et al., 1991; Garfinkel et al., 1994; A. P. Mahowald, personal communication; A. Vincent, personal communication), suggesting that the ovo-specific promoter contributes to at least part of the ovo-svb transcript pool observed in the ovary by in situ hybridization.

The reporter genes are expressed weakly, or not at all, in testes (Fig. 1B,D; also see controls in later figures). Chromogenic staining of the male germ cells in the apex of the testis heterozygous for pCOW+1.9 or pCOW+2.1 ovo:lacZ constructs is weak and inconsistent even with prolonged staining, while staining of coprocessed female germ cells heterozygous for the same constructs is quite distinct. Thus, ovo:lacZ activity in females is much stronger than in males for all reporter lines. In situ hybridization also reveals ovo-svb RNA in the apex of the testis (A. P. Mahowald, personal communication) and an ovo:lacZ construct encoding a fusion protein shows markedly less expression in male than in female germ cells even in larval stages (C. Salles and F. Payre, personal communication). Our data confirm that ovo⁺ is transcriptionally active in the germ cells of both females and males, but is more active in females. More importantly, the reporter activity provides a molecular phenotype to help order genetic requirements for germ-line sex determination.

In the remainder of the paper we determine whether the differential ovo:lacZ expression observed between females and males is dependent on karyotype, somatic signaling, or germ-line sexual identity. These distinctions are critical for placing the ovo⁺ locus within the framework of the germ-line sex determination hierarchy.

Given that there is clear data supporting a role of the somatic sexual identity in addition to the number of X chromosomes for the specification of germ-line sexual identity, we have looked for the expression of the ovo reporters in gonads of mutant flies where the chromosomal sex does not match the somatic sexual identity. In this way we can investigate which sex determination inputs are required for high level reporter gene activity, and by inference, which inputs control ovo⁺.

If an XX karyotype is necessary and sufficient for high levels of ovo:lacZ expression, then female somatic sex determining signals should have no effect on ovo:lacZ in an XY germ cell. XY females can be produced by using a strain of flies bearing a transformer (tra) minigene driven by a heat-shock promoter (tra^hs). In XY flies bearing the tra^hs construct, tra⁺ activity is sufficient to direct female somatic development in flies that would otherwise develop as males (McKeown et al., 1988) but the germ line remains male (SteinmannZwicky et al., 1989). If a female somatic sexual identity is sufficient for the high level reporter activity, then both XX females and XY tra^hs females would be expected to show similar reporter gene activity. The ovo reporter genes are inconsistently and weakly expressed in the germ cells of XY males transformed to female by tra^hs (Fig. 2A,C). This staining is similar to that seen in control XY males processed in the same tube (Fig. 2B,D) and is much weaker than control XX female staining – also processed in the same tube (not shown). These data indicate that the ovo reporter gene is not activated to a high level by a female somatic sexual identity, suggesting that an XX karyotype or female germ-line sexual identity are required for high levels of ovo:lacZ expression.

If an XX karyotype is necessary and sufficient for high ovo reporter activity, then XX males should show high activity. Without transformer-2 (tra-2), XX flies develop as males instead of females (Baker and Ridge, 1980). When XX flies transformed to male and also heterozygous for ovo:lacZ were examined, signal was readily detectable in the germ cells of XX tra-2⁻testes in all allelic combinations tested (Fig. 2E; tra-2^B/tra-2¹ or tra-2^B/tra-2^B or tra-2^B/Df(2)TRIX flies were tested), supporting the idea that an XX karyotype is necessary and sufficient for ovo:lacZ reporter expression.

Intersexual XX flies also show strong ovo:lacZ expression (not shown). Because limited female germ-line development occurs in XX males (Nöthiger et al., 1989; Oliver et al., 1993), it is a formal possibility that the high levels of ovo:lacZ expression are due to partial female germ-line development. In experiments outlined in the next paragraph we show that female germ-line development is not required for high expression of ovo:lacZ, greatly strengthening the idea that the principal regulator of ovo:lacZ activity is the number of X chromosomes.

Given that Sxl⁺ occupies a position downstream of ovo⁺ in the germ-line sex determination hierarchy (Oliver et al., 1990, 1993) mutations in Sxl would not be expected to affect ovo:lacZ expression. However, since the germ cells of Sxl⁻females are sex transformed (Oliver et al., 1988, 1993; Steinmann-Zwicky et al., 1989; Pauli et al., 1993; Wei et al., 1994), Sxl⁻mutants allow us to address the question of whether high ovo reporter activity requires female germ-line differentiation. The ovo reporters show strong activity in XX, Sxl⁻germ cells in all tested allelic combinations (Fig. 3A; Sxl^fs#1/Sxl^fs#1 or Sxl^fs#1/Sxl^fs#3 or Sxl^fs/Df(1)Sxl^7BO). The germarium region is strongly stained and this signal extends beyond the location where decreased reporter activity occurs in wild-type females. Given that the germ cells in these females exhibit both a developmental block and a sex transformation (Pauli et al., 1993; Wei et al., 1994), the more extensive staining is expected and is likely to be a simple function of the expanded region populated by earlier stage germ cells. In brief, the above experiments indicate that high level ovo:lacZ reporter expression depends on the number of X chromosomes, not somatic or germ-line sexual identity.

There are four X-linked genes with clear roles in germ-line sex determination, snf⁺, Sxl⁺, otu⁺ and ovo⁺. Given that the ovo reporters appear to respond to the number of X chromosomes, we have examined the expression of the reporters in backgrounds where the wild-type alleles of these genes are absent. The ovo⁺ and otu⁺ genes are believed to occupy an upstream position in the germ cell autonomous part of the sex determination hierarchy (Oliver et al., 1993; Pauli et al., 1993). Mutations at either locus result in severe reduction in XX germ cell number (King et al., 1986; Oliver et al., 1987), whereas genes like snf⁺ and Sxl⁺ have no known XX germ-line viability functions (Schüpbach 1985; Oliver et al., 1988; Flickinger and Salz, 1994). It is conceivable that otu⁺ is part of the dose dependent X-linked system that regulates ovo reporter expression. We have therefore looked at the activity of ovo reporters in females bearing different otu⁻mutant alleles (otu¹/otu⁸ or otu¹/otu¹⁷ or otu¹/otu¹ or otu¹⁷/otu¹⁷ were tested). In otu⁻germ cells, there was clear and strong activity of the reporter genes (Fig. 3B,C). This chromogenic staining is much stronger than seen in XY flies. The staining pattern in otu⁻ovaries was patchy. In particular we find clusters of germ cells with high reporter activity in very close proximity to germ cells and clusters that show no discernible activity. Even though otu⁺ does not appear to be an obligatory regulator of high level ovo⁺ promoter activity, the absence of otu⁺ may result in a near threshold condition for high ovo:lacZ expression. Alternatively, as XX, otu⁻germ cells are not very healthy, the patchy staining could be a simple consequence of cell death. As outlined above the Sxl⁺ gene is not an obligatory regulator of ovo:lacZ (Fig. 3A). Expression of the reporters in snf⁻flies (snf¹⁶²¹v²⁴/Df(1)JC70 or snf¹⁶²¹v²⁴/snf¹⁶²¹v²⁴) shows the same pattern seen in Sxl⁻(not shown), although there may be a slight decrease in the staining intensity. Thus, we find no evidence that a female germ-line sexual identity is required for ovo:lacZ expression or that the known X-linked germ-line sex determination genes are obligatory (i.e.. non-redundant, see Discussion) counting elements.

We have looked at XX flies mutant for ovo alleles to see if ovo⁺ has any autoregulatory function. Females of the genotype ovo^D1/+ show arrested oogenesis at about stage 4 (Oliver et al., 1987; 1990). There is little reporter gene activity evident in ovo^D1/+ females (Fig. 4). This down-regulation of the ovo reporter is not due to either the absence of viable germ cells or the absence of female differentiation, as ovo^D1/+ flies have basically normal looking germaria and egg chambers until stage 4. Indeed, ovo^D1/+ germ cells show more female character than those of Sxl⁻, snf⁻or otu⁻females, and much greater viability than seen in otu⁻females, which show robust ovo reporter activity. Thus, these data strongly suggest that the ovo^D1 protein is a negative trans-regulator of ovo⁺. This finding may explain why the ovo^D1 mutation acts as a classic antimorph (Busson et al., 1983; indicating that the mutant gene product inactivates or opposes the wild-type gene product). The ovo^D2 allele shows a similar effect on ovo:lacZ expression and the weaker ovo^D3 allele shows a less extreme negative effect on the reporter genes (data not shown). Both of these alleles are also antimorphic (Busson et al., 1983). The complete absence of ovo⁺ does not abolish reporter gene expression as ovo⁻/Y males (ovo^D1rS1/Y or lzl^G/Y) show weak ovo:lacZ expression like wild-type males. Additionally, flies heterozygous for deletions of ovo (Df(1)JC70/+ or Df(1)RC40/+) or ovo⁻insertional mutations (ovo^D1rS1/+ or lzl^G/+) show high levels of enzymatic activity derived from ovo:lacZ reporters (data not shown). While these data implicate ovo⁺ in ovo:lacZ regulation, ovo:lacZ activity is not a simple function of ovo⁺ dose.

The reporter gene data indicate that ovo⁺ activity is differentially regulated in flies with X versus XX karyotypes, but we have not addressed the issue of whether the expression differences are functionally significant. A priori, one might expect that cells showing high reporter activity require ovo⁺, while cells showing low reporter activity may not. This is, in fact, what we find. Before turning to the effect of ovo⁻on the germ line of flies with different combinations of chromosomal and phenotypic sexes, we have re-evaluated the XX ovo⁻phenotype (Fig. 5). The gonads of XX ovo⁻females (newly eclosed or aged for 14 days, grown at 25°C or the more permissive temperature, 18°C) were dissected and stained with antibodies against the germ-line protein, vasa (Lasko and Ashburner, 1990). The ovaries from flies grown at either temperature are characterized by the absence of germ cells (64% in this experiment n=;41; Fig 5A,B,D,G), the rare escaping germ cells are usually found as single cells or small clusters in a single ovariole (20%; Fig. 5C,E,F), but can sometimes form cysts (16%; Fig. 5H). Thus, the atrophic nature of XX, ovo⁻ovaries (Oliver et al., 1987; ovo^D1rS1v²⁴ /ovo^D1rS1 or ovo^D1rS1v²⁴/lzl^Gsn or lzl^Gsn/lzl^Gsn) is due to the absence of germ cells.

If ovo⁺ is required in XX females because of the XX karyotype, then XY females should not be affected by the absence of ovo⁺. We have compared XY, ovo⁻;tra^hs ovaries with XY, ovo⁺;tra^hs ovaries and show that both have a similar phenotype, characterized by abundant germ cells (Fig. 6A,B,C; ovo^D1rS1v²⁴/Y; tra^hs/+ or lzl^Gsn/Y; tra^hs/+). There are some examples of XY, tra^hs females with few or no germline chambers, but about 60% show large numbers of male germ cells (Fig. 6A,C; cf. Fig. 2A,C; Steinmann-Zwicky et al., 1989). Thus, XY, ovo⁻; tra^hs and XY, ovo⁺; tra^hs show similar germ cell abundance in stark contrast to the absence of germ cells in XX, ovo⁻females. The presence of germ cells in XY, ovo⁻;tra^hs flies does not appear to be due to protection by a Y chromosome, as X0, ovo⁻flies are indistinguishable from X0, ovo⁺ males. The germ cells of X0, ovo⁻males (ovo^D1rS1v²⁴/0 or lzl^Gsn/0) have the crystals characteristic of X0 males (Fig. 6D; cf. Meyer et al., 1961), and are fully populated with male germ cells of all stages (Fig. 6E). These data along with previous experiments showing that the germ line is absent from XX, ovo⁻;tra⁻male flies (Steinmann-Zwicky et al., 1989) and XX females homozygous for both ovo⁻and female sterile alleles of Sxl (Oliver et al., 1990) or ovo⁻and otu⁻(Pauli et al., 1993) suggest that ovo⁺ is required in XX germ cells, but not X germ cells. This correlates well with the observation that ovo:lacZ expression is high in flies with an XX karyotype.

A germ cell must initiate either oogenesis or spermatogenesis in essentially all higher organisms, but this commitment step is not well understood. Previous genetic and molecular studies have suggested that the ovo⁺ gene is involved in multiple sexspecific germ-line events, one of which is related to sex determination (Oliver et al., 1987, 1990, 1993; Wei et al., 1991; Pauli et al., 1993). The goals of the experiments reported here were to determine how ovo⁺ is likely to be regulated to perform its female-specific function, and more importantly to determine its position in the germ-line sex determination hierarchy.

Constructs were made to serve as reporters of the ovo⁺-specific part of the ovo-svb locus. The high levels of ovo reporter activity detected in the XX female germ cells and not in XY male germ cells enabled us to use these reporters to ask if ovo⁺ is regulated (directly or indirectly) by the number of X chromosomes or by somatic sex determination cues. Flies with an XX karyotype, require ovo⁺ and show high levels ovo:lacZ expression. There is no correlation of either an ovo⁺ requirement or ovo:lacZ expression with somatic or germ-line sexual identity. These results have allowed us to make some refinements on previous germ-line sex determination models (Fig. 7).

An XX karyotype is sufficient for high ovo:lacZ expression even in the absence of female somatic sex determination signals, suggesting that ovo⁺ is regulated by an XX karyotype. These data suggest ovo⁺ transcription and possibly ovo⁺ function are upstream or independent of the somatic signal (we cannot rule out post-transcriptional regulation by signaling). Both ovo⁺ and the somatic signals are required for femalespecific Sxl⁺ pre-mRNA splicing and good XX germ cell viability (Oliver et al., 1987, 1993; Nöthiger et al., 1989). If one accepts the parsimonious suggestion that both ovo⁺ and the somatic signal affect the same process, there must be a link upstream of Sxl⁺, because Sxl⁺ is not required for XX germ cell viability (Schüpbach, 1985). There are two simple models consistent with data showing both ovo⁺ and the somatic signal upstream of this branch. In the first model the somatic sex determination signal and the number of X chromosomes are interpreted by a group of gene products and a common output regulates both the sexual identity and viability of the XX germ line. The role of ovo⁺ in this model could be the regulation of genes involved in signal reception; making them competent to receive somatic sex determination signals. The fact that XY germ cells are refractory to female somatic sex determination signals (Steinmann-Zwicky, 1994) can be explained by this model. In the second model the two principal cues act independently on both the XX viability pathway and the sexual identity pathway. While, this model is more complicated, it explains some genetic data better than the first model. In particular, there are dose-dependent dominant genetic interactions between ovo^D mutations and somatic line Sxl⁺ function leading to partial suppression of the ovo^D phenotype, and between ovo^D and either otu or snf leading to an enhanced ovo^D phenotype (Oliver et al., 1990; Pauli et al., 1993). Dominant genetic interactions of the latter type are more likely to occur between mutant alleles of genes whose wild-type products regulate a common downstream function (cf. Cline, 1988; Clark, 1991; Szathmáry, 1993). The second model preserves the close link between ovo⁺ and snf⁺. While we do not know the nature of the interaction between ovo^D and snf⁻, the snf⁺ gene encodes a U1A snRNA binding protein (Flickinger and Salz, 1994), suggesting that the snf⁺ role in Sxl⁺ pre-mRNA splicing is direct (Albrecht et al., 1993; Bopp et al., 1993; Oliver et al., 1993). The female lethal(2)d (fl(2)d) gene is also required for Sxl⁺ splicing and fl(2)d mutants show many of the same genetic interactions with Sxl mutations as snf⁻but do not interact with ovo mutations (Granadino et al., 1990, 1992). The Sxl⁺ function(s) lie downstream of the proposed branch in the germ-line sex determination hierarchy and are likely to control female germ-line identity.

The control of germ-line sex determination in males appears to be more simple (note also that this has been less well studied). There are no known mutations resulting in female germ-line development in XY flies, which may indicate that spermatogenesis is ground state. Nonetheless, there may be some effect of somatic sex determination signals on male germ-line development or viability as XY, tra^hs females show some reduction in XY germ cell number (this study) and as mutations in the somatic sex determination gene doublesex result in altered germ cell number in XY intersexes (Steinmann-Zwicky, 1994).

The nature of the chromosomal signal used in germ-line sex determination remains highly speculative. By strict analogy with the soma, the concerted action of proteins, with steadystate levels two-fold higher in XX germ cells as compared to XY germ cells, could elevate ovo⁺ activity in XX germ cells (cf. Cline, 1993). Identification of such partially or fully redundant counting elements is not straightforward. One potentially powerful tool for identifying components of the counting system is to look for modifiers of dominant alleles of ovo. For example, decreasing the dose of an X-linked counting element would be expected to tilt the counting system towards male development and exacerbate the ovo^D phenotype. A number of modifiers of ovo^D have been identified, including the X-linked enhancers, snf⁻and otu⁻(Oliver et al., 1990; Pauli et al., 1993; D. Pauli, B. Oliver and A. P. Mahowald, unpublished data). Combined duplications of X-linked germ-line counting elements might result in up-regulation of ovo:lacZ in males while combined deficiencies of these elements might down-regulate ovo:lacZ in females. This is the same type of effect seen on Sxl⁺ gene expression in the soma when duplications and deficiencies of the sisterless-a and sisterless-b genes are utilized (Cline, 1986, 1988).

Determining whether a gene that responds to the number of X chromosomes is also part of the counting mechanism is difficult (cf. Cline, 1988). Given that ovo⁺ is an X-linked gene, the idea that ovo⁺ participates in counting should be entertained. The striking finding that the dominant alleles of ovo cause the down regulation of ovo:lacZ strongly suggests that ovo^D products inactivate ovo⁺ in trans, but the presence of ovo⁺ is not absolutely required for ovo:lacZ activity. Therefore the other regulators of ovo⁺ can partially substitute for ovo⁺ in the regulation of ovo:lacZ. Regardless of mechanism, any autoregulation of the X-linked ovo⁺ gene would magnify the existing dose difference between XX and XY germ cells and could be used as part of a counting mechanism or for retaining stable sex determination following an initial transient chromosomal signal (cf. Cline, 1984, 1988; Keyes et al., 1992).

In summary, the ovo⁺ locus is required in XX germ cells for viability and for female sex determination and thus operates upstream of a branch in the sex determination hierarchy. The expression of ovo⁺ is higher in XX germ cells than in X germ cells. This level difference is in response to, and might be a part of, a system for counting the number of X chromosomes in the Drosophila germ line.

We thank M. Garfinkel and N. Coré for providing clones, P. Macdonald, W. Mattox and M. Pultz for fly stocks, and P. Macdonald for anti-vasa serum. We are indebted to N. Perrimon, S. Kerridge and an anonymous reviewer for critically reading the manuscript. We especially thank the A. P. Mahowald and A. Vincent labs for discussing unpublished work on the ovo locus. V. L. thanks P. Spierer (Geneva) for support. This work was supported by grants from the CNRS (ATIPE 7), ARC, FRM (to B. O.), and the Swiss Science Foundation (to D. P.).