Abstract
It has been suggested that sexual identity in the germline depends upon the combination of a nonautonomous somatic signaling pathway and an autonomous X chromosome counting system. In the studies reported here, we have examined the role of the sexual differentiation genes transformer (tra) and doublesex (dsx) in regulating the activity of the somatic signaling pathway. We asked whether ectopic somatic expression of the female products of the tra and dsx genes could feminize the germline of XY animals. We find that TraF is sufficient to feminize XY germ cells, shutting off the expression of male-specific markers and activating the expression of female-specific markers. Feminization of the germline depends upon the constitutively expressed transformer-2 (tra-2) gene, but does not seem to require a functional dsx gene. However, feminization of XY germ cells by TraF can be blocked by the male form of the Dsx protein (DsxM). Expression of the female form of dsx, DsxF, in XY animals also induced germline expression of female markers. Taken together with a previous analysis of the effects of mutations in tra, tra-2, and dsx on the feminization of XX germ cells in XX animals, our findings indicate that the somatic signaling pathway is redundant at the level tra and dsx. Finally, our studies call into question the idea that a cell-autonomous X chromosome counting system plays a central role in germline sex determination.
THE choice of sexual identity in somatic tissues of Drosophila melanogaster is a cell-autonomous decision made in response to the ratio of X chromosomes to autosomes (the X:A ratio; reviewed by Cline and Meyer 1996). The initial consequence of the X:A ratio is the unequal expression of several X-linked transcription factors that control the activity of the Sex-lethal (Sxl) establishment promoter, Sxl-Pe (Keyeset al. 1992). Only in nuclei with two X chromosomes does Sxl-Pe become activated. As a consequence of this difference in Sxl-Pe activity, nuclei with a single X chromosome choose male identity, while nuclei with two X chromosomes choose female identity.
The window of opportunity for Sxl-Pe activation is restricted to a short period of time just prior to cellular blastoderm formation (Keyeset al. 1992; Erickson and Cline 1993; Barbash and Cline 1995). Shortly thereafter, Sxl proteins are produced through an autoregulatory feedback loop whereby Sxl directs the productive female-specific splicing of Sxl (Bellet al. 1991). In females, Sxl expression serves to maintain the determined state throughout development. In males, Sxl transcripts are spliced in the default male pattern to incorporate a translation terminating exon and thus do not encode functional proteins (Bellet al. 1988; Salzet al. 1989).
In addition to maintaining the determined state, the Sxl gene orchestrates sexual development by regulating gene cascades responsible for dosage compensation and somatic sexual differentiation. In the latter case, Sxl promotes the productive, female-specific splicing of transformer (tra) pre-mRNAs (McKeownet al. 1987). Tra, together with the sex-nonspecific protein Transformer-2 (Tra-2), directs the female-specific splicing of doublesex. This generates dsx mRNAs encoding the female form of the Dsx protein, DsxF (Nagoshiet al. 1988; Burtis and Baker 1989). In males, tra is spliced in the default mode generating truncated proteins. In the absence of Tra, dsx pre-mRNAs are spliced in the default mode and the resulting mRNAs encode the male form of the Dsx protein, DsxM. The Dsx proteins are transcription factors thought to regulate an array of target genes responsible for sex-specific differentiation (Burtis and Baker 1989). However, the only known direct target for Dsx regulation are the yolk protein genes (Erdman and Burtis 1993; An and Wensink 1995).
Determination of sexual identity in the germline of Drosophila differs greatly from somatic sex determination (reviewed by Wei and Mahowald 1994). First, while germline sexual identity is chosen by the cellular blastoderm stage, sex-specific differences in the germ cells are not evident until mid- to late embryogenesis, after the primitive gonad has formed (Horabinet al. 1995; Poirieet al. 1995). Second, pole cell transplantation experiments and germline mosaics have established not necessary within the germ cells to establish germline sexual identity (Schupbach 1985; Granadinoet al. 1993; Steinmann-Zwicky 1993). Consistent with these findings, Sxl-Pe is not activated in germ cells of blastoderm embryos. Third, while Sxl is required in the female germline for proper oogenesis (Schupbach 1985), it is not understood how the Sxl autoregulatory feedback loop is activated, what role the gene plays in germline development, or what targets it regulates. Recently, Hager and Cline (1997) showed that Sxl can be ectopically expressed in the male germline without having any detrimental effect on spermatogenesis. By contrast, ectopic expression of Sxl in the male soma both feminizes the soma and can be lethal due to upsets in dosage compensation. Fourth, in the soma Sxl orchestrates sexual differentiation by controlling tra and, in turn, dsx. However, neither tra nor dsx is required within the female or male germline for normal development (Marsh and Wieschaus 1978; Schupbach 1982). Although tra-2 is essential in male germ cells for spermatogenesis, it has no known role in the female germline (Wantanabe 1975; Schupbach 1982).
Another difference between the soma and germline is that the choice of sexual identity in the female germline is believed to be non-cell autonomous. Although the proper X:A ratio is important for oogenesis in females (Schupbach 1985), XX germ cells, when transplanted into XY animals, do not undergo oogenesis (Steinmann-Zwickyet al. 1989). Instead, morphological analysis suggests that these XX germ cells assume a male identity and differentiate into early spermatocytes. By contrast, XY germ cells transplanted into XX animals have morphological characteristics suggesting that they differentiate along a spermatogenic rather than an oogenic pathway (Steinmann-Zwickyet al. 1989). On the basis of these findings it has been suggested that female germline sexual identity depends upon both cell-autonomous (e.g., two X chromosomes) and nonautonomous components (e.g., signal from the surrounding soma; see Figure 1A). Male identity, on the other hand, would be a cell-autonomous decision entirely independent of the sexual identity of the soma but dependent upon the number of X chromosomes in the germ cell (Figure 1B).
Autonomous components of the germline sex determination system have not yet been conclusively identified. Candidate genes in the female germline include ovo and ovarian tumor (otu). Mutations in both genes have germline-specific phenotypes and strong loss-of-function alleles result in a severe reduction or a complete elimination of germ cells only in females. Because of their sex-specific effects, these two genes are thought to function in germline sex determination (Oliver et al. 1988, 1990, 1993; Pauliet al. 1993; Staab and Steinmann-Zwicky 1995).
The tra → tra-2 → dsx somatic sexual differentiation pathway has been implicated in the production of the nonautonomous signal in females. Morphological studies on tra, tra-2, and dsx mutant females have suggested that the sexual identity of germ cells is not properly specified, and the gonads of these animals typically have a mixture of germ cells displaying male or female characteristics (Nöthigeret al. 1989; Steinmann-Zwicky 1994). Furthermore, Oliver et al. (1993) detected male Sxl transcripts in the germline of animals mutant in these genes. These findings led to the hypothesis that the somatic sexual differentiation pathway plays a central role in communicating sexual identity from the soma to the germline in females (Figure 1A).
A different idea about the role of the tra → tra-2 → dsx pathway in germline sexual identity derives from a study in which we used sex-specific molecular markers to assess the sexual identity of XX germ cells in tra, tra-2, and dsx mutant females (Horabinet al. 1995, results summarized in Table 1A). Only in females mutant for the sex-nonspecific gene tra-2 did we observe an unambiguous switch in the sexual identity of XX germ cells from female to male. In XX flies whose soma was masculinized by mutations in tra, the sexual identity of the germline appeared to be mixed; both female and male molecular markers were expressed. Finally, for dsx, no change in sexual identity was detected; the germline continued to express female markers when surrounded by an intersexual dsx− soma or by a soma masculinized by the dominant male gain-of-function dsx allele, dsxDom. These findings suggested that a novel tra-2-dependent pathway might be responsible for sending a feminizing signal from the soma to the germline (Figure 1C). In this pathway, tra-2 plays an essential role in directing the female-specific expression of a target gene, “z,” which in turn generates (directly or indirectly) the feminizing signal. Since dsx mutations had no apparent effect on germ cell sexual identity, we proposed that dsx is dispensable for this process. Finally, because germ cells of tra mutant females have a mixed sexual identity, we speculated that there might be another gene upstream of tra-2, called gene “q,” which can partially substitute for tra and vice versa.
With the aim of testing these different models for germline sex determination, we generated artificial gain-of-function alleles of traF and dsxF by ectopically expressing tra and dsx female cDNAs under the control of the constitutive hsp83 promoter. We then examined the effects of these transgenes on germline sexual identity in wild-type and mutant XY animals. If germline sex determination in males is independent of the sexual state of the surrounding soma (Figure 1B), there should be no change in the sexual identity of male germ cells when the soma is feminized. Contrary to this prediction, we find that XY germ cells are not irrevocably committed to male identity. Instead, these cells can be induced to assume female sexual identity in a feminized XY soma. As predicted by the model in Figure 1C, ectopically expressed traF feminizes XY germ cells by a mechanism that depends upon tra-2 function, but not dsx. Surprisingly, however, we found that ectopically expressed dsxF is also capable of feminizing XY germ cells, though less efficiently than traF. Taken together, our experiments suggest that germline sexual identity is determined by overlapping or partially redundant somatic signaling pathways. Finally, we examine the role of the germline genes ovo and otu, as well as Sxl, in responding to the somatic signal.
Previous models for somatic induction pathway of germline sex determination. A and B show a model based largely on transplantation studies for germline sex determination in females (A) and males (B). In this model, germline sex determination in females (A) depends upon a combination of somatic signaling from the tra → tra-2 → dsxF pathway and an unspecified germline cell-autonomous X/A counting system. In males, sex determination in the germline is entirely cell autonomous, depending only on the X/A ratio. C shows a model for germline sex determination in females adapted from Horabin et al. (1995). In this model, dsx is not required for female germline sex determination but instead plays a role in maintaining the determined sexual state. q and z have been changed from X and Y in Horabin et al. (1995), respectively, to avoid confusion with the X and Y chromosomes.
MATERIALS AND METHODS
Fly strains used and transgene construction: Flies were raised on standard cornmeal medium (Cline 1978) and raised at 25° or 29° unless otherwise indicated. The genetic loci used are described in more detail by Lindsley and Zimm (1992). The dsxDf allele used was dsx43 and the tra-2Df allele used was DfTRIX. The dsx stocks used were w1; dsx1/TM3 Ser and w1; dsx43/TM3 Ser. The dsxSwe stock is maintained by crossing w1; dsxDf/TM3 Ser females to w1; dsxSwe/dsxDf males. The tra-2 stocks used were w1; pr cn bw tra-2/Cyo and w1; DfTRIX/Cyo. The tra stocks used, tra1 and traDf, were obtained from J. Belote. The Sxl alleles SxlM1fm3 and SxlM1fm7 are revertant alleles of SxlM1 (Cline 1984). The complete genotypes are w1 SxlM1fm3/Bin and w1 ct SxlM1fm7/Bin. The ovo stocks, ovoD1rs1/FM3 and ovolzlG/FM3, were obtained from B. Oliver. The ovoD1rs1 allele was recombined with w1 for a stock of w1 ovoD1rs1/Bin. The otu alleles used were ct otu1 v/FM7 and y cv otu10 v f/FM6 and were obtained from R. Nagoshi. The ovo reporter line P[4B10C] was provided by B. Oliver (Oliveret al. 1994) and male germline lacZ reporter lines were provided by S. Dinardo (Gonczyet al. 1992).
Summary of somatic P[hsp83-traF] transgene expression effects on germline sexual identity
The hsp83-traF transgene was generated using a tra cDNA provided by M. McKeown. The tra cDNA was cloned behind the hsp83 promoter in a P-element vector containing the miniwhite gene, pHS83Capser (Horabin and Schedl 1993). The transgenes characterized in this analysis are lines 47.5, 5.2, and 5.4. These lines cannot be homozygosed due to the transforming effects observed in males and are maintained unbalanced. All transgenes are selected with each brooding. The hsp83-dsxF transgene line used was 26B as is described in more detail in Waterbury et al. (1999).
Reverse transcriptase (RT)-PCR analysis: Gonads were dissected from animals 0–2 days posteclosion. RNA was prepared from dissected tissues as previously described (Boppet al. 1993). Typically 20–35 mutant gonads were dissected per genotype. Samples were treated with acid phenol to minimize contaminating genomic DNA. All reverse transcriptions were done at 42° for 1.5 hr in the presence of RNAse inhibitor, RNAsin (Promega, Madison, WI).
PCR conditions for Sxl detection are as follows: 94° for 3 min followed by 30 cycles of 94° for 45 sec, 62° for 1 min, and 72° for 1.5 min. This was followed by a second round of PCR using the same reaction conditions and a downstream nested primer. For nested PCR, 1/30 of the first reaction was used as a template. Sxl oligomer sequences: Sxl reverse transcription primer located in exon 7 is 5′ CACCACGAGGACGACCTGTG 3′; Sxl PCR primers are 5′ CGTGTCCAGCTGATCGTCG 3′ located in exon 2; 5′ GATCGTGTCCAGCTGATCGTC 3′ located in exon 7; and nested primer is 5′ CACGCTGCGAGTCCATTTCCG 3′ located in exon 6.
PCR conditions for bruno detection are the same as those described for the first 30 cycles of Sxl amplification. Bruno oligomer sequences: Bruno reverse transcription primer is 5′ CGGAAGCAGGCGTCATCGTCG 3′. Bruno PCR reactions were done separately for male and female transcripts using a common downstream primer and different upstream primers for the two separate transcripts (see Figure 4D). The same pool of cDNAs were used for both bruno PCR reactions. The common downstream primer is 5′ GCGACCACATTGCGGTTGGC 3′. The upstream primer for the transcripts abundant in males is 5′ AGTTCGGTCTCGCGGAGGTG 3′ and the upstream primer for the female-specific transcripts is 5′ ATGTTCACCAGCCGCGCTTCG 3′. Detection of amplified products was done by Southern analysis as described using the Sxl cDNA MS21 (Samuelset al. 1991) or a bruno cDNA provided by P. Webster.
Immunoblot analysis: Samples were prepared for Western analysis as described (Boppet al. 1993). Polyacrylamide gels were prepared at 8% for Orb and Bruno analysis and 10% for Sxl analysis. Anti-Sxl antibody was used at a dilution of 1:10 using hybridoma line mSxl10. Anti-orb antibody was used at a dilution of 1:50 using hybridoma lines m6H4 and m4H8. Anti-bruno antibody was used at a dilution of 1:10,000 using an anti-rat antibody provided by P. Webster. Horseradish peroxidase conjugated secondary antibodies were supplied by Jackson ImmunoResearch Laboratories (West Grove, PA) and detected using the enhanced chemiluminescence system (Amersham, Buckinghamshire, UK).
β-Galactosidase (β-gal) activity analysis: Ovaries, testes, larval gonads, and whole flies were dissected in PBS, fixed in 3% gluteraldehyde, and stained for β-galactosidase activity overnight at room temperature according to the staining protocol of Bellen et al. (1989). Samples were photographed using a Nikon microphot-SA microscope.
Whole mount immunofluorescence: Samples were prepared for whole mount immunofluorescence as described previously (Boppet al. 1993). Gonads were dissected from adults 0–3 days posteclosion and fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) at RT for 20 min. Samples were blocked with 0.5% normal goat serum (NGS; Jackson ImmunoResearch Laboratories) in PBST (PBS, 0.1% Triton X-100). Anti-Sxl antibody was used at a dilution of 1:10 using hybridoma line mSxl114. Rabbit anti-Vasa antibody was provided by P. Lasko and used at a dilution of 1:500. Directly conjugated secondary antibodies were supplied by Molecular Probes (Eugene, OR; Alexa542 and Alexa488) and Jackson ImmunoResearch Laboratories (FITC and Cy3). When double staining was done, FITC and CY3 or Alexa542 and Alexa488 were used. Nuclear staining was done using Hoescht, YOPRO, or TOTO. Samples were analyzed using laser scanning confocal microscopy (Krypton-Argon laser, Biorad MRC 600, or Zeiss LSM510).
RESULTS
Expression of hsp83-traF can restore somatic function to tra− females: Previously, tra cDNAs have been ectopically expressed under the control of the inducible hsp70 promoter (McKeownet al. 1988). However, the hsp70:traF transgene requires multiple rounds of heat shock to partially rescue tra− females or feminize males. To circumvent this limitation, we used the hsp83 promoter to drive the expression of a female tra cDNA. This promoter is constitutively expressed in most tissues under normal growth conditions, and its activity can be increased by higher temperatures. Several independent transgenic lines carrying the hsp83-traF cDNA were isolated, and three of these (47.5, 5.1, and 5.4) were selected for further study.
We used these lines to test whether the hsp83-traF transgene can rescue tra mutant females. Females homozygous for the tra1 allele differentiate as males. For all three lines, we found that the transgene was able to feminize the soma of tra1 mutant females and at least partially restore fertility without heat shock (Figure 2A). We were able to establish and maintain a stable line of tra1/traDf flies with the hsp83-traF transgene as the only source of TraF protein. Although oogenesis appeared to be completely normal in these tra1/traDf; P[hsp83-traF] females, not all of them were able to lay eggs.
We also tested whether the hsp83-traF transgene could rescue the sexual differentiation defects of SxlM1,fm3 and SxlM1,fm7 females. Both of these alleles were isolated as male viable revertants of the gain-of-function Sxl mutation, SxlM1 (Cline 1984). They provide sufficient dosage compensation function to allow a small fraction of homozygous or trans-heterozygous SxlM1,fm3/SxlM1.fm7 mutant females to survive to adulthood; however, the soma of the surviving XX animals is completely masculinized and the gonads resemble pseudotestes (Cline 1984). It is thought that this somatic transformation is due to a failure to regulate tra splicing (Bernsteinet al. 1995; Yanowitzet al. 1999). In spite of the fact that the gonads of SxlM1,fm3/SxlM1.fm7 trans-heterozygous animals are masculinized, female-specific markers are still expressed in the germline (Horabinet al. 1995). To explain this result, we suggested that q might be active in the soma of SxlM1,fm3/SxlM1.fm7 flies. An obvious question is whether the hsp83-traF transgene can feminize the soma of SxlM1,fm3/SxlM1.fm7 and at the same time restore normal oogenic development (Figure 2B and data not shown). We found that the hsp83-traF transgene feminizes the soma not only of SxlM1,fm3/SxlM1,fm7 trans-heterozygotes but also of XX animals homozygous for SxlM1,fm3 or SxlM1.fm7, and in all three cases their morphology is indistinguishable from wild-type females (data not shown). The hsp83-traF transgene also promotes oogenic development in SxlM1,fm3/SxlM1,fm7 and SxlM1,fm7/SxlM1,fm7 animals and their ovaries are filled with maturing oocytes at various stages of development. However, the mature eggs are not laid, suggesting that the tra transgene does not rescue all sexual differentiation defects of these mutant females. Finally, while the hsp83-traF transgene feminizes the soma of SxlM1,fm3/SxlM1,fm3 flies, the ovaries are filled with “tumorous” chambers containing small undifferentiated germ cells (Figure 2, C–E). A similar oogenesis phenotype is found in the Sxl female sterile alleles (Boppet al. 1993), suggesting that SxlM1,fm3 is defective in some essential germline function.
The hsp83-traF feminizes the soma of XY animals and alters the development of the germline: The hsp83-traF transgene feminizes the soma of wild-type XY animals and the external morphology of the adult XY transgene flies is essentially indistinguishable from that of wild-type females. Complete somatic transformation was observed with a single copy of the transgene even at 18°. As can be seen in Figure 2F, the XY pseudofemales lack sex combs and have female genitalia and pigmentation, but are sterile. However, the body size of these animals is not altered by the hsp83-traF transgene; the feminized XY flies are similar in size to wild-type males (Figure 2F). By contrast, wild-type females are typically larger than males. This result is consistent with results of Cline (1984), which suggest that Sxl is responsible for this sex-specific difference.
Expression of the hsp83-traF transgene is sufficient to rescue tra− females and feminize the soma of XY animals. A shows a w1; tra1/traDf; P[hsp83-traF] female. The hsp83-traF transgene is sufficient to rescue all somatic defects caused by loss of tra function. These females undergo normal oogenesis. B shows the ovaries of SxlM1fm7/SxlM1fm7; P[hsp83-traF] females stained for Vasa protein (red) and a nuclear dye (green). Vasa is a germline-specific protein and is used here to designate the germ cells from the surrounding soma. C–E show the arrested development of the germline of SxlM1fm3/SxlM1fm3; P[hsp83-traF] females. The gonads were stained with anti-Vasa (red, C) and anti-Sxl antibodies (green, D). Merged image is shown in E. F shows the complete somatic feminization of a w1/BsY; P[hsp83-traF] pseudofemale on the right in comparison to the w1/BsY male on the left. Pigmentation of the abdominal cuticle of the hsp83-traF transgene male resembles that of wild-type females. G shows the whole mount immunofluorescence of the pseudo-ovaries of hsp83-traF. The germ cells were stained for Vasa protein with anti-Vasa antibodies. Proliferating germ cells are shown to be distributed along ovariole-like filaments.
The somatic tissues in the gonads of the XY transgene pseudofemales are feminized and, like XX females, are subdivided into a series of ovariole-like structures. However, these ovarioles are typically abnormal and resemble the tumorous ovarioles seen in the gonads of XX females carrying mutations in genes (for example, bam or snf) that result in an early oogenesis arrest. Within each ovariole, the somatic follicle cells surround and pinch off a series of pseudo-egg chambers that are filled with a multitude of small, largely undifferentiated germ cells (Figure 2G). When the traF transgenic lines were first recovered, we also observed ovaries in which one or more of the ovarioles had egg chambers that closely resembled the previtellogenic chambers of normal XX females. As can be seen in Figure 4, A and B, the developing egg chambers in these ovarioles contain nurse cells with polyploid nuclei and a cell at the posterior that may correspond to the oocyte. Mature but misshapen eggs were also observed but never laid. While ovarioles containing egg chambers with differentiating nurse cells were found in all of the hsp83-traF lines examined, these “normal” ovarioles were observed only infrequently (2–5% of the ovaries). For unknown reasons, the penetrance of this phenotype has decreased since the lines were first isolated (1992–1993) and is no longer observed.
hsp83-traF expression in XY flies results in the loss of male-specific germline gene expression: On the basis of morphological analysis of germ cells derived from the transplantation of XY pole cell into an XX soma, it has been suggested that the choice of sexual identity in XY germ cells is an entirely cell-autonomous process (Steinmann-Zwickyet al. 1989). If this interpretation is correct, XY germ cells in an XY soma feminized by hsp83-traF transgene should assume male identity and differentiate along the male spermatogenic pathway. However, contrary to this expectation, germ cells in the gonads of XY; P[hsp83-traF] pseudofemales typically resemble the undifferentiated cells found in XX tumorous ovaries, not early spermatocytes.
The germline of XY; P[hsp83-traF] pseudofemales does not express male-specific germline lacZ reporters. The germline β-gal activity staining patterns of three male-specific lacZ reporter constructs are shown (see materials and methods for staining procedure). The lac Z reporters used are P[606] (A–C), P[542] (D–E), and P[817] (G–I) and were initially characterized by Gonczy et al. (1992). A, D, and G, control males; B, E, and H, control females; and C, F, and I, w1/BsY; P[hsp83-traF] pseudofemales. All of the transgenes were in single copy.
Since the sexual identity of the germ cells in XY flies feminized by the hsp83-traF transgene cannot be unambiguously assessed from their morphology, we examined the expression pattern of several male germline-specific markers. If male identity is an entirely cell-autonomous decision, the XY germ cells in these transgenic males should express male germline markers. In contrast, if the sexual identity of these XY germ cells is misspecified by the hsp83-traF transgene, these markers might not be expressed. Three different male germline-specific lacZ reporters lines, P[606], P[542], and P[817], were selected for analysis (Gonczyet al. 1992). For each of these enhancer traps, β-galactosidase expression is detected in the germ cells of wild-type male testes, but not in the germ cells of wild-type female ovaries (see Figure 3). Expression of β-gal in the testes of P[606] and P[542] flies is limited to the germline stem cells and early spermatogonia (Figure 3, A and D), while β-gal expression in P[817] initiates at the spermatocyte stage (Figure 3G). As shown in Figure 3 (C, F, and I), none of the lacZ reporter genes is expressed at detectable levels in the germline of XY; P[hsp83-traF] flies. Thus, these germ cells do not appear to undergo normal male programming.
XY germ cells are feminized by the hsp83-traF transgene: Since the germ cells in the transgenic males do not express early male differentiation markers, an obvious question is whether they have assumed a female identity. To determine if this is the case, we assayed for the expression of several germline markers that are expressed differently in wild-type male vs. female germ cells.
Sxl: The first sex-specific marker we examined was the Sxl gene. In males, Sxl mRNAs are spliced to include a male-specific exon, exon 3, and no Sxl protein can be detected in the germline of wild-type males. In contrast, Sxl protein is present in the germline of wild-type females and the Sxl mRNAs are spliced to exclude the male-specific exon.
To assay for Sxl proteins in the germline of XY; P[hsp83-traF] pseudofemales, we probed whole mounts with anti-Sxl antibodies. Sxl protein was detected in many of the germ cells in the tumorous ovaries, though the level of staining was typically weak (not shown). In the normal XY ovarioles, Sxl protein expression was stronger and showed the characteristic female distribution pattern (Figure 4, A and B; see Boppet al. 1993). As expected, the surrounding soma, having an XY karyotype, did not express Sxl protein (Figure 4B). We next used a RT-PCR assay to examine the splicing of Sxl-Pm mRNAs in the soma and germline of the XY; P[hsp83-traF] pseudofemales. In this assay, the Sxl amplification products from wild-type males are nearly 200 bp larger than the female products because of the inclusion of the male-specific exon. As shown in Figure 4C, only the larger, male-specific amplification products are detected in the soma of gonectamized XY animals feminized by the hsp83-traF transgene. In contrast, a band corresponding to the female Sxl amplification product is observed in RNA isolated from the tumorous ovaries of the XY transgene males. These findings indicate that the hsp83-traF transgene induces female-specific Sxl products in XY germ cells.
bruno: The bruno gene encodes a germline-specific RNA binding protein thought to be involved in translational repression (Websteret al. 1997). As diagrammed in Figure 4D, bruno has two distinct promoters that give rise to transcripts having different 5′ exons and protein coding properties (Websteret al. 1997). The downstream promoter is only active in the female germline and bruno mRNAs derived from this promoter can be detected by RT-PCR amplification in RNA isolated from ovaries but not testes (Figure 4E). This female-specific mRNA encodes a Bruno isoform of ~60 kD that is also observed exclusively in the female germline (see Figure 4F). A somewhat different result is obtained for the upstream promoter. Transcripts derived from this promoter can be detected by RT-PCR amplification in RNA samples from both testes and ovaries (data not shown). Although this result implies that the upstream promoter must function in both sexes, Northern analysis indicates that the level of activity is unequal and transcripts from the upstream promoter are readily detected only in testes RNA (Webster et al. l997). Consistent with this Northern data, the ~80-kD Bruno protein isoform encoded by the message expressed from the upstream promoter is observed only in testes (see Figure 4F). Bruno is introduced here as an additional marker for germline sexual identity.
4 The germline of XY; P[hsp83-traF] pseudofemales expresses female-specific Sxl and Bruno proteins. (A and B) Whole mount immunofluorescence of Sxl protein expression in XY; P[hsp83-traF] pseudofemales showing a developing ovariole containing oogenic-like chambers. Sxl (red) is expressed in a characteristic female germline expression pattern. Nuclei are indicated in green. A is a merged image of Sxl and nuclear staining. B shows Sxl expression alone demonstrating that Sxl is present in the differentiating germline but absent from the surrounding soma. (C) Sxl RT-PCR germline analysis: Each lane is labeled and arrows indicate the sex-specific transcripts. XY and XX are control males and females, respectively. Sxl RT-PCR products of XY; P[hsp83-traF] are shown as gonectamized pseudofemales (s) and dissected gonads (g). (D)Bruno RNA splicing: Rectangles represent exons and the lines connecting the rectangles represent intron splicing. The position of the primers for PCR are depicted by the arrows below the exons (see materials and methods). The middle exon is the female-specific exon. The diagram of bruno is not drawn to scale and does not show the entire bruno transcript (see Websteret al. 1997 for more details). (E) bruno RT-PCR analysis demonstrating that the female splice is sex specific. Female product is detected in the germline of XX; dsxSwe/dsxDf, but not in XX; tra-2/traDf pseudomales. The genotypes of each sample are indicated above each lane. The arrow indicates the amplified female-specific spliced product. The upper band is amplified genomic DNA containing the intronic sequence. Male bruno product was amplified from all samples using the same cDNA preparations (data not shown). (F) Western analysis of Bruno protein expression: Wild-type males (XY) and females (XX) produce sex-specific Bruno isoforms. XY; P[hsp83-traF] pseudofemales express the female-specific Bruno protein isoform. Arrows indicate the sex-specific protein isoforms.
We have found previously that the expression of both Sxl and orb in the female mode in XX germ cells requires tra-2 but not dsx function (Horabin et al. l995). Thus, we tested whether the expression of female bruno gene products in the germline of XX animals exhibits a similar dependence on these two somatic genes. As shown in Figure 4E, tra-2 is required for the expression of the female-specific bruno mRNA; no transcripts from the downstream promoter were detected in the germline of tra-2− pseudomales. On the other hand, when the soma of XX animals was masculinized by a gain-of-function dsx allele, dsxSwe (in dsxSwe/dsxDf), which constitutively expresses the DsxM protein, female bruno transcripts were still observed (Figure 4E). These results support our earlier conclusion that tra-2, but not dsx, is required for feminization of an XX germline (Horabinet al. 1995).
Additional markers are expressed in the female mode in the germline of XY; P[hsp83-traF] pseudofemales. (A) Western analysis of Orb protein expression in control females (XX), control males (XY), and XY; P[hsp83-traF] pseudofemales. Orb has sex-specific protein isoforms: females produce a doublet of Orb protein and males produce a single isoform that migrates immediately above the upper female isoform. The female-specific Orb protein doublet is expressed in the germline of hsp83-traF pseudo-ovaries. (B) Whole mount Anti-Orb in situ analysis in the germline of XY; P[hsp83-traF] pseudofemales. Shown in the figure is a region of the pseudoovaries equivalent to the female germarium. (C–E) β-gal activity staining pattern from the lacZ reporter P[4B10C] is expressed only in the germline of females and hsp83-traF pseudofemales. (C) w′/Y; P[4B10C]/CyO control males, (D) w′/Y; P[4B10C]/CyO control females, (E) w1/BsY; P[4B10C]; P[hsp83-traF] pseudofemales. All transgenes are in single copy.
We used RT-PCR and Western blots to examine the expression of the bruno gene in the germline of XY; P[hsp83-traF] pseudofemales. As would be expected if the transgene feminizes the germline of these XY flies, we observed both female bruno transcripts (data not shown) and the female Bruno protein isoform (Figure 4F).
Orb: Like Bruno, there are sex-specific Orb protein isoforms that differ in their N-terminal domain (Lantzet al. 1994). As illustrated in Figure 5A, there is an Orb protein doublet of ~108 kD in the germline of wild-type females. This doublet is not observed in wild-type testes, and instead a single Orb protein species migrates just above the upper female isoform (Lantzet al. 1994). As found for both Sxl and bruno, the hsp83-traF transgene switches the pattern of orb expression from male to female in XY germ cells, and the characteristic female doublet is observed in Western blots (Figure 5A). Furthermore, the pattern of Orb protein accumulation in the early germ cells of these sex-transformed males closely resembles that found in females (Figure 5B; see Lantzet al. 1994).
ovo-LacZ reporter: One of the ovo-lacZ transgenic lines described by Oliver et al. (1994), P[4B10c], is expressed in the germline of adult females, but is not expressed in the germline of adult males. As shown in Figure 5, C–E, this female-specific lacZ reporter is induced in XY germ cells by the hsp83-traF transgene, and high levels of β-galactosidase can be detected in the pseudo-ovaries. Thus, the female germline-specific marker P[4B10c], together with the female-specific forms of Sxl, orb, and bruno, are all expressed in the female mode within the XY germ cells of XY; P[hsp83-traF] pseudofemales.
Germline feminization by hsp83-traF requires tra expression in the soma: Results presented in the previous sections indicate that it is possible to switch the sexual identity of XY germ cells from male to female by ectopic expression of Tra protein. One hypothesis to explain this finding is that Tra activates the somatic signaling pathway that communicates female sexual identity to XX germ cells in wild-type females. However, since the hsp83 promoter is known to be active in germ cells, it also seemed possible that expression of Tra in XY germ cells, rather than in the surrounding soma, might actually be responsible for changing their sexual identity. If this were true, we should not be able to recover hsp83-traF transformants from males because feminized XY germ cells would be nonfunctional. Contrary to this expectation, our records from the initial injection of the transgene indicated that the first adult germline transmitter was male. To confirm this observation, the transgene construct was reinjected. Several new lines were obtained that originated from male embryos injected with the hsp83-traF construct. Moreover, one Go male was found in which ~70% of his progeny (87/126) were transformants. The high frequency of transformants suggests that a substantial proportion of the germline stem cells in this animal had an integrated copy of the hsp83-traF transgene. These results indicate that Tra protein expression within XY germ cells is not in itself sufficient to initiate female differentiation.
Germline feminization by hsp82-traF is tra-2 dependent: The pathway that signals female sexual identity from the soma to the germline in XX animals requires tra-2 function. If the feminizing activity of the hsp83-traF transgene is due to the ectopic induction of the same somatic signaling pathway in XY animals, it should also depend upon tra-2. To test this, we introduced the hsp83-traF transgene into XY animals that are trans-heterozygous for a tra-2 loss-of-function mutation and a deletion that removes the tra-2 gene. As an additional control, we introduced the hsp83-traF transgene into XX; tra-2/traDf animals. The latter experiment provides another test for the possibility that ectopic expression of Tra in the germline (or in some other female tissue that does not normally express Tra) feminizes the germline by a novel mechanism. We then asked whether the soma and the germline of these tra-2 mutant XY and XX animals are feminized.
Germline feminization of XY; P[hsp83-traF] pseudofemales is tra-2 dependent and dsx independent, but can be blocked by DsxM. (A) RT-PCR analysis of bruno transcripts expressed in the germline of tra-2/traDf; P[hsp83-traF] XX and XY animals and P[hsp83-traF], dsxDf/dsxSwe XX and XY animals. Each lane is indicated by the genotype above the blot. Female bruno transcripts were not detected in XX or XY transgene animals lacking tra-2 function. Female bruno transcripts were detected in XX; P[hsp83-traF], dsxDf/dsxSwe pseudomales, but not in XY; P[hsp83-traF], dsxDf/dsxSwe males. Male bruno transcripts were detected in all samples using the same cDNA preparations (data not shown). Arrow indicates the female amplified product. (B) Germline Orb protein analysis of XY; P[hsp83-traF] pseudofemales. Control females and hsp83-traF pseudofemales express the female-specific isoforms of Orb protein. Only the male isoform is expressed in hsp83-traF pseudofemales lacking tra-2 function or expressing DsxM protein. Each lane is indicated by the genotype above the blot. (C) RT-PCR analysis of bruno expression in XY; P[hsp83-traF] lacking dsx function. Each lane is indicated by the genotype above the blot. Lanes 5 and 6 represent two different P[hsp83-traF] lines (47.5 and 5.4). Male bruno transcripts were detected in all samples using the same cDNA preparations (data not shown). Arrow indicates the female amplified product and arrowhead indicates amplified genomic DNA (see legend to Figure 3).
As expected from previous tra/tra-2 epistasis experiments, the hsp83-traF transgene was unable to feminize the soma of XY or XX animals in the absence of a functional tra-2 gene. In both XY and XX animals, the external morphology of the tra-2/traDf transgenic flies is male-like. In the case of the XY flies, the overall morphology of the gonad is testes-like; however, these flies are sterile. To test whether feminization of the germline by the hsp83-traF transgene also requires Tra-2, we used RT-PCR to examine bruno expression. As can be seen in Figure 6A, female bruno transcripts cannot be detected in the germline of transgenic XX flies in the absence of tra-2 (XX; tra-2/traDf; P[hsp83-traF]). Thus, in XX flies the ectopic expression of Tra protein by the constitutive hsp83 promoter does not provide a mechanism for bypassing the normal requirement for tra-2 function in the feminization of the germline. In addition, we also found that feminization of XY germ cells by the hsp83-traF transgene also depends upon the tra-2 gene. We were unable to detect female bruno transcripts in the germline of XY; tra-2/traDf; P[hsp83-traF] flies. Furthermore, as shown in Figure 6B, only male Orb protein is observed in XY; tra-2/traDf; P[hsp83-traF] males. Similarly, only male Sxl RT-PCR amplification products were detected in the gonads of these flies (data not shown). These findings would be consistent with the idea that the hsp83-traF transgene feminizes XY germ cells by a mechanism that recapitulates, at least in part, the somatic signaling pathway normally used to feminize the germline in XX animals.
Feminization of the XY germline by hsp83-traF is independent of dsx function: Unlike tra-2, dsx is not required to signal female sexual identity to the germline of XX flies. We wondered whether the hsp83-traF transgene could feminize XY germ cells in the absence of dsx function. To answer this question, we compared the pattern of expression of the bruno, Sxl, and orb genes in XY; dsxDf/dsx− males and in XY; P[hsp83-traF], dsxDf/dsx− males. In the germline of XY dsx− males, we did not detect any transcripts from the female-specific bruno promoter, nor did we observe female spliced Sxl transcripts (Figure 6C and data not shown). In addition, only male Orb isoforms were observed (data not shown). A quite different pattern of expression was observed in the XY; P[hsp83-traF], dsxDf/dsx− males. Both female bruno and Sxl transcripts are observed in these dsx− transgene males (Figure 6C and data not shown). These findings indicate that the hsp83-traF transgene does not require dsx function to feminize the germline of XY animals. On the other hand, it should be noted that the soma is not feminized in the absence of dsx function. Instead, the external morphology of these transgenic XY flies is intersexual as is observed in dsx mutant males lacking the transgene.
DsxM can block feminization of the XY germline by hsp83-traF: When the only source of dsx function in XX flies is the DsxM protein provided by the dsxSwe allele, the soma is masculinized and the animals develop as pseudomales. In spite of the overt masculinization of the soma, the germline of these XX; dsxSwe/dsxDf flies assumes a female sexual identity (Horabinet al. 1995 and Figure 4E). Not surprisingly, XX; dsxSwe/dsxDf pseudomales carrying the hsp83-traF transgene also express female bruno transcripts (Figure 6A). However, we obtained an unexpected result when we tested the effects of this gain-of-function dsx allele on the germline feminizing activity of the hsp83-traF transgene in XY animals (XY; P[hsp83-traF]; dsxSwe/dsxDf). As shown in Figure 6A, no female bruno transcripts could be detected, while we did observe bruno male transcripts (data not shown). Similarly, only male Orb protein was observed in Western blots (Figure 6B). These findings indicate that the DsxM is able to counteract the feminizing activity of Tra protein ectopically expressed from the hsp83 promoter.
DsxF is sufficient to feminize an XY germline: The dsx gene is not essential for the feminization of XX germ cells, nor is it required for the feminization of XY germ cells by the hsp83-traF transgene. On the basis of these observations, one might expect that it would not be possible to feminize XY germ cells by ectopically expressing DsxF protein. To test this, we used the hsp83 promoter to drive expression of a female dsx cDNA (Waterburyet al. 1999). P[hsp83-dsxF]/BsY animals with two copies of endogenous dsx develop as males and are fertile with no overt effects on germline morphology. However, when both copies of endogenous dsx are removed, the hsp83-dsxF transgene transforms XY animals into pseudofemales. The external morphology of these animals closely resembles XY; P[hsp83-traF] pseudofemales (or wild-type females); however, their gonads are much less well developed. While the gonads of the hsp83-dsxF pseudofemales are attached to an oviduct, they are much smaller and do not contain the characteristic ovarioles found in hsp83-traF pseudo-ovaries (compare Figure 7D and Figure 2G). Instead, several small, disorganized, capsule-shaped chambers filled with undifferentiated germ cells are present.
To determine if the hsp83-dsxF transgene can feminize an XY germline, we examined the expression of several germline markers in P[hsp83-dsxF]/BsY; dsx−/dsxDf gonads. Surprisingly, we found that all of the markers tested were expressed in the female mode. Figure 7, A–C, shows that the gonads of hsp83-dsxF pseudofemales express the 4B10C reporter and contain both female bruno transcripts and female Orb protein. We also detected female Sxl transcripts (not shown). Although the hsp83-dsxF transgene feminizes the XY germline, both the gonadal morphology and the pattern of 4B10C expression suggest that the extent of germline feminization is less than that observed for the hsp83-traF transgene.
Ovo and otu behave differently in the germline of XY; P[hsp83-traF] pseudofemales: ovo and otu mutations have sex-specific effects on germline development. Females null for either ovo or otu suffer severe germ cell loss and their ovaries are typically agametic (Kinget al. 1986; Oliveret al. 1987; Rodeschet al. 1995; Staab and Steinmann-Zwicky 1995). By contrast, no effects on male germline development are observed. These findings have led to the idea that ovo and/or otu are critical cell-autonomous components of the germline sex determination system (Oliver et al. 1990, 1994; Wei and Mahowald 1994; Hinson and Nagoshi 1999). If this is the case, mutations in either ovo or otu might be expected to prevent the hsp83-traF transgene from feminizing XY germ cells. Alternatively, if these genes act downstream of the autonomous sex determination system, the feminized ovo− or otu− XY germ cells might behave like ovo− or otu− germ cells in XX animals.
Expression of hsp83-dsxF is sufficient to feminize an XY germline. (A and B) β-gal staining pattern from the P[4B10C] LacZ reporter in the gonads of w1, P[hsp83-dsxF]/BsY; dsx1/dsxDf pseudofemales. B is a higher magnification (×50, A vs. ×75, B) of a germ cell cluster. Insets are higher magnifications of stained cell clusters (×100 and ×150, A and B, respectively). C (left) is an RT-PCR for bruno expression and C (right) is a Western analysis for Orb protein expression. Lane 1, control females; lane 2, control males; and lane 3, w 1, P[hsp83-dsxF]/BsY; dsx1/dsxDf pseudofemales. (D) Whole mount immunofluorescence of pseudo-ovaries stained for Vasa protein.
In the case of ovo, two putative null alleles, ovoD1rs1 and ovolzlG (Oliver et al. 1987, 1993; Staab and Steinmann-Zwicky 1995), had no adverse effects on the development of the germline in XY; P[hsp83-traF] pseudofemales; the pseudo-ovaries were filled with a multitude of undifferentiated germ cells (Figure 8, A and B). On the other hand, the strong loss-of-function otu10 allele caused severe germ cell loss and four out of every five pseudo-ovaries were completely agametic (n > 30). Figure 8, C and D, shows an example of one of the pseudoovaries that contains a few remaining germ cells. In females, germ cell loss depends upon the strength of the otu allele. Thus, a weaker otu allele, otu1, which gives a tumorous ovary phenotype in females, does not have an obvious effect on the abundance of germ cells in the XY; P[hsp83-traF] pseudo-ovaries (data not shown). Similar results have been reported by Oliver et al. (1994) using hs70-traF and by Nagoshi et al. (1995) using our hsp83-traF transgene.
XY germ cells of hsp83-traF males are sensitive to loss-of-function alleles of otu but not ovo. (A and B) Whole mount immunofluorescence of the pseudo-ovaries of w1, ovoD1rs1/BsY; P[hsp83-traF] animals. The same results were observed using another ovo loss-of-function allele, ovolzlG. (C and D) Whole mount immunofluorescence of the pseudo-ovaries of ct otu1 v/BsY; P[hsp83-traF] animals. Samples were stained for Vasa protein using anti-Vasa antibodies (B and D) and a nuclear dye (A and C). Germ cells are shown to be abundant in pseudo-ovaries lacking ovo function, but severe germ cell loss is observed in pseudo-ovaries lacking otu function. (E and F) XX females lacking ovo function were stained for Sxl (E) and Vasa (F) protein. Several Vasa-positive cells also stain for Sxl protein. The genotype is w1 ovoD1rs1/w1 ovoD1rs1.
We next asked whether ovo or otu plays a role in determining the sexual state of the germline. The ovaries of females null for ovo (D1rs1 or lzlG alleles) are primarily agametic. However, on rare occasions germ cells survive to adult stages, but do not differentiate into oocytes (Oliveret al. 1994). We stained adult surviving ovo null XX germ cells with anti-Sxl antibodies and discovered that Sxl protein is expressed (Figure 8, E and F). While this result is inconsistent with the idea that ovo is critical for establishing or maintaining female identity in XX germ cells, the pattern of expression of several sex-specific markers in the germline of ovoD1rs1/BSY; P[hsp83-tra5]/+ suggests that ovo might have an ancillary or redundant role in this process. As would be expected if sexual identity of ovo− XY germ cells could still be altered by the hsp83-traF transgene, the male germline enhancer trap P[606] (see above) is not expressed (Figure 9B). However, the ovo− germ cells were less feminized than in wild-type XY; P[hsp83-traF] pseudofemales. We were unable to detect female Sxl or bruno transcripts. Instead, only male mRNA species were observed (data not shown). Expression of Orb protein, on the other hand, was unaffected by the loss of ovo function; female, not male, isoforms of Orb protein were found (Figure 9A).
Ovo is not required for germline feminization of XY; P[hsp83-traF] pseudofemales. (A) Orb Western analysis of germline protein expression in w1 ovoD1rs1/BsY; P[hsp83-traF], ct otu1 v/BsY; P[hsp83-traF], and y w1 Sxl7BO/BsY; P[hsp83-traF] pseudofemales. Female isoforms are shown to be detected in all three genotypes. The male lane was overloaded here. (B) β-gal activity staining pattern of the male-specific lacZ reporter P[606] in hsp83-traF pseudofemales. See Figure 4 for wild-type patterns.
A similar analysis for otu10 in XX animals is complicated by the fact that germ cells do not survive to adult stages while the number of surviving germ cells in otu10 XY pseudofemales is insufficient for analysis. Thus, we could examine only the effects of a tumorous allele, otu1. Neither Sxl nor bruno female transcripts were detected in the germline of otu1/BSY; P[hsp83-traF] pseudofemales (data not shown). However, female Orb isoforms were produced (Figure 9A).
Sxl does not control all aspects of sex determination in the germline: We next asked if the feminization of XY germ cells by the hsp83-traF transgene depends upon Sxl function. As was observed for ovo and otu, the germ cells appear to have a mixed identity as judged by the orb and bruno expression pattern. We found that Sxl7BO/BSY; P[hsp83-traF] germ cells express female Orb protein isoforms (Figure 9A) and male bruno transcripts.
DISCUSSION
Transplantation experiments and clonal analysis have suggested that germline sexual identity in XX animals depends upon a combination of cell-autonomous factors that somehow assess the X/A ratio and nonautonomous factors that signal sexual identity from the soma to the germline. A plausible pathway for linking somatic sexual identity to the mechanism that generates the nonautonomous signal is the well-characterized Sxl → tra/tra-2 → dsx cascade. In previous studies, we tested the effects of mutations in tra, tra-2, and dsx on the sexual identity of germ cells in XX animals (Horabinet al. 1995). Unexpectedly, only in the case of the sex-nonspecific gene, tra-2, did loss-of-function mutations lead to a switch in sexual identity of the XX germ cells from female to male (Table 1A). To account for these findings, we proposed that the somatic signal must be generated by a novel tra-2-dependent regulatory cascade (see Figure 1C). Since dsx is dispensable for this process in XX animals, we postulated that an unidentified tra-2 regulatory target, z, directly or indirectly generates the signal. To explain the fact that XX germ cells retain partial female identity in tra mutants, we suggested that there must be another gene q whose activity overlaps or is redundant with tra. In this view, both tra and q would be able to function with the cofactor tra-2 to promote the female-specific expression of z (see Figure 1C).
Here, we have tested this model by introducing transgenes that ectopically express the female forms of tra and dsx into XY animals and by assaying their effects on germline sexual identity. While our findings are generally consistent with predictions of our original model, there were some unexpected results that altered our understanding of the nature of the germline sex determination process and the role of dsx (compare Figure 1C and Figure 10). Experiments with the tra transgene are considered first.
Tra: According to our model, ectopically expressed tra is predicted to activate the regulatory cascade that signals female identity from the soma to the germline (Figure 10). Activation of this signaling pathway should require tra-2 and the target gene z, while dsx would be dispensable. Our results are generally consistent with these predictions. Ectopically expressed Tra switches the sexual identity of germ cells in XY animals from male to female, turning off male-specific germline markers and inducing female-specific markers. This switch in sexual identity is blocked by mutations in tra-2, but is not prevented by loss-of-function mutations in dsx (see Table 1B).
Revised model for somatic induction pathway of germline sex determination. The somatic pathway for signaling to the germline requires tra-2. The unknown genes are indicated by q and z. Although tra is not essential for proper expression of female-specific germline markers, its role is thought to be redundant and is placed on both sides of the pathway. Gene q is presumed to be a tra-like molecule that would be expressed in a sex-specific manner to interact with tra-2. Gene z is a potential target of q/tra-2 and/or tra/tra-2 regulation. In the model, both dsx and z would be sufficient for germline feminization although we do not know the direct targets of either gene. One possibility is that z and dsx may interact on the same target feminization signal (“fes”), which then signals in some fashion to the germline. While the dsx pathway is not necessary for proper sex determination in the germline, it appears to be required for maintaining the proper developmental decision.
In XX animals, the available evidence indicates that the tra/q → tra-2 → z feminization pathway functions in the soma. Hence, an expectation of our model is that ectopically expressed Tra would also feminize XY germ cells through its action in the soma, not in the germline. However, since the hsp83 promoter is known to be active in both soma and germline, it is possible that Tra protein ectopically expressed in XY germ cells feminizes these cells by a novel mechanism that is independent of the somatic signaling pathway that normally operates in XX animals. Two lines of evidence argue against this. First, since several of our hsp83-traF lines were recovered from males, it would appear that expression of Tra in XY germ cells is not in itself sufficient to feminize these cells. Second, the available evidence suggests that ectopically expressed Tra feminizes the germline in XY animals by a pathway resembling that used in XX animals; it requires tra-2 but is independent of a functional dsx.
Although the hsp83-traF transgene does not require dsx to feminize the germline of XY animals, feminization can be prevented if the only source of Dsx protein is provided by an allele that constitutively expresses DsxM. This result was unexpected since DsxM has no effect on the sexual identity of the germline in XX animals. There are several possible explanations for this discrepancy. We have proposed that there is another gene, q, which occupies the same position in the regulatory cascade as tra (Fig. 10). If this gene is downstream of Sxl, as expected, it would be expressed in the male mode in XY; P[hsp83-traF] animals and hence would not contribute to the production of the feminizing signal. Because DsxM alters the development of the soma surrounding the germline and consequently the cell-cell contacts between soma and germline, the signal produced by tra alone might not be sufficient to feminize. A second possibility is that XY germ cells are intrinsically less responsive to the feminizing signal than XX germ cells. For example, given the lack of strong evidence for germline dosage compensation, the signal could be sensitive to a twofold difference in X-linked genes (see below for further discussion). A third possibility is that DsxM produces a masculinizing signal that is able to counteract the effects of the feminizing signal produced by TraF. At the present, we cannot distinguish between these explanations.
Dsx: Since the dsx gene can be removed or expressed exclusively in the male mode without affecting germline sexual identity in XX animals, we previously suggested that dsx had no role in germline sex determination. However, contrary to this suggestion, we found that ectopic expression of DsxF in XY; dsx− animals can feminize the germline and that this feminizing activity can be blocked if DsxM is also present in the soma.
Why is DsxF capable of feminizing XY germ cells, yet dispensable in XX animals? One way to reconcile these two observations is to postulate that z regulates the synthesis of the feminizing signal (“fes”) instead of encoding the signal itself. If this were the case, both Z and DsxF could independently promote the production of fes. In females, since q and tra would be active, Z would be able to induce sufficient levels of fes to feminize the germline in the absence of DsxF or in the presence of DsxM (Figure 10). Furthermore, since the female and male Dsx proteins recognize the same target sequences, ectopically expressed DsxF would be able to activate fes synthesis in XY animals only when DsxM is absent.
Somatic signaling pathway: In the revised model for the somatic signaling pathway (Figure 10), we have placed Sxl at the top of the regulatory cascade where it is responsible for activating the female-specific expression of both tra and q. While Sxl is known to be required for sex-specific regulation of tra, it should be noted that we have no evidence that it is responsible for controlling the activity of q. However, unless q is itself a target for the X/A counting system, there are no other known mechanisms that could promote female expression. If q is downstream of Sxl, results with SxlM1,fm3 and SxlM1,fm7 suggest that q is regulated by a different mechanism than tra. q and/or tra, together with tra-2, would then activate the female-specific expression of z and dsx. The female products of z and dsx would in turn direct the synthesis of the feminizing signal. By this model, the germline would assume male identity whenever Sxl is off in the soma. However, it is not clear whether the male pathway requires production of a male somatic signal by the male form of Dsx (or Z) or occurs in XY germ cells by default in the absence of a female signal. In favor of the former possibility is the finding that constitutively expressed DsxM prevents Tra from feminizing XY germ cells. On the other hand, functional dsx is not required in XY animals to select male identity.
One interesting feature of the proposed signaling pathway (Figure 10) is that several steps have redundant or partially redundant regulatory factors. Redundancy in critical signaling pathways is not without precedence. For example, in Caenorhabditis elegans, gld-1 and gld-2 are redundant in regulating meiosis. Both promote entry of germ cells into meiosis; however, they differ in their postmeiotic functions (Ferguson and Horvitz 1989; Kadyk and Kimble 1998). A similar scenario in which genes have overlapping but distinct activities may certainly apply to the two regulatory pairs we have suggested, q and tra and z and dsx, to function in germline sex determination.
Nonautonomous vs. autonomous: One question raised by our studies is the role of the postulated autonomous X chromosome counting system in germline sex determination. In particular, it has been argued from pole cell transplantation experiments that this autonomous system overrides input from the soma in XY germ cells, forcing them to assume male identity (see Figure 1). However, we have presented data indicating that the sexual identity of XY germ cells can be switched from male to female by ectopic expression of TraF and DsxF. If TraF and DsxF activate the signaling pathway(s) that normally functions in XX animals, this result would imply that there may be no cell-autonomous system that selects sexual identity by measuring the germ cell X/A ratio. In this view, the autonomous components of the germline sex determination system would play an entirely different role. They would be subordinate to the somatic signaling pathway, being responsible only for responding correctly to the somatic signal and having no role in making the actual choice. Of course, if the default pathway within the germline is male, then this pathway will be followed “autonomously” in the absence of a feminizing signal from the soma.
From a phylogenetic perspective, the simplest solution for germline sex determination is that germ cells strictly follow the same sexual fate as that of the soma securing the development of a fully functional organism. In fact, this appears to be the mechanism for germline sex determination in other dipteran species such as Musca domestica (Hilfiker-Kleineret al. 1994) and Chrysomya rufifacies (Ullerich 1984). In these organisms, somatic sex alone is necessary and sufficient to dictate sexual fate to the germline. Irrespective of their sexual karyotype, when germ cells are surrounded by ovarian tissue, eggs are produced, and when surrounded by testicular tissue, sperm are produced. Studies in the nematode C. elegans and in the mouse further support the idea that somatic sex is widely used to dictate the sexual fate of gametes (McLaren 1995). In Drosophila, we believe that somatic sex is the primary determinant. Why then is this soma-to-germline signaling mechanism insufficient to direct complete female or male germline differentiation independent of the chromosome composition of the germ cells in Drosophila? A likely explanation is that XY and XX germ cells in Drosophila have lost the ability to respond equally well to somatic cues. For example, in XY; P[hsp83-traF] pseudofemales, most of the ovarioles had a tumorous ovary phenotype and ovarioles that had normal-looking egg chambers were observed very infrequently. Given that there is no strong evidence for germline dosage compensation, one plausible explanation for the abnormal development of these sex-transformed XY germ cells is that the dose of X-linked gene products is insufficient to properly execute an oogenic developmental program. The Sxl gene would be a good example of a gene that is required for oogenesis and, because of its autoregulatory activity, is highly sensitive not only to its own dose but also to the dose of other X-linked genes such as the splicing factor snf (Hager and Cline 1997). It is reasonable to suppose that there may be a variety of steps in oogenesis (or spermatogenesis) that are sensitive to the dose of X-linked genes.
Components within the germline: Within the germline, otu, ovo, and Sxl have been identified as candidate genes that respond to the feminizing signal from the soma and determine the sex of the germ cells. Mutations in all three genes have sex-specific effects on germline development. Perhaps the most striking result is the fact that loss-of-function ovo and otu mutations markedly reduce the viability of XX but not XY germ cells. Thus one important question is whether these mutations have similar effects on the viability of XY germ cells feminized by the hsp83-traF transgene. Somewhat surprisingly, we found that otu and ovo mutations behave differently. Strong loss-of-function otu mutations reduce the viability of XY germ cells feminized by the traF transgene. This finding suggests that the lethal effects of strong otu mutations arise because the germ cells assume a female identity, and not because of their number of X chromosomes. In contrast, ovo mutations had no apparent effect on the viability of feminized XY germ cells. One explanation for this difference is that lethal effects are not observed in ovo mutants because the feminizing signal produced by the traF transgene in XY animals is weaker than the feminizing signal found in wild-type XX animals. Alternatively, it is possible that XX germ cell death in ovo mutants does not depend upon the choice of sexual identity, but rather is a function of the X chromosome dose (see Oliveret al. 1994).
If otu, ovo, or Sxl functions as a master sex determination switch within the germline, one would expect to find that mutations in these genes would completely block the feminization of germ cells much like mutations in Sxl prevent feminization in the soma. While our results indicate that none of these genes fits this criterion for a master regulatory switch, we do observe effects on the expression of sex-specific markers. Mutations in all three genes prevent the traF transgene from inducing the expression of female bruno (and Sxl) gene products. On the other hand, in all three cases the transgene still induces the expression of female orb gene products. One interpretation of these findings is that the sex determination pathway in the germline is split into at least two branches, one that contains bruno and Sxl and another that contains orb. For both bruno and orb, sex-specific expression depends upon the activation of distinct sex-specific promoters. If these two genes are in independent branches of the germline sex determination pathway, this would imply that there must be distinct “male” and “female” transcription factors for the four promoters. Moreover, it seems likely that one important function of the somatic signaling system would be to control the expression of these transcription factors. Clearly, it will be important to identify these transcription factors and to learn how they are regulated.
Acknowledgments
The authors thank T. Schupbach, G. Deshpande, S. West, and J. Yanowitz for helpful discussions, J. Goodhouse for confocal support and advice, B. Oliver for the 4B10C reporter line, and S. Dinardo for the male germline-specific LacZ reporter lines P[606], P[542], and P[817]. This work was supported by a grant from the National Institutes of Health. J.A.W. was supported by a fellowship from the New Jersey Cancer Commission.
Footnotes
-
Communicating editor: K. Anderson
- Received November 3, 1999.
- Accepted April 14, 2000.
- Copyright © 2000 by the Genetics Society of America