Sex determination in the germ line of Drosophila depends on genetic signals and inductive somatic factors*

In the fruitfly Drosophila melanogaster, sex is determined by the ratio of X-chromosomes to sets of autosomes. A ratio of 2X: 2A resuls in female, a ratio of IX: 2A in male, and a ratio of 2X:3A in intersexual development (Bridges, 1925). The primary signal of the X: A ratio acts on the key gene Sex-lethal (Sxl, Cline, 1978,1979). This gene controls sex determination and is also involved in dosage compensation (for reviews see Nothiger and Steinmann-Zwicky, 1985; Baker et al. 1987; Cline, 1988). Dosage compensation is the process that regulates the rate of transcription of X-chromosomal genes in such a way that males with one X-chromosome and females with two X-chromosomes have equal amounts of gene products (for review see Lucchesi and Manning, 1987). In the pathway that leads to somatic sexual differentiation, the state of activity of Sxl is transmitted, via a short cascade of subordinate control genes, to the tissue-specific sex-differentiation genes. The differential activity of these genes depends on the subordinate control genes tra (transformer), tra-2 (transformer-2), dsx (double-sex), and ix (intersex) (Bownes and Nothiger, 1981; Belote et al. 1985; Schafer, 1986; Di Benedetto et al. 1987; Chapman and Wolfner, 1988; reviewed in Wolfner, 1988).

In contrast to our detailed knowledge about the genetic control of sex determination in the soma, relatively little is known about the mechanisms that determine the sex of the germ line. Most of the information derives from transplantation of pole cells which are the progenitors of the germ line. These cells can be transplanted from donor embryos into host embryos where they may be incorporated into the developing gonad (Illmensee, 1973; Van Deusen, 1976). Transplantations of pole cells that were mutant for tra-2, tra, ix or dsx have shown that these genes are not needed for the sexual differentiation of germ cells (Marsh and Wieschaus, 1978; Schüpbach, 1982). The role of Sxl was unclear although its function is obviously required in XX germ cells for proper oogenesis, perhaps to regulate dosage compensation (Schüpbach, 1985). Only recently was it shown that Sxl has a sexdetermining effect in germ cells (Steinmann-Zwicky et al. 1989).

Chromosomal females (XX) with the mutant genotypes tra/tra or tra-2/tra-2 or dsx^D/ Df (dsx) are transformed into so-called ‘pseudomales’ (Sturtevant, 1945; Watanabe, 1975; N Öthiger et al. 1980; Nôthiger et al. 1987). These pseudomales are sterile even when a Y-chromosome is present. Their gonads are testes that are best described as rudimentary, uncoiled and small. Brown and King (1961) and Seidel (1963) who studied the gonads of X/X; tra/tra pseudomales found that these contained mostly degenerating cells, but they also observed oogenic stages. In addition, Seidel (1963) noticed that the germ cells in testes of X/X; tra/tra pseudomales could also form spermatocytes. The same germ cells, when transplanted into female hosts, consistently produce normal eggs (Marsh and Wieschaus, 1978). These results suggest that the gonadal soma has an inductive influence on developing XX germ cells.

The aim of our investigation was to analyze the influence of the sex of the soma, of the X: A ratio and of mutations at Sxl on the sexual development of germ cells. For this purpose, we constructed a number of genotypes with 2X;2A or 2X;3A karyotypes that carried various mutations at sex-determining genes, such as Sxl, tra-2, tra, or dsx. In particular, we asked the following questions, (i) Which sexual pathway do the germ cells choose under the various experimental conditions, and to what stages of gametogenesis do they proceed? (ii) How do the X: A ratio, the key gene Sxl, the somatic sex-determining genes, and the genetic background interact to determine the sexual development of germ Une and soma? (iii) To what extent do the gonadal soma and the germ cells mutually influence each other in their development?

Our analyses provide evidence in support of the conclusions drawn by Steinmann-Zwicky et al. (1989) that the sex of the germ cells is determined by the state of activity of Sxl which in turn appears to depend on the X: A ratio and on the phenotypic sex of the soma.

A large number of 2X;2A and 2X;3A animals were constructed that were sexually transformed into pseudomales by mutations at various sex-determining genes. We also studied 2X;3A animals that were normal triploid intersexes and others that were transformed into phenotypical females (pseudofemales). To grasp the results, it is sufficient to study the figures.

Mutations at the loci tra-2 (2 –70), dsx (3 –48.1), or tra (3 –45) allow the construction of three types of pseudomales (=VCT) : (i) X/X; tra-2/tra-2, (ii) X/X; dsx° / dsx or dsx^MaS/dsx (Nôthiger et al. 1980; Baker and Ridge, 1980) and (iii) X/X; tra/tra. In these three types, gametogenesis was analyzed in the gonads of 2137 individuals chosen from 24 different genotypes. Most of these genotypes carried marker mutations such as w or cn bw or st p^p that rendered the testes colorless and transparent so that the germ cells could be inspected in intact testes. Genotypes with attached X-chromosomes were constructed to minimize the possibility of gonial nondisjunction or loss of an X-chromosome. lite detachment frequency was below 1% for any of the used in these experiments. Table 1 shows the complete genotypes and how these were pooled into 10 classes for Table 2 and into 3 classes for Fig. 3. Genotype 11 carries two partially functioning alleles of Sxl (1-19.2); and genotype 12 carries the constitutive allele Sxl^M1, but is a pseudomale because it is homozygous for tra.

For genetic symbols see Lindsley and Grell (1968). The alleles of Sxl were described by Cline (1978; 1984); Sxl^fl results in loss of function, Sxl^Ml in gain of function of Sxl; in this paper, these alleles will be designated as Sxf and Sxl^M, respectively. Two of the deficiencies for tra, Df(3L)st^E23 and Df(3L)st^E73, were generated by X-rays; they extend from 72E1-5 to 74DE and to 74A1, respectively (Butler et al. 1986). Df(3L)st^ss103 was a gift of Dr M. Ashbumer; it extends from 73A1-2 to 74B1-3. The gene tra was localized to 73A9-10 (Butler et al. 1986).

Triploid intersexes (2X;3A) were constructed by crossing chromosomally normal diploid females to males carrying compound second and compound third autosomes C(2L)RM; C(2R)RM; C(3L)RM; C(3R)RM (Fig. 1). The only viable progeny of such a cross are triploid intersexes, and rare triploids females resulting from non-disjunction of the maternal X-chromosomes. Metamales (1X;3A), haploid and aneuploid animals are lethal. The different alleles of Sxl present in some of the 2X;3A animals were introduced from the female parent. Genotypes 13 to 16 are phenotypically male; in this paper, they will be called 2X;3A pseudomales. Genotypes 17 and 18 are triploid intersexes whose soma is a mosaic of male and female cells. Flies of genotype 19 are transformed into pseudofemales by Sxl^M.

As a first step, we constructed a compound chromosome C(3L)RM, mwh tra/mwh tra by irradiating spermatogonia! cells in pupae of the genotype B^SY/y w; mwh tra/mwh tra with 2500 rad, and crossing the emerging adult males to virgins of the compound-3 stock C(3L)RM, st/st; C(3R)RM, p^p/p^p. Out of this cross, B^SY/X males that were phenotypically mwh and p^p were selected. These males carry a new C(3L)RM, mwh tra/mwh tra; C(3R)RM, p^p/p^p. The newly generated C(3L)RM was kept by continuous backcrosses of male carriers to virgins of the compound-3 stock mentioned above.

From now on, we will use the abbreviated form C(3L)RM, mwh tra to denote a C(3L)RM that is homozygous for mwh tra. Other compound chromosomes being homozygous for other mutations will be similarly abbreviated. To introduce our new C(3L)RM, mwh tra into a stock with compound-2 and compound-3 chromosomes, we generated a new compound-2 by irradiating pupae of the genotype B³Y/X; C(3L)RM, mwh tra; C(3R)RM, p^p with 3000 rad, and crossed the adult males with females of genotype y²/y²; C(2L)RM, dp; C(2R)RM, px; C(3L)RM, h; C(3R)RM, +. Out of this cross, males with the genotype B^sY/y²; C(2L)RM, dp; C(2R)RM, px; C(3L)RM, mwh tra; C(3R)RM, + were selected. A stock was maintained by continuous selection of such males, which were then crossed to virgins carrying compound-2; compound-3 chromosomes.

For the production of the experimental 2X;3A animals, we crossed our compound males B^sY/y²; C(2L)RM, dp; C(2R)RM; px; C(3L)RM mwh tra; C(3R)RM, + with virgins of genotype y w/y w; mwh tra/TM3, Sb Ser. Our cross produced the two types of 2X;3A animals listed in Table 1 as genotypes No. 13 (pseudomales, phenotypically y²mwh) and 17 (intersexes, phenotypically y²Sb Ser), respectively.

Flies were raised under uncrowded conditions at 25 ° C in bottles containing standard Drosophila medium (cornmeal, sugar, yeast, agar, Nipagin). The cultures for 2 ×;3A animals were kept at room temperature (20 – 22 ° C). The cultures producing 2 ×;2A pseudomales were raised under three different temperature regimes at 18 ° C, 20 ° C, or 25 ° C. Since neither temperature nor age had any influence on the characters studied, the data were pooled in the tables and figures. Flies were collected 1 to 10 days after eclosion.

Immediately after dissection in Ringer’s solution, the gonads of the animals were squashed under a coverslip, examined with phase-contrast microscopy and photomicrographs were taken. To distinguish young nurse cells from spermatocytes, especially in yellow colored gonads, the multiple nucleoli of the nurse cells were stained with a solution of 2 % orcein-acetic acid-lactic acid=l: 1:1.

Following the indirect immunofluorescence technique described by Hulsebos et al. (1984), squashes of testes of young pseudomales of one genotype (X/X; th st tra cp ri p^p/th st tra cp ri p^p) were fixed and stained with an antiserum that recognizes a spermatocyte-specific protein. The fluorescence pattern was then compared with the pattern obtained with XO and XY spermatocyte nuclei.

Following King (1970), we used the diameters of the nursecell nuclei to classify the oogenic stages from S1 to S10A. An unequivocal identification of oocytes within a nurse-cell cyst was rarely possible. Only those cases were noted where the nucleus of the oocyte was clear and one dark nucleolus and a light karyosome were visible (Fig. 2F,G).

Germ cells were classified as spermatogenic according to morphological criteria. Primary spermatocytes have a characteristic nucleolus of spherical shape with a light center. A characteristic feature of spermatocytes are the so-called crystals, which form when germ cells without a Y-chromosome enter the male pathway (Meyer et al. 1961). The shape of these protein structures depends on the allelic state of an X-chromosomal locus, Stellate (Ste) (Hardy et al. 1984; Livak, 1984). All our X-chromosomes carry the recessive allele, which causes needle-shaped crystals (Fig. 21); only the balancer chromosome FM6 carries the dominant allele, which causes star-shaped crystals. The presence of crystals is a reliable criterion to identify spermatocytes. All spermiohisto-genic stages from spermatids to coiled, non-motüe sperm are easily classified (Fig. 2H, K, L).

With out methods, younger germ cell stages, such as oogonia or spermatogonia that lack the above mentioned oogenic or spermatogenic characteristics, could not be sexed and are classified as sexually undefinable gonia.

It will be necessary to describe briefly the effects of mutations at Sxl, the gene that regulates sex determination and dosage compensation. An active gene product is required in XX zygotes to initiate and maintain the female sexual pathway and to keep the rate of transcription of X-chromosomal genes low; the gene must be inactive in XY animals to bring about male development and to keep transcriptional activity of the single X-chromosome high (Cline, 1978, 1984). The recessive mutation Sxf causes a loss of function which is lethal to XX zygotes, because their X-chromosomes are now hypertranscribed (Lucchesi and Skripsky, 1981; Gergen, 1987); the masculinizing sex-transforming effect can be seen in cell clones or in 2X;3A animals (Sánchez and Nothiger, 1982, 1983; Cline, 1983a). The dominant mutation Sxl^M behaves as a moderate constitutive allele (gain of function) which is lethal to XY males, presumably because their X-chromosome is now hypoactive (Cline, 1978, 1984); the feminizing sex transforming effect is revealed by cell clones and in 2X;3A animals (Cline, 1979, 1983a). This description applies only to the role of Sxl in somatic cells. Its function in the germ line will be discussed later.

In view of the complicated genotypes (see Table 1), we have compiled the data in a synoptical form in Figs 3 and 5 and in Table 2.

All the pseudomales listed in Table 2 and Fig. 3 are morphologically indistinguishable from each other. Their gonads are testes whose size, shape and contents, however, varied widely (Fig. 2A-D). We have classified the gonads into non-gametogenic and gametogenic types. The first category contains degenerated germ cells and debris or gonial cells whose sex we could not determine. The second category contains sexually differentiated germ cells, which reached different stages of oogenesis or spermatogenesis (Fig. 2E-L). But neither oogenesis nor spermatogenesis were ever complete and normal in any of our experimental genotypes.

Fig. 3 subdivides the pseudomales into two groups: the left five columns are pseudomales of karyotype 2X;2A, the four columns on the right are pseudomales of karyotype 2X;3A.

The first three columns on the left represent pseudomales that are transformed by tra-2, dsx^D, or tra and whose X-chromosomes carry a wild-type allele of Sxl. Most of the testes of these pseudomales fell into the non-gametogenic category. A variable proportion (see Table 2), however, contained oogenic or spermatogenic stages of germ cell development. The variability is most likely due to variations in the genetic background rather than to different effects of mutations in the three genes tra-2, dsx and tra of which various alleles have been used in different genetic contexts.

Oogenic differentiation was arrested anywhere between stages S3 and S10A (King, 1970). In about 90% of the oogenic gonads, development was arrested between stage 4 and stage 7 (Fig. 2F), and only in about 4 % did the germ cells reach stages 8 to 10A (Fig. 2G); the rest were arrested around stage 3. Most oogenic testes contained only single nurse cells or nurse cell cysts without identifiable oocytes (Table 2). A single gonad never contained more than 10 oogenic cysts, whereas degenerated cysts were frequent. Among the different pseudomales, the frequency of oogenic gonads varied widely.

Spermatogenic differentiation was frequently arrested at the spermatocyte stage (Fig. 21), but occasionally proceeded to spermatids (Fig. 2K) or even to mature, immotile sperm (Fig. 2L). These advanced stages of spermatogenesis, however, were very rare (Fig. 3). Spermatocytes often appeared loose in the testicular lumen, especially in poorly developed testes. Intact cysts, however, always contained the regular number of 16 spermatocytes. When XX germ cells without a Y-chromosome differentiated spermatocytes, these formed the characteristic needle-shaped crystals (Meyer et al. 1961). But even in gonads with advanced stages of spermatogenesis, degenerating and degenerated cell clusters were abundant. The frequency of spermatogenic gonads varied widely and was not apparently dependent on the particular autosomal gene that was mutant (see Table 2).

Two genotypes (No. 11 and 12, Table 2) carry mutations in the gene Sxl. Genotype Sxl^M1’^fm3/Sxl^fm7’^M1 is a transheterozygote for two mutations that are marginally compatible with survival of XX animals, but whose residual functions are insufficient to bring about female development (Cline, 1984). These animals are transformed into pseudomales. Most of these testes were non-gametogenic; the few gametogenic testes (9%) contained only spermatogenic stages, and no oogenic stages were observed. In contrast, the testes of genotype Sxl^M/ +; tra/tra contained only oogenic stages and displayed no signs of spermatogenesis. Although this genotype gave the highest frequency of oogenic testes, oogenesis was arrested at S3-S6 and none of the cysts reached stage S10A.

Remember that XX pole cells mutant for tra, tra-2, dsx, or Sxl^M1’^fm3/Sxl^fm’^7,M1 can form entirely normal eggs in female hosts (Marsh and Wieschaus, 1978; Schüpbach, 1982, 1985). Our observation that these same cells when left in a pseudomale gonad can enter the spermatogenic pathway points to an inductive, sex determining influence of the somatic environment. In addition, the results obtained with genotypes 11 and 12 suggest that the state of activity of Sxl is instrumental in determining the sexual pathway of germ cells, at least in testes.

At this point, the question arises whether the spermatocytes, spermatids and sperm that we observe in XX pseudomales are really formed by XX cells. Since the frequency of spermatogenic testes of XX pseudomales is variable and sometimes very low (Table 2), one could argue that these rare cases of spermatogenesis are the result of premeiotic loss of an X-chromosome during gonial divisions. Several lines of evidence show that this is not the case.

If XO cells arise, they should develop at least to elongated spermatids and frequently to immotile sperm (Lifschytz and Hare ven, 1977), rather than become arrested at the spermatocyte stage (Fig. 3). Moreover, loss of an X-chromosome should be a rare event occurring in single gonial cells so that an individual testis should occasionally contain spermatogenic and oogenic cells side by side. Such sexually mosaic gonads, however, were not observed; rather, a single gonad contained either oogenic stages or spermatogenic stages (see Discussion).

Three further tests confirmed that XX germ cells can in fact enter the spermatogenic pathway in a testis.

(a) We used compound (attached) X-chromosomes to minimize the occurrence of X-chromosome loss. Spermatogenic gonads were still observed although the frequency did decrease, except for genotype No. 4 where it increased (Table 2). If XO germ cells are to be formed, the compound X must first undergo a detachment which must then be followed by loss of an X-chromosome. The frequency of meiotic detachment of C(1)DX was about 0.1% whereas the frequency of spermatogenic gonads in C(1)DX pseudomales averaged 2.1%. Since C(1)DX is deficient for the rRNA genes, the presence of a Y-chromosome is required for cell viability. Thus, after detachment and mitotic chromosome loss, the resulting karyotype would have to be X/Y;2A which should allow normal development of motile sperm. This was never observed. Instead, spermatogenesis was mainly arrested at the spermatocyte stage as in pseudomales with two free X-chromosomes.

(b) In pseudomales with the karyotype . /X (complete genotype: Y^sX.Y^L,In(1)EN, y/f; th st tra cp ri p^pI th st tra cp ri p^p) the loss of an X-chromosome in the germ line has morphological and physiological consequences that are easily detectable. If the single rod-X chromosome is lost, we expect to see motile sperm. In control animals bearing the unstable ring-X chromosome, ln(l)w^vC (Hinton, 1955), instead of the rod-X chromosome, this was actually the case; and one of these ring-X/ pseudomales was even fertile. If, on the other hand, the attached- is lost, we expect to find spermatocytes with crystals and immotile sperm as in XO males. Among 88 testes of our pseudomales, we found five with spermatocytes, one with spermatids and one with non-motile sperm; in none of these seven testes did we see any crystal or motile sperm. This indicates that neither the nor the X was lost, and therefore that the observed spermatogenic stages were differentiated by germ cells with two X-chromosomes (⁠⁠).

In one testis, however, loss of the attached- apparently did occur in a single gonial cell, as judged by the presence of a few spermatocytes with crystals in a small area of the gonad.

Taken together, these results strongly support our conclusion that XX germ cells can undergo spermatogenesis. On the other hand, the island of crystals observed in one case indicates that a low percentage of spermatogenesis in pseudomales with free X-chromosomes may in fact result from loss of an X-chromosome in a gonial cell.

(c) Finally, we used an antibody directed against a spermatocyte-specific protein (Hulsebos et al. 1984). This antibody, when applied to spermatocytes, produces a characteristic pattern that differs in Y/Y; +/ +, X/O; +/ +and X/X; tra/tra males and pseudomales, respectively (Fig. 4). The immunological test was made with a few testes of X/X; tra/tra pseudomales in which spermatogenesis had advanced beyond the spermatocyte stage. We want to argue that the differences in the immunofluorescence pattern reflect the sex chromosome constitution, and thus we conclude from these results that XX germ cells can enter the spermatogenic pathway.

The procedure for the construction of the 2X;3A animals is schematically shown in Fig. 1. Further details are given in Materials and Methods.

In the absence of any mutations at sex-determining genes, the ratio of 2X:3A represents an intermediate value to which somatic cells respond by differentiating a mosaic pattern of male and female structures (Stern, 1966). It appears that some cells interpret the primary signal of 2X:3A as male, others as female. We have transformed triploid intersexes into pseudomales by mutations causing loss of function of tra or Sxl. The four types of 2X; 3A pseudomales are phenotypically normal males and indistinguishable from each other. Their gonads are well-developed and nicely coiled testes (Fig. 2B). The lack of function of one Sxl allele in Sxf/+; +/ +/ +is apparently sufficient to transform even these animals into pseudomales (Cline, 1983a).

With the four genotypes shown in the right half of Fig. 3, we wanted to test the influence of the X: A ratio on germ cell development in pseudomales. Assuming that the male somatic environment can exert a sexdetermining influence on the germ cells, we expected that 2X;3A germ cells would more easily respond by male gametogenesis than 2X;2A cells. We in fact observed that most testes were filled with spermatogenic germ cells; abortive germ cell development was rare (Fig. 3). But spermatogenesis was mostly arrested at the spermatocyte stage. The spermatocytes formed typical crystals. Spermatids and sperm were not more frequently formed than in 2X;2A pseudomales. Oogenesis, however, was completely absent. Even Sxl^M was unable to direct 2X;3A germ cells into the oogenic pathway, and genotypes Sxl^M/+; tra/tra/tra and Sxf/ +; tra/tra/tra gave very similar results.

Our data show that, in the germ line, 2X:3A is essentially a male signal that does not allow oogenesis to occur in testes and that cannot be overruled by Sxl^M; alternatively, it is possible that an X: A ratio of 2:3 does not permit Sxl^M to become constitutive in germ cells that are surrounded by a (pseudo)male soma (see Discussion).

Fig. 5 presents the four genotypes listed as No. 13, 17, 18, 19 in Table 1. We here compare one class of 2X;3A pseudomales with two groups of triploid intersexes and with 2X;3A pseudofemales. We also examine the consequence of variable doses of tra⁺.

The pseudomales (genotype +/ +; tra/tra/tra) were already described in Fig. 3. They formed well-coiled testes with spermatogenic germ cells.

Flies with the genotype Sxl^M/ +/ + /+/ + are phenotypically female, but sterile (Cline, 1983a). These pseudofemales formed ovaries in which the germ cells invariably entered the oogenic pathway and, except in one case, reached the stage of mature eggs; but eggs were never laid. Most of the cysts were morphologically normal and contained 15 nurse cells plus one oocyte. In about half of the ovaries, however, we occasionally found one or a few cysts with only 8, or with 20 to 30 nurse cells. The number of ovarioles per ovary was lower (10–12) than in wild-type females (16–18).

The triploid intersexes (genotypes +/ +; tra/tra/ + and +/ +; +/ +/ +) formed testes and ovaries. Mosaic gonads, consisting of testicular and ovarian tissue, were rare (3 cases and 18 cases for the two genotypes, respectively). These 21 gonads did not contain any developed germ cells and are not further considered. As shown in Fig. 5, 2X;3A germ cells entered the spermatogenic pathway in testes whereas they entered the oogenic pathway in ovaries. It thus appears that germ cells with the chromosomal signal of 2X:3A can enter either the male or the female pathway as a reaction to somatic stimuli. There was no significant difference between the intersexes with three doses of tra⁺and those with only one dose of tra⁺ in terms of gametogenic development although the latter formed testes more frequently than the former (Fig. 5).

Another set of triploid intersexes, namely the sibs of genotypes 16 and 19, however, gave a somewhat different result. Genotypes 16 and 19 were produced by crossing y cm Sxl^M/FM6 or y cm Sxl^fct f/FM.6 females to y²/Y males carrying compound-2 and compound-3 autosomes. These crosses also yielded 10 FM6/y²; 3A as sibs of genotype 16, and 20 FM6/y²; 3A as sibs of genotype 19. In these intersexes, we found 37 testes that contained undifferentiated germ cells and spermatocytes, and two testes with spermatids; about half of the testes were rudimentary. There were 13 rudimentary ovaries which also contained undifferentiated germ cells and spermatocytes. The spermatocytes formed crystals which in these cases were star-shaped due to the Ste allele on the FM6 chromosome (see Materials and Methods).

The two sets of observations demonstrate the well-known fact that the phenotypic manifestation in triploid intersexes is very variable (Stem, 1966; Laugé, 1968; Cline, 1983a). It also enforces our previous conclusion that 2X:3A corresponds to a male signal in the germ line.

The testes as well as the ovaries of the triploid intersexes were frequently rudimentary, and the testes were only rarely coiled. This is in contrast to the well-developed and usually coiled testes formed by 2X;3A pseudomales listed in Fig. 3.

The gonads of 2X;2A and 2X;3A pseudomales show a strong correlation between form and size of gonads and germ cell development. Differentiation of germ cells is less advanced and more frequently abortive or non-gametogenic in rudimentary than in coiled testes. Non-motile sperm, on the other hand, were only found in coiled testes. Gonads with two turns are much more frequent in 2X; 3A pseudomales than in 2X; 2A pseudomales (see ‘% coiled’ in Fig. 3). The testes of 2X;3A pseudomales also contain many more healthy germ cells than the testes of 2X;2A pseudomales. A high degree of coiling is also displayed by the testes of Sxl™/ +; tra/tra in which germ cells become oogenic (Fig. 3).

The results show that the external morphology of the testes (rudimentary, coiled with one turn or with two turns) and the amount of healthy germ cells inside are strongly correlated, irrespective of the sex of the germ cells. This reveals a developmental interaction between germ cells and gonadal soma. Since gonads of agametic flies such as those produced by females homozygous for grandchildless mutations, remain small and undeveloped, we conclude that the variable degrees of coiling of the testes are the consequence of the mass of developing germ cells.

In those 2X;2A pseudomales in which both types of gametogenesis, female and male, could occur, a single gonad contained, besides degenerating cells, either oogenic or spermatogenic cells, never both. It thus appears that all germ cells in a gonad choose the same sexual pathway. Among those animals in which both gonads differentiated sexually identifiable germ cells, both testes of a pseudomale were either oogenic in 256 animals or spermatogenic in 24 animals. Only two animals had one oogenic and one spermatogenic testis. If a single gonad reacted as a unit independently of the other, the expected number of animals having one oogenic and one spermatogenic testis would be 97. The observed value is about 50-fold lower. This result clearly indicates that the germ cells of both gonads of a pseudomale choose the same sexual pathway. Since no hormonal sex-determining factors exist in Drosophila, we are at present unable to offer a satisfactory explanation for this phenomenon.

It is important to point out that we have not seen any sexually intermediate cell types, such as e.g. oogenic clusters with crystals typical of spermatocytes developing without a Y-chromosome. But we must also mention that gametogenesis in 2X;2A pseudomales was mostly abortive or arrested at sexually unidentifiable stages (Table 2 and Fig. 3). It is conceivable that these germ cells are in fact sexually intermediate, comparable to the phenotype produced by mutations at ix or dsx in somatic cells. More likely, however, the presence of abortive cells is the result of a developmental mismatch between the sex of the germ cells and the sex of the gonad, or between the sexual pathway imposed on the germ cells, and their X: A ratio (see below).

Table 3 presents the main results and conclusions obtained in this report and in earlier studies. It shows that the sexual development of germ cells is controlled by the functional state of Sxl which in turn depends on the X:A ratio and on the phenotypical sex of the gonadal soma. In a complementary study, Steinmann-Zwicky et al. (1989), who transplanted XX and XY pole cells carrying Sxl⁺, Sxl^fl or Sxl^M1, reached the same conclusions. Their results are included in Table 3.

In pseudomales, germ cells with two X-chromosomes enter the oogenic or the spermatogenic pathway. To discuss this result, we must recall that the genes tra⁺, tra-2⁺, and dsx⁺ are irrelevant for sex determination in the germ fine, as shown by transplantation of pole cells mutant for any of these genes (Marsh and Wieschaus, 1978; Schüpbach, 1982). The 2137 animals of genotypes No. 1–10 produced 930 testes with germ cells of identifiable sex; the rest showed abortive gametogenesis. Of the 930 testes, 721 were oogenic and 209 were spermatogenic. This could mean that XX germ cells are inherently female and that this genetic sex can occasionally be overruled by the pseudomale soma. Alternatively, XX germ cells may actually be male and require a female inductive signal from the ovary to enter oogenesis. We would then have to explain why the germ cells in most of the gametogenic gonads of XX pseudomales nevertheless chose the oogenic pathway although they were surrounded by testicular somatic cells. In contrast to the situation in pseudomales, XX germ cells that were transplanted into normal XY males invariably initiated spermatogenesis (Steinmann-Zwicky et al. 1989).

Earlier experiments had already revealed a difference in the soma of normal males and pseudomales. When XY pole cells were transplanted, they developed into normal sperm in XY or XO host males, but did not form functional sperm in X/X; tra/tra pseudomales (Marsh and Wieschaus, 1978) or only rarely so (Schüpbach, 1985; H. Schmid, unpublished data). This indicates that the soma of pseudomales is somehow inferior to the soma of normal males with respect to supporting spermatogenesis. One of the differences is that Sxl⁺ is active in tra and tra-2 pseudomales, but inactive in normal XY males (Bell et al. 1988; Nagoshi et al. 1988; Saiz et al. 1989). In any case, our experiments demonstrate that the initiation of the sexual pathway of 2X;2A germ cells depends on somatic induction, thus confirming the results of Steinmann-Zwicky et al. (1989).

In the somatic cells, Sxl is the key gene that mediates between the primary signal of the X:A ratio and the downstream genes regulating sex determination (tra, tra-2, dsx, ix) and dosage compensation (for review see Baker and Belote, 1983; Nôthiger and Steinmann-Zwicky, 1985; Baker et al. 1987; Wolfner, 1988). The role of Sxl in the germ fine was less clear. Schüpbach (1985) favoured the hypothesis that the relevant function of Sxl in the germ line is to regulate dosage compensation, whereas Steinmann-Zwicky et al. (1989) provided evidence for a sex-determining function. They observed that transplanted XX germ cells mutant for Sxl^fl entered spermatogenesis in ovaries whereas those carrying Sxl^Mi entered oogenesis even in testes.

In our experiments, as long as the XX-pseudomales carried Sxr⁺ alleles, the germ cells could enter the oogenic or the spermatogenic pathway. When Sxl^M was present in X/X; tra/tra pseudomales, all germ cells became oogenic. Thus, the constitutive mutation Sxl^M can dictate the female pathway to XX germ cells despite the male soma, or despite the absence of a female somatic stimulus. In contrast, only spermatogenic and no oogenic development was seen in Sx/^M1,fm3/ Sxl/^fm7,M1, showing that lack of Sxl function forces XX germ cells into the spermatogenic pathway. These results suggest that the functional state of Sxl alone may be capable of determining the sex of the germ cells.

When Sxl^M1,fm3/Sxf^m3M1 germ cells were transplanted into an ovary, they formed functional eggs (Schüpbach, 1985). Since this genotype causes XX zygotes to develop as pseudomales, Cline (1984) inferred that the mutations abolish the sex-determining function of Sxl, but leave its dosage compensation function more or less intact. Therefore, Schüpbach (1985) explained her results by speculating that normal oogenesis depends on an intact dosage compensation. Our experiments lead us to a different conclusion. They suggest to us that Sxl has a sex-determining function not only in the soma, but also in the germ line. This function is affected, but not completely abolished, in Sxl^M1,fm3/Sxf^m3M1, so that the Sxl function can be activated in the germ line, but not in the soma. Thus, germ cells of this genotype are efficiently directed into the female pathway in an ovary, and into the male pathway in a testis. The observations lead us to conclude that the male soma prevents activation of Sxl in germ cells with this heteroallelic combination, whereas the female soma allows it to become active. The Sxl gene is now cloned (Bell et al. 1988) and the proposed hypotheses can be molecularly tested.

In somatic cells, the X:A ratio regulates all aspects of sex determination and dosage compensation in a cell-autonomous manner. Karyotype 1X;2A is male, and 2X;2A is female. In genetic mosaics consisting of XX and X cells, the former differentiate female structures, the latter male structures, directly adjacent to each other. A ratio of 2X:3A represents an intermediate signal which is interpreted as female by some cells, as male by other cells. This results in so-called triploid intersexes (Bridges, 1925), which are phenotypically mosaic animals consisting of variable proportions of female and male structures.

Transplanting chromosomally normal pole cells, van Deusen (1976) found no heterosexual chimeras, indicating that germ cells are incapable of forming sperm, as are XY germ cells of forming eggs. He concluded that the X:A ratio, as in the soma, represents an autonomous sex-determining signal in the germ Une. Steinmann-Zwicky et al. (1989) have now shown that the situation is more complex, and that sex determination in the germ line has a non-autonomous component. Pole cells of karyotype 2X;2A formed eggs in an ovary, but entered spermatogenesis in testes, whereas XY cells autonomously entered the spermatogenic pathway in testes and ovaries (Table 3).

When animals of karyotype 2X;3A were transformed into pseudomales their germ cells became spermatogenic, and only rarely formed abortive gametogenesis (Fig. 3). This either means that 2X:3A approaches the value of the male signal and that cells of this karyotype can more easily be directed into the male pathway by somatic induction than 2X;2A germ cells. Or it could mean that the soma of 2X;3A pseudomales provides a stronger inductive signal than the soma of 2X;2A pseudomales.

Interestingly, genotype Sxl^M/ +; traltraltra does not form any oogenic stages and is in fact indistinguishable from Sxf/ +; tra/tra/tra. This is reminiscent of transplanted XY or XO germ cells carrying which form normal sperm in testes and spermatocytes in ovaries, suggesting that Sxl^M is not truly constitutive with an X: A ratio of 1:2 (Steinmann-Zwicky, 1988; Steinmann-Zwicky et al. 1989). We conclude that in the germ line, 2X:3A is essentially a male chromosomal signal which prevents Sxl^M from directing the germ cells into oogenesis, or from becoming active in germ cells of pseudomales. The observation that germ cells of genotype 2X;3A carrying Sxl^M (see Sxl^M/ +; + / + / T in Fig. 5) undergo almost perfect oogenesis in an ovary argues that the second alternative is correct. Since genotype Sxl^M/ + /+ / + / + develops as female rather than as intersex, an X:A ratio of 2:3 in somatic cells must allow Sxl^M to become active; an ovary is formed, and as a consequence of this somatic environment, Sxl^M now becomes active also in the germ line despite a ratio of 2:3, thus directing the germ cells into oogenesis. In a testicular environment, however, a ratio of 2:3 is not sufficient to allow Sxf^M to become active in germ cells. Only when the ratio is 2:2 can the mutation overcome the influence of the testicular soma, or become independent of an ovarian stimulus, and activate the Sxl gene.

The two types of triploid intersexes (⁠ in Fig. 5) help to complete the picture. In those cases where the gonadal soma happened to develop into an ovary, the germ cells entered the oogenic pathway, whereas they became spermatogenic when a testis was formed. We infer that the Sxl⁺ genes were activated in ovaries, but remained inactive in testes. In another set of our experiments, however, germ cells with an X: A ratio of 2:3 failed to activate Sxl + and entered spermatogenesis even when the somatic cells formed an ovary. Similarly, Steinmann-Zwicky (unpublished data) found only spermatogenic stages in 2X;3A animals that were transformed into pseudofemales by female-specific cDNA of the tra-gene driven by a heat-shock promoter (McKeown et al. 1988). We must recall here that the sexual phenotype of 2X;3A triploid intersexes is variable and susceptible to even slight genetic and environmental changes (Dobzhansky, 1930; Laugé, 1969; Cline, 1983a). Schiipbach (1985), who analyzed 2X;3A germ cells in 3X;3A ovaries, found 35 almost perfect oogenic clusters and 477 clusters in which the germ cells formed what she calls ‘multicellular cysts’. These multicellular cysts looked very similar to those formed by Sxf/Sxf; 2A germ cells transplanted into female hosts.

In summary, the results obtained with 2X;3A germ cells suggest that in the germ line an X: A ratio of 2:3 is marginally capable of allowing Sxl⁺ to become activated if the gonadal soma is female. Alternatively, the Sxl⁺ genes may be active at an intermediate level (Steinmann-Zwicky and Nôthiger, 1985) that is, however, mostly insufficient to direct the germ cells into the oogenic pathway.

If the SxFgene is active in the female, but inactive in the male germ line, as our results and those of Schiipbach (1985), Steinmann-Zwicky (1988) and Steinmann-Zwicky et al. (1989) suggest, then the mechanism of its activation and its target genes must be different from those in the soma. In the soma, activation of Sxl⁺ is a cell-autonomous process that occurs when the X:A ratio is ⩾ 2:3, but only if the maternal products of da⁺ (Cline, 1978; Cronmiller and Cline, 1987) and Dk⁺ (M. Steinmann-Zwicky, E. Bemhardsgrütter, D. Franken and R. Nôthiger, unpublished data) are present; the target gene for the sex-determining function of Sxl is tra (Nagoshi et al. 1988; Amrein et al. 1988); Sxl^M is constitutive in the soma of 2X;3A and 2X;2A, and nearly constitutive in X;2A (Cline, 1978, 1979, 1983a; Steinmann-Zwicky, 1988). In contrast, activation of Sxl in the germ line has a non-autonomous component and does not depend on da (Schiipbach, 1985; Cronmiller and Cline, 1987), and it may fail to take place even with a ratio of 2X: 2A if the germ cells develop in a male soma; the target gene is not tra, but other genes that are not yet known; Sxf^M is not constitutive in X;2A (Cline, 19836; Steinmann-Zwicky et al. 1989) nor in 2X;3A germ cells when these are surrounded by a male gonadal soma.

We want to conclude that the state of activity of Sxl in the somatic cells dictates the formation of an ovary when an active product is made, or of a testis when an inactive or no product is made. The female soma may exert an inductive influence on the germ cells which react by turning on Sxl⁺, but only when the X: A ratio is above 2:3; lower ratios do not allow activation of Sxl⁺. Alternatively, a male soma may prevent Sxl⁺ from becoming active in 2X;3A and even in 2X;2A germ cells. It is also conceivable that both types of induction, positive in a female and negative in a male soma, regulate the differential activity of Sxl in the germ line.

We further propose that the SxZ-protein has the same function in the soma and germ line, namely to achieve a sex-specific splicing of pre-mRNA of target genes (Bell et al. 1988); for the germ line, these target genes are not yet known. We also want to suggest that there is no dosage compensation in the germ fine. For oogenesis to proceed normally, an excess of X-chromosomal gene products is needed; spermatogenesis, on the other hand, requires a low level of such products. When 2X;3A germ cells become spermatogenic, there are too many X-chromosomal gene products for spermatogenesis to proceed normally, and the process is arrested even in the perfect environment of a testis. For oogenesis, the relatively low number of gene products appears to be less deleterious (see Fig. 5). According to our hypothesis, the X-chromosomal gene products in 2X;3A germ cells are under-represented for normal oogenesis, and over-represented for normal spermatogenesis. It seems relevant here to mention that the single X-chromosome of males must apparently be inactivated during a given phase of spermatogenesis (Lifschytz and Lindsley, 1972). It is possible that 2X;2A and 2X;3A germ cells fail to form normal sperm because they are unable to inactivate both X-chromosomes which would lead to an excess of products of X-chromosomal genes.

To understand or interpret our observations, we want to distinguish between the initiation of the sexual pathway and the execution of the initiated developmental program. Initiation of the pathway depends on the state of activity of Sxl which in the germ line is set by the sex of the soma and the X: A ratio of the germ cells. We propose that these two parameters also decide whether the pathway can proceed normally or whether it ends abortively. For normal oogenesis, the somatic environment must be female and the level of X-chromosomal products must be high. Normal spermatogenesis, on the other hand, requires that the somatic environment be male and the level of X-chromosomal products low. Oogenesis is stopped prematurely in a testis, as is spermatogenesis in an ovary, probably due to failure in nutritive or supportive functions of the inappropriate gonadal sex; and spermatogenesis with two X-chromo-somes is arrested even in the matched environment of a testis because there are too many X-chromosomal gene products.

In contrast to the soma where Sxl acts already very early in embryogenesis (Sánchez and Nothiger, 1983; Gergen, 1987), its function in the germ line is apparently only required later (Schiipbach, 1985). We want to suggest that this late requirement corresponds to the time when the gonadal soma exerts its inductive influence on Sxl so that this gene starts to make a functional product when XX germ cells are located in an ovary, and a non-functional product when they are located in a testis, or when their X: A ratio is below 2:3.

We want to thank Drs A. Dübendorfer, C. Grand, G. Morata, H. Schmid for fruitful discussions and many suggestions, and in particular M. Steinmann-Zwicky for critical comments on the manuscript. We are also grateful to Dr W. Hennig in Nijmegen in whose laboratory some of the work was done. We thank Drs M. Ashbumer and T. Cline for sending fly stocks, and M. Eich, S. Hohl-Schlegel and M. Schmet for technical help. The work was supported by the Swiss National Science Foundation, by the ‘Stiftung für wissenschaftliche Forschung an der Universitht Zurich’, and by the ‘Julius Klaus-Stiftung’.