Commitment of abdominal neuroblasts in Drosophila to a male or female fate is dependent on genes of the sex-determining hierarchy

In holometabolous insects, the adult central nervous system (CNS) is generated by embryonic and postembryonic divisions of neuronal stem cells, the neuroblasts (White and Kankel, 1978; Hartenstein and Campos-Ortega, 1984; Booker and Truman, 1987; Truman and Bate, 1988). During embryogenesis in all insects, neuroblasts in the ventral nervous system segregate from an undifferentiated sheet of neuroectodermal cells and subsequently divide to produce neurons for the larval nervous system (Hartenstein and Campos-Ortega, 1984; Doe and Goodman, 1985). Neuroblasts divide unequally producing a series of ganglion mother cells each of which then divides once, equally, to create two neurons (Edwards, 1969). Following their embryonic wave of mitoses, neuroblasts in holometabolous insects become quiescent, re-emerging during larval life to commence a series of postembryonic stem cell divisions which provide adult-specific neurons (Prokop and Technau, 1991). The majority of the embryonic neuroblasts in thoracic segments but only a few abdominal neuroblasts engage in postembryonic neurogenesis (Booker and Truman, 1987; Truman and Bate, 1988).

In Drosophila, the decision to develop as a male or female operates through the activity of a cascade of genes which lead to the appropriate differentiation of sexually dimorphic somatic tissues. These sex-determining genes, extensively characterized both genetically and molecularly, are: Sex-lethal (Sxl), transformer (tra), transformer-2 (tra-2), doublesex (dsx) and intersex (ix) (for reviews, Baker and Belote, 1983; Nöthiger and Steinmann-Zwicky, 1985; Cline, 1985; Baker et al., 1987; Wolfner, 1988; Baker, 1989; Slee and Bownes, 1990; Steinmann-Zwicky et al., 1990). When homozygous for tra or tra-2 loss-of-function mutations, chromosomally female flies (X/X) develop as phenotypic males in all somatic aspects, whereas, the somatic development of chromosomally male flies (X/Y) is unaffected (Watanabe, 1975; Baker and Ridge, 1980; Ota et al., 1981; Belote and Baker, 1982; Belote et al., 1985; McRobert and Tompkins, 1985; Belote and Baker, 1987). Consequently, the genes tra⁺ and tra-2⁺ are necessary for female development. In females, the primary function of the tra⁺ and tra-2⁺ gene products in somatic tissues is the preferential splicing of dsx pre-mRNA into the female-specific dsx mRNA (Nagoshi et al., 1988; Hoshijima et al., 1991; Hedley and Maniatis, 1991). In the absence of the tra⁺ or tra-2⁺ products, as in mutant female flies, dsx pre-mRNA is spliced into the male-specific dsx mRNA (Nagoshi et al., 1988). As a result of this differential splicing, two distinct dsx proteins are made, one type in males and another in females (Baker and Wolfner, 1988; Burtis and Baker, 1989) . Masculine or feminine sexual development depends on which wild-type dsx product is present. In the absence of any functional dsx gene product, as in flies homozygous for dsx loss-of-function mutations, both male and female flies develop as intersexes (Hildreth, 1965; Baker and Ridge, 1980). Dominant alleles of dsx express the male dsx mRNA in both sexes and can be used in combination with other dsx alleles to generate females that develop as phenotypic males or as intersexes (Denell and Jackson, 1972; Baker and Ridge, 1980; Nöthiger et al., 1980, 1987; Nagoshi and Baker, 1990) .

The specification of sexually dimorphic neurons in the adult central nervous system is an attractive system to study the general processes underlying the commitment of neurons and their precursors to a particular fate. Two sources of differential neurogenesis in male and female Drosophila have been described: a postem-bryonic increase in the number of cells and Kenyon fibers in the female mushroom bodies (Technau, 1984), and divisions by terminal abdominal neuroblasts during larval and early pupal stages (Truman and Bate, 1988). We have exploited sex-specific neurogenesis in the abdominal ganglion to examine the role of the sexdetermining genes in the commitment of neuroblasts to a male or female mode of division and to identify the times during development when the decision to enter and maintain a sexual fate are made.

Fly stocks were maintained on a standard cornmeal diet supplemented with live yeast at 25°C. A full description of the mutants included below can be found in Lindsley and Zimm (1985; 1987; 1990).

Control larvae and pupae were derived from either CantonS stocks or from crosses with females homozygous for the recessive mutation cinnamon (cin) (Baker, 1973). Crosses with mutant cin females were used to generate large numbers of single sex larvae, especially for timepoints during the first and second instar when Canton-S larvae cannot be sexed. Female flies that are cin homozygotes are sterile. Progeny can be recovered from these females if a cin⁺ gene is introduced paternally (Baker, 1973). To produce daughters exclusively, cin homozygous mothers were mated to Canton-S fathers, where the wild-type gene is carried on the paternal X-chromosome (Table 1). When cin mothers were mated to y/y⁺cin⁺ Y males, the only cin⁺ gene is carried on the Y-chromosome, and therefore only sons are produced (Baker, 1973; Table 1).

The crosses used to generate mutant progeny are listed in Table 1 and the markers used to score progeny are indicated in Table 2. The tra¹ allele (Sturtevant, 1945) was shown to be a small deletion (McKeown et al., 1987) and was carried on a multiply marked th st tra cp ri p^p chromosome. The deficiency Df(3L)st^j7 (73A1-2; 73B1-2), Ki roe p^p uncovers the tra locus (McKeown et al., 1987). The tra-2 alleles used were: tra-2¹ (Watanabe, 1975), tra-2^B (Belote and Lucchesi, 1980), the deficiency Df(2L)trix (51A1-2;51B6; Belote et al., 1990), tra-2^ts1 (Belote and Lucchesi, 1980) and tra-2^ts2 (Belote and Baker, 1982, 1983). From sequence data, the temperaturesensitive tra-2 alleles have been mapped to single amino acid substitutions either in, as is the case for the tra-2^ts1 allele, or near, as in the tra-2^ts2 allele, the ribonucleoprotein motif found in the tra-2 protein (Amrein et al., 1990). All tra-2 alleles except for tra-2¹ were on chromosomes that were also marked with cn bw. Two dsx recessive mutations were used: dsx¹ (Hildreth, 1965), this chromosome is also marked with p^p; and dsx¹⁹ (also listed as dsx^EFH48, Baker et al., 1991), this chromosome is also marked with the recessive mutations red e. The dsx null created as a trarw-heterozygote of the inversion dsx²³(In(3R)dsx^D+P3, Baker et al., 1991) and the deficiency Df(3R)dsx¹⁵, bx sr e (Df(3R)dsx^M+R15, 84D10-11; 84E8; Baker et al., 1991) has no detectable sex-specific dsx mRNA (R. Nagoshi, personal communication). Two dominant alleles of dsx were used: dsx^D, this chromosome is also marked with Sb e^s, (Denell and Jackson, 1972; Nagoshi and Baker, 1990) and dsx^M (Nöthiger et al., 1980; Nagoshi and Baker, 1990).

For the developmental series, control and mutant larvae were staged independently for each instar. Larvae within an hour of their ecdysis to the next instar were collected on apple/agar plates with yeast paste. The duration of each larval instar was determined for all control and mutant genotypes. For graphic representation, the number of labelled ganglia from a given timepoint during an instar was plotted as the percentage of time spent in that instar at the time the ganglia were isolated and incubated in BUdR.

Timed egg and larval collections were made at both 18°C and 29°C. Egg collections of one to two hours were then shifted at various times until the embryos hatched. Larvae were collected within an hour of their ecdysis and maintained at their rearing temperature until the time of the shift. In some cases, larvae from 2-12 hour egg collections were raised until the second or third instar before the temperature shift; the time of the shift was then computed from the length of time before they became wandering third instars and were dissected. For X/X tra-2^ts animals, the durations of embryogenesis and the first, second and third instars up to white puparium formation were 46, 53, 52 and 139 hours at 18°C and 18, 19, 20 and 41 hours at 29°C. Larval age is given with respect to the time that has elapsed within a given instar. For example, larvae that spent 16 hours as first instars at 18°C before their temperature shift to 29°C will be referred to as 30% first instars.

For wandering third instar and white puparium collections, nervous systems were dissected from animals raised at 25°C. Cells undergoing S-phase were identified by their incorporation of 5-bromodeoxyuridine (BUdR; Gratzner, 1982). Ganglia were incubated in 15 μg ml^-1 of BUdR in Schneider’s Drosophila medium (+Glutamate; Gibco) for 2-8 hours. After fixation with 4% paraformaldehyde and treatment with 2N HCL for 0.5-1 hour, BUdR-labelled cells were detected immunohistochemically with an anti-BUdR antibody (1:200 dilution, Becton-Dickinson, Mt. View, CA) as described in Truman and Bate (1988).

For the developmental series, ganglia removed from timed collections of larvae were incubated in BUdR as described. To ensure the labelling of all of the terminal neuroblasts during their first S-phase of the postembryonic cell cycle, an additional set of collections of late first instar larvae were fed BUdR (0.5 mg ml^-1 food; Truman and Bate, 1988) for 4-12 hours with subsequent fixation and processing for immunohistochemistry. Males and females of the following genotypes were analysed for their first postembryonic S-phase: cin/cin⁺ Y, cin/cin⁺, dsx^M/dsx⁺ and dsx²³/Df(3R)dsx¹⁵.

The number of ganglia examined to ascertain the pattern of terminal neuroblast divisions through embryonic and larval stages were: 78 X/X and 44 X/X Canton-S larvae, plus 52 unsexed first and second instar larvae; 124 X/X cin/cin⁺ and 95 X/Y cin/y⁺Y; 75 X/X and 57 X/X dsx²³/Df(3R)dsx¹⁵; 44 X/X and 48 X/X dsx¹/dsx¹; 69 X/X and 9 X/Y dsx^D/dsx⁺; 145 X/X and 36 X/X dsx^M/dsx⁺ 328 X/X and 48 X/Y tra-2^ts1/tra-2^ts2shifted from 18°C to 29°C; 129 X/X and 32 X/X tra-2^ts1/tra-2^ts2 shifted from 29°C to 18°C. The number of late larval or early pupal ganglia examined is listed in Table 3.

Neurogenesis to produce adult specific neurons was monitored in larvae and pupae by the incorporation of BUdR into the DNA of dividing cells. After a brief exposure to BUdR, the ventral CNS contains small clusters of labelled cells, including the progenitor neuroblast, a few ganglion mother cells (GMCs) and, occasionally, a pair of neurons (Truman and Bate, 1988). Within these clusters, the neuroblast is distinguishable by its large nucleus and more dispersed labelling compared to the smaller, often darker, nuclei of the ganglion mother cells, or the diminutive nuclei of neurons.

The most conspicuous sexual difference in neurogenesis in Drosophila melanogaster is found in a set of four neuroblasts in the terminal abdominal ganglia of males that continue dividing (Fig. 1A,B; Table 3) after neuroblasts in female ganglia have stopped mitosis (Fig. 1C,D; Table 3; Truman and Bate, 1988). These neuroblasts are located near the seventh and eighth abdominal nerve roots in the larval ventral nervous system (Figs 1A,B, 2B). They are often positioned as bilaterally symmetric pairs, but in some ganglia one or more neuroblasts are displaced to the opposite side of the midline.

The developmental history and characterization of the cell cycle of the sex-specific neuroblasts was carried out using two sources of control flies: Canton-S larvae, which can be sexed beginning in the mid-second instar by the size of the larval testes, and single sex progeny from crosses with cin/cin mutant females and appropriate males (Table 1). Comparable results were obtained from both of these two types of control male and female larvae.

The time course of divisions by terminal neuroblasts was determined by incubating male and female nervous systems in BUdR at various larval and pupal stages. The developmental series of neuroblast labelling in cin/cin⁺ females and cin/cin⁺ Y males was established for animals raised at 25°C, and is shown in Fig. 2A. Labelled terminal neuroblasts first appeared in both sexes during the molt from the first to the second larval instar. Twelve neuroblasts incorporated BUdR in their first postembryonic S-phase but did not undergo mitosis until between 8 and 12 hours after ecdysis (33-50% second instars, Fig. 2A; Truman and Bate, 1988). Continuous divisions by all twelve neuroblasts began by 50% of second instar and persisted until 32% of third instar.

The twelve terminal neuroblasts appeared to have the same location whether they were found in males or females. These neuroblasts constituted the most posterior group of the segmentally arrayed neuroblasts in the abdominal ganglia. Within this group, they appeared to be organized into a lateral and a medial set of three pairs of neuroblasts each, although this order was not discernible in every preparation. In addition, two to four neuroblasts tended to have a more dorsal location compared to the remainder of the terminal neuroblasts. More anterior abdominal segments have between three and nine pairs of bilaterally symmetric neuroblasts that divide postembryonically (Truman and Bate, 1988; Fig. 2B,C).

The cessation of divisions by terminal neuroblasts began during the midthird instar in both sexes. By 40% of the third instar, around 10 terminal neuroblasts are labelled with BUdR in a 2 hour pulse (Fig. 2B,C). The failure of some neuroblasts to label after this time suggests that these terminal neuroblasts have stopped dividing, which has been confirmed by longer pulses (data not shown). In female ganglia, neuroblasts from both anterior and terminal abdominal segments have ceased mitosis by 56-64% of the third instar stage. Two or three terminal neuroblasts located posteriorly and medially divide slightly longer than neuroblasts from more anterior abdominal segments suggesting that there may be a very small late phase of divisions for some female terminal neuroblasts. In the male CNS, neuroblasts from anterior abdominal segments as well as eight terminal neuroblasts stopped mitoses over the same (56-64%) period in the third instar.

Four terminal neuroblasts in male ganglia can be distinguished from other abdominal neuroblasts by their exclusive incorporation of BUdR beginning at about the 75% third instar and continuing until 12 hours (12%P) after white puparium formation (Fig. 2A,B; Table 3). The duration of the cell cycle of male-specific terminal neuroblasts and GMCs was estimated by exposing nervous systems from wandering third instar larvae to pulses of BUdR of varying durations and counting the number of labelled cells. A similar analysis of cell cycle durations for thoracic and abdominal neuroblasts and GMCs was done by Truman and Bate (1988).

The data obtained for Canton-S and cin/cin⁺ male ganglia is summarized in Fig. 3. In Canton-S ganglia, all four terminal neuroblasts were labelled with the shortest pulse, 2 hours. GMCs are only present transiently before they subsequently divide to generate two neurons. To determine the number of GMCs that label with successively longer pulses, we calculated the number of labelled “GMC equivalents”, equalling the number of GMCs and one-half the number of labelled neurons. The number of GMC equivalents labelled with pulses of 2, 4 and 6 hours was 2.8, 3.2 and 4.2 for Canton-S larvae. The increase in GMC equivalents after longer pulses is due to the addition of neurons to the labelled clusters. The slope of the regression line fitted to these data shows that a new GMC is added about every three hours in Canton-S male ganglia. Around three GMCs are labelled for each of the three pulses, this stability in GMC numbers is due to the equilibrium reached between the addition of new GMCs and their loss through division. To account for the numbers of GMCs labelled along with the calculated rate of their addition, the interval between the birth and division of a GMC must be around 9-10 hours. The length of the G₂M phase of the GMC cell cycle can be determined by the time it takes for the first labelled neuron to appear, which occurred after four hours of labelling. Thus, the S-phase duration for GMCs was around 5.5 to 6 hours and is given by the X-intercept of the regression line. The Y-intercept indicates the number of GMCs (around 2) that would be labelled under very short exposures to BUdR because of the overlapping durations of the S-phases of different GMCs.

In cin/cin⁺ Y ganglia, four neuroblasts were also labelled after a 2 hour pulse of BUdR (Fig. 3B). The number of GMC equivalents were 3.2, 3.6 and 4.0 for pulses of 2, 4 and 6 hours. The rate of GMC addition by the sex-specific neuroblasts at one GMC every 5 hours was slower than for neuroblasts in Canton-S ganglia. In addition, the S-phase of the GMCs was lengthened to about 14 hours with a concomitant increase in the number of GMCs that would label with instantaneous pulses (Y-intercept). The G₂M phase for the GMCs in cin/cin⁺ Y ganglia, however, was around 4 hours. The differences in the cell cycle durations primarily in the S-phase reflect normal variations, since cin/cin⁺ Y flies are normal and fertile, and are partly compensated by the slower development of cin/cin⁺ Y larvae compared to Canton-S larvae. These cell cycle durations were obtained by averaging together the numbers of GMCs and neurons for each of the four neuroblasts per ganglion; thus, any systematic differences between the two bilaterally symmetric pairs of neuroblasts are obscured by this analysis.

To determine whether terminal abdominal neuroblasts had entered the male pathway in female flies sexually transformed into phenotypic males, wandering larval or white prepupal nervous systems from tra¹ homozygotes and tra¹/Df(3L)st^j7trans-heterozygotes were incubated in BUdR. At these stages, the only abdominal neuroblasts still dividing are in the terminal portion of male ganglia (Fig. 2A). As shown in Fig. 4A, labelled terminal neuroblasts were present in ganglia derived from these chromosomally female tra^- mutants. The posterior location of the neuroblasts and their clusters of progeny, near the 7th and 8th nerve roots, matched the position of the male-specific neuroblasts found in either X/Y tra^- mutant (data not shown) or control male nervous systems (Figs 1B,2B). The average number of neuroblasts labelled in X/X tra¹ homozygotes or tra¹/Df(3L)st^j7 was similar to the number present in X/Y tra¹ homozygotes or tra¹/Df(3L)st^j7 or control males (Table 3). The number of GMCs and neurons produced by the late-dividing neuroblasts were similar in both male and female tra^- ganglia. Each neuroblast from X/X tra¹/Df(3L)st^j7 flies was associated with 2.5 ± 0.3 (mean ± s.e.m.) and from X/Y flies with 2.2 ± 0.2 GMCs during a 5-6 hour incubation period.

Mitotically active terminal neuroblasts were also present in abdominal ganglia from late larval and early pupal stages in X/X tra-2¹ homozygotes and tra-2^B/Df(2R)trix flies (Table 3; Fig. 4B). To show that the sex-specific neuroblasts from X/X tra-2¹ homozygotes continued to divide later into the pupal stage, ganglia from mutant pupae up to 16 hours postpupariation were incubated in BUdR. The posterior neuroblasts from X/X tra-2¹ homozygotes divided up to 12 hours postpupariation (n=27, data not shown) which was similar to the temporal limits found in control male ganglia (Fig. 2A). The number of GMCs labelled after a four hour pulse in both female tra-2¹/tra-2¹ ganglia was similar to the number seen in control males (3.0 ±0.1 GMCs; Fig. 3).

Both male and female flies bearing loss-of-function dsx mutations simultaneously express masculine and feminine attributes in many of their sexually dimorphic tissues (Hildreth, 1965; Baker and Ridge, 1980; Baker and Belote, 1983; Nöthiger et al., 1987). To ascertain if any terminal neuroblasts from dsx null male and female ganglia had a male-specific division pattern, wandering larvae or white prepupae of three different genotypes were dissected and their nervous systems incubated in BUdR. There was no incorporation of the label into terminal neuroblasts from male or female ganglia from dsx¹ homozygotes, dsx²³/Df(3R)dsx¹⁵ (Fig. 5A,B), a irans-heterozygote combination that eliminates sexspecific transcripts on northern blots (R. Nagoshi, personal communication), or dsx¹⁹/Df(3R)dsx¹⁵ (Table 3). That dsx is epistatic to tra has been shown by incubating wandering larval nervous systems from male and female tra¹dsx¹ double mutants; late-dividing neuroblasts were not found in these doubly mutant abdominal ganglia, the same phenotype as expressed by dsx null mutants alone (Table 3).

The absence of late neurogenesis in loss-of-function dsx mutants suggests that the male dsx⁺ product is required for terminal neuroblasts to display the male pattern of divisions. Chromosomally male or female flies mutant for dsx dominant alleles in trans with a dsx deficiency express only the male dsx mRNA (Nagoshi and Baker, 1990). Females of this genotype are phenotypically male but sterile, whereas, males are normal and fertile flies (Duncan and Kaufmann, 1975; Baker and Ridge, 1980; Nöthiger et al., 1980, 1987, 1989). To test the hypothesis that the male-specific product of dsx is necessary for late divisions by abdominal neuroblasts, wandering third instar nervous systems from both X/X and X/Y dsx^D/Df(3R)dsx¹⁵ animals were incubated in BUdR. Four terminal neuroblasts were labelled in X/X and XfY dsx^D/Df(3R)dsx¹⁵ ganglia showing that the male dsx⁺ product is required at least for the malespecific phase of divisions (Fig. 5C,D; Table 3).

The parameters of the mitotic cycles of neuroblasts and GMCs were determined for dsx^D/Df(3R)dsx¹⁵ larval ganglia. As was true for nervous systems from control males, four neuroblasts were labelled with the shortest pulse of 2 hours. The number of GMCs labelled per neuroblasts in X/X dsx^D/Df(3R)dsx¹⁵ ganglia was 2.5, 3 and 2.9 with pulses of 2, 4 and 6 hours, and for X/Y ds^D/Df(3R)dsx¹⁵ was 2.6 and 2.8 with pulses of 2 and 6 hours, respectively. The number of GMC equivalents calculated for X/X dsx^D/ Df(3R)dsx¹⁵ nervous systems was 2.5, 3.2 and 3.4 and for X/N dstP/Df(3R)dsx¹⁵ nervous systems 2.6 and 3.4 for the time points listed above. From the regression line fitted to the data from X/X dsx^D/Df(3R)dsx¹⁵ ganglia, the rate of GMC addition was found to be around one GMC every four hours and the GMC S-phase (X-intercept) lasts for about 9 hours (data not shown). Labelled neurons first appear after a four hour pulse suggesting that GMCs have a G₂M-phase of about this length. Thus, the GMC cell cycle has a duration of about thirteen hours in these mutants. These values are in between those from control males (Fig. 3) showing that the male dsx product from the dsx^D allele is necessary and sufficient to support the normal male pattern of divisions by the terminal sex-specific neuroblasts.

In dsx null flies, the failure of neuroblasts to be mitotic during the stages when male-specific divisions take place reveals an earlier block in neurogenesis. The developmental time course of sex-specific neuroblast divisions in dsx¹/dsx¹ and dsx²³/Df(3R)dsx¹⁵ mutant ganglia was determined by exposure to BUdR through out larval life (Fig. 6). Nervous systems dissected from late first instars of dsx²³/Df(3R)dsx¹⁵trans-heterozygotes and incubated in BUdR over the time that they would have ecdysed to the second instar showed labelling in only eight terminal neuroblasts rather than the twelve stem cells normally seen in wild-type larvae. The absence of four neuroblasts in both male and female dsx mutant ganglia strongly suggests that the sex-specific neuroblasts require the presence of one or the other dsx⁺ protein to generate any portion of their normal postembryonic lineage. These eight neuroblasts began dividing about half-way through the second instar and continued into the middle of the third instar but stop dividing at a time when the neuroblasts of other more anterior abdominal segments ceased mitosis (Fig. 6). In ganglia from mid-third instar larvae (4048%), the last abdominal neuroblasts in mitosis were lateral and close to the 6th nerve root. Because of their more anterior location, these neuroblasts could either be lateral terminal neuroblasts or associated with neuroblasts of the sixth abdominal segment (Fig. 6B). The position of these lateral neuroblasts in dsx null mutants contrasts with the primarily medial and posterior location of the last dividing neuroblasts in both female and male controls (Fig. 2B,C) and may be indicative of a role for dsx⁺ in the divisions of other abdominal neuroblasts.

In both male and female CNSs, four terminal neuroblasts require dsx⁺ function for their postembryonic divisions (Fig. 6). Whether the neuroblasts in female ganglia correspond to those in male ganglia was elucidated by examining ganglia from female dsx^D/dsx⁺ and dsx^M/dsx⁺ larvae. These animals exhibit intersexual differentiation resulting from the presence of both male and female dsx⁺ products (Denell and Jackson, 1972; Duncan and Kaufmann, 1975; Baker and Ridge, 1980; Nöthiger et al., 1980, 1987, 1989). Twelve terminal neuroblasts incorporated BUdR when ganglia of dsx^M/dsx⁺ females were exposed throughout the period when first instar larvae molt into second instars (Fig. 7A). The same number of terminal neuroblasts were found in both male or female wild-type larvae at these times (Fig. 2A). If the sex-specific neuroblasts in female ganglia were distinct from those in male ganglia then sixteen neuroblasts would have been expected in dsx^M/dsx⁺ female ganglia. The twelve neuroblasts found in dsx^M/dsx⁺ ganglia began their postembryonic divisions in mid-second instar larvae (50%) and continued dividing until the middle of the third instar (Fig. 7A).

During the third instar when female neuroblasts normally become quiescent, some neuroblasts persisted in both X/X dsx^D/dsx⁺ and dsx^M/dsx⁺ ganglia (Table 3; Figs 7, 8). In these mutants, neuroblasts and their progeny could be labelled until the time when divisions would normally stop in control male ganglia. The number of neuroblasts labelled in pulses during the late third instar and through the early pupal period was quite variable. On average, less than one neuroblast per ganglia was found to be dividing (Table 3); in a number of ganglia no neuroblasts were labelled with BUdR while in others two or three neuroblasts were present.

For the neuroblasts in dsx^M/dsx⁺ females displaying the male pattern of divisions, their cell cycle properties were similar to those of control cin/cin⁺ Y male neuroblasts and GMCs. The number of labelled GMCs per neuroblast were 3.1, 3.6 and 3.3 for pulses of 2, 4 and 5.5-6 hours. The GMCs equivalents were calculated to be 3.1, 3.9 and 3.8 for the three pulses tested. From the plotted regression line, the addition of new GMCs occurred at a rate of one for every five hours and the S-phase (X-intercept) duration was around fourteen hours (data not shown). Labelled neurons first appeared after a four hour pulse indicative of a G₂M duration of this length.

We found no evidence that neuroblasts in these mutants once they had entered the male pathway were capable of premature cessation of mitosis rather than being stably committed for the entire period of malespecific divisions. GMCs were labelled only in association with a neuroblast; and neuroblasts, when labelled, had GMCs and neurons. Once neuroblasts from mutant ganglia have entered the male pathway, they seem to perform as efficiently as those in ganglia from male siblings or control males irrespective of the presence of a normal dsx⁺ allele.

The time when neuroblasts adopt a particular sexspecific developmental program can be discerned by using conditional tra-2 mutations. When X/X tra-2^ts1/ira-2^ts2 (tra-2^ts) flies were raised at the femaledetermining temperature (18°C), no neuroblasts in the terminal abdominal region were labelled by BUdR incorporation in ganglia from wandering larvae or white prepupae. In contrast, in ganglia from flies of the same genotype raised at the male-determining temperature, (29°C) four neuroblasts were labelled (Table 3; Figs 9,10A). Mutant animals were shifted between the female-determining temperature and the male-determining temperature during embryogenesis and larval life. The presence or absence of late-dividing neuroblasts was then monitored by in vitro BUdR incorporation in ganglia removed when animals became wandering third instars or white puparia (Fig. 9). In these experiments, only labelled neuroblasts can unambiguously be assigned a fate, which is that of male-specific neuroblasts; unlabelled ganglia could represent cases where neuroblasts were unable to incorporate BUdR because they were in the female-specific pathway or simply absent. Nervous systems of male tra-2^ts larvae raised under the temperature regimes employed in these experiments and incubated as wandering third instar larvae had four labelled neuroblasts (data not shown).

Temperature shifts from the female- to the maledetermining temperature indicate the latest time in development when neuroblasts could enter the malespecific pathway. Late-dividing neuroblasts were found when embryos or first instar larvae were shifted to the male-determining temperature (Fig. 9). Flies shifted as embryos have from three to four neuroblasts expressing the male pattern whereas flies shifted as first instars generally have two labelled neuroblasts (Figs 9,10B). Because of the normal variability in the location of the four male-specific neuroblasts, we could not establish whether these two labelled neuroblasts represented one of the two bilaterally symmetric pairs or a random sample of the possible pairs. The latest time when larval shifts could lead to the activation of the male-specific pathway was at the end of the first instar. Ganglia from animals shifted to the male-determining temperature as late first instar larvae still had some late-dividing neuroblasts, whereas, ganglia from larvae transferred five hours later as freshly ecdysed second instars, had neuroblasts that were not labelled (Fig. 10C). Similarly, when tra-2^ts larvae were shifted as second or third instars, there was no incorporation of BUdR in neuroblasts when tested as wandering larvae (Figs 9,10D).

The opposite set of temperature shifts, from the male- to the female-determining temperature, identify the earliest time when neuroblasts could be irreversibly committed to the male-specific pathway. When larvae were shifted to the female-specifying temperature as first, second or early third instar larvae and the ganglia assayed from wandering third instar larvae, terminal abdominal neuroblasts did not incorporate BUdR, indicating that no neuroblasts were in the male-specific mode of divisions (Fig. 9). To examine more closely the fate of neuroblasts that had entered the male-specific pathway, X/X tra-2^ts animals were shifted to the femaledetermining temperature during the mid-third instar and examined at two different time points. Twenty-four hours after the shift (about 17% of the duration of the third instar), four sex-specific neuroblasts could still be labelled (n=9; Fig. 11A), but by 72 hours at 18°C, at most only one terminal neuroblast still labelled (n=4; Fig. 11B).

To determine if the commitment of sex-specific neuroblasts to a sex-specific pathway could be affected at any time during their postembryonic divisions, mutant larvae raised at one temperature were shifted after ecdysis to the second instar and their nervous systems were incubated in BUdR as soon as they became third instars. In control ganglia, both male and female flies have twelve terminal neuroblasts that are mitotically active at the beginning of the third instar (Fig. 2). In ganglia from larvae raised at the male-specifying temperature then shifted to the femaledetermining temperature, only eight neuroblasts were dividing in ganglia from recently ecdysed third instar larvae (Fig. 12A). Likewise, larvae raised at the femalespecifying temperature and shifted to the male-determining temperature showed that only eight terminal neuroblasts were still dividing in newly ecdysed third instars (Fig. 12B). These experiments show that the commitment of neuroblasts to either a male or female fate must be maintained throughout the period when their sex-specific lineages are being produced.

Temperature-pulse experiments were used to assay whether the male-specific neuroblasts could recover from treatment at the female-specific temperature. When larvae were shifted as early to mid-second instars from the male- to the female-specifying temperature and then returned to the male-specifying temperature in the mid-third instar, the male-specific neuroblasts did not resume divisions when examined by BUdR incorporation as white puparia (data not shown). The exposure of neuroblasts that had been committed to the male division pattern to the female-specifying temperature, under the conditions employed here, appears to irreversibly block the expression of the male mode of late divisions by these neuroblasts.

In the terminal region of the abdominal CNS of larvae, twelve neuroblasts are found in both males and females. Terminal neuroblasts exhibit the same behavior in both sexes as gauged by BUdR incorporation until the middle of the third instar. They enter their first “S” phase during the late first instar and begin divisions during the mid-second instar. Eight of these twelve neuroblasts in males and all twelve neuroblasts in females stop dividing in the middle of the third instar.

In male larvae, four neuroblasts continue to divide until around 12 hours after pupariation (12% P). With BUdR labelling alone, it is not possible to identify which of the twelve terminal neuroblasts enters the male pathway until the other eight neuroblasts cease divisions. During the male-specific period of divisions in Canton-S ganglia, each neuroblast divides approximately once every three hours over this roughly 30 hour period, adding about twenty extra neurons to the adult nervous system of male compared to female flies (schematized in Fig. 13A). From preliminary experiments, when BUdR is fed to males as late third instar larvae or injected into white prepupae, labelled neurons are present in the terminal part of the adult abdominal ganglia suggesting that the male-specific cells persist into the adult (data not shown).

The genes of the sex-determining hierarchy are essential for the appropriate development and divisions of sex-specific neuroblasts. The male-specific product of dsx⁺ is needed for four neuroblasts to enter the male pathway, which has been monitored by their division pattern during the later part of the third instar through the early pupal stages. In females sexually transformed into phenotypic males as in tra^-, tra-2^- or dsx^D/Df(3R)dsx¹⁵ mutants, four terminal neuroblasts have the male pattern of persistent divisions with roughly the same mitotic rate as control males (Fig. 13B). In dsx^D/Df(3R)dsx¹⁵ males, the male dsx⁺ product, provided by the dsx^D allele, is necessary and sufficient for four terminal neuroblasts to enter the male pathway.

In both sexes, wild-type dsx⁺ gene products are necessary for the generation of a postembryonic lineage by the sex-specific neuroblasts. In dsx^- homozygotes, four terminal neuroblasts never incorporate BUdR at any time during the second or third instar stages when all other terminal neuroblasts are mitotically active. The failure to specify the sexual identity of these neuroblasts results in the complete inability of these neuroblasts to perform their normal divisions, not just to engage in a series of sexually indifferent mitoses. From these experiments, it cannot be determined whether the neuroblasts are present and unable to divide, or whether they fail to initiate any postembryonic development such as re-enlarging during the first instar (White and Kankel, 1978; Truman and Bate, 1988).

The absence of sex-specific neuroblasts in dsx null ganglia is unusual compared to that of other sexually dimorphic tissues that require the activity of the dsx⁺ gene. For example, in the genital disc, both genital primordia differentiate in dsx null mutants (Hildreth, 1965). Other structures, like the sexcombs bristles on the foreleg, differentiate with an intermediate number and intersexual morphology in dsx^- mutants (Hildreth, 1965; Baker and Ridge, 1980). Biochemical sexual dimorphisms show similar effects; for example, yolk proteins, normally a female-specific product of fat body cells, are synthesized by both male and female fat body cells in dsx^- flies (Postlethwait et al., 1980; Ota et al., 1981; Bownes and Nöthiger, 1981). Consideration of these findings led to the hypothesis that the products of the dsx⁺ gene suppress the differentiation of the opposite sexual phenotype in these tissues (Baker and Ridge, 1980; Baker and Belote, 1983). Although the molecular mechanisms by which dsx⁺ controls the specification of sex-specific neuroblasts is unknown, the dsx⁺ gene products appear to act by stimulating neuroblast activity, not solely by suppressing inappropriate sexual differentiation.

For many tissues the intersexual development that results from loss-of-function dsx mutations is very similar to that obtained when both male and female dsx products are active, as in X/X dsx^D/dsx⁺ or dsx^M/dsx⁺ flies (Nöthiger et al., 1980, 1987; Baker and Ridge, 1980). For the sex-specific neuroblasts, their development and differentiation is very distinct in these two type of mutants. Whereas the sex-specific neuroblasts in dsx null mutants completely fail to divide, those in dsx^D/dsx⁺ or dsx^M/dsx⁺ appear to adopt either the female or male pathway and these decisions seem to be stable (Fig. 13C,D). Less than one-quarter of the sexspecific neuroblasts in X/X dsx^M/dsx⁺ ganglia entered the male-specific pathway and late-dividing neuroblasts were found in these mutant ganglia up until the time when the male-specific divisions normally cease.

At this time, there is no separate marker for the male- and female-specific neuroblasts. They can be distinguished only on the basis of the temporal limits of their incorporation of BUdR. Nevertheless, the analysis of the different mutants in the sex-determining genes has provided some insight into the relationship between the sex-specific neuroblasts found in both sexes. Two general models can be proposed to account for these sexually dimorphic neuroblasts. On one hand, the same four neuroblasts in the late first instar larva could opt for either a male or a female pathway depending on the activity of the dsx⁺ gene. An alternative hypothesis would be that first instar larvae possess two distinct sets of four neuroblasts with one set becoming active and the other dormant depending on the activity of the sexdetermining genes.

The results from a phenotypic analysis of dsx mutants supports the first hypothesis. In the absence of dsx⁺ function, four neuroblasts in both male and female ganglia fail to make any postembryonic divisions. When both male and female dsx⁺ products can be produced in the same individual only twelve neuroblasts are found at times, from the mid-second to mid-third instar, when all terminal neuroblasts are normally dividing. This finding is consistent with the hypothesis that a single set of four neuroblasts exist that can choose either the male- or female-specific pathway depending on which dsx⁺ gene product is present. Because additional terminal neuroblasts are not labelled in X/X dsx^M/dsx⁺ ganglia, the other possibility that there are two sets of terminal neuroblasts is less tenable. However, a condition of imperfect cross-inhibition between the male- and female-specific dsx⁺ products acting in two distinct sets of sex-specific neuroblasts, which results in four active sex-specific neuroblasts per ganglia, cannot be entirely resolved until individual markers for the male- and female-specific neuroblasts are discovered.

An additional complication in the interpretation of the dsx^M/dsx⁺ results would occur if the male-specific neuroblasts were a separate group from the twelve terminal neuroblasts and only became mitotically active late in the third instar. The assignment of the four late-dividing neuroblasts in male ganglia as members of the twelve terminal neuroblasts is supported by the findings that only eight terminal neuroblasts divide in male dsx^- nervous systems, and that temperature-shifts from the male to the female-specifying temperature after neuroblast commitment to the male fate resulted in only eight dividing neuroblasts when assayed by BUdR incorporation at the beginning of the third instar (see Fig. 12), a time when twelve neuroblasts are dividing in control male central nervous systems.

During development, there are three likely decision points for control of sex-specific neurogenesis. The first is the initial emergence of neuroblasts during embryogenesis when the divisions necessary to create neurons for the larval nervous system occur. The second point is at the end of the embryonic neurogenic period when the neuroblasts decide between becoming dormant or permanently postmitotic. The third point is late in the first instar when the abdominal neuroblasts enter their first postembryonic S-phase in preparation for mitosis.

By performing temperature-shift experiments using tra-2^ts larvae, we determined that the time when sexspecific neuroblasts become committed to a male or female pathway was late in the first instar (Fig. 14). The commitment event then immediately precedes or is coincident with the first S-phase associated with the postembryonic divisions of these neuroblasts. Temperature shifts of X/X tra-2^ts flies cause the type of dsx mRNA produced to change; a temperature shift from the female- to the male-specifying temperature results in the expression of the male-specific form of dsx mRNA at the expense of the female-specific form of dsx mRNA (Nagoshi et al., 1988). We conclude that the presence of either the male- or female-specific dsx proteins at the end of the first instar determines whether sex-specific neuroblasts enter the male or female pathway (Fig. 14).

For all of the neuroblasts that have been studied to date, all postembryonically active neuroblasts also have embryonic divisions (Prokop and Technau, 1991). If neuroblasts that will become the sex-specific neuroblasts segregate during embryogenesis and produce an embryonic lineage, this initial emergence is independent of the functioning of the sex-determining genes (Fig. 14). These neuroblasts, then, may be unique in having two distinct determination steps, one necessary for their embryonic and another required for their postembryonic divisions, controlling their fate and that of their progeny. The genetic and molecular mechanisms that control the embryonic phase of neuroblast segregation and identity are beginning to be estab-fished. For embryonic neurogenesis, several genes are known to contribute to these processes (for review, Campos-Ortega and Jan, 1991). Expression of three genes within the achaete-scute complex correlates with neuroblast segregation and when absent, as in embryos with a deficiency for the locus, at least the posterior median neuroblast does not appear to segregate (Cabrera et al., 1987). Another gene, prospero, has been found to be important in designating neuroblast identity, resulting in the proper differentiation and axonal pathfinding of its neurons, but not necessary for the segregation of these neuroblasts, their divisions or that of their GMCs (Doe et al., 1991). The requirements for the activity of the dsx gene in sex-specific neuroblasts may encompass both reactivation as well as identity functions for neuroblasts and their progeny.

The commitment of sex-specific neuroblasts to a male or female mode of division must be maintained throughout the period when these neuroblasts are engaged in generating their postembryonic lineages. Following an initial determinative event during the late first instar, there appears to be no other single time point after which neuroblasts become independent of the activity of tra-2 and other downstream sexdetermining genes to express a sex-specific pattern of divisions. In X/X tra-2^ts larvae, when sex-specific neuroblasts are exposed to the temperature conducive to the differentiation of the opposite sex, they cease dividing, even at times when these neuroblasts would be expected to be mitotic by comparison to control male and female CNSs (see Figs 11,12). In addition to their inability to execute a particular developmental program, this finding suggests that neuroblasts, once committed to a sex-specific neuroblast pathway, cannot switch to the neuroblast pathway of the opposite sex. For male-specific neuroblasts, the instability of the commitment to the male-specific mode of divisions occurs under conditions where the female-specific form of dsx mRNA comes to predominate, as happens when X/X tra-2^ts animals are shifted from the male- to the female-specifying temperature (Nagoshi et al., 1988). Indeed, under different conditions where both male and female dsx products can be made, as in X/X dsx^M/dsx⁺ animals, the commitment to the malespecific pathway appears to be stable. The decisions as to the differentiated state of sex-specific neuroblasts differs from the irreversible commitment that exists for a set of male-specific transcripts made by the male accessory glands (Chapman and Wolfner, 1988) or the reversible expression of the yolk protein genes in fat body cells in X/X tra-2^ts flies (Belote et al., 1985).

We wish to thank Drs John Ewer, Steven Robinow, Angela Giangrande, Michael McKeown and Kenneth Burtis for their careful reading of the manuscript and helpful suggestions. The Drosophila stocks were generously provided by Drs Bruce Baker, Rodney Nagoshi, Bruce Chase, John Belote and Michael Mckeown. The anonymous reviewers contributed many thoughtful comments that improved the manuscript. B.J.T. was supported by an NIH postdoctoral fellowship, 5F32NS08618.