achaete-scute feminizing activities and Drosophila sex determination

Sexual fate in Drosophila is determined by a two-fold difference in X-chromosome number: 2X diploid flies are female; 1X diploid flies are male (reviewed in Baker, 1989; Hodgkin, 1990; Steinmann-Zwicky et al., 1990; Parkhurst and Ish-Horowicz,, 1992). This difference in X-chromosome dosage achieves sex-specific expression of the Sexlethal (Sxl) gene that regulates sexual development and dosage compensation (Cline, 1978, 1980; Bell et al., 1988). Sxl activity is continuously required in XX-animals for female development and a low transcription rate of X-chromosome genes. In XY-animals, Sxl must be inactive, eliciting male development and a higher rate of X-chromosomal transcription. Mutations leading to inappropriate Sxl expression (Sxl^– in females; Sxl^constitutive in males) cause sex-specific lethality (Lucchesi and Skripsky, 1981; Cline, 1983, 1984; Sánchez and Nöthiger, 1983).

Sxl expression throughout development is maintained by an autoregulatory pathway of differential RNA splicing (Bell et al., 1988, 1991). However, initiation of Sxl activity relies on early transcripts derived from an embryonic promoter (Sxl P_E; Keyes et al., 1992) that is responsive to dosage-sensitive transcriptional activators encoded by dispersed X-linked loci (‘numerator elements’; Cline, 1986, 1988). The Sxl P_E promoter is only activated in females, resulting in the synthesis of sufficient Sxl protein to trigger Sxl autoregulation (Bopp et al., 1991; Keyes et al., 1992). Cline has defined two numerator loci, sisterless-a (sis-a) and sisterless-b (sis-b), that are required for Sxl activation in females (Cline, 1986, 1988). Genetic analysis shows that sis-b maps within the achaete-scute complex (AS-C) and corresponds to AS-C T4 (scute- ). Mutations within T4 affect sex determination, T4 transposons rescue sis-b mutations, and T4 mis-expression activates Sxl expression in male embryos (Torres and Sánchez, 1989; Parkhurst et al., 1990; Erickson and Cline, 1991).

T4 is one of four proteins encoded within the AS-C that belongs to the basic helix-loop-helix (bHLH) class of transcriptional regulators (Villares and Cabrera, 1987; Alonso and Cabrera, 1988; Murre et al., 1989a). Another bHLH protein, encoded by the daughterless (da) gene, is also required for Sxl activation in females (Caudy et al., 1988b). In vivo and in vitro interactions are consistent with direct heterodimeric interactions between Da and AS-C proteins. Da/AS-C heterodimers bind specifically to candidate target site DNA in vitro and selectively activate transcription from reporter constructs assayed in yeast (Murre et al., 1989b; Van Doren et al., 1991; Cabrera and Alonso, 1991). Genetic interactions suggest that Da and AS-C heterodimers also act to promote neural cell fates later during Drosophila embryogenesis (Caudy et al., 1988a).

We previously presented evidence for the regulation of sex determination by bHLH heterodimers using Drosophila fusion genes that mis-express the hairy (h) bHLH pair-rule segmentation gene (Fig. 1A; Parkhurst et al., 1990). We used the hunchback (hb) gap-gene promoter to drive h expression in the anterior half of the blastoderm embryo (hb-h; Parkhurst et al., 1990; Parkhurst and Ish-Horowicz, 1991). hb-h carrying embryos are female-lethal due to premature h expression that inhibits Sxl expression in the anterior of the embryo (Fig. 1B). Although the h bHLH protein does not normally function in sex determination, it may be mimicking the action of Deadpan, a bHLH protein recently shown to function during sex determination (Younger-Shepherd et al., 1992; see Discussion). Extra doses of sis-a⁺ or sis-b⁺ rescue hb-h females, as does constitutive Sxl expression, showing that ectopic expression of the h protein in the anterior of the embryo interferes with numerator activity and prevents the initiation of Sxl expression (Parkhurst et al., 1990). We also showed that ectopic T4 expression from a hb-T4 fusion gene activates Sxl expression in the anterior of male embryos, leading to male-lethality (Fig. 1C).

Two other AS-C bHLH genes, T3 (lethal-of-scute) and T5 (achaete), whose bHLH domains substantially resemble a female phenotype in the absence of detectable Sxl protein expression. Their mis-expression does not affect the viability of either sex (Parkhurst et al., 1990) suggesting that, alone, they neither activate nor repress Sxl expression. In this paper, we present an extended study of the effects of endogenous and ectopic AS-C expression on wild-type and mutant embryos. We show that all three AS-C proteins have feminizing activity, but that T4, and not T3 or T5, is a numerator gene. The latter two are only weakly expressed in early embryos, and are weaker activators of early Sxl transcription. We also examine Sxl expression in detail and show that female embryos homozygous for sis mutations can be rescued by maternal AS-C expression. We discuss our results in terms of in vivo interactions between bHLH proteins, and discuss evidence that female (XX) cells may be viable and display a female phenotype in the absence of detectable Sxl protein expression.

Flies were cultured and crossed on yeast, maizemeal, molasses, agar medium, at 25°C unless otherwise stated. The stocks used in this study are listed below:

Construction and germline transformation of the hb-h and hb-AS-C fusion genes was as previously described (Parkhurst et al., 1990; Parkhurst and Ish-Horowicz, 1991). All second chromosome lines are marked with the recessive mutation, brown, and balanced over CyO. All third chromosome lines are marked with the recessive mutation, scarlet, and balanced over TM3.

Embryos were prepared and immunohistochemical detection of Sxl was performed as previously described (Parkhurst et al., 1990). The mouse anti-Sxl monoclonal antibodies were kindly provided by D. Bopp (Bopp et al., 1991). Rabbit anti-T3, anti-T4 and anti-T5 polyclonal antibodies were kindly provided by C. Cabrera (Cabrera, 1990 and personal communication). The mouse anti-ac (T5) monoclonal antibodies and the rabbit anti-sc (T4) polyclonal antibodies were kindly provided by J. Skeath and S. Carroll (Skeath and Carroll, 1991). All secondary antibodies were obtained from Jackson ImmunoResearch Labs (West Grove, PA). The stained embryos were dehydrated in 100% ETOH and mounted under a coverslip in methacrylate mounting medium (JB-4, Polysciences) that was polymerized under CO₂ for 1-2 hours at room temperature.

To compare Sxl staining intensities of rescued sis-female embryos with control embryos that lack the transgene, embryos from sis-b; hb-T3 or sis-a; hb-T4 were mixed with embryos from oskar mothers before fixation and processed together. The latter, recognized by their lack of pole cells or aberrant gastrulation morphology, have wild-type levels and patterns of Sxl expression. The staining reactions were allowed to proceed until the control embryos began to show background staining. No specific Sxl staining could be detected in the posterior of such rescued sis embryos, suggesting that possible low-level staining is not within the limits of detection of our staining protocol.

Immunohistochemical whole-mount in situ hybridization was performed according to the protocol of Tautz and Pfeifle (1989). The genomic fragments containing the AS-C genes were a gift of C. Cabrera (Villares and Cabrera, 1987; Alonso and Cabrera, 1988). Digoxigenin-substituted probes were obtained by Polymerase Chain Reaction (PCR) amplification to generate the following fragments: hb-T3: –85 bp to +1380; hb-T4: –125 bp to +1485 bp; and hb-T5: –65 bp to +1050 bp.

The precise staging of the embryos was determined using a combination of Dapi nuclear staining and a series of carefully timed short embryo collections.

Adult cuticle was split on the ventral surface and cleared in lactoacetic acid/Hoyer’s medium (1:1) at 65°C for several hours.

DNA was isolated by the method of Gloor and Engels (1992). For adults, one fly was ground up in 50 μl ‘squishing buffer’ (10 mM Tris-HCl pH 8.2, 1 mM EDTA, 25 mM NaCl, and 200 μg/ml proteinase K), incubated at 37°C for 20 minutes, then heated to 95°C for 2 minutes. After visualizing Sxl staining with alkaline phosphatase-coupled secondary antibodies, individual embryos were ground up in 5 μl squishing buffer then processed as described above for adults. PCR reactions were done using 1 μl (adults) or 5 μl (embryos) of the DNA mix according to the manufacturer’s reaction conditions (Cetus) and the following timings: 95°C -60 seconds; 58°C -90 seconds; 72°C -120 seconds.

Primers specific to the wild-type h gene were included in all reactions as an internal control for PCR efficiency.

Previous analyses have demonstrated that AS-C is transcribed during the syncytial blastoderm stage (nuclear cycles 11/13), consistent with a role in sex determination (Romani et al., 1987; Cabrera et al., 1987; Cabrera, 1990; Martín-Bermudo et al., 1991). However sex-specific differences in Sxl protein expression are already established by nuclear cycle 9 (Bopp et al., 1991; Keyes et al., 1992), so we used in situ hybridization to identify AS-C transcripts in preblastoderm stage embryos. Fig. 2A shows that lowlevel but ubiquitous T4 expression is detectable as early as cycle 3. Transcript levels increase during cycles 7/8, peak during cycles 11/12, and rapidly decay during cycle 13 until no longer detectable by cycle 14 (Fig. 2B-D).

Two other AS-C HLH transcripts, T3 and T5, are more weakly expressed than T4, and are only detectable between blastoderm cycles 7/8 and 12/13 (Fig. 2E-L). Their inability to rescue sis-b mutations may be due to reduced and delayed expression, or to qualitative differences with T4.

The functional significance of early T4 expression was confirmed by analyzing embryos mutant for sis-b^3-1, a female-lethal T4 allele that retains neural function and, therefore, cannot encode an inactivated gene-product (García-Bellido, 1979; Campuzano et al., 1985; Cline, 1988; C. Cabrera, personal communication). In situ hybridization to sis-b^3-1 embryos reveals a defect in early transcription, T4 transcripts being undetectable before cycle 11 (Fig. 3A) and expressed only weakly during cycles 11 to 13 (Fig. 3B). Neural T4 expression is normal in later embryos (not shown), indicating that the 3-1 lesion selectively affects early T4 expression. T3 and T5 transcription are unaffected in sis-b^3-1 embryos (Fig. 3C-D), showing that the female-lethality is due to failed T4 expression, and that T3 and T5 expression is independent of early T4 activity.

The above results confirm that T4 is the major numerator in the AS-C, i.e. that it acts together with da in activating the early Sxl P_E promoter and that, despite their close similarities, the T3 and T5 genes are largely inactive in normal sex determination. However, T3 and T5 proteins, if expressed at higher levels, might also be able to activate the Sxl P_E promoter. Therefore, we examined the effects of ectopic T3 or T5 expression on sex determination using two previously described fusion genes, hb-T3 and hb-T5, that express their respective AS-C proteins in the anterior region of early blastoderm embryos (Parkhurst et al., 1990). hb- AS-C fusion genes were constructed with a promoter fragment that contains both hb promoters (Tautz et al., 1987; Schröder et al., 1988; Tautz and Pfeifle, 1989), but lacks the 5′ cis-regulatory sequences necessary for proper expression of the distal maternal promoter (Parkhurst et al., 1990; Parkhurst and Ish-Horowicz,, 1991).

hb-T3 and hb-T5 embryos are fully viable (88-98%) and show normal cuticular pattern, and homozygous stocks are readily maintained (Table 1; Parkhurst et al., 1990). This is not due to poor expression of the fusion constructs, the hb-AS-C fusion transcripts being expressed at much higher levels than endogenous AS-C transcripts (cf. Fig. 4). Neither hb-T3 or hb-T5 embryos nor embryos including both fusion constructs show ectopic Sxl activation (not shown). In contrast, a single copy of hb-T4 causes male-specific lethality (26-40%) and activates ectopic Sxl expression in the anterior of male embryos (Parkhurst et al., 1990).

Nevertheless, interactions between the various hb-fusion genes show that the T3 and T5 proteins can influence sex determination. Like hb-T4, single copies of either hb-T3 or hb-T5 rescue the female-lethality of hb-h (Table 2A). In addition, a single copy of hb-T3 or hb-T5 strongly reduces the viability of heterozygous hb-T4 males (Table 3A), indicating that T3 and T5 proteins have feminizing activity that can co-operate with T4 in activating Sxl. Single copies of hb-T3 or hb-T5 reduce the viability of hb-T4/+ males 4-to 7-fold, whereas sibling females remain fully viable. Extra wild-type da copies do not reduce the viability of hb-T4/+ males (Table 3B), showing that X-linked numerator activities, not Da levels, are normally limiting in sex determination.

The above results could be due either to direct interactions between h and AS-C proteins as previously proposed to explain the mutual antagonism between hb-h and hb-T4 (Table 2B; Parkhurst et al., 1990), or to feminizing activity of T3 and T5 not revealed in previous assays. The feminizing activity of T3 and T5 is clearly not sufficient to kill males, which have only one copy of each sis gene, but might be revealed in female embryos.

We therefore tested if ectopic T3, T4 or T5 expression rescues sis-b^3-1 females, which lack early T4 expression but retain their wild-type complement of other sis genes. hb-T4 rescues sis-b females (60-87%; Table 4A). More unexpectedly, a single copy of either hb-T3 or hb-T5 also restores sis-b^3-1 female viability (Table 4A). hb-T5 rescues almost as effectively as hb-T4, while sis-b; hb-T3 females survive about half as often (19-47%; Table 4A). A low frequency (1-5%) of the rescued sis-b; hb-T3 females show partial sex transformations – small patches of male pigmentation on the last two abdominal tergites (not shown). Presumably, Sxl activation is incomplete in some cells of these individuals, sufficient to establish dosage-compensation and tissue-viability, but not female sex differentiation (see below and Discussion).

hb-AS-C rescue is not due to cross-activation of the sis-b^3-1 gene since early T4 expression is not detected in sis-b; hb-T5 or sis-b; hb-T3 embryos (not shown). The expression levels and patterns of the AS-C genes do not show cross-regulatory interactions in AS-C mutant embryos during embyogenesis (C. Cabrera, personal communication; S. M. P., unpublished results), in contrast to the cross-reg ulatory interactions of AS-C genes reported during imagi nal disc development (Martínez and Modolell, 1991). We confirmed these results by rescuing In(1) sc^8Lsc^4R females (Table 4B). In(1) sc^8Lsc^4R is a rearrangement within the AS-C region that deletes the T4 bHLH gene, but does not affect T3 or T5 (Villares and Cabrera, 1987).

In a further assay for feminizing activity, hb-T4 readily rescues sis-a females (25-48% viable; Table 5), consistent with previously reported rescue by sis-b⁺ duplications (Cline, 1988; Torres and Sánchez, 1989; Erickson and Cline, 1991), and with the ability of sis-a⁺ duplications to rescue hb-h female-lethality (Parkhurst et al., 1990). How ever, T3 and T5 proteins are unable to substitute for sis-a: sis-a; hb-T3 and sis-a; hb-T5 females are <1% viable (Table 5). Either only T4 is a sufficiently potent feminizing pro tein to rescue sis-a embryos, or T4 numerator activity dif fers qualitatively from that of the other AS-C proteins, despite the resemblance in bHLH domains.

The rescue of sis females by hb-fusion genes is somewhat surprising; Sxl is required in all cells, yet the hb promoter is zygotically expressed only in the anterior half of the embryo. Indeed, Sxl is activated only in the anterior half of hb-T4 male embryos (Parkhurst et al., 1990; S. M. P., unpublished data). We used monoclonal antibodies against full-length ‘female-specific’ Sxl protein (Bopp et al., 1991) to examine Sxl staining in sis female embryos whose lethal ity is rescued by a hb-AS-C fusion gene. All such embryos (sis-b females rescued by all three hb-AS-C fusion genes and sis-a; hb-T4) show strong anterior Sxl activation (Figs 5, 6). We confirmed that the Sxl-staining embryos are geno typically female using PCR with primers specific for Y-chromosome sequences (see Materials and Methods). Ante rior Sxl expression is not restored in sis-a; hb-T3 and sis-a; hb-T5 female embryos (not shown), consistent with their lack of viability.

The patterns of posterior Sxl staining differ according to genotype, and fall into two classes that we consider sepa rately: (1) Sxl protein is also expressed in the posterior half of sis-b; hb-T4 and sis-b; hb-T5 female embryos. (2) Sxl is activated only in the anterior half of sis-b; hb-T3 and sis a; hb-T4 female embryos.

Female sis-b; hb-T4 and sis-b; hb-T5 embryos show the first class of staining pattern in which posterior Sxl stain ing is readily detectable (Fig. 5). This generalized Sxl expression explains their embryonic viability and is seen regardless of which parent contributes the transgene, con sistent with activation by uniformly distributed maternal transcript or by uniform zygotic low-level activity from the hb promoter (see below and Discussion). sis-b; hb-T5 embryos show lower levels of Sxl expression in the posterior half compared to anterior expression (Fig. 5D). Such stable maintenance of intermediate Sxl levels has previously been noted for hb-T4 and sis-b^3-1 embryos (Parkhurst et al., 1990; Bopp et al., 1991).

Female sis-b; hb-T3 and sis-a; hb-T4 embryos show the second class of staining pattern in which Sxl is expressed normally in the anterior, but lacking in the posterior half (Fig. 6). PCR analysis-with Y-specific primers shows that such embryos are genotypically female (see Materials and Methods). Posterior Sxl staining is undetectable compared to control embryos, even under saturating staining conditions (see Materials and Methods). Anterior Sxl expression is maintained throughout embryogenesis-(Fig. 6D,H). Yet, mutant female embryos are fully rescued (>85% hatching) and morphologically normal. Such viability is unexpected because lack of Sxl expression in female tissue is usually sex-specific lethal due to inappropriate dosage compensation (Cline, 1983; Sánchez and Nöthiger, 1983). The viability of female cells lacking apparent Sxl expression, raises the possibility that female devel-opment can proceed in cells that do not express active Sxl protein (see Discussion).

One explanation of the posterior rescuing activity is that our hb promoter retains low levels of maternal activity, resulting in maternal AS-C transcripts that would be distributed throughout the embryo (Tautz et al., 1987; Schröder et al., 1988). Fusion transcripts lack the 3′ hb- sequences that confer sensitivity to nanos-induced degradation (Wharton and Struhl, 1991), so any such maternal transcripts would be expected to persist until the late blastoderm stage.

Maternal T4 expression is sufficient to rescue sis-b females and their viability does not require a zygotic hb-T4 gene (Table 4A). About half of the rescued females (7 of 16 tested) lack hb-T4, so their rescue is due solely to maternal T4 expression (presence of hb-T4 transgene determined in individual rescued sis-b flies using PCR – see Materials and Methods). Maternal hb-T5 also rescues sis-b females, again with about half of the rescued females (7 of 14 tested) lacking the hb-T5 transgene.

In contrast, zygotic AS-C expression appears to be essential for rescue of sis-b by hb-T3, and sis-a by hb-T4. All rescued females retain the fusion gene when analyzed by PCR (7/7 sis-b; hb-T3. 13/13 sis-a; hb-T4), and rescue frequency is only 50% (Tables 4, 5). These results are consistent with T3’s weaker feminizing activity and with sis-a mutations affecting sex determination more severely than sis-b.

Thus, although maternal hb-AS-C efficiently rescues sis flies, paternal hb-AS-C expression is also effective (Tables 4, 5). The genotypes susceptible to maternal rescue correspond to sis-b; hb-T4 and sis-b; hb-T5, the class of rescued females exhibiting posterior Sxl expression. This posterior expression is also seen in embryos rescued paternally, despite the anterior specificity of early zygotic hb expression.

Despite the structural similarities between the AS-C bHLH proteins and their demonstrated overlapping neural functions (Jiménez and Campos-Ortega, 1979; Cabrera et al., 1987; Romani et al., 1987; Cubas et al., 1991; Martínez and Modolell, 1991; Skeath and Carroll, 1991), only T4 functions as a strong feminizing element. The T3 and T5 genes do not behave like T4 for two reasons. First, they are only weakly expressed in early embryos, transcripts being undetectable prior to nuclear cycles 7/8 (Fig. 2). In contrast, T4 is transcribed by nuclear cycle 3, ensuring that T4 protein is present to activate Sxl expression by nuclear cycle 9. Second, T3 and T5 proteins are weaker activators of early Sxl transcription than T4 (see below).

The requirement that Sxl switching be error-free requires that slight changes in X:A ratio have major effects on sex determination. Our experiments exclude models in which Sxl’s non-linear response to changes in numerator dosage is due to numerators regulating their own transcription. Expression of T3, T4 and T5 is independent of each other and of sis-a during the time at which sex is determined. Also, we find no evidence that T4 autoactivates its own expression in females. We are unable to discriminate T4 expression levels in female and male embryos, consistent with gene dosage causing 2-fold differences that would not be reliably distinguished by in situ hybridization. We presume that threshold Sxl activation reflects combinatorial interactions among diverse numerators at multiple DNA-binding sites within the Sxl P_E promoter.

From our detailed examination of wild-type patterns of AS-C expression, we envisage at least two distinct stages of establishing Sxl state (ON or OFF) during normal development. First, Sxl protein becomes differentially expressed prior to nuclear cycle 9, at which time Sxl activity is first detected in female, but not male, embryos (Bopp et al., 1991; Keyes et al., 1992). This period corresponds to the time of early T4 expression and to the temperature-sensitive period for the female-specific lethality of the sis-b^3-1 allele (Erick-son and Cline, 1991; Torres and Sánchez,, 1991). The earliest T4 transcripts are detectable at nuclear cleavage cycle 3 and are likely to be zygotic rather than maternal because they are lacking in sis-b^3-1 embryos (Fig. 3A). Sxl tran-scriptional initiation appears to be particularly sensitive to this early numerator activity, being already differentially expressed in males and females before maximal T4 expression in cycles 10 to 13. By cycle 14, Sxl expression is insensitive to endogenous h and AS-C expression, so a second step must lock the Sxl ‘switch’, presumably inactivating the Sxl P_E promoter (Keyes et al., 1992).

The three AS-C proteins differ in their abilities to activate early Sxl transcription when ectopically expressed, T4>T5>T3. Only hb-T4 is male lethal, and hb-T5, but not hb-T3, restores Sxl expression in the posterior half of the sis-b female embryos. The different feminizing activities of these three AS-C gene products must arise from differences in protein structures. bHLH proteins are characterized by two adjacent protein motifs: the basic domain required for sequence-specific DNA-binding, and the HLH domain required for protein oligomerization (Murre et al., 1989a; Murre et al., 1989b; reviewed in Weintraub et al., 1991). The AS-C proteins include identical basic and highly related HLH domains, so it is expected that all can recognize and bind the early Sxl promoter.

Cabrera and Alonso (1991) have shown differential stability of heterodimers between Da and each of the AS-C proteins, as well as variable affinities of Da/AS-C het-erodimers for different DNA target sites. Differences in loop structure could also alter binding to bHLH partner proteins (e.g. Da) or to ‘co-activator’ proteins such as those believed to modulate homeoprotein DNA-binding specificity in vivo (Hayashi and Scott, 1990; Smith and Johnson, 1992). The relative transcriptional activation potentials of the AS-C proteins will also depend on structural differences outside the bHLH domain. Experiments studying in vivo activities of hybrid AS-C proteins with exchanged and modified structural domains will begin to define the determinants of feminizing activity and their roles in transcriptional activation.

The rescue of hb-h female-lethality by sis⁺ duplications led us to propose that the mis-expressed h protein inhibits Sxl transcription by forming inactive heterodimers with T4 protein (Fig. 1B; Parkhurst et al., 1990). Such in vivo competition between bHLH proteins is consistent with the reciprocal interactions between hb-h and hb-T4 (Table 2), and with the rescue of hb-h females by ectopic expression of the T3 and T5 genes. When exhibiting feminizing activity, what are T3 and T5 telling us about the normal mechanism of sex determination?

One model for T3 and T5 antagonism of hb-h is that they sequester the ectopic h as heterodimers, thereby freeing endogenous T4 to act together with Da to activate Sxl. Equally, T3 and T5 could antagonize denominator ‘masculinizing’ proteins. Autosomally encoded bHLH proteins have recently been identified that are expressed during early embryogenesis, exhibit male-specific lethality and alter Sxl expression (Younger-Shepherd et al., 1992; D. Barbash and T. Cline, personal communication; S. M. P. and D. I-H, unpublished). hb-h may be mimicking the action of such elements in causing female-lethality.

However, several attempts to demonstrate h/AS-C heterodimers in vitro have been unsuccessful (cf. Van Doren et al., 1991), suggesting that h might act through other bHLH proteins. The possibility remains that T3 and T5 proteins act directly on the Sxl promoter and that repression by h is a more active process. Evidence for this view comes from assaying h activity using site-directed mutagenesis of hb-h during sex determination (S. M. P., unpublished) and analysis of the molecular lesions associated with h mutations (Wainwright and Ish-Horowicz, 1992). These experiments indicate that h protein minimally requires an intact basic domain, the HLH domain, and a C-terminal ‘WRPW’ domain. The first two domains imply that h repression acts via bHLH heterodimers that bind DNA; the last is indicative of interactions with other proteins. Thus, h may partner a maternally derived bHLH protein, forming heterodimers that compete with or over-ride numeratormediated activation.

As Sxl is required throughout the female embryo, we did not expect the hb fusion genes to rescue sis female-lethality. In two cases (sis-b; hb-T4 and sis-b; hb-T5 – Fig. 5/Table 4A), viability is due to Sxl expression throughout the embryo. PCR analysis demonstrates that rescued embryos need not inherit a zygotic fusion gene, showing that viability is primarily due to maternal transcripts deposited into the egg. These are present at low levels (indeed, are not detected by our in situ hybridizations), but should persist throughout the blastoderm stage, while the embryo is sensitive to numerator activity.

Nevertheless, zygotic fusion gene expression alone can rescue sis female embryos (Tables 4, 5). Rescued sis-b; hb-T4 and sis-b; hb-T5 embryos exhibit ubiquitous Sxl expression. In this case, posterior Sxl activation could result from diffusion of anterior sis transcripts or protein, or from low-level zygotic transcription in the posterior hb domain. Sex determination in sis-b embryos would be sensitive to early generalized AS-C expression from the hb promoter, although we have not detected such transcripts by in situ hybridization.

More puzzling is the viability and normal morphology of sis-b; hb-T3 and sis-a; hb-T4 females despite their apparent lack of posterior Sxl expression. Sxl is required in female cells to suppress hyperactivation of the X chromosomes and to maintain the female pathway of development, so surviving cells with Sxl^off were expected to die due to overtranscription of X-chromosomal genes, or be sexually transformed (Lucchesi and Skripsky, 1981; Sánchez and Nöthiger, 1982, 1983; Cline, 1983). The lack of male structures is not due to delayed Sxl expression because the anterior-specific pattern of Sxl expression persists throughout larval life and is evident in adults. In any case, there is abundant evidence that the state of Sxl in somatic cells is only sensitive to the X:A ratio during early developmental stages (summarized above and in Sánchez and Nöthiger, 1983; Cline, 1984).

There are several plausible explanations for the rescued posterior cells being viable and untransformed. First, Sxl may be expressed at immunologically undetectable but physiologically significant levels sufficient to initiate and maintain the dosage compensation pathway. Although Sxl expression is autoregulatory, the threshold for autoactivation is not yet understood and intermediate level expression can be stable (Parkhurst et al., 1990; Bopp et al., 1991; this paper). The effects of persistent low-level Sxl expression are unclear, but would be expected to result in immunologically detectable protein, since a pulse of Sxl protein can establish stable Sxl activation (Bell et al., 1991).

Second, the posterior cells could indeed lack any Sxl protein but be insensitive to aberrant dosage-compensation. This would contradict previous clonal analysis in which Sxl cells are poorly viable unless induced late in development, in which case they are sexually transformed (Cline, 1979; Sánchez and Nöthiger, 1982, 1983).

Finally, a low rate of X-chromosome transcription could have become initiated (perhaps by maternal AS-C expression) and be maintained via a Sxl-independent mechanism. There is no precedent for this model, but it is not excluded by current data. Further experiments are required to discriminate among the alternative models. In particular, analysis of genetic mosaics should determine whether posterior cell survival and morphology are indeed Sxl-independent.

Our results are consistent with the notion of functional interchangeability among numerator elements. hb-h females are rescued by either sis-a⁺ or sis-b⁺ duplications, and hb-T4 rescues both sis-a and sis-b mutations. Recent evidence indicates that numerator genes show diverse character, although all are likely to be transcriptional regulators. The runt segmentation gene behaves as a weak numerator element (Duffy and Gergen, 1991; Torres and Sánchez, 1992). runt encodes a nuclear protein that has been implicated in transcriptional regulation during segmentation (Kania et al., 1990). The molecular characterization of the sis-a gene product has not yet been reported, though Cline (1988) has presented evidence suggesting qualitative differences between it and sis-b. Lack of identical molecular motifs does not preclude direct molecular interactions among different numerator proteins. The involvement of bHLH proteins with other proteins including different classes of transcriptional regulators has recently been reported. For example, direct molecular interaction has been demonstrated between the MyoD bHLH domain and c-Jun bZIP domain (Bengal et al., 1992).

Our results also demonstrate clear quantitative differences in the strengths of feminizing activity. Sex is determined by limiting concentrations of numerator proteins, so relative abilities to activate Sxl transcription will be very sensitive to minor affinity differences for co-factor proteins and DNA-binding sites. The three AS-C proteins induce posterior Sxl expression in sis embryos with differing efficiencies. runt’s interactions with other sis genes show it to be a weaker numerator than T4, but have stronger feminizing activity than T5 or T3 (Cline, 1988). As originally proposed by Dobzhansky, the X-chromosome should include many dispersed numerator genes of widely varying strengths (Dobzhansky and Schultz, 1934). The minor feminizing activity of genes such as T3 and T5 may be the basis for the genetic modifiers of sex determination, as interstrain variation of such genes would become significant under conditions of near-threshold X:A ratio (Cline, 1988).

This paper is dedicated to the memory of Carlos Cabrera, whose major contributions to studying the achaete-scute Complex underly much of the current and future work in the field. We thank Suki Parks, Min Han and Dave Turner for numerous discussions and valuable suggestions. We also thank Carlos Cabrera, Susan Shepherd, Yuh Nung Jan, Daniel Barbash and Tom Cline for sharing unpublished data; Steve Russell for sharing unpublished Y-specific sequences; Daniel Bopp, Carlos Cabrera, Jim Skeath and Sean Carroll for antibodies; and Carlos Cabrera, Tom Cline, Claire Cronmiller, Jim Erickson, Juan Modolell, Harald Vässin and the Bowling Green Stock Center for fly stocks used in this study. We are very grateful to Keith Blackwell, Karen Blochlinger, Helen Frances-Lang, Min Han, Mike Krause, Bob Levis, Phil Meneely, Suki Parks, Ze’ev Paroush, Matt Thayer, Dave Turner, Mark Wainwright and Hal Weintraub for reading and critical comments on the manuscript. We thank our colleagues at the ICRF, Caltech and the Hutch for discussion and advice during the course of this work. SMP was supported by a fellowship from the Helen Hay Whitney Foundation, by a Developmental Biology grant from the Lucille P. Markey Charitable Trust (to HDL) and by New Development Funds from the Fred Hutchinson Cancer Research Center.