intersex, a gene required for female sexual development in Drosophila, is expressed in both sexes and functions together with doublesex to regulate terminal differentiation

A single, multi-branched regulatory hierarchy controls all aspects of somatic sexual differentiation in D. melanogaster (Fig. 1) (reviewed by Cline and Meyer, 1996; Marín and Baker, 1998). This hierarchy functions, via a cascade of alternative pre-mRNA splicing steps, to generate the sex-specific products of the doublesex (dsx) and fruitless (fru) genes, which head two parallel branches. Here, we are concerned with the dsx branch of the sex hierarchy. Wild-type dsx function is necessary for all somatic sexual development outside the central nervous system (CNS) in males and females (Baker and Ridge, 1980), as well as some aspects of sexual development in the CNS (Jallon et al., 1988; Taylor and Truman, 1992; Villella and Hall, 1996). The regulated splicing of the dsx pre-mRNA in females results in the production of a female-specific mRNA that encodes DSX^F, whereas in males default splicing of the dsx pre-mRNA generates a male-specific mRNA which encodes DSX^M. DSX^M and DSX^F are zinc-finger transcription factors with identical DNA-binding domains, but different C termini (Burtis and Baker, 1989; Burtis et al., 1991; Erdman and Burtis, 1993). The dsx gene is the last sex determination regulatory gene in its branch of the hierarchy, as its proteins bind to, and regulate the transcription of, the Yolk protein 1 (Yp1) gene (Burtis et al., 1991; Coschigano and Wensink, 1993). Functioning together with dsx in females are the intersex (ix) (Baker and Ridge, 1980; Chase and Baker, 1995) and hermaphrodite (her) (Li and Baker, 1998a; Li and Baker, 1998b; Pultz and Baker, 1995; Pultz et al., 1994) genes.

Previous studies have provided some insights into the functional relationships of the her and ix genes to dsx (Baker and Ridge, 1980; Pultz and Baker, 1995). The her gene is required maternally for the initial expression of the Sex-lethal (Sxl) gene at the top of the sex determination hierarchy, and in addition is required zygotically for female somatic sexual differentiation and some aspects of male somatic sexual differentiation (Li and Baker, 1998a; Li and Baker, 1998b; Pultz et al., 1994). Furthermore, epistasis analysis places the zygotic function of the her gene in parallel to, or downstream of, the dsx gene (Li and Baker, 1998b; Pultz and Baker, 1995). The ix gene is required for female, but not male, somatic sexual development (Baker and Ridge, 1980; Chase and Baker, 1995). Genetic epistasis studies indicate that ix also acts in parallel to, or downstream of, dsx in the sex-determination hierarchy (Baker and Ridge, 1980). Moreover, molecular data indicate that neither ix nor her is required for the sex-specific splicing of dsx pre-mRNA (Nagoshi et al., 1988; Pultz and Baker, 1995). Therefore, the genetic and molecular data suggest that ix, her and dsx function at, or near, the end of the hierarchy to regulate the terminal differentiation genes in females.

Comparisons of the phenotypes of her, dsx double mutant flies with those of flies that are mutant at just one of these genes showed that her and dsx act independently to regulate some aspects of sexual differentiation and function interdependently to control other aspects of sexual differentiation in females. Thus, the DSX^F and HER proteins independently activate Yolk protein (Yp) gene expression in females. They also independently promote development of the vaginal teeth and anal plates in females (Li and Baker, 1998b). However, these proteins function interdependently to regulate female-specific differentiation of foreleg bristles and pigmentation of tergites 5 and 6 (Li and Baker, 1998b). The effect of her on Yp gene expression is not through the fat body element (FBE), to which the DSX proteins bind (Burtis et al., 1991; Coschigano and Wensink, 1993), but rather through Yp DNA sequences outside the FBE, consistent with the finding that these proteins control the Yp genes in an independent manner. That HER and DSX^F act independently in regulating some aspects of sex and interdependently with respect to other aspects of sex could be due to different organizations of the regulatory elements of the genes being controlled in these tissues, or to differences between the arrays of other factors regulating these genes together with HER and DSX^F.

There are also previous genetic data bearing on the relationship between dsx and ix. First, it has been reported that simultaneous heterozygosity for specific mutant alleles of ix and dsx in diplo-X flies results in a cold-sensitive intersexual phenotype (S. E. Erdman, PhD thesis, University of California at Davis, 1994) (Erdman et al., 1996). As cold-sensitive nonallelic noncomplementation is frequently indicative of protein-protein interactions (Hays et al., 1989; Stearns and Botstein, 1988), it has been suggested that there may be a physical interaction between the IX and DSX proteins. Second, it has been shown that a dsx^F transgene promotes female differentiation in an XY individual that is otherwise wild type, but not in an XY individual lacking ix function (Waterbury et al., 1999). These findings led Waterbury et al. (Waterbury et al., 1999) to suggest that ix and dsx function interdependently and that IX is either constitutively expressed (and therefore present in males), or directly under the control of DSX^F.

To understand how the female-specific function of the ix gene is established and how ix regulates terminal differentiation in females, we have cloned the ix gene. ix was localized to the cytological region 47F by complementation with deficiencies and further localized to a 65-kb region by restriction fragment length polymorphism (RFLP) mapping. A clone containing the ix gene was identified by its ability to rescue ix mutant phenotypes when introduced into flies by P-element-mediated germline transformation. The ix protein has sequence similarity to proteins proposed to act as transcriptional activators, but does not contain a known DNA-binding domain. Additionally, the ix pre-mRNA is not alternatively spliced, suggesting that the ix protein is present in both sexes and may interact with one or more female-specific proteins to regulate female differentiation. As IX and DSX^F are proposed to act at the bottom of the sex determination hierarchy, the possibility that these proteins cooperate to regulate female terminal differentiation genes was investigated. Analysis of females mutant for ix, dsx, or both, demonstrated that IX and DSX^F function interdependently to activate Yp gene expression and to regulate differentiation of vaginal teeth, anal plates, foreleg bristles and sixth-tergite pigmentation. Therefore, unlike the DSX^F and HER proteins, which cooperate to control some terminal differentiation genes and function independently to regulate others, IX and DSX^F function together to control somatic sex differentiation in all female structures analyzed. A possible mechanism for the interdependence of IX and DSX^F is revealed by our demonstrations that IX interacts with DSX^F, but not DSX^M, in yeast 2-hybrid and co-immunoprecipitation assays, and that IX and DSX^F form a DNA-binding complex, as assayed by gel shift.

Mutations and chromosomes not referenced are described elsewhere (Lindsley and Zimm, 1992). Crosses were carried out at 25°C unless another temperature is indicated.

Deficiency breakpoints were analyzed in polytene larval salivary gland chromosomes dissected in 0.7% NaCl and stained with orcein (Ashburner, 1989). The deficiency stocks Df(2R)17 and Df(2R)27 (gift of R. Burgess) were crossed to wild-type flies and the Df/+ chromosomes were analyzed. The distal breakpoints for each deficiency were determined. The insertion sites of the P elements #4412 and #13403 (Torok et al., 1993) were confirmed by in situ hybridization to polytene chromosomes following a standard protocol (Ashburner, 1989). Two changes to the procedure were made: the chromosomes were dissected in 0.7% NaCl and the acetylation step was skipped.

Genomic DNA was isolated, electrophoresed, transferred and probed using standard techniques (Sambrook et al., 1989).

To localize ix by RFLP mapping, pairs of closely linked markers flanking the ix locus were employed. The P elements P[w⁺]4412, inserted at 47D, and P[w⁺]13403, inserted at 48A, were used (Torok et al., 1993). To generate recombination events proximal or distal to the ix² mutation, w/w; P[w⁺]4412 ix²/CyO females were crossed to w; P[w⁺]13403/CyO males, and the Cy⁺ female progeny (w/w; P[w⁺]4412 ix²/P[w+]13403) were collected as virgins and crossed to w; Sp/CyO; Sb/TM2 males. The male progeny of the latter cross were scored by eye color. Males with white eyes (no P element) and males with darker eye pigmentation (two P elements) were crossed to w/w; Sp/CyO; Sb/TM2 virgin females to establish stocks of the recombinant chromosomes. To determine whether the recombinant chromosomes carried ix², and thereby to determine the location of the crossover relative to the ix locus, males carrying the recombinant chromosomes were crossed to w/w; ix²/CyO virgin females.

DNA samples isolated from P[w⁺]4412 ix²/CyO and P[w⁺]1340/CyO flies were digested with 24 restriction enzymes and probed with DNA fragments from the ix chromosomal walk (Fig. 2B) to identify RFLPs between the two parental chromosomes. DNA samples isolated from the fly stocks established for 22 recombinant chromosomes balanced with the CyO chromosome were then analyzed using the restriction enzymes and DNA probes that identified RFLPs. This analysis indicated that six crossover events mapped proximal and 16 crossovers mapped distal to the ix² mutation.

DNA polymorphisms between the parental chromosomes were identified and used to analyze the recombinant chromosomes. Southern analysis with four DNA fragments (R16, 2G, 4C, P6.1) distributed across the 100 kb DNA walk (Fig. 2B) detected DNA polymorphisms between the P[w+]4412 ix² and P[w+]13403 parental chromosomes. All six crossover events proximal to ix were also proximal to the R16 fragment, located within 5 kb of the Df(2R)27 breakpoint, which indicated that all of the proximal recombination events isolated fail to further localize ix.

Analysis of the distal crossovers was more informative. Using the 4C and P6.1 DNA clones as probes detected 3 recombination events proximal to these probes, and the remaining crossovers occurred distal to these fragments. Results with the 4DXR DNA fragment as a probe demonstrated that all three of the crossovers proximal to phage 4C are distal to this probe. Unfortunately, because of the uneven distribution of recombination events, the location of ix cannot be ascertained by regression of crossover frequency on a physical map. However, the results of this RFLP analysis further localized ix to the region proximal to the 4C phage clone and distal to the Df(2R)27 breakpoint, a 65 kb region.

Wild-type (Canton-S) polyA+ RNA (5 μg per lane) from females and males was electrophoresed on a 1% agarose/1.85% formaldehyde gel, then transferred to a Hybond-N+ membrane and fixed by alkali treatment. For sizing bands detected by autoradiography, an RNA ladder (Life Technologies, Rockville, MD) was also run on the gel and visualized by staining with ethidium bromide. Hybridization with an antisense ix probe labeled with [α-³²P]UTP was carried out overnight in 5×SSPE, 5×Denhardt’s solution, 0.4% SDS, 10 μg/ml salmon sperm DNA, 50% formamide at 60°C. Two washes in 2×SSC/0.1% SDS at room temperature for 15 minutes were followed by one wash in 1×SSC/0.1% SDS at 65°C for 15 minutes, two washes in 0.1×SSC/0.1% SDS at 65°C for 10 minutes, two washes in 0.1×SSC/0.1% SDS at 70°C for 10 minutes, two washes in 0.1×SSC/0.1% SDS at 78°C for 10 minutes, and two washes in 0.1×SSC/0.1% SDS at 85°C for 10 minutes. The blot was then exposed to film. For probe-making, the ix cDNA was cloned into the HincII and EcoRI sites of pBluescript KS II + (Stratagene, La Jolla, CA) as a MscI-EcoRI fragment. The plasmid was linearized by digestion with ClaI, then transcribed by T7 RNA polymerase in the presence of [α-³²P]UTP.

As a loading control, the same blot was subsequently hybridized with a probe from the ninaE gene, which encodes the major rhodopsin, RH1 (O’Tousa et al., 1985). The ninaE gene was chosen as a control because the transcript is not expressed sex-specifically (data not shown). The rp49 gene (O’Connell and Rosbash, 1984), typically used as a loading control, was found to be expressed at higher levels in females than in males (data not shown) and therefore was determined not to be a good loading control for comparing the two sexes. The ninaE probe was labeled with [α-³²P]dCTP by extension of random hexamers, using as a template a PCR-amplified region of the gene from the beginning of exon 2 through the beginning of exon 5 (nucleotide positions 365 through 1716 in GenBank Accession Number K02315). Hybridization conditions were as above for the ix probe. Two washes in 2×SSC/0.1% SDS at room temperature for 15 minutes were followed by one wash in 1×SSC/0.1% SDS at 42°C for 15 minutes, two washes in 0.1×SSC/0.1% SDS at 68°C for 15 minutes, and two washes in 0.1×SSC/0.1% SDS at 75°C for 15 minutes. Relative intensities of male and female signals for the ix hybridization and the ninaE hybridization were determined by analyzing scanned autoradiographs with NIH Image 1.62 software.

To determine which one of the genes in the region to which ix had been localized was ix, two genomic rescue constructs were made and tested for their ability to rescue the ix phenotype. The 2GB construct was made by subcloning the 5.8-kb BamHI-EcoRI genomic fragment from phage 2G into the CaSpeR4 vector (Pirrotta, 1988). The 1GS construct was made by subcloning the 12-kb SalI fragment from phage 1G into the XhoI site of the CaSpeR4 vector.

A knockout construct for each gene present in the 2GB construct, designated R, G and H, was generated to test for the inability to rescue the ix phenotype. The knockout construct, 3GBRΔ, which deletes the R gene after amino acid (aa) 16, was derived from the 2GB construct by the following procedure. The 2GB DNA was digested with SpeI and EcoRI, the ends were filled in with the Klenow fragment of DNA polymerase I (New England Biolabs, Beverly, MA), and the 9.7 kb SpeI-EcoRI CaSpeR4 vector + genomic DNA fragment was isolated. In a separate reaction the 2GB plasmid was digested with SpeI, the ends were filled in with Klenow, and the 1.7-kb SpeI genomic DNA fragment was isolated. The 1.7-kb SpeI blunt-ended DNA fragment was ligated to the 9.7-kb SpeI-EcoRI blunt-ended DNA fragment. To identify the desired construct, the DNA from candidate clones was digested with PstI to determine the orientation of the 1.7 kb SpeI fragment and to confirm the presence of the 0.7 kb deletion. The knockout constructs for the G and H genes, 3GBG* and 3GBH* respectively, were derived from the 3GB construct, which contains a 1.0 kb deletion that removes the adjacent tRNA:SeC and trypsin iota genes by the following procedure. The 3GB construct is a derivative of the 2GB construct. The 2GB DNA was digested with BglII and EcoRI, the ends were filled in with Klenow, and then the DNA was ligated to itself generating a 1.0 kb deletion. For the 3GBG* construct, a stop codon was inserted at amino acid 91 by digesting the 3GB plasmid with SstII, recessing the ends with T4 DNA polymerase (New England Biolabs, Beverly, MA), and ligating a 12 bp linker, NheI* (New England Biolabs, Beverly, MA), containing an NheI site and stop codons in all three reading frames, to the blunt-ended 3GB DNA. Candidate clones were screened for the presence of the unique NheI site and the absence of the unique SstII site. The 3GBH* construct was made using the same steps as those used to make the 3GBG* construct, except the unique SfiI site was used instead of the SstII site and the stop codon was inserted at aa 44. Both 3GBG* and 3GBH* knockout constructs were further verified by DNA sequencing using the primers: G*1 (5′-CTCGCGGACAACTTAAAGAG) and H*1 (5′-GACAAGTTTTACGTGGAC).

Heat-shock-inducible cDNA (hscDNA) constructs were made to test rescue of the ix phenotype. The G and H cDNAs were subcloned into the HpaI and NotI sites of the CaSpeRhs vector. cDNA H was inserted as a PvuII-NotI fragment, and cDNA G2 was inserted as a HincII-NotI fragment into the same vector.

The 2GB, 1GS, and hscDNA constructs (0.3 μg/μl) were injected separately into w¹¹¹⁸ embryos with the transposase source Δ2-3 (0.1 μg/μl) (Laski et al., 1986), following standard techniques (Rubin and Spradling, 1982; Spradling and Rubin, 1982). The knockout constructs were injected at the concentration 0.4 μg/μl with the transposase source Δ2-3 (0.1 μg/μl) following the same method. G0 adults were crossed to w¹¹¹⁸; Sp/CyO; Sb/TM2 flies of the opposite sex to identify transformants. All F1 progeny with pigmented eyes were crossed to w¹¹¹⁸; Sp/CyO; Sb/TM2 flies of the opposite sex to determine into which chromosome the construct inserted. The transgenes inserted into either the X or third chromosome were tested for rescue of the ix² mutation. Larvae carrying the hscDNA constructs were grown at 29°C continuously or heat shocked at 37°C for 1 hour each day during larval growth to assay rescue of the ix² mutation.

The genomic sequence of gene G was determined by sequencing the RΔ construct genomic DNA from gene H to gene R on both strands by cycle sequencing using dye termination reactions (Applied Biosystems, Foster City, CA). The primers used in the sequencing reactions were:

G#1, 5′-GAAAACAATTCGCGGCTGTTCAATATTTT;

G#4, 5′-TGCGCGGCACTAATCAGAGTGTCGTGT;

G#7, 5′-TTCACCCTGGAAATGTTGTCCAATTTTTCGGCCT;

G#8, 5′-CAAGGACTACCCAATATTTCATATTGTTACATACATAAAAGT;

G#S1, 5′-CTCGCGGACAACTTAAAGAG;

G#S2, 5′-TCACACGCATGCACTTAAGTTAAG;

R#S2, 5′-CTTCATTGCAGGTGGGTG; and

5′UTR#2, 5′-ATGAGATGACAGCTCTTTCCGGTCGGTTGACATTAGCTA.

Total RNA from wild-type (Oregon-R) females and males was isolated from 4- to 5-day-old adult flies using TRIzol reagent (Life Technologies, Rockville, MD) according to the manufacturer’s instructions. mRNA was purified from total RNA by binding of polyA+ RNA to dC₁₀T₃₀ oligonucleotides linked to polystyrene-latex beads (Qiagen, Valencia, CA). For probe-making, the 457 bp MscI-PstI genomic DNA fragment containing the ix translation initiation codon was subcloned into the HincII and PstI sites of pBluescript KS II + (Stratagene, La Jolla, CA). The plasmid was linearized by digestion with XhoI, then transcribed by T7 RNA polymerase in the presence of [α-³²P]UTP, producing a uniformly labeled 527-bp antisense ix probe. The full-length probe was excised from a 5% acrylamide (19:1 acrylamide:N,N′-methylenebisacrylamide)/8 M urea gel and eluted in 350 μl of 0.5 M ammonium acetate/1 mM EDTA/0.2% SDS for 2 hours at 37°C. The RNase protection assay was performed using reagents supplied by Ambion (Austin, TX) according to the manufacturer’s instructions. PolyA+ female and male RNA (1.8 μg) were each combined with 15 μl of the probe eluate. Two control tubes containing 50 μg of total yeast RNA and 15 μl of probe eluate were also prepared. Sample and probe were allowed to hybridize 16 hours at 42°C. The female and male fly RNA hybridization and one of the yeast control hybridization reactions were then digested with a 1:100 dilution of RNase mix (250 U/ml RNase A, 10,000 U/ml RNase T1); one yeast control hybridization was left undigested. The RNases were then inactivated and the nucleic acids precipitated and resuspended in 10 μl of gel loading buffer. The protected fragments were resolved on a 5% acrylamide/8 M urea gel. Only 10% of the yeast control without RNase digestion was loaded. φX174 DNA, digested with HaeIII and labeled with [α-³²P]dATP by fill-in with T4 DNA polymerase, was also loaded as a size marker. After electrophoresing, the gel was dried on Whatman 3MM chromatography paper and autoradiographed.

PolyA+ RNA (200 ng per reaction) from w¹¹¹⁸ females and males was reverse transcribed using gene-specific primer ix4 (5′-TGCGCGGCACTAATCAGAGTGTCGTGT). The 5′ RACE system (Life Technologies, Rockville, MD) was used to amplify the 5′-end sequence of the ix transcript, by first adding an oligo-dC tail to the 3′ end of the cDNA with terminal transferase, then performing PCR with gene-specific primer ix12 (5′-CGATGGCGAGGATTGCATTACCTGCATCAT) and an anchor primer complementary to the oligo-dC, followed by nested PCR amplification with gene-specific primer ixL (5′-GGCATCATGTTCATGTTGGGATTCAT) and a second anchor primer. The PCR conditions were 94°C for 1 minute; 35 cycles of 94°C for 1 minute, 55°C for 1 minute, 72°C for 2 minutes; then 72°C for 7 minutes. These amplification products were cloned and sequenced. A corresponding RT-PCR experiment using the same PCR conditions was performed on the same first-strand cDNA reactions (without dC-tailing) using a gene-specific primer, 5′UTR3 (5′-AATGCTAAATGAAACATTACACATCGTTTTTTATTTGGGA), instead of the RACE anchor primers, for the two nested amplification reactions. RT-PCR products appearing to be splice variants because of their smaller size than predicted from genomic sequence, were cloned and sequenced.

Wild-type (Canton-S) total RNA (5 μg per reaction) from females and males was reverse transcribed using an oligo-dT-containing adapter primer (Life Technologies, Rockville, MD). This first-strand cDNA was then subjected to PCR amplification: 94°C for 1 minute; 30 cycles of 94°C for 1 minute, 56°C for 1 minute, 72°C for 1 minute; then 72°C for 7 minutes, using gene-specific primer ix943U (5′-TTAAAGAGGGACACGGGTGC) and a universal amplification primer with sequence matching the non-oligo-dT segment of the adapter primer (Life Technologies, Rockville, MD). A second, nested PCR amplification reaction was performed using gene-specific primer ix1032U (5′-CTTGAAGACGGCGATGCAGT) and the same universal amplification primer. These amplifications yielded single products of approximately 350 bp for both female and male RNA. The amplification products were cloned and sequenced.

The lacZ activities were measured according to a previously published protocol (Coschigano and Wensink, 1993) incorporating published modifications (Li and Baker, 1998b).

The Yp data were analyzed using a two-factor analysis variance (ANOVA), with ix genotype and transgene presence/absence as fixed main effects for the pML-58 experiments and with ix genotype and dsx genotype as fixed main effects for the pCR1 experiments. To detect interactions between ix and dsx, the pCR1 experiment data were log-transformed before ANOVA, as multiplicative effects in the raw data become additive in the transformed data. Bristle counts for vaginal teeth, LTRB and sixth sternite were found to have heterogeneous variances among genotype classes, so the non-parametric Mann-Whitney U test (1-tailed) was used to detect differences between the classes. A G-test with the Yates correction was used to analyze the dorsolateral anal plate data. For sixth-tergite pigmentation, arcsin-transformed data were analyzed by one-tailed Student’s t-test.

The Matchmaker Gal4 two-hybrid system (BD Biosciences Clontech, Palo Alto, CA) was used according to the manufacturer’s protocols. Briefly, full-length IX-, DSX^F- and DSX^M-coding sequences were cloned into the pAD and pBK vectors, which were co-transformed in pairs into the AH109 yeast strain and plated on -Ade, -His, -Leu and -Trp restrictive medium. Transformants that grew on this restrictive medium were further assayed for positive interactions by a colony-lift β-galactosidase assay.

IX, DSX^F and DSX^M coding sequences were cloned in frame into the pAc5.1/V5-HisB or pMT5.1/V5-His (Invitrogen Corporation, Carlsbad, CA) vectors. To generate AU1-tagged constructs, the V5 epitope and polyhistidine regions of the V5-tagged constructs were replaced by digestion with BstBI and PmeI and ligation with an oligonucleotide dimer (5′-GTT/CGAAGACACCTATCGCTATATACGTA/CCGGTCA) containing an in-frame AU1 epitope (Covance, Princeton, NJ). Drosophila S2 cells were cultured in Schneider’s Drosophila medium (Gibco) in 10% FCS. Transfections were performed using Effectene reagent (Qiagen, Valencia, CA) according to the manufacturer’s protocol and as described elsewhere (Mosher and Crews, 1999). Briefly, cells were washed in PBS pH 7.4 and plated in media at a density of 5-7×10⁵ cells/ml in a six-well plate (1.6 ml/well). Plasmid (∼2 μg) and Effectene mixture was added and cells were grown for 24 hours prior to induction with 500 mM copper sulfate for an additional 24 hours.

Cells were washed in PBS and nuclear extracts (200 μl/well) were prepared as described elsewhere (Huang and Prystowsky, 1996). Extracts were normalized to equal protein concentrations and 100 μl samples were incubated with 2 μl monoclonal anti-AU1 antibody (Covance, Princeton, NJ) for 1 hour at room temperature. Bovine serum albumin was added to 2% final volume and lysates were incubated with Protein G Sepharose beads (Amersham Pharmacia Biotech, Piscataway, NJ) for an additional 2 hours at room temperature. Beads were pelleted, washed and transferred to SDS loading buffer. Proteins were resolved on a 12% SDS gel and probed via western blot using rabbit polyclonal anti-V5 antisera (Medical & Biological Laboratories, Nagoya, Japan) according to standard protocols.

Drosophila S2 cells were cultured and transfected as described above. Nuclear extracts (200 μl/well) were prepared as described above. Probe fragments were made by ³²P end-labeling the 185 bp ClaI-BglII FBE fragment of the Yp promoter region described previously (Burtis et al., 1991). Extracts (10 μl) were incubated in binding buffer [4% glycerol, 1 mM MgCl₂, 0.5 mM EDTA, 0.5 mM DTT, 50 mM NaCl, 10 mM Tris-HCl pH 7.5, 0.05 mg/ml poly-dI-dC, protease inhibtor cocktail (Roche Applied Science, Indianapolis, IN, catalog number 1697498)] and 2 μl (50-100K cpm) probe was added for 20 minutes at room temperature. Monoclonal anti-V5 (Invitrogen Corporation, Carlsbad, CA) or anti-AU1 (Covance, Princeton, NJ) antibody was added in samples for super-shift where indicated. Proteins were resolved via native PAGE (4% acrylamide, 5% glycerol, 0.5×TBE) and complexes were visualized via autoradiography.

Complementation tests with deficiencies and loss-of-function alleles of ix localized ix to the cytological region 47E-47F11-18 (Chase and Baker, 1995). In addition to the previously characterized deficiencies, two new deficiencies were tested (Fig. 2A). Df(2R)17 fails to complement ix and Df(2R)27 complements ix. These complementation results place ix in the cytological region 47F between the Df(2R)27 breakpoint at 47F1 and the Df(2R)ixⁱ³ breakpoint at 47F11-18. However, one complementation test with Df(2R)ix⁸⁷ⁱ³ gave a result that was not consistent with that localization of ix. Df(2R)ix⁸⁷ⁱ³ complemented a temperature-sensitive allele, ix4, at the nonpermissive temperature, although it failed to complement all other ix alleles tested (Chase and Baker, 1995). After ix was molecularly identified, it was determined that the Df(2R)ix⁸⁷ⁱ³ chromosome contains a more complex rearrangement. In addition to the deletion of 47D-47F11-18, at least 6 kb of DNA containing ix (which is located approximately 70 kb from the 47F11-18 deletion breakpoint) was transposed to cytological position 50 on chromosome arm 2R. Thus, the complementation of the temperature-sensitive allele of ix by this deficiency chromosome was due to the transposed ix locus. A chromosomal walk was completed through the 47F interval and the relevant deficiency breakpoints were mapped to the DNA in the walk (Fig. 2B). The region between the Df(2R)27 and Df(2R)ix⁸⁷ⁱ³ breakpoints, within which ix is located, is 100 kb. To localize ix in this 100 kb region, restriction fragment length polymorphism (RFLP) mapping was carried out (see Materials and Methods). This narrowed the region of interest to 65 kb.

To locate candidate genes in the 65 kb region identified by the RFLP mapping, phage clones covering the entire region were used as probes to isolate cDNAs (B. C. W., C. M. G.-E., E. Williams and M. L. Goldberg, unpublished). Five cDNA classes were identified (Fig. 3A). However, for most of the classes only one cDNA was isolated, raising the possibility that other genes may reside in this region and were not detected. Analysis of the Drosophila genome sequence (Adams et al., 2000) indicated that four additional transcripts (CT25954, CT25948, CT32444 and CT25938) are predicted in this region. Some or all of these predicted transcripts represent potential additional candidate genes. In addition, a tRNA:SeC gene and a cluster of trypsin genes (Wang et al., 1999) were previously mapped to the ix region.

P-element-mediated germline transformation experiments were undertaken to determine whether one of the genes identified in the 65 kb region was ix. The genomic rescue construct 2GB, which contains a 6 kb segment from the proximal part of the 65 kb ix walk, encompassing three genes of unknown function, as well as the tRNA:SeC and trypsin iota genes, was tested first for rescue of the ix phenotype (Fig. 3B). Two 2GB lines with insertions on the third chromosome were tested for rescue of the ix² mutation, and one line with the transgene inserted on the second chromosome was recombined onto the Df(2R)en^B chromosome and then tested for rescue of the ix²/Df(2R)en^B phenotype. The somatic phenotype of the ix mutant females carrying the transgenes ranged from fully rescued (normal female development) to no rescue of the intersexual phenotype, depending on the lines tested and whether one or two copies of the transgenes were present. Because the ix phenotype was rescued for some of the ix mutant females, ix is one of the genes contained within the 2GB construct (Fig. 3B). The variable rescue of the ix phenotype suggested that this 6 kb genomic construct containing the ix gene was sensitive to position effect. Transformants with an overlapping construct 1GB were tested, and this construct also rescued the ix²/Df(2R)en^B phenotype (Fig. 3B). The region of overlap between these two constructs is 4.5 kb and contained the three candidate genes R (CG12384), G (CG13201) and H (CG12352), but not the tRNA:SeC and trypsin iota genes.

To ascertain whether the DNA sequence of these three genes might indicate which is most likely ix, a cDNA representing each gene was sequenced. The amino acid sequence of each predicted protein was compared with sequences in the GenBank CDS translation, PDB, SwissProt, PIR and PRF databases using the PSI Blast program. Gene R encodes a protein that is 43% identical and 52% similar to the human death associated protein 1 (DAP1) (Deiss et al., 1995). The predicted H protein is 19% identical and 37% similar to the Saccharomyces cerevisiae ARD-1 (arrest-defective-1) protein (Whiteway and Szostak, 1985), and 22% identical and 43% similar to a N-acetyltransferase ARD-1 human homolog. Gene G encodes a novel protein. The sequence similarities of the candidate genes did not indicate that one gene was a better ix candidate than the others.

P-element-mediated germline transformation experiments were carried out with additional genomic constructs, to determine which candidate gene is ix. The approach taken was to knock out each candidate gene individually, while leaving the other two genes intact and to assay each of these derivatives of 2GB for the inability to rescue the ix phenotype. The 3GBRΔ construct deletes all but the first 16 amino acids of the R protein, the 3GBG* construct introduces a stop codon in the middle of the G protein at amino acid position 91, and the 3GBH* construct introduces a stop at amino acid position 44 of the H protein. For the 3GBRΔ construct, 21 lines were isolated with 13 insertions on the third chromosome, for the 3GBH* construct 43 lines were isolated with 18 insertions on the third chromosome, and for the 3GBG* construct 21 lines were isolated with 13 insertions on the third chromosome. The transgenic lines with insertions on the third chromosome were tested for rescue of the ix² phenotype. All 13 of the 3GBG* lines failed to rescue the ix phenotype (Fig. 3B, Fig. 4). For comparison, only one out of the 13 3GBRΔ lines and 1 of the 10 3GBH* lines tested failed to rescue the ix phenotype. This analysis of the knockout transgenes indicated that candidate G was ix.

To establish unequivocally that candidate G was ix, a heat-shock-inducible cDNA construct for gene G was tested for rescue of the ix phenotype. One of the five lines tested for the hscDNA G construct partially rescued the ix² phenotype when the larvae were grown continuously at 29°C (Fig. 4). Females carrying two copies of the transgene were rescued for the somatic defects but were not fertile, and females with one copy of the transgene were partially rescued. The results of experiments with the 3GBG* and hscDNA G transgenes showed that gene G is ix.

The ix gene encodes a protein of 188 amino acids (Fig. 5A,B). Analysis of the GenBank, EMBL and DDBJ EST databases using the Gapped Blast program identified mammalian ESTs and predicted proteins with significant similarity to the ix protein. The functions of the genes represented by these ESTs are unknown. From amino acids 15 to the C terminus of IX, the longest EST, a mouse EST (AA388092), is 37% identical and 52% similar to the ix protein; this mouse EST does not show similarity to the N-terminal 15 amino acids of IX. This similarity is highest in a 35 amino acid region of these proteins from amino acid 95 to amino acid 129. The sequence in the 35 amino acid region is 55% identical and 74% similar between the ix protein and either the mouse EST or a very similar human EST (U46237) (Fig. 5D). The stop codon introduced in the 3GBG* rescue construct is located just before this region. If the truncated G* protein is stable, the 35 amino acid region or a region after it must be required for ix function. Additionally, from amino acid 20 to the C terminus of IX, two predicted human proteins (XP_046121 and DKFZp434H247.1) are 35% identical and 50% similar to IX.

Comparison with sequences in the GenBank CDS translation, PDB, SwissProt, PIR and PRF databases using the PSI Blast program with aa position 3 to 47 in the N-terminal region of the ix protein revealed sequence similarity to the human synovial sarcoma translocation (SYT) protein (Clark et al., 1994), mouse SYT protein (de Bruijn et al., 1996) and the C. elegans suppressor of ras protein (SUR-2) (Singh and Han, 1995). In the 44 amino acid region of similarity, the ix protein is 45% identical and 51% similar to the human SYT protein, 50% identical and 52% similar to the mouse SYT protein, and 42% identical and 47% similar to SUR-2 (Fig. 5C).

The sur-2 gene was identified as a suppressor of the ras multivulva phenotype (Singh and Han, 1995). Genetic epistasis analysis placed sur-2 at the same position as transcription factors in the vulval signal transduction pathway (Singh and Han, 1995), suggesting that the sur-2 protein may function as a transcription factor.

The SYT protein is proposed to act as a transcriptional activator (Brett et al., 1997). In vitro analysis of SYT indicates that the 155 amino acid region of SYT with the highest transcriptional activation function contains the 44 amino acid sequence with similarity to ix (Brett et al., 1997). The sequence similarity of the IX protein to a region of the SYT protein that is capable of activating transcription raises the possibility that ix may function as a transcriptional activator.

Because the ix phenotype is female specific and some genes in the somatic sex-determination hierarchy are regulated at the level of splicing, it was conceivable that the ix pre-mRNA would be sex-specifically spliced. However, no introns were identified by comparing the genomic sequence with the ix cDNA sequence, and Northern analysis did not detect sex-specific transcripts (Fig. 6A). In both males and females, a single hybridizing RNA species of approximately 750 bp was observed, consistent with the expected transcript size as determined by 5′ and 3′ RACE, which is 734-766 bases (start position 612-626, end position 1332, tail 28-46 bases, Fig. 5A). The start position determined by 5′ RACE is variable in both males and females but does not show a sex-specific difference. The relative signal intensity in males and females for the ix northern hybridization was normalized by the relative signal intensity in males and females for hybridization to ninaE. The ix transcript is ∼8.7 times as abundant in wild-type females as in wild-type males. Preliminary data (not shown) from ix mutant germline clones in females, and from RNA analysis of females lacking a germline, suggest that the difference in transcript levels between females and males may be due to high ix expression in ovaries. The northern hybridization, cDNA analysis and 5′ RACE results suggest that the ix transcript is not sex specific and is not spliced.

However, sequence analysis of the genomic region just upstream of the transcription start site of the ix gene identified by 5′ RACE revealed a potential exon and intron. The putative exon would encode 33 amino acids and contain a consensus donor splice site (Fig. 5A). RT-PCR experiments, using a 5′ PCR primer that begins upstream of and extends into the putative exon, detected products that were of the size expected from the genomic DNA and smaller, apparently spliced, products were sometimes observed (data not shown). The RT-PCR result raises the possibilities that the transcription start determined by 5′ RACE is not correct or that a transcript initiating from an upstream start site is also expressed but at a much lower level, and was not detected by northern analysis or in the cDNAs isolated.

To confirm the ix pre-mRNA was not sex-specifically processed, RNase protection assays of polyA+ RNA isolated from males and females were performed using a probe that could distinguish between the spliced and unspliced products (Fig. 6B). RNase protection assays depend on neither reverse transcription nor amplification of the RNA as RT-PCR does, and RNase protection assays are more sensitive than northern analysis and could detect a rare transcript. The major protected fragment is approximately 200 bp (Fig. 6C), as expected for an unspliced transcript that begins at the site indicated by 5′ RACE. Additionally, no qualitative difference between male and female protected fragments was observed. These results agree with the northern data, cDNA analysis and 5′ RACE results, and indicate the ix pre-mRNA is not spliced. Therefore, alternative processing of the ix transcript is not responsible for the female-specific ix phenotype, suggesting that ix functions together with one or more female-specific proteins to achieve the sex-specificity of the ix phenotype.

As ix functions at approximately the same position in the sex determination hierarchy as dsx, we carried out genetic experiments to ascertain whether ix cooperates with, or functions independently of, dsx to control female sexual differentiation. We examined how ix and dsx function relative to one another in controlling Yp gene expression and the development of an array of sexually dimorphic cuticular structures.

We first focused on the role of ix in controlling Yp gene expression. Previous studies identified the Fat Body Enhancer (FBE) in the Yp1 and Yp2 intergenic region as necessary and sufficient for the sex-specific expression of both Yp1 and Yp2 (Garabedian et al., 1986). DSX regulates Yp gene expression through three DSX binding sites in the FBE (An and Wensink, 1995; Burtis et al., 1991; Coschigano and Wensink, 1993) and northern analysis suggests that ix⁺ was required for DSX^F mediated activation of Yp1 transcription (Waterbury et al., 1999). To confirm that ix regulates Yp gene expression and to determine whether ix activates Yp expression through the same regulatory region as dsx, the expression of Yp reporter gene constructs was assayed in wild-type and ix mutant females.

Our analysis of the expression levels of the pML-58 Yp reporter construct (provided by M. Lossky and P. Wensink), which contains the FBE and 196 bp of the Yp1 and Yp2 intergenic region fused to the lacZ gene, indicates that this region is sufficient for ix regulation of the Yp genes in females. Chromosomal females either homozygous or heterozygous for an ix mutation and either carrying or not carrying an ix⁺ transgene were compared. Including the transgene in the analysis allows definitive assignment of an effect on reporter expression to ix and not to a linked locus. A 1.9-3.5-fold reduction in lacZ activity from pML-58 reporter-construct expression was observed comparing homozygous and heterozygous ix-mutant females (Fig. 7A, ANOVA genotype main effect P<0.0001). Therefore there is a significant effect of the ix genotype on expression of the Yp reporter construct regardless of the presence or absence of the ix transgene. Additionally, the effect of the ix transgene on expression of the Yp gene reporter construct is to increase lacZ activity (ANOVA transgene main effect P<0.0001). There is no significant interaction between the ix genotype and the presence of the ix transgene (ANOVA interaction P=0.74), indicating that adding one wild-type copy of the ix gene, either at the ix locus or via the transgene, increases Yp reporter gene expression equivalently. As the ix-mutant females assayed are heteroallelic (ix²/ix³) and one copy of the ix transgene rescues the decreased Yp reporter construct expression observed in these ix-mutant females, the reduction in Yp gene expression is due to the ix mutation and not another mutation on the second chromosome. These results indicate that the ix protein activates transcription of the Yp gene reporter construct in females through a region that contains the DSX^F DNA-binding sites, raising the possibility that IX interacts with DSX^F to regulate expression of the Yp genes.

To investigate whether ix regulation of the Yp genes is dependent on dsx activity, expression of the pCR1 Yp reporter construct (Lossky and Wensink, 1995), which contains the entire Yp1 and Yp2 intergenic region fused to a lacZ reporter gene, was analyzed in ix, dsx and ix; dsx double mutant flies. The use of the full intergenic region, including the her responsive region (HRR) outside the FBE (Li and Baker, 1998b), increases the resolution of the analysis, as her upregulates Yp expression approximately fivefold, except when DSX^M is present, thereby amplifying expression differences due to the presence or absence of DSX^M activity. pCR1 expression was reduced in both ix mutant females (Fig. 7B, ANOVA ix genotype main effect P=0.0001) and dsx mutant females (ANOVA dsx genotype main effect P<0.0001). If ix and dsx act independently to regulate Yp expression, then the combined effect of the ix; dsx mutant would be the product of the individual mutant effects. Log-transforming the data makes multiplicative effects additive, so the ix×dsx interaction term in the ANOVA is an indicator of the independence of the effects of the two loci. The interaction between ix and dsx is highly significant (P<0.0001), indicating a strong dependent relationship between the two loci. In males, the ix genotype has no effect on the level of pCR1 expression (Fig. 7B, P=0.4), but the dsx genotype has a significant effect on pCR1 expression (P<0.0001). The interaction between ix and dsx is not significant (P=0.71). These results indicate that ix does not function in males to regulate Yp gene expression. Therefore, the ix protein only functions in females and cooperates with DSX^F to activate Yp gene expression.

In addition to regulating Yp expression, DSX^F controls the development of sexually dimorphic cuticular structures (Baker and Ridge, 1980; Li and Baker, 1998b). Because the ix phenotype is indistinguishable from the dsx female phenotype, IX may also interact with DSX^F to regulate these aspects of female differentiation. However, her, another gene with a phenotype similar to ix and dsx, cooperates with dsx to control female differentiation of foreleg bristles and tergites 5 and 6, but functions independently of dsx to regulate development of vaginal teeth and anal plates in females (Li and Baker, 1998b). If ix acts independently of dsx to regulate some aspect of terminal sexual differentiation, then in ix; dsx mutant females that aspect of sexual differentiation would be masculinized compared with the individual mutants. However, if the genes function together then the phenotype of ix; dsx double mutants would be the same as that of the single mutants. To test these possibilities, the phenotypes of five sexually dimorphic cuticular structures in ix, dsx and ix; dsx mutant flies were assayed.

The first cuticular phenotype examined was the number of vaginal teeth in females. There are on average 26.6 vaginal teeth on ix/+; dsx/+ females (Table 1, row 1) and 0 vaginal teeth on ix/+; dsx/+ males (Table 1, row 5). The intersexual ix and dsx single mutant females had on average 9.7 vaginal teeth and 6.0 vaginal teeth, respectively, significantly fewer than wild-type females [Table 1, compare rows 1 and 2 (P<0.0001), and rows 1 and 3 (P<0.0001)]. The ix; dsx mutant females formed on average 6.45 vaginal teeth, indicating that loss of wild type ix function does not masculinize dsx mutant females (Table 1, rows 3 and 4, P=0.77). This result indicates that dsx is dependent on wild-type ix activity for vaginal teeth development in females. Elimination of wild-type dsx function appears to weakly masculinize the ix mutant females (Table 1, rows 2 and 4, P=0.032). However, this effect may be due to the fact that the EMS-induced ix³ allele is a strong loss-of-function allele but is not completely null (Chase and Baker, 1995). The nucleotide sequence of ix³ is consistent with this inference, in that the only difference between ix³ and the ix⁺ allele of its progenitor stock is a T to A substitution in nucleotide 1221, which results in a Ser to Arg amino acid substitution (data not shown). We thus conclude that ix and dsx act interdependently to regulate differentiation of vaginal teeth in females.

The next sexually dimorphic structure analyzed was the anal plates. Females have one dorsal and one ventral anal plate, and males have two lateral anal plates. Intersexual flies have a pair of dorsolateral plates that are often fused at the dorsoanterior side. When collecting the data for all genotypes, the wild-type female dorsal anal plate was considered fused. All ix/+; dsx/+ females had fused dorsal anal plates (DLAP) compared with 65% of the dsx mutant females [Table 1, compare rows 1 and 3 (P=0.0073)] and 90% of the ix mutant females [Table 1, compare rows 1 and 3 (P=0.46)] that had fused DLAP. Although 90% of the ix mutant females had fused DLAP, these anal plates appeared intersexual because the two anal plates were not completely fused into one anal plate. Thus the absence of statistical significance should not be taken as evidence against a sex-transforming effect of ix on the anal plates of females. Similar to the single mutant phenotypes, 85% of the ix; dsx mutants had fused DLAP [Table 1, compare rows 3 and 4 (P=0.27) and rows 2 and 4 (P=1)], indicating that the ix; dsx double mutant phenotype is not stronger than the single mutant phenotypes. Therefore, ix and dsx do not act independently in the anal-plate precursor cells to control female-specific differentiation of anal plates.

Two other sexually dimorphic cuticular phenotypes analyzed were the extent of pigmentation of the sixth tergite and the development of the last transverse row of bristles (LTRB) on the basitarsus, which form the sex combs in males. With respect to pigmentation of tergite 6, the ix/+; dsx/+ females had on average 53% pigmentation (Table 1, row 1) and the ix/+; dsx/+ males had 97% pigmentation (Table 1, row 5). The loss of wild-type ix or dsx activity increased pigmentation of the sixth tergite significantly, to 92% and 96%, respectively [Table 1, compare rows 1 and 2 (P<0.0001), and rows 1 and 3 (P<0.0001)]. Similarly, in the ix; dsx double mutant females, the extent of pigmentation was 95% [Table 1, compare rows 3 and 4 (P=0.65), and rows 2 and 4 (P=0.045)]. Therefore, the double mutant phenotype is comparable with the single mutant phenotypes. The results for the analysis of differentiation of LTRB were the same as for pigmentation of tergite 6. In ix/+; dsx/+ females, an average of 5.28 bristles formed compared with 10.6 bristles on ix/+; dsx/+ males (Table 1, rows 1 and 5). The number of bristles increased significantly, to 6.8 in ix mutant females and 7.75 in dsx mutant females [Table 1, compare rows 1 and 2 (P<0.0001), and rows 1 and 3 (P<0.0001)], and the ix; dsx double mutant females had 7.25 bristles, similar to the single mutants [Table 1, compare rows 3 and 4 (P=1.0) and rows 2 and 4 (P=0.004)]. For both sixth-tergite pigmentation and LTRB differentiation, the ix; dsx mutant females were weakly masculinized compared with the ix mutant females but not the dsx mutant females. Again, this observation is presumably due to the fact that the ix³ allele is not completely null and residual ix activity is eliminated in the double mutant by loss of dsx function. These results indicate that for female-specific development of the LTRB and pigmentation of the sixth tergite, ix and dsx function cooperatively.

The last cuticular phenotype examined was the number of bristles formed on the sixth sternite. Females have an average of 20.85 bristles (Table 1, row 1) and males have an average of 0.30 bristles (Table 1, row 5) on the sixth sternite. Previous analysis indicated that bristles form on the sixth sternite of females independent of the activity of dsx and her (Li and Baker, 1998b). A comparison of the phenotypes of ix mutant females and ix/+; dsx/+ females demonstrates that the development of these bristles is also independent of ix function [Table 1, compare rows 1 and 2 (P=0.73)]. In this analysis the dsx mutant females had a slight reduction in the number bristles, from 20.85 to 19.50 [Table 1, compare rows 1 and 3 (P=0.0045)]. However, this weak effect of the dsx genotype was not observed in previous studies and may represent differences in the genetic background of these flies.

The analysis of these sexually dimorphic cuticular structures in males indicated that ix does not function in males. Unlike the dsx mutant males, ix mutant males did not develop vaginal teeth nor do they have fused dorsal lateral anal plates (Table 1, rows 6 and 7). The extent of pigmentation of the sixth tergite and differentiation of the LTRB on the basitarsus were also unaffected (Table 1, rows 5 and 6). The wild-type activity of both dsx and her is required in males to prevent bristle formation on the sixth sternite (Li and Baker, 1998b). However, loss of ix function did not increase bristle formation on S6 [Table 1, compare rows 5 and 6 (P=0.18)]. These results confirm previous results that ix only functions in females (Chase and Baker, 1995).

In all cases, we examined in which ix and dsx regulate female differentiation of cuticular structures – formation of vaginal teeth, development of the dorsal anal plate, development of the LTRB and pigmentation of the sixth tergite – the phenotype of the ix; dsx double mutant females was not masculinized compared with the ix and dsx single mutant phenotypes. Although the analysis of vaginal teeth differentiation, pigmentation of the sixth tergite, and formation of LTRB revealed that elimination of wild type dsx activity weakly masculinized ix mutant females, this effect probably represents the elimination of residual ix activity because the ix alleles may not be complete loss of function alleles (Chase and Baker, 1995). Therefore, the results of the phenotypic analysis of sexually dimorphic cuticular structures and the Yp gene reporter constructs indicate that ix and dsx act interdependently to regulate all aspects of female terminal differentiation.

Given the sex-specific effect of ix mutants, and the dependence of DSX^F function on ix genotype, we sought to determine if IX interacts directly with DSX^F to regulate transcriptional targets such as the Yp genes. As a preliminary test for such an interaction, we used a yeast 2-hybrid assay to look for an interaction between IX and either the DSX^F or DSX^M proteins. Constructs fusing full-length IX, DSX^F or DSX^M proteins with the Gal4 activation or DNA-binding domains were created and co-transformed into a yeast strain containing metabolic and enzymatic reporters for Gal4 function. Positive interactions were assayed by growth on restrictive medium as well as by lacZ expression (see Materials and Methods). By these criteria, the IX fusion proteins exhibit positive homomeric interaction, and exhibit heteromeric interaction with DSX^F, but not DSX^M, fusion proteins (Fig. 8A). Control transformants, containing only single constructs, failed to demonstrate either growth on restrictive medium or lacZ expression.

To confirm the results of our two-hybrid assay, we performed co-immunoprecipitation of tagged IX and DSX proteins expressed in Drosophila S2 cells. Constructs capable of expressing IX tagged with an AU1 epitope and either DSX^F or DSX^M tagged with a V5 epitope were co-transfected into S2 cells, extracts of which were subsequently immunoprecipitated using monoclonal anti-AU1 antibody. The immunoprecipitates and supernatants were resolved via SDS-PAGE and analyzed by western blot with polyclonal anti-V5. The AU1-epitope-tagged IX is able to co-immunoprecipitate DSX^F-V5 but not DSX^M-V5, indicating that IX specifically forms a stable complex with DSX^F in vivo (Fig. 8B). Analysis of supernatants confirmed that all proteins were expressed upon induction.

Because DSX^F functions as a transcription factor, we sought to determine if the complex between IX and DSX^F proteins is able to bind DNA effectively. We performed electrophoretic-mobility shift assay (EMSA) using as probe the previously characterized 185 bp FBE region of the Yp enhancer, which contains DSX-binding sites (Burtis et al., 1991). Nuclear extracts from S2 cells transfected with the epitope-tagged constructs discussed above were incubated with ³²P end-labeled FBE fragments and resolved by native PAGE. A stable DNA-binding complex was seen in extracts containing IX and DSX^F (Fig. 8C, lanes 1-5). To confirm that this complex contained IX and DSX^F, extracts were incubated with probe in the presence of one of three antibodies – anti-AU1, anti-V5 or nonspecific mouse IgG. That the predominant DNA-binding complex is specifically super-shifted by antibodies to the individual tags indicates that the complex contains minimally IX and DSX^F (Fig. 8C, lanes 6-8).

To begin to understand how ix regulates the terminal differentiation genes in females and how the sex-specificity of the ix phenotype is achieved, we have cloned the ix gene. The N-terminal 44 amino acids of IX share sequence similarity with the human and mouse synovial sarcoma translocation (SYT) proteins (Clark et al., 1994; de Bruijn et al., 1996) and C. elegans SUR-2 (Singh and Han, 1995). The remainder of IX has sequence similarity to mammalian ESTs. The function of the genes represented by these ESTs is unknown.

The SYT and SUR-2 proteins are proposed to function as transcription factors (Brett et al., 1997; Singh and Han, 1995). Human SYT was first identified as a chimeric protein resulting from a chromosomal translocation that is implicated in synovial sarcomas (Clark et al., 1994), and the region of the SYT protein that activates transcription in in vitro assays (Brett et al., 1997) contains the region with similarity to the ix protein. The SYT chimeric protein is nuclear, as expected for a transcription factor (Brett et al., 1997; dos Santos et al., 1997). As SYT does not contain a known DNA-binding motif (Clark et al., 1994), it is thought to form a complex with a DNA-binding protein to activate transcription.

The sequence similarity of IX to the human and mouse SYT proteins and to SUR-2 suggests that ix may also act as a transcription activator. Additionally, like the SYT proteins, IX does not contain a recognizable DNA-binding domain. dsx and her function at the same position in the hierarchy as ix and these genes encode proteins with zinc-finger DNA-binding domains (Erdman and Burtis, 1993; Li and Baker, 1998a). However, neither DSX^F nor HER proteins can activate transcription alone in 2-hybrid assays (this paper; H. Li, data not shown), suggesting these proteins lack activation domains and interact with additional proteins to regulate the expression of the terminal differentiation genes in females. The DSX^M protein has a 152 amino acid male-specific C terminus, whereas the smaller DSX^F protein has only 30 unique amino acids at its end (Burtis and Baker, 1989). Therefore, the DSX^F protein may need to interact with a co-factor for female-specific activity. The genetic results in this paper, indicating that dsx and ix act interdependently to regulate female-specific differentiation, and the biochemical results, indicating the DSX^F and IX physically interact, suggest that IX may be this co-factor. It remains to be determined whether the specific interaction of DSX^F and IX is mediated through the 30 amino acid C terminus of DSX^F.

As the ix phenotype is female specific and expression of other genes in the somatic sex determination hierarchy is controlled sex-specifically, expression of the ix gene could have been sex-specifically regulated. However, XY flies expressing a cDNA corresponding to DSX^F are phenotypically female (Waterbury et al., 1999) instead of intersexual, suggesting that ix protein is present in these chromosomal males. Our analysis of ix cDNAs, northern hybridization and RNase protection assays demonstrated that the ix pre-mRNA is not sex-specifically spliced. Therefore, the female-specific phenotype is not achieved through alternative processing of the ix transcript. The previous genetic results and our molecular results suggest that the ix protein is present in both females and males and its female-specific function is mediated through interactions with the female-specific protein DSX^F.

Analysis of Yp gene expression demonstrated that dsx, her and ix control Yp gene expression in the fat body (An and Wensink, 1995; Burtis et al., 1991; Coschigano and Wensink, 1993; Li and Baker, 1998b; Waterbury et al., 1999). Our results indicate that ix acts through the Yp intergenic region that contains the DSX-binding sites. Additionally, expression of DSX^F in ix mutant males is not sufficient to activate Yp expression (Waterbury et al., 1999), suggesting DSX^F requires IX to regulate Yp expression. Our analysis of Yp reporter constructs in ix; dsx mutant females also suggests that IX and DSX^F act together to control Yp gene transcription. Therefore, DSX^F may require IX as a co-factor to directly regulate Yp gene expression in females. This possibility is supported by the observation that IX and DSX^F are present in a complex that binds the region of the Yp FBE that contains DSX-binding sites.

Phenotypic analysis of ix; dsx mutant females demonstrated that ix and dsx also cooperate to regulate female-specific differentiation of sexually dimorphic cuticular structures. The ix mutation failed to masculinize the dsx mutant females, indicating that dsx is dependent on ix activity in the precursor cells that differentiate into the vaginal teeth, dorsal anal plates, last transverse row of bristles on the basitarsus and sixth tergite pigment-producing cells. Additionally, the phenotypic analysis of ix mutant males confirmed that ix does not function in males. The possibility that ix also functions with her to control female-specific differentiation of some sexually dimorphic structures remains to be tested. The tight interdependence of DSX^F and IX suggests that the relationship between HER and IX is likely to be the same as that between HER and DSX in females.

Understanding of the role of the sex determination hierarchy in sex-specific differentiation has been substantially revised and enhanced by recent studies that have begun to illuminate how information from the sex determination hierarchy is integrated with information from other developmental hierarchies. In particular, it had been thought that dsx played a mainly permissive role in the development of the internal and external genitalia. These structures develop from the genital imaginal disc, which is composed of three primordia deriving from embryonic abdominal segments A8, A9 and A10. The classical view of the genital disc was that the A8-derived primordium differentiated into female genital structures in females and was repressed in males, whereas the A9-derived primordium differentiated into male genital structures in males and was repressed in females; the A10-derived primordium differentiates into anal structures appropriate to the sex of the individual. Thus, whereas the differentiation of the anal primordium requires an instructive cue from the sex hierarchy, the differentiation of the appropriate genital primordium was inferred to require only a permissive function of the sex hierarchy, with segmental identity determining the structures that ultimately developed. This classical view was overturned by the finding that the ‘repressed’ genital primordium in each sex actually develops into adult structures: the ‘repressed’ female (A8) primordium produces a miniature eighth tergite in males and the ‘repressed’ male (A9) primordium produces the parovaria in females (Keisman et al., 2001). Consistent with its instructive role, the sex hierarchy actively modulates the regulation by other developmental pathways of sex-specifically deployed genes. The dachshund (dac) gene is differentially expressed in the male and female genital discs, and the sex hierarchy mediates this sex-specific deployment by determining cell-autonomously whether dac is activated by wingless signaling (in females) or by decapentaplegic signaling (in males) (Keisman and Baker, 2001). Fibroblast growth factor (FGF) signaling in the genital disc is also regulated cell-autonomously by the sex hierarchy (Ahmad and Baker, 2002). DSX^F represses the FGF-encoding branchless (bnl) gene, thus restricting bnl-expressing cells to the male genital disc. FGF signaling from these cells recruits into the disc mesodermal cells expressing the FGF receptor encoded by the breathless (btl) gene. Once inside the male genital disc, these btl-expressing cells become epithelial and eventually give rise to the paragonia and vas deferens, components of the internal male genitalia. An instructive role for the sex hierarchy is also evident in an adult tissue not derived from the genital imaginal disc. The bric à brac (bab) locus integrates signals from the homeotic genes, as well as the sex hierarchy to repress pigmentation of tergites 5 and 6 in females (Kopp et al., 2000).

Although the Yp genes, which are activated by DSX^F and repressed by DSX^M, are the only known direct target of dsx, it is likely that DSX^F acts in some cases to repress transcription and that DSX^M acts in some cases to activate transcription. Indeed, if the examples above represent cases of direct regulation, then it is clear that the effect of DSX^F or DSX^M is dependent upon both the cellular context and the promoter organization of the target gene. Such context-dependent duality of function finds precedent in several well characterized transcription factors. The mechanisms that determine whether a bi-functional transcription factor is in an activating or repressing state are diverse, and include binding of ligand co-factors, differential organization of binding sites in promoters, interaction with other DNA-binding factors, and concentration-dependent structural changes (Roberts and Green, 1995). The DSX proteins provide an especially interesting case of dual regulatory activity because not only are DSX^F and DSX^M each capable of activating some target genes and repressing others, but the two isoforms often have opposite effects, with DSX^F repressing those genes that DSX^M activates and vice versa. It may be that IX, functioning as a co-factor for DSX^F, plays a key role in effecting this symmetry of dual regulatory activities.

The authors thank Mike Simon, Margaret Fuller, Pam Carroll, Lisa Ryner, Axel Franke and members of the Baker laboratory for helpful discussions; Guennet Bohm for the preparation of culture media and fly food; and Pieter Wensink, Marie Lossky, Ken Burtis and Delphine Fagegaltier for reagents. This work was supported by an NIH Developmental and Neonatal Training Grant (C. M. G.-E.), a NSF/Sloan Foundation Postdoctoral Fellowship and NIH NRSA Postdoctoral Fellowship (M. L. S.), the Medical Scientist Training Program (D. S. M.) and by an NIH grant to B. S. B.