An N-Terminal Truncation Uncouples the Sex-Transforming and Dosage Compensation Functions of Sex-lethal

Abstract

In Drosophila melanogaster, Sex-lethal (Sxl) controls autoregulation and sexual differentiation by alternative splicing but regulates dosage compensation by translational repression. To elucidate how Sxl functions in splicing and translational regulation, we have ectopically expressed a full-length Sxl protein (Sx.FL) and a protein lacking the N-terminal 40 amino acids (Sx-N). The Sx.FL protein recapitulates the activity of Sxl gain-of-function mutations, as it is both sex transforming and lethal in males. In contrast, the Sx-N protein unlinks the sex-transforming and male-lethal effects of Sxl. The Sx-N proteins are compromised in splicing functions required for sexual differentiation, displaying only partial autoregulatory activity and almost no sex-transforming activity. On the other hand, the Sx-N protein does retain substantial dosage compensation function and kills males almost as effectively as the Sx.FL protein. In the course of our analysis of the Sx.FL and Sx-N transgenes, we have also uncovered a novel, negative autoregulatory activity, in which Sxl proteins bind to the 3′ untranslated region of Sxl mRNAs and decrease Sxl protein expression. This negative autoregulatory activity may be a homeostasis mechanism.

Sex-lethal (Sxl) encodes an RNA recognition motif (RRM) class RNA binding protein that serves as the developmental switch for sex determination in Drosophila melanogaster (6). Sxl is expressed only in females, where it controls sexual differentiation and dosage compensation by posttranscriptional regulatory mechanisms that affect pre-mRNA splicing and mRNA translation. Misregulation of Sxl results in sex-specific lethality and sex transformations (see reference 15 and references therein).

Female sexual identity is maintained by an autoregulatory feedback loop in which Sxl proteins promote their own synthesis by directing the female-specific splicing of Sxl pre-mRNAs (Fig. 1A; references 5, 6, and 14). Functional female Sxl mRNA is generated by joining exon 2 to exon 4, skipping the third (male-specific) exon, which contains in-frame translation stop codons. Male identity is maintained by the default splicing machinery, which incorporates the third exon into the mature mRNAs, ensuring that no Sxl protein is produced. Sxl-dependent posttranscriptional regulation also controls the gene cascades that direct the different aspects of female or male development (Fig. 1A). Sxl protein promotes female differentiation by directing the female-specific splicing of transformer (tra) pre-mRNAs (25, 45, 46). In the absence of Sxl protein, the default splicing of tra results in mRNAs that do not encode functional protein. Sxl also regulates dosage compensation, which is responsible for equalizing the expression of X-linked genes in the two sexes. One component of the dosage compensation system is the hyperactivation of X-chromosome gene expression in males by the male specific lethal (msl) genes (1, 7, 32, 33). Sxl proteins prevent hyperactivation in females by blocking both the splicing and translation of transcripts from one of these genes, msl-2 (3, 18, 28, 51). A second component of the dosage compensation system is msl independent (8, 19, 20) and is thought to function in females to reduce X-chromosome gene expression. Kelley et al. (28) recently suggested that Sxl itself mediates this dosage compensation by repressing the translation of mRNAs expressed from X-linked genes.

FIG. 1.

Regulatory activities of Sxl. (A) Models for Sxl splicing, tra splicing, and msl-2 translational control in Drosophila males (top) and females (below). The default splicing of both Sxl and tra occurs in males and results in transcripts with in-frame stop codons so that no functional protein can be made. In females, Sxl and tra are alternatively spliced due to the action of Sxl protein. The mRNAs made in females code for functional protein. msl-2 splicing and translation are regulated by Sxl proteins which bind to sites within the 5′ and 3′ UTRs of msl-2 mRNA. Figures are not drawn to scale. Exons are shown by boxes; introns are shown by straight lines; Sxl binding sites are shown by black circles. Arrows indicate the positions of primers used in RT-PCR experiments. (B) Schematic representation of transgenes. hsp83::Sxl.FL-MS3 (Sx.FL) is the full-length Sxl cDNA of the short C-terminal Sxl isoform, ending in exon 8 and with the long form of exon 5. It has an almost full-length 3′ UTR. hsp83::Sxl-N40aa (Sx-N) is a deletion construct in which 40 aa have been removed from the extreme amino terminus of the Sx.FL cDNA. Sx.FLΔ and Sx-NΔ are identical to Sx.FL and Sx-N, respectively, except that the 3′ UTRs have been truncated to remove all but two of the putative Sxl binding sites. All transgenes are expressed under the control of the constitutive hsp83 promoter. The translation start sites have been changed to the Kozak consensus sequence for D. melanogaster. Shown with the schematics are the functional Sxl domains: the ∼120-aa N terminus, the two RRM RNA binding domains (R1 and R2), and the C-terminal domain.

The posttranscriptional regulatory activities of the Sxl gene depend on direct interactions between Sxl proteins and target RNAs. RNA binding activity is provided by Sxl’s two RRM domains, R1 and R2 (26, 37, 43, 49). The two RRM domains recognize poly(U) runs of seven or more nucleotides, and all of the known Sxl regulatory targets have one or more of these arrays. In the case of tra, in vivo and in vitro studies indicate that Sxl protein directs female-specific splicing by binding to a poly(U) run in the polypyrimidine tract of the default 3′ splice site (Fig. 1A) (25, 45, 46). It has been proposed that this prevents the generic splicing factor U2AF from binding to the default polypyrimidine tract and forces the assembly of a U2AF-U2 snRNP splicing complex on the weaker, female-specific 3′ splice site downstream (22, 48).

While a direct competition for overlapping binding sites accounts for what is known about tra splicing, the RNA binding activity of Sxl is not sufficient to explain either Sxl autoregulation or the repression of msl-2 translation. The key targets for Sxl autoregulation are located in the introns upstream and downstream of the male exon at distances of 200 or more nucleotides from the regulated 3′ and 5′ splice sites (24). Hence, instead of a direct blockage mechanism, Sxl must indirectly prevent the assembly of productive splicing complexes at the male exon. One possibility is that homotypic interactions between Sxl proteins sequester the male exon from the splicing machinery (24, 35). Consistent with this possibility, Sxl proteins interact in vitro, and these interactions stabilize Sxl complexes on RNA (26, 36, 43, 49). These Sxl-Sxl interactions are mediated by the two RRM domains (37, 42, 50). A second model postulates that Sxl interacts with and inactivates components of the splicing machinery assembled at the male exon splice sites (e.g., U1 and U2 snRNPs [16, 40]). Consistent with this model, Sxl proteins in vivo are found in large complexes which contain both U1 and U2 snRNPs and Sxl pre-mRNAs (16, 43). In addition, mutations in the sans-fille (snf) gene, encoding the fly homologue of two mammalian snRNP proteins, U1A and U2B", disrupt autoregulation and exacerbate the female-lethal effects of Sxl mutations (38, 40). This synergism may be attributed to interactions between these two proteins; Snf-Sxl complexes can be detected in vivo and in vitro, and this interaction is mediated by the first Sxl RRM domain (16, 43). Finally, efficient translational repression of msl-2 mRNA requires Sxl protein binding sites in both the 5′ and 3′ untranslated regions (UTRs) (4, 18, 29). In a manner analogous to autoregulation, interactions between Sxl proteins upstream and downstream of the open reading frame could sequester the msl-2 mRNA from the translational machinery. Alternatively, Sxl might interact with and poison this machinery.

While the two Sxl RRM domains have been implicated in both RNA binding and protein-protein interactions, much less is known about the functions of the N- and C-terminal domains. Though there are no known mutations in the N-terminal domain, there are some indications that it may be important for the regulatory functions of the Sxl protein. A truncated Sxl protein was detected in the heads of adult D. melanogaster males (11). This smaller isoform appears to result from translation initiation at an AUG codon in exon 4, downstream of the male exon (exon 3), and gives a protein lacking the first 40 amino acids (aa). A slightly larger male-specific protein is also detected in the related drosophilid, D. virilis (12). Although the D. melanogaster and D. virilis male proteins contain both RRM domains and appear to bind appropriate target RNAs, they do not have detectable feminizing activities. It was initially thought that the concentration of the truncated proteins might be too low to induce feminization. This explanation was called into question by Wang and Bell (49), who found that a truncated Sxl protein (SxlN1), similar to one observed in D. melanogaster male heads, was impaired in Sxl autoregulation when transiently expressed in tissue culture cells.

To learn more about the regulatory functions of the different Sxl protein domains, we have compared the biological activities of the full-length Sxl protein (Sx.FL) and Sx-N in vivo. We were able to uncouple the splicing and translational regulatory activities of Sxl protein. Sx-N is impaired in autoregulatory function, and contrary to the expectations of the U2AF blockage model, the N terminus plays an essential role in the regulation of tra splicing. However, these amino acids are not required to regulate msl-dependent dosage compensation. Finally, we provide evidence that Sxl is controlled by both positive and negative autoregulation.

MATERIALS AND METHODS

Fly stocks.

All fly stocks, unless otherwise noted, are referenced by Lindsley and Zimm (31). Flies were grown on standard Drosophila medium and maintained at room temperature (22°C) unless otherwise specified. The Binsinscy (Bin) chromosome was the first chromosome balancer in all crosses.

Plasmid construction and Drosophila transformation.

The Sxl MS3 cDNA described by Samuels et al. (44) was the starting point for construction of the Sxl transgenes. The MS3 cDNA encodes a 38-kDa Sxl protein isoform; in the N-terminal domain it has the eight additional amino acids which arise from the alternative splice in exon 5, while the C terminus is encoded by exons 7 and 8. Consensus translational start sites (13) were engineered into the MS3 cDNA by PCR mutagenesis to generate the Sx.FL and Sx-N transgenes. In the case of Sx.FL, the entire 5′ UTR was deleted and the sequence just upstream of the normal Sxl start site was changed to an XhoI site followed by the consensus initiation sequence (underlined; GTCGACCAACATGTACGGC). In the case of Sx-N, a methionine codon at the very 3′ end of exon 4 was used as the translation start site. Again this ATG was preceded by a consensus initiation sequence (GTCGACCAACATGTCACGT) and an XhoI restriction site. Sx.FLΔ and Sx-NΔ were generated by removing sequences 3′ of the HindIII site in the Sxl MS3 3′ UTR. All constructs were then cloned into the hsp83 mini-white vector (24) as Xho/Not fragments. Germ line transformations (47) were performed by injecting plasmid DNA and helper vector pTurbo into white¹ (w¹) embryos.

RT-PCR analysis and Southern blotting.

RNA was prepared as described by Bell et al. (5). Reverse transcription (RT) was performed according to the procedure of Frohman et al. (17). To make Sxl cDNAs, primer MES21 was used in the RT reaction. PCR was then performed with primers T41-3 and BellA1 spanning the sex-specific Sxl splice junctions (primers described in reference 5). The transgene mRNA (or cDNA) is not amplified since the BellA1 primer is upstream of the transgene breakpoint (Fig. 1). In the PCR, 1.5% of the cDNA was amplified for 30 cycles of 95°C for 1 min, 65°C for 45 s, and 72°C for 30 s. One percent of the reaction volume was diluted, loaded on 2% agarose gels, and Southern blotted onto nitrocellulose. Blots were hybridized with randomly primed Sxl MS3 cDNA (44).

To make tra cDNAs, RT was performed with the tra-3′ primer (GATCTGGAGCGAGTGCGTCTG). Approximately 2% of tra cDNA was amplified with tra-5′ and tra-2 primers (GGTCACACTGAGGAAAGTGC and CTTCTCACCCGATCCTGTTCTC). PCR conditions were 1 cycle of 95°C for 5 min, 50°C for 2 min, and 72°C for 10 min; 15 cycles of 95°C for 1.5 min, 52°C for 1 min, and 72°C for 1 min; and 17 or 20 cycles of 95°C for 1.5 min, 55°C for 1 min, and 72°C for 1 min. Ten percent of the reaction volume was separated on 2% agarose gels and Southern blotted onto nitrocellulose. Blots were hybridized with randomly primed DNA from a DraI/BamHI subclone of the 5′ end of tra cDNA.

Immunoprecipitations.

Mouse anti-Sxl and mouse anti-β-galactosidase (β-Gal) antibodies were cross-linked to protein A-Sepharose beads as described in reference 23. Beads were stored at 4°C in immunoprecipitation buffer. Total embryonic extracts were made according to the protocol of Bopp et al. (11). One milliliter of extract was added to 100 μl of antibody-coupled beads and shaken overnight at 4°C. Beads were washed five times in immunoprecipitation buffer with protease inhibitors (aprotinin, phenylmethylsulfonyl fluoride, and pepstatin). RNA was prepared from beads by extraction with 200 μl of 1:1 phenol (pH 4.0)-chloroform and then ethanol precipitated with tRNA carrier (1.5 μg/ml). RNA pellets were suspended in 85 μl of double-distilled water with 10 μl of 10× DNase buffer and 5 μl of DNase. Following a 2-h incubation at 37°C and 15 min at 65°C, the RNA was reextracted and ethanol precipitated. Pellets were resuspended in 20 μl of double-distilled water; 2 μl each of RNA and control white female RNA (also DNase treated) was used for RT.

For Sxl RT-PCR, primers were 3BFill3 (GATTCAGTCCTTGGTTG) for RT and 8BCK (GTTATTGTGCTGTATCCG) and 8FOR (CTTGAGAGTGTTTACATCTG) for PCR. Conditions for PCR with 2% of cDNAs were 35 cycles of 95°C for 1 min, 52°C for 1 min, and 72°C for 1.5 min; a 5% aliquot of the first amplification was then amplified for 1 cycle of 95°C for 1 min, 46°C for 1.5 min, and 72°C for 4 min, followed by 20 cycles of 95°C for 1 min, 52°C for 1 min, and 72°C for 1.5 min.

For msl-2 and mle 3′ UTR amplification, RT was performed with a poly(A) tail-specific primer, ADPTT (GCGAGCTCCGCGGCCGCGTTTTTTTTTTTT). msl-2 3′ UTR primers were M2-1 (CGGGTATCTGATAGTCGGG) and M2-2 (CTTAATCCTAGGTTCACGTGCTC). mle 3′ UTR primers were mle1 (ACACAGGTTCACCAAAAGCCT) and mle2 (GAGGATCAGGCGGCGGCTTTAG). cDNAs were amplified as described above for the Sxl 3′ UTR.

One-tenth of the reaction products were run on 2% agarose gels and visualized by ethidium bromide staining.

Western analysis.

Five flies of each genotype were collected and frozen on dry ice; 50 μl of 2× Laemmli sample buffer was added to the flies, which were then homogenized with a hand-held Dounce homogenizer. Samples were boiled for 5 min and spun for 3 min at 14,000 rpm, and then 10 μl of sample (equivalent to one fly) was loaded on sodium dodecyl sulfate (SDS)–12% acrylamide gels. Western blotting was performed as described elsewhere (11). Blots were prehybridized in PBST–5% nonfat dry milk and probed with mouse anti-Sxl antibody M114 (1:10) overnight at 4°C. Blots were washed three times for 10 min each in PBST and hybridized with horseradish peroxidase-conjugated anti-mouse antibody (1:5,000) in PBST–5% milk for 2 h at room temperature. Blots were again washed three times for 10 min each in PBST and visualized with enhanced chemiluminescence reagent. For loading controls, blots were then reprehybridized and probed with mouse anti-Snf (1:10) or rat anti-GAGA (1:2,000) antibody and then processed as described above with anti-mouse and anti-rat horseradish peroxidase-conjugated antibodies as secondary probes.

RESULTS

Sxl transgenes.

To address the function of the amino terminus of the Sxl protein, we used the Sxl MS3 cDNA (44) to generate two transgenes, hsp83::SxlFL-MS3 (Sx.FL) and hsp83::Sxl-N40aa (Sx-N) (Fig. 1B). Sx.FL encodes a full-length Sxl isoform (see Materials and Methods), while Sx-N encodes a 40-aa acid truncation which corresponds to the species believed to present in the heads of D. melanogaster males. To try to ensure equal translation, we deleted the 5′ UTRs from the Sx.FL and Sx-N cDNAs and changed the translation start regions to match the Drosophila Kozak consensus sequence (13). Both cDNAs retain the long 1.6-kb 3′ UTR of the original MS3 cDNA. To drive expression in transgenic animals, the two cDNAs were placed under the control of the constitutive hsp83 promoter.

Full-length Sxl transgenes.

We analyzed the biological functions of the Sx.FL transgene and compared its activity to that of the previously analyzed hsp70::SxlcF1 transgene (5). The cF1 transgene differs in two important respects from our transgenes. First, the Sxl cDNA is expressed under the control of the inducible hsp70 promoter. Second, while the cF1 cDNA encodes the same Sxl protein isoform as Sx.FL, it has a nearly full length 5′ UTR (with the normal Sxl translation initiation sequence) and a much shorter (220-bp) 3′ UTR. The cF1 lines were classified into three groups based on the phenotypic effects of the transgene in wild-type (Sxl⁺) males. In the first group, males with either one or two copies of the transgene had no obvious phenotype at either 25 or 29°C. In the second, the viability of males with a single copy of the transgene at 25°C was reduced by about one half and the survivors were often, but not always, intersexual. At 29°C, viability of the transgenic males dropped to less than one-fourth; however, the extent of sex transformation remained about the same. This is due to the linkage of sexual differentiation and dosage compensation: an increase in Sxl activity which would make the sex transformation of these males more complete would also enhance male lethality by further disrupting the dosage compensation system. In the third group, viability of males with a single copy of the transgene was reduced to only a few percent, even at 25°C. Our three Sx.FL transgenic lines resemble the strongest cF1 lines in that they are completely male lethal in single copy even at 18°C.

Several experiments show that the lethal effects of Sx.FL are due to the ability of the transgene to both initiate autoregulation and repress dosage compensation. To test the contribution of autoregulation, we examined whether removing the endogenous Sxl locus mitigates the effects of Sx.FL. It does. In a Sxl deficiency background (Sxl^7B0 [39]), Sx.FL males are ∼15% (143 of 923) as viable as their siblings without the transgene (Fig. 2A). Since the transgene associated lethality is less in the Sxl⁻ background than in the Sxl⁺ background, the transgene must be able to transactivate the endogenous gene.

FIG. 2.

Viability of males carrying Sxl transgenes. (A) Viability (x axis) of Sx.FL line 93 males in different genetic backgrounds (y axis). Sxl⁺ refers to the viability (eclosion) in a wild-type Sxl background, 7B0 refers to the Sxl^f7B0, deficiency background, M2 refers to the males rescued by the constitutive msl-2 transgene, H83M2 (28), and 7B0; M2 refers to males with both the Sxl deficiency and the msl-2 constitutive transgene. The number after each box is the number of transgene males compared to the number of sibling males or females of each genotype. Crosses and viability calculations are as follows. For Sxl⁺, the cross was w/w; P[w⁺, Sx.FL]/+ ♀ × w/Y ♂. Viability was calculated as the number of w; P[w⁺, Sx.FL] males compared to w; +/+ sibling brothers. The same value was attained if transgene⁺ males were compared to their transgene⁺ sisters. For 7B0, the cross was y w Sxl^7B0 ct/ Bin; P[w⁺, Sx.FL]/+ ♀ × y w Sxl^7B0/Y ♂. Viability was calculated as the number of y w Sxl^7B0 ct/Y; P[w⁺, Sx.FL]/+ males compared to y w Sxl^7B0 ct/Y siblings. For M2, the cross was w; P[w⁺, Sx.FL]/TM3 Sb Ser ♀ × y w/Y; P[w⁺, H83M2]/TM3 Ser ♂. Here viability is the number of w; P[w⁺, H83M2]/P[w⁺, Sx.FL] males compared to w; P[w⁺, H83M2]/TM3 Sb Ser brothers. The viability of +/TM3 males is equal to that of males from our wild-type w¹ stock. For 7B0; M2, the cross was y w Sxl^7B0/Bin; P[w⁺, Sx.FL]/TM3 Ser ♀ × w/Y; P[w⁺, H83M2]/TM3 Ser ♂. The number of y w Sxl^7B0; P[w⁺, Sx.FL]/P[w⁺, H83M2] males was compared to the number of sibling males of the genotype y w Sxl^7B0; P[w⁺, H83M2]/TM3 Ser. For all crosses, P values were calculated and found to be <0.05. (B) Viability of Sx-N transgenic males shown as heterozygotes (grey bars) and homozygotes (black bars). Viability is expressed as the percentage of viable transgene males relative to viable transgene sibling females. Line 152 is homozygous lethal. Viability of males homozygous for lines 23 and 11 was also tested in the Sxl null background, y w Sxl^7B0 (hatched bar). The cross for viability of males versus females with a single copy of the transgene was w; P[w⁺, Sx-N]/BAL ♀ × w; P[w⁺, Sx-N]/BAL ♂. The balancer chromosome was either CyO or TM3 Ser. For viability of males relative to females homozygous for the Sx-N transgene, the cross was w; P[w⁺, Sx-N]/P[w⁺, Sx-N] ♀ × w/Y; P[w⁺, Sx-N]/ P[w⁺, Sx-N] ♂. For viability of transgene males versus females in the Sxl null background, the cross was y w Sxl^7B0 ct/Bin; P[w⁺, Sx-N]/ P[w⁺, Sx-N] ♀ × y w Sxl^7B0/Y; P[w⁺, Sx-N]/P[w⁺, Sx-N] ♂. For all crosses, P < 0.05. (C) Viability of Sx.FLΔ and Sx-NΔ males was determined for the same genetic backgrounds as described in for panel A. Numbers of animals for each relevant genotype are shown next to the bars. For all crosses, P < 0.05.

The lethal effects of the Sx.FL transgene in the absence of a wild-type Sxl gene argue that the transgene protein itself is able to repress dosage compensation. We tested this hypothesis by providing Sx.FL males with constitutive msl-2 function from the H83M2 transgene (28). As shown in Fig. 2A, the constitutive msl-2 transgene relieves the Sx.FL-associated male lethality in both Sxl⁺ and Sxl^7B0 backgrounds. In the Sxl⁺; msl-2^c background, the relief is only partial. In contrast, in the Sxl^7B0; msl-2^c background, the relief is complete: sibling males with and without the Sx.FL transgene are equally viable. The suppression of lethality by the H83M2 transgene suggests that the Sx.FL transgene kills males largely by inactivating msl-2 and that it does not have sufficient activity to kill males through the msl-independent dosage compensation system.

Amino-terminal deletion of Sxl. (i) The Sx-N transgene has substantially reduced activity.

Deletion of 40 aa from the amino terminus significantly impairs the activity of the Sxl protein. Whereas a single copy of the Sx.FL transgene is fully male lethal, no effects on male viability or morphology were evident in eight of the nine Sx-N lines isolated (Fig. 2B). In the exceptional line, 152, viability was reduced about one-third. Since Western blots indicate that the truncated proteins are expressed (data not shown), the apparent lack of activity of a single copy of the Sx-N transgene indicates that the N-terminal 40 aa are important for the regulatory functions of Sxl protein.

The truncated protein is not, however, completely inert. This can be seen by increasing the dose of Sx-N transgene. In all but one line (line 111), males homozygous for the Sx-N transgene were only 5 to 50% as viable as their homozygous Sx-N sisters (cf. 23 and 11 in Fig. 2B). The reduction in male viability for particular Sx-N inserts is not due to inactivation of vital genes by insertion of the transgene since similar effects are observed in males transheterozygous for different Sx-N transgene inserts (data not shown). Although two copies of the Sx-N transgene reduce male viability, the surviving males are phenotypically normal.

(ii) Truncated Sxl protein is capable of initiating autoregulation.

As was the case for the full-length protein, the male-lethal effects of the Sx-N protein could be indirect, arising from activation of the endogenous Sxl gene, or could be direct, through repression of dosage compensation. Consistent with the former possibility, viability of Sx-N males carrying two copies of the transgene improves when the Sxl⁺ gene is removed. In the Sxl⁺ background, only 5% of the expected Sx-N line 11 males eclose, whereas in the Sxl^7B0 background, 50% of the expected male progeny eclose. The increase in viability observed with removal of the endogenous Sxl gene indicates that the Sx-N protein must retain some autoregulatory activity. The remaining lethality (40 to 50%) suggests that the Sx-N transgenes also have dosage compensation activity.

To confirm that the Sx-N transgenes were able to initiate the autoregulatory feedback loop, we analyzed the splicing of endogenous Sxl⁺ transcripts in an RT-PCR assay (5). As shown in Fig. 3, splicing of Sxl transcripts in wild-type flies is sex specific: only the exon 2:4 splice is observed in females, while only the exon 2:3:4 splice is detected in males (Fig. 1A). If the Sx-N transgene activates the Sxl autoregulatory feedback loop, we should be able to detect some female-specific splice products in Sxl⁺; transgene males. As can be seen in Fig. 3, female-specific Sxl transcripts are present in males from the five Sx-N transgenic lines examined. Consistent with the presence of female spliced mRNAs, Western blots indicate that the two major Sxl protein isoforms are expressed in these Sx-N males (data not shown). Significant levels of male-specific Sxl transcripts are also present in all but line 152 (which shows lethal effects as a heterozygote). The presence of both female and male splice forms in Sx-N transgenic males implies that the autoregulatory feedback loop either is not activated in every cell or is not activated fully in all cells. Additionally, since these males escaped the lethal effects of the transgene, the autoregulatory feedback loop may not have been turned on in these animals until relatively late in development.

FIG. 3.

Autoregulation is activated in Sx-N males. Sxl RT-PCR was performed with primers flanking the Sxl sex-specific splice sites (the RT primer resides in exon 7; PCR primers are from exons 2 and 7 [Fig. 1A]). Results were visualized by probing Southern blots with radioactively labeled Sxl cDNA. Six independent Sx-N homozygous lines (indicated by line number above each lane) were assayed for the profile of Sxl splice forms. w¹ females and males served as controls. Flies were raised at 29°C to optimize expression of the transgene.

To test whether timing of Sx-N protein expression accounts for the range of phenotypes of the different Sx-N transgene lines, we stained embryos with Sxl antibodies. We found that lines with the highest lethality had Sxl staining in most embryos. For example, Sx-N line 11 is 95% male lethal when present in two copies. Virtually all of the early embryos from this line display uniform Sxl staining. There is a corresponding decrease in the percentage of embryos staining with the Msl-2 antibody. Conversely, in line 111, which has little or no transgene-associated lethality, only patchy Sxl staining is observed in some late (presumptive) male embryos, while most of the (presumptive) male embryos show a wild-type Msl-2 staining pattern (data not shown). In these studies, we also observed that the subcellular distribution of the Sxl proteins in embryos carrying the Sx.FL and Sx-N genes was the same and that the staining pattern was identical to that seen in wild-type females (data not shown). Thus, the differences in the activity of the two proteins are probably not due to differences in protein localization.

(iii) Truncated Sxl protein is impaired for tra splicing.

While the results described above indicate that the Sx-N protein retains at least some autoregulatory and dosage compensation activities, the most striking finding is that the protein seems to lack the ability to regulate the tra sexual differentiation pathway. Shown in Fig. 4A is a comparison of the feminization activities of the Sx.FL and Sx-N transgenes in males which do not contain the endogenous Sxl locus. This background allows us to unambiguously ascribe alterations in somatic sexual differentiation to the Sxl proteins expressed by the transgenes. The extent of sex transformation that we see in escaper Sx.FL males again appears to resemble that for the strong cF1 lines. The males are intersexual; they have lighter abdominal pigmentation, rotated genitalia, and fewer sex combs, and they are sterile (Fig. 4A). In sharp contrast, Sxl⁻; Sx-N males are morphologically indistinguishable from wild-type males and are fertile (compare flies in Fig. 4A).

FIG. 4.

Sex-transforming function of Sxl transgenes. (A) Phenotype of transgenic males in the absence of the endogenous Sxl gene shows that males carrying the Sx.FL cDNA display sex transformations (middle). These males have lighter pigmentation, rotated and often intersexual genitalia, and fewer sex combs than their sibling brothers without the Sxl transgene (left). In contrast, males homozygous for the amino-terminal Sxl truncation, Sx-N, and the Sxl null allele (right) have no intersexual features and are morphologically identical to their sibling brothers without the transgene. Sex combs of males of each genotype are also shown. The precise genotypes of these males are y w Sxl^7B0 (left), y w Sxl^7B0; P[w⁺, Sx.FL line 93] (middle), and y w Sxl^7B0; P[w⁺, Sx-N line 11]/P[w⁺, Sx-N line 11] (right). (B) Sex-transforming function of Sxl transgenes in females homozygous for the Sxl loss-of-function allele, Sxl^fm3. From left to right: w¹ female; y w Sxl^fm3/w Sxl^fm3; P[w⁺; Sx.FLΔ line 2] female; y w Sxl^fm3/ w Sxl^fm3; P[w⁺; Sx.FL line 93] female; y w Sxl^fm3/w Sxl^fm3; P[w⁺; Sx-NΔ line 28] female; y w Sxl^fm3/w Sxl^fm3 female; w¹ male.

The absence of sex transformations in Sx-N males most likely reflects an inability to regulate the tra pathway. To determine if this is the case, we used an RT-PCR assay to examine the splicing pattern of tra mRNAs. Three amplification products can be detected in this assay, and these correspond to unspliced RNA, default (male) spliced RNA, and female spliced RNA (Fig. 5). Of these, only unspliced and default spliced tra RNAs are observed in wild-type males, while all three species are found in females. In wild-type females, the yield of the female spliced amplification product at either 32 or 35 cycles (Fig. 5) is greater than the yield of products corresponding to either the unspliced or the default spliced tra mRNA. In Sxl^7B0; Sx.FL-52 males, the major amplification product is also the female spliced RNA. However, the yield of the female product relative to the other tra RNAs is not as high as in wild-type females. Thus, while the Sx.FL transgene has tra regulatory activity, it is not as effective as the endogenous Sxl gene.

FIG. 5.

tra splicing is impaired in Sx-N males. RT-PCR was performed with tra-specific primers located in exon 1 and downstream of the female-specific splice in exon 2 (Fig. 1A). Aliquots were removed from PCRs after 32 or 35 cycles, run on 2% Tris-borate-EDTA gels, Southern blotted, and probed with radioactively labeled tra cDNA (see Materials and Methods for specific primers, PCR conditions, and probe). Analyses were performed in w¹ females (F), w¹ males (M), and transgenic males of the genotypes y w Sxl^7B0; P[w⁺, Sx.FL line 52] (lanes 1) and y w Sxl^7B0; P[w⁺, Sx-N line 11]/P[w⁺, Sx-N line 11] (lanes 2). The PCR identifies three products: the female-specific and sex-nonspecific splice forms and the unspliced RNA.

A different profile of PCR products is observed in Sx-N line 11 males. At 32 cycles, the profile resembles that of wild-type males; only the unspliced and the default splice products are detected. After 35 cycles, low levels of the female splice product appear; however, their yield is much less than that of either the default or unspliced products. Very similar results are obtained for two other Sx-N transgene lines (data not shown). These findings confirm the suggestion that the Sx-N protein is defective in its tra regulatory function. In this context, it should be noted that our PCR assay most likely underestimates the extent of reduction in tra regulatory activity. We find that the female splice product is preferentially amplified: as the number of amplification cycles increases, the amount of female product relative to the default and unspliced products also increases (Fig. 5; compare wild-type female products after 32 and 35 cycles).

The Sxl 3′ UTR is a cis regulatory region. (i) Sxl protein binds to the 3′ UTRs of msl-2 and Sxl mRNAs in vivo.

Sxl is thought to block the expression of Msl-2 protein by binding to multiple target sites in the 5′ and 3′ UTRs of the msl-2 mRNA (4, 29). While full repression requires both UTRs, removal of either the 5′ or the 3′ UTR alone does not lead to complete derepression. These observations (as well as the msl-independent dosage compensation function of Sxl proposed by Kelley et al. [28]) led us to question whether our Sx.FL and Sx-N transgenes were providing an accurate assessment of the regulatory potential of the full-length and truncated proteins. The long, ∼1.6-kb 3′ UTRs of the Sx.FL and Sx-N cDNAs contain 10 sequences which closely match the consensus Sxl binding sites, as well as several additional lower-affinity sites (26, 36, 43, 49). If Sxl proteins bind these sites in the transgene mRNAs, they might repress translation by a mechanism analogous to that used in msl-2 regulation.

To explore this possibility, we examined whether Sxl proteins are associated with the 3′ UTRs of Sxl and msl-2 RNAs in vivo. RNA was isolated by immunoprecipitating wild-type embryo extracts with either Sxl or β-Gal antibodies and was RT-PCR amplified by using primers specific to the 3′ UTRs of Sxl RNA, msl-2 RNA, and, as a negative control, mle RNA. As shown in Fig. 6B and C, 3′ UTR sequences from both Sxl and msl-2 mRNAs are present in Sxl immunoprecipitates but are not found in β-gal immunoprecipitates. Confirming the specificity of Sxl immunoprecipitation, mle RNA can be readily detected in bulk RNA but is not present in the Sxl (or β-Gal) immunoprecipitate (Fig. 6A). In previous studies (16), we have found that most RNAs are partially hydrolyzed during the immunoprecipitation procedure even in the presence of RNase inhibitors. As a consequence, RNA sequences distant from the (presumptive) protein binding sequences are usually absent or present in much lower yields. As would be expected from the fact that Sxl protein binding sites are located in the 3′ UTR of Sxl mRNAs but not in the translated sequences, we were unable to detect amplification products in Sxl immunoprecipitates by using primers for exon sequences encoding the C terminus of the Sxl protein (data not shown).

FIG. 6.

Sxl protein binds to its own 3′ UTR. Sxl and β-Gal immunoprecipitates were analyzed by RT-PCR, run on 2% agarose gels, and visualized by ethidium bromide staining. RT of immunoprecipitates was performed with primers specific to the poly(A) tail or to the Sxl 3′ UTR. PCR was carried out with primers specific to the 3′ UTRs of mle (A), Sxl (B), and msl-2 (C). Shown are three gels with identical lanes: 1, total female RNA (control for RT-PCR), 2, anti-Sxl immunoprecipitate; lane 3, anti-β-Gal immunoprecipitate; 4, no-RT control; and 5, DNA control. Primers and conditions are described in Materials and Methods.

(ii) Deleting sequences from the 3′ UTR increases expression of the transgene protein.

The in vivo association of Sxl proteins with the 3′ UTRs of Sxl mRNAs is consistent with the possibility that they negatively regulate their own expression. This negative autoregulation model makes two predictions. First, the activities of the Sx.FL and Sx-N transgenes should be increased by removing target sites for the Sxl protein. Second, an increase in biological activity should be accompanied by a corresponding increase in expression of transgene proteins. To test these predictions, we generated two new transgenes, Sx.FLΔ and Sx-NΔ, by deleting ∼1 kb from the 3′ UTRs of Sx.FL and Sx-N, respectively. These deletions eliminate 8 of the 10 consensus Sxl binding sites in the 3′ UTR (Fig. 1B).

Consistent with the first prediction, the regulatory activities of the Sx.FLΔ and Sx-NΔ transgenes are substantially enhanced. This is most obvious in the case of the Sx-NΔ transgene. Whereas a single copy of Sx-N had no male-lethal effects, males are killed by a single copy of the Sx-NΔ transgene. Two of the Sx-NΔ lines (21 and 28) are fully male lethal (Fig. 2C), while in the third line, 26, less than 10% of the males survive (data not shown). The activity of the Sx-FLΔ transgene is also greater than that of the Sx.FL transgene; however, since all of the original Sx.FL lines (as heterozygotes) are completely male lethal in a Sxl⁺ background, it is possible to detect the increased activity only in special genetic backgrounds (see below).

Consistent with the second prediction, deletion of the 3′ UTR increases expression of the transgene proteins. Again this is most easily illustrated for the Sx-NΔ transgene. Figure 7 compares the amount of Sx-N protein in Sxl⁻ males from the Sx-NΔ lines 21 and 28 with that in two Sx-N lines, 152 and 11. Of the nine Sx-N lines, the highest level of Sx-N protein is found in line 152, while the other lines resemble line 11 (data not shown). The amount of Sx-N protein in the two Sx-NΔ lines is about twofold higher than that of Sx-N line 152 and nearly threefold higher than that of line 11. Even more pronounced differences in Sx-N expression (four to fivefold) between the Sx-NΔ and Sx-N lines are observed in females. Western blots of Sx-FLΔ and Sx-FL females suggest that there is also a significant increase in protein expression when the 3′ UTR is deleted (see Fig. 9). However, exact differences in protein expression cannot be quantified since the Sx.FL protein comigrates with one of the endogenous proteins.

FIG. 7.

Expression of Sxl transgenes is enhanced by removing the 3′ UTR. Western blot analysis was performed on Sx-NΔ and Sx-N males in the Sxl deficiency background. Blots were probed with anti-Sxl antibody and anti-Snf antibody. Genotypes for flies in each lane are as follows: F, w¹ females; M, w¹ males; Δ28, y w Sxl^7B0; P[w⁺, Sx-NΔ line 28]; P[w⁺, H83M2]; Δ21, y w Sxl^7B0; P[w⁺, Sx-NΔ line 21]; 152, y w Sxl^7B0; P[w⁺, Sx-N line 152]; 11, y w Sxl^7B0; P[w⁺, Sx-N line 11]. Sx-NΔ line 28 males are viable only when the H83M2 transgene is also present. The H83M2 transgene does not alter the expression of the Sxl transgenes (data not shown). The positions of the female-specific Sxl doublet and of the Sx-N transgene are indicated. Below each lane for Sx-NΔ males is the ratio of the Sx-N transgene to the Snf loading control. Quantitation was performed with NIH Image software.

FIG. 9.

Western blot analysis of Sx.FL protein expression in whole flies (A) or in the soma and the germ line (B) indicates that the Sxl 3′ UTR acts as a cis-regulatory sequence. Extracts were run on SDS–12% gels, transferred to nitrocellulose, and probed with anti-Sxl antibody and anti-Snf antibody. Lanes: F and w¹, w¹ females, Δ2, w; P[w⁺; Sx.FLΔ line 2] females; Δ10, w; P[w⁺; Sx.FLΔ line 10] females; Δ16, w; P[w⁺; Sx.FLΔ line 16] females; 93, w; P[w⁺; Sx.FL line 93] females; 141, w; P[w⁺; Sx.FL line 141] females. As shown in lane F, Sxl normally runs as a doublet of 38 and 36 kDa. The Sx.FL protein comigrates with the upper, 38-kDa band. In carcasses and ovaries, the Sxl signal from the 36-kDa lower band of the Sxl doublet was quantitated by using NIH Image software. Samples were normalized for the amount of Snf protein expressed in each lane, and the relative amount of the 36-kDa isoform in Sx.FLΔ transgene females was compared to that wild-type females. In carcasses, the 36-kDa isoform ranged between 40 and 65% of wild-type levels, depending on the line. The ratio of the Sxl to Snf was 0.60 in wild type, 0.25 in Sx.FLΔ line 2, 0.27 in Sx.FLΔ line 10, and 0.39 in Sx.FLΔ line 16. These reductions were consistent over multiple experiments. In ovaries, there was a more dramatic reduction in the 36-kDa isoform representing less than 15 to 33% of wild-type levels, again depending on the line examined. The ratios of Sxl to Snf were 1.7 for wild type, 0.26 for Sx.FLΔ line 2, and 0.55 for Sx.FLΔ line 16. Due to the very high levels of the Sx.FLΔ line 10 protein, it was not possible to visualize the 36-kDa isoform as a distinct band, and therefore the extent reduction in this isoform could not be calculated. The image presented has been lightened so that the 38-kDa transgene protein could be visualized as a distinct band.

Autoregulatory and dosage compensation activities of the Sx.FLΔ and Sx-NΔ transgenes.

To more precisely compare the autoregulatory and dosage compensation activities of Sx.FLΔ and Sx-NΔ with those of their Sx.FL and Sx-N counterparts, we introduced the UTR deletion transgenes into different genetic backgrounds. As shown for two of the Sx.FLΔ lines in Fig. 2C, removal of the Sxl⁺ gene (by the 7B0 mutation) provides no relief from the lethal effects of the transgene. Similar results were obtained for other lines (not shown). By contrast, Sx.FL requires the endogenous Sxl locus for full male lethality. Sx.FLΔ also has more dosage compensation activity than Sx.FL, as shown by the finding that Sx.FLΔ males are not rescued by the H83M2 transgene. Moreover, even when the H83M2 transgene was combined with the Sxl^7B0 mutation, the viability of the Sx.FLΔ transgenic males was increased only to around 20% (Fig. 2C). These results suggest that the Sx.FLΔ transgene may kill males by regulating both the msl-dependent and msl-independent dosage compensation systems.

Like Sx.FLΔ, the male-lethal effects of Sx-NΔ do not depend on a functional Sxl⁺ gene. As shown in Fig. 2C, the Sxl^7B0 mutation did not rescue Sx-NΔ line 28 males and resulted in only a very small (0.1%) increase in viability of Sx-NΔ line 21 males. In fact, by this measure the dosage compensation activity of the Sx-NΔ transgene is greater than that of the Sx.FL transgene (Fig. 2A). The rescuing activity of the H83M2 transgenes shows that the lethal effects of the Sx-NΔ transgene are largely due to a down-regulation of msl-dependent dosage compensation. Unlike the case for Sx-FLΔ, viability of Sx-NΔ males is substantially increased by the H83M2 transgene. In the case of Sx-NΔ line 28, male viability increases to about 50%, while in Sx-NΔ line 21 it is nearly 40%. The fact that Sx-NΔ males can be rescued by the H83M2 transgene would also suggest that Sx-NΔ is less effective than Sx-FLΔ in activating the splicing autoregulatory feedback loop. On the other hand, Sx-NΔ does have significant positive autoregulatory activity since Sx-NΔ males are fully viable when the H83M2 transgene is combined with the Sxl^7B0 mutation but are only partially viable in either background alone.

The Sx-NΔ transgene lacks feminization activity.

While the male-lethal effects of Sx-N are enhanced substantially by removal of the 3′ UTR, we were surprised to discover that there is no evidence of a corresponding increase in tra activity. Sxl⁻; Sx-NΔ males which are rescued by expression of the msl-2 constitutive transgene are morphologically indistinguishable from wild-type males and are fertile (not shown). In contrast, the few Sx.FLΔ males that manage to survive in the Sxl⁻; H83M2 background have a typical intersexual phenotype (defective or missing sex combs, female abdominal pigmentation, and rotated or feminized genitalia [14]) and are sterile (not shown). These results again indicate that the tra regulatory activity of the truncated Sx-N protein is severely impaired.

To confirm this conclusion, we examined the tra regulatory activity of the different Sx.FL and Sx-N transgenes in females homozygous for two Sxl partial loss-of-function mutations, Sxl^f3,M1 (fm3) and Sxl^M1,f7 (fm7). fm3 and fm7 are second-site mutations in the constitutive Sxl allele, Sxl^M1 (9). Females homozygous for either of these alleles survive at a low frequency, presumably because the mutant Sxl gene has some residual ability to regulate the dosage compensation system. On the other hand, the mutations have all but eliminated tra regulatory activity: the surviving females are completely sex transformed and are virtually indistinguishable from wild-type males even with respect to their smaller body size. Hence, we should be able to assay the tra regulatory activities of the different transgenes in fm3 or fm7 mutant females without the potential complications that arise in males from down-regulation of X-linked gene expression.

As shown in Fig. 8, the transgenes rescue the viability of fm7 homozygous females to a much greater extent than fm3. For both fm3 (Fig. 4B) and fm7 (not shown), the females rescued by the Sx.FL transgene (line 93 in Fig. 4B) had female-like morphology and light body pigmentation; however, feminization was usually incomplete, and male structures such as sex comb teeth were often observed. In contrast, feminization of both fm3 and fm7 mutant females by Sx.FLΔ is nearly complete. As illustrated in Fig. 4B for Sx.FLΔ line 16, Sx.FLΔ transgene fm3 females are essentially indistinguishable from wild-type females with respect to external morphology and pigmentation. A quite different result is obtained for the Sx-NΔ transgenes. Although the Sx-NΔ transgenes increase the viability of the fm3 and fm7 females as well as if not better than the Sx.FLΔ transgenes (Fig. 8), all of the surviving females remain completely masculinized. As shown for Sx-NΔ line 28 in Fig. 4B, the rescued fm3 females resemble wild-type males. This result provides an additional demonstration that the Sx-N protein is defective in regulating tra pre-mRNA splicing in vivo.

FIG. 8.

Activity of Sxl transgenes in females. Sxl transgenes (tg) rescue females from loss-of function Sxl alleles. Shown is the percent viability (y axis) of Sxl homozygous females carrying Sx.FLΔ, Sx.FL and Sx-NΔ transgenes (x axis). Percent viability (as measured by eclosion) reflects the ratio of the number of Sxl⁻/Sxl⁻; transgene females to the number of Sxl⁻/Bin; transgene sibling sisters for each genetic background. Crosses were performed as follows: 2593, w Sxl^f2593/Bin; P[w⁺; Sxl] ♀ × w Sxl^f2593/Y ♂; fm3, y w Sxl^fm3/Bin; P[w⁺; Sxl] ♀ × w Sxl^fm3/Y ♂; fm7, w Sxl^fm7/Bin; P[w⁺; Sxl] ♀ × w Sxl^fm7/B^sY ♂. In the fm3 and fm7 crosses, males and females could be distinguished from each other because of the yellow and Bar stone markers, respectively. In the f2593 crosses, however, males could not be distinguished from females based on secondary markers. Consequently, we performed preliminary w Sxl^f2593/Bin; P[w⁺; Sxl] ♀ × y Sxl^f2593/Y ♂ crosses to confirm that no Sxl^f2593; P[w⁺; Sxl] males survived. For all crosses, P < 0.05.

Sxl transgenes alter expression of the endogenous Sxl gene.

One unexpected finding was that the Sx.FLΔ transgene is, if anything, less able to rescue females from the lethal effects of the fm3 and fm7 mutations than is the original Sx.FL transgene (Fig. 8). Since the phenotypic effects of the Sx.FLΔ transgene are, by all other measures, much greater than those of Sx.FL, it should also have been much more effective in improving the viability of these Sxl mutant females. Interestingly, the same result was obtained for the Sxl^f2593 allele (14), a hypomorphic, temperature-sensitive Sxl mutation. Under restrictive growth conditions, we find that the Sx.FLΔ transgene does not rescue female viability as well as the Sx-NΔ transgene (Fig. 8), although the few Sx-FLΔ survivors are completely feminized and are fertile (data not shown). A more complete rescue of viability is observed for the Sx.FL transgene, but the survivors are intersexual.

One possible explanation for the lack of correlation between the activities of the Sx.FLΔ transgenes and their ability to rescue hypomorphic Sxl mutations is that the high levels of the Sx.FL protein produced by these transgenes negatively regulate the endogenous Sxl gene. A prediction of the negative autoregulation model is that an increase in the expression of the transgene protein (by deleting the 3′ UTR) should be accompanied by a concomitant decrease in the expression of Sxl proteins from the endogenous gene. To test this, we examined the levels of Sxl protein in Sx.FL and Sx.FLΔ transgene females by Western blotting. The Sx.FL and Sx.FLΔ transgenes encode full-length Sxl proteins which comigrate with the 38-kDa (upper) band observed in wild-type females. Like the Sx-N transgenes (see above), we found that removal of the 3′ UTR sequences enhanced expression of the full-length protein: females transgenic for the Sx.FLΔ construct have higher levels of the 38-kDa isoform than lines with the full-length 3′ UTR (Fig. 9A; compare lanes 2, 3, and 4 with lines 5 and 6).

Additionally, in the Sx.FLΔ lines (which express very high levels of the 38-kDa Sx.FL transgene protein), there is a reduction in the amount of the 36-kDa isoform encoded by the endogenous gene to almost undetectable levels (Fig. 9A). To confirm and extend these results, we examined the Sxl protein profile in adult carcasses, in ovaries, and at different stages of development. In carcasses from adult females, as in whole extracts, only the 38- and 36-kDa Sxl isoforms are detected. In Sx.FLΔ females, there is a 1.5- to 3-fold reduction in the yield of the 36-kDa species compared to the wild type (see the legend to Fig. 9B for quantitation of data for each line). A more dramatic reduction (three- to sevenfold) of this isoform is observed in ovaries from Sx.FLΔ females. In addition to the isoforms which comigrate with the 36- and 38-kDa species seen in total extracts, two minor bands of 41 and 43 kDa are typically present in ovaries. These larger isoforms arise from the use of a 5′ donor site in the middle of exon 8 (44). They are typically found at much lower yield and can be difficult to resolve. As shown in Fig. 9B, these isoforms do not appear to be affected by the Sx.FLΔ transgenes. Furthermore, the transgenes did not have much effect on the expression of endogenous Sxl proteins (the 36-kDa species) in either 0- to 12-h or 12- to 24-h embryos, but there was a modest reduction in first-instar larvae (data not shown).

We also examined Sxl protein expression in females carrying the Sx-NΔ transgenes. The strongest Sx-NΔ line (line 28) shows a modest reduction in the levels of the endogenous 36- and 38-kDa Sxl proteins. However, in other lines that express less Sx-N protein, there is little effect on the expression of the endogenous Sxl proteins (data not shown). These findings suggest that the Sx-N protein is less able to repress Sxl protein expression from the endogenous gene than is Sx.FL.

DISCUSSION

Sxl belongs to a large family of RRM domain RNA binding proteins (6). Family members play significant roles in all aspects of RNA processing, including splicing, polyadenylation, and translational regulation (30). The strong similarity between family members and their involvement in multiple processes raises the question of how functional specificity is attained. The Drosophila Sxl protein is one of the best-characterized members of this family, and it provides a useful model for understanding what features of the protein contribute to its target specificity and regulatory functions. In this study, we examined the role of the Sxl N-terminal domain by comparing the biological activities of transgenes expressing a full-length Sxl protein and an N-terminal truncation. We show that deletion of the N terminus has much more severe consequences for the splicing functions of the Sxl protein than for its translation functions.

Regulation of the tra pathway.

In the current model for tra regulation, Sxl protein prevents the use of the default 3′ splice site by binding to the polypyrimidine tract and thereby blocking the binding of U2AF (48). A strong prediction of this model is that RNA binding activity should be sufficient for efficient tra regulation. Since the RNA binding activity of the full-length Sxl protein is reconstituted in vitro by a protein containing only the two RRM domains (26, 37, 42, 50), deletion of the N terminus should have little or no effect on tra splicing. Contrary to this expectation, our in vivo results demonstrate that the truncated Sx-N protein is defective in tra splicing regulation. Thus, additional activities of the Sxl protein are required for tra regulation.

Three lines of evidence support this conclusion. First, the Sx-N transgenes have no effect on the sexual differentiation of Sxl⁻ males. Even under conditions which suppress the strong male-lethal effects of the most active Sx-NΔ transgene (the Sxl⁻; msl-2^c transgene combination) and hence would be favorable for detecting sexual transformations, transgenic males are phenotypically wild type and fertile. Second, although the Sx-NΔ transgenes rescue the viability of females carrying several hypomorphic Sxl mutations at least as well as the Sx.FLΔ transgene, they do not rescue the sexual phenotype, and survivors are still phenotypically male. By contrast, the animals rescued by the Sx.FLΔ transgene resemble wild-type females. Finally, consistent with the apparent lack of biological activity, tra pre-mRNAs are spliced in a male-like pattern in Sxl⁻; Sx-N males. In this assay the Sx-N protein is not, however, completely inert, and very small amounts of female spliced tra mRNA can be detected after more extensive PCR amplification. In this context, it should be noted that Granadino et al. (22) generated an hsp70 transgene that expresses a Sxl protein lacking 94 aa from the N terminus and examined the state of tra splicing in these males following heat shock and recovery. After 40 cycles of PCR amplification, they found significant levels of female spliced tra in these males. They therefore concluded that the truncated protein has normal tra activity. However, we have found that the tra female splice product is preferentially amplified: the relative amount of the female product increases with each round of amplification, and after very extensive amplifications, its yield from the Sx-N transgenic lines can be equal to or greater than the yield of the unspliced and default PCR products. Therefore, Granadino et al. may have overestimated the tra regulatory activity of their much larger N-terminal truncation. It should also be noted that these authors did not determine whether their very large N-terminal truncation has biological function and can sex transform males.

Our finding that the first 40 aa of the Sxl protein play an essential role in the regulation of tra splicing is inconsistent with the current version of the U2AF blockage model. One possible explanation for the discrepancy is that the N terminus is essential for Sxl protein binding to tra pre-mRNAs in vivo, even though it is not critical in vitro. However, if this is the case, it is not clear why this special requirement for RNA binding would apply to tra but not to either msl-2 or Sxl RNA. An alternative, and we believe more plausible, hypothesis is that this region of the N terminus plays a critical role in some protein-protein interaction(s). In the simplest revision of the U2AF competition model, this interaction would be necessary to stabilize the association of Sxl with the default 3′ splice site. A precedent for this suggestion is the finding that protein-protein interactions are required to stabilize the binding of Tra, Transformer-2, and SR proteins to the dsx splicing enhancer in vitro (34). It is also possible that the N terminus is required in vivo because tra regulation involves mechanisms different from those envisioned in the U2AF competition model. For example, instead of simply competing with U2AF, Sxl might poison snRNP complexes associated with the default 3′ splice site via contacts mediated by its N-terminal domain. Alternatively, Sxl could have a positive role, promoting the assembly of snRNP complexes on the downstream female-specific splice site. Supporting the idea that tra regulation requires specific protein-protein interactions involving the N terminus, we have found that a chimeric protein consisting of the entire Sxl N terminus (but not the RRM domains) fused to β-Gal can weakly promote the female-specific splicing of tra pre-mRNAs in vivo in males (16a). Since both of the known Sxl protein-protein (Sxl-Sxl and Sxl-Snf) interactions are mediated by the RRM domains, we suppose that the N terminus would have to interact with some other, as yet unidentified protein. A good candidate would be the Fl(2)d protein, which Granadino et al. (21) have shown to be essential for tra pre-mRNA splicing.

Sxl positive autoregulation.

In agreement with the in vitro results of Wang and Bell (49), we find that the N-terminal truncation substantially impairs Sxl splicing regulation in vivo. As suggested above for tra, we presume that the female-specific splicing of Sxl pre-mRNAs is facilitated by a protein-protein interaction(s) mediated by the N-terminal domain. However, the fact that the Sx-N transgenes retain some autoregulatory activity argues that this interaction either is not absolutely essential for Sxl splicing or is redundant. With respect to the latter possibility, the truncated Sx-N protein should be able to participate in both Sxl-Sxl and Sxl-Snf interactions since it retains both RRM domains.

Dosage compensation.

The strong male lethality of the Sx-NΔ transgene even in the absence of a Sxl⁺ gene indicates that the N-terminal truncation retains substantial dosage compensation function. Since these male-lethal effects are suppressed by the constitutive H83M2 transgene, it would appear that the male-lethal effects of the truncated protein arise from a down-regulation of msl-2 translation. It has been suggested that repression of msl-2 depends on synergistic interactions between Sxl proteins bound to the 5′ and 3′ UTRs of the mRNA (3). The Sx-N protein should be capable of such synergistic interactions since the two RRM domains can mediate contacts between Sxl proteins (36, 42, 50). Although the Sx-N protein retains substantial dosage compensation function, its overall activity is less than that of Sx.FL. This can be seen by comparing the male-lethal effects of the Sx-NΔ and Sx.FLΔ transgenes in a Sxl^7B0; H83M2 background: while Sx-NΔ males are fully rescued in this genetic background, Sx.FLΔ males are only partially rescued. This difference could be due, at least in part, to a reduced activity of the Sx-N protein in the msl-independent dosage compensation system, where Sxl is thought to down-regulate X-linked gene expression in females by binding to poly(U) runs in the 3′ UTRs of mRNAs from X-linked genes (27). If this were correct, it would suggest that full dosage compensation activity may require interactions between Sxl and the translational machinery that depend on sequences in the N terminus.

Negative autoregulation.

The finding that Sxl translationally regulates msl-2 expression prompted Kelley and Kuroda (27) to examine the occurrence of multiple Sxl binding sites in transcripts from other genes. They discovered that mRNAs from many X-linked, but not autosomal, genes have three or more Sxl consensus binding sites within their 3′ UTRs. Surprisingly, one of these X-linked genes is Sxl itself. This observation raised the possibility that Sxl proteins might negatively regulate their own expression by associating with the 3′ UTR of the Sxl mRNA. This hypothesis is supported by several of our findings. First, we have shown that Sxl proteins are bound to the 3′ UTR of Sxl mRNAs in vivo. Second, the regulatory activities of the Sx.FL and Sx-N transgenes are substantially enhanced by deleting most of the 3′ UTR from the transgenes. Third, expression of the endogenous Sxl protein is reduced when the Sx.FL and, to a lesser extent, Sx-N proteins are highly expressed. Fourth, even though the activity of the Sx.FLΔ transgene is by many criteria much stronger than its Sx.FL counterpart, it is impaired in its ability to rescue females from the lethal effects of several hypomorphic Sxl mutations. This paradoxical result could be explained by the fact the Sx.FLΔ transgene may be more efficient than Sx.FL in repressing Sxl protein expression from the endogenous gene (upsetting the normal balance of Sxl protein isoforms).

While the positive (splicing) autoregulatory feedback loop provides a mechanism for maintaining the Sxl gene in the “on” state in females, it is possible that this feedback loop, operating unchecked, would produce toxic levels of Sxl protein (especially if the protein directly down-regulates expression of X-linked genes). A negative autoregulatory feedback loop would maintain homeostasis, keeping the levels of Sxl protein just high enough to maintain the positive autoregulatory feedback loop but below the level where the proteins could begin to have detrimental effects. Favoring the idea that this feedback loop is likely to have a role in fine-tuning the amount of Sxl protein, the Sx.FLΔ and Sx-NΔ transgenes do not have dominant effects in wild-type females (data not shown). Such a model is not without precedent; it is thought that the snf homolog in mammals (U1A) and the poly(A) binding protein in yeast control their own rates of accumulation by binding to the 3′ UTRs of their respective mRNAs and down-regulating translation (2, 10). It is also possible that Sxl negative autoregulation is a vital process. In this case, the two- to threefold induction over background of our transgenes would not be sufficient to reveal this essential role. Perhaps by removing the additional Sxl binding sites from the 3′ UTR of Sx.FLΔ we might obtain drastically higher levels of Sxl protein and be able to test this hypothesis.

This model would also help explain why the 3′ UTR profile of the Sxl mRNAs changes during development (41, 44). The Sxl mRNA profile is dynamic throughout development. During early embryogenesis, when Sxl protein must be rapidly synthesized to ensure that the positive (splicing) autoregulatory feedback loop is activated in all female cells (5), Sxl mRNAs with short 3′ UTRs—and few Sxl protein binding sites—predominate. Later in development when the Sxl gene is stably activated and a high rate of Sxl protein accumulation would no longer be required, the major Sxl mRNA species have long 3′ UTRs. Negative autoregulation mediated by Sxl protein binding to multiple sites in the long 3′ UTRs of these RNAs would ensure that the concentration of Sxl protein is maintained at a constant level. This concentration should be high enough to sustain the positive autoregulatory feedback loop but low enough to avoid toxic effects.

While our results are consistent with the idea that Sxl proteins negatively regulate their own synthesis through binding sites in the 3′ UTRs of the Sxl mRNAs, the molecular mechanism(s) of repression remain unclear. Northern blots indicate that the level of RNA from the transgenes with short 3′ UTRs (Sx.FLΔ and Sx-NΔ) is higher than from the transgenes with long 3′ UTRs (data not shown). This could mean that the RNAs with the longer UTRs turn over more rapidly (in our model this would be a consequence of Sxl protein binding). Alternatively, the reduction might be an indirect consequence of reduced translation. This puzzle is not unique to the Sxl transgene RNAs; for example, the amount of msl-2 RNA is less in females than in males (27). An additional complication with the transgene data is that the RNAs encoded by these constructs are not spliced. Since 3′-end processing is often coupled to splicing, it is possible that the postulated regulation of the transgene RNAs by Sxl proteins follows a pathway that is different in some respects from that of RNAs (like msl-2 or the endogenous Sxl mRNAs) which are subject to splicing. Further studies will clearly be required to elucidate how Sxl is able to reduce protein expression and to show conclusively that Sxl negatively autoregulates its own expression.