Recruitment of the Proneural Gene scute to the Drosophila Sex-Determination Pathway

Corresponding author: Thomas W. Cline, University of California, 11 Koshland Hall, Mail Code 3204, Berkeley, CA 94720-3204., sxlcline{at}uclink.berkeley.edu (E-mail)

Communicating editor: T. C. KAUFMAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In flies, scute (sc) works with its paralogs in the achaete-scute-complex (ASC) to direct neuronal development. However, in the family Drosophilidae, sc also acquired a role in the primary event of sex determination, X chromosome counting, by becoming an X chromosome signal element (XSE)—an evolutionary step shown here to have occurred after sc diverged from its closest paralog, achaete (ac). Two temperature-sensitive alleles, sc^sisB2 and sc^sisB3, which disrupt only sex determination, were recovered in a powerful F₁ genetic selection and used to investigate how sc was recruited to the sex-determination pathway. sc^sisB2 revealed 3' nontranscribed regulatory sequences likely to be involved. The sc^sisB2 lesion abolished XSE activity when combined with mutations engineered in a sequence upstream of all XSEs. In contrast, changes in Sc protein sequence seem not to have been important for recruitment. The observation that the other new allele, sc^sisB3, eliminates the C-terminal half of Sc without affecting neurogenesis and that sc^sisB1, the most XSE-specific allele previously available, is a nonsense mutant, would seem to suggest the opposite, but we show that housefly Sc can substitute for fruit fly Sc in sex determination, despite lacking Drosophilidae-specific conserved residues in its C-terminal half. Lack of synergistic lethality among mutations in sc, twist, and dorsal argue against a proposed role for sc in mesoderm formation that had seemed potentially relevant to sex-pathway recruitment. The screen that yielded new sc alleles also generated autosomal duplications that argue against the textbook view that fruit fly sex signal evolution recruited a set of autosomal signal elements comparable to the XSEs.

NATURAL selection can generate new regulatory gene functions by co-opting existing genes (reviewed recently in TRUE and CARROLL 2002 ). The X chromosome signal elements (XSEs) of Drosophila melanogaster provide an excellent opportunity for studying this basic evolutionary process, since (a) these sex-determination genes all have non-sex-specific functions that almost certainly predate their recruitment to the sex-determination pathway; (b) their recruitment took place relatively recently and rather precipitously; and (c) the molecular action of XSEs is understood fairly well (reviewed in SACCONE et al. 2002 ). Four XSEs determine fruit fly sex by communicating X chromosome dose to the master sex-determination switch gene, Sex-lethal (Sxl; reviewed in CLINE and MEYER 1996 ). The two copies of these X-linked genes present in wild-type XX embryos specify female development and the female rate of X chromosome dosage compensation by activating transcription of Sxl. The single copy of each XSE in wild-type XY embryos leaves Sxl silent, so they develop as males instead.

To function as XSEs, these genes had to acquire two abilities: to be expressed at an extremely early point in embryonic development before most zygotic genes are first expressed and to interact with their target, Sxl_Pe, the "sex-pathway establishment" promoter. What cis-acting information drives such early expression? Was it present in connection with these genes' non-sex-specific functions prior to recruitment, or did it instead arise in the course of recruitment as information specific to sex determination? Were changes in protein sequence required for XSE gene products to interact with Sxl_Pe, and if so, are these new residues important only for sex determination? These questions are explored below in connection with scute (sc), the strongest (WALKER et al. 2000 ) of the XSEs, and the one whose non-sex-specific role in development has received by far the greatest attention for the longest time (CAMPOS-ORTEGA 1998 ). sc was first studied to determine the nature of genes (CARLSON 1966 ), but interest continued due to the gene's relevance to pattern formation and neurogenesis (see below), to population (LONG et al. 2000 ) and evolutionary (SKAER et al. 2002 ) biology, and, most recently, to sex determination (CLINE and MEYER 1996 ).

In its non-sex-specific (proneural) role, sc works with the adjacent gene, achaete (ac), to help specify sense-organ mother cells (SMCs; reviewed in SKEATH and CARROLL 1994 ; CAMPOS-ORTEGA 1998 ; DAMBLY-CHAUDIERE and VERVOORT 1998 ; MODOLELL and CAMPUZANO 1998 ; CALLEJA et al. 2002 ). SMCs generate the highly stereotyped pattern of bristles on the adult. sc and ac encode basic-Helix-Loop-Helix (bHLH)-family transcription factors and are the most closely related of four highly homologous adjacent genes that comprise the achaete-scute complex (ASC) at the distal tip of the X chromosome (ALONSO and CABRERA 1988 ). One question addressed in the current study is whether the duplication event that generated sc and ac from a common ancestor was associated in time with the acquisition by sc of a sex-determination role.

Despite the attention it has received, sc has proven remarkably difficult to understand, even with the help of modern molecular tools. Handicaps impeding its analysis include the considerable functional redundancy between Sc and Ac proteins (BALCELLS et al. 1988 ; RODRIGUEZ et al. 1990 ; HINZ et al. 1994 ; PARRAS et al. 1996 ; SKEATH and DOE 1996 ) and the fact that the unusually small transcription units that encode these two proteins share enhancers that are scattered over nearly 100 kb of DNA (GOMEZ-SKARMETA et al. 1995 ) in a chromosomal region that does not undergo meiotic recombination. No transgene yet generated recapitulates the regulation of sc in its proneural role. These factors frustrate the standard tests by which one would define the null phenotype (not yet known after almost a century of study) and conspire against establishing allelism by means other than DNA sequence analysis. The discovery reported here regarding the unexpected molecular nature of sc^sisB1, an allele used extensively in previous studies of XSE function, illustrates how misleading analysis of sc mutant alleles can be.

As a consequence of the unusually high ratio of regulatory to coding DNA for sc, previous screens for sc mutations have generated a vast excess of gross chromosome rearrangements over point or pseudopoint mutations in or near the transcription unit, except when carried out in genetic backgrounds that include other lesions in the ASC from which new mutants cannot be separated (ALONSO and GARCIA-BELLIDO 1986 ). Here we describe a powerful screen based on suppression of male-specific lethality that overcomes this problem because the regulatory information for XSE function is far more compact and nearer the sc transcription unit than that for neurogenesis. Two of the new sc alleles generated by this approach proved to be defective only for XSE function and hence seemed particularly relevant to the question of how sc was co-opted for sex determination. These new alleles are compared to the most sex-specific allele known previously, sc^sisB1, and to sc^M6, the best candidate for a simple null.

The nature of the lesion in one of these new sex-specific alleles led us to ask whether the evolution of the C-terminal half of Sc protein was influenced by sc's acquisition of a sex-determination role. To address this question, we isolated new ASC orthologs and paralogs from a variety of fly species and generated the most detailed sequence comparison yet presented for the three most closely related members of the ASC. To test the functional significance of Sc residues highlighted by this comparative sequence analysis, we determined whether Sc protein lacking these residues could provide XSE function and whether Ac could substitute for Sc in sex determination. Because these experiments were designed so that the timing and level of expression of the tested proteins would be wild type, they provide a more meaningful test of interchangeability than do previous studies.

The lesion in the other new sex-specific allele led us to a cis-acting regulatory region downstream of the transcription unit that is likely to help drive the extremely early expression specifically required for XSE function. Strong functional synergism was observed between a deletion of this 3' sequence and lesions engineered in a heptameric sequence upstream of sc that was identified earlier (ERICKSON and CLINE 1998 ) as potentially important for XSE evolution only on the basis of DNA sequence comparisons.

The screen that generated these new sc alleles also produced autosomal duplications. These duplications do not support the textbook view of the Drosophila sex-determination signal (see SCHUTT and NOTHIGER 2000 ) as a ratio between the known XSEs and a corresponding set of analogous "denominator elements" that, with one interesting exception (YOUNGER-SHEPHERD et al. 1992 ; BARBASH and CLINE 1995 ), are simply presumed to exist. To avoid using terminology that makes the same presumption, we adopted the term XSE from nematode sex-determination studies (CARMI et al. 1998 ) to replace "numerator element" (CLINE 1988 ).

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks and culture:
Drosophila were cultured in uncrowded conditions on a standard cornmeal, yeast, sucrose, and molasses medium. Transgenes were tested for sisB⁺ activity only after five generations of backcrossing to a standard sc^M6 stock to generate homogeneous genetic backgrounds. Markers and balancers are described at http://flybase.bio.indiana.edu. D. virilis and Scaptodrosophila lebanonensis were from the Drosophila Species Stock Center (Bowling Green State University), while Musca domestica was from Carolina Biologicals.

A powerful F₁ genetic selection scheme for mutations in elements of the primary sex-determination signal:
The scheme diagrammed in Fig 1 was used to recover loss-of-function mutations in XSEs as well as duplications of autosomal signal elements (ASEs) as suppressors of male-specific lethality caused by simultaneous duplication of sisB⁺ and its target, Sex-lethal⁺. The extra copy of Sxl⁺ was present as a tandem duplication on the X chromosome [Dp(1;1)jnR1-A, Sxl⁺ l(1)7Aa^a Sxl⁺l(1)7Aa^c], while the extra sisB⁺ copy was carried on an insertion of most of the ASC and the nearby marker allele yellow⁺ into chromosome II. Table 1 describes the yield from two rounds of such a screen with gamma rays as the mutagen (4- to 5-day-old males exposed to 2750 rad from a ¹³⁷Cs source). We also examined 33,000 and 115,000 X chromosomes that had been mutagenized instead by ethyl methanesulfonate (EMS) or diepoxybutane (DEB), respectively (see Table 1 legend).

View larger version (31K): [in this window] [in a new window] [Download PPT slide]   Figure 1. An F1 positive selection scheme for loss-of-function mutations in X chromosome signal elements and gain-of-function mutations or duplications in autosomal signal elements. Such mutations suppress the male-specific lethality caused by simultaneous increases in the dose of scute+ [as Dp(1;2)sc19] and Sxl+ [as Dp(1;1)Sxl+]. See MATERIALS AND METHODS for details.

Fewer than 0.06% of unmutagenized males carrying the sisB⁺ duplication (y⁺ sons from cross 2, Fig 1) survived (data not shown); however, such a low escaper rate was observed only if growth temperature was kept at 18° not only for the sons themselves during early embryogenesis, but also for their mothers from the time that they pupated until they laid their eggs—a remarkably long time, considering that no mutant maternal effect is involved. Regardless of the mutagen used, the escaper rate for mutagenized males was several times higher than that for unmutagenized animals, but always <1% if maternal and zygotic culture temperatures were kept low. The four times higher escaper rate in round A vs. round B in Table 1 reflects the use in A of a shi^ts-based scheme for generating virgins in large numbers that exposed developing mothers to 29°. Once the importance of maternal temperature was appreciated, the pn-Kpn scheme for making virgins shown in Fig 1 was used instead. Because rare y⁺ males were immediately subjected to a simple retest cross (Fig 1, cross 3) in which most proved to be sterile, even the relatively high escaper rate in round A was not a serious inconvenience. Although this retest cross was run at 25° for convenience, the distinction between escapers and suppressors was unambiguous even at this higher temperature. Lines were established and maintained only for suppressors rescuing >30% of males in this retest because experience taught us that genetic characterization of suppressors less effective than this was impractical. Even with the 30% cutoff, many partial-loss-of-function, fully viable alleles of the fairly weak XSE outstretched (os), an essential gene, were recovered.

The standard mutagen EMS was ineffective in this F₁ screen, undoubtedly because EMS induces mosaic genetic change (LEE et al. 1970 ), yet male survival required suppression throughout the embryo. Application of EMS to the flies was clearly effective, as 53 white mutants were observed (1/623 mutagenized X chromosomes, with most being mosaics). DEB was more effective than EMS, but still three times less effective than gamma rays for generating suppressors and hence not worth the greater effort that its application entailed.

Originally, only y sons from positive retest crosses (Fig 1, cross 3) were sent through cross 4, the purpose being to discover whether survival of their y⁺ brothers in the retest cross might have been due to a spontaneous mutation in Dp(ASC), rather than to an induced mutation elsewhere. By not also applying the test to sons from apparently negative retest crosses, we knew we would miss suppressors induced on chromosome II, since they would segregate from the Dp(ASC) used to show suppression in the retest. However, when it became apparent that we were unlikely to recover any suppressors on chromosome III, we felt it worthwhile to apply the cross 4 test to progeny from all fertile retest crosses. By removing the chromosome II blind spot in this way, we could learn whether the screen was at least able to detect duplications of the one known autosomal signal element, deadpan (dpn). Only if we recovered changes in dpn could we attach significance to a failure to recover suppressors on III. This expanded test was applied to 35% of the DEB screen and to all of round B of the gamma-ray screen (Table 1).

Cross 5 (Fig 1) generated lines from progeny of males that had passed the retest hurdle and could be maintained with minimum effort pending further characterization. In these lines, the Dp(ASC) chromosome [homozygous male sterile in our stock even in the absence of Dp(Sxl⁺)] was balanced against an In(2)NS chromosome carrying a recessive lethal. This arrangement forced males to keep Dp(ASC), and hence each new suppressor, each generation.

Using variously marked Dp(Sxl⁺) chromosomes in trans, we roughly mapped suppression of Dp(ASC),sisB⁺ male lethality for each line relative to pn (0.8), cm (18.9, very near Sxl at 19.1), sn (21.0), v (33.0), and f (56.7). This effort also told us (a) whether the suppressor behaved as a single gene trait, (b) whether it was X linked or autosomal, and (c) whether it might be associated with a chromosome rearrangement. With this information, we could remove Dp(Sxl⁺) from X chromosomes carrying suppressors in genes other than Sxl to ascertain whether those suppressors displayed the female-specific lethality seen previously for mutations in XSEs.

When suppressors mapped near cm, the marker closest to Sxl, or when recombination relative to cm was blocked, indicating a rearrangement that would preclude separating the suppressor from the tandem duplication of Sxl, we used a simple genetic test to determine if the survival of y⁺ males was due to mutations in one of the two tandem copies of Sxl⁺ on the mutagenized X chromosome. This test discriminates between one vs. two copies of Sxl⁺ and is based on the antimorphic allele splicing-necessary-factor¹⁶²¹ (snf¹⁶²¹), which specifically interferes with Sxl autoregulation (DESHPANDE et al. 1996 ; SALZ and FLICKINGER 1996 ; CLINE et al. 1999 ). Even when only heterozygous, snf¹⁶²¹ causes Sxl⁺ to be haplo-insufficient in females (OLIVER et al. 1988 ; STEINMANN-ZWICKY 1988 ). The dominant zygotic interaction between snf¹⁶²¹ and Sxl^- is more extreme at a low culture temperature, and, owing to the maternal effect of snf, is enhanced when mothers for the test are themselves snf¹⁶²¹/+. Indeed, only with snf¹⁶²¹/+ mothers does the assay test functioning of Sxl during early embryogenesis, and even then the test is probably not definitive for the earliest function of Sxl—counting X chromosomes (T. W. CLINE, unpublished results). If either of the two tandem copies of Sxl⁺ on the chromosome to be tested is nonfunctional, snf⁺ "Dp(Sxl⁺)"/snf¹⁶²¹ Sxl^- females will be functionally equivalent to snf⁺ Sxl⁺/snf¹⁶²¹ Sxl^- females and will suffer sterility, masculinization, and/or reduced viability at 18°. Using this snf test, we determined that 63% of X-linked suppressors were hits in Sxl (Table 1). This result is expected both because the Sxl⁺ alleles are far larger targets than the other XSEs and because complete loss of Sxl function is not deleterious to males, while null alleles of all XSEs other than sisB would be lost due to male lethality arising from these genes' non-sex-specific functions. Unanticipated difficulty extracting these potentially interesting new Sxl mutations from the tandem duplication by recombination limited their usefulness (data not shown).

DNA and RNA analysis:
Genomic DNA was isolated using either DNAzol (Molecular Research Center, Cincinnati) or the Quick fly genomic DNA prep from the Berkeley Drosophila Genome Project Protocol (SPRADLING et al. 1999 ). RNA was prepped according to the Trizol protocol (GIBCO BRL, Gaithersburg, MD). 5'-Rapid amplification of cDNA ends (5'-RACE) and RT-PCR were carried out according to FROHMAN 1994 using Superscript II (Bethesda Research Laboratories, Gaithersburg, MD) and TdT (Promega, Madison, WI). Except when noted, all PCR reactions were carried out with Tfl (Epicenter, Madison, WI) or Taq (Perkin-Elmer, Norwalk, CT) polymerase using standard reaction conditions.

Molecular analysis of sc mutations:
For sc^sisB1, sc^sisB2, and sc^sisB3, PCR fragments were generated from genomic DNA and both strands of the coding region were sequenced. Each allele was sequenced independently twice. For sc^sisB2 and sc^sisB3, a 5.2-kb Xba-EcoRI fragment extending 0.6 kb 5' and 3.2 kb 3' to the transcription unit was sequenced in its entirety. Southern analysis (DIG System protocol, Boerhinger Mannheim, Indianapolis) was carried out for all sc alleles and covered 35 kb of genomic DNA including scute and lethal-of-scute (l'sc; from 13.4 kb upstream of sc to 6.9 kb downstream of l'sc) and 9 kb flanking ac (4.9 kb upstream and 3.3 kb downstream).

RNase protection analysis of sc mutants:
Total RNA was made from pooled collections of 1- to 3-hr embryos grown at 25°. Oligotex (QIAGEN, Valencia, CA) beads were used to isolate poly(A)+ mRNA. RNase protections were carried out according to the RPA III protocol (Ambion, Austin, TX) using 1 µg of poly(A)+ RNA with an excess of labeled probes (2 fmol each of scute and tubulin) and loading the entire reaction mixture. Two different dilutions of an RNase A/T1 mix were used. The sc probe was generated from a 453-bp XhoI-NciI genomic fragment subcloned into pBluescript (Stratagene, La Jolla, CA) overlapping the transcriptional start sites. The tub probe was a 400-bp fragment that extended across an intron (HEDLEY and MANIATIS 1991 ).

Cloning ASC genes from non-melanogaster species:
On the basis of the sc protein-coding sequence from D. melanogaster and D. subobscura (BOTELLA et al. 1996 ), degenerate oligonucleotide primers were designed to two regions unique to sc that had high conservation and low codon degeneracy: degsc3 5'-TG(A/G)AA(A/G)TG(T/C)TC(A/G)TA(G/T/C)GG(T/C)TC-3' and degsc4 5'-CC(G/A/C)TA(T/C)AA(T/C)GT(G/C)GA(T/C)CA-3'. sc orthologs from D. virilis were amplified using these primers in standard PCR reactions at 42° annealing temperature for 35 cycles, followed by a second PCR round at a 55° annealing temperature for 35 cycles. Bands were gel purified and subcloned into pCR2 (Invitrogen, San Diego), transformed, and colony screened (MANIATIS et al. 1989 ) using a melanogaster sc probe. For S. lebanonensis sc, a 37° annealing temperature for 35 cycles, followed by a 50° annealing temperature for 30 cycles, was used instead. On the basis of alignments incorporating this new sequence information, new degenerate oligomers to conserved regions were designed: MASC1 [5'-GT(A/C/G)AAGCAGGTGAACAA-3'] with degsc3 amplified Musca sc while MASC3 [5'-AT(A/G)TA(A/G)TCCA(A/G) (A/G)ATCTCCTC-3'] with MASC1 amplified ac from lebanonensis and l'sc from Musca. Oligomers AS1[5'-CG(G/A/T/C)CG(G/A/T/C)AA(T/C)GC(A/C)CG(G/A/T/C)GA(A/G)CG-3'] and AS4[5'-TC(C/T)TGCCA(C/G)AG(G/A/T/C)GA(T/G/A)AT(G/A)TA(A/G)TC-3'] amplified l'sc from lebanonensis and AS3 [5'-AA(G/A)GT(G/T)(G/A)A(T/C)AC(C/G)(T/C)T(G/A) CGCAT-3'] plus AS4 amplified ac in Musca. RT-PCR and 5'-RACE products were used to identify the 5' and 3' ends of each of the genes, and the sequences were subsequently confirmed for genomic DNA amplified using primers based on this information.

Isolation of D. virilis genomic DNA flanking the sc transcription unit:
An 5-kb PCR fragment of genomic DNA including information 5' to the sc transcription unit was generated using one primer internal to sc and another corresponding to an SMC enhancer reported previously (CULI and MODOLELL 1998 ). Southern analysis (Genius System, Boerhinger Mannheim) with this 5-kb fragment as a probe identified a single 12-kb Xba-Xba sc-containing fragment, X1.1B2, which was gel purified and directly cloned into a phage vector (Lambda FIX II, Stratagene). After sequencing the region 3' to the sc coding region in X1.1B2, a probe made from this region was used in Southern analysis to identify a 10-kb EcoRI genomic fragment containing this region. This recovered clone, C1.22, shared 2 kb of sequence that was identical to the 3' end of X1.1B2.

Site-directed mutagenesis and germline transformation:
Standard techniques were used for germline transformation (SPRADLING 1986 ). All transformation constructs were engineered into a vector derived from the w⁺ Casper P-element clone p[B22 T4(fs), w⁺] (ERICKSON and CLINE 1991 ). All site-directed mutagenesis (Quick-Change site-directed mutagenesis kit, Stratagene) was carried out on a 4-kb Xho-Kpn wild-type sc-containing fragment from p[B20 T4, w⁺] (ERICKSON and CLINE 1991 ) and cloned into pBluescript (Stratagene). All DNA used for transformation was sequenced prior to injection. DNA from wild-type and mutant clones was moved in several steps into the w⁺ Casper P-element vector (PIRROTTA 1988 ) containing an 8.8-kb fragment of sc genomic DNA that included 2.4 kb 5' and 5.3 kb 3' of the transcription unit. For the heptamer triple knockout, the two closely linked repeats (sites 2 and 3) located 245 and 236 bp, respectively, upstream of the scute translation start site were eliminated first using two 46-bp complementary primers with mismatches to the heptamers: 5'-CACTCACTTCGAGTTCCAATAAACTGCATTTATCTCTTGCCGTCAC-3' (changed from wild type: —TCCCTACCTGTGCAGGTAGCTC—). The double knockout clone served as a template for a second reaction to alter heptamer site 1 located at -274: 5'-CAGAGAAAGAGAGAGAGAGTAGGTCTGGCTCACTCAC-3' (wild type: —AGA CTACCTGTGG—).

To facilitate replacement of the melanogaster sc coding region with a variety of other sequences as NotI-AflII fragments, we used site-directed mutagenesis to place restriction sites at the beginning and end of the sc protein-coding region. The melanogaster and simulans sc translation start sites are unusual in beginning upstream of those in all other sequenced sc genes. Hence, to avoid disrupting potential regulatory information that might be specific for melanogaster sc, the foreign protein-coding regions to be tested were inserted at the site of a methionine 12 residues downstream of the melanogaster start site. This methionine is conserved in all sc genes sequenced and is the initiating methionine in most. Introducing a NotI site at this methionine in the cassette changed the next residue from a conserved serine (a signature residue for Sc) to alanine for the D. melanogaster and M. domestica Sc transgenes. D. melanogaster Ac was unchanged, since it already had alanine at this position. This change was introduced using two complementary 38-bp oligomers (scNcoF and scNcoR) with the following single-nucleotide change: 5'-CAACGAAAAGCACTACCATGGCATCGAGTGTGCTGTCC-3' (wild type: —CTACCATGTCAT—).

Following introduction of the N-terminal change, we used two complementary 40-bp oligomers (scAflF and scAflR) to introduce a single change just after the TGA stop at the C terminus that introduced an AflII site: 5'-CTCTATGGCAGGAGCAGTGACTTAAGCCCCAAAATTTACC-3' (wild type: —TGACTTAATCCC—). By positioning the AflII site this way, in all cases the termination triplet would belong to the inserted coding region of interest, but everything more 3' would be from the melanogaster gene. The only change caused by the AflII site was conversion of a nonconserved T to G six nucleotides distal to the stop codon.

Inserts with an NcoI site at the start codon and an AflII site just following the stop codon were constructed for melanogaster ac and Musca sc using a single forward primer for the NcoI site and a single reverse primer for the AflII site (the complement of the reverse primer is shown). For the ac gene, the protein generated was wild type, while for the sc gene, the serine following methionine was changed to alanine. For melanogaster ac: acNcoF, 5'-TCTTACCATGGCTTTGGGCAGC-3' (changed from wild type: —AAAATGGCT—); acAflR, 5'-GACCTGTAACTTAAGAGATCAAATC-3' (wild type: —TAAAAAAACAGA—). For Musca scute: MscNcoF, 5'-CAAATACGACCATGGCAAGTGTTAG-3' (wild type: —CGAGAATGTCAA—); MscAflR, 5'-GGCAGGAACAGTAACTTAAGAACACAAAATC-3' (wild type: —TAAAAACAAAAC—). PCR products were subcloned into the pCRII vector (Invitrogen), cut with NcoI and AflII, and subcloned into the wild-type cassette vector described.

For the sc^sisB2 deletion-mimic transgene, site-directed mutagenesis introduced an AvrII restriction site at the exact location of the 5' end of the deletion breakpoint (indicated as "/" below) The oligomer (and complement not shown) for G15Avr2REV was: 5'-GAAATCAAGGCAGCGAC/CTAGGTTCACAGGACTCGCG-3' (sc^sisB2 parental was: —GCGAC/TCAGTTTC—). The Casper clone was cut with EcoRI and gel purified, cut with AvrII, and then treated with mung bean nuclease. The fragment was gel purified, mixed with DNA ligase, transformed, and screened using PCR. Positives were sequenced for the presence of the desired junction: 5'-genomic GAAATCAAGGCAGCGAC/CTAGTATGTATGCasper polylinker-3'. For the transgenes containing both the heptamer triple knockout and the sc^sisB2 deletion, a small band from the mutant heptamer construct was cut and inserted into the sc^sisB2 deletion transformation vector.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

This study of the relationship between the sex-determination and proneural functions of scute began with the recovery of two new sc alleles that were unlike any previously described with respect to the degree of specificity of their functional defect: they seemed to be defective only for sex determination. Consequently, they seemed likely to reveal aspects of sc gene structure uniquely involved in XSE function. These alleles were generated along with a number of other mutations shown in Table 1 in a powerful F₁ suppressor screen for mutations in zygotically acting regulators of Sxl (see MATERIALS AND METHODS).

Recovery of mutations in XSEs:
Five new mutations from this screen identified a previously unknown regulator of Sxl, the vital gene os (alias unpaired). Their characterization and experiments establishing that os is a bona fide XSE are reported elsewhere (SEFTON et al. 2000 ).

Five gamma-ray-induced and one DEB-induced X-linked mutation listed in Table 1 were identified as likely lesions in the X-linked ASC on the basis of map position and/or bristle phenotype. One of these, In(1)ac^T1, displayed a classic ac bristle phenotype; however, since this chromosome had an inversion with one breakpoint in or near the ASC and the other in the vicinity of Dp(Sxl), the possibility that its recovery was due to disruption of Sxl rather than of the ASC could not be excluded, despite passing a test for Sxl⁺ function described in MATERIALS AND METHODS. Protein-coding regions for ac and sc in this chromosome were wild type (data not shown). Of the five other mutants, three displayed bristle phenotypes consistent with ASC lesions. Two of these had especially severe bristle phenotypes and were <25% viable, indicating a gross lesion in the ASC affecting more than one gene, and were not studied further. The third, ultimately named sc^T1, was molecularly characterized because it had high male viability and displayed a classic strong scute phenotype. The last two mutants in this group, sc^sisB2 and sc^sisB3, generated the most interest because they displayed no bristle defects.

Two y⁺ males in the gamma-ray mutagenesis, and two in the DEB screen, came through the selection because the ASC⁺ duplication Dp(1;2)sc¹⁹ contributed by their mothers had mutated spontaneously. Two of the four mutant duplications were likely to be gross disruptions, since they failed to complement Df(1)sc¹⁹, the corresponding deletion of most of the ASC, for viability in males. The other two mutants, sc^T2 and sc^T3, were viable in this test and were characterized molecularly.

Recovery of mutations in ASEs:
Five suppressors on chromosome II were recovered (Table 1). All were large tandem duplications. Three duplicated overlapping regions in 2R: MR1 (42C–44D), MR2 (42A–45E), and MR3 (41F–44D). For all three lines, suppression segregated from a w⁺-tagged P-element insertion at 44D in trans, indicating that suppression was caused by the rearrangements. This point was established directly for MR2 when the loss of suppressing activity in one line was seen to be associated with spontaneous loss of the duplication. All three 2R rearrangements increase the dose of deadpan (at 44C), the only ASE identified in an earlier, extensive, genome-wide screen for loss-of-function mutations in ASEs (BARBASH and CLINE 1995 ). The other two autosomal suppressors were associated with even larger tandem duplications in 2L: MR4 (33A–37E) and MR5, a complex rearrangement with four breakpoints between 34C–D and 38C that duplicate most of that region. Extensive subsequent efforts to identify the specific dose-sensitive region(s) in MR4 and MR5 responsible for their phenotypes were unsuccessful. Noise from genetic background differences among the defined duplication and deficiency stocks employed in this effort overwhelmed effects of individual mutant chromosomes so that no consistent picture emerged (data not shown).

Viability effects of sc mutants reveal Sxl activation by XSEs to be intrinsically heat sensitive:
The two new mutant alleles sc^sisB2 and sc^sisB3 that appeared wild type for bristle formation (neurogenesis) behaved as recessive, female-specific lethals (crosses A and B, Table 2). In both cases, lethality was temperature dependent, as was sc^sisB1 (cross E), the most female-specific sc allele previously known (GARCIA-BELLIDO 1979 ; CLINE 1988 ; TORRES and SANCHEZ 1989 ; TORRES and SANCHEZ 1991 ; ERICKSON and CLINE 1993 ). Even the much less functional nonsense mutant allele sc^M6, the best candidate for point-mutant null on a chromosome that is otherwise wild type (GOMEZ-SKARMETA et al. 1995 ), proved to be heat sensitive with respect to female viability effects (Table 2, cross F). Crosses C and D of Table 2 show that the female-specific lethality of the two new sc mutants is due to their failure to activate Sxl's feminizing functions, since Sxl^M4, a gain-of-function allele locked in the female expression mode (BERNSTEIN et al. 1995 ), fully rescued both mutants even when they were hemizygous. The fact that every sc mutant examined is heat sensitive with respect to female viability suggests that temperature sensitivity is an intrinsic feature of this system.

The fact that sc^sisB2 and sc^sisB3 were less lethal than sc^M6 suggested that the new alleles are hypomorphic rather than amorphic for XSE function. Data in Table 3 confirm this point. In this experiment, the relative strength of each mutant's sex-determination defect is assayed as a dominant, female-specific lethal synergism with a mutation in sisA, the other strong XSE. By this assay, the defect in sc^sisB2 is comparable to that for sc^sisB1, while that of sc^sisB3 is somewhat more severe, but not as strong as sc^M6. This phenotypic series for XSE activity is the same as that from the different assay in Table 2. Three new sc alleles with strong bristle defects are also included in Table 3 (crosses 6, 7, and 8) for comparison. These mutants are as defective as sc^M6 for sex determination.

Two different crosses (4 and 5) for assaying the strength of sc^M6 are shown to illustrate the variability (9 vs. 1%) that one typically encounters as a consequence of uncontrolled differences in genetic background and maternal age. In light of this variability, the most reliable comparisons for assessing the relative strength of two XSE mutations are made when both alleles are transmitted to the affected daughters from the same set of parents, as in crosses 9 and 10. Again, daughters with the new sisB alleles were more viable than their sisters with sc^M6, arguing again that the mutants have not lost all sc XSE function. And again, sc^sisB2 appeared less impaired than sc^sisB3, despite the fact that the magnitude of the lethal interaction was different in crosses 9 and 10 than in crosses 2 and 3 (69 and 36% vs. 30 and 14%, respectively).

Even by more stringent tests, sc^sisB2 and sc^sisB3 are essentially wild type for neurogenesis:
Unless an allele is devoid of function, hemizygous mutant females are expected to have half as much sc proneural activity as homozygous mutant females or hemizygous mutant males; hence, a mild defect in proneural function for sc^sisB2 or sc^sisB3 might be exposed in hemizygous females that would not be apparent in homozygous females or in males. Fig 2 compares hemizygous and homozygous mutant females with respect to bristle formation. Potential complications from viability effects were minimized by the use of Sxl^M4 to provide the required Sxl⁺ function, regardless of XSE activity. In(1)sc8Lsc4R (abbreviated scute) eliminates 22 kb of ASC DNA that includes the sc transcription unit, while Df(1)sc¹⁹ ( ASC in Fig 2) deletes this and 68 kb more, including the two flanking transcription units ac and l'sc (CAMPUZANO et al. 1985 ).

View larger version (49K): [in this window] [in a new window] [Download PPT slide]   Figure 2. Comparison of homozygous vs. hemizygous female bristle phenotypes for various sc mutants. The phenotypes shown are for females (n > 20) grown at 22° that carry the constitutive allele SxlM4 to suppress viability effects that would otherwise arise from failure to activate Sxl. Macrochaete abbreviations are standard (GARCIA-BELLIDO 1979 ). Solid circles indicate bristles present in 90–100% of the cases, open circles indicate bristles present in 50–89%, and "x" indicates bristles present in <50%. scute is In(1)sc8Lsc4R, which eliminates the sc transcription unit, while ASC is Df(1)sc19, which eliminates the achaete and lethal-of-scute transcription units as well as sc.

The data for sc⁺ in Fig 2 show that a single copy of the ASC normally suffices for bristle formation in females: only a few bristles were occasionally abnormal in females hemizygous for the complex. In contrast, females with sc^sisB1 in trans to scute were clearly abnormal, and in trans to ASC the phenotype approached that of homozygotes for the putative null, sc^M6. Clearly, sc^sisB1 is more impaired for proneural activity than is apparent from homozygous females or hemizygous males. Bristle defects with sc^sisB1 in trans to scute had been reported earlier (GARCIA-BELLIDO 1979 ), but in those studies the possibility of bias caused by viability effects could not be eliminated.

For sc^M6 as well, the bristle phenotype of scute hemizygous females was considerably more extreme than that of homozygotes. By itself, this observation would indicate that sc^M6 is not null; however, the observation that the phenotype of the ASC hemizyotes was even worse showed that the level of products from ac and/or l'sc is relevant to the sc mutant bristle phenotype when the activity of sc itself is low. Hence, the difference between sc^M6 homozygotes and sc^M6/ scute hemizygotes may reflect effects of this sc^- deletion on expression of ac and/or l'sc (GOMEZ-SKARMETA et al. 1995 ); however, the possibility cannot be excluded that sc^M6 is not null or, alternatively, that the chromosome carries an undiscovered additional lesion affecting another member of the ASC. Since sc was named for effects on the scutellar bristles (CARLSON 1966 ), it is ironic that the scutellars (ASC and PSC) were unaffected in sc^M6 homozygotes and only moderately affected even in sc^M6/ ASC hemizygotes.

In contrast to sc^sisB1 and sc^M6, the mutants sc^sisB2 and sc^sisB3 displayed wild-type bristle patterns even in trans to ASC, strong evidence that these two XSE-defective alleles are wild type with respect to proneural activity. Because this conclusion is so important, its validity was tested by a second, very different assay that we believed would be even more sensitive to reductions in proneural function. This assay is based on the fact that extramacrochaetae (emc) is a negative regulator of sc (reviewed in CAMPUZANO 2001 ). Partial loss-of-function mutations in emc cause excess sc⁺ activity that induces ectopic macrochaetae (large bristles). A mutation in sc that lowers the level of sc activity should reduce the number of ectopic bristles. Of the three regions of adult cuticle scored for ectopic bristles (Table 4), the dorsal notum had the largest number (18), so data from that region are likely to be most reliable.

Cross 1 in Table 4 shows that the putative null allele sc^M6 had the expected dramatic effect on ectopic bristle formation. Not only did it eliminate most ectopic bristles in hemizygous males, but also it reduced their number to below half of the sc⁺ control value even when only heterozygous in females. Hence a mutation that reduced sc activity by only half in males would be expected to have an unambiguous effect in this assay. Even by this sensitive test, sc^sisB2 is essentially wild type. Mutant males showed a wild-type level of ectopic bristles in the notum and scutellum. Ectopic bristles in the head were somewhat reduced but still above half of the wild-type reference value. The effect of sc^sisB3 was a bit stronger, but even this allele must have more than half the activity of a wild-type allele, since its effects on ectopic bristle number in males were less than those of sc^M6 in heterozygous females in all but the dorsal head region where few ectopic bristles were observed even in the sc⁺ control.

Another negative regulator of the ASC that we exploited for tests of functional specificity is hairy (h). This repressor was originally thought to be specific for ac with respect to its regulation of the ASC (BOTAS et al. 1982 ; MOSCOSO DEL PRADO and GARCIA-BELLIDO 1984 ; OHSAKO et al. 1994 ; VAN DOREN et al. 1994 ), in part because h mutants generate only ectopic microchaetae (small bristles), a type of bristle that sc loss-of-function mutations do not affect. However, as the relationship between the classic ac and sc phenotypes and the functioning of the ac and sc transcription units has become less clear, workers in the field have begun to hedge on the regulatory specificity of h for ac vs. sc (MODOLELL and CAMPUZANO 1998 ). Data in Table 5 show that the putative null allele sc^M6 had an effect in the h-suppression assay: sc^M6 eliminated nearly 75% of ectopic microchaetae in the dorsal scutellum. Even sc^sisB1 had a significant effect on ectopic bristle frequency in the h assay. Either both alleles are somewhat defective with respect to ac as well as sc, or sc as well as ac contributes to the h ectopic bristle phenotype. In contrast, sc^sisB2 and sc^sisB3 were indistinguishable from sc⁺ by this assay.

Surprises in the molecular characterization of old as well as new sc alleles:
Fig 3 shows the molecular lesions associated with the five new suppressors of Dp(XSE)-induced male lethality that mapped to the ASC and were not associated with gross chromosomal rearrangements or greatly reduced male viability. The figure also shows the lesion reported for the strong mutant sc^M6 (which we confirmed), and the lesion we discovered for sc^sisB1, the most sex-specific sc mutant available prior to this study. All five of the new sc alleles selected for defective XSE function had lesions in the region defined by the smallest (5.2 kbp) transgene with full XSE activity. The three new mutants with strong sc bristle phenotypes were associated with gross changes in the local DNA. sc^T2 carried an insertion of a b104 transposable element in the protein-coding region. We did not determine the specific nature of the alterations at the sc^T1 and sc^T3 breakpoints, nor did we define the location of those breakpoints more precisely than shown by the regions bracketed in Fig 3.

View larger version (23K): [in this window] [in a new window] [Download PPT slide]   Figure 3. The molecular nature of new and relevant older sc mutant lesions presented in the context of other molecular aspects of the gene. The shaded region (bottom) shows the extent of the smallest transgene (5.2 kbp) that provides full X chromosome signal element function (ERICKSON and CLINE 1991 ). Regions of homology between D. virilis and D. melanogaster that satisfy the following criteria are diagrammed to scale as hanging solid bars: they must (a) contain a core of at least eight consecutive identical residues, (b) have no breaks of nonhomology beyond that core that are larger than two adjacent residues, (c) terminate in regions of at least two identities, and (d) contain guanine or cytosine residues. The virilis region sequenced for comparison extended from the SMC enhancer involved in proneural function on the 5' end (CULI and MODOLELL 1998 ) to 1.2 kb past a block of homology just beyond the 3' end of the melanogaster XSE+ transgene shown.

The molecular lesions in the two new alleles with reduced XSE activity but normal proneural function were remarkably different from each other. sc^sisB2 was deleted for 2594 bp starting 2.15 kbp downstream of the end of the sc transcription unit; hence, this mutant identifies cis-acting sequences needed for, and possibly only for, XSE function. In contrast, for sc^sisB3, five foreign base pairs were substituted for six endogenous residues beginning at Gly172, introducing an immediate translation termination codon and a frameshift that would garble translational readthrough products. The lesion is predicted to eliminate more than half of the Sc protein, including the highly conserved C terminus.

Although the phenotype of sc^sisB1 is comparable to that of sc^sisB2 and sc^sisB3 with respect to XSE activity and includes only rather subtle defects in neurogenesis, we were surprised to discover that its lesion more closely resembles that of the far more defective allele, sc^M6: a nonsense mutation at Lys136 within the bHLH region, not far downstream from the sc^M6 nonsense mutation (Gln114). Lys136 is conserved among all the higher Diptera examined and helps distinguish Sc from its two closest paralogs in the ASC (see below). The higher activity of sc^sisB1 relative to sc^M6 is most likely due to the fact that the sc^sisB1 chromosome carries a tandem duplication that doubles the dose of the sc^sisB1 allele and the adjacent wild-type l'sc transcription unit (Fig 4). The breakpoints of this duplication did not fall within 5.3 kb of either side of the sc transcription unit or within 2.6 kb upstream or 1.5 kb downstream of the l'sc transcription unit (data not shown). The simplest interpretation of our Southern data would limit this duplication to the vicinity of the sc and l'sc transcription units.

View larger version (17K): [in this window] [in a new window] [Download PPT slide]   Figure 4. Southern analysis reveals a duplication associated with the scsisB1 nonsense-mutant allele. (A) A PstI digest of genomic DNA from the three different scsisB mutant alleles probed with a fragment shown schematically below, which is expected to generate wild-type bands of 11.0 and 16.2 kbp, with the larger band containing sc. The nature of the large mutant band observed for scsisB1 in A is explored in B, which shows a PstI digest of iso-X male DNA from the wild-type and two separate scsisB1 lines (q and r), this time probed with a fragment corresponding to only the sc coding region. Data in C test whether lethal-of-scute is also duplicated in scsisB1. DNA from the same males as in B was digested instead with SphI and probed with the l'sc coding region.

RNAse protection analysis presented in Fig 5 supports this view of sc^sisB1. The level of mutant RNA extracted from 1- to 3-hr-old sc^sisB1 embryos is clearly higher than that of either the wild-type or the two new sex-specific mutants. In contrast, the mRNA level for sc^sisB2 is significantly lower than that for wild type during very early development, as expected if the DNA deleted in this allele influences the timing or rate of transcription. The mRNA level for the protein-coding-defective allele sc^sisB3 appeared normal.

View larger version (52K): [in this window] [in a new window] [Download PPT slide]   Figure 5. RNase protection analysis shows that scsisB2 reduces transcript level while scsisB1 increases it. Poly(A)+ mRNA from 1- to 3-hr embryos of each mutant was analyzed. The image shown is of a single gel. The "No RNA" control lanes contain probe but no fly RNA. Two sc protection products were observed corresponding to the two predicted sc transcriptional start sites (VILLARES and CABRERA 1987 ). Equal loading of the lanes was confirmed by the tubulin internal control.

Protein sequence conservation patterns for ASC genes do not account for the sex-specific phenotype of sc^sisB3 but do show that sc diverged from ac before becoming an XSE:
The lesion in sc^sisB3 suggested that the C-terminal half of Sc is far more important for sex determination than for neurogenesis and led us to ask whether the pattern of Sc protein sequence conservation in this region might suggest which, if any, residues were required only for sex determination and hence were relevant to the recruitment of sc as an XSE. Residues that are completely conserved among flies that use sc as a sex signal but differ from the corresponding residues in Sc from flies that do not would be the best XSE-specific candidates. If residues satisfying this criterion were involved in recruitment, we would expect their frequency to be higher in sc than in the paralogous ASC genes ac and l'sc for which no comparable role in sex determination is expected. Although there is relatively little direct information on which species use sc as part of their sex signal beyond D. melanogaster and D. virilis (MEGNA 2001 ; DORSETT et al. 2003 ), it seems reasonable to assume that sc will be used only as a sex signal in species that display sex-specific regulation of Sxl. Sex-specific Sxl regulation has been shown for species throughout the genus Drosophila (BOPP et al. 1996 ; DINES 2001 ) and more recently for S. lebanonensis (DINES 2001 ), a member of another genus in the family Drosophilidae. In contrast, a medfly (Ceratitis capitata) from the related family Tephritidae and a housefly (M. domestica) and blowfly (Calliphora vicina), both from the related section Calyptratae, do not use an X chromosome dose sex-determination signal and do not show sex-specific regulation of Sxl (reviewed in SACCONE et al. 2002 ; SHEARMAN 2002 ). Thus, the residues we seek to identify are those that are conserved within the family Drosophilidae but differ from the corresponding residues in other species within the division Schizophora. The relationships among the species compared are illustrated at the bottom of Fig 6.

View larger version (68K): [in this window] [in a new window] [Download PPT slide]   Figure 6. A comparison of the predicted protein sequence for Sc, Ac, and L'Sc from several Drosophila species and other higher Diptera. In the "species compared" list for each protein, names in black indicate sequences new to this study, while names in gray indicate sequences previously published (referenced in text). Sequences for species in parentheses are incomplete (virilis Ac is missing 40% from the N terminus, while capitata L'Sc lacks 30% at the N terminus and 21 residues from the C terminus). The alignments on which this figure is based are provided as supplemental Figures S1–S4 available at http://www.genetics.org/supplemental/. Conserved regions of homology common to all three paralogs are boxed, and the basic-Helix-loop-Helix transcription factor motif is shaded. Short solid boxes identify residues that are identical for all orthologs examined. Tall solid boxes identify "signature" residues that not only are invariant among all orthologs, but also are different from the corresponding residues in the other two ASC paralogs. To minimize ambiguity in identifying corresponding positions among paralogs, residues were considered signature only if they were adjacent to at least one other completely conserved residue. Tall shaded boxes indicate residues that are conserved within all species of the family Drosophilidae examined, but are different from all those outside of this family. The number of residues in each conservation class found distal to the actual or analogous position of the scsisB3 lesion is given for each protein in the boxed area at the lower right of each section. Phylogenetic relationships among these fly species are shown schematically at the bottom of the figure, with branch lengths drawn proportional to the estimated time of divergence (BEVERLEY and WILSON 1984 ; KWIATOWSKI et al. 1994 ; POWELL and DESALLE 1995 ; RUSSO et al. 1995 ; NIRMALA et al. 2001 ). Dro.mel., D. melanogaster; Dro.sim., D. simulans; D.yak., D. yakuba; D.sub., D. subobscura; D.psu., D. pseudoobscura; D.vir., D. virilus; Sca.leb., S. lebanonensis; Cer.cap., C. capitata; Cal.vic., C. vicina; Mus.dom., M. domestica.

Using lack of conservation of residues among sc paralogs to argue for the significance of any Drosophilidae-specific pattern of conservation in sc would be reasonable only if the duplication event that gave rise to the divergence between ac and sc occurred before sc was recruited as a sex signal. Prior to our study, ac, the closest relative of sc, had not been found outside of the Drosophila genus (SKAER et al. 2002 ), raising the possibility that the duplication event that generated ac and sc might have been contemporaneous with or after the acquisition by sc of a role in sex determination. However, our recovery of ac from the Calyptrate species M. domestica, (see below) argues that this duplication event occurred long before sc became a sex-signal gene.

For the purposes of ASC protein comparisons, we sequenced ac, sc, and l'sc protein-coding regions from S. lebanonensis and M. domestica. We also sequenced sc and ac from D. virilis, but could not recover l'sc from that species. We drew on information already available for sc from D. simulans, D. yakuba, and D. subobscura (TAKANO 1998 ); for sc and l'sc from C. capitata (WULBECK and SIMPSON 2000 ) and C. vicina (PISTILLO et al. 2002 ); and for ac, sc, and l'sc from D. pseudoobscura (ftp://ftp.hgsc.bcm.tmc.edu/pub/data/Dpseudoobscura/fasta/). The lower dipteran Anopheles gambiae, a mosquito, has only a single gene corresponding to sc, ac, and l'sc and hence is not relevant to this analysis (WULBECK and SIMPSON 2002 ).

Fig 6 presents a summary of the pattern of conservation among Sc protein sequences from the seven species within the family Drosophilidae and the three species outside it mentioned above. A similar comparison was made for Ac and L'sc with somewhat fewer species. Information is presented in the context of the D. melanogaster sequences. Surprisingly, the region of Sc past the position of the change in sc^sisB3, whose deletion has so little effect on neurogenesis, includes two regions of homology (boxed in Fig 6) that are apparent in all three paralogs, both of which contain "signature" residues that distinguish Sc from its paralogs in all species examined, regardless of whether they are likely to use sc as a sex signal. As many signature residues are in the region eliminated by sc^sisB3 as are in the highly conserved basic-Helix-Loop-Helix domain (shaded box in Fig 6), which is the functional heart of all three transcription factors. Although the region eliminated by sc^sisB3 includes 11 candidate sex-determination-specific residues, 6% of the total in this region, the corresponding region of L'sc, which is considerably shorter, has nearly twice the frequency of such residues—11% (12/106)—despite having no apparent involvement in sex determination. The corresponding amount for Ac is also 11% (12/105), but this is uncorrected for the fact that only one sequence outside the Drosophilidae was available for comparison. If only this one "outside" species had been available for Sc and L'sc, the frequencies of sex-determination residues for them would have been 14 and 21%, respectively. Thus, with respect to the number of residues with characteristics one might expect for involvement in sex determination, Ac has the same frequency as Sc, and L'sc has significantly more, which is not the expected result. The extensive comparisons presented here show that Sc, L'sc, and Ac proteins acquired unique identities long before sc was recruited as a sex signal, but the effects, if any, on protein sequence of sc's recruitment to the sex-determination hierarchy are not apparent.

Musca Sc, and to a much lesser extent, D. melanogaster Ac, can substitute for D. melanogaster Sc in sex determination:
The protein sequence comparisons described above suggest that if recruitment of sc to the sex-determination pathway involved changes in the sc gene, those changes are likely to be in cis-acting regulatory sequences rather than in protein-coding sequences. By this hypothesis, Musca Sc protein should be able to substitute for melanogaster Sc in regulating Sxl. But to infer the significance of a positive result in such an experiment, one would also need to know whether Drosophila ASC paralogs can substitute for Sc.

We answered these questions using a melanogaster transgene cassette in which essentially all the melanogaster protein-coding region on a 9.1-kb Pst-Sal genomic fragment containing sc could be replaced easily by any coding region of choice as a NotI-AflII fragment. This allowed us to express any protein under the control of wild-type melanogaster sc regulatory sequences that we knew were sufficient to provide full XSE activity (see MATERIALS AND METHODS). The transgenes' XSE activity was assayed by their ability to suppress the female-specific lethality of sc^M6, the putative null allele. Because the specific level of sc^M6 female viability measured in any cross where rescue is incomplete is extremely sensitive to genetic background, we normalized the genetic backgrounds prior to all assays by backcrossing transgenic animals to animals from a reference balanced sc^M6 stock for many generations. Only in this way could meaningful comparisons be made.

We first had to determine whether the altered Sc signature residue in the sisB protein-coding exchange cassette (see MATERIALS AND METHODS) had any adverse effect on sisB⁺ activity. The top two rows of Fig 7 show that the behavior of the cassette transgene was indistinguishable from that of the parental transgene: each raised the viability of females relative to their sc⁺ sisters to an average of 100% while, in the absence of either transgene, female viability averaged well below 1%.

View larger version (48K): [in this window] [in a new window] [Download PPT slide]   Figure 7. Effects of replacing the protein-coding region of a sc transgene that has normal XSE function but no proneural function on its ability to rescue scM6 mutant females—a measure of XSE activity. Viability of scM6/scM6 females was measured relative to that of their scM6/Binsinscy sisters (n > 200) from crosses of the form scM6/Binsinscy x scM6/Y; P{sisB+w+mC}/+ at 25°. NcoI and AflII restriction sites were introduced into a 9.1-kb D. melanogaster sc fragment present in the scsisB+ transgene to generate a protein exchange cassette used to assay the ability of the coding regions from M. domestica sc and D. melanogaster ac to function in sex determination when expressed under the control of melanogaster scsisB+ regulatory information.

Despite differing from its D. melanogaster ortholog at 52% of the Drosophila residues, Sc protein from the housefly was nearly as effective as the endogenous melanogaster protein at rescuing females: relative viability of transgenic sc^M6 females averaged 90% (Fig 7, third row), and all nine lines rescued more than half the mutant females. In contrast, Ac protein from melanogaster was far less effective: relative viability averaged only 9%, and two of the eight lines exhibited no rescue. On the other hand, the fact that females were rescued to some extent by six of the eight Ac-expressing lines, with one line even achieving 46% rescue, shows that other proteins of the ASC can substitute for Sc to some degree when expressed at levels and at times that are normal for sc. Taken together, the results from Fig 7 show that the protein sequence changes that distinguish Sc from Ac are far more relevant to Sxl regulation than those that distinguish Musca Sc from melanogaster Sc. This observation is strong evidence that the acquisition by sc of a role in sex determination did not involve significant changes in Sc protein sequence.

sc^sisB2 identifies a regulatory region downstream of the sc transcription unit that is involved in sex determination:
Having found no evidence that changes in Sc protein were relevant to this gene's acquisition of a sex-determination role, we turned our attention to cis-acting regulatory sequences. The deletion associated with sc^sisB2 indicated that such sequences might lie beyond the 3' end of the transcription unit. As shown in Fig 3, this deletion eliminates a region of sc within the smallest transgene known (ERICKSON and CLINE 1991 ) to provide full XSE⁺ activity. The deleted region includes two notable stretches of DNA sequence identity between melanogaster and virilis (Fig 3), including one run of 72 identities out of 77 bp. To confirm the significance of this 3' region for sex determination, we eliminated it from the control 9.1-kb transgene and assayed for sc^M6 rescue. As Fig 8 illustrates, loss of this region reduced the transgene's ability to rescue mutant females by over two-thirds. The fact that the viability of sc^M6 mutant females carrying this truncated transgene was higher for some lines than was the viability of homozygous mutant sc^sisB2 females (Table 2) may reflect genetic background differences or insertion-site position effects or instead may indicate that sc^M6 is not a true null.

View larger version (43K): [in this window] [in a new window] [Download PPT slide]   Figure 8. Effects of mutations in flanking untranscribed regions on the ability of a scsisB+ transgene to rescue scM6 mutant females. Viability was measured as in Fig 7, and the wild-type control is from that figure.

The heptamer CAGGTAG is involved in sex determination:
The discovery that all D. melanogaster XSEs, except the somewhat atypical element runt, have the sequence CAGGTAG or its complement repeated three times within 500 bp upstream of their transcription start sites suggested that this sequence might be important for driving the extremely early expression specifically required for XSE function (ERICKSON and CLINE 1998 ; our earlier report that sc had only two such heptamers was due to an error in the published sequence). We subsequently found that clusters of three such heptamers within 500 bp occur only 26 times in the sequenced genome of the fly—24 times if one merges the three overlapping clusters at sisA, the only supercluster. Only for bottleneck (bnk) and Neurotactin (Nrt) are the clusters within 500 bp of a known transcription start site that does not belong to an XSE (J. TENBOSCH, unpublished results). Although bnk and Nrt are autosomal and do not appear to be involved in sex determination, they share with the XSE genes the feature of being expressed prior to the onset of general transcription in nuclear cycle 14 (SCHEJTER and WIESCHAUS 1993 ; LECUIT and WIESCHAUS 2000 ) and hence support our hypothesis for the role of this heptamer. Even the XSE runt has a cluster of two CAGGTAG heptamers and one degenerate (CAGGCAG), but the cluster is unlike those for other XSEs in being farther (3 kb) upstream of the transcription start site (J. TENBOSCH, unpublished results). To test the functional significance of the CAGGTAG cluster upstream of sc, we mutated all three copies in a sisB transgene and assayed for rescue of sc^M6 females. Fig 8 shows that elimination of this cluster did not abolish rescue, but it did have an effect comparable to that of the sc^sisB2 deletion.

Although neither the sc^sisB2 deletion nor the triple-heptamer knockout eliminated sc sex-determination activity by themselves, data in Fig 8 show that the combination of the two alterations on opposite sides of the transcription unit did: not a single line among 17 independent transgene inserts carrying both mutations could rescue sc^M6 females. On the other hand, when we remobilized one of these nonrescuing mutant transgenes and selected for rescue of sc^M6 females, at a low frequency we could recover reinsertions that had regained full rescuing activity, presumably by coming under the influence of those rare enhancers able to drive very early gene expression (data not shown). This resurrection of XSE activity established that lack of rescue in the 17 lines of Fig 8 was not an artifact of transgene misconstruction and shows that this mutant transgene can be used as an enhancer trap to identify cis-acting regulatory information capable of driving or boosting preblastoderm gene expression—enhancers perhaps belonging to genes whose extremely early zygotic expression has been masked by maternal transcripts.

A test for lethal dominant interactions does not support the proposal of a third general role for sc in early development:
RNA in situ hybridization studies suggested that sc may work with twist and snail to generate mesoderm and neurectoderm in the very young embryo (GONZALEZ-CRESPO and LEVINE 1993 ). Although this function had not been discussed in reviews or confirmed by others, we were intrigued by the possibility that the earlier timing and broader spatial extent of this activity would make it a more plausible evolutionary antecedent of XSE functioning than sc proneural function (YANG et al. 2001 ). We wondered if our sc⁺ transgenes would provide this early vital function, and if so, whether the mutations that seemed to specifically disrupt XSE function would also disrupt this function.

This idea that sc might be involved in mesoderm formation arose from the observation of a triple dominant interaction between loss-of-function mutations in the genes sc, twist (twi), and maternal dorsal (dl), which together, and only together, appeared to block ventral furrow formation and prevent the establishment of mesoderm. Conclusions were clouded somewhat by the fact that the null allele of sc used was on a chromosome that was also somewhat impaired for ac; however, the fact that the effect of this chromosome was indistinguishable from that of a chromosome deleted for all members of the ASC pointed to a strong involvement of sc. Although the 1993 study was based on the earlier observation by