Our earlier data placed the swi locus in region 2B6/B7-8 of the X chromosome, between the proximal break points of Df(1)S39 and Dp(1;Y)y²67g24.2Y 7. This region corresponds to 1421800–1466900 bp of the Release 3.1 X chromosome sequence 101112. This interval is overlapped by the EDGP cosmids 73D1 and 9D2 (Fig. 1). Saturation mutagenesis of this region revealed only two loci 7deep orange (dor) (see 13) and singed wings (swi). Subsequently, Makunin and co-authors 14 identified two further genes (b6 and a6) (Fig. 1) and the computational analysis of the genomic sequence identified a further six 12 (Fig. 1). cDNA sequences are available for all of the genes predicted by sequence analysis, with the exception of CG14797 15.

In order to narrow the list of possible candidate loci for swi a genomic library made from the DNA of swi⁺ flies in the CosPeR cosmid vector (see Methods) was screened using the DNA of the clone λ1487 as a probe. A 30 Kb clone (5581) was recovered and mapped to the genome (Fig. 1). Two lines independently transformed by this cosmid (5581.118 and 5581.4), each with a single independent autosomal insertion, were obtained and tested for complementation of swi^t476. Neither of them was able to complement swi^t476, implying that a6, b6, CG14798 and EG:9D2.4 are not swi.

To choose between remaining four loci we made a further set of transposons (Fig 1). SWI1 contained the very 3'-end of dor, a complete EG:171E4.4 gene and the 5'-half of EG:171E4.2. SWI3 overlapped the 3'-end of dor, and the full sequences of EG:171E4.4 and EG:171E4.2 while SWI4 included only EG:171E4.2. Five independent transformed lines with insertions of SWI1, seven lines containing insertions of SWI3 and sixteen lines with insertions of SWI4 were established. All insertions of SWI1 were autosomal; none complemented the swi^t476 mutation. On the contrary, all six autosomal insertions of SWI3 (Table 1) and all 16 insertions of SWI4 completely rescued this mutation (Table 2). These results unambiguously indicate that the swi gene is identical to EG:171E4.2.

To confirm this conclusion we had isolated the DNA from homozygous y¹swi^t476w^pn/y¹swi^t47w^pn third instar larvae and sequenced the region spanning from bases 1425864 to 1431743 (Release 3 coordinates). There are nineteen differences between this sequence (deposited in EMBL-Bank as AJ626646) and the reference sequence (AE003421). Only one of these differences affects the predicted protein product of EG:171E4.2, a G to A transition at base 1428762, that would change Cys379 to Tyr379.

The transformation rescue of the mutant swi allele by a construct that includes only EG:171E4.2, and the fact that the mutant allele carries a very non-conservative amino acid substitution prove that swi corresponds to this predicted gene.

The 4.9 kb HpaI-HpaI fragment from SWI4 (Fig. 1) is capable of fully rescuing the swi^t476 phenotype; this fragment, therefore, must include not only the coding sequence of swi, but also all of its necessary regulatory sequences. The swi (EG:171E4.2) mRNA as judged from the longest found cDNA sequence (RE03173) spans the region from position 1426365 to 1429883. We suspect that the true length of the transcript is slightly greater, since neither a consensus Drosophila Initiator (INR) element nor any known Drosophila promoter element could be found in the immediate vicinity of RE03173 5'-end. At the same time there is a perfect TATA box consensus (TATAAA), at position 1426228 (seen with PromH, 16) and CA and T nucleotides at positions 30 and 36 downstream of this, characteristic of INR elements 17 (Additional file 1). The latter suggests that swi may have a strong TATA promoter, initiating 107 bp upstream of RE03173 and that is consistent with the fact that only a 150 bp upstream region is required for swi transgene function.

We have performed BLAST search of the Drosophila pseudoobscura genomic sequence database with the sequence of the 4.9 kb HpaI-HpaI fragment. This identified a region (coordinates 5225–8898) of contig 2140 that aligns with the D. melanogaster sequence from coordinate 1426109 to 1429698 and a short region of contig 2139 complementary sequence that aligns with the very 3' end of D. melanogaster HpaI-HpaI fragment (Additional file 1). The alignment of D. melanogaster sequence with that of contig 2140 corresponds to almost the entire swi locus with only last 185-bp of the 3'-UTR missing. The degree of conservation between these sequences is not uniform: there is high nucleotide conservation (78%) in the central coding region of swi (coordinates 1428116 – 1429529) (Additional file 1A), but less (44%) elsewhere (Additional file 1). In addition, there are short highly conserved regions within the first intron (coordinates 1427144 – 1427515) (Additional file 1). This region includes a number of short, but highly conserved, motifs, often AT rich and resembling the consensus homeobox binding consensus sequence (NNATTA) 18. These motifs may well have a regulatory function.

Translation of the swi open reading frame (ORF) predicts a 611 amino acid protein with threonine and histidine rich N-terminal domains, two leucine rich repeats and a cysteine rich C-terminus (Fig. 2A). The predicted amino acid sequence of the Dpse\SWI protein has 71% identity and 73% similarity to that of its D. melanogaster homolog.

The predicted SWI protein was used to search public protein databases for known or predicted proteins of similar sequence. The sequences of the protein of the predicted CG14485 (TrEMBL:Q8SXT3) gene (e = 8e-24) and of the predicted Anopheles gambiae protein (agCP6178; protein_id:EAA11923.1) (e = 1e-14) showed significant similarity to that of SWI. It is also important to note that this proteins appeared to be more similar to each other than to SWI itself suggesting Q8SXT3 to be designated as SWI2, the second member of Drosophila SWI family of proteins, and EAA11923.1 as its Anopheles gambiae simalog Agam\SWI2.

The second large group of sequences showing much lower, but evident, similarity to SWI (e ~1e-07) consists of virtually all known homologs of decorin and biglycan proteins. These proteins are characteristic members of the Class I small leucine-reach proteoglycan protein family. These proteins possess a protein core substituted with a single glycosaminoglycan chain near their N-terminus 19. The most salient feature of decorins and biglycans is the presence of 10 leucine-reach repeats flanked by cysteine-rich regions 20. Despite the relatively small size of the protein core (~36 kDa), these proteins possess several distinct protein binding activities. Originally, they were shown to bind collagen fibers, playing a role in their assembly. More recently decorin and biglycan were also discovered to interact with the Transforming Growth Factor-β and Epidermal Growth Factor (EGF) receptors, participating in cellular proliferation control 19.

The third group of sequences similar to SWI contains two members of Toll receptor superfamily: Toll related protein Ae\Toll1B from Aedes aegypti (P_rm = 7e-04) and Toll-4 (P_rm = 9e-04) from Drosophila melanogaster. Both of these were discovered in course of the systematic characterization of Toll-like proteins 2122. Although some members of Toll protein family are transmembrane receptors, with a well established role in innate host defense, no evidence is available for the function of either Ae\Toll1B or Toll-4 21.

We have made all possible pairwise combinations and multiple alignments of these proteins. As expected, the best multiple alignment is between the SWI, Dpse\SWI, SWI2 and Agam\ SWI2 protein sequences. (Fig. 2B). Examination of this alignment shows that the C-terminal 211 amino acids of SWI are conserved between four proteins (20% identity; 52% similarity) with much higher homology between SWI and Dpse\SWI (90% identity). This alignment also clearly shows that Agam\SWI2 is more similar to SWI2 than to SWI.

The region of significant similarity between SWI and the decorin/biglycan family proteins is narrower, it is limited to 90 amino acids between positions 396–485 of SWI. In decorins this region is N-terminal (residues 54–141 of murine decorin). This region contains the very end of N-terminal cysteine-rich flanking region and the first two leucine-rich repeats 20. The region of homology between the proteins does not correspond to the leucine-rich repeats of SWI. The two stretches of amino acids especially conserved between SWI and decorins are also well conserved between SWI, SWI2, Agam\SWI2 and Dpse\SWI (Fig. 2B). Interestingly enough, such conserved regions have no documented role in known decorin functions, implying that decorins might have some yet undiscovered properties that these proteins have in common with SWI.

The region of homology between the SWI and Ae\Toll1B or Toll-4 proteins is situated between positions 249–580 of the SWI amino acid sequence. This corresponds to positions 307–626 of Toll-4 and includes one of its N-terminal cysteine-rich flanking region and two leucine-rich repeats; this region of similarity does not overlap the Toll/IL-1R (TIR) domain of Toll4. This cytoplasmic homology domain is a key feature common of all Toll-like proteins and is indispensable for the activation of the antifungal response in flies by Toll and 18-wheeler (Toll-2) 23. Comparison of SWI and Ae\Toll1B or Toll-4 sequences revealed two distinct conserved CysXCys motifs (Fig. 2B). Both of them are also conserved in SWI2, Agam\SWI2 and Dpse\SWI. The swi^t476 mutation affects one of the cysteines in the first CysXCys motif. These CysXCys motifs may be part of a yet uncharacterized functional protein domain (or two distinct but similar protein domains) indispensable for SWI activity and, perhaps, also important for Toll-4 function.

We have used a number of prediction programs in an attempt to learn more about the possible cellular location of the swi protein. von Heijne's algorithm 24, as implemented by PSORTII, indicates that the protein may have a cleavable signal sequence (residues 1–19) (a prediction confirmed by the TargetP Server v1.01 25. This is consistent with predictions from the Pastuer Institute's implementation of Claros and von Heijne's TopPred algorithm, which indicate a predominantly extracellular protein anchored by cytoplasmic N- and C-termini.

As an alternative approach we have searched the database of Drosophila protein interactions recently built on the basis of genome-wide yeast two-hybrid screening 26. SWI was found to have only one interacting partner, the product of CG9025 gene. This interaction seems to be very specific and has a very high confidence score (0.9085 with 1.0 being the maximum). CG9025 is an uncharacterized protein similar to mammalian sex-determination protein encoded by Fem1b gene. CG9025 in turn has eight high confidence interactors from which all with known cellular localization are cytoplasmic. It is interesting to note that one of CG9025 interacting proteins, the product of TBPH (CG10327) has a putative RNA binding domain.

The phage genomic clones overlapping the 2B6/B7-8 region have been described by 10. The coordinates of the FlyBase Drosophila genome annotation (Release 3.1) are used in this paper. The cosmid 73D1 was provided by F. Galibert. The genomic library made from DNA of isogenic y¹; cn¹bw¹sp¹ flies in the CosPeR cosmid vector, the generous gift of J. Tamkun. CosPeR contains P-element termini and the white gene as a marker, which allows the direct use of a clone for Drosophila transformation. The stock y¹swi^t476w^pn/FM6B, carrying the lethal allele of singed wings, has been described earlier 7. Lindsley and Zimm 27 designated an EMS-induced mutation swi^t467 as hfw⁷. However, complementation tests of the hfw¹ allele described by Rayle 9 with any of the swi alleles was never done, since the hfw¹ stock had been lost by the time any swi mutant allele was obtained. We use the swi^t467 allele from the Novosibirsk stock collection. The y¹swi^t476w^pn mutant chromosome was also rebalanced over FM7, y^93jsc⁸w oc pt B, P{w^+mC, act::GFP=pActGFP}, a "green" balancer (the generous gift of J.-M. Reichhart) to make the y¹swi^t476w^pn/FM7, y^93jsc⁸w oc ptg B, P{w^+mC, act::GFP=pActGFP} stock. This strain allows the easy selection of y¹swi^t476w^pn/y¹swi^t476w^pn larvae as non-GFP animals.

All general molecular methods were after Sambrook and co-authors 28. Plasmid and cosmid DNA or PCR products were directly sequenced using the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction kit (Perkin Elmer). To amplify fragments of genomic DNA isolated from y¹swi^t476w^pn/y¹swi^t476w^pn flies the following pairs of primers were used:

Swi-m1 5'-CCGTGCGGCAGTATGACG-3', Swi-m2 5'-TCTGCAGGCTCGTTCGGT-3';

Swi-m3 5'-CAGCGAGTGCTCTGCTGC-3', Swi-m4 5'-GGCTTCGGCATGGAGTGG-3';

Swi-m5 5'-CCTCGTATCCTCACACGC-3', Swi-m6 5'-CTTCCTTCAGCGGCGCAT-3';

Swi-m7 5'-CAGGGCACATGGCTCCTG-3', Swi-m8 5'-GGCGTTCACCAGTCGAGT-3';

Swi-m9 5'-TGCCAGTGCAGGGAGCAC-3', Swi-m10 5'-TGGGGTAGCTCCTCCAGG-3';

Swi-m11 5'-CGAACTACGGCAGCTGCA-3', Swi-m12 5'-AGGGCAGATAGCCAGCCG-3';

Mut1 5'-GAACAGCCCGAGTATCTG-3', Mut2 5'-GTCGTAGGACACCTTGGT-3';

Mut3 5'-TACCTGGCAGGCAACAAG-3', Mut4 5'-CAACAGTGAGCCGTATCC-3';

Mut5 5'-TCGAGCGCGCACCTAAGT-3', Mut6 5'-ATCCAGAGGCCAGGACCT-3'.

The raw sequence data were analyzed using BLAST 29 and ClustalW 30 computer programs. Screening of Drosophila pseudoobscura genome sequence for homology to the swi sequence was done with BLAST at http://www.hgsc.bcm.tmc.edu/projects/drosophila/. To look for promoter elements, genomic DNA sequence corresponding to the region of the putative transcription start, i.e. the 5' end of the BDGP cDNA clone RE03173 (accession number AY070931) was visually inspected for sequences corresponding to the TATA, INR and DPE consensus sequences, as described by Kutach and Kadonaga 17. In addition, the entire sequence of swi locus was analyzed with PromH program 16. Unless specially stated the default parameters of all software programs were employed. To browse D. melanogaster genome annotations the Apollo (v1.3.4) software was used 31. The parameters of protein primary structure were calculated using the ProtParam tool at 32 and a protein motif search was performed with InterProScan software 33. To predict protein localization the following programs were used. PSORTII at 34, the TargetP Server v1.01 25 at 35 and the Pastuer Institute's implementation of Claros and von Heijne's TopPred algorithm 36. The Drosophila protein yeast two-hybrid interaction network 26 was searched at 37.

To make the SWI1 construct the 4.4 kb BamHI-BamHI fragment from the λ1479 clone (Fig. 1) was inserted into the CaSpeR3 vector 38 digested with BamHI. The SWI3 transposon was constructed as follows: The 9.4 kb XbaI-EagI fragment from the cosmid 73D1 clone was ligated into pBluescript KS (Stratagene), digested with XbaI and EagI, to give 73D1-XE-pBl-KS. The 73D1-XE-pBl-KS construct was further digested with XbaI and NheI and the resulting 8.5 kb XbaI-NheI fragment was directly inserted into the CaSpeR2 vector (Pirrotta, 1988), digested with XbaI. The SWI4 transposon was constructed in similar way: The 73D1-XE-pBl-KS construct was cut with HpaI and the resulting 4.9 kb fragment was inserted into the CaSpeR4 vector 38, digested with the same restriction enzyme. All constructs were injected into the host strain y¹, Df(1)w^67c23 as described by Spradling 39.

To test complementation of the swi^t476 mutation by various transformed fragments of genomic DNA y¹swi^t476w^pn/FM6B females were crossed to transformed y¹, Df(1)w^67c23/Y males carrying an autosomal insertion of the genomic fragment under study. The rescue of the swi lethal phenotype was judged by the appearance of yellow males in the progeny.