Generation of New Hairless Alleles by Genomic Engineering at the Hairless Locus in Drosophila melanogaster

Hairless (H) is the major antagonist within the Notch signalling pathway of Drosophila melanogaster. By binding to Suppressor of Hairless [Su(H)] and two co-repressors, H induces silencing of Notch target genes in the absence of Notch signals. We have applied genomic engineering to create several new H alleles. To this end the endogenous H locus was replaced with an attP site by homologous recombination, serving as a landing platform for subsequent site directed integration of different H constructs. This way we generated a complete H knock out allele H attP, reintroduced a wild type H genomic and a cDNA-construct (H gwt, H cwt) as well as two constructs encoding H proteins defective of Su(H) binding (H LD, H iD). Phenotypes regarding viability, bristle and wing development were recorded, and the expression of Notch target genes wingless and cut was analysed in mutant wing discs or in mutant cell clones. Moreover, genetic interactions with Notch (N 5419) and Delta (Dl B2) mutants were addressed. Overall, phenotypes were largely as expected: both H LD and H iD were similar to the H attP null allele, indicating that most of H activity requires the binding of Su(H). Both rescue constructs H gwt and H cwt were homozygous viable without phenotype. Unexpectedly, the hemizygous condition uncovered that they were not identical to the wild type allele: notably H cwt showed a markedly reduced activity, suggesting the presence of as yet unidentified regulatory or stabilizing elements in untranslated regions of the H gene. Interestingly, H gwt homozygous cells expressed higher levels of H protein, perhaps unravelling gene-by-environment interactions.


Introduction
Communication amongst cell neighbours is made possible by the Notch signalling pathway, driving cell specification and differentiation [1][2][3]. In Drosophila, Notch signalling is activated by the binding of one of the membrane-tethered ligands Delta (Dl) or Serrate (Ser), present on one cell, to the Notch receptor present on the neighbouring cell. This interaction results in the cleavage of the Notch receptor and the release of its intracellular domain (ICN). ICN mediates the formation of an activator complex together with the DNA-binding protein Suppressor of Hairless [Su(H)] and the co-activator Mastermind (Mam), entailing the transcriptional activation of Notch target genes [4][5][6]. In the absence of Notch signalling activity, Notch target genes are silenced by a repressor complex consisting of Su(H) bound to Hairless (H) and the two corepressors Groucho (Gro) and C-terminal binding protein (CtBP) [7][8][9][10].
Hairless (H) is the major antagonist of the Notch signalling pathway in Drosophila, negatively regulating the majority of Notch mediated events during imaginal development (overview in [11]). The role of H has been investigated in the past with classical mutations and with the overexpression of wild type and mutant H constructs. The latter has been used for a systematic structure-function analysis of the H protein and allowed the localization of the Su(H)-, the Groucho-and the CtBP-binding domains, respectively (SBD, GBD, CBD) [7,8,10,[12][13][14][15][16][17][18][19]. Moreover, it was shown that the mutation of a single residue within the SBD (Leucine 235 to Aspartate) completely abolished the H-Su(H) interaction [15]. Tissue specific overexpression of H results in a plethora of phenotypes as a consequence of the downregulation of Notch activity [7,10,[12][13][14][15][18][19][20][21]. Albeit very helpful for a detailed analysis of specific processes, the overexpression of H is quite dramatic and does not compare to an analysis of mutants expressed under wild type regulation. Therefore, we made use of the recently established method of genomic engineering in Drosophila [22] to replace the endogenous H locus with an attP site by homologous recombination. The resultant H attP founder line was genetically and molecularly verified as a complete H knock out allele. It further served as landing platform for the subsequent site-directed integration of various H constructs, two wild type and two mutants affecting Su(H) binding. A detailed characterization of these new H alleles is presented, furthering our insight into the structure and function of the H gene.

Generation of the Hairless knock out line
Genomic engineering according to Huang et al. [23] was employed to generate the Hairless knock out founder line H attP as outlined in Fig 1. To this end, genomic DNA fragments flanking the Hairless locus were cloned into pGX-attP and the resultant construct pGX-H was introduced by P-element mediated transformation into the fly genome (Fig 1A and 1B). Homologous recombination [24] generated the Hairless founder line H attP w+ where a white + gene replaced the Hairless (H) gene ( Fig 1B). Subsequent elimination of the white + marker by Cre-lox mediated recombination resulted in the final Hairless knock out line H attP containing the attP site and one loxP site in place of the original H locus (Fig 1B). The genotype was confirmed by PCR and sequence analysis ( Fig 1C).
As H attP carries a complete deletion of the H coding sequence it is expected to be a complete null allele. Accordingly, H attP is a recessive lethal and displays the typical haplo-insufficient H phenotypes, i.e. loss of macro-and microchaetae, frequently accompanied by a transformation of bristle shaft into socket, causing a double-socket phenotype (Fig 1D) [25][26][27].
Rescue of the H attP knock out line with wild type H DNA We took advantage of the attP site within the Hairless locus to re-integrate wild type forms of the H gene with the help of PhiC31-integrase as outlined before in [28] (S1 Fig). We used genomic as well as H cDNA to generate H gwt and H cwt alleles, respectively (Fig 2A). Due to the construction, the attR side resides within the 5', and the loxP site within the 3' untranslated region (UTR) of the H gene. None of them affects known regulatory motifs like e.g. polyadenylation sites, the described binding site of miR-305 [29] or of other potential micro-RNA targets (Fig  2A and S1 Fig). Both genomic and cDNA were apparently fully functional: H gwt and H cwt homozygous stocks were viable and fertile and phenotypically indistinguishable from wild type    [15]. Conserved amino acids are highlighted in blue or red; red marks the mutated positions. Protein binding of H-NTCT wild type and mutant constructs (in pEG) to full length Su(H) (in pJG) was probed in a yeast two-hybrid assay. Empty vectors served as control. H-NTCT comprises amino acids 171-357 and overlaps the Su(H) binding domain [15,30]. In NTCT-L235D, Leucine at position 235 and in NTCT-I244D, Isoleucine at position 244 was each changed to Aspartate. The L235D mutation was shown before to destroy Su(H) binding [15]. Likewise does the I244D mutation.
indicating that both alleles allow normal fly development (Fig 3 and S2 Fig). As expected, the heterozygotes were indistinguishable from wild type (Fig 3). Moreover, the hemizygotes and the balanced heterozygous siblings hatched at similar rates (H gwt /H attP 97.4% of H gwt /TM6B [n = 995], and H cwt /H attP 104.4% of H cwt /TM6B [n = 967]). The hemizygous condition, however, uncovered a subtle to weak loss of activity of both rescue alleles: H gwt /H attP flies developed statistically 20% less bristles on head and thorax compared to +/H attP flies (Fig 3 and S2 Fig), and H cwt /H attP flies had just 50% of the bristles seen in the hemizygotes (Fig 3 and S2 Fig). As bristle development in H mutants is highly susceptible to genetic background (Nash 1969), the difference between wild type and H gwt may be negligible. Compared to H gwt , however, H cwt apparently had reduced H activity, given that they were induced in the same genetic background (Fig 3 and S2 Fig). The only difference between H gwt and H cwt are the introns and for technical reasons, a small sequence duplication in the 3'UTR, raising the possibility of regulatory or stabilizing elements residing within these sequences (Fig 2A and S1 Fig).

H mutants defective of Su(H) binding
The Su(H) binding domain of H contains two highly conserved boxes, the NT-and the CT-Box [30]. The contact to Su(H), however, requires only the NT-Box [15,30]. In a yeast twohybrid assay using the overlapping NTCT construct, Leucine at position 235 was shown to be critical for the Su(H) contacts, since its mutation to Aspartate destroyed the H-Su(H) binding [15] (L235D; see also Fig 2B). A second point mutation was introduced in NTCT, replacing Isoleucine at position 244 with Aspartate (I244D; Fig 2A). A yeast two-hybrid experiment confirmed that the I244D mutation suppressed the binding of the H-NTCT domain to the full length Su(H) protein ( Fig 2B) [15]. The mutants were introduced in the H cDNA and inserted into the H locus using the H attP founder as outlined above (S1 Fig and Fig 2A). The resultant mutant alleles, H LD and H iD , were recessive lethal and lethal in trans over H attP as well as over the defined alleles H 2 and H P8 [27,31]. Heterozygotes displayed the typical bristle loss (Fig 4A). In the hemizygous trans-combination over H attP pharate adults were obtained on rare occasions that lacked bristles altogether ( Fig 4B). Overall, the phenotypes were very similar to the H attP and H 2 alleles [25] (Fig 4A and 4B), indicating that both H LD and H iD had lost most of H activity.

Growth defects and protein expression in the Hairless mutants
Both H LD and H iD were designed as single amino acid substitutions, which are expected to produce normal amounts of H proteins, whereas H attP is expected to be protein null. H protein expression is notoriously difficult to detect on Western blots due to the low amount of expression [31]. We therefore performed a clonal analysis of the respective alleles using the Flp/FRT technique [32], allowing a comparison of H protein expression levels in the homozygous mutant cells with those of the heterozygous or homozygous wild type cells ( Fig 5). As expected, H protein was nearly absent in the H attP clones-residual staining may result from background or from leftover parental protein ( Fig 5). In contrast, H protein was well expressed in cell clones homozygous for either H gwt or H cwt and likewise for either H LD or H iD at levels matching those of the heterozygous or of the wild type sibling cells ( Fig 5). Unexpectedly, H protein appeared enriched in cells homozygous for H gwt (Fig 5), suggesting a stronger expression or stability of the protein despite the apparent subtle reduction in activity (Fig 3 and S2 Fig).
Loss of the Notch antagonist H results in a gain of Notch activity, effecting for example a size increase in the wing imaginal discs [20,33,34], which we observed in H attP as well as in H LD or H iD homozygous mutant animals, whereas H gwt wing discs appeared wild type ( Fig 6A). In addition Wingless protein expression, which is under the control of Notch signalling activity along the dorso-ventral boundary of the wing anlagen [34][35][36], was monitored. As expected, the Wingless stripe appeared slightly broader in the mutants compared to the rescue flies or the control (Fig 6A).
We next analysed the expression of Cut protein: normally Cut accumulates in a 2-3 cells wide strip along the dorso-ventral boundary of the wing imaginal disc in response to Notch signalling activity [20,37,38] (Fig 6B). To this end, homozygous mutant cell clones were generated and stained for Cut protein. Whereas Cut protein expression was undisturbed in homozygous H gwt or H cwt cells, the stripe appeared broader in homozygous mutant H attP cell clones touching the dorso-ventral boundary (Fig 6B). A similar broadening of Cut expression was seen in cells homozygous for either H LD or H iD , demonstrating lack of H repressor activity in the mutants despite normal H protein expression (Figs 5 and 6B).

Phenotypic classification of the Hairless mutants
The H 2 allele has been often used as a reference in the past [25,[39][40][41]. H 2 has been described as an amorphic allele with regard to bristle loss and wing venation phenotype [25,41] or as a weak allele with regard to lethality in trans over other H alleles [39]. To address pupal lethality, the mutant alleles were crossed inter se, to the null allele H attP as well as to H 2 and H P8 [39], and the weak hypomorphic H 22 allele [25], included as a reference. Pupae were counted in relation to their TM6B siblings; homozygotes were rare and not recorded ( Table 1). As observed before, the H P8 allele was strongest with regard to lethality giving only a quarter of the expected H attP / H P8 trans-heterozygotes, whereas nearly 90% of the expected H attP /H 2 pupae appeared. H 22 combinations even outnumbered the TM6B siblings, however, unexpectedly no H 22 homozygotes were obtained in our culture [25]. Both, H LD and notably H iD gave higher numbers of trans-heterozygotes than H attP in this assay, suggestive of residual H activity ( Table 1).
As second criterion for a classification of the new H alleles, the loss of macrochaetae was recorded in the heterozygous mutants (S3 Fig). As described before this phenotype is highly dependent on genetic background [41], perhaps explaining the differences between H 2 and H attP (S3 Fig). However, H LD and H iD alleles have formally the same genetic background; both Control is H gwt /+ Oregon R and H gwt /H attP . Only a few of the hemizygous H 2 , H LD and H iD animals developed to pharate adults that were dissected from the pupal case. They are completely bald, except for a few interommatidial bristles in H iD /H attP (arrowhead points to an example). Asterisk marks example of remains of the pupal envelope. (C) Trans-heterozygous combination of the various H alleles with the weak allele H 22 . Note the complete shaft to socket transformation of both macro-and microchaetae on head and thorax. Examples are highlighted by arrows. In addition, transformation of outer into inner cell fates results in a partial or complete disappearance of the bristle organs [25]; some examples are marked by arrowheads. Incomplete transformation is best seen on the head. Note that interommatidial bristles are largely unaffected. Size bars in (A-C), 100 μm.
doi:10.1371/journal.pone.0140007.g004 derived from H attP . Again, both H LD and H iD appeared somewhat weaker compared to H attP or H 2 , with H LD being more similar to the two references (S3 Fig). Finally, all alleles were crossed with the hypomorphic allele H 22 [25]. Trans-heterozygotes were viable lacking all macrochaetae and many microchaetae on head and thorax ( Fig 4C). Overall, these analyses allow arranging the alleles into the following phenotypic series: OreR % H gwt > H cwt >> H 22 >> H iD > H LD > H attP .

Genetic interactions of new Hairless alleles with Notch and Delta mutants
Genetic interactions between H and Notch-pathway mutants have been amply described [13,19,39,42,43]. Heterozygous mutants of either the Notch (N), Delta (Dl) or H locus display dominant wing phenotypes, i.e. wing margin incisions and thickened wing veins, which were name giving for the N and Dl gene, respectively, and shortened longitudinal wing veins for the H gene [25,39,40,42,44] (Fig 7). It has been observed before that a loss of one H gene copy ameliorates or even abrogates either N or Dl mutant wing phenotypes and vice versa [39,42,45]. In our culture conditions, about 91% of the amorphic N 5419 allele displayed notching of one or both wings (Fig 7A and 7A' and S4 Fig). Combined with each H loss of function allele tested here, wing notching was very rare (Fig 7D-7H and 7D'-7H' and S4 Fig). Whereas the H gwt allele was largely indistinguishable from the wild type Oregon R control, we observed a lowered frequency of notches in N 5419 /+; H cwt /+ trans-heterozygotes, supporting the earlier observation of a subtle loss of H activity in this allele (Fig 7A-7C and 7A'-7C' and S4 Fig). At the same time, the other typical phenotype of heterozygous N mutant wings, namely thickening of L3 and L5 veins, was likewise rescued by a mutation in H, no matter which allele was used, but not by the rescue alleles ( Fig 7A-7H'). Moreover, the shortening of the L5 vein typical of H mutants (Fig 7D-7G), disappeared in the trans-heterozygous combinations as well (Fig 7D'-7G').
Heterozygous Dl mutants display thickened and knotted veins that broaden distally to form deltas at the wing margin [40,42], exemplified in Fig 7A" for the amorphic allele Dl B2 . It has been shown before that the Dl/H trans-heterozygotes display a more normal wing venation pattern [42]. Rescue of the Dl wing phenotype was indeed observed in the trans-heterozygous combination of Dl B2 with any of the tested H alleles, except for the rescue constructs H cwt or H gwt as was expected (Fig 7A"-7H"). Quantitative evaluation of the differences confirmed the impression given by the pictures and the phenotypic series established above (S4 Fig). These data demonstrate that the engineered mutants H attP , H LD and H iD are indeed loss of function H alleles, albeit H LD and notably H iD apparently retain some H activity and may hence be classified as strong hypomorphic mutants.

Discussion
In this work we have applied the method of genomic engineering to generate new alleles in the H locus. Genomic engineering permits defined genetic modifications at the locus of choice to eventually study its regulation or function. The directed replacement of the locus with any desired allele is straightforward once the attP-founder line is established. The newly arising Crispr/Cas9 system is going to speed up this approach for the simplicity to replace a given locus for example with an attP landing site [46,47]. We have shown that the H attP founder line can be used to generate new wild type or mutant H alleles. Alternatively, H could be replaced by other known inhibitors of the Notch pathway to address their repressive potential during fly development, be it fly or mammalian components [11,48]. It will be interesting to see for example, whether mammalian Notch-repressors like KyoT2, whose structural interaction with the mammalian Su(H) homologue is known [49], are able to replace H function in Drosophila.
The H attP founder line deletes the entire H coding region and may serve henceforth as a reference for a H null allele. Both H LD and H iD mutants retain some H activity compared to H attP , albeit the differences are subtle. The fact that H iD appears weakest suggests some residual Su (H) binding of this mutant, which however could not be detected in the yeast two-hybrid experiments. We have to keep in mind that both mutations were generated in the H cDNA, which in its wild type form is not sufficient to provide H activity completely. Unfortunately, we Assuming that H LD and H iD fail to bind to Su(H), the residual activity of these alleles implies additional, Su(H)-independent activities of H. This is, however, not unexpected since H has been shown by genetic and molecular means to interact with other factors but Su(H), apart from the binding to its corepressors Groucho and CtBP [7,9,10,12,14,[16][17][18]29,42,[50][51][52][53][54]. For example, the binding of Pros26.4 to H results in a destabilization of H protein, thereby promoting Notch signalling activity independent of Su(H) (Müller et al. 2006). More recently it was shown that H interacts with Runt thereby participating in Runt-dependent transcriptional regulation of segmentation genes during Drosophila embryogenesis [52]. A complete loss of H is hence expected to affect fly development more profoundly than just the lack of Su(H) binding, i.e. repression of Notch signalling activity.
It came to our surprise that the H cDNA was not able to fully rescue the H mutant phenotypes. The H gene is expressed at low to moderate level in all tissues throughout development with some enrichment in the larval brain (http://www.flybase.org; http://www.flyatlas.org) [39,55]. Accordingly, a H cDNA under heat shock promoter control was able to rescue the haplo-insufficient dominant phenotype of H heterozygotes even at ambient temperatures (Bang and Posakony 1992), suggesting that a moderate H expression may suffice for full H activity. Our analysis does not support this view; instead it favours the presence of as yet unidentified regulatory elements within the H introns. In search of such regulatory elements we compared H intron sequences within the genus Drosophila, comprising species of the Sophophora and Drosophila subgenera that have been split about 40 million years ago (http:// www.flybase.org/blast). This distance is large enough to detect regulatory elements on ground of their conservation as exemplified for the fushi tarazu gene [56,57]. No conservation outside of the subgenus Sophophora was however observed suggesting a lack of intronic regulatory elements. Maybe the elements are too small to be detected this way. Likewise, the sequences in the H 5' or 3' UTR that were altered in the course of genomic engineering, are not conserved between D. melanogaster and other Drosophila species except for sibling species within the melanogaster subgroup. This leaves us with no simple explanation for the restricted activity of the H cDNA construct. Perhaps the manipulation of the genome has resulted in a sensitized genetic background uncovered in the hemizygous condition.
Both H cwt and H gwt homozygotes are fully viable without apparent phenotype, which demonstrates the functionality of the constructs. Notably, the reduced activity is only observed in the hemizygous condition as an enhancement of the H haplo-insufficient phenotype. Most other fly genes are fully recessive, i.e. normally such a subtle decrease in activity may go unnoticed. H heterozygotes in contrast appear to provide a highly sensitized genetic background regarding bristle development. This has been observed decades ago during systematic analyses of the H loss of bristles phenotype, which was shown to be highly background dependent [41]. In fact, three copies of H result in gain of function phenotypes in support of the notion that H activity is strictly dose sensitive [39,42,58]. Accordingly, H belongs to the small number of known quantitative trait loci implicated in natural variation of bristle number [59,60]. Quantitative traits are generally sensitive to the environment, effected by the simultaneous segregation of alleles at multiple loci. This explains easily the strong influence of the genetic background on the H bristle loss phenotype.
Against expectations we observed an enrichment of H protein in homozygous H gwt cells compared to cells bearing the GFP-marked third chromosome used in the FRT-mediated mosaic analysis in one or two copies (Fig 5). In fact the H locus shows extraordinary fluctuations of transcriptional activity within but not between species [61]: in this work 10 inbred D. simulans strains were compared with a pool of isogenic D. melanogaster strains. Perhaps, our mosaic analysis visualized these results in a direct way, i.e. differential levels of H expression in different genetic backgrounds. Usually gene expression is remarkably precise and reproducible within cell populations, despite the random molecular fluctuations expected to occur between individual cells. It has been proposed that this transcriptional noise is filtered through the coordinated activity of the two signalling pathways Wnt and Notch [62,62,63]. Hairless, as the major antagonist of Notch signalling activity in Drosophila, may be at the heart of this regulation, since its activity has a direct influence on Notch signalling output [11]. Perhaps the fluctuation in the expression of H uncovers the regulatory mechanisms implemented in filtering transcriptional noise.

Genome engineering methods
Genome engineering was performed as outlined before [23]. At first a 3 kb Bam HI/Kpn I 5' flanking fragment (5'arm) and a 6.4 kb Eco RV/Xho I 3' flanking fragment (3'arm) were excised from genomic subclones [39], and cloned into the 5' and 3' MCS of pGX-attP. Five independent transgenic pGX-H fly lines were established by random P-element mediated germ line transformation. To induce homologous recombination, a second chromosomal insertion line was chosen and crossed with y 1 w 1118 ; P{ry Ã , 70Flp}11P{v + ,70I-SceI}2B Sco/CyO (BL6934) under a heat shock regimen as outlined before [24]. One recombination event H attP w+ was obtained from roughly 1000 virgins bearing mosaic eye colour, crossed with w Ã ; P{70FLP} 10 (BL6938). It was confirmed by phenotype and PCR analysis using the primer pair P1 (upstream of 5' arm) and P2 (within attP). Vector sequences including the white + marker gene were deleted by CreI-mediated recombinase using y 1 w 67c23 P{Crey}1b; D Ã /TM3 Sb (BL851) as described before [24]. The resultant H attP mutant was stabilized over balancer chromosomes and confirmed by PCR using primer pair P3 and P4. In addition, the 1.3 kb amplificate was subcloned and sequence verified.
Generation of the H integration constructs followed largely the same strategy: all constructs cover the Kpn I/Eco RV fragment deleted in H attP (position 415 to 4187 according to FlyBase). The H-cWT construct contains the entire H cDNA (h9/h7) [39], starting with the Kpn I site in the 5' UTR and ending at position 4323 within the 3' UTR [14,39], i.e. 136 nt downstream of the Eco RV site (S1 Fig). At first the MCS of the original pBT H cDNA subclone [14] was extended by insertion of a Kpn I site between the 3' flanking Xba I and Sac I site using annealed primers. This way, the entire 4.2 kb insert could be excised as one Kpn I fragment and inserted into likewise opened pGE-attB GMR vector [23]. The same strategy was used for H-LD and H-iD cDNAs. Orientation was tested by confirmative restriction digests. Moreover, the H-LD and H-iD replacements were confirmed by diagnostic digests.
The genomic 5.1 kb H-gWT construct spans from Kpn I to Eco RV containing all the introns. It was gained from the HBS genomic subclone [39], which however, lacks the 5' Kpn I site due to a polymorphism (genotype Kr SB10 /SM1). Hence, we replaced an overlapping 1.9 kb Eco RI fragment with one containing the Kpn I site, derived from a different genomic library (H-clones, genotype Oregon R; [39]). Shuttling into pGE-attB GMR vector was as described above using the flanking Kpn I sites. Fly stocks were confirmed by PCR before and after floxing using primer pairs P3/P5 and P6/P4, respectively (S1 Fig). The genotypes of H gwt and H cwt were further confirmed by sequencing the junctions of a P7-P4 amplificate with primers S8 and S9. Primers are listed in S1 Table. Insertion of DNA constructs into the H attP genome was performed exactly as outlined before [23]. To this end, about 500-1000 embryos derived from a cross of w Ã ; H attP /TM6B males with y 1 M {vas-int.Dm}ZH-2A w Ã (BL40161) females expressing the phiC31 integrase [28], were injected and transgenic lines established. The frequency of recombinants ranged between 0.7%-23.3% of first generation crosses. The pGE-attB GMR and white + marker sequences were eliminated with Cre-recombinase by crossing in y 1 w 67c23 P{Crey}1b; D Ã /TM3 Sb (BL851). White eyed stocks lacking the recombinase and balanced over TM6B were established for each construct. The correct genotype was confirmed by PCR analyses and sequencing.

Yeast two-hybrid experiments
Yeast two-hybrid experiments were performed according to standard protocols [64,65], using pJG Su(H) [10] as prey and pEG NTCT [30], pEG NTCT-L235D [15] or pEG NTCT-I244D (this work) as bait. Empty vectors served as control. QuikChange 1 II XL Site-Directed Mutagenesis Kit (Agilent Technologies, Böblingen, Germany) was used to introduce the I244D exchange in NTCT [30] according to the manufacturer's protocol. Primers iDup and iDlo are listed in S1 Table. The clone was sequence verified and was shuttled as Bam HI/Xho I fragment into likewise opened pEG202 vector [66] for yeast two-hybrid experiments.
Clones were induced using the Flp/FRT technique as outlined before [17,32,68] using either H gwt , H cwt , H attP , H iD or H LD allele recombined with P{neo FRT}82B (BL2050) and crossed with P{neo FRT}82B P{Ubi-GFP S65T nls}3R/TM6B (BL32655). FLPase was induced by a 1h heat shock at 37°C in first to second instar larvae, to be dissected as wandering third instar.

Immuno-staining of wing imaginal discs
Immuno-staining of imaginal discs was according to standard protocols using Hairless anti-A (1:500, from guinea pig) [21], mouse anti-Cut (1:25) or anti-Wingless (1:25) (developed by G. Rubin and S.M. Cohen, respectively; obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Dept of Biology, Iowa City, IA 52242) and rabbit anti-GFP (1:200, Santa Cruz Biotech, Dallas, USA). Goat secondary antibodies coupled to FITC, Cy3 or Cy5 were from Jackson Immuno-Research (Dianova, Hamburg, Germany). Tissue was mounted in Vectashield (Vector Labs, Eching, Germany), and examined using a BioRad MRC1024 confocal microscope and Laser-Sharp 2000TM software (Carl Zeiss, Jena, Germany).
Supporting Information S1 Fig. Reintegration of H variants. Schematic representation of the integration events at the H locus (not to scale). A) The pGE-attB GMR vector, containing the H construct of interest (H Ã ) and a white + marker, is recombined via its attB site into the attP site present in place of the H locus in the H attP allele. B) As an example, the resulting flies carry the entire vector at the H locus, which was confirmed by a 5'H and a 3'H PCR (see D). Due to the recombination event, attP and attB have been changed to attR and attL. C) Vector sequences and the white + marker were removed by Cre-mediated recombination at the loxP sites. In the end the H gene is replaced by the given allele; only the attR and the loxP sites remain in the 5' and the 3'UTR, respectively. D) Control PCR reactions performed on single flies before floxing (see B). M, marker (appr. size in kb); 1, H attP w+ /TM6B; 2, H cw w+t ; 3, H LD w+ /TM6B; 4 H iD w+ /TM6B, 5, H gwt w+ . For the 5'H PCR a 2.28 kb is expected for the wild type chromosome (arrow); a 2.39 kb fragment after introduction of the genomic H-gwt construct (size increase due to attR); a 1.64 kb fragment after the introduction of H cDNA constructs lacking the introns (arrowhead) (primer pair P3/P5). Asterisk, unspecific priming. Using primer pair P6/P4, the 3'H PCR is expected to give a 1.25 kb fragment in case of a successful integration event (arrowhead). D') Control PCR reactions performed on single flies after floxing (see C). The 5'H PCR is expected to give the same results as in D). With the 3'H PCR no amplificate is expected; as control unfloxed H gwt w+ was included (arrowhead). Asterisk, unspecific priming. M, marker (appr. size in kb); 1, H attP /TM6B; 2, H cwt ; 3, H LD /TM6B; 4, H iD /TM6B; 5, H gwt ; 6, H gwt w+ . E) Scheme of the H 3' UTR to show position of changes relative to known or potential functional elements like polyadenylation sites [p(A)] and micro-RNA binding sites (miR). Target sites for several miRs are indicated as predicted by TargetScanFly (release 6.2; http://www.targetscan.org/fly_ 12). Purple, 8mer; red, 7mer-m8; blue, 7mer-1A). Numbering is according to FlyBase. Ends of major transcripts RA-RD (B/D, A/C), as well as of the cDNA NB15 [39] are indicated by an arrowhead. Polylinker sequences are shown in black for pBT and grey for pGX-attP; loxP is in yellow. The 136 nt sequence duplication, flanking the foreign sequences, is marked by an orange box; it does not contain known elements. Wild type (OreR) and the given H allele, respectively, were crossed with N 5419 /FM7c virgins, and female offspring was evaluated for notched wings. 33-95 animals were analysed. B) Wild type (OreR) and the given H allele, respectively, were crossed with Dl B2 /TM6C Sb virgins. Wings of female offspring were evaluated for vein thickening at 13 positions indicated in the scheme. Thickening was recorded with a value of 1, no thickening with 0. Total number of analysed wings is given for each combination. Statistical relevance was determined with a one tailed Student's T-test (ns, not significant with p>0.05; ÃÃ p<0.01; ÃÃÃ p<0.001; n, 17-28). (TIF) S1 Table. List of primers, given in 5' -> 3'. (DOC)