Complex genetic interactions of novel Suppressor of Hairless alleles deficient in co-repressor binding

Throughout the animal kingdom, the Notch signalling pathway allows cells to acquire diversified cell fates. Notch signals are translated into activation of Notch target genes by CSL transcription factors. In the absence of Notch signals, CSL together with co-repressors functions as a transcriptional repressor. In Drosophila, repression of Notch target genes involves the CSL homologue Suppressor of Hairless (Su(H)) and the Notch (N) antagonist Hairless (H) that together form a repressor complex. Guided by crystal structure, three mutations Su(H)LL, Su(H)LLF and Su(H)LLL were generated that specifically affect interactions with the repressor H, and were introduced into the endogenous Su(H) locus by gene engineering. In contrast to the wild type isoform, these Su(H) mutants are incapable of repressor complex formation. Accordingly, Notch signalling activity is dramatically elevated in the homozygotes, resembling complete absence of H activity. It was noted, however, that heterozygotes do not display a dominant H loss of function phenotype. In this work we addressed genetic interactions the three H-binding deficient Su(H) mutants display in combination with H and N null alleles. We included a null mutant of Delta (Dl), encoding the ligand of the Notch receptor, as well as of Su(H) itself in our genetic analyses. H, N or Dl mutations cause dominant wing phenotypes that are sensitive to gene dose of the others. Moreover, H heterozygotes lack bristle organs and develop bristle sockets instead of shafts. The latter phenotype is suppressed by Su(H) null alleles but not by H-binding deficient Su(H) alleles which we attribute to the socket cell specific activity of Su(H). Modification of the dominant wing phenotypes of either H, N or Dl, however, suggested some lack of repressor activity in the Su(H) null allele and likewise in the H-binding deficient Su(H) alleles. Overall, Su(H) mutants are recessive perhaps reflecting self-adjusting availability of Su(H) protein.


Introduction
The Notch signalling pathway is instrumental for a multitude of cell fate decisions during the development of higher metazoan animals. The principle outcomes of Notch activity are cells of a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 different fate arising from a direct intercellular communication of cell neighbours [1,2]. A prime example is the process of lateral inhibition, where single cells are selected from a cell group of originally equal potential. The cells selected retain their primary fate, whereas their neighbours are directed into a secondary fate. The selection of sensory organ precursor cells giving rise to mechano-sensory bristle cells, or the refinement of a wing vein from a field of cells with provein potential, are classical examples for lateral inhibition taking place during the development of Drosophila melanogaster (for review: [3][4][5]). Failure of this process, for example as consequence of mutations in Notch signalling components, results in too many bristles or in thickened veins [6]. The opposite phenotypes, lack of bristles or veins, are observed when Notch activity is gained, and primary cell fate is completely inhibited as a consequence [3,7,8]. In addition to the process of lateral inhibition, Notch activity is also required for the formation of the dorso-ventral boundary in the wing anlagen that eventually forms the wing margin [9][10][11]. Accordingly, downregulation of Notch activity causes failure of wing margin formation, giving rise to wing incisions, i.e. name-giving wing 'notches' [3,6]. Moreover, specification of the sensory organ precursor cell's daughters requires differential Notch activity. The outer shaft differentiates from the socket cell by a specific Notch signal. Again, loss of Notch activity may result in a double shaft, and gain of Notch activity in a double socket phenotype (for review: [12][13][14]).
The Notch signalling pathway, simplified, consists in Drosophila of the following core components (for review: [2,11,15]): two transmembrane ligands, Delta (Dl) and Serrate (Ser) presented on the signalling cell, the transmembrane receptor Notch on the signal receiving cell, plus the transcription factor Suppressor of Hairless (Su(H)) that assembles activator or repressor complexes on Notch target genes, depending on the activation status of the receptor. Once Notch is bound by Dl or Ser, it is cleaved within the membrane, and the intracellular domain -i.e. activated Notch-is released. By binding to Su(H) the Notch intracellular domain (NICD) assembles an activator complex together with Mastermind (Mam), resulting in a burst of transcriptional activity from Notch target genes (for review: [2,5,11,16]). The unligated Notch receptor remains at the membrane, leaving the cell under the rule of its antagonist named Hairless (H) (for review: [2,16,17]). H binds to Su(H), and by recruitment of general corepressors Groucho and C-terminal binding protein, it causes the silencing of Notch target genes [16][17][18][19][20][21][22][23]. Su(H) can therefore be considered a molecular switch: activating or repressing Notch target genes depending on the bound cofactors and the cellular context. Su(H) binds the two cofactors Notch and H with similar affinity at nanomolar range [24,25]. The structure of either activator or repressor complex has been determined by X-ray crystallography [24,26]. It was shown that two structural domains of Notch contact Su(H) at the surface of its beta-trefoil and C-terminal domains [ [6,28,29].
Comprehensive genetic analyses have revealed that accurate Notch signalling relies on a well-balanced presence of its core components. In fact, among only 26 haplo-insufficient loci discovered in Drosophila melanogaster by extensive chromosomal deletion studies [30] three comprise the Notch core components N, Dl and H, whereas the majority affects ribosomal proteins. Heterozygous null mutants N, Dl or H display dominant phenotypes mostly affecting bristles and wings. As described above, N heterozygotes are characterized by wing incisions, and display thickened longitudinal veins in addition [6]. Likewise, Dl heterozygotes develop thickened and knotted veins, whereas H mutants are characterized by loss of mechano-sensory bristles or hairs and shortened and thinned wing veins [6]. Since H-binding deficient Su(H) alleles are expected to behave similar to H null mutants, we might have expected a likewise reduced bristle number or vein defects. Instead, there are no indications of a dose sensitivity for Su(H) activity. We therefore used genetic interaction studies to define gain and loss of function activities in these novel Su(H) mutants. Firstly, our studies confirm earlier reports on a gain of Notch activity in heterozygous Su(H) null alleles, which is in agreement with a loss of repressor function. Hence Su(H) attP resemble H attP heterozygotes in the genetic combinations similar to H-binding deficient Su(H) alleles. We also observed allele specific interactions of Su (H) LLF and Su(H) LLL with N 5419 or H attP . The results suggest that Su(H) LLF , in addition to its failure to assemble repressor complexes with H, may also be hampered in activator complex formation. Su(H) LLL however, is the one allele that resembles closest a H loss of function. The name giving suppression of the double socket phenotype of H heterozygotes, however, was only observed in combinations with the Su(H) attP null allele. In contrast, H-binding deficient Su(H) alleles behaved more like wild type control with little influence on socket cell formation. We conclude that the dose sensitive interaction uncovered in Su(H), H doubly heterozygotes is a result of the specific auto-regulatory activity of Su(H) during socket cell development.

H-binding deficient Su(H) alleles are neither suppressors nor enhancers of Hairless
As indicated by its name, the Suppressor of Hairless (Su(H)) locus was originally identified by virtue of its dosage-sensitive, antagonistic interaction with Hairless (H) mutants: the haploinsufficient bristle defect of H loss of function alleles is suppressed in trans-heterozygous combinations with Su(H) loss of function alleles, whereas it is enhanced by duplications of the locus [28,29,31-33]. According to the role as Notch antagonist, mutations in H result in a gain of Notch activity, expected to be ameliorated by a concomitant loss of the signal transducer Su (H) [12,28,29 Fig 1A) [27]. In fact, we have no indications for a dose dependency of these mutants as tissue hypertrophy is indistinguishable between homo-and hemizygotes, i.e. independent of whether one or two mutant gene copies are present [27]. In the heterozygous condition, bristle shafts and sockets of H-binding deficient Su(H) alleles, Su (H) LL , Su(H) LLF and Su(H) LLL , appear wild type, indicating normal regulation of Notch signalling effecting regular cell type specification (Fig 1A). Two controls were used, wild type Oregon R flies and Su(H) gwt . The latter was generated by introducing a genomic wild type copy of In heterozygous H attP /+ females about 12 bristles are affected on average, a value that drops to about 5 in the absence of one Su(H) gene copy as in Su(H) attP (Fig 1B). These numbers are Numbers of affected macrochaete (either missing or transformed to a socket) were determined in female flies (n!20), and are represented as box plot: centre lines depict the medians, box limits indicate the 25 th and 75 th percentiles, whiskers extend 1.5 times the interquartile range, outliers are shown by dots. Statistical significance was determined by ANOVA two-tailed Tukey-Kramer approach for multiple comparisons; significant differences are indicated-color code referring to correspondingly colored box (highly significant ÃÃÃ , p<0.001; very significant ÃÃ , p<0.01).  Fig 1B). Similar numbers are observed in combinations with the H-binding deficient alleles that share the identical genetic background ( Fig 1B). Whereas Su(H) LL and Su(H) LLL have little impact on the haplo-insufficient phenotype resulting from a reduction of H gene dose, Su(H) LLF had a mild but significant suppressor effect ( Fig 1B). The latter may result from structural deficits in the Su(H) LLF protein uncovering some loss of Su(H) activity [26,27]. We had expected an enhancement of the H mutant bristle phenotype in the trans-heterozygous combinations, since in a homozygous condition the three new alleles gain Notch activity similar to H mutants [27]. Instead, the three alleles behaved more like the control, and only Su(H) LLF behaved like a weak hypomorphic Su(H) allele in this context.

Impact of H-binding deficient Su(H) alleles on the haplo-insufficient H attP mutant wing phenotype
Apart from the name giving bristle phenotype, H mutants are characterized by a distal shortening of longitudinal L5 with high, and of L4 veins with low penetrance (Fig 2A) [6]. Loss of wing veins results from increased lateral inhibition, reflecting gain of Notch activity (overview in [4,34] LL and Su(H) LLF heterozygotes: it occurs at very low penetrance and not significantly different from control. Su(H) LLL heterozygotes, however, display shortening of L5 in a substantial fraction of wings (Fig 2A and 2B).
It has been reported that in combination with Su(H) loss of function alleles, the venation defects of H heterozygotes are enhanced [6,29,35]. We confirmed this observation: not only L5 but also L4 was affected in flies doubly heterozygous for the null mutations Su(H) attP and H attP , albeit at low frequency ( Fig 2B). Apparently in the context of wing development, a reduced repressor activity is uncovered in the Su(H) attP null allele. The H-binding deficient Su(H) alleles, however, gave different results: when combined with H attP , wing phenotypes of the doubly heterozygotes were nearly unchanged regarding frequency and gap size, except for Su (H) LLL (Fig 2B and 2C). Not only was the size of the L5 gap significantly increased in the Su (H) LLL /+; H attP /+ combination (Fig 2C), the L4 was affected as well in a high proportion of the wings ( Fig 2B). Overall, we were surprised by these results, as we might have expected an enhancement of venation defects, affecting L4 and L2 as well also in the other combinations with Su(H) LL and Su(H) LLF . Such an enhancement of the H wing venation phenotype can be seen for example in combinations with N Ax mutations that gain Notch activity [36][37][38]. In sum, with exception of Su(H) LLL , enhancement of the heterozygous H wing phenotype was absent or rather mild, indicating that the remaining wild type copies of the Su(H) and H genes were sufficient for normal repression of Notch activity.

Genetic interactions between H-binding deficient Su(H) alleles and the Notch loss of function allele N 5419
Notch mutants were originally picked up by their characteristic notched wing phenotype observed in heterozygous females [6,39] (Fig 3). This phenotype is exquisitely sensitive to genetic background [6,[39][40][41], and can be completely rescued by reducing the gene dose of Delta or Hairless [42][43][44]. We reasoned that H-binding deficient Su(H) alleles might behave similar to H mutants in this context, as we expected them to gain Notch activity, whereas Su (H) null alleles were expected to enhance loss of Notch activity. To our surprise we found that the deficiency Su(H) Δ47 rescued wing notching of N 5419 heterozygous null mutants very well, whereas the null allele Su(H) attP did not (Fig 3). We noted a high degree of variations, however, between independent experiments as well as between lines with formally identical genetic background. Similarly, high variability was observed amongst the three controls N 5419 /+ (derived from N 5419 /FM7c x Oregon R), N 5419 /y 1 w 67c23 , and N 5419 /+; Su(H) gwt /+ (Fig 3) in line with the background sensitivity of N phenotypes [6,[41][42][43][44]. In fact, Su(H) attP did not vary significantly from these controls, and neither did Su(H) LLF . As noted in the context of bristle development, the Su(H) LLF mutant appears a hypomorphic mutant in the N/+ mutant background as well (Fig 3). The two other H-binding deficient Su(H) alleles, Su(H) LL and Su(H) LLL , however, gained Notch activity and rescued the wing defects very well, thereby resembling the H attP null mutant and the Su(H)-binding deficient H LD allele ( Fig 3B). Notably, H alleles had the greatest potential in rescuing Notch defects (Fig 3B). We cannot decide, however, whether these results reflect meaningful differences in biological activity or are due to genetic background [41].

Genetic interactions between H-binding deficient Su(H) alleles and the Delta loss of function allele Dl B2
Delta heterozygous mutant flies display a thickening and knotting of veins, notably along the longitudinal veins as well as the cross veins. Moreover, veins frequently run out in a delta at the wing margin, which is the name giving phenotype [6,45] (Fig 4). Interestingly, this phenotype is ameliorated by both gain and loss of Notch activity. Doubly heterozygous Dl /H flies show nearly wild type wings [42,44,46], i.e. both dominant phenotypes are compensated for, which can be explained by a loss of Notch activity in the absence of Dl and a gain of Notch activity in the absence of H. Both Dl and N wing phenotypes are ameliorated in doubly heterozygous N/+; Dl/+ flies as well [42]. This phenotype has been accounted for by balancing the gene dose of ligand and receptor, or more likely by cis-inhibition [3,47]. We have combined the null allele Dl B2 with the H-binding deficient Su(H) alleles, and noted an amelioration of vein thickening (Fig 4). In this context, the three H-binding deficient Su(H) alleles were indistinguishable from the null allele Su(H) attP . The rescue, however, was clearly less effective than that obtained with the H attP null allele (Fig 4B). Unexpectedly, we also observed a significantly weakened phenotype in the Su(H) gwt /+; Dl B2 /+ combination compared to Dl B2 /+ alone ( Fig  4B). The effect was clearly weaker than with the other Su(H) alleles or with H attP , however ( Fig  4B). Whether it results from genetic background or whether the Su(H) gwt allele lacks some Su (H) activity cannot be distinguished at the moment [41].

Discussion
In this work, we have studied the genetic interactions between H-binding deficient Su(H) alleles and mutations affecting several Notch pathway components. The results are complex, perhaps reflecting the dual role of Su(H) in Notch signal transduction as activator and repressor alike [11,16,17]. In the activator complex together with NICD and Mam, Su(H) acts as a column). Columns of the doubly heterozygotes are highlighted in orange. Shortening of L4 is indicated in addition, and is shown in purple. (C) The length of distal L5 (indicated as a magenta line in the control wing in (A)) was measured and recorded as fraction of the expected length. Standard deviation is given (n!16). Statistical significance was determined by ANOVA two-tailed Tukey-Kramer approach for multiple comparisons; significant differences are color labeled with blue relative to control and magenta relative to Su(H) LLL /+; H attP /+ (highly significant ÃÃÃ , p<0.001; significant Ã , p<0.05). Note little variation from control in the heterozygous Su(H) alleles except for Su(H) LLL (Fig 1). In fact, in combination with N 5419 heterozygotes it was undistinguishable from the Su (H) attP null allele (Fig 3). We conclude that Su(H) LLF , in addition to its incompetence for repressor complex formation, has a reduced ability to form Notch activation complexes perhaps as a consequence of structural deficits [26]. In contrast, Su(H) LLL is the one allele that resembles mostly the H attP null mutant also in the heterozygous condition, judged by the wing venation defects this allele displays on its own and in combination with H (Fig 2). Other than that, H-binding deficient Su(H) alleles are fully recessive, unlike H. Apparently, one wild type copy of Su(H) is sufficient for normal Notch pathway repression, and the mutant Su(H) gene copy does not increase Notch output in any detectable way.
The core components of Notch signalling in Drosophila, the ligand Delta, the receptor Notch and the co-repressor Hairless are dose sensitive, i.e. mutations are haplo-insufficient and cause dominant phenotypes [6]. Curiously, mutations in the two central players, the coactivator Mastermind (Mam) and the transcription factor Su(H), are recessive, suggesting that these two factors are not limiting in the process of Notch signal transduction [6,29,30]. Mam together with NICD and Su(H) assemble the activator complex [24]. As recruitment of Mam strictly relies on the presence of NICD, it has no effect on Notch target gene expression on its own [11,24]. Hence it can be hold available in excess without influencing signalling output. The picture is quite different for Su(H): in the absence of NICD, Su(H) may engage in other transcriptional complexes impacting gene expression [11,16,52]. The classical example is Quantitative analysis summarizing percentage of notched wings derived from females of the given genotype (n, total number of wings is given in each column). Standard deviation is given from 2-4 independent experiments. Statistical significance was determined on the total by ANOVA two-tailed Tukey-Kramer approach for multiple comparisons; significant differences are color coded correspondingly (highly significant ÃÃÃ , p<0.001; very significant ÃÃ , p<0.01; significant Ã , p<0.05). Note high variations in the three control crosses N 5419 /+ (derived from N 5419 /FM7c x Oregon R), N 5419 /y 1 w 67c23 , and N 5419 /+; Su(H) gwt /+. The two null alleles Su(H) Δ47 and Su(H) attP show drastically different genetic interactions-the former rescuing N wing phenotypes, the latter behaving rather like a wild type allele. Moreover, N 5419 /+; Su(H) LLF /+ does not differ from the controls.  Su(H) alleles have been identified originally by their dominant suppression of the bristle defects normally observed in heterozygous H mutants [28,29,32] (Fig 1). H heterozygotes either completely lack the entire bristle organ or just the bristle shaft, which is instead transformed into a bristle socket [12,53,54]. The resultant double sockets are observed in place of normal bristle organs [44,[54][55][56]. Loss of complete bristle organs in H mutants originates from increased Notch signalling within the proneural field, preventing selection of the bristle founder cell [12,54]. The double socket phenotype reflects the role of Notch in asymmetric cell type specification within the bristle lineage: the sensory organ precursor cell divides asymmetrically into a cell pair of outer and inner fate (pIIa and pIIb), with pIIa eventually giving rise to shaft (trichogen) and socket (tormogen) cells [12,14,54,57]. Gain of Notch activity prevents the asymmetry, and drives for example trichogen into tormogen fate, resulting in the typical double socket phenotype observed in heterozygous H mutant flies [44,54,57,58]. Interestingly, suppression of H phenotypes by Su(H) is restricted to the double sockets, i.e. the decrease of Su (H) levels improves trichogen formation in this sensitized genetic background [29,32].
Su(H) plays an active role within the tormogen, whereas the trichogen forms by default repression of Notch signals [12,51,59,60]. Su(H) is required within the socket cell, as a pIIa cell lacking Su(H) gives rise to shaft cells only [33,61]. Su(H) protein accumulates to high levels within the tormogen by means of an autoregulatory element (ASE) driving Su(H) expression in response to Notch activity [32, [59][60][61][62]. One of Su(H) targets is Sox15, and together the two proteins entail socket cell differentiation as well as normal electrophysiology of the bristle organ [59,60]. Moreover, they inhibit the transcription of shaven, a gene required for shaft cell differentiation [60,63]. shaven is expressed in the late sensory organ precursor cell and its daughters, but later on expression is only maintained in cells protected from Notch signal [63]. Protection from Notch signals is the prime role of the two proteins Numb and Hairless that antagonize Notch activity [12,14,17,57,61]. By enabling shaven expression, they set a bias for the shaft fate [60,63]. Accordingly, the double socket phenotype of H is strongly enhanced by shaven mutants [53], and that of numb is suppressed by Su(H) [61].
In the heterozygous H mutant, commitment to the trichogen fate has become unstable, since Notch activity cannot be completely abolished. Spurious Notch signals may switch on Su (H) expression via the ASE, resulting in the activation of Sox15 and the repression of shaven, eventually driving socket cell differentiation. In this H/+ context, Su(H) appears dose sensitive: with centre lines representing the median, box limits indicate the 25 th and 75 th percentiles, whiskers extend 1.5 times the interquartile range and outliers are represented by dots. Statistical significance was determined by ANOVA twotailed Tukey-Kramer approach for multiple comparisons; significant differences are color labeled correspondingly (highly significant ÃÃÃ , p<0.001; very significant ÃÃ , p<0.01). https://doi.org/10.1371/journal.pone.0193956.g004

Complex genetic interactions of novel Su(H) alleles
The lowered Su(H) dose may be insufficient for spurious Notch activity to trigger ASE activation, and shaven expression-i.e. trichogen bias-is maintained. H-binding deficient Su(H) alleles have little influence on the trichogen-tormogen fate selection (Fig 1). In these mutants, the NICD binding sites are untouched [24,26]. Activator complex assembly is expected to occur normally [26], as is the response to Notch receptor activation within the presumptive tormogen. These Su(H) alleles should be similar to wild type with regard to the activation of Notch target genes including ASE regulation, which is in agreement with our observations. Overall, the heterozygous condition of Su(H) is fully recessive except for the sensitized H/ + genetic background regarding the tormogen-trichogen cell lineage. If any, a slight gain of Notch activity is uncovered in Su(H) heterozygotes, reflecting the important role of Su(H) in the default repression of Notch target genes.

Materials and methods
Flies were maintained on standard fly food at 18˚C; crosses were raised at 25˚C. The following stocks were used: Oregon R (BL5), y 1 [65], Su(H) Δ47 /CyO [66]. Further information on fly strains is found in Flybase (flybase.org) and the Bloomington Stock Centre (http://flystocks.bio.indiana. edu/). Standard genetics were applied for the re/combination of fly stocks. Where applicable, genotypes were confirmed by PCR plus diagnostic restriction digests where applicable. Mutant phenotypes were documented as outlined before [44]. Adult wings from female flies were dehydrated in ethanol and mounted in Euparal (Roth, Karlsruhe, Germany). Pictures were taken with an ES120 camera (Optronics, Goleta CA, USA) mounted to a Zeiss Axiophot (Carl Zeiss AG, Jena, Germany) using Pixera Viewfinder software, version 2.0. Scanning electron micrographs were captured with a table-top scanning electron microscope (Neoscope JCM-5000; Nikon, Tokyo, Japan) of uncoated animals. Pictures were assembled using Photo Paint and Corel Draw software. Image J was used for measurements of vein length. Statistical significance was determined by ANOVA using a two-tailed Tukey-Kramer approach for multiple comparisons (highly significant ÃÃÃ , p<0.001; very significant ÃÃ , p<0.01; significant Ã , p<0.05; not significant ns, p>0.05). Box blots were compiled with the online plotting tool BoxPlotR. The raw data are contained in S1 Table. Supporting information S1