Genetic and Molecular Interactions between HΔCT, a Novel Allele of the Notch Antagonist Hairless, and the Histone Chaperone Asf1 in Drosophila melanogaster

Cellular differentiation relies on the highly conserved Notch signaling pathway. Notch activity induces gene expression changes that are highly sensitive to chromatin landscape. We address Notch gene regulation using Drosophila as a model, focusing on the genetic and molecular interactions between the Notch antagonist Hairless and the histone chaperone Asf1. Earlier work implied that Asf1 promotes the silencing of Notch target genes via Hairless (H). Here, we generate a novel HΔCT allele by genome engineering. Phenotypically, HΔCT behaves as a Hairless gain of function allele in several developmental contexts, indicating that the conserved CT domain of H has an attenuator role under native biological contexts. Using several independent methods to assay protein–protein interactions, we define the sequences of the CT domain that are involved in Hairless–Asf1 binding. Based on previous models, where Asf1 promotes Notch repression via Hairless, a loss of Asf1 binding should reduce Hairless repressive activity. However, tissue-specific Asf1 overexpression phenotypes are increased, not rescued, in the HΔCT background. Counterintuitively, Hairless protein binding mitigates the repressive activity of Asf1 in the context of eye development. These findings highlight the complex connections of Notch repressors and chromatin modulators during Notch target-gene regulation and open the avenue for further investigations.


Introduction
The highly conserved Notch signaling pathway mediates cell-cell communication in the process of cellular specification during development and in the context of tissue homeostasis (reviewed in Refs [1,2]). Given its impact on human health, a comprehensive understanding of its regulation is of critical importance. To this end, model systems, such as Drosophila melanogaster, play a pivotal role in defining the complex interrelations of the many factors involved in signaling. Notch signaling occurs between two neighboring cells-one cell presenting a membrane bound ligand to the other cell presenting the Notch receptor. As a consequence of receptor-ligand binding, Notch target genes are activated in the signal receiving cell, directing a specific cell fate or provoking cellular change (reviewed in Refs. [2][3][4]). Central to Notch signal transduction lies the DNA binding transcription factor CSL (acronym of mammalian CBF1/RBPJ, fly Su(H) and worm Lag1 orthologs), Ref. [42]). In Drosophila, Asf1 was found to contribute to the selective silencing of Notch target genes via the general Notch antagonist Hairless (H) [28]. In this context, Asf1 is part of a large protein complex, including several different histone modifiers [33].
Here, our work demonstrates that the Asf1-H relationship is much more complex than previously appreciated. An H allele (H ∆CT ) lacking the Asf1 binding site behavescontrary to expectations-as a stronger, not weaker, inhibitor of Notch signaling, suggesting that, under native conditions, H attenuates Asf1 repressive activity during Notch signal regulation. This finding is interesting in the context of the exquisite sensitivity of Notch signaling activity and changes in the chromatin landscape, as both Asf1 and H have been implicated in the modulation of local chromatin structure. As our work fits within the general framing of Asf1 function, it may open the avenue for investigating the complex connections of Notch repressors and chromatin modulators during Notch target-gene regulation.

Genome Engineering of the H ∆CT Allele
The generation of the H ∆CT allele followed the principles outlined before [43]. Genomic full-length Hairless DNA in pBT vector (5.1 kb Kpn I/Eco RV fragment) [43] was Xba I/Sac I digested and ligated with an annealed primer pair containing a Kpn I recognition sequence and compatible Xba I/Sac I sticky ends. The Eco RI restriction site was deleted from the polylinker by a Sma I/Eco RV digest followed by re-ligation. Afterward, the CT domain coding sequences were removed together with the separating intron by Eco RI/Bsp EI digestion. An annealed primer pair carrying compatible sticky ends for Eco RI and Bsp EI was ligated to circulate the clone; it included some sequences eliminated before (see Figure 1a,c): HdCt up 5 AAT TCT ACG CTT GGC CGA GCT GCT CT 3 and HdCt lo 5 CCG GAG AGC AGC TCG GCC AAG CGT AG 3 . The third intron was PCR amplified; the lower primer contained an Eco RI site and splice sites: Hint-dCT up 5 CGT CGG CGG TGA CCA CAG CGT AT 3 and HdCT lo 5 GAG CTG TTG GAA TTC GAA ACT GCA ATG AAG AG 3 . The amplicon was reintroduced by Eco RI digestion and ligation. The final construct was shuttled into the pGE-attB GMR transformation vector [44] via the Kpn I sites and sequence verified. The H ∆CT allele was inserted into the H attP founder genome as described earlier [43,44]. Successful transgenic lines were confirmed by PCR. Subsequently, the pGE-attB GMR and white + marker sequences were eliminated via Cremediated recombination [43,44]; the final line was PCR genotyped.

Tissue-Specific Expression Using the Gal4/UAS-System
Tissue-specific expression was performed using the Gal4/UAS system [45]. In order to avoid position effects and allow a direct comparison, the H-overexpression constructs were all placed at the identical chromosomal position at 3 L 68E via PhiC31 integrase-based insertion using the ΦX-68E strain as outlined earlier [19,46]. To this end, the deletion ∆CT (codons 271-320) was introduced in a Cla I fragment of the Hairless cDNA by using the ExSite PCR Based Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). The truncated Cla I fragment was reinserted into the full-length Hairless cDNA and cloned into pUAST-attB-vector [46] as Acc 65I/Xba I fragment; it was sequence verified. The successful transformant UAS-attB-H∆CT line was confirmed by PCR genotyping. A fulllength UAS-attB-H transgenic line, likewise inserted in 68E, served as control [19]. Further UAS lines used were UAS Asf1/CyO (kind gift of S. Bray) [47], UAS-CtBP (kind gift of S. Parkhurst) [48], UAS-Gro (kind gift of C. Delidakis) [49], UAS-GFP (BL4775) [50], UAS dsAsf1 (VDRC 23737) [51], UAS dsCtBP [14], UAS dsGro [52], UAS dsH [14]. The driver lines were ey-Gal4 (BL5534) [53], gmr-Gal4 (BL1104) [54], Eq1-Gal4 (kind gift of H. Sun) [55] and omb-Gal4 [56]. Moreover, the combined lines ey-Gal4 UAS Asf1/CyO (gift of S. Bray) [47] and ey-Gal4 UAS dsAsf1/CyO were used; the latter was generated by meiotic recombination followed by PCR-based genotyping. For the ectopic expression of ey-Gal4 UAS Asf1 in the homozygous H ∆CT background, we crossed y 1 w 67c23 ; H ∆CT /H ∆CT with ey-Gal4 UAS Asf1/CyO flies. Red-eyed offspring carries one copy of ey-Gal4 UAS Asf1 on the second and one allele of H ∆CT on the third chromosome. To avoid recombination, we crossed the respective males back to homozygous virgin y 1 w 67c23 ; H ∆CT /H ∆CT females and selected again for red-eyed offspring, half of which were expected to be homozygous for H ∆CT in the background of ey-Gal4 UAS Asf1. Flies were categorized by phenotype and subsequently genotyped by PCR. Figure 1. Generation of H ∆CT flies by genome engineering. (a) Full-length genomic H DNA was cloned into pBT between Kpn I and Eco RV polylinker sites [17]. Coding sequences are shown in green, intron sequences in yellow and leader and trailer sequences in white. Sequence coding for Su(H) interaction domain is shown in red (NT, N-terminal domain); the conserved C-terminal domain (CT) is depicted in black. The Eco RI site (red) was subsequently removed from the polylinker by re-ligation after Eco RV and Sma I digestion. A Kpn I site (green) was added at the 3 end by Xba I and Sac I digest followed by ligation of an annealed primer with compatible sticky ends. This was necessary for final cloning in the Kpn I site of the pGE-attB-GMR vector [44]. The CT domain, including the third intron, was deleted by an Eco RI/Bsp EI double digest. A primer pair with Eco RI/Bsp EI sticky ends as well as the deleted sequences of the third intron was annealed and ligated into the opened construct; see (c). The third intron was PCR amplified with an upper primer 5 of the endogenous Eco RI site and a 3 lower primer containing an Eco RI site and splice sites; see (c). The construct was then Eco RI digested, and the Eco RI digested PCR amplificate containing the third intron was reintroduced. Afterward, the genomic clone was ligated into the pGE-attB-GMR vector as Kpn I fragment and inserted into the H attP founder line to generate H ∆CT flies [43,44]. (b) Relevant parts of the H amino acid sequence containing the NTCT domain are shown. The upper amino acid sequence displays the wild-type H with the Su(H) binding motive NT in red and the conserved CT domain in italics; magenta residues are conserved in A. mellifera, the bold ones are identical, and underlined ones also in D. pulex [23,24]. Conserved basic sequences are shown in purple [22]; those identical in the honeybee are underlined. Framed residues (300-337) are deleted in H ∆CT . (c) Relevant genomic sequence in H ∆CT DNA with restriction sites used for cloning. Introns are marked in yellow. The introns conform to the GT-AG rule (marked in red). Underneath the amino acid sequence: in red, the Su(H) interaction domain NT; in italics and underlined, the remaining conserved residues between Drosophila and Daphnia. The newly introduced Eco RI restriction site (red) did not change the asparagine and serine codons (N,S in light blue) from wild type. Arrows indicate the position of the primer pair for intron amplification. The 5 start bases of the lower primer correspond to wild-type sequences deleted after Eco RI digestion of the PCR amplification product. The sequences between the Bsp EI and Eco RI restriction sites were introduced with the annealed pair of primers (bold and underlined) indicated in (a).

Adult Phenotypes
Adult flies of the respective genotype were collected and etherized. Pictures were taken with an ES120 camera (Optronics, Goleta, CA, USA) mounted to a Wild M5 stereomicroscope (Leica, Wetzlar, Germany). Dehydrated wings of female flies were embedded in Euparal (Carl Roth, Karlsruhe, Germany) and documented with an ES120 camera (Optronics, Goleta, CA, USA) coupled to a Zeiss Axiophot (Carl Zeiss, Jena, Germany). Pictures were recorded using Pixera viewfinder software version 2.0. Alternatively, uncoated etherized flies were pictured with a table top scanning electron microscope NeoScope (JCM-5000 SEM; Nikon, Tokyo, Japan) using proprietary software. The wing or eye area was determined using the freehand tool of Image J (open source). Only the SEM pictures of female flies were used for the quantification of micro-and macrochaetae, respectively. The numbers of bristles were determined as described before [57][58][59][60]. Statistical analysis was conducted by ANOVA two-tailed test for multiple comparisons using Tukey-Kramer's or Dunnett's approach, as indicated: *** p ≤ 0.001 highly significant; ** p ≤ 0.01 very significant; * p ≤ 0.05 significant; not significant (n.s.) p > 0.05. The figures were assembled using Corel Photo Paint, Corel Draw and BoxPlotR software.

Clonal Analysis
The FLP/FRT system [61] was applied for generating twin clones as described [43,62,63] by crossing the FRT82B H ∆CT fly line with y 1 w* hs-flp; FRT82B Ubi-GFP S65T nls/TM6B (BL32655, obtained from BDSC, Bloomington, IN, USA); the wild-type allele hence expressed GFP. Clones were induced by a 1 h heat shock at 37 • C in first to second instar larvae. Wing imaginal discs were dissected from wandering third instar and subjected to antibody staining according to standard procedures [62,63] using primary rat anti-Deadpan antibodies (Dpn, 1:100; ab19573 Abcam, Cambridge, UK) and goat secondary antibodies coupled to Cy3 (Jackson Immuno-Research via Dianova, Hamburg, Germany). GFP signals were recorded without further enhancement. Fluorescently labeled tissue was embedded in Vectashield (Vector labs, Eching, Germany) and pictures taken with a BioRad MRC1024 confocal microscope coupled to a Zeiss Axioskop using LaserSharp 2000 TM software (Carl Zeiss, Jena, Germany). The figures were assembled using Image J, Corel Photo Paint and Corel Draw software.

Yeast Two-Hybrid Experiments
Yeast two-hybrid experiments were performed as previously described using standard protocols [63][64][65][66][67][68]. As bait, we used full-length pEG H [14] and the deletion constructs pEG NTCT, pEG NT and pEG CT [23], pEG NTCT ∆ NT and pEG NTCT ∆ CT [19]. pJG Su(H) [14] and pJG Asf1, which was amplified from genomic DNA and inserted as Eco RI/Xho I fragment into pJG4-5 [66], served as prey. To this end, the following primers were used: Asf1-up 5 GCA ATC TCC GAG GTG AAT TCA TGG 3 and Asf1-lo 5 CGG ACT GCC TCG AGC TCT CAA CA 3 . Empty vectors pEG202 [67] and pJG4-5 [66] served as control. The expression of the lacZ reporter from pSH18-34 resulted from productive interaction [68]. The Asf1-HA tagged region was derived from the pJG4-5 yeast vector providing an N-terminal HA-tag by PCR using the upper primer 5 GGA GAT AGA TCT TAC CCT TAT GAT G 3 with a Bgl II restriction site and the M13 reverse primer. The Bgl II/Hind III digested PCR product was inserted into the Bam HI/Hind III sites of the pMAL vector.

Protein Expression
Maltose-binding protein MBP fusion proteins encoded by pMAL vectors and Glutathione S-transferase GST fusion proteins encoded by pGEX-2T vector [69] were expressed in Escherichia coli according to standard procedures [70][71][72]. To this end, pMAL constructs encoding the NTCTmyc variants and Asf-HA, respectively, and Su(H) (codons 288-594) in pGEX [11] were transformed into E. coli UT580 [71] and grown to log phase. Expression was induced with 1 mM IPTG at 18 • C overnight and bacteria lysed by French press. MBP fusion proteins were affinity purified using amylose beads (New Englands Biolabs, Ipswich, MA, USA) and eluted with maltose in 500 µL fractions as outlined in Ref. [73]. Likewise, the Su(H)-GST fusion protein was affinity purified on glutathione-sepharose 4B beads (GE Merck, Darmstadt, Germany) as described before [74]. The protein content of the fractions was measured with a Bradford assay and correct protein expression confirmed by PAGE.

Myc-Trap Binding Assay
The Myc-trap binding assay was performed according to the manufacturer's protocol (ChromoTek, Planegg-Martinsried, Germany) using anti-Myc-tag Nanobody/V H H coupled magnetic agarose resin (Myc-Trap ® Nanobodies). Purified proteins, Asf1-HA-MBP or Su(H)-GST were gently mixed with respective NTCTmyc-MBP peptides in a 1:1 ratio for about 10 min at 4 • C in a total volume of 500 µL of wash solution (150 mM NaCl, 50 mM Tris pH7.5, 0.1% SDS, 1 tablet of cOmplete TM protease inhibitor per 10 mL (Roche Merck, Darmstadt, Germany)). Subsequently, 25 µL prewashed Myc-Trap magnetic beads were added per assay for 1 h at 4 • C under rotation. Beads were separated on a magnetic stand and washed three times before the elution of the proteins with 50 µL of 3× SDS-loading dye plus 0.125 M DTT. Probes were boiled for 10 min, spun briefly, before loading 15 µL of the supernatant onto SDS gel for electrophoresis and subsequent Western blotting. Blots were probed with the following antisera, mouse anti-Myc (1:500, 9B11 Cell Signaling Technology, Danvers MA, USA), rat anti-HA (1:500, 11867423001 Roche Merck, Darmstadt, Germany) and goat anti-GST (1:5000, 27,457,701 GE Healthcare, Munich, Germany), to detect the tagged H, Asf1 and Su(H) proteins, respectively.

In Vivo Protein Analysis
For in vivo protein detection on Western blots, proteins were extracted from 15 wandering third instar larvae each, homozygous for H gwt , H ∆CT and y 1 w 67c23 , exactly as outlined before [22]. PageRuler TM Plus prestained protein ladder, 10-250 kDa, was used (Thermo Scientific, Waltham MA, USA), allowing blots to be cut at around 100 kDa. The upper part was treated with polyclonal rat anti-H antisera (h5, 1:250 [75]), the lower with monoclonal mouse anti-β-Tubulin A7 (1:500; developed by M. Klymkowsky, obtained from the Developmental Studies Hybridoma Bank DSHB, created by the NICHD of the NIH and maintained at The University of Iowa, Iowa City, IA, USA) as loading control. Secondary goat antibodies coupled to alkaline phosphatase (1:1000) were used for detection (Jackson Immuno-Research Laboratories via Dianova, Hamburg, Germany).

Isothermal Titration Calorimetry
Isothermal calorimetry was performed as described earlier [19,76]. D. melanogaster Asf1 protein (amino acids 1-154) was overexpressed and purified from bacteria using a modified pET-14p construct containing an N-terminal His-tag and C-terminal Strep-tag. Asf1 was purified using a combination of affinity and size exclusion chromatography. Hairless constructs (232-358, 315-358, 232-338 and 339-358) were overexpressed and purified as His-SUMO fusion proteins from a modified pET-28b+ vector and purified using affinity and size exclusion chromatography as previously described [20]. ITC experiments were carried out using a MicroCal VP-ITC calorimeter. All experiments were performed at 25 • C in a buffer composed of 20 mM HEPES pH 7.5 and 150 mM NaCl. The purified proteins were degassed and buffer-matched using dialysis and/or size exclusion chromatography. A typical experiment consisted of 100 µM Hairless in the syringe and 10 µM Asf1 in the cell. Protein concentrations were determined by UV absorbance at 280 nm. At least three independent experiments were performed. The data were analyzed using ORIGIN software and fit to a one-site binding model.

Gene Engineering to Generate the H ∆CT Allele, Specifically Lacking the CT Domain Only
Based on previous studies, the functions of the conserved domains of the Hairless protein were assigned to the binding of Su(H), the co-repressors Gro and CtBP and to nuclear translocation [7,[11][12][13][14]22]. However, the function of the highly conserved CT domain, located directly C-terminal to the Su(H) binding domain NT, remained unknown ( Figure 1a). To elucidate the role of the CT domain, we first generated an UAS-attB-H ∆CT line and used it in tissue-specific overexpression studies via the Gal4-UAS system [45]. However, the results were inconclusive, since no conspicuous phenotypic differences were observed compared to wild-type H control ( Figure S1), suggesting that there is little difference between the two transgenes. Nonetheless, ectopic expression leads to dramatically altered phenotypes and tissue loss, such that subtler functional differences may go unnoticed ( Figure S1) [19,23,26,[77][78][79][80]. Hence, we engineered a novel H allele H ∆CT , lacking the CT domain, for a detailed functional analysis of the CT domain on H activity. Since an intron is splitting the CT box in two parts, the cloning design was complex ( Figure 1). In the first step, we removed the intron plus large parts of the CT box to eventually reintroduce the intron sequences including splice sites (Figure 1a,c). Hence, the resulting gene construct is identical to the wild type apart from the small deletion of 38 codons, covering most of the conserved CT domain (Figure 1b). After sequence verification, H ∆CT genomic DNA was integrated into the H attP knockout line by genome engineering to generate the new H ∆CT allele as outlined previously [43]. A stable line was generated after floxing the white+ gene marker; it turned out that H ∆CT is homozygous viable, indicating that the CT domain is not absolutely required for the repressive H function, whereas a loss of H activity results in larval/pupal lethality [17,43,81,82]. A Western Blot analysis revealed the expression of the shortened H ∆CT protein isoform, including the smaller variant derived from internal ribosome entry, indicative of correct splicing ( Figure S2) [75,80,83].

The H ∆CT Allele Displays H Gain of Function Phenotypes
Homozygous H ∆CT flies are viable and fertile. Only upon close inspection, the phenotypic alterations became apparent. As a control, we used y 1 w 67c23 representing the genomic background of the stocks, as well as H gwt , which is a genome-engineered control harboring the genomic wild-type DNA [43]. Compared to these controls, we observed an increase in the number of microchaetae, the small mechano-sensory bristle organs that cover the fly thorax [59,84,85]. To analyze this phenomenon more precisely, scanning electron micrographs were taken and the microchaetae counted in a field between intrascutal suture and posterior dorsocentral macrochaetae (Figure 2a,b). The quantification showed a highly significant increase in microchaetae numbers in the H ∆CT flies compared to the controls (Figure 2c). In contrast to macrochaetae, whose number and position is strictly defined, the number of microchaetae is also variable in wild-type flies, determined, e.g., by thorax size [59,[84][85][86]. During pupal development, microchaetae are positioned in distinct rows; their spacing is controlled by lateral inhibition governed by Notch signal activity [59,87,88]. Accordingly, a downregulation of the Notch activity results in more densely spaced bristles, whereas its upregulation causes balding [12,82,[89][90][91]. Apparently, the H ∆CT allele displayed a mild gain of function phenotype in agreement with a slight downregulation of Notch activity.  in (a,b). (c) Microchaetae counts from y 1 w 67c23 and H gwt for control (n = 14) and from H ∆CT (n = 25) were plotted. Center lines show the medians, empty circles the means; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, respectively. Significance was determined with ANOVA two-tailed Dunnett's test for multiple comparisons; significant differences are observed between both the controls and H ∆CT (*** p ≤ 0.001).
The second remarkable phenotype concerns wing development, since a large fraction (81%) of H ∆CT homozygotes developed additional wing vein material. Small veinlets appeared mostly within the marginal cell, i.e., between the first and the second longitudinal veins, and within the second posterior cell next to the posterior cross vein (Figure 3a,b). In about 3% of the flies, wings displayed several ectopic veinlets also within the submarginal cell, i.e., between the second and third longitudinal veins, and within the third posterior cell next to L5 (Figure 3a-c). The wing size appeared somewhat smaller than in the control; however, the differences were not significant. Similar to the spacing of microchaetae, vein formation is controlled by Notch signaling activity applying mostly to vein thickness but also to veinlet formation [92][93][94]. For example, heterozygous null alleles of H lack veins partially due to the gain of Notch activity, whereas Notch heterozygotes display thickened veins. A second wing phenotype results from a downregulation of Notch activity, namely notches in the wing margin, which the pathway derives its name from [1,82]. The cells giving rise to the wing margin form the dorso-ventral boundary of the wing anlagen and act as a source for morphogens regulating patterning and growth. These cells are fated by Notch activity, which induces the expression of several specific target genes, including Deadpan (Dpn) [93,95]. As no wing incisions were observed, we asked whether the H ∆CT mutation does affect Dpn expression. To this end, we applied a mosaic analysis [61], generating homozygous H ∆CT cell clones to be compared with homozygous wild-type cells for Dpn protein expression (Figure 3d-d"). As expected for a gain of H repressive activity, Dpn protein accumulation was lowered in H ∆CT mutant cells compared to the sibling cells. In sum, the phenotypes displayed by the H ∆CT homozygotes are in line with a stronger repressive activity, suggesting that the CT domain somehow attenuates normal H repressor function.  , d,d"). The H ∆CT mutant cell clones are unmarked (outlined by a dashed line). Dpn protein expression along the dorso-ventral boundary is shown in red (arrowheads in d,d'). Compared to wild-type clones (open arrowhead), Dpn protein is lowered in the H ∆CT mutant cell clones (closed arrowhead). Scale bar, 20 µm.

The H ∆CT Allele Attenuates Cell Fating Defects
Apart from the above phenotypes concerning microchaetae and wing development, the H ∆CT homozygotes appeared to be wild type. For example, the number of large bristles, i.e. the macrochaetae, was normal (Figure 4a-c). If the H ∆CT allele were in fact a gain of H function, it should show a positive genetic interaction with a null allele of H. To test this idea, we performed crosses with H attp lacking any H activity [43] and compared the resultant phenotypes with H attp crossed to H gwt for a control. In the case of a gain of function, we expected an ameliorated phenotype in the H attp /H ∆CT flies compared to the H attp /H gwt heterozygotes. H heterozygous flies display a dominant loss of bristle phenotype, affecting both micro-and macrochaetae. In addition to the loss of an entire mechano-sensory organ, sometimes, a transformation of the outer shaft to socket cell is observed, giving rise to a so-called double socket phenotype [43,58,91,96] (Figure 4d,d'). In our analysis, we concentrated on the forty large bristles, as their number and position are a constant trait [84][85][86]. About fifteen macrochaetae were affected in H attp /H gwt control flies, on average, but only about nine in the H attp /H ∆CT flies, in agreement with a significant gain of H activity (Figure 4e,f). A closer look, however, revealed that bristle loss was indistinguishable between H attp /H ∆CT and control (Figure 4g). Accordingly, the phenotypic difference incorporated almost exclusively cell transformations, which were virtually absent in H attp /H ∆CT heterozygotes (Figure 4e,h). This result was rather unexpected, as it may hint at a specific role for the CT domain during cell fate selection. Alternatively, dose differences could also be the explanation for these phenotypic differences. In this case, a minute increase in H activity may suffice to allow cell fate distinction but not sensory organ precursor selection.

Complex Genetic Interactions between Asf1 and Hairless
The above results demonstrate that the conserved CT domain somehow mitigates H activity, raising the possibility that it may serve as a binding site for respective factors. One binding partner of H to be considered in this context is Asf1. Earlier, it was shown that Asf1 promotes Notch repression together with H, albeit the molecular mechanisms are unknown [28]. The ectopic expression of Asf1 within the eye anlagen was shown to reduce eye size, ameliorated by the concomitant loss of one H gene dose, suggesting that Asf1 and H collaborate in repressing N activity [28]. In fact, Drosophila eyes are somewhat enlarged in H attP heterozygous null mutants and likewise upon tissue-specific downregulation of H activity during eye development by RNA interference ( Figure S3). This phenotype is in accordance with an overactivation of the Notch pathway inducing growth [97][98][99]. On the other hand, induction of H in the eye anlagen confounds eye development, resulting in pupal lethality ( Figure S3). In this instance, growth is inhibited, and cell death is induced [23,[100][101][102]. Moreover, we observed a significant amelioration of the Asf1-induced small eye phenotype by a loss of H activity ( Figure S3), in agreement with the published data [28]. Whether this apparent rescue is based on direct genetic interactions or on additive effects, however, remains unclear. Hence, we asked the question whether a downregulation of Asf1 by RNA interference would allow for a more comprehensive analysis. We also included the other known co-repressors that bind H, namely Gro and CtBP, in the hope of gaining additional functional insights ( Figure S4). However, the results were inconclusive because the RNAi-mediated downregulation of Asf1 in the developing eye resulted in smaller eyes similar to its ectopic expression. Perhaps the loss of Asf1 activity induced cell death. This explanation seems likely, firstly because Asf1 homozygosity was reported to be cell lethal [28], and secondly because the respective eyes display typical signs of cell death ( Figure S4b,c). Specificity of the Asf1-RNAi was confirmed by the concurrent induction of Asf1, and these flies displayed near wild-type eyes (Figure S4b,c). Quantification of the phenotypes indicated some rescue of the Asf1-RNAi-induced small eye phenotype by loss of H activity, which again could be an additive effect or a specific counteraction of a gain of Notch activity against cell death induction. Moreover, downregulation of either co-repressor Gro or CtBP slightly enhanced the eye phenotype as well, perhaps due to increased cell death or a further block of tissue growth ( Figure S4).

Mapping the H-Asf1 Interaction Domain by Yeast Two-Hybrid Assays
Earlier reports demonstrated a physical interaction of the full-length Asf1 and H proteins by pull-down assays [28]. We aimed to define the Asf1 interaction domain in H by yeast two-hybrid analyses using a number of deletion constructs for H. The analysis, however, was hampered by the extremely weak interactions between the two when compared to the positive control Su(H). We were unable to see any binding between the full-length proteins and only extremely weak binding of Asf1 to the isolated NTCT domain even after prolonged exposure ( Figure 5). The NTCT domain is highly conserved among insect species and is even found in the H protein from the arthropod D. pulex [23,24,103]. NTCT can be split into N-terminal NT and C-terminal CT domains; the NT domain is sufficient for the binding to Su(H), whereas no previous function has been assigned to the CT domain [19,20,23] (Figure 5). We noted a weak binding of Asf1 to the isolated CT domain and a rather robust binding to the NTCT ∆ NT peptide. Specificity of the interaction was confirmed by a lack of binding to the NTCT ∆ CT peptide, which, on the contrary, bound well to Su(H) as expected ( Figure 5). In fact, Asf1 is the first protein identified to potentially contact this conserved region of the H protein. The weak interaction between the two, however, is puzzling in light of the previous reports [28]. One reason could be the high conservation of Asf1 proteins in eukaryotes. Even the Asf1 orthologs from yeast and fly share almost 60% sequence identity in the active domain (residues 1-156) ( Figure S5) [104,105]. Potentially, the endogenous yeast Asf1 outcompetes the Drosophila Asf1-AD protein expressed from pJG, thereby lowering productive reporter activation. Hence, we sought other technical approaches for mapping the H-Asf1 interface.  in (b)). GBD, Groucho-binding domain; CBD, binding domain for CtBP. Numbering refers to amino acids of full-length H protein [17,75]. (b) Interaction tested between Asf1 and specific fragments of Hairless as indicated in the schemes. Size of fragments is given in codons for NTCT, NT and CT; size of deletion within NTCT given in codons for NTCTD ∆ NT and NTCT ∆ CT. Empty vectors (pEG, pJG) served as negative and Su(H) as positive controls. Protein interaction is revealed by blue coloration of yeast colonies. Note the different incubation times to observe conspicuous coloration: overnight for the Su(H)-H interactions, however, 48 h incubation for the Asf1-H interactions.

Interaction Assays Using Tagged Hairless and Asf1 Proteins
Next, we generated several Hairless NTCT variants that were tagged with Myc, as well as full-length Drosophila Asf1 tagged with HA. The respective proteins were expressed in bacteria as MBP fusion proteins and purified by affinity chromatography. As a positive control, a GST Su(H) fusion protein containing the H-binding C-terminal domain CTD was used [11]. For binding assays, the purified Asf1 or Su(H) proteins were mixed with H peptides and the Myc-tagged H proteins trapped with magnetic beads. After washing and elution, the eluates were analyzed in Western blots. The presence of the respective proteins in the H precipitates was probed with antibodies directed against myc, HA or GST ( Figure 6). As predicted from the above experiments, we found robust binding of Asf1 to NTCTmyc but not to NT ∆ CTmyc, whereas both peptides co-precipitated with Su(H) as expected (Figure 6a,b). To map the binding domain more exactly, we subdivided the conserved CT domain into several shorter peptides, either containing the most highly conserved sequences only (aa 295-317) or starting further in the C-terminal, including conserved basic residues (aa 315-358 and aa 315-369) (Figure 6c,d; compare with Figure 1b). Unexpectedly, only the two C-terminally extending peptides bound to Asf1, despite their moderate expression, whereas the small peptide containing the N-terminal part of the conserved CT domain failed to do so (Figure 6c). From the current analysis, we tentatively map the H residues involved in Asf1 binding to amino acid residues 315-338. This is based on the assumption that (1) fragmentation does not interfere with higher order protein structure (i.e., every peptide assumes its correct fold) ( Figure S6) and (2) that the binding site is restricted to a few isolated amino acids.

Thermodynamic Analysis of the Hairless and Asf1 Interaction
To provide an orthogonal approach to validate H/Asf1 binding, as well as further map the interacting regions and quantitatively measure binding, we used isothermal titration calorimetry with purified recombinant proteins of H and Asf1. As shown in Figure 7 and Table 1, H (232-358) bound to Asf1 (1-154) with~4 µM affinity and a 1:1 stoichiometry. Similarly, H (315-358) bound Asf1 (1-154) with a comparable K d of~2 µM. However, when we further truncated our H construct to residues 232-338 or 339-358, respectively, we were no longer able to detect binding by ITC. Taken together, these data confirm the above results: the domain containing aa 315-338 appears absolutely required for the binding of Asf1; however, it is not sufficient on its own to support binding. Only H (315-358) was firmly established to contain the Asf1 binding domain.

The Asf1-Induced Small Eye Phenotype Is Enhanced in H ∆CT Background
Earlier work suggested that Asf1 overexpression promoted Notch target gene repression most likely via Hairless [28]. Accordingly, an increased repression of Notch activity during eye development resulting from ectopic Asf1 protein could affect eye growth, resulting in smaller eyes. If indeed this phenotype depended on the combined activity of Asf1 and H, it should no longer be observed in the H ∆CT background, where Asf1-H complexes cannot form due to the lack of the Asf1 binding site in H. To address this idea, we combined H ∆CT with the ey-Gal4 UAS-Asf1 expression line and noted enhanced phenotypes instead of a rescue, independent of sex ( Figure 8). Quantification of the eye size confirmed the visual impression; whereas the eye size of H ∆CT matched the control, the ectopic expression of Asf1 during eye development reduced eye size by roughly ten percent, and more than twenty percent in the H ∆CT background (Figure 8), again independent of sex. These data clearly illustrate genetic interaction between H and Asf1-yet, in an unexpected manner. Evidently, H ameliorates Asf1 repressive activity, since in the absence of H binding, Asf1 repressive activity increased, resulting in reduced growth (Figure 8).

Discussion
In this work, we applied genome engineering to generate a novel H ∆CT allele, encoding an H protein lacking the conserved CT domain, allowing us to specifically address CT's biological function. Unexpectedly, H ∆CT flies were viable without restriction, suggesting that the CT domain is not essential for H repressor function during Notch signaling. Moreover, the subtle gain of function phenotypes displayed by H ∆CT flies imply that the CT domain somehow attenuates normal H repressor activity. Of note, these gain of function phenotypes are largely restricted to the process of lateral inhibition during the selection of mechano-sensory organ precursors and the specification of wing veins (Figures 2 and 3). During lateral inhibition, cells are selected from a field of equipotential cells with the same fate, i.e., pro-neural or pro-vein fate, to differentiate correspondingly. The remaining cells refrain from the respective fate due to Notch signaling activity, involving a transcriptional feedback loop that amplifies small differences between the cells [87,[106][107][108]. In this context, H protects the presumptive sensory organ precursors or pro-vein cells from spurious Notch signals [109][110][111]. Our data suggest that H ∆CT is more potent in doing so, thereby allowing more cells to choose the primary fate, resulting in extra sensory organs, as well as ectopic veinlets (Figure 3a-c). This may reflect a specific activity of H ∆CT , for example, in feedback regulation, or just a slightly stronger activity overall. A simple explanation could be that CT deletion helps stabilize the H protein. Dose sensitivities for Notch signaling components including H are well described [17,82,112,113]. Notably, bristle numbers are highly susceptible to phenotypic plasticity due to genomic fluctuations, where H plays an important role [114][115][116]. Our recent studies, however, give no indications that the CT domain might contribute to H stability or might contain a degron ( Figure S2) [32,62]. An additional hint of a specific regulatory role of the CT domain comes from the observation that the H ∆CT heterozygotes did not display the characteristic double socket phenotype of the H heterozygotes ( Figure 4). This phenotype is caused by a transformation of prospective shaft into socket cells [91,96]. Shaft and socket cells are siblings derived from an asymmetric cell division; they are differentially fated via a Notch signal in the presumptive socket cell, which induces a specifying, auto-regulatory Su(H) activity [96,[117][118][119]. In the H heterozygotes, the Su(H) auto-activation also takes place in presumptive shaft cells, causing a fate change, and hence, apparent transformation into a socket cell [89,96,118,119]. The H ∆CT heterozygous condition, however, allows for shaft cell development, i.e., the Su(H) autocatalytic feedback loop is restricted to the presumptive socket cell. Hence, H ∆CT clearly differs from the wild type regarding its regulatory capacity. Whether this is based on dose or activity differences remains to be established.
Our in vitro protein interaction experiments demonstrated that H ∆CT lacks Asf1 binding (Figure 6a,d). The ∆ CT deletion comprises only 38 amino acids, of which codons 315-338 are indispensable for H-Asf1 binding. The minimal sequences sufficient for Asf1 binding are contained within 44 amino acids extending further C-terminal up to codon 358, with an overlap of 23 highly conserved amino acids (Figure 1b) [23]. Splitting this region apart did not support any Asf1 binding, suggesting an underlying structural entity; however, structural analyses of H-Asf1 complexes are currently lacking. In silico models of the H NTCT domain predict a β sheet followed by an α helix, none of which are sufficient for the binding (Figure 7 and Figure S6) [120]. As the CT domain does not influence the binding of Su(H), the repressor complex formation itself should be unaffected. This raises the possibility that CT may influence the binding dynamics of the repressor complex on chromatin, for example, via Asf1. Earlier reports showed that Asf1 contributes to the repression of Notch target genes, since RNAi-mediated knock-down caused an upregulation of Notch target genes, notably of E(spl)m3, E(spl)m7 and E(spl)mγ [28]. Moreover, local induction of Asf1 in the eye and wing anlagen was accompanied by tissue loss, in line with a downregulation of Notch growth-promoting activities [28,97]. The physical interactions between H and Asf1 prompted a model, whereby Asf1 was recruited to the Notch target genes via H, promoting transcriptional silencing [28]. Clearly, based on our studies herein, the regulation mediated by H and Asf1 is more complex. Overall, our genetic data reveal a mutual negative relationship between H and Asf1, i.e., both attenuate each other's repressive activity in specific developmental contexts. On the one hand, growth retardation induced by the ectopic expression of Asf1 was enhanced in the H ∆CT background, implying that, normally, H impairs Asf1 repressor activity in the context of overexpression ( Figure 8). Perhaps H, in contrast to H ∆CT , is able to sequester ectopic Asf1, thereby reducing the detrimental effects on growth. On the other hand, Asf1-binding-defective H ∆CT is a stronger repressor than wild-type H, suggesting that Asf1 binding may reduce H repressor activities (Figures 2 and 3). Conceivably, Asf1 impedes the dynamics of H-Su(H) repressor complex exchange by chromatin modifications [28,34,38]. The counteractive function of Asf1 may be restricted to the transient period of interactions between the H-Su(H) repressor complex and the promoter region before stable silencing occurs. We envisage that during the process of lateral inhibition, when cell fates are not yet fixed, a flexible regulatory input is needed to allow the transcriptional feedback loop to take place for signal amplification, eventually generating the differential outcome [87,106]. The activity of Asf1, however, may induce premature stable chromatin states no longer amenable to a dynamic regulation. Alternatively, the recruitment of Asf1 to the Su(H) repressor complex is primarily mediated by Skip (Bx42 in Drosophila) and, to a lesser degree, by H [28]. In mammalian cells, Skip binds to the Su(H) homolog CBF1, acting as a hub for activating and repressing co-factors [121,122]. Subsequently, Asf1 may spread into the transcribed gene region, increasing nucleosome density, in accordance with its well-documented role in nucleosome assembly in ORFs during gene silencing [42]. Accordingly, RNAi-mediated depletion of H caused Asf1 disappearance not only from the promoter regions but also from the ORF of E(spl)m7 and E(spl)m3 genes [28]. Overall, our work adds to the growing understanding of the concerted action of transcriptional regulators, histone modifiers and chromatin remodelers to organize chromatin and transcription.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/genes14010205/s1, Figure S1: Overexpression of H and H ∆CT ; Figure S2: In vivo H ∆CT protein expression; Figure S3: Small eyes induced by ectopic Asf1 expression are rescued by a loss of H activity; Figure S4: Eye-specific downregulation of Asf1 results in smaller eyes; Figure S5: Asf1 is extremely well conserved in eukaryotes from yeast to fly; Figure S6: Secondary structure prediction for NTCT.  Data Availability Statement: Not applicable. All data are contained within the manuscript. collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.