A Drosophila Su(H) model of Adams-Oliver Syndrome reveals cofactor titration as a mechanism underlying developmental defects

Notch signaling is a conserved pathway that converts extracellular receptor-ligand interactions into changes in gene expression via a single transcription factor (CBF1/RBPJ in mammals; Su(H) in Drosophila). In humans, RBPJ variants have been linked to Adams-Oliver syndrome (AOS), a rare autosomal dominant disorder characterized by scalp, cranium, and limb defects. Here, we found that a previously described Drosophila Su(H) allele encodes a missense mutation that alters an analogous residue found in an AOS-associated RBPJ variant. Importantly, genetic studies support a model that heterozygous Drosophila with the AOS-like Su(H) allele behave in an opposing manner to heterozygous flies with a Su(H) null allele, due to a dominant activity of sequestering either the Notch co-activator or the antagonistic Hairless co-repressor. Consistent with this model, AOS-like Su(H) and Rbpj variants have decreased DNA binding activity compared to wild type proteins, but these variants do not significantly alter protein binding to the Notch co-activator or the fly and mammalian co-repressors, respectively. Taken together, these data suggest a cofactor sequestration mechanism underlies AOS phenotypes associated with RBPJ variants, whereby the AOS-associated RBPJ allele encodes a protein with compromised DNA binding activity that retains cofactor binding, resulting in Notch target gene dysregulation.

Introduction Notch signaling is a highly conserved pathway that mediates cell-to-cell communication in all metazoans [1][2][3]. During embryogenesis, Notch plays a crucial role in vasculogenesis, hematopoiesis, neurogenesis, and cardiac development [4]. Additionally, Notch regulates tissue homeostasis, including epidermal differentiation and maintenance, lymphocyte differentiation, muscle and bone regeneration, and angiogenesis [4]. Intriguingly, Notch regulates these diverse processes using a common molecular cascade [5] (Fig 1A). Signaling is activated when Notch receptors on a cell interact with DSL (Delta, Serrate, Lag-2 for mammalian, Drosophila, and C. elegans orthologs, respectively) ligands on a neighboring cell. Mammals encode four Notch receptors (Notch1-4) and five DSL ligands (Dll1,3,4 and Jag1,2), whereas Drosophila has one Notch receptor and two ligands (Delta and Serrate). The Notch-DSL interaction triggers DSL endocytosis, resulting in force induced Notch cleavage [6], which is mediated by ADAM10 and the γ-secretase complex. Once cleaved, the Notch intracellular domain (NICD) is freed from the cell membrane, transits to the nucleus, and forms a complex with the transcription factor CSL (CBF1/RBPJ, Su(H) and Lag-1 for mammalian, Drosophila, and C. elegans orthologs, respectively) and the co-activator Mastermind (MAM) [7]. The NICD/CSL/MAM (NCM) complex binds to enhancer and promoter DNA elements to activate gene expression [8]. However, CSL can also function as a transcriptional repressor by forming complexes with co-repressor proteins, such as SHARP and Hairless in mammals and Drosophila, respectively [9][10][11]. Hence, Notch signal strength in a cell is largely defined by the balance of NCM activating complexes and CSL/co-repressor complexes that regulate target gene expression.
Genetic studies have shown that a subset of Notch-dependent processes are highly sensitive to gene dose. The term Notch derives its name from the original notched wing phenotype identified in Drosophila, resulting from Notch receptor haploinsufficiency. Heterozygous Notch phenotypes are also observable in the number and spacing of bristles on the fly notum. In knockout mice studies, Dll4 heterozygous-null alleles produce a severe phenotype with most pups dying in utero [12]. Moreover, human birth defects and developmental syndromes Overview of the Notch signaling pathway. Signal sending cells that express Notch ligands (DSL) interact with adjacent signal receiving cells that express Notch receptors. Upon receptor-ligand binding, the Notch receptor is cleaved by γ-secretase, resulting in the release of the soluble Notch Intracellular Domain (NICD). NICD binds to the CSL transcription factor and recruits the co-activator Mastermind (MAM) to activate transcription of Notch target genes. The CSL transcription factor can also engage co-repressors (CoR) to inhibit transcription. B. Sequence traces from PCR amplified genomic DNA have been linked to haploinsufficiency for three out of the four Notch receptors (NOTCH1/2/ 3), two out of five ligands (DLL4 and JAG1), and the RBPJ transcription factor (human CSL) [13]. As an example, variants in the Notch pathway underlie Adams-Oliver Syndrome (AOS), a rare disease defined by scalp aplasia cutis congenita (missing skin and skull tissue) and limb malformations. Additionally, AOS patients can suffer cardiovascular abnormalities, such as cutis marmorata telangiectatica (dilated surface blood vessels), hypertension, and heart defects [14]. Genetic studies have revealed that~40% of AOS patients inherit dominant mutations within the NOTCH1 receptor, the DLL4 ligand, or the RBPJ transcription factor. In contrast, haploinsufficiency mutations in the NOTCH2 receptor and JAG1 ligand are associated with Alagille Syndrome, a disease characterized by liver, eye, kidney, heart, skeleton, and vasculature defects [15]. Hence, genetic variants within the Notch signaling pathway can cause developmental defects in organisms from flies to humans.
In 2012, Hassad et. al. reported two autosomal dominant variants in RBPJ in separate families with AOS [14]. Interestingly, this study was the first to report germline variants in RBPJ that cause a disease in humans. These variants (E63G & K169E) mapped to the DNA binding region of RBPJ (Fig 1C and 1D), and the authors used cell extracts to show each variant impaired DNA binding [14]. Hence, these variants were classified as loss-of-function alleles. Here, we show that a previously identified allele in Su(H) (Drosophila CSL), Su(H) T4 , encodes an E137V mutation at the same highly conserved glutamic acid residue (human RBPJ E63G) as seen in AOS patients. However, prior genetic studies showed that Su(H) T4 has conflicting "gain-of-function" activity as heterozygotes in Drosophila and "loss-of-function" activity in homozygous mutant clones [16,17]. To better explain the discrepancy between these conflicting biochemical and genetic results, we used a combination of quantitative Drosophila genetic studies, DNA binding assays, and proteinprotein interaction assays. Our data show that in heterozygous flies the AOS-like Su(H) allele and a Su(H) null allele behave in a largely opposing manner, as the Su(H) AOS-like allele exacerbated both Notch and Hairless phenotypes, whereas the Su(H) null allele partially suppressed Notch and Hairless phenotypes. Moreover, we found that the Su(H) alleles and the mammalian RBPJ AOS variant alleles encode proteins defective in DNA binding. However, the AOS-like variants bound both the Notch activation complex and the Hairless/SHARP repression complexes with similar affinities as wild type Su(H)/RBPJ. Altogether, these data suggest a sequestration mechanism, in which RBPJ/Su(H) AOS variants efficiently bind the Notch pathway co-activator and co-repressor proteins, but their reduced DNA binding activity excludes these complexes from DNA and results in the mis-regulation of Notch target genes.

Identification of a Drosophila Su(H) allele with an analogous amino acid change as a human RBPJ variant associated with AOS
Genetic studies in Drosophila have previously identified several Su(H) alleles that were described as having dominant gain-of-function phenotypes [17]. To better understand the isolated from flies heterozygous for the Su(H) T4 and Su(H) O5 alleles reveal missense mutations (arrows) in each respective fly line. C. Top: Linear domain layout of CSL. NTD = N-terminal domain, BTD = beta-trefoil domain, CTD = C-terminal domain. Grey regions are disordered and poorly conserved between species. Bottom: Multiple sequence alignments of a portion of the human, mouse, and Drosophila NTD of CSL with residues that are 100% conserved highlighted in cyan. Yellow highlighting indicates the residues altered due to missense variants in two separate families with AOS (specific amino acid changes listed in red text) and the missense variants found in the Su(H) alleles (specific amino acid changes listed in green text). D. Middle: Ribbon diagram of RBPJ bound to DNA (PDBID 3IAG) [50], with the same domain coloring as described in C and the DNA shown in purple. Left: Close up view of the K169/195/243 residue in yellow, numbered according to human (red), mouse (blue), and Drosophila (green) respectively. Right: Close up view of the E63/89/137 and K58/84/132 residues in yellow, numbered according to the human (red), mouse (blue), and Drosophila proteins respectively. Black dashed lines indicate hydrogen bonding.
https://doi.org/10.1371/journal.pgen.1010335.g001 PLOS GENETICS molecular function of these alleles, we sequenced the Su(H) coding regions of the Su(H) T4 and Su(H) O5 alleles and found that both contain a single missense mutation within the same highly conserved domain (i.e. the N-terminal domain, or NTD) of Su(H), which is involved in DNA binding (Fig 1B). Su(H) O5 encodes a lysine to a methionine mutation (K132M), whereas Su (H) T4 encodes a glutamic acid to valine mutation (E137V). Intriguingly, the E137 Su(H) residue corresponds to E63 in human RBPJ, which was found to be encoded by a missense variant (E63G) in a family with Adams Oliver Syndrome (AOS) [14] (Fig 1C). However, unlike the dominant gain-of-function activity ascribed to the fly Su(H) T4 allele in heterozygous animals, the RBPJ E63G variant was proposed to be a loss-of-function allele due to loss of DNA binding activity. Based on the crystal structures of RBPJ and Su(H) bound to DNA, both AOS variants identified by Hassed et. al. [14] (E63G & K169E) and both alleles identified by Fortini et. al. [17] (K132M and E137V) are residues involved in specific and non-specific interactions with DNA (Fig 1D). It should be mentioned that while E63 and E137 correspond to the same glutamate residue in human and Drosophila RBPJ/Su(H), K169 and K132 are not homologous. Nonetheless, the discrepancy between the predicted gain-of-function activity of the Su(H) E137V variant in Drosophila and the predicted loss-of-function activity of the analogous human E63G variant provided an opportunity to reexamine the mechanisms underlying how the homologous Su(H) and RBPJ variants impact Notch signaling.

Su(H) and RBPJ AOS variants disrupt DNA binding
To elucidate how the RBPJ/Su(H) variants affect function, we first performed comparative DNA binding assays using EMSAs with a probe containing a RBPJ/Su(H) binding site and purified recombinant proteins corresponding to Drosophila Su(H) and mouse Rbpj (Fig 2A  and 2B, note we will refer to the mouse Rbpj protein with only the first letter capitalized, whereas human RBPJ is in all capital letters). We expressed and purified WT and AOS variants of mouse Rbpj  and Drosophila Su(H) (98-523), which corresponds to the highly conserved structural core of CSL proteins [7]. We used differential scanning fluorimetry to determine if the variants affect folding and overall stability of Rbpj/Su(H). Consistent with these residues being surface exposed, none of the variants significantly affected folding/stability of Rbpj/Su(H) compared to WT proteins except for mouse Rbpj E89G (analogous to human RBPJ E63G), which had a modest effect on its melting temperature (ΔT m = -3.1˚C, S1 Fig). EMSAs were performed with equimolar concentrations of WT Su(H) and the E137V [encoded by the Su(H) T4 allele] and K132M [encoded by the Su(H) O5 allele] variants, which revealed that both variants decreased, but did not abolish, DNA binding to a high affinity site (Fig 2A). Quantitation of the EMSA data revealed that Su(H) E137V consistently had a stronger impact on DNA binding than the Su(H) K132M variant. Similarly, purified Rbpj variants carrying analogous mouse amino acid variants (E89G and K195E) as those residues found in human associated AOS variants (E63G and K169E) also strongly weakened but did not abolish DNA binding (Fig 2B). In this case, however, K195E had an even greater impact on DNA binding than E89G (Fig 2B). These results are further supported by EMSAs performed with the fulllength mammalian proteins produced by cell-free in vitro transcription/translation (S2 Fig). Next, we used isothermal titration calorimetry (ITC) to quantitatively assess the changes in affinity and thermodynamics of these variant proteins binding to DNA (Fig 2C and 2D, S1 Table). Each binding experiment was performed by titrating DNA containing a single Rbpj/Su (H) binding site from a syringe into a calorimetric cell containing either purified Su(H) or Rbpj. WT Su(H) bound to DNA with a K d of 188 nM (Fig 2C, S1 Table); whereas Su(H) K132M bound to DNA with a 690 nM K d , showing an~3.5 fold reduction in binding, and Su (H) E137V bound to DNA with a K d of 842 nM displaying an~4.5 fold reduction in binding from WT (Fig 2C, S1 Table). WT mouse Rbpj bound DNA with a 74 nM K d , while the E89G and K169E variants had an~6-and 16-fold reduction in binding, respectively (Fig 2D, S1  Table). These trends in reduced DNA binding are consistent with our EMSA data that similarly showed that the Su(H) E137V is more severe than K132M and Rbpj K169E is more adversely affected than E89G (Fig 2A and 2B). From these experiments, we can conclude that the effects of these variants on DNA binding are conserved between the fly and human proteins, suggesting that the known Drosophila Su(H) alleles can be used as a developmental model to study RBPJ variants that cause AOS in humans.

The Su(H) AOS-like alleles exacerbate phenotypes associated with imbalanced Notch co-activator and Hairless co-repressor levels
Fortini and Artavanis-Tsakonas originally identified the Su(H) T4 and Su(H) O5 alleles in a genetic modifier screen of Notch pathway phenotypes [17]. In that study, they showed that flies heterozygous for these two Su(H) alleles altered Notch pathway phenotypes in sensitized backgrounds, and often in an opposing manner compared to flies that were heterozygous for Su(H) loss-of-function alleles. Hence, they classified these alleles as having "gain-of-function" activity. To better understand how each Su(H) allele impacts Notch and Hairless phenotypes in light of our new sequencing results, we quantitatively reanalyzed two well established Notch (N) sensitive phenotypes: wing nicking or "notches" that are found in Notch heterozygotes flies (N +/-); and macrochaetae sensory bristles, which increase in number with lower Notch levels (i.e. N +/-) but decrease in number with lower Hairless (H) co-repressor levels (i.e. H +/-).
First, we directly compared the impact of having either a "gain-of-function" Su(H) T4 or Su (H) O5 allele or the "loss-of-function" Su(H) IB115 allele encoding a non-functional protein due to a premature stop codon (K138-stop) [17,18]. Flies heterozygous for the Su(H) IB115/+ null allele in an otherwise WT genetic background developed normal wings (no nicks observed, N = 174; Fig 3A). However, consistent with prior studies [17], we found that a small subset of flies heterozygous for either the Su(H) T4 or Su(H) O5 allele had a dominant wing nicking phenotype and quantitative analysis revealed that flies with the Su(H) T4 allele had both a higher penetrance and increased wing nicking severity than the Su(H) O5 allele (Fig 3A and 3B). Thus, consistent with our quantitative in vitro DNA binding assays (see Fig 2), we found that the Su (H) T4 allele encoding the E137V variant has both a more dramatic impact on DNA binding and a stronger impact on wing nicking when compared to the Su(H) O5 allele encoding the K132M variant.
To determine if the wing nicking phenotype in the more severe Su(H) T4 allele could be suppressed by compensatory changes in the Notch pathway, we first used GFP-tagged BAC transgenes to add an extra copy of either WT Su(H) (Su(H)-GFP) or Notch (Notch-GFP) into flies heterozygous for the Su(H) T4 allele. Importantly, increasing the gene dose of both Su(H) (Su (H) T4 /+;Su(H)-GFP) and Notch (Su(H) T4 /+;Notch-GFP) suppressed the wing nicking phenotype (Fig 3A). Similarly, we found that removing a single copy of the Hairless gene that encodes a co-repressor that specifically antagonizes Notch pathway activation, also suppressed wing nicking in Su(H) T4/+ flies (Su(H) T4/+ ;H 1/+ ) (Fig 3A and 3C)  Next, we compared the ability of the Su(H) T4 , Su(H) O5 , and Su(H) IB115/+ alleles to modify the wing nicking phenotype in Notch heterozygous animals that are known to have significant wing nicking in an animal with two WT Su(H) alleles (Fig 3D and 3E). As expected, both the Su(H) T4 and Su(H) O5 alleles dramatically enhanced the wing notching phenotype in Notch heterozygous flies (i.e. N 55e11/+ ;Su(H) T4/+ and N 55e11/+ ;Su(H) 05/+ ) with the Su(H) T4 allele having a more dramatic impact on wing morphology (Fig 3D, 3F and 3G). In sharp contrast, the Notch haploinsufficiency wing phenotype was weakly, but significantly, suppressed in flies heterozygous for the Su(H) IB115 null allele (N 55e11/+ ;Su(H) IB115/+ ) (Fig 3D). Thus, while lowering the dose of Su(H) using a null allele can partially alleviate the wing phenotype caused by too little Notch, the presence of the DNA binding compromised Su(H) T4 and Su(H) O5 alleles exacerbates this phenotype.

PLOS GENETICS
These data show that flies heterozygous for the Su(H) T4 allele are susceptible to wing notching with a penetrance and severity that can be either enhanced in N 55e11 heterozygotes ( Fig  3D) or suppressed by adding an extra Notch allele (N-GFP in Fig 3A). Similarly, we found that the ability of the Su(H) T4 allele to induce wing nicking could also be suppressed by removing a Hairless allele (i.e. in Su(H) T4/+ ;H 1/+ compound heterozygotes in Fig 3A). To further test how Hairless gene dose impacts the wing nicking phenotype induced by the Su(H) T4 allele, we used a previously published Hairless BAC rescue transgene (H-GFP) [19] to add an extra Hairless allele in either wild type flies, Su(H) T4 Fig 3A). Taken together, these findings highlight how the gene dose of both Notch and Hairless can have a profound impact on the wing phenotypes generated by the Su(H) T4 allele.
The wing nicking phenotype in Notch heterozygous flies is thought to be caused by decreased Notch signal strength in the Cut-positive wing margin cells during larval development [20][21][22]. To better understand the impact of the Su(H) T4 and Su(H) O5 alleles on Notch signal strength within these cells, we performed comparative expression analysis of the E(spl) mα-GFP reporter gene in wing imaginal discs isolated from either WT, Su(H) IB1115 /+, Su (H) T4 /+ or Su(H) O5 /+ larva. E(spl)mα is a known Notch target in the wing imaginal disc and is expressed in Cut+ wing margin cells as well as in additional cells that contribute to sensory organ development [23,24] (Fig 4A). To assess Notch signal strength in wing margin cells, we performed quantitative GFP expression analysis using confocal microscopy (see Methods for details) and used a Cut-antibody to specifically label wing margin cells. Since E(spl)ma-GFP is activated very strongly in the sensory organ lineage that develops along the anterior wing margin by proneural transcription factors [25], we focused our analysis on the easily isolated and visualized posterior Cut+ margin cells (labeled posterior in Fig 4A, see diagram in Fig 4A' for location of wing margin cells and nearby sensory organs). Importantly, quantitative GFP analysis revealed that replacing a WT Su(H) copy with the Su(H) IB115 null allele did not significantly impact GFP levels, whereas both Su(H) O5 and Su(H) T4 heterozygotes had significantly lower GFP levels relative to WT and Su(H) IB115 heterozygotes (Fig 4B). These data are consistent with the AOS-like Su(H) T4 and Su(H) O5 alleles, but not the Su(H) IB115 null allele, being sufficient to lower Notch signal strength in wing margin cells, and thereby induce wing nicks.
Prior studies of flies heterozygous for either the Su(H) T4 or Su(H) O5 allele also found that both enhance the loss of macrochaetae sensory organs in flies heterozygous for the Hairless (H) co-repressor gene [17]. In contrast, and as its name implies, heterozygosity for a Su(H) loss-of-function allele significantly suppresses the decrease in macrochaetae formation in Hairless heterozygous animals [26,27]. We similarly tested each of these Su(H) alleles in flies with either two WT alleles of H (i.e. a yw 1118 stock) or in H 1/+ heterozygotes. In the WT background, flies heterozygous for the Su(H) IB115 null allele invariably developed the expected 40 macrochaetae on the head and thorax of each adult fly (Fig 5A). Flies heterozygous for the Su (H) O5 and Su(H) T4 alleles also developed an average of 40 macrochaetae. However, unlike in Su(H) IB115 heterozygotes, we observed a small number of flies in both Su(H) O5 and Su(H) T4 heterozygotes that either had 1 or 2 missing macrochaetae or had 1 extra macrochaetae ( Fig  5A). While the penetrance of these phenotypes is very low and the direction of change is variable (both increases and decreases in macrochaetae are observed), these data suggest that the process of macrochaetae selection and/or development may be less robust in Su(H) O5 and Su (H) T4 heterozygotes than in Su(H) IB115 heterozygotes.
Next, we tested each allele in the H 1 /+ heterozygotes that lose an average of~12 macrochaetae (Fig 5B-5F). As expected, replacing a WT Su(H) allele with the Su(H) IB115 allele significantly suppressed the loss of macrochaetae in compound heterozygous animals (Su(H) IB115 /+; H 1 /+, Fig 5B and 5D) [26,27]. In contrast, both the Su(H) O5 and Su(H) T4 alleles enhanced the loss of macrochaetae in H 1 /+ compound heterozygotes, and the Su(H) T4 allele again had a stronger impact than the Su(H) O5 allele (Fig 5B, 5E and 5F). Next, we analyzed the impact of each of these Su(H) alleles on macrochaetae phenotypes in flies sensitized to produce too

Fig 4. Flies with an AOS-like Su(H) allele have dramatically lower expression of a Notch reporter gene in wing margin cells. A-A'.
Representative expression of the E(spl)mα-GFP reporter in larval wing discs from WT flies (far left) or from flies heterozygous for either Su (H) IB115 , Su(H) O5 , or Su(H) T4 , as indicated at top. Note, the wing discs were immunostained for Cut (red) to mark the wing margin cells and only the posterior margin cells were quantified in B. We focused on the posterior margin cells to avoid the strong GFP staining in the neighboring sensory organ cells because the E(spl)mα-GFP reporter is strongly activated by proneural transcription factors in a Notch-independent manner within the sensory organs [25]. Bottom panel shows GFP expression in grayscale. A'. Schematic of the larval wing disc with the wing margin cells (red) and the sensory organs (green) highlighted. Note, the absence of macrochaetae cells in the posterior compartment. B. Quantification of GFP levels from wing discs of the indicated genotypes. Each dot represents the average posterior wing margin cell pixel intensity as measured from an individual imaginal disc. � denotes significance (p<0.01) using a two-sided Student's t-test.
https://doi.org/10.1371/journal.pgen.1010335.g004 PLOS GENETICS many sensory bristles. Notch heterozygous flies produce a weakly penetrant, but significant increase in macrochaetae formation in animals with two WT alleles of Su(H) (Fig 5G). Intriguingly, we found that flies heterozygous for either the Su(H) T4 or the Su(H) O5 allele further enhanced the number of extra macrochaetae in N 55e11/+ flies (Fig 5G). In contrast, flies heterozygous for the Su(H) IB115 null allele had the opposite effect and resulted in a small but significant suppression in the increase in macrochaetae in N 55e11/+ flies.
Taken together, these data support two ideas: First, even within the same tissue, the presence of either a Su(H) T4 or Su(H) O5 allele can significantly increase or decrease the number of sensory organs formed, but primarily only in sensitized genetic backgrounds that alter the amount of the Notch pathway co-activator (i.e. Notch) or co-repressor (i.e. Hairless) proteins. Second and in sharp contrast to the Su(H) T4 or Su(H) O5 alleles, we found that flies heterozygous for a Su(H) null allele (i.e. Su(H) IB115/+ ) partially suppressed each of these phenotypes in the same genetic backgrounds. These data suggest that simply lowering the amount of Su(H) transcription factor can suppress imbalances in Notch co-activator to Hairless co-repressor levels, whereas having a DNA binding compromised Su(H) molecule enhances these imbalances to make phenotypes worse. To further investigate how Su(H) gene dose impacts sensory organ formation, we generated a series of flies containing an additional allele of WT Su(H) (i.e. Su(H)-GFP BAC). In an otherwise WT background, flies containing a third copy of Su(H) were phenotypically normal and did not change the number of macrochaetae (Fig 5H).  Fig 5H), consistent with further increasing the levels of Su(H) transcription factor causing even stronger phenotypes. Next, we replaced one of the WT Su(H) alleles with the Su(H) T4 allele (i.e. Su(H) T4/+ ;H 1 /Su(H)-GFP) and found that these flies had an even larger loss in macrochaetae (Fig 5H). In contrast, replacing one of the WT Su(H) alleles with the Su(H) null allele (i.e. Su(H) IB115/+ ;H 1 /Su(H)-GFP) resulted in flies with a similar number of macrochaetae as H 1/+ heterozygotes (Fig 5G and 5H). Altogether, these data suggest that the amount of Su (H) transcription factor is critical for proper Notch signaling and if one of these alleles encodes a Su(H) transcription factor with compromised DNA binding activity the phenotypes are dramatically enhanced, especially in flies with imbalanced levels of Notch and Hairless.

Su(H) and RBPJ AOS variants bind normally to the NICD co-activator and the Hairless/SHARP co-repressor proteins
Integrating the DNA binding and genetic data suggests the following model: the Su(H) T4 and Su(H) O5 alleles produce proteins that sequester the NICD co-activator and Hairless co-repressor off DNA. Hence, flies heterozygous for either a Su(H) T4 or Su(H) O5 allele can exacerbate both wing notching and the gain in macrochaetae when Notch is limiting (N 55e11/+ ), and the loss in macrochaetae when Hairless is limiting (H 1/+ ). To further test this idea, we performed ITC assays to directly measure Su(H) and Rbpj binding to both co-activators and co-repressors. We hypothesized that the DNA binding mutations in Su(H)/RBPJ would not impact cofactor binding because these sites are distant from the interfaces of Su(H)/RBPJ involved in co-activator and co-repressor binding (Fig 6A). To test the Su(H) variants' ability to bind coactivators, we performed ITC with a construct that corresponds to the RAM domain of the Drosophila Notch receptor, which binds the β-trefoil domain (BTD) of Su(H) with high affinity [28] (Fig 6A). WT Su(H) bound to RAM with 187 nM affinity, whereas we observed no statistically significant differences in RAM binding to Su(H) E137V [T4] or K132M [O5] (Fig 6B,  S1 Table). To test the Su(H) variants' ability to bind co-repressors, we performed ITC with Hairless, residues 232-358, which correspond to the region of Hairless that binds the C-terminal domain (CTD) of Su(H) with high affinity [29,30] (Fig 6A). WT Su(H) bound Hairless with a 3 nM K d and we observed no significant differences in Hairless binding to Su(H) E137V or K132M (Fig 6C, S1 Table).
To test if mouse Rbpj proteins with the AOS variants showed similar trends with respect to co-activator/repressor binding, we performed ITC with the mouse RAM domain from Notch1 and the mouse co-repressor SHARP [31][32][33]. WT Rbpj and the E89G and K195E variants similarly bound RAM with an~20 nM K d (Fig 6D, S1 Table). For Rbpj-SHARP complexes, WT and E89G bound the mouse SHARP protein with~5 nM affinity, whereas K169E bound with a 21 nM K d; however, this difference was not found to be statistically significant (T-test p value = 0.079). To further confirm that co-activator/repressor binding was unaffected for the mouse Rpbj protein containing analogous AOS variants, we performed coimmunoprecipitation experiments in HEK293 cells transfected with epitope-tagged versions (either GFP-or FLAG-tagged as indicated) of each Rbpj variant in the presence of either NICD1, NICD1 + MAML1, or SHARP (Fig 6F-6H). NICD1 coimmunoprecipitated with each mouse Rbpj AOS variants to a similar extent as WT Rbpj (Fig 6F) and MAML1 coimmunoprecipitated with NICD1 and Rbpj to similar levels with WT and variant Rbpj constructs (Fig 6G). Both mouse Rbpj AOS variants also coimmunoprecipitated equally well with SHARP compared to WT Rbpj (Fig 6H). Similarly, using a well-established mammalian two-hybrid assay that is independent of DNA binding by Rbpj, the variants E89G and K195E interact with the corepressor SHARP equivalently to WT Rbpj (S3 Fig). Taken together, these data support the hypothesis that the Su(H) variants and mouse Rbpj AOS variants specifically disrupt DNA binding without affecting co-activator/repressor binding, corroborating a model of cofactor sequestration.

RBPJ AOS variants alter Notch-mediated transcription in mammalian cells via a sequestration mechanism
To confirm that the mouse Rbpj AOS variants do not impact protein subcellular localization, we transfected HeLa cells with GFP-tagged Rbpj (WT and variants) and monitored protein distribution using fluorescence microscopy. Like WT Rbpj, both AOS variants were primarily localized in the nucleus (Fig 7A). Additionally, we treated Rbpj transfected cells with cycloheximide and monitored protein half-life. The degradation rates of both Rbpj AOS variants were  S4 Fig). We performed luciferase assays using a reporter containing 12 CSL binding sites in HeLa RBPJ-KO cells, in which endogenous RBPJ had been deleted using CRISPR-Cas9 [34], to test whether the AOS variants could bind and activate the reporter in the presence and absence of NICD1 (Fig 7B and 7C). First, we expressed mouse Rbpj-VP16 fusion proteins, WT or AOS variants, to determine whether the AOS variants could bind and activate the reporter in a NICD1 independent manner. Both K195E and E89G Rbpj-VP16 fusions showed significant reduction in reporter signal compared to WT Rbpj-VP16 (Fig 7B), consistent with the above DNA binding experiments that show each variant protein has decreased DNA binding activity (Fig 2). To investigate the effect of these AOS variants on Notch-mediated transcription, we performed luciferase assays by co-transfecting with mouse NICD1 and Rbpj. In this case, we observed robust signaling with WT Rbpj and NICD1, whereas Rbpj K195E and E89G activated the reporter to a much lesser extent than WT. Moreover, it should be noted that the K195E variant showed a stronger negative impact on transcriptional responses when either fused to VP16 or co-transfected with NICD1 than the E89G variant, consistent with the K195E variant have a more dramatic impact on DNA binding.
Next, we assessed the ability of each Rbpj variant to repress target genes using a mature T (MT) cell line in which Notch signaling is in the "off" state. In this previously established system, CRISPR/Cas9-mediated deletion of Rbpj leads to upregulation of the Notch target genes Hey1 and Hes1, due to de-repression, but not the Tbp control, and this phenotype can be efficiently rescued by reintroducing WT mouse Rbpj expression via lentiviral transduction [31]. We generated WT Rbpj, E89G, and K195E MT cell lines and performed qRT-PCR to quantify each protein's ability to repress transcription at the Hey1 and Hes1 Notch target sites (Fig 7D). As expected, expression of WT Rbpj rescued repression of Hey1 and Hes1, whereas the Rbpj K195E protein that more strongly affects DNA binding (Fig 2), failed to repress transcription at both Hey1 and Hes1 sites. However, Rbpj E89G, which has a lesser impact on DNA binding than K195E (Fig 2), restores repression at Hey1 and Hes1 similar to WT RBPJ. Taken together, these results suggest that the severity of the mutation on DNA binding directly corelates to the level of both transcriptional activation and repression.
Lastly, to test the sequestration mechanism in mammalian cells, we once again used our luciferase assays in HeLa RBPJ-KO cells (Fig 7E). Because transfection of increasing amounts of WT Rbpj results in reduced reporter activity (Figs 7E and S5B), which has been reported previously [35,36] and is likely due to titration of NICD, we designed three additional Rbpj mutants: (1) Rbpj KRS (K195E/R218H/S221D), which combines the variant K195E with two other previously described Rbpj mutants that affect DNA binding [37,38] and is completely devoid of any DNA binding (S5 Fig); (2) Rbpj NBM (NICD Binding Mutant, F261A/V263A/ R422E/E425R/E398R), which is unable to interact with NICD1 but retains its ability to bind DNA (Figs 7E & S5); and (3) Rbpj NBM/KRS , which is neither able to bind DNA nor NICD1. It should also be mentioned that Rbpj NBM has been tested in Drosophila, using the fly ortholog Expression of GFP-Rbpj (middle blot) was detected using an anti-GFP antibody. Expression of the Flag-NICD protein (bottom blot) was detected using an anti-Flag antibody. The asterisk in the upper blot marks the heavy chain of the antibody used for immunoprecipitation. G. Mouse Rbpj WT and Rbpj-AOS variants similarly form the NICD co-activator complex that includes Mam1. HEK293 cells were transfected with the plasmids expressing Maml1 in the absence or presence of Flag-Rbpj WT, Flag-Rbpj E89G, or Flag-Rbpj K195E and Flag-NICD. Co-immunoprecipitation was performed with the anti-Flag antibody agarose beads and detected by Western blotting using an anti-Maml1 antibody. Expression of Flag-Rbpj proteins (middle blot) and Flag-NICD was detected using an anti-Flag antibody. Expression of the Maml1 protein (bottom blot) was detected using the anti-Maml1 antibody. H. CoIP of mouse SHARP with WT and AOS variants of Rbpj. HEK293 cells were transfected with the plasmids GFP-Rbpj WT, GFP-Rbpj E89G, or GFP-Rbpj K195E in the absence or presence of Flag-SHARP-(RBPID, Rbpj-interaction domain). Immunoprecipitation was performed with the anti-Flag antibody agarose beads and detected by Western blotting using an anti-GFP antibody. Expression of GFP-Rbpj (middle blot) was detected using an anti-GFP antibody. Expression of the Flag-SHARP-(RBPID) protein (bottom blot) was detected using an anti-Flag antibody. The asterisk in the upper blot marks the heavy chain of the antibody used for immunoprecipitation. Luciferase reporter assay using an Rbpj-VP16 fusion to test for DNA binding capacities (schematic lower panel). Both Rbpj K195E-VP16 and E89G-VP16 show reduced reporter signal compared to WT, consistent with the observation that both have impaired DNA binding. HeLa RBPJ-KO cells were transfected with the Notch/RBPJ dependent reporter 12xCSL-RE-Luc (250 ng) without or with the indicated Rbpj-VP16 constructs (10 ng). Luciferase activity was measured 24 hours after transfection. Bars represent mean values from six independent experiments; error bars indicate standard deviation, ��� p< 0.001. C. Transactivation capacities of RBPJ WT and AOS variants together with NICD. HeLa RBPJ-KO cells were transfected with the Notch/ RBPJ dependent reporter 12xCSL-RE-Luc (250 ng) without or with the indicated Rbpj constructs (50 ng) and the NICD expression plasmid (10 ng) (schematic lower panel). In RBPJ depleted HeLa cells, NICD is not able to transactivate the reporter. Together with NICD, the AOS Su(H), and in these in vivo assays was also shown to be completely lacking in NICD binding and Notch-mediated transcriptional activation [19]. For these experiments, a constant amount of WT Rbpj (50 ng) and NICD1 (10 ng) was co-transfected into cells to activate the reporter (Fig 7E, panel i), and increasing amounts (50, 150, 250 ng) of either WT Rbpj (Fig 7E, panel  ii), Rbpj NBM (Fig 7E, panel iii), Rbpj KRS (Fig 7E, panel iv), Rbpj NBM/KRS (Fig 7E, panel v), or the AOS variant K195E (Fig 7E, panel vi) were additionally transfected into the cells. These experiments demonstrate the following: (1) Increasing cellular concentrations of the Rbpj DNA binding mutants KRS and K195E both significantly reduce reporter activity more than WT Rbpj, albeit the reduction by the AOS variant K195E is modest compared to the KRS mutant; (2) Increasing cellular concentrations of Rbpj NBM , which cannot bind NICD1 but can still bind DNA, has only a very minor negative effect on reporter activity, likely by competing with WT Rbpj for DNA binding sites on the reporter; and (3) Increasing cellular concentrations of Rbpj NBM/KRS , which cannot interact with NICD1 or DNA, has no effect on reporter activity. Taken together, these data provide proof-of-principle that Rbpj variants that are defective in DNA binding reduce Notch signaling via a cofactor sequestration mechanism.

Discussion
Human genetic studies previously revealed that Adams Oliver Syndrome (AOS) is associated with autosomal dominant alleles in the NOTCH1, DLL4, and RBPJ genes [39]. Sequence analysis of these Notch pathway variants has led to the prediction that each variant is likely to compromise Notch signaling in specific tissues, and thereby cause a developmental syndrome that affects multiple organ systems. In knockout mouse studies, homozygous null Notch1, Dll4, or Rbpj mice are all embryonic lethal [40][41][42] and only Dll4 heterozygous mice display a severe phenotype [12]. Here, we took advantage of the unexpected finding that a previously described dominant Su(H) allele in Drosophila [17] contains a missense variant in an analogous residue found in a family with AOS [14]. By combining quantitative genetic studies with quantitative DNA binding, protein-protein interaction, and transcriptional reporter assays, our findings provide evidence that Su(H)/Rbpj variants that compromise DNA binding can result in Notch target gene misregulation and phenotypes due to the sequestration of either the Notch signal or the antagonistic co-repressor proteins. Moreover, our findings reveal how increasing or decreasing the genetic dose of Su(H) can either suppress or exacerbate Notch pathway phenotypes depending on genetic background. Taken together with prior studies using these and similar Su(H) alleles as well as the human genetic data, our findings have several important Rbpj variants show decreased transactivation capacities compared to Rbpj WT. Luciferase activity was measured 24 hours after transfection. Bars represent mean values from six independent experiments; error bars indicate standard deviation, ��� p< 0.001. D. Rbpj-WT but not the AOS variant K195E are able to rescue transcription repression of Notch target genes. Left: Western blot against Rbpj (WT and variants) in reconstitution experiments using CRISPR/Cas9-mediated depletion of Rbpj in mature T (MT) cells. Rbpj protein expression levels were analyzed in Rbpj-depleted cells (control, line 1), Rbpj WT or Rbpj E89G and Rbpj K195E mutants, (lanes 2,3 and 4 respectively) and TBP was used as a loading control. Right: qRT-PCR: Expression of Rbpj WT and Rbpj E89G, but not Rbpj K195E rescue the repression of Hey1 and Hes1 Notch target genes. Total RNA was extracted from Rbpj depleted mature T cells reconstituted with either empty vector (control), Rbpj WT and Rbpj E89G or Rbpj K195E mutants. Data shown were normalized to the housekeeping gene Bact and represent the mean ± SD of three independent experiments ([ ��� ] P < 0.001, [ �� ] P < 0.01, [ � ] P < 0.05 and [NS] not significant, unpaired Student's t-test). E. Luciferase reporter assays demonstrating that the cofactor sequestration mechanism occurs in mammalian cells. HeLa RBPJ-KO cells were transfected with the Notch/RBPJ dependent reporter 12xCSL-RE-Luc (250 ng), WT Rbpj (50 ng), and NICD (10 ng) as shown in panel i. Cells were additionally transfected with increasing concentrations (50, 150 & 250 ng) of either WT, NBM (F261A/V263A/R422E/E425R/E398R), KRS (K195E/R218H/S221D), NBM/KRS, or K195E Rbpj constructs in panels ii-iv. Rbpj NBM is unable to bind NICD and activate the reporter, and Rbpj KRS is completely ablated for DNA binding (S5 Fig). Increasing concentrations of Rbpj KRS strongly inhibit reporter activity due to a complete lack of DNA binding, but still able to bind cofactors (panel iv); however, increasing concentrations of Rbpj KRS/NBM have no effect on reporter activity (panel v) due to the inability of this construct to interact with NICD and titrate it off the reporter. (

PLOS GENETICS
implications for better understanding the potential molecular defects underlying both AOS phenotypes and Notch pathway dysregulation.
First, our studies in Drosophila highlight how Su(H) allelic variants that specifically encode proteins with altered DNA binding, but not co-activator/co-repressor binding, behave differently than null Su(H) alleles that fail to generate any functional protein. In particular, we found that flies heterozygous for a Su(H) allele encoding a DNA compromised protein strongly enhances Notch and Hairless phenotypes, whereas simply lowering the genetic dose of Su(H) using a protein null allele partially suppresses these phenotypes. These findings support the idea that the Su(H) T4 (E137V) and Su(H) O5 (K132M) variants act as dominant negative proteins due to sequestering cofactors, especially in sensitized backgrounds where either Notch or Hairless are limiting. It should be noted that a similar mechanism was proposed and demonstrated by over-expressing a DNA binding deficient Rbpj (R218H) molecule in mammalian cell culture [43], which is also consistent with our cellular reporter assays using the Rbpj DNA binding mutants KRS and K195E (Fig 7E). This mechanism also likely explains the previously observed dominant gain-of-function activity of other described Su(H) alleles, such as Su (H) eBC11 (Y134N) [17,44] and Su(H) S5 (R266H) [27,38]. While the impact of the Su(H) eBC11 Y134N variant on DNA and cofactor binding has yet to be experimentally tested, its location within the DNA binding domain and the reported dominant genetic behavior of the Su (H) eBC11 allele support the idea this variant is also likely to selectively compromise DNA binding. Moreover, the Su(H) S5 R266H variant has been previously proposed to have dominant negative activity based on prior studies showing that it qualitatively disrupts DNA binding and was found to still interact with the Hairless protein in a yeast-two hybrid assay [45]. Our quantitative molecular and genetic studies further support these ideas and also help to explain how homozygous mutant clones of the Su(H) T4 allele cause a Notch loss-of-function phenotype [16]; namely, the loss of DNA binding by the Su(H) T4 variant will result in the failure to robustly bind key target genes downstream of Notch activation. Thus, the Su(H) T4 allele should be viewed as both a dominant negative allele that can interfere with wild type Su(H) function in heterozygous cells by binding limiting co-factors, and a hypomorphic loss-of-function allele due to decreased DNA binding and the subsequent failure to activate key Notch target genes in homozygous mutant cells.
Interestingly, since the discovery of the AOS-associated E63G and K169E human RBPJ variants that were originally identified in 2012 [14], additional RBPJ AOS variants have been identified, including R65G, F66V, and S332R [39]. Importantly, none of the AOS-associated RBPJ alleles are predicted to be a protein null allele and given the role of R65 and F66 in DNA binding, as shown in previous Rbpj structures [33], and their close proximity to E63 in the NTD of RBPJ, these missense variants will also likely reduce DNA binding, while having little to no effect on cofactor binding. S332, which is currently classified as a variant of unknown significance [46], does not directly contact DNA, but rather is located in a long β-strand that connects the BTD and CTD of RBPJ. However, it is possible that the S332R AOS variant will incur structural changes that indirectly decrease the overall binding affinity for DNA, resulting in a similar disease mechanism. Nonetheless, we have established a rigorous platform to test these and future AOS variants in vitro and in vivo.
Intriguingly, an analogous cellular mechanism that compromises Su(H) DNA binding has previously been identified in Drosophila [37,47]. In this case, phosphorylation of Su(H) residue S269, which interacts with DNA, blocks DNA binding, and plays a role in fly hematopoiesis. Moreover, it is tempting to speculate that the severity of AOS disease is potentially linked to the impact the variant has on DNA binding. Certainly, our genetic studies in Drosophila support this premise, i.e. the Su(H) T4 allele encoding the E137V variant had a stronger impact on DNA binding in our in vitro assays than the Su(H) O5 allele encoding the K132M variant (Fig  2), and the Su(H) T4 allele consistently led to stronger mutant phenotypes in vivo (Figs 3-5). Perhaps some of the patient variability can be explained by the severity of the RBPJ variant; however, there are likely other genetic and environmental modifiers that also contribute to an individual's disease presentation, obscuring the relative effect a variant has on DNA binding.
Second, our studies, as well as the original publication that identified the Su(H) T4 and Su (H) O5 alleles [17], revealed strong genetic interactions with not only Notch, but also the antagonistic Hairless co-repressor gene. These findings raise the possibility that human AOS-associated RBPJ alleles with compromised DNA binding may cause developmental defects due to the sequestration of mammalian co-repressor proteins and not simply due to sequestration of the Notch signal. However, it should be noted that repression mechanisms are not as well conserved across species, as there is no direct Hairless ortholog in mammals and mammalian RBPJ can bind multiple different co-repressors [10]. Moreover, we found that only the stronger AOS variant Rbpj K195E was defective in the MT cell repression assay (Fig 7D), whereas Rbpj E89G, which has a milder defect in DNA binding, functioned similarly to WT in this repression assay. Additionally, no known variants in Notch pathway co-repressor genes have been associated with AOS, whereas numerous NOTCH1 and DLL4 variants have been associated with this developmental syndrome [39], and several of the autosomal dominant NOTCH1 and DLL4 AOS alleles encode nonsense variants that are unlikely to generate a protein with dominant-negative effects. Thus, the clinical similarities between AOS patients with variants in NOTCH1, DLL4, and RBPJ suggest that the majority of developmental defects associated with RBPJ variants that encode proteins with compromised DNA binding are likely due to decreased NOTCH1 signal strength and not due to an impact on the antagonistic co-repressors. Nevertheless, since the genetic etiology of human patients with AOS is far from complete, future studies may want to include sequence analysis for variants in co-repressor genes known to interact with RBPJ.
Third, our Drosophila genetic data generated by systematically testing the impact of both increasing and decreasing Notch and Hairless gene dose in the presence of different Su(H) alleles highlight how genetic background can dramatically sensitize tissue-specific phenotypes. For example, flies heterozygous for the Su(H) T4 allele only have a weakly penetrant wing notching phenotype in a genetic background with two wild type alleles of Notch and Hairless. However, simultaneously lowering the Notch gene dose or increasing the Hairless gene dose dramatically enhanced the wing phenotype, whereas increasing Notch gene dose or decreasing Hairless gene dose suppressed the wing phenotype. Moreover, we similarly found that while the Su(H) T4 allele on its own did not dramatically impact macrochaetae development, simultaneously changing the relative dose of Notch and Hairless could significantly exacerbate macrochaetae phenotypes in compound heterozygotes with the Su(H) T4 allele. Collectively, these data suggest that genetic modifiers in the presence of the dominant-negative AOS-like Su(H) allele can have tissue-specific impacts based on whether they either strengthen or weaken the Notch co-activator complex versus Hairless co-repressor complex. Lastly, we non-intuitively found that simply changing the gene dose of the Su(H) transcription factor can modify (i.e. partially suppress) each of these Notch-and Hairless-dependent phenotypes. Currently, we do not have a good molecular explanation for how decreasing the expression levels of the common transcription factor could suppress both Notch-and Hairless-dependent phenotypes, and future experiments are needed to define the complex relationships that exist between Notch, Hairless, and Su(H) gene dose in each specific tissue.
Fourth, our studies and proposed sequestration disease mechanism raise an interesting unanswered question-since RBPJ is the sole downstream nuclear effector of Notch signaling, why do AOS patients harboring RBPJ variants have distinct and nonoverlapping phenotypes with Alagille syndrome patients that have variants in NOTCH2 or JAG1 [15]? Current data suggests that the NOTCH signal (i.e. the intracellular domains) generated by both NOTCH1 and NOTCH2 interact with RBPJ with similar affinities [48,49], suggesting that the DNA binding compromised RBPJ proteins should similarly sequester the signals generated by either NOTCH1 or NOTCH2. Hence, it is unclear why the AOS-associated RBPJ alleles that cause similar phenotypes as the AOS-associated NOTCH1 variants do not also cause Alagille-like phenotypes that have been associated with NOTCH2 variants that cause decreased NOTCH2 activity. While answering this question is beyond the scope of this study, developing vertebrate models of AOS will certainly help address this question and provide further insights into the molecular mechanism of AOS, Alagille, and Notch signaling in general.

Bacterial expression constructs
Su(H) amino acid residues 98-523 corresponding to the conserved and structurally ordered core domain were cloned into the pGEX-6P1 bacterial expression vector [28]. Based on previous studies [30], additional mutations R155T and N218G, which are surface exposed and distal from sites of DNA and co-regulator binding, were introduced to improve upon protein purification, yield, and stability. These mutations were previously shown to have identical binding to Notch, Hairless, and DNA as WT Su(H) [30]. Quick-change site directed mutagenesis was used to introduce the mutations K132M and E137V into the pGEX-6P1-Su(H) construct. For the mammalian ortholog, mouse Rbpj amino acids 53-474, corresponding to the structural core, were cloned into pGEX-6P1 and the E89G and K195E variants were introduced as mentioned above [33].

Bacterial recombinant protein expression and purification
As previously described [28,33], competent BL21 Tuner cells transformed with pGEX-6P1-Su (H) or pGEX-6P1-Rbpj were grown at 37˚C in LB + ampicillin to an OD of 1.5 followed by overnight IPTG induction at 20˚C. Cell pellets were resuspended in lysis buffer, sonicated, and centrifuged at 15,000 g. 60% w/v ammonium sulfate was added to the supernatant to precipitate the protein and then centrifuged at 15,000 g. The protein pellet was resuspended in PBS + 0.1% triton and incubated overnight with glutathione affinity resin. GST-Su(H) or -RBPJ tagged protein was eluted from the column and then cut with PreScission Protease. An SP ion exchange column was used to separate the cut GST from the Su(H) or RBPJ protein, which was then concentrated, flash frozen in liquid nitrogen and stored at -80˚C. Drosophila Hairless residues 232-358 and human SHARP residues 2776-2833, corresponding to the regions that bind Su(H) and Rbpj, respectively, were cloned into pSMT3 to produce recombinant protein with an N-terminal SMT3 and His tag and purified as previously described [29][30][31].

Electrophoretic Mobility Shift Assays (EMSAs) using purified proteins
EMSAs using bacterially purified proteins were performed essentially as previously described [52][53][54]. In brief, a fluorescent labeled probe containing a high affinity CSL binding site (shown in bold) was generated by annealing an unlabeled 5'-CGAACGAGGCAAACC-TAGGCTAGAGGCACCGTGGGAAACTAGTGCGGGCGTGGCT-3' oligonucleotide with a 5'IRDye-700 labeled 5'-AGCCACGCCCGCACT-3' oligonucleotide that is complementary to the underlined region of the longer oligonucleotide. The Klenow enzyme was used to make the double-stranded probe and the indicated amounts of each Su(H) and Rbpj proteins were incubated with 3.4 nM of the labeled probe for 10 minutes at room temperature prior to being loaded onto native acrylamide gel electrophoresis. Acrylamide gels were run at 150V for 2 hours and imaged and quantified using the LICOR Odyssey CLx scanner.

In vitro protein translation
The in vitro protein translations were performed using the TNT-assay (L4610) from Promega according to manufacturer's instructions. Prior to EMSAs the in vitro translations of Rbpj (WT) and mutant proteins were monitored by western blotting using an anti-Flag antibody (M5, Merck).

EMSA from in vitro protein translation system
Reticulocyte lysates from in vitro translations were used for electromobility shift assays (EMSAs) in a binding buffer consisting of 10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.1 mM EDTA, 0.5 mM DTT, and 4% glycerol. For the binding reaction, 10 ng (0.02 U) poly(dI-dC) (GE healthcare) and approximately 0.5 ng of 32 P-labeled oligonucleotides were added. The sequence of the double-stranded oligonucleotide with 2 RBPJ binding sites (underlined): 5 -CCT GGA ACT ATT TTC CCA CGG TGC CCT TCC GCC CAT TTT CCC ACG AGT CG-3 and reverse strand: 5´-CTC GCG ACT CGT GGG AAA ATG GGC GGA AGG GCA CCG TGG GAA AAT AGT TC-3´. Super shifting of complexes was achieved by adding 1 μg of anti-Flag (M5, Merck) antibody. The reaction products were separated using 5% polyacrylamide gels with 1x Tris-glycine-EDTA at room temperature. Gels were dried and exposed to X-ray films (GE Healthcare).

Oligonucleotide Preparation for ITC
The following 20-mer oligonucleotide sequences were ordered from Eurofins: 5'-GGCACCGTGGGAAACTAGTG-3' and the reverse compliment 5'-CACTAGTTTCC-CACGGTGCC-3' showing the Su(H)/RBPJ binding site in bold and underlined. Single stranded oligos were further purified on a Resource Q ion exchange column and then buffer exchanged into an oligo annealing buffer containing 10 mM Tris 8.0, 500 mM NaCl, and 1 mM MgCl 2 . Single stranded oligos were combined in equal molar amounts, boiled for 10 minutes, and then slowly cooled to room temperature to allow for proper annealing.

Isothermal titration calorimetry
All proteins, peptides, and oligos were dialyzed overnight in a buffer containing 50 mM sodium phosphate and 150 mM sodium chloride. Experiments were conducted using the VP-ITC MicroCalorimeter manufactured by MicroCal. All protein/DNA binding experiments were performed at 10˚C, while protein/peptide experiments were performed at 20˚C. Each experiment was performed in triplicate with 20 injections, 14 μL per injection. Heat of dilution experiments were performed by injecting syringe samples into a cell containing only buffer, and all analysis was performed with the heat of dilution subtracted before fitting. The raw data were analyzed using ORIGIN and fit to a one site binding model. A two sample T-test was used to compare WT proteins to each mutant, with a p-value <0.05 indicating a significant difference.

Differential scanning fluorimetry
Rbpj/Su(H) samples were prepared in triplicate at 5 μM in buffer containing 50 mM sodium phosphate and 150 mM sodium chloride. DSF was performed by heating samples containing 10X SYPRO Orange in 1˚C increments from 25˚C to 100˚C using the StepOne real-time PCR machine and using the Protein Thermal Shift software (ThermoFisher) to fit the data.

Genomic DNA isolation and sequencing
For the original determination of the mutations in the Su(H) T4 and Su(H) O5 alleles, chromosomal DNA was isolated by protease K digestion of fly lysates. Genomic DNA was subjected to PCR with primers specific to the Su(H) coding region. Due to the presence of a large intron in the coding region, the first exon was amplified and sequenced separately using two primers TTGCAGCCTTAAACAGAAGCCAGC (forward) and AGCCGGTATTAT-CAGGTGCTTGGT (reverse). The resulting PCR fragment was sequenced using the same primers as well as an internal primer, ACAAATGCAGATGTCCTTGCTGCC. Exons 2-4 were amplified using two primers, CAAAGCTGCATTGCTTGCGGTT (forward) and TCAATCTACAAACTAAGGTCTTCG (reverse). The resulting PCR fragment was sequenced using these primers as well as two additional primers (ACAGTCAAACTGGTGTGCTCCG-TAA and ATGTAGAAGGCGCATTTGTGCAGC). Since the original fly stocks were heterozygous for the Su(H) T4 and Su(H) O5 alleles, the mutations were identified by the presence of double peaks in the resulting chromatogram. To confirm specific coding variants in each allele prior to our genetic phenotyping assays, genomic DNA was prepared using Qiagen's DNeasy blood purification kit. 10 adult males were used per genotype. DNA was eluted in 100ul Buffer AE. We used two primers (5'-GTCCAGTCCGCAATGAAAAT-3' (forward) and 5'-TGCTGCAACATCTCCTCGTA-3' (reverse)) to cover approximately 300bp surrounding each point mutation. The same primers were used for DNA sequencing and ExPasy.org was used to translate nucleotide to amino acid sequence. Note, the sequencing of these Su(H) alleles were previously reported to Flybase via personal communication by Dr. Patrick Dolph [44].

Fly husbandry
All flies were cultured on standard cornmeal food at 25˚C. The following alleles were obtained from the Bloomington Drosophila Stock Center: Su(H) T4 [19].

Reporter assay in wing disc
Homozygous E(spl)mα-Gal4, UAS-GFP male flies were mated to either WT, Su(H) IB115 /cyo, act-GFP, Su(H) T4 /cyo,act-GFP or Su(H) O5 /cyo,act-GFP virginal females. Imaginal wing discs from non-cyo-act-GFP wandering 3 rd instar larvae were dissected in PBS and fixed in 4% formaldehyde for 15min. Discs were subsequently washed 4 times with PBX (0.3% TritonX-100 in PBS), and incubated with the primary antibody, anti-Cut (Mouse 1:50, DSHB). No antibody was used for GFP detection. A fluorescent secondary antibody (Goat anti-Mouse 555 Alexa Fluor, Molecular Probes) was used to detect the Cut-positive wing margin cells. For quantitative purposes, all imaginal discs were dissected, fixed, and imaged at the same time on a Nikon A1R inverted confocal microscope (40x objective) with identical exposure settings. Zstack images were analyzed using Imaris software. A Student's t-test was used to determine significance.

Genetic assays
We analyzed the wing nicking and macrochaetae phenotypes essentially as previously described [52]. In brief, flies of the appropriate genotypes were mated and transferred to fresh food every other day to avoid overcrowding. Offspring of the listed genotypes were selected and the number of nicks on each wing and/or the number of macrochaetae on the dorsal head and thorax was recorded. Proportional odds model with Bonferonni adjustment and twosided Fisher's exact test were used to determine significance for wing nicking. Student's t-test and two-sided Fisher's exact test were used to determine significance for bristle number.

Cell culture and preparation of cell extracts
Rbpj-depleted hybridoma mature T (MT) cell line was grown in Iscove's Modified Dulbecco Medium (IMDM, Gibco) supplemented with 2% FCS, 0.3 mg/l peptone, 5 mg/l insulin, nonessential aminoacids and penicillin/streptomycin. Cells were grown at 37˚C with 5% CO 2 . Phoenix packaging cells (Orbigen, Inc. San Diego, CA, USA) were cultivated in Dulbecco's Modified Eagle Medium (DMEM, Gibco) supplemented with 10% fetal calf serum (FCS) and penicillin/streptomycin. CRISPR/Cas9-mediated Rbpj depleted hybridoma mature T cells were generated as previously described [31]. Rbpj-depleted mature T-cell lines stably expressing Rbpj WT , Adams-Oliver mutant Rbpj E89G or Adams-Oliver mutant Rbpj K195E were generated as follows: 5 x 10 6 Phoenix cells were transfected with 20 μg of the retroviral plasmid DNA mixed with 860 μl of H 2 O and 120 μl of 2x HBS buffer (50 mM HEPES pH 7.05, 10 mM KCl, 12 mM Glucose, 280 mM NaCl, 1.5 mM NaHPO 4 ) while vortexing and the solution was incubated 20 min at room temperature. In the meantime, 25 μM Chloroquine solution (Sigma-Aldrich) was added to the Phoenix cells (1 μl/ml) and the cells were incubated for 10 min. The DNA solution was added to the cells and 12 h later the medium was replaced. After 24 h of incubation, the medium containing the retroviral suspension was filtered and Polybrene (Sigma-Aldrich) solution was added. Fresh medium was added to the Phoenix cells that were maintained in culture for further infections. The retroviral solution was used for spin infection of Rbpj-depleted MT cells by centrifuging 45 min at 1800 rpm at 37˚C. In total, four spin infections were performed over two days. Positively infected cells were selected with Blasticidin (Gibco).

Preparation of protein extracts and western blotting from Rbpj-depleted mature T cells
The nuclear extract (NE) from MT cells overexpressing the Rbpj constructs or control cells containing empty vector was prepared as follows. Briefly, 10 x 10 6 cells were washed with PBS and resuspended in 200 μl of Buffer 1 (10 mM HEPES pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1mM beta-mercaptoethanol, supplemented with PMSF). The cell suspension was incubated 10 min on ice, 5 μl of 10% NP-40 were added and mixed by vortexing. After 10 s of centrifugation at 13000 rpm at 4˚C, the nuclei pellet was washed twice in 500 μl of Buffer 1 and resuspended in 100 μl of Buffer 2 (20 mM HEPES pH 7.9, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM beta-mercaptoethanol, supplemented with PMSF). After 20 min of incubation on ice, the nuclei suspension was centrifuged 10 min at 13000 rpm at 4˚C. and the supernatant was collected for further analysis. Protein concentration was measured by Bradford assay (Biorad) and samples were boiled after adding SDS-polacrylamide gel loading buffer. Samples were resolved by SDS-Page and analyzed by Western blotting using antibodies against RBPJ (Cosmo Bio Co. LTD) or TBP (Santa Cruz Biotechnology).

Gene expression analysis as measured by qRT-PCR in hybridoma mature T-cells
Total RNA was purified using TRIzol reagent accordingly to manufacturer's instructions. 1 μg of RNA was reverse transcribed into cDNA using random hexamers and M-MuLV reverse transcriptase (New England Biolabs). Quantitative PCRs were assembled with Absolute QPCR ROX Mix (Thermo Scientific, AB-1139), gene-specific oligonucleotides and double-dye probes and analyzed using StepOne Plus Real Time PCR system (Applied Biosystem). Data were normalized to the housekeeping gene Beta actin (Bact) and represent the mean ± SD of three independent experiments ( ��� p<0.001; �� p<0.01, � p<0.05 and [ns], not significant, unpaired Student's t-test).

Luciferase assays
HeLa cells and HeLa RBPJ-KO cells were seeded in 48-well plates at a density of 2.5 × 10 5 cells. Transfection was performed with Lipofectamine 2000 reagent (ThermoFisher) using 0.5 μg of reporter plasmid alone or together with various amount of expression plasmids (given in the corresponding figure legends). After 24 h luciferase activity was determined from at least four independent experiments with 20 μl of cleared lysate in a Centro LB 960 Microplate luminometer (Berthold) by using the Dual-Glo Luciferase Assay System from Promega.

Cyclohexamide assays
Mouse mK4 cell line was maintained in DMEM/F12 media supplemented with 10%FBS, penicillin, and streptomycin. Transfection of 6xMyc-tagged RBPJ constructs (pCS2-RBPJ-6xMyc, WT and AOS variants) was carried out with TransIT-293 transfection reagent (Mirus) following the manufacture's instruction. After 24 h of the transfection, the cells were treated with cycloheximide (500ng/ml) and harvested at 48, 72, and 96 h. Protein samples were subjected to SDS-PAGE and subsequently transferred to PVDF membranes. The following primary antibodies were used at 1:1000 dilution in PBST (0.05% Tween-20) and 5% non-fat dry milk and incubated overnight at 4˚C: Myc (Cell Signaling Technology #2276), RBPJ (Cell Signaling Technology #5313), and actin (Cell Signaling Technology #5125). Protein bands were visualized by SuperSignal West Pico PLUS enhanced chemiluminescent reagent (ThermoFisher) and detected by ChemiDoc MP system (BioRad).

S3 Fig. Mammalian two-hybrid assay to analyze WT and Rbpj AOS variants binding to SHARP in cells.
HeLa cells were cotransfected with the indicated Gal4-SHARP (300ng) and Rbpj-VP16 (50ng) constructs together with the pFR-Luc reporter (500ng), which contains five Gal4 DNA binding sites upstream of the luciferase gene. Relative luciferase activity was determined after cotransfection of the pFR-Luc reporter construct alone. WT, E89G, and K195E activate the reporter similarly, suggesting that the AOS variants bind SHARP similar to WT Rbpj, whereas the Rbpj double-mutant F261A/L388A, which is compromised for SHARP binding, does not activate the reporter. ] P < 0.0001, unpaired Student's t-test). C. Luciferase reporter assays demonstrating that Rbpj mutants NBM and KRS are unable to bind NICD and DNA, respectively. HeLa RBPJ-KO cells were transfected with the Notch/RBPJ dependent reporter 12xCSL-RE-Luc (250 ng) without or with the indicated Rbpj or Rbpj-VP16 constructs (50 ng), and with or without NICD (10 ng). Luciferase activity was measured 24 hours after transfection. While Rbpj-VP16 NBM is able to activate the reporter, addition of NICD does not result in an addition increase in reporter activity due to Rbpj NBM unable to bind NICD. Rbpj-VP16 KRS and NBM/KRS are unable to activate the reporter with or without NICD. Bars represent mean values from six independent experiments and error bars indicate standard deviation. ([ ��� ] P < 0.0001, [NS] not significant, unpaired Student's t-test). (TIFF)