Making sense out of missense mutations: Mechanistic dissection of Notch receptors through structure-function studies in Drosophila

Abstract

Notch signaling is involved in the development of almost all organ systems and required post-developmentally to modulate tissue homeostasis. Rare variants in Notch signaling pathway genes are found in patients with rare Mendelian disorders, while unique or recurrent mutations in a similar set of genes are identified in cancer. The human genome contains four genes that encode Notch receptors, NOTCH1–4, all of which are linked to rare diseases and cancer. Although some mutations have been classified as clear loss- or gain-of-function alleles based on cellular or rodent based assay systems, the functional consequence of many variants/mutations in human Notch receptors remain unknown. In this review, I will first provide an overview of the domain structure of Notch receptors and discuss how each module is known to regulate Notch signaling activity in vivo using the Drosophila Notch receptor as an example. Next, I will introduce some interesting mutant alleles that have been isolated in the fly Notch gene over the past >100 years of research and discuss how studies of these mutations have facilitated the understanding of Notch biology. By identifying unique alleles of the fly Notch gene through forward genetic screens, mapping their molecular lesions and characterizing their phenotypes in depth, one can begin to unravel new mechanistic insights into how different domains of Notch fine-tune signaling output. Such information can be useful in deciphering the functional consequences of rare variants/mutations in human Notch receptors, which in turn can influence disease management and therapy.

Graphical Abstract

Missense alleles in the fly Notch gene isolated over the past >100 years of research have facilitated the understanding of Notch biology. By identifying unique alleles of the fly Notch gene through forward genetic screens, mapping their molecular lesions and characterizing their phenotypes in depth, one can begin to unravel new mechanistic insights into how different domains of Notch fine-tune signaling output. Such information can be useful in deciphering the functional consequences of rare variants/mutations in human Notch receptors, which in turn can influence disease management and therapy.

1. Unique aspects of the Notch signaling pathway

Notch signaling is conserved in most metazoan species and regulates many developmental decisions (Artavanis-Tsakonas and Muskavitch 2010; Yamamoto et al. 2014b). More recently, Notch signaling has been shown to be required in many post-developmental contexts, including tissue homeostasis (Bigas and Porcheri 2018) and neural physiology (Ables et al. 2011). Several features make the canonical Notch signaling pathway unique compared to other evolutionarily conserved core developmental signaling pathways, such as Wnt, Hedgehog, and BMP/TGF-β (Bier 2005; Salazar and Yamamoto 2018). For example, while many signaling pathways transmit signals to remotely located cells (paracrine or endocrine signaling), Notch signaling typically occurs between two juxtaposed cells (Hoppe and Greenspan 1986). One reason for the juxtacrine nature of this pathway is that both ligands [Delta or Serrate in Drosophila, Delta-like (Dll) and Jagged (Jag) family ligands in mammals] and Notch receptors are type-I transmembrane proteins that are anchored to the plasma membrane or found in other membranous compartments within the cell (Wharton et al. 1985a; Kopczynski et al. 1988; Fleming et al. 1990). In addition, unlike many cell-cell communication pathways in which the physical interaction between ligands and receptors is sufficient to trigger signal activation, Notch signal activation further requires several additional events that take place between the neighboring signal sending and receiving cells such as endocytosis that takes place in the signal sending cell (Parks et al. 2000; Nichols et al. 2007). Sequence of events following the ligand-receptor interaction ultimately causes the release of the intracellular domain of Notch (NICD) from the plasma membrane (Kopan and Ilagan 2009). Nuclear localization signals that are present in the NICD target this fragment to the nucleus where it forms a transcriptional activation complex with CSL (CBF-1, Suppressor of Hairless, Lag-2) transcription factors and Mastermind-family coactivators (Kovall and Blacklow 2010). Basic helix-loop-helix (bHLH) transcriptional repressors of the Enhancer of split [E(spl)] family proteins are representative examples of Notch signaling target genes that are evolutionarily conserved (Bray 2006). To prevent constitutive activation of the pathway, NICD becomes phosphorylated and polyubiquitinated for proteasomal degradation through its carboxyl-terminal PEST (proline, glutamic acid, serine, and threonine-rich) domain (Bray and Gomez-Lamarca 2018).

Another unique feature of Notch signaling is its stringent dosage sensitivity. In Drosophila, the Notch gene encodes the sole Notch receptor for this organism (Wharton et al. 1985a; Kopan and Ilagan 2009). Removal of one copy of Notch using a deletion or addition of an extra copy of Notch using a duplication both cause morphological alterations (Artavanis-Tsakonas and Muskavitch 2010). Similarly in mice that have four Notch paralogs (Notch1–4), Notch1 (Zhang et al. 2000; Koenig et al. 2017) and Notch2 (Witt et al. 2003) are also known to be haploinsufficient loci. Furthermore, haploinsufficiency of NOTCH1 in humans has been linked to Adams-Oliver syndrome (OMIM #616028) (Stittrich et al. 2014) and aortic valve diseases (OMIM #109730) (Garg et al. 2005). Considering that the amount of NICD that actively engages in transcription is tightly coupled to the amount of Notch receptors that are synthesized, activated and degraded due to the lack of any signal amplification steps (Salazar and Yamamoto 2018), even a modest reduction in the amount of receptors can have a significant impact on signal output.

2. Variants and mutations in human NOTCH genes that are linked to rare Mendelian diseases and cancer

Of the four genes encoding Notch receptors in the human genome, variants in NOTCH1, NOTCH2 and NOTCH3 have been associated with rare Mendelian diseases (Louvi and Artavanis-Tsakonas 2012; Salazar and Yamamoto 2018), whereas single nucleotide polymorphisms in NOTCH4 have been associated with increased risk of schizophrenia (Wei and Hemmings 2000) and systemic sclerosis (Cardinale et al. 2016; Zhou et al. 2019) in certain populations (Table 1). Some disease associated human NOTCH variants have been experimentally categorized as pathogenic loss- (LOF) or gain-of-function (GOF) alleles. For example, de novo or dominantly inherited missense or early frameshift variants in NOTCH2 that are linked to Alagille syndrome 2 (OMIM #610205) behave as LOF alleles (McDaniell et al. 2006). In contrast, late nonsense and frameshift alterations in the same gene are linked to Hajdu-Cheney syndrome (OMIM #102500), and these variants behave as GOF alleles because they result in formation of functional and more stable NICDs that lack the PEST domain (Isidor et al. 2011; Simpson et al. 2011). Interestingly in some cases, the functional consequences of disease-linked missense variants in NOTCH genes are still under debate. For example, a few variants in NOTCH3 linked to CADASIL (cerebral autosomal dominant arteriopathy with subcortical infarct and leukoencephalopathy, OMIM #125310) have been reported to behave as hypomorphic (partial LOF) alleles, while others have been reported to have no effect on Notch signaling (Joutel et al. 1996; Rutten et al. 2014). As whole-exome (WES) and whole-genome sequencing (WGS) become part of the general medical practice in the era of personalized medicine (Posey et al. 2019), the number of variants of unknown significance (VUS) in NOTCH and related genes will likely increase at a rapid pace. Understanding the precise mechanism of how each mutation affects protein function is critical since different therapeutic strategies may need to be employed based on the specific type of variant an individual carries (Wangler et al. 2017; Bellen et al. 2019; Harnish et al. 2019).

Table 1.

Human NOTCH receptor genes and their links to rare and common diseases. This table was assembled based on information gathered through Online Mendelian Inheritance in Man (OMIM, https://www.omim.org/) (Amberger et al. 2019), MARRVEL (http://marrvel.org/) (Wang et al. 2017; Wang et al. 2019a; Wang et al. 2019b)and PubMed (https://www.ncbi.nlm.nih.gov/pubmed). The term “early truncation” refers to nonsense or frameshift variants/mutations that are found prior to or within the ANK domain, whereas the term “late truncation” refers to such variants/mutations that are found after this domain.

Gene	Disease Study	Disease Name	OMIM#	Types of variants/mutations	Proposed mechanism	Representative References
NOTCH1	Mendelian diseases	Adams-Oliver syndrome 5	#616028	early truncation or missense	LOF (haploinsufficiency)	(Stittrich et al., 2014)
		Aortic valve disease 1	#109730	early truncation or missense	LOF (haploinsufficiency)	(Garg et al., 2005)
	Cancer	T-cell acute lymphoblastic leukemia (T-ALL)	-	missense in the NRR and late truncation	GOF (oncogene)	(Weng et al., 2004)
		B-cell Chronic Lymphocytic Leukemia (B-CLL)	-	late truncation	GOF (oncogene)	(Sportoletti et al., 2010)
		Squamous cell carcinoma	-	early truncation, in frame deletions or missense	LOF (tumor suppressor)	(Agrawal et al., 2011; Wang et al., 2011)
		Bladder transitional cell carcinoma	-	early truncation, in frame deletions or missense	LOF (tumor suppressor)	(Rampias et al., 2014)
		Small-cell lung carcinoma	-	early truncation or missense	LOF (tumor suppressor)	(George et al., 2015)
		Lower-grade glioma	-	early truncation, in frame deletions or missense	LOF (tumor suppressor)	(Cancer Genome Atlas Research et al., 2015)
NOTCH2	Mendelian diseases	Alagille syndrome 2	#610205	early truncation, missense	LOF (haploinsufficiency)	(McDaniell et al., 2006)
		Hajdu-Cheney syndrome	#102500	late truncation	GOF (hypermorphic)	(Isidor et al., 2011; Simpson et al., 2011)
	Cancer	Splenic marginal zone lymphoma	-	late truncation, missense	GOF (oncogene)	(Kiel et al. 2012; Rossi et al. 2012)
		Chronic myelomonocytic leukaemia	-	missense	LOF (tumor suppressor)	(Klinakis et al., 2011)
		Bladder transitional cell carcinoma	-	early truncation, in frame deletions or missense variants	LOF (tumor suppressor)	(Rampias et al., 2014)
		Small-cell lung carcinoma	-	early truncation or missense variants	LOF (tumor suppressor)	(George et al., 2015)
		Squamous cell carcinoma	-	early truncation, missense variants	LOF (tumor suppressor)	(Wang et al., 2011)
NOTCH3	Mendelian diseases	Cerebral arteriopathy with subcortical infarcts and leukoencephalopathy 1 (CADASIL)	#125310	missense in EGFr affecting cysteine residues (majority)	? (neomorphic? hypomorphic?)	(Joutel et al., 1996)
		Lateral meningocele syndrome	#130720	missense	GOF? (hypermorphic?)	(Gripp et al., 2015)
		Infantile myofibromatosis 2	#615293	late truncation	GOF (hypermorphic)	(Martignetti et al., 2013)
	Cancer	Bladder transitional cell carcinoma	-	early truncation or missense variants	LOF (tumor suppressor)	(Rampias et al., 2014)
		Squamous cell carcinoma	-	early truncation, in frame deletions or missense variants	LOF (tumor suppressor)	(Wang et al., 2011)
NOTCH4	GWAS	Susceptibility to schizophrenia	-	single nucleotide polymorphisms	?	(Wei and Hemmings, 2000)
		Susceptibility to systemic sclerosis	-	single nucleotide polymorphisms (missense)	?	(Cardinale et al., 2016)
	Cancer	Squamous cell carcinoma	-	early truncation, missense	GOF? (oncogene?)	(Wang et al., 2011)

In addition to the involvement in a handful of genetic diseases, somatic mutations in human NOTCH genes as well as alterations in their expression are found in different types of cancer (Nowell and Radtke 2017; McCarter et al. 2018; Aster et al. 2017). Similar to variants linked to rare diseases, in vitro and in vivo studies of cancer-associated NOTCH alleles have led to the classification of some mutations as LOF or GOF (Table 1). For example, late-truncation mutations in the NOTCH1 gene that remove the PEST domain are often found together with missense mutations that affect the LNR (Lin-12/Notch repeat) domain (see below) in T-cell acute lymphoblastic leukemia (T-ALL) (Weng et al. 2004). These two types of mutations act synergistically to activate Notch signaling in a ligand-independent manner to produce a NICD that is highly stable (Aster et al. 2009). Hence, these are GOF alleles and NOTCH1 acts as an oncogene in this context. Similar late-truncating and missense GOF mutations in NOTCH1 and NOTCH2 are also frequently found in B-cell chronic lymphocytic leukemia (B-CLL) (Sportoletti et al. 2010; Fabbri et al. 2011; Puente et al. 2011) and splenic marginal zone lymphoma (Kiel et al. 2012; Rossi et al. 2012), respectively. On the contrary, LOF mutations in NOTCH genes are frequently found in chronic myelomonocytic leukemia (Klinakis et al. 2011), bladder transitional cell carcinoma (Rampias et al. 2014), small-cell lung carcinoma (George et al. 2015), lower-grade glioma (Cancer Genome Atlas Research et al. 2015), and squamous cell carcinoma of the head, neck, skin and lung (Agrawal et al. 2011; Wang et al. 2011), indicating that Notch signaling acts as a tumor suppressor in these tissues. Therefore, while pharmacological strategies to inhibit Notch activity may be useful to treat certain types of leukemia, drugs that activate Notch signaling may be effective against other types of leukemia and specific types of solid tumors. Notch signaling is highly context specific, and can even perform completely opposite tasks in different tissues, cell types, and developmental stages (e.g. promoting growth in one context while inducing cell death in another). Thus in order to facilitate precision medicine, it is critical to understand the functional consequences of diverse NOTCH receptor mutations identified from cancer tissue sequencing efforts to design appropriate treatment strategies (Letai 2017).

3. Domain structure of Notch receptors and their known function

Notch receptors are composed of relatively large extracellular and intracellular domains and a single transmembrane domain that connects the two. The extracellular domain of the fly Notch receptor is ~180kDa in size, whereas the intracellular domain is ~120kDa. A number of evolutionarily conserved domains and features are found along the entire length of this ~300kDa protein (Figure 1). Here, I highlight each functional unit within the fly Notch receptor starting from the amino-terminal of the protein and discuss their roles in signal transduction.

Figure 1.

Schematic diagram of the Drosophila Notch receptor and mutations found in the epidermal growth factor-like repeats. Note that the three dimensional conformation of the full Notch protein has not yet been elucidated, and that this diagram is not drawn to scale. For simplicity, the Notch receptor is depicted here as a monomer but it may function as a dimer or multimer in vivo (see main text). Abbreviations: SS (signal sequence), EGFr (epidermal growth factor-like repeat), LNR (LIN-12/Notch repeat), HD (heterodimerization domain), NRR (negative regulatory region), RAM (RBP-jκ Associated Molecule domain), ANK (ankyrin repeat), NLS (nuclear localization signal), TAD (transactivation domain), PEST (proline, glutamic acid, serine, and threonine-rich domain).

i). Epidermal growth factor-like repeats (EGFr)

The amino acid sequence of the Drosophila Notch protein starts with a signal sequence [52 amino acids (a.a.) according to UniProt (The UniProt 2017), https://www.uniprot.org/uniprot/Q32UW5] followed by 36 epidermal growth factor-like repeats (EGFr). Human NOTCH1 and NOTCH2 also contain 36 EGFr but NOTCH3 and NOTCH4 have fewer EGFr: 34 in NOTCH3 and 29 in NOTCH4, respectively (Kopan and Ilagan 2009). Each EGFr is approximately 40 a.a. and contains six cysteine residues that form three disulfide bonds to stabilize their modular structures (Stenflo 1991). Some EGFr contain Ca⁺² binding motifs at their amino-terminus and these are specifically referred to as calcium binding EGFr (cbEGFr) (Handford et al. 1991). When Ca⁺² is bound to cbEGFr, it allows the EGFr to be packed into a rigid, rod-like conformation, reducing the flexibility between the cbEGFr and the preceding EGFr (Hambleton et al. 2004). In Drosophila Notch, 21 EGFr are predicted to be cbEGFr and the positions of these cbEGFr are highly conserved across evolution (Jafar-Nejad et al. 2010; Rana and Haltiwanger 2011), suggesting this pattern may help Notch receptors adopt a certain three-dimensional conformation in vivo.

Many EGFr are also subjected to post-translational O-glycosylation on the hydroxyl “O” of certain serine (S) or threonine (T) residues (Stanley and Okajima 2010). EGFr that are subjected to O-fucosylation carry a consensus sequence motif, C₂-X-X-X-A/G/S-(S/T)-C₃ (C₂ and C₃ indicate the 2^nd and 3^rd cysteine of an EGFr, X can be any amino acid; A, G, S, T are alanine, glycine, serine and threonine, respectively; O-fucosylation occurs on the S/T), which is the target site for the protein O-fucosyltransferase 1 (O-fut1). In Drosophila Notch, 22 EGFr are predicted to contain the O-fucosylation motif and most of these sites are indeed O-fucosylated at high stoichiometry in cultured cells (Harvey et al. 2016). O-fucosylation of Notch takes place in the endoplasmic reticulum (ER) (Okajima et al. 2005), and O-fut1 can only recognize properly folded EGFr (Wang and Spellman 1998), serving as one of the quality control mechanisms during Notch biosynthesis to ensure the production of Notch proteins with the correct structural conformation. LOF mutations or RNAi mediated knockdown of O-fut1 cause strong loss of Notch signaling phenotypes in a variety of contexts (Okajima and Irvine 2002; Sasamura et al. 2003). In O-fut1 mutant cells, Notch accumulates in the ER and fails to be exocytosed. Since this trafficking defect can be rescued by over-expression of an enzymatically inactive form of O-fut1, this protein is thought to also function as a molecular chaperone that assists in the proper folding of Notch in the ER in addition to its catalytic role as a protein glycosyltransferase (Okajima et al. 2005; Okajima et al. 2008). However, since several reports in flies and mice suggest that Notch can be delivered to the cell surface in the absence of O-fut1 (Pofut1 in mouse), the role of O-fut1 and its chaperone function is still under debate (Okamura and Saga 2008; Stahl et al. 2008; Yao et al. 2011; Sasamura et al. 2007; Sasaki et al. 2007). Furthermore, by analyzing a catalytically dead O-fut1 allele generated through a gene knock-in methodology in Drosophila, Notch receptors that lack O-fucosylation were shown to exhibit defects in endocytic trafficking at high temperature (30°C) (Ishio et al. 2015). Together with O-glucosylation that takes place on specific EGFr (see below), monosaccharide modification of Notch on multiple EGFr seems to function as a mechanism to ensure proper Notch signaling activity under certain conditions. Considering that the role of endocytic trafficking has been proposed to function as a buffering mechanism to provide robustness to Notch signaling under different temperature conditions (Shimizu et al. 2014), it would be interesting to see how changes in the structure of the extracellular domain of Notch impacts endocytic trafficking.

O-fucosylation on EGFr of Notch mediated by O-fut1 can be further elongated in the Golgi apparatus by Fringe (Fng), a β−1,3-N-acetylglucosamintransferase that catalyzes the addition of N-acetylglucosamine (GlcNAc) to O-fucosylated EGFr (Munro and Freeman 2000; Moloney et al. 2000a; Bruckner et al. 2000). The primary function of Fng is to change the affinity of the interactions between Notch and its ligands. When Notch is not Fng modified, it can bind strongly to Serrate but weakly to Delta. On the contrary, when Notch becomes Fng modified, this ligand selectivity becomes reversed and Notch can now bind strongly to Delta but only weakly to Serrate. In vitro experiments suggest that multiple EGFr that are modified by O-fut1/Fng contribute to ligand-receptor interactions and ligand selectivity (Xu et al. 2007; Harvey et al. 2016). In addition to studies based on genetics and biochemistry, recent crystallography and nuclear magnetic resonance (NMR) spectroscopy based structural biological approaches are beginning to unravel how Notch molecularly distinguishes Delta family ligands from Serrate/Jagged family ligands (Baron 2017; Handford et al. 2018), which will be discussed later in this article.

In addition to the O-fut1 mediated O-fucosylation and subsequent elongation by Fng, some EGFr of Notch can be covalently modified by O-glucose and xylose. EGFr that are O-glucosylated carry a motif, C₁-X-S-X-P-C₂ (C₁ and C₂ indicate the first and second cysteine of an EGFr, X can be any amino acid; S, P are serine and proline, respectively; O-glucosylation occurs on the S), that is recognized by Rumi, a protein O-glucosyltransferase located in the ER (Harvey and Haltiwanger 2018). Mutant alleles of rumi were originally isolated as temperature sensitive mutations in Drosophila that show severe Notch LOF phenotypes through a forward genetic screen (Acar et al. 2008). O-glucosylation of Notch is dispensable for ligand-receptor interactions but essential for S2 (site-2) cleavage (see below) of Notch at restrictive temperatures (above 25°C). At permissive temperatures (below 18°C), however, S2 cleavage and signal activation occurs normally, suggesting that alterations in thermodynamics may affect local or overall structure of the extracellular domain of Notch that is required for proper processing by ADAM proteases (Kuzbanian in Drosophila, ADAM10 in human) that mediates this cleavage. In mice, Poglut1 (rumi ortholog) knockout animals are embryonic lethal and exhibit a number of Notch signaling related developmental defects (Fernandez-Valdivia et al. 2011). In Drosophila Notch, 18 EGFr carry the consensus O-glucosylation motif and these sites are indeed efficiently modified when tested in cultured cells (Jafar-Nejad et al. 2010; Takeuchi and Haltiwanger 2010; Harvey et al. 2016). Through elegant structure-function studies performed in vivo in Drosophila using bacterial artificial chromosome (BAC) based genomic rescue constructs, O-glucosylated residues were found to contribute to robust signal activation at high temperature in an additive fashion, and no single site was found to be solely responsible for the rumi mutant phenotype (Leonardi et al. 2011).

Studies in mammalian cells as well as Drosophila have shown that the O-glucose that is added by Rumi can be further elongated by two contiguous α−1,3-linked xylose residues (Bakker et al. 2009; Moloney et al. 2000b). The addition of a xylose to glucose is mediated by a UDP-D-xylose:β-D-glucoside α−1,3-D-xylosyltransferase that is encoded by the shams gene in Drosophila (Lee et al. 2013). shams negatively regulates Notch signaling by modulating trafficking and ligand-receptor interaction of the Notch receptor (Lee et al. 2013; Lee et al. 2017). The first xylose added by Shams can be further extended by the Xyloside xylosyltransferase (Xxylt) (Pandey et al. 2018). While no obvious morphological defects are detected in Xxylt null animals, these mutants can genetically suppress the dominant haploinsufficient wing notching phenotype seen in heterozygous Notch null alleles, suggesting a modulatory role in Notch signaling that can be revealed in a sensitized genetic background (Pandey et al. 2018). Furthermore, studies on fly mutants defective in UDP-xylose (source of xylose added to EGFr) synthesis indicated that xylosylation and O-fucosylation have redundant roles in regulating Notch trafficking (Matsumoto et al. 2016), indicating that O-glycosylation of Notch has numerous unique as well as overlapping roles in fine-tuning Notch signaling. To further complicate things, biochemical studies have shown that both Drosophila and mammalian Rumi can act as a protein O-xylosyltransferase to mediate the addition of O-xylose instead of O-glucose on EGFr, establishing a dual substrate specificity (Takeuchi et al. 2011). Additionally in mammals, two Rumi-related proteins, Poglut2 and Poglut3, have been identified to O-glucosylate specific EGFr in NOTCH1 and NOTCH3 independent of Rumi/Poglut1 (Takeuchi et al. 2018). Although the function of the fly homolog of mouse Poglut2/3 encoded by the CG31139 gene has yet to be characterized, these and other studies reveal the intricate and complicated nature of post-translational modification of the extracellular domain of Notch receptors.

ii). LIN-12/Notch repeat (LNR) and heterodimerization (HD) domains

The LNR domain consists of three LNR motifs (Figure 1) (Lovendahl et al. 2018). Each LNR motif consist of a ~40 amino acid module that contains six cysteine residues that form three intramolecular disulfide bonds and a Ca²⁺ binding motif, similar to cbEGFr (Aster et al. 1999; Vardar et al. 2003). This evolutionarily conserved LNR domain is always present in the extracellular domain immediately following the long stretch of EGFr in all Notch receptors studied to date (Kopan and Ilagan 2009). Other than a few genes in the pappalysin family of metalloproteinases that contain LNR-like motifs (Overgaard et al. 2003), LNR motifs are uniquely found in Notch receptor family proteins, suggesting that this structure may have evolved to carry out a very specific function in Notch signaling regulation. The region between the LNR domain and the transmembrane domain is also highly conserved across evolution and is referred to as the HD domain. The HD domain is cleaved [S1 (site-1) cleavage] by Furin or Furin-like proteases during the biosynthesis process of Notch receptors in the Golgi apparatus (Logeat et al. 1998; Lake et al. 2009). The two pieces of Notch cleaved at the S1 site are still held together by strong noncovalent bonds within the HD domain, hence forming a “heterodimer” (Blaumueller et al. 1997). The three tandem LNR motifs together with the subsequent HD domain are often referred to as the negative regulatory region (NRR) because this structure protects the S2 cleavage site, located within the HD domain, from ectopic cleavage by ADAM proteases in the absence of the ligand (Gordon et al. 2007). Upon ligand binding and trans-endocytosis of the Notch extracellular domain into the signal sending cell, a mechanical pulling force opens up the NRR by dissociating the LNR domain from the HD domain (Lovendahl et al. 2018). This conformational change allows ADAM proteases to access the HD domain and execute the S2 cleavage. The extracellular domain of S1/S2 cleaved Notch is co-endocytosed into the signal sending cell together with the ligand and becomes degraded through the lysosomal pathway (D’Souza et al. 2010; Yamamoto et al. 2010). The remaining membrane bound S2 cleaved Notch is often referred to as NEXT (Notch extracellular truncation) and contains a very short extracellular domain, the transmembrane domain (TMD) and NICD. NEXT is subjected to subsequent S3 (site-3) cleavage by the γ-secretase complex (Levitan and Greenwald 1995; De Strooper et al. 1999; Struhl and Greenwald 1999; Ye et al. 1999). Ca²⁺ in the LNR domains also contributes to the stabilization of the NRR because chelation of extracellular Ca²⁺ leads to ligand-independent activation of Notch signaling in cultured cells (Rand et al. 2000).

iii). Transmembrane Domain (TMD)

The TMD of Drosophila Notch is composed of a short (21 a.a. according to UniProt) stretch of hydrophobic amino acids and the S3 cleavage site is located within the TMD proximal to the intracellular domain (Figure 1) (Kopan and Ilagan 2009). NEXT is recognized by γ-secretase, a multiprotein protease complex that mediates the intra-membrane cleavage of Notch and other transmembrane proteins (Jorissen and De Strooper 2010). Following the S3 cleavage, γ-secretase performs an additional site-4 (S4) cleavage within the TMD proximal to the extracellular domain, generating small ~20 residue peptides called Nβ (Okochi et al. 2002). This is analogous to Aβ (amyloid-β) peptides generated from amyloid precursor protein (APP) that aggregate and accumulate as amyloid plaques in brains of Alzheimer’s disease patients (Selkoe and Hardy 2016). Interestingly, in vitro data suggest that familial Alzheimer disease linked variants in Presenilin genes, encoding the catalytic subunits of the γ-secretase complex, cause a shift not only in γ-cleavage site of APP but also the S4 cleavage site of Notch to generate Nβ peptides that are longer than usual (Okochi et al. 2006). However, the in vivo significance of this S4 cleavage and Nβ peptide is still unknown.

iv). Notch Intracellular Domain (NICD)

NICD consists of a single RAM (RBP-jκ Associated Molecule) domain, seven ankyrin (ANK) repeats, a transactivation domain (TAD) and a PEST sequence (Figure 1) (Bray and Gomez-Lamarca 2018). In Drosophila, a poly-glutamine (Q)-rich domain called the OPA domain is present within the TAD (Wharton et al. 1985b) but this feature is not shared among mammalian Notch receptors. In addition, NICD carries multiple nuclear localization sequences (NLS) (Kopan et al. 1994), as well as target sites for multiple post-translational modifications including phosphorylation (Espinosa et al. 2003; Ranganathan et al. 2011), acetylation (Guarani et al. 2011; Palermo et al. 2012), methylation (Hein et al. 2015), and ubiquitination (Fryer et al. 2004). NICD released from the plasma membrane translocates into the nucleus via the NLS mediated nuclear import machinery (Huenniger et al. 2010; Sachan et al. 2013).

In the nucleus, fly NICD interacts with Suppressor of Hairless [Su(H)], a DNA binding transcription factor, through the RAM and ANK domains (Kovall and Hendrickson 2004; Bray 2006). Su(H) and its orthologs are often referred to as CSL (CBF1/Su(H)/LAG-1) proteins and the human CSL is encoded by the RBPJ (Recombination signal binding protein for immunoglobulin kappa J region) gene (Tanigaki and Honjo 2010). CSL proteins bind to consensus DNA sequences (5’-TGGGAA-3’) in both the absence or presence of NICD. In the absence of NICD, Su(H) that is bound to DNA recruits transcriptional co-repressors such as Hairless, CtBP (C-terminal Binding Protein) and Groucho, which in turn recruit histone deacetylases (HDACs) and other repressive chromatin regulators to negatively regulate the expression of Notch target genes (Nagel et al. 2005; Morel et al. 2001; Kao et al. 1998; Oswald et al. 2005). Upon Notch signal activation and NICD binding to Su(H), the co-repressor complex disassembles, leading to de-repression of target gene transcription (Morel and Schweisguth 2000; Barolo et al. 2002). Mastermind, a co-activator that interacts with both NICD and Su(H), is also recruited to the complex to form a tripartite complex (Nam et al. 2006; Wilson and Kovall 2006). NICD, Su(H) and Mastermind together recruit histone acetyltransferases (HATs) such as p300 and other chromatin remodeling complexes to actively promote gene transcription and expression of Notch target genes (Kurooka and Honjo 2000; Oswald et al. 2001; Fryer et al. 2002). Signal termination is mediated by phosphorylation and subsequent ubiquitin proteasome pathway mediated degradation of NICD. Although this process has not been extensively studied in Drosophila, mammalian studies suggest that phosphorylation of NICD by kinases such as CDK8 (Cyclin-dependent kinase 8) (Fryer et al. 2004) and poly-ubiquitination by E3 ubiquitin ligases such as FBXW7 (F-Box And WD Repeat Domain Containing 7) (Oberg et al. 2001; Wu et al. 2001; Gupta-Rossi et al. 2001) facilitate the degradation of NICD in the proteasome (Schweisguth 1999; Kopan 1999).

4. Unique missense mutations in the fly Notch gene

Since the isolation of the first female fly with a notched wing phenotype in 1913 by John S. Dexter in Thomas Hunt Morgan’s laboratory (Dexter 1914), over 350 alleles of Notch (official gene symbol: N) have been reported to date in FlyBase (http://flybase.org/reports/FBgn0004647#alleles_main). Other than a few mutants such as the original N allele that appeared spontaneously while housing a large number of flies in culture (Morgan and Bridges 1916), most alleles were obtained through diverse mutagenesis techniques including physical mutagenesis using X-rays and γ-rays, chemical mutagenesis using EMS (ethyl methanesulfonate) and ENU (ethyl nitrosourea), and mobilization of transposable elements such as P-elements and piggybac (Lindsley and Zimm 1992). Surprisingly, the molecular identity of majority of reported Notch alleles are still unknown because most were identified and characterized based on their phenotypes and basic genetic experiments such as complementation tests and rescue experiments. Many mutants including N^55e11 and N^54l9 [also known as Df(1)N-54l9] that have been frequently used in genetic experiments behave as null alleles (amorph) (Kidd et al. 1983; Mohler 1956). Amorphic N alleles were critical for the mapping and cloning of the Notch gene in 1985 (Wharton et al. 1985a), a key event in Notch signaling research history that catalyzed the dramatic expansion of the field in the following three decades. Amorphic Notch alleles in Drosophila are recessive lethal during embryogenesis and exhibit numerous developmental defects including a characteristic “neurogenic” phenotype that leads to epidermal-to-neuronal transformation during cell fate specification of the ectoderm (Poulson 1937; Lehmann et al. 1981). In fact, this neurogenic phenotype of N became the first bridge between two academic fields that were considered to be incompatible with one another at the time: genetics and developmental biology (Artavanis-Tsakonas and Muskavitch 2010). FLP/FRT based clonal analysis of amorphic N alleles have also revealed the involvement of Notch signaling in various post-embryonic developmental contexts (Salazar and Yamamoto 2018).

Currently (as of July 2019), 106 out of 367 Notch alleles curated in FlyBase (http://flybase.org) have been reported to have a defined molecular or cytological lesion. Out of these 106 defined Notch alleles, only 18 are missense alleles (Table 2) (Hartley et al. 1987; Kelley et al. 1987; Brennan et al. 1997). Despite this small number, genetic, phenotypic, biochemical and cell biological studies on a subset of these mutations have provided novel insights into how Notch signaling is regulated in vivo. In this section, I will discuss how studies of these missense alleles in Drosophila have improved our understanding of ligand-receptor interactions, higher-order protein complex formation and non-canonical signaling.

Table 2.

Missense mutations in the Drosophila Notch gene reported to date in FlyBase (http://flybase.org/) (Thurmond et al. 2019). Abbreviations: EGFr (epidermal growth factor-like repeat), LNR (LIN-12/Notch repeat), HD (heterodimerization), ANK (ankyrin repeat), TAD (transactivation domain).

Drosophila Notch allele	FlyBase ID	Affected Domain(s)	Amino acid change(s)	Notes
nd-3	FBal0012892	EGFr-3	p.C105F	Weak hypomorphic allele (affects wing)
jigsaw	FBal0280339	EGFr-8	p.V361M	Separation of function allele (inhibits Notch-Serrate binding)
M1	FBal0029949	EGFr-12	p.E491V	Antimorphic allele (likely have an impact on both Notch-Delta and Notch-Serrate binding)
spl-1	FBal0012900	EGFr-14	p.I578T	Weak hypomorphic allele (affects eye and bristle), but may also be hypermorphic or antimorphic
Mcd5	FBal0130887	EGFr-18	p.C739Y	Separation of function allele (affects non-canonical signaling)
Ax-59b	FBal0012853	EGFr-24	p.C972G	Complecated alleles (Abruptex alleles that have mixed features of gain- and loss-of-function)
Ax-59d	FBal0012854		p.C972YS
Ax-9	FBal0012851		p.D948V
Ax-1	FBal0012850	EGFr-25	p.N986I
Ax-M1	FBal0029940		p.C999Y
Ax-71d	FBal0028450	EGFr-27	p.S1088I
Ax-E2	FBal0012858	EGFr-29	p.H1167Y
Ax-16	FBal0012852		p.G1174A
ts1	FBal0012887	EGFr-32	p.G1272D	Temperature sensitive hypomorphic alleles
l1N-B	FBal0012886	LNR3	p.G1573V	Hypermorphic alleles
414	FBal0066009	LNR3 & HD	p.D1577G & p.Q1641R
su42c	FBal0033873	ANK-5	p.A2060V	Strong hypomorphic alleles
1081	FBal0058512	TAD	p.G2490C

i. N^M1: an antimorphic allele in the ligand-receptor interaction domain

For Drosophila Notch and human NOTCH proteins that have 36 EGFr (NOTCH1 and NOTCH2) in their extracellular domains, EGFr-11 and -12 are absolutely essential for ligand-receptor binding and signaling (Tien et al. 2009; Kovall and Blacklow 2010). Through cell culture based studies in Drosophila, Notch without EGFr-11/12 was shown to be completely defective in ligand-receptor binding (Rebay et al. 1991). In mice, an internal deletion of multiple EGFr including EGFr-11/12 of Notch1 (EGFr-8~12) also exhibited strong LOF phenotypes that resulted in embryonic lethality indistinguishable from Notch1 null mutant embryos (Ge et al. 2008). Genetic, biochemical, and cell biological experiments all pointed to the importance of these specific EGFr in ligand-receptor interactions, and molecular structures of some portions of Notch and its ligands were determined through crystallography (Cordle et al. 2008) and NMR (Hambleton et al. 2004). However, the actual structural configuration of Notch receptors bound to its ligands was not solved until very recently due to the large size of the proteins, transmembrane nature of both ligands and receptors, and relatively weak molecular interactions that had been difficult to capture. In 2015, the first co-crystal structure of a Notch receptor binding to a Delta family ligand was reported, providing direct structural evidence that EGFr-11 and EGFr-12 are indeed directly engaging in ligand-receptor interactions (Luca et al. 2015). Subsequent study by the same group further showed that EGFr-11/12 are also directly involved in the interaction between Notch receptors to Serrate/Jagged family ligands (Luca et al. 2017). Although these studies are still based on relatively small fragments of Notch and its ligands compared to the full-length proteins (EGFr-11~13 for Notch1-Dll4 complex, EGFr-8~12 for Notch1-Jag1 complex), they are indeed critical milestones towards a full structural understanding of the Notch receptor in action in vivo.

The N^M1 allele was first reported in 1993 through an EMS-mediated chemical mutagenesis screen and carries a p.E491V mutation in the Ca²⁺ binding domain of EGFr-12 (de Celis et al. 1993; Baron 2017). Hemizygous and homozygous N^M1 mutants exhibit recessive embryonic lethality and neurogenic phenotypes similar to amorphic alleles (Couso and Martinez Arias 1994; de Celis et al. 1993). Furthermore, heterozygous N^M1 mutant flies also exhibit dominant morphological phenotypes in the wing and notum. Interestingly, these dominant phenotypes are stronger than amorphic alleles and cannot be fully rescued by introducing a wild-type copy of Notch, suggesting that this allele encodes a dominant negative (antimorphic) protein (de Celis et al. 1993). Although the structural consequences of this allele have not been further investigated, the fact that this mutant carries a missense change in EGFr-12 suggests that the N^M1 mutant may have altered ligand binding properties. Since one would expect that a simple loss of ligand binding ability should behave as an amorphic allele similar to the Notch1 knock-in mouse that lacks the ligand binding domain (Ge et al. 2008), understanding how this allele behaves as an antimorphic allele will likely provide additional structural insights into ligand-receptor binding and signal activation.

ii. N^jigsaw: a separation of function allele for ligand selectivity

Although EGFr-11 and EGFr-12 are necessary for ligand-receptor interactions, they are not fully sufficient. For example, although a small fragment of Notch containing only EGFr-11/12 was shown to be able to interact with its ligands based on cellular assays, this interaction is significantly weaker compared to the full length Notch (Xu et al. 2005). Additional experiments that deleted different domains of Notch and biochemically assessed their binding ability showed that EGFr-6~9 and EGFr-25~36 strengthen the ligand-receptor binding mediated by EGFr-11/12, while EGFr-1~5 are dispensable for or may even play a negative role in these interactions (Xu et al. 2005). In addition to some EGFr being involved in regulating both Notch-Delta and Notch-Serrate interactions, other domains within the extracellular domain were predicted to function in distinguishing the two ligand types based on data obtained from fng mutants (Irvine and Wieschaus 1994; Fleming et al. 1997; Panin et al. 1997). However, considering that Fng modifies many EGFr scattered across the protein (Harvey et al. 2016), whether a single or multiple EGFr were responsible for this function was unclear.

We reported the isolation of N^jigsaw in 2012, a peculiar missense allele (p.V361M) in EGFr-8 that was obtained through a somatic mosaic chemical mutagenesis screen using EMS (Yamamoto et al. 2012). This screen was originally designed to isolate recessive lethal mutations on the fly X-chromosome that exhibit developmental or neurological phenotypes in homozygous mutant clones induced in a tissue specific manner using the FLP/FRT system (Yamamoto et al. 2014a; Deal and Yamamoto 2018). In addition to discovering a new gene that regulates Notch signaling (Charng et al. 2014b), we isolated 42 new recessive lethal alleles of Notch. Through molecular and cell biological characterization of these mutants, we found 20 of the 42 alleles were missense alleles that affect different domains of Notch (Yamamoto et al. 2012). N^jigsaw was distinct from other alleles for several reasons. First, flies with homozygous N^jigsaw clones exhibited strong wing phenotypes while having no effect on the development of mechanosensory bristles. Considering that flies with amorphic N clones exhibit strong phenotypes in both wings and bristles, this suggested that N^jigsaw is a separation of function allele. Second, in the flies with homozygous N^jigsaw clones that exhibited wing defects, we observed both loss (notching) and gain (ectopic) of wing margin tissue even within the same wing, analogous to a piece of a jigsaw puzzle. Because the former phenotype is associated with loss of Notch signaling while the latter is linked to gain of Notch signaling, this result suggested that this mutant may not be a simple LOF or GOF allele. Third, considering that a wild-type copy of Notch was able to rescue all phenotypes we observed in this mutant, we hypothesized that N^jigsaw was some sort of a LOF allele that was missing a specific function.

Through series of experiments in vivo and in vitro, we demonstrated that this specific missense mutation in EGFr-8 strongly diminishes the ability of Notch to bind to Serrate without affecting its interactions with Delta. We further determined that N^jigsaw was exhibiting both loss and gain of Notch signaling phenotypes in mosaic tissues because this mutation prevented both trans and cis interactions between Notch and Serrate. In addition to binding to ligands presented from neighboring cells in trans, Notch receptors can interact with ligands that are expressed in the same cell in cis (del Alamo et al. 2011; Sprinzak et al. 2010). Ligands presented in cis and trans both compete for the same ligand binding domain of the receptor (Cordle et al. 2008). Since cis-interactions cannot trigger the S2 cleavage event due to lack of mechanical tension that is provided by trans-endocytosis of the ligand, Notch signaling diminishes cell autonomously when ligands and receptors are bound in cis. This “cis-inhibition” was first documented through ligand over-expression experiments during wing development in Drosophila (Micchelli et al. 1997), and was subsequently shown to fine-tune the level of Notch signaling in vivo during the development of the fly eye (Miller et al. 2009) and oogenesis (Palmer et al. 2014).

A variation of “cis-inhibition” between Notch and Serrate (but not Delta) has been reported and referred to as “ligand cis-inhibition” (Becam et al. 2010). In this scenario, Notch receptors regulate the amount of Serrate present at the cell surface through cis-interactions and subsequent endolysosomal degradation. Since N^jigsaw cannot bind to Serrate, the level of Serrate in N^jigsaw mutant cells increases, which in turn mediates ectopic activation of Notch signaling in neighboring wild-type cells in a mosaic animal (Yamamoto et al. 2012). By solving this “N^jigsaw puzzle”, we were able to identify the first residue outside of EGFr-11/12 in Notch that was responsible for ligand binding and discrimination, a finding that was explained structurally when the first co-crystal structure of a Notch receptor binding to a Serrate/Jagged family ligand was published five years later (Luca et al. 2017). Considering that a variant analogous to N^jigsaw in mouse Notch2 (p.V327M) also exhibited defects in Jag1 mediated signaling without affecting Dll1 mediated signaling in cultured cells (Yamamoto et al. 2012), similar variants and alleles in human NOTCH orthologs may impact ligand binding and selectivity in both cis and trans.

iii). Abruptex (Ax) mutations: complicated allelic series that suggest Notch receptors functions as dimers or multimers in vivo

The region between EGFr-24 and EGFr-29 of Drosophila Notch is referred to as the Abruptex (Ax) domain (Kelley et al. 1987). This name originates from a series of missense alleles that cluster in this region, first documented in 1930 (Nasarenko 1930). All N^Ax mutants exhibit shared dominant phenotypes that are characterized by shortening (abrupt) of the fifth longitudinal wing vein and mild reduction of mechanosensory bristles on the dorsal thorax (Lindsley and Zimm 1992). Because these dominant phenotypes can be explained by mild increases in Notch signaling (de Celis and Garcia-Bellido 1994; Heitzler and Simpson 1993; de Celis and Bray 2000), this domain likely plays an inhibitory role in Notch signaling in certain contexts. Molecularly, the Ax domain can physically interact with a region in the extracellular domain that contains EGFr-11/12 (Pei and Baker 2008), suggesting they may mediate intra- or inter-molecular interactions within or between Notch proteins. Hence, ligands may need to physically compete with the Ax domain and overcome this inhibitory interaction to be able to bind and activate the receptor. In Notch receptors that carry Ax alleles, the interaction between the ligand binding domain and Ax domain may be lost or reduced, allowing Notch to be more easily activated. However, structural evidence is currently lacking to support this model since a crystal or cryo-electron microscopy (EM) structure of a Notch receptor that includes the Ax domain is still lacking.

It is important to note, however, that many aspects of Ax mutations cannot be explained by a simple GOF mechanism, and we still do not have a full understanding of what this domain actually does in vivo. Some Ax mutants are recessive lethal (lethal Ax), while others are homozygous viable. Homozygous viable Ax mutants can be further subdivided into two classes based on how they behave when crossed to amorphic Notch alleles. The first class of Ax alleles is called “enhancer of Notch” because transheterozygous flies (N^Ax/N^null) have stronger wing notching phenotypes compared to flies heterozygous for an amorphic allele (N⁺/N^null). The second class of mutants is called “suppressor of Notch” because N^Ax/N^null animals exhibit milder wing notching phenotypes compared to N⁺/N^null. Ax alleles that are “suppressor of Notch” are found in EGFr-24 and EGFr-25 whereas “enhancer of Notch” are located in EGFr-27 and EGFr-29 (Kelley et al. 1987), suggesting that the Ax domain can be further functionally subdivided. Surprisingly, transheterozygous (compound heterozygous) flies with one “enhancer of Notch” Ax allele and another “suppressor of Notch” Ax allele are lethal or semi-lethal (Foster 1975; Portin 1975). This extremely rare genetic phenomenon is called “negative complementation” because it is completely opposite of the well-known “complementation” phenomenon in which two recessive lethal mutants become viable when put in trans (often used as a genetic evidence that two mutations affect different genes or different functional domains within a single protein).

One molecular explanation for this mysterious “negative complementation” seen among certain allelic combinations of N^Ax mutants is that the true functional unit of a Notch receptor in vivo is not a monomer, but a dimer or multimer (de Celis and Garcia-Bellido 1994). Notch receptors within one class of Ax mutation may still be able to form functional complexes with themselves, but these mutants may lack compatibility with Notch receptors that carry the other class of Ax alleles. Hence, when two classes of Ax Notch receptors become expressed in a single cell in transheterozygous animals, most multi-Notch receptor containing complexes may become nonfunctional. This idea is supported by several biochemical experiments including co-immunoprecipitation (Vooijs et al. 2004), denaturing gel electrophoresis (Kidd et al. 1989) and cross-linking experiments (Sakamoto et al. 2005). Recently, a single particle EM based study visualized the extracellular domain of Notch as a dimer that can be found in different conformational states (Kelly et al. 2010). It is interesting to note that the NICD has also been shown to form a dimer when bound to paired CSL-binding sites on DNA to boost transcription of certain target genes (Arnett et al. 2010). Hence, it is important to understand the structural basis by which different domains within both extracellular and intracellular regions of Notch contribute to higher-order complex formation to obtain a full understanding of how this receptor works at the molecular level. Further structural biological studies of Notch and its ligands in their native states at atomic resolution using techniques such as cryo-EM (Yang et al. 2019) may allow scientists to observe how these proteins actually interact in vivo.

iv). Microchaetae defective (Mcd) mutations: series of alleles that implicate Notch in a non-canonical signaling pathway

A series of Microchaetae defective (Mcd) alleles of Notch exhibit a dominant phenotype that is characterized by reduction of the number of small mechanosensory bristles (microchaetae) present on the dorsal thorax of adult flies (Brennan and Gardner 2002; Heitzler 2010). N^Mcd alleles are also recessive lethal and form an allelic series in terms of their phenotypic strength (Ramain et al. 2001). Interestingly, all but one allele carries late truncation mutations in the intracellular domain distal to the ANK domain, deleting part of the TAD and PEST domains. Similar to late truncation variants in NOTCH2 and NOTCH3 that are linked to Hajdu-Cheney syndrome (OMIM #102500) (Isidor et al. 2011; Simpson et al. 2011) and lateral meningocele syndrome (OMIM #615293) (Gripp et al. 2015), respectively, these mutations are predicted to increase the stability of the NICD, leading to a GOF phenotype. As mentioned earlier, analogous truncation mutations in NOTCH1 are also frequently found in T-ALL (Weng et al. 2004) and B-CLL (Puente et al. 2011). Interestingly, one N^Mcd allele, named Mcd5 carries a p.C739Y missense mutation in one of the six cysteine residues that form the backbone of EGFr-18 (Ramain et al. 2001). Dominant phenotypes seen in N^Mcd5 and other N^Mcd truncation alleles were reported to be similar, suggesting that this missense allele somehow mimics the effect of late truncation alleles. How this may occur structurally and mechanistically has not been further investigated.

Sensory organ precursor cells (SOPs) that form mechanosensory bristles are specified from proneuronal clusters through lateral inhibition mediated by Notch, Delta, Su(H) and E(spl) (Charng et al. 2014a; Shaya and Sprinzak 2011). GOF of Notch signaling can cause a decrease in the number of SOPs by specifying more epithelial cells at the expense of SOPs during this process, which is why some bristles are missing in N^Ax mutants (Brennan et al. 1999). Surprisingly, however, the loss of bristle phenotype seen in N^Mcd alleles have been reported to be independent of Delta, Su(H), and E(spl), suggesting that N^Mcd alleles affect a non-canonical aspect of Notch function (Ramain et al. 2001; Alfred and Vaccari 2018). Phenotypes seen in N^Mcd alleles are dependent on deltex, an endocytic trafficking gene that regulates both ligand-dependent and -independent Notch signaling events (Xu and Artavanis-Tsakonas 1990; Fuwa et al. 2006). In addition, N^Mcd phenotypes depend on several proteins that are considered to function as core components of the canonical Wnt signaling pathway in Drosophila including dishevelled (dsh, a cytoplasmic phoshoprotein), shaggy (sgg, a cytoplasmic kinase GSK3β) and armadillo (arm, also known as β-Catenin) (Ramain et al. 2001). Although the precise mechanisms of how N^Mcd alleles cause the loss of bristle phenotype is unknown, yeast-two-hybrid experiments have revealed that Dsh can bind to the carboxy-terminal end of Notch, which is missing in all of the truncated N^Mcd alleles (Axelrod et al. 1996; Ramain et al. 2001). Therefore, N^Mcd alleles may be pointing us to a molecular interface where Notch signaling intersects with Wnt signaling. Cross-talk between Notch and Wnt signaling pathways have been reported in diverse developmental and pathophysiological contexts (Borggrefe et al. 2016; Collu et al. 2014). Indeed, Wingless (Wg), a homolog of mammalian WNT1 and the major ligand of the Wg/Wnt pathway in flies, has also been reported to interact with the extracellular domain of Notch to modulate its function (Wesley 1999; Couso and Martinez Arias 1994). By investigating how the domains mutated in N^Mcd alleles affect canonical as well as non-canonical aspects of these two signaling pathways in vivo, we will likely have a better understanding of Notch-related diseases that are caused by late truncating NOTCH alleles, such as Hajdu-Cheney syndrome, lateral meningocele syndrome, and certain forms of leukemia.

v). N^nd-3, N^spl-1 and N^ts1: Additional mutations in EGFr with unique characteristics

In addition to N^M1, N^jigsaw, N^Ax and N^Mcd5 that I have discussed so far, several additional alleles of Notch have been reported to carry missense mutations in EGFr. These alleles also exhibit unique features, providing us with a glimpse of how pleiotropic the extracellular domain of Notch actually is. A rare genetic phenomenon referred to as “intragenic complementation”, in which two different alleles of Notch complement each other’s phenotype, can be observed between certain allelic combinations, suggesting that these mutations impact different functions within the Notch receptor. Even though many alleles were isolated in the early days of genetics, very little is known about how these mutations alter the function and structure of Notch. Phenotypic description of many classical Notch alleles can be found in “The Red Book” (Lindsley and Zimm 1992) as well as in FlyBase (Thurmond et al. 2019).

notchoid (nd) alleles of Notch are a group of recessive visible mutations that exhibit wing notching. Of these, N^nd-3 has been reported to carry a missense mutation in one of the six cysteine residues that form the core disulfide bonds in EGFr-3 (p.C105F) (Lyman and Young 1993). Since N^nd-3 is hemizygous/homozygous viable and these animals only exhibit a mild defect in wing margin (wing notching) and wing vein (thickening of the ends of wing veins, referred to as “wing delta” which is also seen upon Delta haploinsufficiency) development, it is considered as a weak hypomorphic (mild LOF) allele. If this mutation only disrupts the local structure of this specific EGFr, it is possible that EGFr-3 is dispensable for most Notch signaling processes. Interestingly, the majority of NOTCH3 variants that are linked to CADASIL are alleles that decrease (5 cysteine/EGFr) or increase (7 cysteine/EGFr) the number of cysteine residues in EGFr, similar to N^nd-3 and N^Mcd5 mutations (Joutel et al. 1996). Odd numbers of cysteines per EGFr leads to an unpaired cysteine that may form abnormal disulfide bonds within NOTCH3 or with other proteins that it may interact with. Although some CADASIL alleles behave as LOF alleles when tested in cells or in mice (Arboleda-Velasquez et al. 2011), other variants do not have an obvious effect on canonical Notch signaling, suggesting that this disease may be caused a neomorphic mechanism. For example, a cysteine residue in NOTCH3 that failed to form intramolecular disulfide bonds may form covalent bonds with other extracellular/luminal molecules in its vicinity, thereby modulating the function of these unidentified target proteins. In support of this idea, one key diagnostic criteria of CADASIL in addition to genetic testing is the accumulation of granular osmophilic material (GOM), an electron dense structure that contains the extracellular domain of NOTCH3 and other proteins, in vascular tissues (Baudrimont et al. 1993; Ishiko et al. 2006). A similar but recessively inherited condition called CARASIL (Cerebral autosomal recessive arteriopathy with subcortical infarcts and leukoencephalopathy, OMIM #600142) is caused by mutations in HTRA1, a gene encoding a protease involved in TGF-β signaling (Hara et al. 2009). Thus, it would be interesting to explore whether NOTCH3 with CADASIL mutations may aberrantly interact with TGF-β signaling factors or other HTRA1 substrates in order to probe the molecular mechanism of these disorders. Another interesting but unexplained aspect of CADASIL alleles is that although NOTCH3 has 34 EGFr, most of the variants linked to this disease are found in EGFr-1~5 (corresponding to EGFr-1~6 in Drosophila Notch which has 36 EGFr) (Joutel et al. 1997; Masek and Andersson 2017). Understanding the effect of cysteine mutations of these early EGFr compared to other EGFr may provide us with some hints to why such a bias exists. Hence, functional studies of N^nd-3 (EGFr-3) as well as other N alleles that affect cysteine residues in EGFr such as N^Mcd5 (EGFr-18) may provide useful information that benefits patients who suffer from this devastating neurodegenerative disease.

split (spl) alleles of Notch are recessive visible mutations that primarily manifest in eyes and bristles (Lindsley and Zimm 1992). Homozygous females and hemizygous males have small rough eyes and their bristles are often split, doubled or sometimes missing (Lees and Waddington 1942). Interestingly, N^spl alleles do not exhibit any wing defects on their own, and show intragenic complementation with a number of other recessive visible N mutations including N^nd and N^Ax alleles (Shellenbarger and Mohler 1975; Brennan et al. 1997). N^spl-1 is caused by a missense mutation in EGFr-14 (p.I578T), but the exact genetic effect of this allele is controversial. Some studies refer to this allele as a hypomorph based on complementation with strong N LOF alleles (Brennan et al. 1997), whereas others document this as a hypermorph (Hing et al. 1999) or an antimorph (de Celis and Garcia-Bellido 1994) based on other genetic tests. The reason why this single mutation can be genetically classified as different allele types based on different experiments may be due to the complicated role of this residue and domain in different developmental contexts. Although the precise function of EGFr-14 and this specific amino acid in Notch is still unknown, the study of N^spl alleles was critical for Notch research because genetic screens to identify intergenic modifiers of N^spl led to the identification of the E(spl) complex, a group of genes encoding evolutionarily conserved bHLH transcriptional repressors that function downstream of the canonical Notch signaling pathway (Von Halle 1965; Lehmann et al. 1983). Without the pioneering study of N^spl and E(spl) mutations, the mammalian HES (Hairy and Enhancer-of-Split) (Akazawa et al. 1992; Sasai et al. 1992) and HEY (Hairy/Enhancer-of-split related with YRPW motif protein) (Leimeister et al. 1999) genes that are homologous to Drosophila E(spl) genes would have been named differently.

Some Notch alleles are known to be temperature sensitive (Shellenbarger and Mohler 1975; Lindsley and Zimm 1992). This could be due to the variant causing destabilization of the protein at restrictive (high) temperatures, or it could be due to other reasons such as those revealed through studies of genes involved in O-glycosylation of Notch (see above). N^ts1 (also known as N^l1N-ts1) carries a missense mutation in an O-fucosylation motif of EGFr-32 (p.G1272D) and behaves as a temperature sensitive hypomorphic allele (Xu et al. 1992). Independent mutations named N^ts2 (Shellenbarger and Mohler 1975) and N^ts3 (Li et al. 2009; Nicholson et al. 2011) have been reported to function in a similar manner but their molecular lesions have yet to be determined. Although the mechanisms by which these mutations inhibit Notch signaling at their restrictive temperatures are unknown, these alleles have been useful in assessing the function of Notch signaling in diverse developmental processes post-embryonically. By keeping the N^ts animals at a permissive temperature (18°C) during embryogenesis, and shifting the animals to a restrictive temperature (29°C) during specific time points of interest, one can perform “conditional knockout” experiments of Notch in a semi-reversible manner. Such experiments elucidated that Notch signaling is used not only once but reiteratively during the formation of variety of tissues such as the fly eye (Cagan and Ready 1989), leg (Bishop et al. 1999), wing (Micchelli and Blair 1999), and mechanosensory bristles (Hartenstein and Posakony 1990). Furthermore, these alleles have allowed researchers to explore the post-developmental roles of Notch signaling in adult flies, revealing its roles in tissue homeostasis through regulation of stem cells (Song et al. 2007; Ohlstein and Spradling 2006; Micchelli and Perrimon 2006; Chaturvedi et al. 2017), stereotypic behavior including learning and memory (Presente et al. 2004), and longevity (Presente et al. 2001). Our understanding of Notch signal regulation will likely advance if we can precisely understand how the N^ts1 missense mutation in EGFr-32 causes these temperature sensitive phenotypes at the molecular level. In addition, mapping of the N^ts2 and N^ts3 alleles as well as identifying other missense mutants in Notch that cause conditional phenotypes will help decipher how the Notch receptor works in different environmental settings in vivo.

5. Conclusions

In contrast to most signaling pathways that harbor the name of the major ligand in the pathway (e.g. Wnt, Hedgehog, BMP/TGF-β, EGF, FGF, IGF/Insulin, PDGF, VEGF…), Notch seems to be one of the few major developmental signaling pathways that bears the name of the receptor, implicating the central role of this protein in the entire signaling network (Guruharsha et al. 2012). Since the isolation of the first Notch allele in the early 20^th century, hundreds of Notch alleles with different molecular lesions and unique phenotypic and molecular characteristics have been reported. While amorphic alleles of Notch have been critical tools to assess which biological events depend on Notch signaling, various missense alleles have provided insights into how this pleiotropic and complicated pathway is fine-tuned in vivo. In this review, I primarily focused on alleles that affect the EGFr of the Notch receptor and discussed how genetic analyses have provided insights into the workings of this molecular pathway. Several missense mutations that are found in other domains of the Notch receptor are also worth mentioning. N⁴¹⁴ and N^l1N-B carry a p.D1577G and a p.G1573V change in the third LNR repeat, respectively (Brennan et al. 1997; Lyman and Young 1993). Similar to some of the missense mutations associated with T-ALL in NOTCH1 (Weng et al. 2004) and a variant linked to infantile myofibromatosis 2 in NOTCH3 (OMIM #615293) (Martignetti et al. 2013; Xu et al. 2015), these mutations behave as GOF alleles likely due to ligand-independent activation mediated by unfolding of the NRR. Some other interesting missense mutations are also found in the intracellular domain of Notch, including N^su42c that carries a p.A2060V change in the fifth ANK repeat (Diederich et al. 1994) and N¹⁰⁸¹ carries a p.G2490C change within the TAD (Brennan et al. 1997). These mutations behave as strong LOF alleles, providing functional information that may be useful to understand of how NICD binds to its interaction partners in the nucleus. Finally, in addition to the N^Mcd truncation alleles discussed earlier in this article, there are a number of nonsense and frameshift alleles of N that exhibit unique phenotypes. By comparing of phenotypes seen in mutants such as N^Co (nonsense allele within the RAM domain, deleting most of the intracellular domain) (Gottschewski 1935; Lyman and Young 1993), N^60g11 (frameshift allele immediately after the ANK repeats) (Welshons et al. 1963; Brennan et al. 1997), N^l1N-3 (frameshift allele deleting PEST and part of the TAD including the OPA domain identical to N^Mcd3) (Welshons 1965; Brennan et al. 1997), and N^nd-2 (introduces a frameshift 14 residues before the carboxy-terminal end of the protein) (Welshons 1965; Xu et al. 1990), we can begin to understand how different domains within the NICD function in both canonical and non-canonical modes of Notch signaling. Since the intracellular domain of type-I transmembrane receptors are often required for proper intracellular trafficking (Yamamoto et al. 2010), it would be interesting to assess what aspect, if any, of the phenotypes seen in these intracellular domain mutations can be attributed to defects in exocytosis, endocytosis or endolysosomal trafficking (Schnute et al. 2018). Considering such knowledge cannot be obtained by studying amorphic alleles of Notch or through gene knockdown studies using RNAi, studies of unique missense, nonsense and frameshift mutation will continue to provide novel insights into this pathway.

Although advancement of genome editing technologies such as the CRISPR/Cas9 system are allowing scientists to introduce virtually any mutation into any gene in any species (Knott and Doudna 2018), forward genetic screens using chemical mutagens such as EMS will still be one of the most powerful ways to reveal specific residues or domains in Notch and other proteins that have unique functions in vivo (Yamamoto et al. 2012). Although the N gene has been extensively studied by numerous laboratories in the past, interesting mutations in this gene have not yet reached saturation. In addition, assessments of disease associated missense variants/mutations in human NOTCH genes may provide novel structural and functional information on how Notch receptors are regulated. If human NOTCH receptors can be shown to function in vivo in Drosophila, “humanized” Notch flies may allow functional studies of disease-linked variants/mutations in the context of the human protein in an in vivo environment (Bellen and Yamamoto 2015; Harnish et al. 2019).

Lastly, there are a number of non-canonical Notch signaling events that need further mechanistic studies. For example, Drosophila Notch has been shown to be involved in axon guidance through Abl (Abelson murine leukemia viral oncogene homolog) tyrosine kinase in a manner that does not depend on Su(H) (Giniger 1998; Kannan et al. 2018). In another example in mammalian cells, NICD was found in the mitochondria where it prevented apoptosis in an RBPJ independent fashion (Perumalsamy et al. 2010). How these and other non-canonical functions of Notch are mechanistically achieved and whether such functions are evolutionarily conserved are still open questions that need to be investigated (Alfred and Vaccari 2018). By screening for Notch alleles that exhibit unique phenotypes, uncovering the molecular lesions of such mutations, and further understanding the functional impact of these changes on diverse context of Notch biology, researchers will likely continue to discover unanticipated and exciting aspects of the Notch signaling pathway. The fact that Drosophila is the only genetic model organism that possesses a single gene that encodes a Notch receptor (even C. elegans has two) affords fly geneticists a great advantage. By extrapolating the findings obtained in Drosophila into mammalian systems and combining this knowledge with precise structural information and large-scale human genomics data sets, we will likely identify new biological knowledge that will become useful in understanding and treating genetic disorders or cancers that are linked to defective Notch signaling.