Multifaceted regulation of Notch signaling by glycosylation

Abstract

To build a complex body composed of various cell types and tissues and to maintain tissue homeostasis in the postembryonic period, animals use a small number of highly conserved intercellular communication pathways. Among these is the Notch signaling pathway, which is mediated via the interaction of transmembrane Notch receptors and ligands usually expressed by neighboring cells. Maintaining optimal Notch pathway activity is essential for normal development, as evidenced by various human diseases caused by decreased and increased Notch signaling. It is therefore not surprising that multiple mechanisms are used to control the activation of this pathway in time and space. Over the last 20 years, protein glycosylation has been recognized as a major regulatory mechanism for Notch signaling. In this review, we will provide a summary of the various types of glycan that have been shown to modulate Notch signaling. Building on recent advances in the biochemistry, structural biology, cell biology and genetics of Notch receptors and the glycosyltransferases that modify them, we will provide a detailed discussion on how various steps during Notch activation are regulated by glycans. Our hope is that the current review article will stimulate additional research in the field of Notch glycobiology and will potentially be of benefit to investigators examining the contribution of glycosylation to other developmental processes.

Introduction

Protein glycosylation is the addition of carbohydrate molecules to proteins, which can occur co-translationally and posttranslationally and contributes to structural and functional characteristics of the glycosylated proteins. It involves the addition of different forms of glycan to the specific sites onto the protein and occurs in the endoplasmic reticulum (ER) and the Golgi apparatus for secreted and transmembrane proteins. The two major classes of glycosylation are N- and O-linked glycosylation. In N-linked glycosylation, the glycan is attached to an asparagine (N) residue mostly present in an NX(S/T) sequence, where X can be any amino acid but proline (Stanley et al. 2015). In O-linked glycosylation, the glycan is attached to a serine or threonine residue in specific consensus sequences. Mutations in components of both the N- and O-linked glycosylation pathways play causative roles in congenital disorders of glycosylation, a heterogeneous group of diseases whose list is rapidly growing (Ng and Freeze 2018; Peanne et al. 2018; van Tol et al. 2019). Glycosylation regulates various cell signaling pathways during animal development. One of the most conserved cell signaling pathways which plays diverse roles during animal development and tissue homeostasis is the Notch signaling pathway (Artavanis-Tsakonas and Muskavitch 2010; Kovall et al. 2017). Given the indispensable role of Notch signaling in embryonic development and adult maintenance and the association of aberrant Notch signaling with various human diseases (Masek and Andersson 2017; Siebel and Lendahl 2017), understanding the mechanisms involved in the regulation of Notch signaling is of great interest.

The foundation of Notch Glycobiology was laid in year 2000, when a key component of the Notch pathway in Drosophila called fringe (fng; (Irvine and Wieschaus 1994)) and its mammalian homologs were shown to be glycosyltransferase enzymes which add N-acetylglucosamine (GlcNAc) to O-fucose on Notch EGF repeats to modulate Notch–ligand interactions (Bruckner et al. 2000; Moloney, Panin, et al. 2000). During the last two decades, glycans (mostly O-glycans) have been shown to influence multiple aspects of the Notch signaling pathway (Haltom and Jafar-Nejad 2015; Harvey and Haltiwanger 2018; Varshney and Stanley 2018). Moreover, mutations in a number of glycosyltransferases involved in Notch pathway regulation have been identified in several human diseases with distinct phenotypes (Sparrow et al. 2006; Li et al. 2013; Shaheen et al. 2013; Basmanav et al. 2014; Servian-Morilla et al. 2016; Takeuchi, Wong, et al. 2018). Here, we will review the roles of different forms of O-glycosylation in the regulation of various steps of Notch pathway activation, including receptor folding, trafficking and ligand interactions. We will cover recent progress in the field on specific aspects of Notch signaling such as dosage-sensitive effects of O-glycosyltransferases, non-cell autonomous regulation of signaling and the in vivo contribution of O-glycans on specific sites to cis- and trans-interactions between Notch and ligands. We will also highlight the contribution of other forms of glycosylation such as N-glycosylation to Notch signaling.

Sugar decorations of Notch pathway components

Notch signaling is a well-known, highly conserved cell-to-cell communication pathway in multicellular organisms. The Notch receptors are large type I transmembrane proteins that following synthesis in the ER, are cleaved by the Furin convertase (S1 cleavage) in the Golgi apparatus. There are four Notch paralogs (NOTCH1–4) in mammals, while C. elegans has two (LIN12 and GLP1) and Drosophila has only one Notch receptor. The extracellular domain (ECD) of Notch receptors consists of a negative regulatory region and 29–36 tandem epidermal growth factor-like (EGF) repeats. These EGF repeats harbor the sites for addition of both N-linked glycans and O-linked glycans (Harvey and Haltiwanger 2018; Varshney and Stanley 2018). The intracellular domain of Notch receptors (NICD) is a transcriptional co-activator that mediates the transcriptional changes downstream of Notch activation.

In order to initiate signaling, NECD interacts with the Delta/Serrate/LAG-2 (DSL) ligands from neighboring cells (Kovall et al. 2017). Similar to Notch receptors, canonical Notch ligands are also type I transmembrane proteins, with tandem EGF repeats in their ECDs which facilitate their interaction with Notch receptors (Bray 2006). In Drosophila, there are two canonical ligands: Delta and Serrate. C. elegans has four ligands LAG-2, APX-1, ARG-1 and DSL-1, while in mammals there are three Delta homologs (Delta-like 1 or DLL1, DLL3 and DLL4) and two Serrate homologs (jagged-1 or JAG1 and JAG2) (Bray 2006). Upon binding of a ligand to Notch, endocytosis of the ligand by the signal-sending (ligand-expressing) cells creates a pulling force, which leads to a series of proteolytic cleavages which ultimately release the NICD from the membrane (Kovall et al. 2017). Cleaved NICD is translocated to the nucleus and binds the Notch pathway effector protein RBPJ (recombination signal binding protein for immunoglobulin kappa J region) to induce the expression of the Notch target genes (Figure 1A).

Fig. 1.

(A) Cartoon representation of canonical Notch signaling. In the ER, the newly synthesized Notch protein is glycosylated. In the Golgi, the S1 cleavage occurs which generates a heterodimeric form of Notch containing extracellular (NECD, gray) and intracellular (NICD, red) domains. Further elongation of the glycans takes place in the Golgi. Upon reaching the cell surface in the signal-receiving cell, NECD binds with ligand from the signal-sending cell. Endocytosis of ligand into the signal-sending cell exerts a pulling force on Notch, leading to the dissociation of the Notch heterodimer and the S2 cleavage of Notch by ADAM proteins. The γ-secretase complex then mediates the S3/S4 cleavages, upon which the NICD is released from the membrane, translocates into the nucleus and forms the Notch transcriptional activator complex together with RBPJ/CSL, Mastermind-like (MAML1) and other co-activators. (B) Diagram showing a generalized version of a full Notch protein having an NECD with 36 EGF repeats (gray boxes) and a negative regulatory region comprised of three LNR motifs and a heterodimerization domain (HD). The extended LBD is marked by the horizontal black line under EGF8–12. NICD contains an RBPJ association module (RAM), seven Ankyrin repeats (ANK) and a PEST domain. Magnified view of an EGF repeat showing the location of different types of O-glycan and the structure of each O-glycan along with responsible enzymes is depicted in the diagram. B4GALT1, β-1,4-galactosyltransferase 1; ST3GAL, ST3 β-galactoside α-2,3-sialyltransferase. (C) Schematic showing the EGF repeats in the ECDs of Drosophila Notch, its human homologs and their ligands. The ECDs of the fly Notch, human NOTCH1 and human NOTCH2 have 36 EGF repeats, while human NOTCH3 and NOTCH4 have 34 and 29 EGF repeats, respectively. Blue and red boxes indicate EGF repeats with sites for O-glucosylation and O-fucosylation, respectively. The black dashes under EGF repeats indicate the O-GlcNAcylation sites. Orange stars indicate those O-glucosylated-EGF repeats whose xylose-extension by Gxylt/Xxylt has been experimentally verified. Blue squares indicate O-fucosylated EGF repeats with GlcNAc-elongation by fringe proteins. DLL, Delta-like; DSL, Delta/Serrate/LAG-2; JAG, Jagged; TM, transmembrane.

The three major types of O-linked glycan found on EGF repeats of Notch pathway components are O-fucose, O-glucose and O-GlcNAc, which can be as monosaccharides or extended forms (Figure 1B). In the following subheadings, we will introduce the enzymes responsible for the addition of these three types of O-glycosylation to EGF repeats of Notch receptors and their ligands.

O-fucosylation of EGF repeats

In O-fucosylation, an O-fucose residue is added on the consensus site C²X_4–5(S/T)C³ (where S/T is serine or threonine, and C² and C³ are the second and third cysteines of the EGF repeat; X can be any amino acid) by a glycosyltransferase called “protein O-fucosyltransferase 1” (POFUT1 in mammals and O-fut1 in Drosophila) (Wang et al. 1996; Wang and Spellman 1998; Wang et al. 2001). POFUT1 is an ER-localized enzyme (Luo and Haltiwanger 2005). Over a hundred mammalian proteins harbor EGF repeats with a consensus O-fucosylation site, with Notch receptors harboring the largest number of sites (Figure 1C). Mass spectrometric analysis of Drosophila Notch and mouse NOTCH1 has shown that the majority of the O-fut1/POFUT1 target sites in these proteins are efficiently modified (Harvey et al. 2016; Kakuda and Haltiwanger 2017). A recent study by the Rini group has determined the X-ray crystal structure of mouse POFUT1 in complex with different EGF domains including Notch EGF12 and EGF26 and defined its EGF-domain binding properties (Li, Han, et al. 2017). Another sugar residue, GlcNAc, is added in a β1,3-linkage to O-fucose on Notch EGF repeats by the Golgi-localized Fringe family of glycosyltransferases (Bruckner et al. 2000; Moloney, Panin, et al. 2000). Drosophila has one fringe protein while mammals have three homologs: manic (MFNG), radical (RFNG) and lunatic fringe (LFNG) (Irvine and Wieschaus 1994; Cohen et al. 1997; Johnston et al. 1997). Two other enzymes, β-1,4-galactosyltransferase 1 (B4GALT1) and ST3 β-galactoside α-2,3-sialyltransferase (ST3GAL), are reported to extend the GlcNAc-fucose-O disaccharide on EGF repeats of the mammalian proteins to tri- and tetrasaccharides by the addition of a galactose and a sialic acid (Neu5Ac) residue successively (Harvey and Haltiwanger 2018). However, in flies, extension of the GlcNAc-fucose-O disaccharide to the tri- or tetrasaccharide has not been reported (Xu et al. 2007; Harvey and Haltiwanger 2018).

O-glucosylation of EGF repeats

In O-glucosylation, O-glucose residue is added to serine residues within the consensus sequence C¹-X-S-X-(P/A)-C² (where the modified serine is underlined, X represents any amino acid, and C¹ and C² are the first and second cysteine residues of the EGF repeat) by “protein O-glucosyltransferase 1” (POGLUT1 in mammals, Rumi in Drosophila (Acar et al. 2008; Takeuchi et al. 2011)). This gene was originally discovered by Wang and colleagues from human hematopoietic stem/progenitor cells isolated from a patient with myelodysplastic syndrome who had undergone leukemic transformation and was named hCLP46 (human CAP10-like protein 46 kDa), in reference to the similarity of its major recognizable domain with the CAP10 protein of Cryptococcus neoformans (Teng et al. 2006). Upon later discovery of the enzymatic activity of Drosophila Rumi and its mammalian homologs (Acar et al. 2008; Fernandez-Valdivia et al. 2011; Takeuchi et al. 2011), hCLP46 and its mouse homolog (formerly called KTELC1) were renamed to POGLUT1. It is worth mentioning that in addition to its O-glucosyltransferase activity, POGLUT1 can also add an O-xylose residue to the same consensus motif (Takeuchi et al. 2011). However, the O-xylosyltransferase function of POGLUT1 is less efficient than its O-glucosyltransferase activity and the biological significance of O-xylosylation in Notch signaling is not known (Li, Fischer, et al. 2017; Takeuchi et al. 2011). Therefore, our discussion of POGLUT1 for the remaining of this work will focus on its O-glucosyltransferase activity.

Structural studies on human POGLUT1 bound to an EGF repeat have shown that the conformation of the O-glucosylation motif is the reason why POGLUT1 cannot add O-glucose to threonine (unlike POFUT1) (Li, Fischer, et al. 2017). Similar to POFUT1, POGLUT1 is a soluble enzyme localized to the ER and only modifies properly folded EGF repeats (Takeuchi et al. 2012). POGLUT1 has two homologs in mammals (KDELC1 and KDELC2). It was previously shown that human KDELC1 and KDELC2 are not able to add O-glucose to EGF repeats harboring the POGLUT1 consensus sequence (Takeuchi et al. 2011). A recent study has shown that these proteins can add O-glucose onto a specific serine residue in a motif between the 3rd and 4th cysteine residues in EGF repeats, which is distinct from the POGLUT1 target site (Takeuchi, Schneider, et al. 2018). Similar to POGLUT1 and POFUT1, these two enzymes, which were renamed to POGLUT2 and POGLUT3, only modify properly folded EGF repeats (Takeuchi, Schneider, et al. 2018). POGLUT1 and POFUT1 have many target sites on all Notch proteins and a number of other proteins, including Notch ligands (Figure 1C) (Fernandez-Valdivia et al. 2011; Haltom et al. 2014). However, POGLUT2/3 have so far been shown to only glucosylate one EGF repeat in human NOTCH1 and NOTCH 3 (EGF11 and EGF10, respectively) and none in human NOTCH2 (Takeuchi, Schneider, et al. 2018).

The O-glucose monosaccharides added by POGLUT1 to EGF repeats can be further extended to xylose-glucose-O disaccharides by glucoside xylosyltransferase enzymes (GXYLT) (Whitworth et al. 2010). Mammals have two GXYLT enzymes (GXYLT1 and 2), while Drosophila has a single GXYLT called Shams (Sethi et al. 2010; Lee et al. 2013). The disaccharide can be further elongated to xylose-xylose-glucose-O trisaccharide by xyloside xylosyltransferase enzymes (XXYLT1 in mammals, Xxylt in Drosophila (Sethi et al. 2012; Pandey et al. 2018)). In mouse NOTCH1, all predicted POGLUT1 target sites are O-glucosylated, and most if not all of them can be efficiently extended to a trisaccharide form, although some EGF repeats show cell type-specific variations in the degree of O-glucosylation and xylosylation (Rana et al. 2011). In contrast, only a subset of O-glucosylated EGF repeats in Drosophila Notch is xylosylated (Lee et al. 2013; Harvey et al. 2016).

O-GlcNAcylation of EGF repeats

The O-GlcNAc modification was first discovered on Drosophila Notch EGF20 by the Okajima group (Matsuura et al. 2008). O-GlcNAc is added to a serine or threonine residue within the putative consensus sequence C⁵-X-X-G-X-(T/S)-G-X-X-C⁶ by “EGF domain-specific O-linked GlcNAc transferase” (EOGT), although more recent mass spectral analyses on Drosophila Notch and mouse NOTCH1 have shown that the amino acid preceding the O-GlcNAcylated S/T (the underlined X in the consensus sequence) is usually an aromatic amino acid (Ogawa et al. 2018, 2020). EOGT is localized to the ER and similar to POFUT1 and POGLUT1, it only modifies properly folded EGF repeats (Sakaidani et al. 2011). Despite the presence of O-GlcNAc modification consensus sites on 18 EGF repeats of the Drosophila Notch, mass spectrometric analysis on purified Notch from Drosophila Schneider 2 (S2) cells and Drosophila embryos revealed robust O-GlcNAc modification only on five sites (Harvey et al. 2016). Mouse NOTCH1 possesses 17 EGF repeats with O-GlcNAc consensus sites, the majority of which are modified by O-GlcNAc (Ogawa and Okajima 2019). O-GlcNAc can be further extended to a Gal-GlcNAc-O disaccharide and a sialic acid (Neu5Ac)-Gal-GlcNAc-O trisaccharide in mammals (Sakaidani et al. 2012; Ogawa et al. 2018), although this elongation appears to occur only on a subset of O-GlcNAcylated EGF repeats of the mouse NOTCH1 (Ogawa and Okajima 2019).

As mentioned above, O-glycans are added to EGF repeats of Notch and other proteins in the ER and can be extended in the Golgi, as the target proteins move through the secretory pathway. Moreover, no enzyme has yet been discovered that can remove any of these O-glycans once they are added to an EGF repeat. Accordingly, the ECDs of Notch proteins harbor O-glycans for most of their lifetime in the cell. In keeping with this, O-glycosylation of Notch has been reported to affect Notch signaling at various steps, including folding, trafficking of the receptor to the cell surface, its stability and interaction with DSL ligands and its cleavage. In the following sections, we will discuss the contribution of O-glycans to each of the above-mentioned steps.

Regulation of Notch trafficking and surface expression by O-glycosylation

O-Fucosylation in Notch Folding and Trafficking

O-fucosylation of Notch occurs in the ER (Luo and Haltiwanger 2005) and is required for Notch signaling in Drosophila and mammals (Okajima and Irvine 2002; Sasamura et al. 2003; Shi and Stanley 2003). Although both Notch receptors and ligands are modified by O-fucose glycans, studies in Drosophila and mouse models have revealed that Notch signaling is primarily regulated by O-fucosylation of Notch receptors, although O-fucosylation of Notch ligands might also have functional consequences in some contexts (Okajima and Irvine 2002; Panin et al. 2002; Stahl et al. 2008; Muller et al. 2014; Serth et al. 2015). Studies in mammalian cell lines and mice have provided evidence for the importance of the enzymatic activity of POFUT1 in Notch signaling (Zhou et al. 2008). Cell culture studies have shown mixed results on the effects of fucosylation and POFUT1 on the cell surface expression of mammalian Notch receptors. Fucosylation-deficient Chinese hamster ovary cells and Pofut1^−/− murine embryonic stem cells showed normal surface levels of endogenous NOTCH1–4 receptors (Stahl et al. 2008). Moreover, Pofut1^−/− hematopoietic progenitors showed a mild decrease in surface expression of NOTCH1 and NOTCH2 which mirrored a similar decrease in the total levels of these two receptors in these cells and showed no decrease in NOTCH3 cell-surface levels (Yao et al. 2011). However, in POFUT1^−/− HEK293T cells, around 50% of the endogenous NOTCH1 was trapped in the ER (Takeuchi et al. 2017). This is in agreement with the partial ER entrapment of endogenous NOTCH1 in presomitic mesoderm of Pofut1^−/− mouse embryos (Okamura and Saga 2008). In contrast, recent in vivo reports showed reduced Notch signaling in Pofut1 mutant endocardial cells in mice, while surface levels of NOTCH1 were not altered (Wang et al. 2017). Together, these data suggest that although loss of POFUT1 affects the surface expression of mammalian NOTCH1 in some contexts, the effects of POFUT1 loss on mammalian Notch signaling cannot be primarily explained by reduced surface expression of Notch receptors. It is worth mentioning that localization of Notch receptors to specific membrane subdomains can affect their signaling properties. For example, recruitment of NOTCH3 to lipid rafts at the surface of thymocytes prevents the ubiquitination and proteasomal degradation of the NOTCH3-ICD, which will lead to enhanced NOTCH3 signaling and will help sustain NOTCH3’s oncogenic role in T-cell leukemia (Checquolo et al. 2010). Therefore, although detecting Notch at the cell surface in a given mutant background indicates that the receptor is not trapped in the ER, it does not guarantee that Notch has reached the proper subdomain in the membrane required for its optimal signaling.

Studies in Drosophila have suggested nonenzymatic roles for O-fut1 as a chaperone in Notch protein folding and its ER exit (Okajima et al. 2005) and also in endocytic trafficking of Notch (Sasamura et al. 2007). We recently reported that point mutations in individual O-fucosylation sites of the Drosophila Notch (EGF8, EGF9 and EGF12) and their pairwise combinations can strongly affect Notch signaling without reducing the cell surface expression of Notch (Pandey et al. 2019). In the context of embryonic neurogenesis, which is a Delta-mediated process, loss of fucose from EGF12 caused a phenotype very similar to the Notch-null phenotype in this context (Pandey et al. 2019). Moreover, analysis of mice with a knock-in allele ablating the EGF12 O-fucosylation site resulted in embryonic lethality on a C57BL/6 genetic background (Varshney et al. 2019). Even if these point mutations affect the binding of O-fut1/POFUT1 to the mutated EGF repeats, they should not affect the overall recognition of the fly Notch and mouse NOTCH1 by these enzymes, given that more than 20 intact O-fucosylation sites remain in these alleles. These reports indicate that loss of even a single O-fucose residue in Notch proteins can dramatically affect fly and mammalian Notch signaling despite the presence of wild-type copies of O-fut1/Pofut1 in these animals, in agreement with the notion that the critical, evolutionarily conserved role of Ofut1/POFUT1 is its enzymatic activity.

O-glucosylation in Notch folding and trafficking

O-glucosylation is important for Notch signaling (Acar et al. 2008; Fernandez-Valdivia et al. 2011; Leonardi et al. 2011; Perdigoto et al. 2011). Loss of the Drosophila Poglut1 (encoded by rumi) and mutations in the O-glucosylation sites of the Drosophila Notch result in a temperature-sensitive loss of Notch signaling in all contexts studied (Acar et al. 2008; Leonardi et al. 2011; Perdigoto et al. 2011). However, analysis of the surface expression of Notch in imaginal disks harboring rumi mutant clones or expressing Notch with mutations in its O-glucosylation sites did not show a reduction in the cell surface level of Notch (Acar et al. 2008; Leonardi et al. 2011). In fact, rumi^−/− clones showed an accumulation of Notch both intracellularly and at the cell surface at high temperature (Acar et al. 2008), indicating that the loss of Drosophila Notch signaling caused by lack of O-glucosylation cannot be due to impaired exocytic trafficking of Notch. Moreover, RNAi-mediated knockdown (KD) of rumi in Drosophila S2 cells did not reduce the binding of Notch to Delta, suggesting that a defect in Notch–ligand binding cannot explain the rumi loss-of-function phenotype (Acer et al. 2008). Notably, analysis of the endogenous Notch cleavage products by immunoblotting showed that the Notch cleavage pattern in rumi-KD S2 cells is identical to the pattern observed in S2 cells with KD for Kuzbanian (Acar et al. 2008), which encodes the Drosophila ADAM10 (a disintegrin and metallopeptidase domain 10) protein responsible for the S2 cleavage of Notch (Lieber et al. 2002). Moreover, activation of Notch signaling mediated by overexpression of a mutant version of Notch which is ligand-independent but depends on the Kuzbanian-mediated S2 cleavage was fully suppressed by simultaneous deletion of rumi (Leonardi et al. 2011). Collectivity, these studies suggest that O-glucose on Drosophila Notch is not required for its surface expression and ligand-binding, but is essential for its S2 cleavage. While the precise molecular mechanism for the regulation of the S2 cleavage by O-glucosylation remains to be elucidated, the data strongly suggest that O-glucose residues maintain the Notch ECD in a conformation that prevents the S2 cleavage of Notch in the absence of ligands but helps expose the S2 cleavage site upon ligand-binding (Jafar-Nejad et al. 2010; Leonardi et al. 2011).

Analysis of Poglut1^−/− mouse embryos prior to their demise by embryonic day 9.5 shows phenotypes compatible with loss of NOTCH1 signaling and a severe reduction in the level of cleaved NOTCH1 (Fernandez-Valdivia et al. 2011; Ramkumar et al. 2015) (we note that Poglut1^−/− embryos also show a Crb2 loss-of-function phenotype due to impaired CRB2 trafficking (Ramkumar et al. 2015)). Moreover, homozygous and compound heterozygous mutations in human POGLUT1 have recently been identified in patients with a new form of limb-girdle muscular dystrophy, who show reduced NOTCH1 cleavage and signaling in muscle biopsies (Servian-Morilla et al. 2016, 2020). These observations indicate that similar to Drosophila Notch, mammalian NOTCH1 signaling requires the function of POGLUT1. shRNA-mediated KD of Poglut1 in the C2C12 mouse myoblast cell line results in a reduction in the expression of Notch target genes and a severe decrease in the level of cleaved (activated) NOTCH1 (Fernandez-Valdivia et al. 2011; Pelisse et al. 2018). However, Poglut1-KD C2C12 cells showed normal binding to JAG1 and DLL1 and only a modest decrease in the surface levels of NOTCH1 (Fernandez-Valdivia et al. 2011). POGLUT1 KD in several human cell lines also resulted in reduced NOTCH1 activation (Ma et al. 2011; Chu et al. 2013), without reducing its ligand-binding capacity (Ma et al. 2011). However, POGLUT1 KD did not reduce the ligand-independent activation of NOTCH1 induced upon EDTA incubation in U937 human leukemia cell line (Ma et al. 2011). Finally, POGLUT1 overexpression increased NOTCH1 activation and signaling in 293 T cells (Chu et al. 2013). Together, these observations indicate that similar to Drosophila Notch, O-glucosylation regulates mammalian NOTCH1 signaling, likely downstream of ligand-binding and upstream of or at the level of S2 cleavage. It is also worth mentioning that overexpression of an enzyme-dead version of human POGLUT1 fails to rescue the Poglut1-KD phenotypes in Neuro2A neuroblastoma mouse cell line, indicating that similar to flies, the effects of POGLUT1 on Notch signaling can be explained by its enzymatic activity (Acar et al. 2008; Fernandez-Valdivia et al. 2011; Leonardi et al. 2011).

Redundant and additive functions of O-glycans in Notch folding and trafficking

Although the above-mentioned studies indicate a clear role for O-glucosylation downstream of ligand-binding, a combination of in vitro, cell culture and Drosophila studies from Haltiwanger and Matsuno laboratories draws a more nuanced picture of the contribution of these glycans to Notch signaling. As mentioned before, all enzymes that add an O-linked glycan to Notch receptors can only glycosylate properly folded EGF repeats (Wang and Spellman 1998; Sakaidani et al. 2011; Takeuchi et al. 2012; Takeuchi, Schneider, et al. 2018), suggesting that they are not required for the initial folding of EGF repeats. However, unfolding assays on a single EGF repeat harboring O-fucose and O-glucose consensus sites showed that these two glycans help stabilize the folded EGF repeat in an additive manner (Takeuchi et al. 2017). Similarly, presence of O-GlcNAc on EGF20 of the Drosophila Notch was shown to play a stabilizing role for the folded EGF repeat (Ogawa et al. 2020). In HEK293T cells, loss of POGLUT1 or POFUT1 each reduced the ER exit and cell surface level of endogenous NOTCH1 by ~ 50%, and loss of both further reduced the NOTCH1 surface levels to less than 10%, again suggesting an additive effect on NOTCH1 folding and ER exit for the loss of these two glycans (Takeuchi et al. 2017). Moreover, although loss of EOGT by itself did not reduce the surface expression of NOTCH1 (Sawaguchi et al. 2017; Ogawa et al. 2020), when a number of O-fucosylation and O-glucosylation sites were mutated in mouse NOTCH1, O-GlcNAcylation of NOTCH1 by EOGT became essential for NOTCH1 surface expression (Ogawa et al. 2020). These observations indicate that similar to O-glucose and O-fucose, O-GlcNAc residues on NOTCH1 also contribute to its proper folding and trafficking, albeit in a fully redundant fashion.

As mentioned before, recent work has identified two novel enzymes which can add O-glucose to EGF repeats in a location distinct from the POGLUT1 target site (Takeuchi, Schneider, et al. 2018). Transgenic expression of human POGLUT1 in flies fully rescued the Notch-dependent phenotypes caused by loss of rumi, but POGLUT2 and POGLUT3 did not show any rescue in this in vivo assay, despite comparable expression levels (Takeuchi et al. 2011). Moreover, loss of their only identified modification site on human NOTCH1 (EGF11) did not affect NOTCH1’s surface expression, DLL1 binding and DLL1-mediated activation by itself. However, when this mutation was combined with the elimination of the O-fucosylation site in EGF8, the NOTCH1 surface level and DLL1-mediated signaling was reduced (Takeuchi, Schneider, et al. 2018). These observations suggest that although POGLUT2/POGLUT3 do not play a nonredundant role in mammalian NOTCH1 signaling, the O-glucose residue added by these enzymes to NOTCH1 might contribute to NOTCH1 folding in cooperation with other O-linked glycans.

In vivo studies in Drosophila support the notion that O-linked glycans on Notch function together to promote its ER exit and cell-surface trafficking. Deletion of rumi (Poglut1), a missense mutation that abolishes its enzymatic activity without affecting its expression, and Notch genomic transgenes with mutations in Rumi target sites all affect Notch signaling in a temperature-sensitive manner (Acar et al. 2008; Leonardi et al. 2011). At low temperature (18°C), these animals are viable and only show a mild decrease in Notch signaling in some contexts. As the temperature increases, the level of Notch signaling gradually declines in these mutants, with a partial lethality and intermediate phenotypes at 25°C and a full lethality and severe loss of Notch signaling at 28–30°C. Notably, even at high temperatures the Notch protein lacking O-glucose reaches the cell surface (Acar et al. 2008; Leonardi et al. 2011). However, when a null allele of rumi is combined with a knock-in allele of O-fut1 with negligible in vivo glycosyltransferase activity, the double-mutant animals show a severe loss of Notch signaling combined with ER entrapment of Notch even at 18°C (Ishio et al. 2015; Matsumoto et al. 2016). Altogether, these reports uncover an evolutionarily conserved role for O-fucose and O-glucose residues added to Notch proteins (and potentially other EGF repeat-containing proteins) to ensure that Notch molecules can pass the ER folding quality control system and traffic to the cell surface. The fact that the effects of these glycans on Notch trafficking range from a fully redundant role in some contexts to a nonredundant, additive role in others suggests that factors like the level of receptor expression and the capacity of each cell type’s ER quality control machinery ultimately determine the minimum number of O-glycans required for a given Notch protein to be able to exit the ER and reach the cell surface in each cell type.

Regulation of Notch–ligand interactions by O-glycosylation

Upon reaching the cell surface, the ECDs of Notch receptors from the signal-receiving cell interact with the trans-ligand (ligands from the neighboring signal-sending cell). This interaction is followed by the endocytosis of the trans-ligand by the signal-sending cell, which exerts a pulling force on Notch and leads to the exposure of the S2 cleavage site (Gordon et al. 2015). Upon S2 cleavage by ADAM proteins and the subsequent cleavages by the γ-secretase complex, the NICD is released into the cytoplasm, translocates to the nucleus and functions as a coactivator of the RBPJ protein to induce the transcription of the Notch target genes (Artavanis-Tsakonas and Muskavitch 2010). This process is called trans-activation of Notch by ligands (Figure 2A). In contrast, binding of the Notch receptor with ligands expressed by the same cell prevents its activation in most cases (de Celis and Bray 1997; Klein et al. 1997; Micchelli et al. 1997; Jacobsen et al. 1998; Glittenberg et al. 2006), and this process is referred to as cis-inhibition of receptor by ligand (Figure 2B). The relative levels of the Notch receptor(s) and ligands expressed in a given cell and also in its neighboring cells, combined with the strength of the Notch–ligand binding, are thought to affect the balance between trans-activatory and cis-inhibitory interactions among Notch receptors and ligands and thereby determine whether a cell behaves as the signal-sending or signal-receiving cell (Sprinzak et al. 2010; LeBon et al. 2014). It is worth noting that a recent report has documented cis-activation of Notch by ligands in several mammalian cell lines (Nandagopal et al. 2019). Figure 2 represents a schematic diagram showing possible interactions between receptor and ligand and their effect on Notch signaling. In this section, we will discuss how glycosylation regulates Notch signaling by modulating both trans- and cis-interactions of ligands with Notch receptors.

Fig. 2.

Schematics showing possible interactions between Notch receptor and ligands and signaling consequence. (A) Interaction between ligand from signal-sending cell (trans-ligand) and Notch from signal-receiving cell leads to activation of Notch signaling, hence called trans-activation. (B, C) Interaction of Notch with ligands expressed on the surface of the same cell (cis-ligands) affect signaling base on their relative expression levels. When the levels of cis-ligands are relatively higher than Notch, they prevent the activation of Notch by trans-ligands (ligands from neighboring cell) and result in cis-inhibition of the Notch receptor by ligand in the same cell (B). When the level of Notch is relatively higher than cis-ligands, Notch prevents the cis-ligands from activating Notch receptor expressed by neighboring cells, thereby causing cis-inhibition of ligand by receptor (C). (D) Notch receptor and ligand from the same cell can interact with one another in the endosomes in a configuration similar to trans-interaction and lead to cis-activation.

The role for O-fucose glycans in Notch–ligand interactions

O-fucosylation of Notch is essential for signaling. Loss of O-fucosylation in O-fut1/Pofut1 mutants results in embryonic lethality with typical Notch loss-of-function phenotypes in both flies and mice (Okajima and Irvine 2002; Shi and Stanley 2003). Genetic experiments indicate that GlcNAc elongation of O-fucose by fringe proteins is also important for Notch signaling, as loss of Drosophila fringe and mouse Lfng both result in severe Notch loss-of-function phenotypes in specific contexts (Evrard et al. 1998; Zhang and Gridley 1998; Correia et al. 2003). However, the fringe loss-of-function phenotypes in model organisms are much more limited than the O-fut1/Pofut1 loss-of-function phenotypes (even when all three mouse fringe genes are mutated (Moran et al. 2009)). Moreover, recessive mutations in LFNG cause a skeletal disorder called spondylocostal dysostosis in human patients (Sparrow et al. 2006), but a recessive mutation in POFUT1 has recently been reported in a patient with a multisystem developmental disorder (Takeuchi, Wong, et al. 2018). Together, these observations suggest that in addition to functioning as a docking site for GlcNAc residues added by fringe proteins, the O-fucose monosaccharide itself has key roles in Notch signaling independently of GlcNAc-elongation. In the following sections, we will discuss the contributions of the O-fucose monosaccharides and its GlcNAc-elongated version to Notch–ligand binding, with an emphasis on more recent studies.

O-Fucose Monosaccharides on Specific EGF Repeats are Essential for Some Aspects of Notch Signaling In Vivo

RNAi-mediated O-fut1 KD in Drosophila S2 cells results in a severe reduction in the binding of Notch to both Delta and Serrate without affecting the surface expression of Notch (Okajima et al. 2003; Sasamura et al. 2003). Mass spectral analysis of Notch expressed in these cells identified O-fucose monosaccharides on all Notch EGF repeats with an O-fucosylation consensus sequence, with the exception of EGF5, on which low levels of GlcNAc-fucose-O disaccharide were also observed (Harvey et al. 2016). This indicates that S2 cells have very low fringe activity and that the effects of O-fut1 KD on Notch–ligand binding must be due to the loss of O-fucose monosaccharides from Notch. The first study to examine the role of Notch O-fucose sites in vivo came from Ken Irvine’s group, who used transgenic overexpression of several mutant Notch proteins lacking individual O-fucosylation sites to begin to address this issue (Lei et al. 2003). In this study, overexpression of Drosophila Notch with O-fucosylation site mutation at EGF12 was shown to rescue the neurogenic phenotype in Notch mutant embryos. Drosophila embryonic neurogenesis does not depend on the function of fringe and Serrate and is controlled by Delta-mediated Notch singling (Fleming et al. 1990; Haines and Irvine 2003; Ishio et al. 2015). Therefore, this observation indicates that Notch can respond to Delta in the absence of O-fucose on EGF12.

We recently revisited this issue by using the advances in Drosophila genetic engineering technologies, which allowed us to generate O-fucose mutant Notch transgenes that are expressed at near endogenous levels (Venken et al. 2006; Leonardi et al. 2011; Pandey et al. 2019). Surprisingly, we found that a Notch transgene with a single serine-to-alanine mutation in the EGF12 O-fucosylation site completely lost the ability to rescue the embryonic neurogenesis and lethality of Notch-null animals (Pandey et al. 2019). A transgene with serine-to-threonine mutation fully rescued these phenotypes, indicating that it was the loss of O-fucose and not the amino acid change that impacted the function of Notch. Importantly, cell aggregation assays indicated that the O-fucose mutation in Notch EGF12 significantly reduced but did not abolish Notch–Delta interaction (Pandey et al. 2019). This observation provides a mechanistic explanation for the previous report that overexpression of this mutant version could bypass the requirement for EGF12 O-fucose in Delta–Notch signaling and fully rescue the embryonic neurogenesis phenotype of a Notch-null allele (Lei et al. 2003). It also indicates that the remaining Delta–Notch binding in the absence of this O-fucose is below the threshold required for sufficient Notch signaling in Drosophila embryos. Of note, a single O-fucose mutation in EGF9 also decreased the ability of Notch to rescue the embryonic neurogenesis and lethality by ~ 20%, but an O-fucose mutation in EGF8 did not affect the functionality of Notch in this assay (Pandey et al. 2019). Together, these data indicate that in the absence of fringe, O-fucose monosaccharides of EGF12 and to some extent EGF9 of Notch play nonredundant roles in Delta–Notch signaling in vivo (an essential role for O-fucose on EGF12, a minor but statistically significant role for O-fucose on EGF9).

The in vivo importance of O-fucosylation on Notch EGF12 was also examined by the Stanley lab on mouse NOTCH1 (Ge and Stanley 2008; Varshney et al. 2019). This group generated a knock-in allele in mouse Notch1 (Notch1^12f) which changed the fucose-binding threonine in EGF12 to an alanine (Ge and Stanley 2008). Animals homozygous for this allele were viable and fertile but showed defects in T cell development (Ge and Stanley 2008), a process in which LFNG plays an important role (Visan et al. 2006). However, further analysis of this allele indicated that the role of NOTCH1 EGF12 O-fucosylation site is not limited to fringe-dependent processes. First, the original work showed compound heterozygosity for Notch1^12f and a null allele of Notch1 resulted in embryonic lethality with phenotypes similar to Notch1^−/− embryos (albeit with a somewhat later lethality) (Ge and Stanley 2008). Secondly, a more recent study reported that on a pure C57BL/6 genetic background, Notch1^12f/12f becomes embryonic lethal, with phenotypes similar to but somewhat milder than the Notch1^−/− and Pofut1^−/− phenotypes (Varshney et al. 2019). Since some of the Notch1^12f/12f and Notch1^12f/− phenotypes reported in these studies are not seen in fringe-mutant mice, the data indicate that similar to Drosophila, the O-fucose monosaccharide on NOTCH1 EGF12 plays a critical role in mammalian Notch signaling distinct from serving as the docking site for GlcNAc.

GlcNAc Elongation of O-Fucose by fringe Regulates Notch Signaling

As mentioned before, fringe proteins are the first family of glycosyltransferase enzymes for which a role in Notch signaling was documented (Irvine and Wieschaus 1994; Bruckner et al. 2000; Moloney, Panin, et al. 2000). Biochemical, cell culture and genetic experiments indicate that fringe modifications of Notch promote Delta–Notch interaction and signaling but reduce Serrate–Notch interactions and signaling in Drosophila (Panin et al. 1997; Bruckner et al. 2000; Xu et al. 2005, 2007). An early study from the Irvine lab found that no single O-fucosylation site on Drosophila Notch can fully explain the fringe-mediated modulation of Notch–ligand interactions and concluded that GlcNAc elongation on multiple sites are required for the fringe effect in flies (Xu et al. 2005). A recent collaborative effort between our group and the Haltiwanger lab examined the critical fringe target sites of the Drosophila Notch by combining systematic biochemical and cell-based assays with in vivo studies (Pandey et al. 2019). We found that of the 21 O-fucosylated EGF repeats of Drosophila Notch, EGF8, 9 and 12 play key roles in mediating the effects of fringe on Notch–ligand binding. In the context of wing vein formation, which is primarily regulated by Delta–Notch signaling, GlcNAc residues on all of these three EGF repeats are important in a partially redundant manner. It is interesting to note that in contrast to embryonic neurogenesis, even mutating the O-fucosylation site on both EGF9 and EGF12 did not fully abolish the ability of a Notch transgene to rescue the haploinsufficient wing vein phenotypes of Notch^+/− animals (Pandey et al. 2019). This observation suggests that although in the absence of fringe, O-fucose on EGF12 is essential for Delta–Notch signaling during embryonic neurogenesis, addition of GlcNAc to O-fucosylated EGF8 by fringe can partially sustain Delta–Notch signaling during wing vein development, even when O-fucosylation of EGF12 and 9 are both ablated.

Unlike flies, which have one Notch, one Delta, one Serrate and one fringe, mammals harbor several paralogs for each of these pathway components. Therefore, not surprisingly, the regulation of mammalian Notch signaling by fringe proteins is more complicated and nuanced. Studies with mammalian proteins have shown that LFNG, MFNG and RFNG all promote DLL1-NOTCH1 binding and signaling (Yang et al. 2005; LeBon et al. 2014; Kakuda and Haltiwanger 2017). Systematic analysis of NOTCH1 O-fucosylation sites indicated that FNG proteins increase DLL1-NOTCH1 interaction by adding GlcNAc to NOTCH1 EGF8 and 12 (Kakuda and Haltiwanger 2017), which is in agreement with in vivo studies in flies (Pandey et al. 2019). Similar to their Drosophila homolog, LFNG and MFNG reduced JAG1-NOTCH1 signaling (LeBon et al. 2014; Kakuda and Haltiwanger 2017). However, RFNG promoted JAG1-NOTCH1 signaling (LeBon et al. 2014; Kakuda and Haltiwanger 2017). Moreover, unlike what has been observed in Drosophila, all three FNG proteins increase JAG1-NOTCH1 binding, suggesting that the LFNG- and MFNG-mediated inhibition of NOTCH1 activation by JAG1 occurs downstream of receptor–ligand binding (Kakuda and Haltiwanger 2017). Given that at a biochemical level, all three FNG proteins have the same function (Rampal et al. 2005), how can the effect of RFNG and LFNG/MFNG on JAG1-NOTCH1 signaling be in the opposite direction? The answer to this question came from a thorough analysis of the FNG target sites on NOTCH1 by the Haltiwanger lab (Kakuda and Haltiwanger 2017). This study indicated that LFNG and MFNG reduce JAG1-NOTCH1 signaling by adding GlcNAc to NOTCH1 EGF6 and EGF36, two of the sites which are not modified by RFNG. Although the in vitro catalytic efficiency of LFNG is ~ 10 times that of MFNG and RFNG, the latter enzymes have similar in vitro catalytic efficiencies (Rampal et al. 2005). Therefore, differences in the amino acid sequence of some EGF repeats might favor binding to and/or modification by LFNG/MFNG over RFNG.

While the bulk of information on the role of mammalian FNG proteins is related to how they modify NOTCH1 to regulate its response to DLL1 and JAG1, there are also reports on the impact of FNG modification on the function of other Notch pathway components. For example, the effects of FNG modifications of NOTCH2 on its response to different ligands might not mirror those of NOTCH1 (Hicks et al. 2000; Shimizu et al. 2001). Another recent report indicated that RFNG and LFNG promote the surface expression of DLL1 and DLL4 in the intestinal epithelium to promote Notch signaling in the neighboring cells in a non-cell autonomous manner (Kadur Lakshminarasimha Murthy et al. 2018). Therefore, understanding the diverse effects of FNG proteins on mammalian Notch signaling and their underlying mechanisms will await further investigation.

The Role of Fringe Proteins in Notch–Ligand Cis-Interactions

Cis-inhibitory interactions between Notch receptors and their ligands have emerged as a key regulatory mechanism for Notch signaling in both vertebrates and invertebrates (de Celis and Bray 1997; Klein et al. 1997; Micchelli et al. 1997; Jacobsen et al. 1998; Miller et al. 2009; Palmer et al. 2014; Baek et al. 2018). The original reports highlighted cis-inhibition of Notch by ligands, which would result in a cell autonomous loss of Notch signaling (de Celis and Bray 1997; Klein et al. 1997; Micchelli et al. 1997). However, later studies in Drosophila and zebrafish showed Notch receptors can also cis-inhibit a ligand expressed in the same cell, thereby leading to a non-cell autonomous loss of signaling in neighboring cells (Matsuda and Chitnis 2009; Becam et al. 2010). Using quantitative cell-based studies and mathematical modeling, the Elowitz lab established that the cis-inhibitory interactions between Notch and ligands are mutually inhibitory, i.e. the receptor and ligand molecules engaged in cis-interaction both fail to participate in a trans-activatory interaction with molecules on neighboring cells (Sprinzak et al. 2010). This meant that the relative levels of Notch and ligand molecules expressed in a cell play a key role in determining whether the cell will assume the signal-receiving or signal-sending role upon Notch–ligand cis-interactions: An excess of Notch molecules would favor a signal-receiving status, while an excess of ligand molecules would favor a signal-sending status.

Given the well-established role of fringe proteins in regulating Notch–ligand interactions and signaling in trans, it was of interest to know whether and in which direction fringe proteins affect the cis-interactions between Notch and ligands. Indeed, quantitative experiments in CHO cells showed that the effects of mammalian FNG proteins on the binding of NOTCH1 to DLL1 and JAG1 in cis mirror the FNG effects on those bindings in trans (LeBon et al. 2014). Specifically, LFNG, MFNG and RFNG all enhanced the cis-interaction between NOTCH1 and DLL1. Moreover, while LFNG and MFNG reduced NOTCH1-JAG1 interactions in cis, RFNG promoted their interaction. These observations suggested that FNG proteins can expand the repertoire of the signal-sending and signal-receiving states that a cell can adopt. For example, in a cell that co-expresses NOTCH1 and JAG1, presence of LFNG can decrease the cis-interaction between these two proteins, thereby allowing the cell to respond to DLL1 from a neighboring cell (signal-receiving role) and simultaneously send a JAG1 signal to the neighboring cell (signal-sending role).

The notion that in general FNG proteins promote NOTCH1-DLL1 cis-interaction and oppose NOTCH1-JAG1 cis-interaction (with the exception of RFNG) was indeed supported by dosage-sensitive genetic interaction studies in Drosophila in the same study. It was reported a hundred years ago that removing one copy of Drosophila Notch results in two distinct phenotypes in the adult wing, wing margin loss and wing vein expansion (Mohr 1919). The relative decrease in Notch levels compared to ligand levels in Notch^+/− animals seems to make some Notch-dependent processes susceptible to cis-inhibition of Notch by ligands (de Celis and Bray 2000). Examining the effects of increasing the gene dosage of Serrate and Delta in Notch^+/− animals led to the conclusion that their wing margin loss is primarily caused by cis-inhibition of Notch by Serrate, but the wing vein loss is primarily caused by cis-inhibition of Notch by Delta (LeBon et al. 2014). Interestingly, simultaneous removal of one copy of fringe in Notch^+/− animals had opposite effects on these two phenotypes: it enhanced the wing margin loss, compatible with increased Notch–Serrate cis-interaction, and it suppressed the wing vein phenotype, compatible with reduced Notch–Delta cis-interaction (LeBon et al. 2014). Together, these cell-based and in vivo observations indicate that by changing the level of fringe expression, cells can simultaneously adjust the response of Notch to each ligand in cis and thereby modulate the degree of Notch activation.

As mentioned in the previous section, there is one report on non-cell autonomous regulation of Notch signaling by RFNG and LFNG, where they increase the cell surface expression of DLL1 and DLL4 in the intestinal epithelium and thereby affect Notch signaling in neighboring cells (Kadur Lakshminarasimha Murthy et al. 2018). However, most if not all other FNG effects on Notch signaling are known to be mediated through modification of Notch receptors and are considered to be cell autonomous. For example, the above-mentioned effects of fringe heterozygosity on the Notch^+/− phenotypes were due to altering the cis-inhibition of Notch by ligands, not the other way around and were therefore cell autonomous in nature (LeBon et al. 2014). Theoretically, and supported by quantitative cell-based data (LeBon et al. 2014), if there is excess of Notch compared to ligands in a given cell, the activity of fringe can relieve (or strengthen) the cis-inhibition of ligand by Notch in that cell, which would affect the ability of that cell to activate Notch signaling in neighboring cells non-cell autonomously. But are there in vivo contexts wherein FNG plays such a role? Our recent work indicates that during Drosophila wing margin formation, fringe prevents the cis-inhibition of Serrate by Notch, thereby allowing Serrate to activate Notch signaling in neighboring cells and induce normal wing margin formation (Pandey et al. 2019). In the course of in vivo analysis of the effects of mutating specific Notch O-fucosylation sites on Notch signaling and fringe effects in Drosophila, we noticed that a Notch genomic transgene with O-fucosylation site mutations in both EGF8 and 12 results in a dominant negative wing margin loss, accompanied by enhanced cis-interaction between Notch and Serrate in cell culture studies (Pandey et al. 2019). Notably, adding one genomic copy of Delta did not modify this phenotype, but adding one copy of Serrate suppressed it. Together, these data indicate that addition of GlcNAc to Notch EGF8 and 12 by fringe decreases the cis-inhibition of Serrate by Notch in cells of the dorsal compartment of the wing pouch during wing margin development and leaves sufficient levels of Serrate available to trans-activate Notch in the ventral compartment, resulting in normal wing margin formation (Figure 3). Based on in vivo experiments in the context of mouse inner ear hair cell development, it has been suggested that LFNG and MFNG in the prosensory domain cells prevent NOTCH1 from cis-inhibiting JAG1 and JAG2, which in turn activate Notch signaling in neighboring cells to prevent them from adopting a sensory fate (Basch et al. 2016). Our data in flies lend support to this model and suggest, together with the study on the regulation of DLL surface expression by RFNG and LFNG (Kadur Lakshminarasimha Murthy et al. 2018), that at least in some contexts, the FNG loss-of-function phenotypes arise from non-cell autonomous perturbation of Notchsignaling.

Fig. 3.

Schematic model showing the non-cell autonomous inhibition of Notch signaling upon loss of fringe-modification sites on specific Notch EGF repeats during Drosophila wing margin formation. Close-up circles show the extended LBD. In the left-most panel, fringe modifications of Notch in dorsal compartment cells of the wing pouch prevent its interaction with Serrate and thereby allow Serrate to trans-activate Notch in ventral compartment cells, leading to normal wing margin formation. The middle panel shows that upon loss of fringe, Notch interacts with Serrate in the same cell and thereby cis-inhibits Serrate, leading to non-cell autonomous inhibition of Notch signaling in the ventral compartment. This results in wing margin loss. The right-most panel shows that loss of O-fucosylation sites on EGF8 and 12 (shown in the close-up circle) allows the cis-interaction between Notch and Serrate even in the presence of fringe (adapted from Pandey et al. 2019). This indicates that GlcNAc-elongation of O-fucose on Notch EGF8 and 12 by fringe is essential for its role in blocking the cis-inhibition of Serrate by Notch. Dl, Delta; N, Notch; Ser, Serrate.

Carbohydrates as Surrogate Amino Acids in the Notch–Ligand Interface

One of the key questions in the field has been the mechanism(s) through which (GlcNAc)-fucose-O residues regulate the binding affinity between Notch and ligands. Despite the advances in structural biology tools and techniques, the large size (~300 kDa) of Notch receptors hampers their structural analysis. Another obstacle in the analysis of Notch–ligand interactions is the inherently low affinity of the corresponding binding domains of Notch receptors and their ligands (Luca et al. 2015). To overcome these limitations, several groups have used small NOTCH1 fragments comprised of one to several EGF repeats and/or used directed evolution to increase the Notch–ligand affinity (Hiruma-Shimizu et al. 2010; Taylor et al. 2014; Luca et al. 2015; Hayakawa et al. 2016; Luca et al. 2017). The Nishimura group used nuclear magnetic resonance to show that addition of extended versions of O-fucose and O-glucose glycans to NOTCH1 EGF12 stabilizes the antiparallel β-sheet in the ligand-binding domain (LBD) of this EGF repeat (Hayakawa et al. 2016). The Handford laboratory used unglycosylated, O-fucosylated and fringe-modified O-fucosylated forms of prokaryotically expressed human NOTCH1 EGF11–13 to assess the effects of these O-glycan modifications on its binding to different mammalian Notch ligands using surface plasmon resonance and a flow cytometry-based assay (Taylor et al. 2014). They also determined the crystal structures of O-fucosylated and fringe-modified O-fucosylated human NOTCH1 EGF11–13. This study concluded that GlcNAc elongation of O-fucose on EGF12 by fringe markedly increases its affinity for the ligands JAG1 and DLL1 without altering its backbone structure. A major advance in the field came from a study by the Garcia lab, who performed directed evolution on the Notch-binding fragment from DLL4-ECD to increase its affinity for human NOTCH1 EGF11–13 and generated co-crystals of this fragment with NOTCH1 EGF11–13 (Luca et al. 2015). Interestingly, this study found that O-fucose on EGF12 is directly embedded in NOTCH1-DLL4 interface and functions as a surrogate amino acid by making specific and essential contacts with DLL4. These authors did not observe the fringe-added GlcNAc on the O-fucosylated NOTCH1 EGF12 in their structure. However, by modeling how the added GlcNAc would behave based on their available O-fucosylated structure, the authors predicted that GlcNAc will further enhance the NOTCH1-DLL4 binding by serving as a molecular bridge between the two molecules (Luca et al. 2015). More recently, another study from the same group showed a similar role for O-fucose residues at the NOTCH1-JAG1 interface (Luca et al. 2017). Altogether, these studies indicate that the O-fucose glycans at the NOTCH1 LBD directly participate in ligand-binding by contacting amino acids in the ligands and do not contribute to binding by allosteric effects.

The Extended Notch–Ligand Binding Domain

In Drosophila Notch, EGF11 and EGF12 were shown by the Artavanis-Tsakonas group to be necessary and sufficient for Notch–ligand interactions (Rebay et al. 1991), leading to the consideration of EGF11–12 as the LBD of Notch. However, data presented in that seminal report already indicated that EGF11–12 were not the only EGF repeats to show interactions with ligands (Rebay et al. 1991). Additional studies since then have shown that multiple EGF repeats in the ECD of Notch are involved in mediating ligand interactions, in particular those between EGF6–15, such as EGF8 (Xu et al. 2005; Yamamoto et al. 2012; Andrawes et al. 2013; Kakuda and Haltiwanger 2017). Furthermore, using cell-based assays, fringe modifications at EGF8 and EGF12 were shown to enhance DLL1-mediated NOTCH1 activation by increasing binding (Kakuda and Haltiwanger 2017). Additional evidence for the potential importance of O-fucose residues on EGF8 and EGF12 in Notch–ligand binding was provided by one of the above-mentioned studies, which solved the co-crystal structure of the NOTCH1 EGF8–12 region bound to the Notch-binding domain of JAG1 (Luca et al. 2017). This report showed a direct contact between the O-fucose residues on NOTCH1 EGF8 and EGF12 with amino acids from JAG1 and that mutations in EGF8 and EGF12 O-fucosylation sites dramatically decreased NOTCH1-JAG1 interaction in cell-based assays (Luca et al. 2017). Finally, in support of these biochemical and cell culture analyses, a recent report from our lab has highlighted the importance of (GlcNAc)-fucose-O glycans on EGF 8, 9 and 12 of the Drosophila Notch for Notch–ligand interactions and ligand-mediated signaling in flies (Pandey et al. 2019). Together, these studies support the notion that the core LBD of Notch receptors consists of EGF repeats 8–12 (Figure 1B).

The role of O-glucose glycans in Notch–ligand interactions

As mentioned before, cell-based and genetic experiments suggest that O-glucose modifications of NOTCH1/Notch EGF repeats by POGLUT1/Rumi do not directly impact Notch–ligand interactions and regulate Notch signaling at a step downstream of ligand-binding (Acar et al. 2008; Fernandez-Valdivia et al. 2011; Leonardi et al. 2011; Taylor et al. 2014). This is in agreement with DLL4-NOTCH1 co-crystal structure (Luca et al. 2015). Specifically, unlike the O-fucose glycan on NOTCH1 EGF12 which was in direct contact with DLL4, the two O-glucose residues attached to NOTCH1 EGF12 and EGF13 did not contact DLL4 (Luca et al. 2015). Of note, the solved structure suggested that each of these two O-glucose residues protect two hydrophobic amino acids in their corresponding EGF repeat from solvent exposure. It is worth mentioning that the O-glucose added to EGF11 by POGLUT2/POGLUT3 directly contacts amino acids in DLL4 (Luca et al. 2015), suggesting that it might directly impact ligand-binding. However, cell-based assays so far have not shown a role for this glycan in NOTCH1-DLL1 binding (Takeuchi, Schneider, et al. 2018).

Xylosylation of Drosophila Notch Specifically Affects Notch Trans-Activation by Delta

The above-mentioned studies provide strong evidence that addition of O-glucose to Notch EGF repeats by POGLUT enzymes does not play a key role in Notch–ligand binding. However, our studies on the role of the Drosophila glucoside xylosyltransferase (Shams) in Notch signaling indicate that addition of xylose to O-glucose monosaccharides on Notch by this enzyme affects Notch–ligand binding (Lee et al. 2017). Loss of both copies of shams did not result in lethality but led to loss of specific sensory bristles in the fly head and partial loss of wing veins, phenotypes compatible with a gain of Notch signaling (Lee et al. 2013). Mass spectral analysis of Drosophila Notch showed that all 18 EGF repeats with a predicted O-glucosylation site are modified efficiency by Rumi/Poglut, but xylose extension is only seen on a subset of EGF repeats (primarily in EGF15–20) (Lee et al. 2013; Harvey et al. 2016). Phenotypic analysis of transgenic flies expressing near endogenous levels of Notch with O-glucose mutations in EGF16–20 and also animals overexpressing human GXYLT1 or XXYLT1 confirmed that addition of xylose to these EGF repeats negatively regulates Notch signaling in flies (Lee et al. 2013). Loss of shams in the pupal wings resulted in an increase in the surface expression of Notch but did not affect Notch surface expression in the third instar wing imaginal disk (Lee et al. 2013). Therefore, the precise mechanism for the negative regulation of Notch signaling in specific contexts by xylose residues remained to be elucidated.

To provide mechanistic insight into how xylosylation affects Drosophila Notch signaling, we used in vivo and cell-based approaches and found that shams regulates Notch signaling by modulating Notch–ligand binding (Lee et al. 2017). The data suggest that (1) the shams loss-of-function phenotypes are caused by enhanced Delta-mediated signaling, not Serrate-mediated signaling, and (2) addition of xylose to Notch by Shams decreases the binding of Notch to and its activation by trans-Delta without affecting the Notch-trans-Serrate binding or binding of Notch to and its inhibition by cis-ligands. This way, xylosylation determines the balance between Notch receptor’s responses to trans-Delta vs. cis-ligands and modulates the pathway activation in contexts like wing vein development (Lee et al. 2017), which primarily depend on Delta (de Celis et al. 1997; Huppert et al. 1997; Zeng et al. 1998). As mentioned before, the critical Notch glycosylation sites involved in the regulation of Delta–Notch signaling by fringe and Shams are different from each other (EGF8, 9 and 12 for fringe, EGF16–20 for Shams). Therefore, it is not surprising that the two enzymes have distinct loss-of-function phenotypes (Correia et al. 2003; Lee et al. 2013) and affect Notch–ligand binding differently (fringe promotes Delta–Notch binding both in trans and in cis, Shams opposes Delta–Notch binding in trans and does not affect Delta–Notch binding in cis; fringe opposes Notch–Serrate binding both in trans and in cis; Shams does not seem to affect Notch–Serrate binding). These observations suggest that (1) EGF repeats outside of the extended LBD can exclusively or preferentially modulate the binding of Notch to one of its ligands and (2) Notch might use different structural conformations or domains to bind trans- vs. cis-ligands.

Do Mammalian Xylosyltransferases Regulate Notch Signaling Similar to their Fly Homologs?

Only a subset of Drosophila Notch EGF repeats are xylosylated by Shams and even a smaller subset harbor a second xylose at low stoichiometry, which is added by Xxylt (Lee et al. 2013; Harvey et al. 2016; Pandey et al. 2018). Mass spectral analysis of mouse NOTCH1 shows that similar to Drosophila Notch (Harvey et al. 2016), all predicted sites are modified by POGLUT1, although some degree of site-specific and cell type-specific variation exists (Rana et al. 2011). However, in contrast to Drosophila Notch (Lee et al. 2013; Harvey et al. 2016), most if not all O-glucosylated EGF repeats show an efficient elongation to a xylose-xylose-glucose-O trisaccharide (Rana et al. 2011). The limited distribution and low stoichiometry of xylosylation on fly Notch compared to mammalian NOTCH1 could be due to the presence of two GXYLT proteins in mammals compared to one in flies. It might also result from more efficient enzymatic activity and/or broader EGF repeat substrate recognition of the mammalian enzymes compared to the fly enzymes. For example, when Shams and Drosophila Xxylt were co-expressed with Drosophila Notch EGF16–20 region in Drosophila S2 cells lacking endogenous Shams, they were only able to generate a trisaccharide on EGF16 (Pandey et al. 2018). However, co-expression of human GXYLT1 and XXYLT1 in the same system generated trisaccharides on all four EGF repeats examined (EGF16, 18, 19 and 20) (Pandey et al. 2018). Moreover, overexpression of human XXYLT1 in the fly wing resulted in a much more severe loss of Notch signaling compared to overexpression of fly Xxylt using the same driver at the same temperature, even though both phenotypes fully depended on intact O-glucosylation sites on EGF16–20 (Lee et al. 2013; Pandey et al. 2018). Regardless of the mechanism(s), since the negative regulation of Delta–Notch trans-activation by Drosophila Shams is mediated via xylosylation of EGF16–20, it is difficult to predict the impact of loss or gain of GXYLT1, GXYLT2 and XXYLT1 on mammalian Notch signaling.

Overexpression of human GXYLT1 in Drosophila reduced Notch activation and restored the wing vein loss phenotype in shams mutants, which is caused by increased Notch signaling (Lee et al. 2013). This suggests that GXYLT1 is capable of replacing the function of Shams by adding xylose to O-glucosylated EGF16–20. However, at this time no reports exist on the effect of GXYLT1 on mammalian Notch signaling. Similarly, although expression of human XXYLT1 in flies resulted in loss of Notch signaling by adding xylose residues to Notch EGF16–20 (Lee et al. 2013), no functional studies in mammals have yet linked XXYLT1 to the Notch pathway. It is worth noting that analysis of the publicly available cancer genomic database cBioportal (Cerami et al. 2012) showed significant XXYLT1 gene amplification in squamous cell carcinomas of the lung and head and neck (Yu et al. 2015), which are commonly associated with reduced Notch signaling (Siebel and Lendahl 2017). This suggests that human XXYLT1 might indeed reduce Notch signaling.

A recent report on GXYLT2 has used siRNA-mediated KD and overexpression studies in human breast cancer (MDA-MB-231) and gastric adenocarcinoma (BGC823) cells to show that GXYLT2 promotes NOTCH1 signaling (Cui et al. 2019). GXYLT2 KD dramatically reduced in vivo tumor growth and the number of metastatic nodules in the livers of mice injected with MDA-MB-231 cells compared to animals injected with control MDA-MB-231 cells. Moreover, GXYLT2 overexpression in BGC823 cells promoted the proliferation, migration and invasion of these cells in culture and increased N1-ICD and HES1 levels. All of these phenotypes were suppressed upon treatment with the γ-secretase inhibitor DAPT (Cui et al. 2019), which blocks the activations of all Notch receptors. These observations suggest that at least in these contexts, GXYLT2 promotes NOTCH1 signaling, an effect opposite to that of Drosophila Shams (Lee et al. 2013, 2017). It is not known, though, whether the observed effect is mediated at the level of Notch–ligand interaction. In summary, given the presence of more than one GXYLT in mammals and the potential importance of specific xylose modification sites on Notch pathway activity (like EGF16–20 in flies), the effects of mammalian xylosyltransferases on Notch signaling are likely to be more complex and nuanced that those of their fly counterparts.

The role of O-GlcNAcylation in Notch–ligand binding and Notch signaling

Loss of both maternal and zygotic functions of Eogt in Drosophila results in embryonic lethality (Sakaidani et al. 2011). However, the mutant embryos do not exhibit the classic neurogenic phenotype observed in Notch pathway mutants. Instead, the lethality in these animals seems to arise from impaired cell–matrix or cell–cell interaction due to loss of O-GlcNAc from another Eogt target called Dumpy (Sakaidani et al. 2011). Interestingly, wing-specific KD of Eogt results in a wing blistering phenotype which is suppressed by simultaneous removal of one copy of Notch or several other Notch pathway components. These observations suggest that although Eogt is not essential for Drosophila Notch signaling, the Eogt loss-of-function phenotype in the wing can be improved by reducing Notch signaling, suggesting that enhanced Notch signaling might contribute to the phenotype (Muller et al. 2013).

Eogt null mice are viable, fertile and do not show gross morphological abnormalities (Sawaguchi et al. 2017; Varshney and Stanley 2018). However, germline or endothelial-specific loss of both copies of Eogt resulted in mild retinal angiogenesis defects including enhanced blood vessel branching and leaky blood vessels similar to animals with decreased Notch signaling, including Notch^+/− and Rbpj^+/− mice (Sawaguchi et al. 2017). Furthermore, Eogt-KD and Eogt^−/− mammalian cell lines showed reduced binding of NOTCH1 to DLL1 and DLL4 (but not to JAG1), accompanied by reduced NOTCH1 signaling (Sawaguchi et al. 2017). Importantly, mutations in the O-GlcNAc sites of several conserved EOGT targets in NOTCH1 EGF repeats recapitulated the effects of loss of EOGT on NOTCH1-DLL4 binding, but neither loss of Eogt nor these mutations affected the surface expression of NOTCH1. Together, these observations provide compelling evidence that specific O-GlcNAc residues added by EOGT to the NOTCH1 receptor promote DLL1/4-mediated NOTCH1 signaling without affecting JAG1-mediated NOTCH1 signaling. It is interesting to note that the phenotypic similarity between homozygous mutant Eogt and heterozygous mutant Notch1 and Rbpj mirrors the relationship between EOGT and NOTCH1/RBPJ/DLL4 mutations in human patients with Adams–Oliver syndrome (AOS). The disease is characterized by aplasia cutis congenital and terminal transverse limb defects, together with variable penetrance of cardiovascular defects and brain anomalies (Adams and Oliver 1945; Snape et al. 2009; Algaze et al. 2013). While homozygous mutations in EOGT cause an autosomal recessive form of AOS (Shaheen et al. 2013; Cohen et al. 2014), heterozygous mutations in RBPJ, NOTCH1 and DLL4 have all been identified in autosomal dominant AOS patients (Hassed et al. 2012; Algaze et al. 2013; Stittrich et al. 2014). Altogether, these data indicate that although EOGT is not essential for Notch signaling, it makes important contributions to some aspects of Notch signaling in mice and humans.

Dosage-sensitive contributions of glycosylation to Notch signaling and human disease

Animal development is highly sensitive to the gene dosage of the Notch pathway components. Studies in animal model organisms and human patients with a number of Notch-related diseases indicate that normal development requires both copies of most of the core pathway components (Penton et al. 2012; Masek and Andersson 2017). In fact, loss of one copy of NOTCH1, NOTCH2, DLL1, DLL4, JAG1 and RBPJ each causes a human developmental disorder (Li et al. 1997; Oda et al. 1997; McDaniell et al. 2006; Hassed et al. 2012; Stittrich et al. 2014; Meester et al. 2015; Fischer-Zirnsak et al. 2019). Therefore, identification of modifier genes which can modulate the Notch pathway strength can suggest potential therapeutic avenues. This is especially the case if removing one copy of a given modifier improves Notch pathway haploinsufficient phenotypes without affecting Notch signaling in an otherwise wild-type background (i.e. with normal gene dosage of core Notch pathway components).

Dosage-sensitive modification of Jag1^+/− phenotypes by glycosyltransferases: implications for Alagille syndrome

Work over the last decade or so has documented a number of instances where reducing the gene dosage of an EGF repeat glycosyltransferase affects Notch signaling. A study by the Loomes group (Ryan et al. 2008) examined fringe genes as potential mediators of phenotypic variability of Alagille syndrome (ALGS), which is associated with bile duct paucity and abnormalities in other organs and is caused by JAG1 dominant mutations in most cases (Alagille et al. 1975; Warthen et al. 2006). The authors did not report any bile duct defects in Jag1^+/− animals used in this study (Ryan et al. 2008), in agreement with the original study that generated and analyzed this strain on a mixed genetic background (Xue et al. 1999). However, they found significant biliary proliferation in Jag1^+/−; Rfng^+/− and Jag1^+/−; Lfng^+/− animals. In contrast, adult Jag1^+/−; Mfng^+/− animals showed an increase in the bile duct to portal vein ratio compared to control animals (Ryan et al. 2008). Of note, Rfng^−/−, Lfng^−/− and Mfng^−/− adult animals did not show any biliary phenotypes in a wild-type (Jag1^+/+) background (Ryan et al. 2008). Together, these results indicate that heterozygosity for Jag1 renders the liver development sensitive to the dosage of fringe genes, suggesting these genes as candidate modifiers of the liver phenotypes in ALGS patients.

A more recent study from our group sought to determine whether Poglut1 genetically interacts with Jag1 (Thakurdas et al. 2016). During the course of this study, we serendipitously observed that on a pure C57BL/6 genetic background, Jag1^+/− animals show significant bile duct paucity, reminiscent of ALGS patient livers. Interestingly, Jag1^+/−; Poglut1^+/− animals showed a significant improvement in biliary development and liver histology compared to their Jag1^+/− siblings, suggesting that wild-type levels of POGLUT1 decrease JAG1-mediated signaling in Jag1^+/− mouse livers (Thakurdas et al. 2016). The mechanism for this dosage-sensitive rescue has not been fully elucidated, although the available data, including efficient modification of all four predicted POGLUT1 target sites of JAG1, suggest that decreasing O-glucose glycans on the JAG1 protein itself might mediate the observed effect (Thakurdas et al. 2016). It remains to be seen whether JAG1 is indeed a biologically relevant target of POGLUT1 or if the observed effects are mediated through reduced Notch receptor glycosylation. It would also be interesting to examine whether altering the levels of the first and/or second xylose residue in the xylose-xylose-glucose-O glycans results in a similar improvement in Jag1^+/− phenotypes or not.

Dosage-sensitive regulation of Delta-mediated Notch signaling by Drosophila shams (Gxylt)

As mentioned above, loss of the fly shams (Gxylt) results in increased Delta-mediated Notch signaling in specific contexts (Lee et al. 2013; Lee et al. 2017). More recently, we characterized the function of Drosophila Xxylt and asked whether shams or Xxylt show any dosage-sensitive effects on Notch signaling (Pandey et al. 2018). Loss of Drosophila Xxylt did not affect animal survival and fertility and did not result in any gross morphological phenotypes compatible with altered Notch signaling (Pandey et al. 2018). This is in line with the limited distribution and low stoichiometry of the xylose-xylose-glucose-O trisaccharides on fly Notch (Lee et al. 2013; Harvey et al. 2016). We next tested whether the Notch haploinsufficient phenotypes in Drosophila wing (i.e. wing margin loss and wing vein thickening) can be modified by simultaneous decrease in xylosyltransferases (shams and Xxylt). Indeed, loss of one copy of shams resulted in the partial rescue of both of these phenotypes. Moreover, loss of one copy of shams significantly increased the ability of overexpressed Delta to trans-activate Notch in vivo (Pandey et al. 2018). Since shams heterozygous animals do not exhibit any phenotypes in a Notch^+/+ background (Lee et al. 2013), these data provide another example for dosage-sensitive effect of a glycosyltransferase mutation on a Notch pathway haploinsufficient phenotype despite lack of a similar effect on a wild-type background. Collectively, the Drosophila and mouse studies discussed above indicate that heterozygosity for Notch receptors and ligands can make Notch signaling sensitive to the gene dosage of specific glycosyltransferases. Since at least some of these effects are receptor paralog- or ligand-specific, targeting these enzymes might provide novel strategies for therapeutic manipulation of Notch signaling in disease contexts.

Distinct effects of POGLUT1 and POFUT1 heterozygous vs. homozygous mutations in human patients

Dominant mutations in POGLUT1 and POFUT1 have been reported in patients with Dowling–Degos syndrome, which is a hereditary hyperpigmentation disorder (Dowling and Freudenthal 1938; Degos and Ossipowski 1954; Li et al. 2013; Basmanav et al. 2014). Also, dominant POGLUT1 mutations have been identified in a related hyperpigmentation syndrome called Galli–Galli (Wilson et al. 2017; Kono et al. 2019). A number of the above-mentioned mutations in POGLUT1 and POFUT1 were nonsense and frameshift, and several POGLUT1 missense mutations were shown by enzymatic assays to severely reduce its enzymatic activity (Li et al. 2013; Basmanav et al. 2014; Wilson et al. 2017; Ralser et al. 2019), in line with the notion that the Dowling–Degos/Galli–Galli mutations in these genes are loss-of-function. Morpholino-induced KD of POFUT1 in zebrafish resulted in hypopigmentation and abnormal melanin distribution, accompanied by a reduction in the expression of NOTCH1, NOTCH2 and the Notch pathway target HEY1 (Li et al. 2013). SiRNA-mediated KD of POGLUT1 in a melanocyte cell line resulted in altered expression of some Notch pathway components (including decreased NOTCH1 and NOTCH2; increased JAG1), but did not affect the expression of the Notch pathway target HES1 (Ralser et al. 2019). In addition, dominant mutations in the γ-secretase component PSENEN are also found in patients with Dowling–Degos phenotypes associated with inflammatory acne (Ralser et al. 2017). Moreover, studies in mice have demonstrated a key role for Notch signaling in melanocyte development and hair color maintenance (Moriyama et al. 2006; Schouwey et al. 2007; Kumano et al. 2008). Together, these studies indicate a dosage-sensitive role for both POGLUT1 and POFUT1 in the regulation of Notch signaling in human pigmentation, although the precise effect of these mutations on the activity of individual Notch receptors remains to be identified.

More recently, the Paradas group reported recessive POGLUT1 mutations in multiple families with a limb-girdle type form of muscular dystrophy, LGMD-R21 (Servian-Morilla et al. 2016; Servian-Morilla et al. 2020). The first family reported with this disease was homozygous for the D233E missense mutation, which dramatically reduced the enzymatic activity of POGLUT1 although some residual activity remained (Servian-Morilla et al. 2016). Analysis of muscle biopsies from the patients showed a severe decrease in NOTCH1 cleavage and signaling, but the patients did not exhibit any skin phenotypes (Servian-Morilla et al. 2016). Another recent study reported a homozygous missense mutation in POFUT1 in a human patient with a multi-system developmental syndrome (Takeuchi, Wong, et al. 2018). This mutation severely reduced POFUT1’s enzymatic activity and diminished the ability of POFUT1 to restore Notch signaling upon transfection into POFUT1-deficient cells. However, neither the patients nor the heterozygous parents showed any skin abnormalities (Takeuchi, Wong, et al. 2018). These observations suggest an intriguing genotype–phenotype relationship for POGLUT1 and POFUT1 mutations, as while some dominant loss-of-function mutations cause a specific skin disease, homozygous mutations for the same genes affect other organ systems without involving the skin. Dowling–Degos and LGMD-R21 are very rare, and only one patient has so far been reported with POFUT1 homozygous mutation. Therefore, identification of new patients, combined with additional mechanistic and genetic studies are required to determine whether these differences can be explained by the genetic background.

Association of POGLUT1 and POFUT1 overexpression with cancer

As mentioned, POGLUT1 (hCLP46) was originally isolated from the hematopoietic stem/progenitor cells of a human patient with a hematological malignancy (Teng et al. 2006). Additional studies reported that POGLUT1 is overexpressed in a number leukemia cell lines and in the peripheral blood cells of patients with T-cell acute lymphoblastic leukemia and acute myelogenous leukemia (Wang et al. 2010). Given the important roles played by Notch signaling in hematological malignancies, these studies suggested a potential role for POGLUT1 in the pathophysiology of these diseases. More recently, analysis of a human cancer gene expression database and staining of human cancer samples indicated POGLUT1 overexpression in nonsmall cell lung cancer (NSCLC) (Chammaa et al. 2018). Importantly, there was a statistically significant correlation between higher POGLUT1 levels and worse prognosis in these patients. Of note, POGLUT1 KD reduced Notch target genes in NSCLC cells, suggesting a possible mechanism in worsening of the disease outcome upon POGLUT1 overexpression (Chammaa et al. 2018).

POFUT1 overexpression has also been reported in several malignancies associated with dysregulated Notch signaling. Hyperactivation of the Notch pathway is thought to contribute to several solid tumors, including breast cancer (Siebel and Lendahl 2017). Interestingly, a recent study has found that high POFUT1 expression level is strongly correlated with increased NOTCH1-ICD expression and poor prognosis in breast cancer patients (Wan et al. 2017). Similarly, POFUT1 was shown to be overexpressed in cell lines and tissue samples obtained from hepatocellular carcinoma (HCC) patients, and higher POFUT1 levels were correlated with poor prognosis and higher recurrence rate in these patients (Ma et al. 2016). As expected, POFUT1 overexpression increased Notch signaling and POFUT1 KD reduced Notch signaling in HCC cell lines, providing a potential mechanism for the observed negative correlation between POFUT1 level and HCC disease outcome (Ma et al. 2016). More recent studies have established a potential functional role for POFUT1 overexpression in colorectal cancer (CRC) and have shown that higher POFUT1 levels in stage I CRC is associated with metastasis (Chabanais et al. 2018; Du et al. 2018). Altogether, these reports suggest that not only reducing POGLUT1 and POGLUT1 gene dosage can affect Notch signaling in human patients, increasing their expression level might also contribute to some forms of cancer.

Additional glycosylation-related mechanisms for Notch pathway regulation

Although the bulk of studies on the regulation of Notch signaling by glycosylation has focused on the three types of O-glycan discussed in previous section, several studies have provided a link between other types of glycosylation and also a specific lectin and the regulation of Notch signaling. In this section, we provide three exciting examples of these glycosylation-related mechanisms.

MGAT3 promotes NOTCH1 signaling, likely at the level of receptor trafficking

The glycosyltransferase mannoside acetylglucosaminyltransferase 3 (MGAT3, also known as GnT-III) adds a β1,4-linked GlcNAc to the mannosyl core of N-linked glycans (Figure 4A) to generate a structure commonly known as a bisecting GlcNAc (Narasimhan 1982; Nishikawa et al. 1992). Multiple minimal consensus sites for the attachment of N-glycans are present in ECD of the NOTCH1 protein (Figure 4B), and both Drosophila Notch and mouse NOTCH1 have been shown to be N-glycosylated (Johansen et al. 1989; Moloney, Shair, et al. 2000). MGAT3 mRNA levels are significantly increased in epithelial ovarian cancers (EOCs) and result in the generation of aberrant, truncated N-linked glycan structures in these tumors (Abbott et al. 2008; Allam et al. 2017). Importantly, EOC patients with lower MGAT3 levels show improved outcomes. Moreover, shRNA-mediated MGAT3 KD reduced tumor growth and metastasis in a xenograft model of ovarian cancer in mice, suggesting a critical role for one or more targets of this enzyme in determining the prognosis of ovarian cancer (Allam et al. 2017). Given the identification of several Notch pathway modifiers as differentially glycosylated from human ovarian cancer tissues and the established role of aberrant Notch signaling in ovarian cancers (Allam et al. 2015; Cancer Genome Atlas Research Network 2011), the Abbott group examined whether MGAT3 regulates Notch signaling in spheroids generated from several ovarian cancer cell lines. They found that suppression of MGAT3 mRNA levels reduces expression levels of NOTCH1 and NOTCH3 transcripts and their downstream targets HEY1 and HES1 (Allam et al. 2017). Importantly, KD of this gene in EOCs led to blockage of NOTCH1 activation more efficiently than a γ-secretase inhibitor. The authors also identified the NOTCH1 receptor itself as a target of MGAT3. This study provided evidence to support a model (Figure 4C) in which MGAT3 decreases the trafficking of NOTCH1 to late endosome/lysosome and thereby allows a bigger pool of signaling-competent NOTCH1 to remain in the cell (Allam et al. 2017). These findings suggest that targeting MGAT3 and the resulting decrease in NOTCH1 signaling may provide a therapeutic opportunity for the treatment of EOCs. Further studies are needed to examine whether the effects of MGAT3 on NOTCH1 trafficking is mediated by altering the structure of N-glycans on NOTCH1 itself or by affecting the function of other MGAT3 targets.

Fig. 4.

(A) Example of an N-glycan highlighting the bisecting GlcNAc and the enzyme responsible for its addition, MGAT3 (GnT-III). Gal, galactose; GlcNAc, N-acetylglucosamine; Man, mannose; Neu5Ac, N-acetylneuraminic acid (sialic acid). (B) Schematic showing the ECD of the human NOTCH1 with 36 EGF repeats (gray boxes) followed by three LNR motifs (pink ovals) and an HD. Red asterisks indicate the minimal consensus sites for the attachment of N-glycan (N-X-S/T, where X is any amino acid other than proline). Blue asterisks represent the confirmed glycosylation sites for GALNT11. Note that N-glycan and O-linked N-acetylgalactosamine (O-GalNAc) are also present outside the EGF region. (C) Model showing the proposed role of MGAT3 in NOTCH1 signaling. Left: High levels of MAGT3 inhibit the trafficking of N1ICD to lysosomes, thereby ensuring that high levels of N1ICD are maintained and strong NOTCH1 signaling is achieved. Right: In cells expressing low levels of MAGT3, N1ICD is redirected to lysosomes and degraded, thereby decreasing the N1ICD level and NOTCH1 signaling. MAML1, Mastermind-like 1. (D) Addition of GalNAc to serine or threonine (S/T) residues by GALNT family enzymes including GALNT11. The resulting O-GalNAc can be extended by the addition of galactose, GlcNAc and/or sialic acid.

Regulation of Notch signaling by galectin-3

Galectin-3 (official symbol: Lgals3), is a lectin which binds β-galactoside-containing glycoconjugates and is found in both intracellular and secreted forms (Liu and Rabinovich 2005). Galectin-3 is expressed in many tissues, including heart, kidney, liver, lung, stomach, colon, ovary and uterus (Kim et al. 2007) and is implicated in carcinogenesis and immune response (Liu and Rabinovich 2005). Compared to control mice, galectin-3 mutant mice showed an increase in the expression of multiple Notch pathway components upon experimental infection with Leishmania major (Fermino et al. 2016). Even in the absence of Leishmania infection, bone marrow-derived dendritic cells (BMDCs) from the galectin-3 mutant mice exhibited an increase in the expression of DLL4 and JAG1 ligands. Moreover, loss of galectin-3 differentially affected the ability of BMDCs to respond to Notch ligands, with an overall increase in JAG1 sensitivity but a variable response to DLL4 (increase in some contexts, decrease in others) (Fermino et al. 2016). The authors also showed that addition of recombinant human galectin-3 to the culture media reduces the activation of Notch signaling in BMDCs by both JAG1 and DLL4. Overall, this report suggested that galectin-3 has an inhibitory role for the activation of Notch signaling in immune cells, especially in response to JAG1 (Fermino et al. 2016).

In contrast to BMDCs, galectin-3 helps maintain ovarian cancer stem cells (CSCs) by increasing NOTCH1 signaling and thereby enhances tumorigenesis (Kang et al. 2016). Using RNAi-mediated galectin-3 KD and galectin-3 overexpression in ovarian cancer cell lines, this group showed that galectin-3 promotes the proliferation, cell viability, migration and invasion capacity of these cells. These experiments also showed that galectin-3 promotes the expression of HEY1 and HES1, which are among the key transcription factors induced downstream of Notch signaling (Kang et al. 2016). Interestingly, this study used immunoprecipitation experiments in ovarian cancer cells to show that galectin-3 is in a complex with N1ICD through its C-terminal carbohydrate recognition domain (CRD) (Kang et al. 2016).

Additional studies have suggested a role for galectin-3 in the regulation of Notch signaling. One study showed that co-culturing human fetal osteoblast cells with human breast cancer or prostate cancer cell lines expressing and secreting galectin-3 leads to inhibition of osteoblast differentiation (Nakajima et al. 2014). Galectin-3 in the culture media promoted NOTCH1 cleavage and the expression of the Notch target HEY1 in fetal osteoblasts in a CRD-dependent manner, suggesting that recognition of specific carbohydrate residues by galectin-3 mediates this effect (Nakajima et al. 2014). This effect was not limited to fetal osteoblasts, as incubation of HEK293 cells with galectin-3 also resulted in NOTCH1 activation. This study reported that full-length galectin-3 and its CRD can both pull down full-length NOTCH1, suggesting that binding of galectin-3 to NOTCH1-ECD might promote the activation of NOTCH1 cleavage and signaling. Another study examined the role of hypoxia-induced secretion of galectin-3 by cancer cells in inducing angiogenesis and found that direct binding of galectin-3 to JAG1 increases JAG1’s half-life and preferentially promotes JAG1-Notch signaling over DLL4-Notch signaling (Dos Santos et al. 2017). The effect of galectin-3 on JAG1-mediated signaling in this context seems to be the opposite of what has been reported in immune cell, as described above (Fermino et al. 2016). Altogether, these data suggest context-dependent modulation of Notch signaling by the lectin galectin-3. It will be interesting to know whether galectin-3 regulates Notch signaling by directly binding carbohydrate residues on Notch receptors and/or ligands or whether its effects on Notch signaling are indirect. It will also be important to examine the effect of galectin-3 on Notch signaling during normal development or tissue maintenance.

GALNT11 promotes NOTCH1 signaling, potentially by increasing S2 cleavage

Polypeptide N-acetylgalactosaminyltransferase 11 (GALNT11) is a member of the evolutionarily conserved family of O-GalNAc transferases, which catalyze the transfer of GalNAc to serine/threonine residues (Figure 4D) in target proteins to initiate mucin-type O-glycosylation (Bennett et al. 2012). Khokha, Brueckner and colleagues had previously identified GALNT11 as a candidate gene responsible for a congenital developmental disease called heterotaxy, which is a disorder of left–right body patterning (Fakhro et al. 2011). Given the similarity between the left–right asymmetry phenotypes caused by reduced NOTCH1 and reduced GALNT11, the investigators asked whether GALNT11 glycosylates NOTCH1 and affects NOTCH1 signaling. Mass spectrometry analysis showed that GALNT11 indeed O-glycosylates human NOTCH1 at EGF6, EGF36 and a site in the juxtamembrane region (Figure 4B) close to the S2 cleavage site (Boskovski et al. 2013). In vitro cleavage studies on a peptide harboring the cleavage site of the NOTCH1 receptor and its flanking amino acids showed that addition of O-GalNAc to a neighboring amino acid promotes the S2 cleavage of NOTCH1 (Boskovski et al. 2013). These findings led the investigators to propose a new mechanism for Notch regulation, in which site-specific GALNT11 glycosylation of a non-EGF region in NOTCH1 promotes its S2 cleavage by ADAM proteins and thereby increases NOTCH1 activation (Boskovski et al. 2013).

Future perspectives/conclusions

Research into the glycobiology of Notch has made it clear that sugar modifications of Notch receptors and their ligands are not mere decorations, but play important regulatory roles. Advances in the analysis of site-specific glycan structures on EGF repeats, solving the crystal structure of Notch glycosyltransferases and Notch–ligand complexes, and in vivo analysis of Notch and its glycosyltransferases enabled by modern genetic and genome engineering technologies have all contributed to the recent progress in this field and will likely continue to do so. Strategies for ligand-specific modulation of Notch signaling can have important therapeutic potential. For example, several fucose analogs were recently shown to specifically inhibit DLL-mediated, but not JAG1-mediated, Notch activation (Schneider et al. 2018). Moreover, as discussed above, a number of EGF repeat glycosyltransferases can modify specific aspects of Notch signaling in a dosage-sensitive manner. Accordingly, we expect further work in the area of Notch glycobiology to provide potential avenues for therapeutic modulation of this pathway in human patients. Although the biological roles of O-fucose, O-glucose and O-GlcNAc have primarily been studied in the context of Notch signaling, other important targets and processes are also regulated by these glycans. Examples include the regulation of cell–matrix interactions via O-GlcNAcylation of the Drosophila Dumpy (Sakaidani et al. 2011), regulation of gastrulation in mouse embryos by O-glucosylation of CRB2 (Ramkumar et al. 2015), regulation of Drosophila photoreceptor morphogenesis by O-glucosylation of the Eyes shut protein (Haltom et al. 2014), regulation of acetylcholine receptor clustering by O-fucosylation of muscle agrin (Kim et al. 2008), and regulation of the SMAD3 protein stability by POGLUT1 independently of its role in Notch signaling (Xing et al. 2015). Therefore, as we continue to examine the consequences of glycosyltransferase mutations in model organisms and human patients, it is important to consider proteins other than Notch pathway components as potential biologically relevant targets of these fascinating enzymes.

Abbreviations

ADAM10

a disintegrin and metallopeptidase domain 10;

ALGS.

Alagille syndrome;

AOS,

Adams–Oliver syndrome;

BMDCs,

bone marrow-derived dendritic cells;

CRC,

colorectal cancer;

CRD,

carbohydrate recognition domain;

DLL,

Delta-like;

DSL,

Delta/Serrate/LAG-2;

EGF,

epidermal growth factor-like;

EOC,

epithelial ovarian cancer;

EOGT,

EGF domain-specific O-linked GlcNAc transferase;

ER,

endoplasmic reticulum;

FNG,

fringe;

GALNT11,

Polypeptide N-acetylgalactosaminyltransferase 11;

GalNAc,

N-acetylgalactosamine;

GlcNAc,

N-acetylglucosamine;

GXYLT,

glucoside xylosyltransferase;

HCC,

hepatocellular carcinoma;

hCLP46,

human CAP10-like protein 46 kilo-Dalton;

JAG,

jagged;

MGAT3,

mannoside acetylglucosaminyltransferase 3;

NECD,

Notch extracellular domain;

NICD,

Notch intracellular domain;

NRR,

negative regulatory region;

O-fut1,

O-fucosyltransferase 1 (Drosophila);

POFUT1,

protein O-fucosyltransferase 1;

POGLUT,

protein O-glucosyltransferase;

XXYLT,

xyloside xylosyltransferase.

Funding

This work was primarily funded by grants from the National Institutes of Health (R35 GM130317 and R01 GM084135).

Conflicts of interest statement

The authors declare no conflict of interest. The funders had no role in the design of the study, in the writing of the manuscript or in the decision to publish the manuscript.