Glycosylation and behavioral symptoms in neurological disorders

The complex cellular events that occur during the development of multicellular eukaryotes are coordinated by a relatively small number of evolutionarily conserved signalling pathways. One such pathway, the Notch pathway, controls a broad range of cell-fate decisions throughout the metazoa (reviewed in Ref. 1). These cell-fate decisions can be categorized at a cellular level into a relatively small number of distinct modes (reviewed in Ref. 2; Fig. 1). The best known of these is lateral inhibition, during which Notch signalling inhibits all but one of a group of equivalent precursor cells from adopting a particular fate -- for example, becoming a NEUROBLAST. Notch signalling also functions during other processes including asymmetric cell division, inductive signalling, and neurite extension and guidance (Fig. 1). Given the broad range of processes that require normal Notch signalling, it is not surprising that a number of human genetic diseases result from mutations in Notch-pathway components. These include T-cell leukaemia (caused by mutations in translocation-associated Notch homologue (TAN-1)) and a series of congenital diseases that are associated with defects in bone, vasculature or internal organs (Alagille syndrome, cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), and spondylocostal dystoses; reviewed in Ref. 3). Dysregulation of Notch signalling has also recently been implicated in the pathogenesis of multiple sclerosis, and in transformation by one of the most frequently mutated oncogenes in human cancers, RAS.

The core Notch signalling pathway takes a direct route from the cell membrane to the nucleus (Fig. 2; reviewed in Refs 1,6). The receptor for this pathway is Notch, a large single-pass transmembrane protein. There is only one Notch gene in Drosophila melanogaster, but mammalian genomes encode four Notch receptors (Table 1). The ligands for Notch are also transmembrane proteins, and in both arthropods and vertebrates they fall into two classes: Delta and Serrate (which is known as Jagged in mammals; Table 1). Binding of ligands to Notch receptors in neighbouring cells triggers proteolytic processing of Notch, which results in the release of the intracellular domain of Notch (N-IC) from the plasma membrane. N-IC then travels to the nucleus and associates with a CSL (CBF-1/Suppressor of Hairless/Lag-1) DNA-binding protein (Fig. 2). This association converts CSL-containing complexes from transcriptional repressors to transcriptional activators, and thereby changes the transcriptional programme of these cells. Superimposed on this relatively direct core Notch signalling pathway are a wide array of modulators of Notch signalling, most of which are specific for only a subset of Notch modes of action. Appreciation of the importance of glycosylation in regulating the Notch pathway stems from the determination that one such modulator, Fringe, is a GLYCOSYLTRANSFERASE.

In addition to the standard N-linked glycosylation that is commonly found on extracellular protein domains, Notch is also subject to two unusual types of O-linked glycosylation; the addition of O-linked glucose (O-glucose) and O-linked fucose (O-fucose; Fig. 3). Whereas the functional significance of O-glucosylation remains unknown (Box 1), studies of Fringe have established that elongated forms of O-fucose can modulate Notch signalling in certain contexts (reviewed in Refs 11-13). More recent studies have now also uncovered an essential role for the O-fucose monosaccharide in circumstances other than those that require only Fringe. Additional studies are continuing to provide insights into the nature, role, regulation and importance of O-fucose glycans in the Notch signalling pathway. In this review, we first describe our current understanding of the requirements for the different glycosyltransferases that participate in the synthesis of O-fucose glycans. We then consider the implications of recent investigations into the actual distribution of O-fucose glycans on Notch, and discuss possible mechanisms by which O-fucose glycans might exert their influences on the Notch signalling pathway.

O -glycosylation of EGF domains

O-glucose and O-fucose were first identified on epidermal growth factor (EGF) DOMAINS of serum glycoproteins that are involved in regulating blood clotting and FIBRINOLYSIS (reviewed in Ref. 18). The bulk of the extracellular domains of Notch and its ligands consists of a series of tandemly repeated EGF domains, and, 3 years ago, Moloney and colleagues reported that murine Notch1 is subject to both O-glucosylation and O-fucosylation. Both O-glucose and O-fucose can then be elongated by the sequential action of other glycosyltransferases to yield, respectively, a trisaccharide or a tetrasaccharide structure (Fig. 3). The actual structure of the O-glucose trisaccharide that is attached to Notch has not been determined, but by analogy with the O-glucose trisaccharide that has been identified on EGF domains of serum glycoproteins (Box 1), it is presumed to be Xylα1,3Xylα1,3Glc (where Xyl is xylose and Glc is glucose, in which glucose is attached to serine). The tetrasaccharide attached to Notch1 in Chinese hamster ovary (CHO) cells has been identified as Siaα2,3Galβ1,4GlcNAcβ1,3Fuc (where Sia is sialic acid (also known as N-acetylneuraminic acid), Gal is galactose, GlcNAc is N-acetylglucosamine and Fuc is fucose, in which fucose is attached to serine or threonine). Elongation of the O-fucose monosaccharide is dependent on Fringe, which is an EGF-O-fucose β1,3 N-acetylglucosaminyltransferase -- that is, an enzyme that catalyses the transfer of N-acetylglucosamine to fucose-O-EGF in a β1,3 linkage.

Cloning and characterization of an EGF- O -fucosyltransferase. An activity in extracts of CHO cells that can O-fucosylate EGF domains was first described by Spellman and colleagues. Purification of this enzyme, O-FucT-1, led to the cloning of a human gene that has been called protein O-fucosyltransferase 1 ( POFUT1 ), and that was shown to possess O-FucT-1 activity. Genomic analyses indicate that there is only a single POFUT1 gene in mammals, as well as single potential homologues in Caenorhabditis elegans and Drosophila. Biochemical studies have shown that O-FucT-1 requires a properly folded EGF domain as a substrate, and that O-fucosylation occurs at a single unique residue of an EGF domain, within a loose consensus sequence(Box 2).

The first indication that the O-fucose monosaccharide might be required for Notch signalling came from the observation that expression of a Notch-dependent reporter gene was reduced in cells that were deficient in the biosynthesis of GDP-fucose, the donor sugar for fucosyltransferases. The subsequent identification of the gene encoding O-FucT-1 (Ref. 22) then made it possible to determine the significance of O-fucose by genetic studies. A requirement for Ofut1 in Notch signalling was first shown in Drosophila using RNAi to reduce Ofut1 expression. More recently, mutations in Drosophila Ofut1 have been identified. Mutation or RNAi treatment of Ofut1 results in loss of Notch signalling, and the phenotypes can be as severe as those observed in the complete absence of the Notch receptor. Characterization of the requirements for Ofut1 in different tissues has shown that Ofut1 is broadly required for Notch signalling in Drosophila. Moreover, it is required for each of the main modes of Notch signalling, including lateral inhibition, cell-lineage decisions and inductive signalling.

In mice, a targeted mutation of the Pofut1 gene results in embryonic lethality, with defects in somitogenesis, cardiogenesis, vasculogenesis and neurogenesis. This phenotype resembles that observed in mice that are mutant for the only murine CSL gene, or mice that are double mutants for the two murine presenilin genes, which are required for the proteolytic processing associated with Notch activation (Fig. 2). So, the absence of O-FucT-1 in mice also results in phenotypes that resemble the complete absence of Notch signalling. Notably, this phenotype is more severe than that observed in mice that are mutant for individual Notch genes, which implies that O-fucose is required on all Notch receptors.

All of the phenotypes that have so far been characterized in fucosyltransferase mutants -- including Pofut1-mutant mice, flies that are mutant for Ofut1 or flies that have been treated with Ofut1 RNAi -- can be explained by the loss of Notch signalling. The absence of other phenotypes is surprising, as both the fly and mouse genomes encode other proteins with EGF domains that include the consensus sequence for O-fucosylation (reviewed in Ref. 13). In some cases, additional requirements for O-fucose might be masked by the severe developmental defects that are associated with inactivation of the Notch pathway. However, mutation of murine Cripto -- an EGF-CFC (for Cripto, Frl-1 and Cryptic) gene -- results in embryonic phenotypes at earlier stages than those observed in Pofut1-mutant mice. Cripto is O-fucosylated, and mutagenesis of the attachment site had indicated that O-fucosylation was required for its activity. The reason for this discrepancy is uncertain, and further studies will be required to determine the other processes that are influenced by O-fucosylation.

Regardless of the uncertainty surrounding other potential requirements for O-fucose, the results from studies in both flies and mice indicate that O-fucose is positively required for most or all Notch signalling. Therefore, the enzyme, O-FucT-1, that catalyses O-fucose synthesis can be considered to be a core component of the Notch signalling pathway. By contrast, studies of Fringe (see below) indicate that elongated forms of O-fucose are only required in certain contexts. Interestingly, C. elegans has a candidate O-FucT-1 gene, but does not seem to encode a Fringe gene. So, the modulatory effects of O-fucose elongation seem to be a more recent modification of an ancient requirement for O-fucose in Notch signalling.

O -FucT-1 expression and Notch regulation

The determination that O-FucT-1 is required for Notch signalling raises the question of whether this enzyme is simply a permissive factor that is required for Notch activation, or instead whether regulation of its distribution contributes to the pattern of Notch activation. Although both the Drosophila Ofut1 gene and the murine Pofut1 gene are broadly expressed, in situ hybridization and northern blotting experiments show that the levels of messenger RNA are regulated both spatially and temporally. Intriguingly, overexpression of Ofut1 in Drosophila can actually inhibit the lateral inhibition and inductive modes of Notch signalling. Elevated expression of OFUT1 has been reported to exert both positive and negative effects on Notch-ligand binding. Although the biological significance of these effects remains to be determined, the observation that either too little or too much O-FucT-1 can alter normal Notch activation indicates that the observed transcriptional regulation of Ofut1 is likely to influence Notch signalling.

Fringe elongates EGF O -fucose

The fringe gene was first identified for its mutant phenotypes rather than its enzymatic activity. Indeed, it was originally suggested that Fringe might function as a secreted molecule, and some authors have reported activities for secreted Fringe proteins, or have suggested that Fringe might influence Notch signalling by physically associating with it, rather than by enzymatically modifying it. However, a weak similarity in protein sequence between Fringe and a bacterial galactosyltransferase, lex-1, was uncovered by bioinformatic analysis. This observation, combined with the determination that fringe influences the Notch signalling pathway, and the finding that Notch contains consensus sequences for O-glycan modifications, raised the possibility that Fringe might be involved in the synthesis of O-glycans on Notch. This possibility was subsequently confirmed by both cell culture and in vitro experiments. Lec1 CHO cells (derivatives of CHO cells that are defective in N-LINKED GLYCAN biosynthesis and are commonly used to facilitate characterization of O-LINKED GLYCANS) normally express only low levels of Fringe, and the O-fucose on Notch is mostly in the monosaccharide form in these cells. However, when Lec1 CHO cells are transfected with a Fringe expression plasmid, there is a shift in the profile of O-fucose glycans on Notch. Biochemical characterization of purified Fringe proteins from both flies and mice confirmed that Fringe catalyses the second step in the synthesis of the O-fucose tetrasaccharide that is found on EGF domains -- it transfers GlcNAc onto fucose in a β1,3 linkage(Fig. 3). As Fringe uses fucose-O-EGF as a substrate, prior fucosylation of the EGF domain by O-FucT-1 is a prerequisite for Fringe activity (Fig. 3). Like O-FucT-1, Fringe has a strong preference for a properly folded EGF domain as a substrate, and can glycosylate EGF domains on both Notch and its ligands.

Several observations indicate that the glycosyltransferase activity of Fringe is essential for its ability to influence Notch signalling. Fringe contains a highly conserved DXD motif that in other glycoslytransferases has been shown to be essential for enzymatic activity. Mutation of this motif in Fringe eliminates the enzymatic activity of Fringe in vitro, and eliminates the biological activity of Fringe in vivo and in cell-culture assays. Two additional point mutations have been identified in other sequence motifs of Fringe that are shared among a superfamily of known and putative UDP-sugar β1,3-glycosyltransferases. In cultured-cell assays (Fig. 4), the ability of Fringe to influence Notch signalling is impaired in cells that are deficient in the synthesis of fucose-containing glycans, but not in cells that are deficient in the synthesis of normal N-linked glycans, which is consistent with a requirement for the Fringe acceptor substrate, O-linked fucose. Finally, mutation of the gene fringe connection , which encodes a nucleoside sugar transporter that can transport the Fringe donor sugar, UDP-GlcNAc (Box 3), produces the same phenotype that is seen in fringe mutants.

Fringe influences Notch-receptor activation

The requirement for Fringe in Notch signalling was first shown by studies of Drosophila wing development. In the wing, Fringe potentiates the ability of Delta to activate Notch, and at the same time inhibits the ability of Serrate to activate Notch(Fig. 5). Subsequent studies have shown that Fringe has essential functions in regulating Notch signalling in a number of different Drosophila tissues, including eyes, legs and ovaries. The requirements for fringe in these other tissues are consistent with the general conclusion that, in Drosophila, it acts as a positive regulator of Delta signalling and as a negative regulator of Serrate signalling.

In vertebrates, the influence of Fringe on Notch signalling is complicated by the existence of several Notch proteins, several Notch ligands and several Fringe proteins (Table 1). In cultured-cell assays (Fig. 4), Lunatic Fringe (L-Fng), has been reported to inhibit signalling by Jagged1 to Notch1 (Refs 10,25,53), and to potentiate signalling by Delta1 to Notch1 (Ref. 53), which parallels the situation in Drosophila. However, L-Fng has been reported to potentiate signalling by Delta1 and by Jagged1 to the Notch2 receptor. By contrast, a different study reported that L-Fng inhibited signalling by Jagged1 to Notch2, and had no effect on Delta1 signalling to Notch2 (Ref. 54). Although it is not yet clear whether the conflicting results are due to the different cell lines or different expression constructs used, or to differences in the design of the assays, the combined results do indicate that the influence of Fringe on Notch signalling in vertebrates might depend not only on the ligand, but also on the receptor. The influence of Manic Fringe (M-Fng) has been reported to be qualitatively similar to that of L-Fng, although, in some cases, differences in the strength of the effects are observed. The influence of Radical Fringe (R-Fng) has yet to be tested in these cultured-cell assays.

Fringe modulates a subset of Notch modes

In contrast to the broad requirement for O-FucT-1 in Notch signalling, Fringe is required in only a subset of Notch signalling processes. In Drosophila, the contexts in which Fringe is required can generally be classified as inductive signalling events (Fig. 1c). In the Drosophila wing, where both Serrate and Delta are expressed, the opposing effects of Fringe on signalling by these two Notch ligands help to position a stripe of Notch activation along the edge of fringe expression (Fig. 5). This stripe of activation is then essential for the growth, patterning and compartmentalization of the wing (reviewed in Refs 55,56).

In contrast to the requirements for Fringe during inductive signalling processes in Drosophila, little or no requirement for Fringe is observed during lateral inhibition or asymmetric cell divisions. Even ectopic expression of Fringe has only very mild effects on these processes (V. Panin, K. Consolino and K.D.I, unpublished observations). Serrate and Delta have redundant functions during the asymmetric cell divisions of sensory-organ cells, and in this context it might be that Fringe has little effect because its opposing effects on signalling by the two ligands cancel each other out. However, Delta is the only Notch ligand that is required during lateral inhibition and during oogenesis in Drosophila. Fringe has very little influence on lateral inhibition, but it is required for normal Notch activation during oogenesis.

In vertebrates, ectopic-expression experiments have shown that Fringe can influence the development of limbs, teeth, the neural crest, haematopoietic cells and the forebrain. However, the existence of three mammalian Fringe genes complicates genetic analysis of their normal requirements, and the extent to which Fringe participates in different modes of Notch signalling remains unclear. Gene-targeted mutations have been reported for L-Fng and R-Fng, but only L-Fng mutations result in a visible phenotype. L-Fng is expressed in a complex spatial and temporal pattern during somitogenesis, and is required, along with other components of the Notch pathway, for the subdivision of the PRESOMITIC MESODERM into somites (see Box 4).

Common to all of the situations in which Fringe functions is that differential glycosylation of Notch effectively creates different forms of the receptor, which differ in their responsiveness to Notch ligands. The distribution of these different forms of Notch is determined by the transcriptional regulation of fringe genes. Strikingly, in many cases it is the distribution of this glycosyltransferase, rather than of the Notch receptors or their ligands, that is the key determinant of where Notch signalling will be active.

Elongation of the GlcNAc-Fuc O -glycan

Subsequent to the action of Fringe, the GlcNAc-Fuc disaccharide on Notch can be extended in CHO cells by a β1,4 galactosyltransferase (β4GalT) and a sialyltransferase to yield the tetrasaccharide Siaα2,3Galβ1,4GlcNAcβ1,3Fuc (Ref. 19; Fig. 3). This raises the question of what the relevant glycan is for the modulation of Notch signalling by Fringe. Experiments to determine the minimal structure that is required for the modulatory effect of Fringe on Notch signalling have been carried out by taking advantage of mutants derived from CHO cells that are defective in specific steps in glycan biosynthesis, together with a Notch signalling assay that involves co-culture of ligand-expressing cells with Notch-expressing cells (Fig. 4). So far, the effects of glycan elongation have only been assessed for Jagged1-induced Notch signalling. In CHO cells that are defective in galactose addition (Lec20 or Lec8 CHO cells), Fringe is unable to inhibit Jagged1 to Notch1 signalling. Importantly, this effect can be rescued in Lec20 CHO cells by transfection with β4GalT-1, which implicates this enzyme as a key effector. As Notch signalling is unaffected in Lec1 CHO cells, which are defective in complex and hybrid N-glycan (the mature forms of N-glycans, which are galactosylated) biosynthesis, the requirement for β4GalT-1 must be related to O-glycan or glycolipid biosynthesis. Given the known structure of the O-fucose glycan, it seems most likely that the effect is directly attributable to the inability to elongate the GlcNAc-Fuc disaccharide (Fig. 3). In contrast to the requirement for galactose, cell lines that are defective in the addition of sialic acid are still sensitive to Fringe. Together, these observations imply that the minimal structure required for inhibition of Jagged1 to Notch1 signalling is the trisaccharide Galβ1,4GlcNAcβ1,3Fuc. In the future, it will be important to confirm these results in whole animals, and to extend them to the evaluation of the sugars that are required for the influence of Fringe on Delta signalling.

Although murine β4GalT-1 is dispensable for embryonic development, mammalian genomes encode seven members of the β4GalT family, six of which can transfer Gal to GlcNAc (reviewed in Ref. 69). So, a requirement for elongation of the GlcNAc-Fuc disaccharide in mice might be masked by genetic redundancy. The Drosophila genome encodes three members of the β4GalT family, two of which (CG14517 and CG8536) are predicted by sequence to transfer Gal or GalNAc to GlcNAc, and one of which (CG11780/β4GalT7) uses principally xylose as an acceptor and participates in proteoglycan biosynthesis. It is not yet known which, if any, of these Drosophila β4GalTs participate in elongation of the GlcNAc-Fuc disaccharide, nor whether they are required for Fringe-dependent modulation of Notch signalling.

Relevant sites for O -fucosylation

Ligand or receptor? As O-fucose glycans are found on both Notch and its ligands, in principle, glycosylation of either substrate could be responsible for the influences of Fringe and O-FucT-1 on Notch signalling. However, genetic studies in Drosophila have shown that the normal requirements for both O-FucT-1 and Fringe in Notch signalling are CELL AUTONOMOUS and occur in the cell that receives the signal, which indicates that Notch, rather than Notch ligands, is probably the biologically relevant target. Similarly, Notch signalling assays in cultured mammalian cells have shown that Fringe functions in cells that receive Notch signals, and not in the cells from which the signals originate. Moreover, Notch-ligand binding experiments (see below) implicate modification of Notch rather than its ligands in the modulation of receptor-ligand interactions. So, although the modification of ligands is intriguing, and the presence of potential O-fucosylation sites is highly conserved, so far there is no direct evidence for a functionally relevant modification of the Notch ligands.

Potential variations in Notch O-fucosylation. Although Notch proteins might be the only biologically relevant target of O-fucosylation in the Notch pathway, the current consensus sequence for O-fucosylation (Box 2) predicts that between one-half and two-thirds of the EGF domains in most Notch receptors could be modified with O-fucose (Figs 6,7). In fact, recent studies of Notch1 O-fucosylation in Lec1 CHO cells confirm that a large number of potential O-fucosylation sites are used, and that these are distributed throughout the extracellular domain of Notch1 (Ref. 24). As both O-FucT-1 and Fringe recognize a folded EGF domain as a substrate, it would not be surprising if different EGF domains differ in the efficiency with which they are glycosylated. Together with variations in the levels of O-FucT-1 and Fringe expression, this could potentially result in an additional layer of Notch regulation -- developmentally regulated differences in the levels of expression of these enzymes could result in differences in the profile of O-fucose glycans on Notch. Although the extent of O-fucosylation of individual EGF domains has not yet been determined, differences in the efficiency of elongation of different EGF domains by a single Fringe have been observed. Moreover, L-Fng and M-Fng can differ in their ability to glycosylate different EGF repeats. Such differences in the efficiency of O-fucosylation and elongation of different EGF repeats have the potential to generate many different forms of Notch, and it is tempting to speculate that this could be used as a mechanism for fine-tuning the responsiveness of Notch receptors to their ligands.

Biologically relevant sites. The large number of potential sites of O-fucosylation poses a challenge to the characterization of biologically relevant sites of modification. Two regions of Notch have been proposed to be potentially important sites of fucosylation on the basis of prior genetic and structure-function studies. EGF repeats 11 and 12 of Drosophila Notch are necessary and can be sufficient for binding to ligands in a cell-aggregation assay. Intriguingly, the presence of a potential O-fucosylation site in EGF repeat 12 (EGF12) is conserved in all Notch receptors (Fig. 6), and direct assessment of its glycosylation status confirms that EGF12 is a substrate for both O-FucT-1 and Fringe.

The isolation of a missense mutation in EGF12 of Drosophila Notch that results in a strong loss-of-function phenotype (N) is consistent with its apparent importance in ligand binding. By contrast, point mutations that change specific amino acids in EGF24, 25, 27 or 29 of Drosophila Notch result in a gain-of-function phenotype, which reflects an increased sensitivity of Notch to its ligands. The isolation of these Abruptex mutations (N) indicates that EGF24-EGF29 might inhibit Notch activation. Intriguingly, this overlaps with an array of predicted sites of O-fucosylation (Figs 6,7), at least several of which can be fucosylated in cell culture and in vitro. Genetic interactions between N alleles and mutations in fringe and fringe connection raised the possibility that N mutations influence Notch by affecting its biochemical interactions with Fringe. However, Shao et al. have tested this hypothesis by introducing N mutations into EGF repeats of mouse Notch1. Although one N mutation inhibited O-fucosylation, another did not. So, whereas an influence on Fringe-dependent elongation of O-fucose could contribute to the N phenotype in some cases, an influence on O-fucose glycans does not seem likely as a general explanation for how these mutations affect Notch signalling.

In contrast to those EGF repeats that have conserved O-fucose sites, at other EGF repeats it is the absence of a potential O-fucose site that is conserved (Fig. 6), which indicates that it might be important for some EGF domains to remain free of O-fucose. Interestingly, in one such repeat, EGF14, a point mutation, split (N), that elevates Notch activity in certain contexts has been isolated. This mutation results in the creation of a potential O-fucose modification site, and in vitro studies confirm that the N mutation makes EGF14 a substrate for O-FucT-1 and Fringe. Intriguingly, the N mutant phenotype can be suppressed by a reduction of scabrous levels. The scabrous gene encodes a secreted protein with a domain that is related to fibrinogen, and normally functions during certain lateral inhibition processes, but can also inhibit Notch signalling in the Drosophila wing when it is expressed ectopically. Scabrous has been shown to physically associate with a complex that includes Notch. The portion of Notch that is required for this physical association maps to EGF19-EGF26, and so overlaps with an array of potential O-fucose sites between EGF20 and EGF31 (Figs 6,7). Together, these observations support a model in which the normal association of Scabrous with Notch depends on O-fucosylation of EGF20-EGF26, and inappropriate association of Scabrous with O-fucosylated EGF14 results in abnormal Notch activity.

What steps are affected by O -fucosylation?

EPISTASIS experiments in flies and in cultured mammalian cells indicate that both O-FucT-1 and Fringe function upstream of the activated (cleaved) form of the Notch receptor. This positions the requirement for O-fucose glycans somewhere upstream of the cleavages that liberate Notch from the plasma membrane (Fig. 2). In principle, several possibilities exist for the step or steps that are influenced by O-fucosylation (reviewed in Refs 12,13), but the simplest is a direct influence on Notch-ligand binding affinities.

Notch-ligand binding. Genetic results that are consistent with an influence of Fringe on Notch-ligand binding were initially obtained by Fleming and colleagues, who observed that when the amino-terminal domain of Serrate is replaced with that of Delta, the ability of Fringe to inhibit Serrate signalling is abrogated. As this amino-terminal region is essential for binding to Notch, the co-localization of a region that is required for Fringe inhibition is consistent with Fringe affecting Notch-Serrate binding. In addition, a secreted, dominant-negative form of Serrate, which inhibits Notch signalling, is also inhibited by Fringe. The observation that Fringe both blocks the ability of the wild-type (agonist) form of Serrate to activate Notch, and blocks the ability of a dominant-negative (antagonist) form of Serrate to inhibit Notch activation, argues for an effect at the ligand-receptor binding step, and against an influence on downstream processing events.

Notch and its ligands are normally transmembrane proteins, and initial investigations of Notch-ligand binding took advantage of the observation that the affinity between ligand and receptor mediates the aggregation of cultured cells. Initial experiments to confirm an influence of Fringe on Notch-ligand binding using this assay were unsuccessful (V. Panin and K.D.I., unpublished observations). More recently, the ability of soluble ligands or receptors to bind to cultured cells has been assayed. Evidence in favour of an influence of Fringe on Notch-ligand binding came first from the observation that Fringe can enhance binding between a soluble, tagged form of Notch or Delta to membrane-bound Delta or Notch, respectively, expressed on the surface of cultured Drosophila cells. More recent experiments have shown that Fringe can inhibit Serrate-Notch binding, and that OFUT1 is required for both Serrate and Delta binding to Notch (Refs 16,17). Binding is influenced by the expression of Fringe and OFUT1 in the Notch-expressing cell, and not the ligand-expressing cell, which is consistent with these enzymes glycosylating Notch. In mammals, evidence in favour of an influence of Fringe on Notch-ligand binding has come from the observation that M-Fng or L-Fng could decrease the binding of a soluble form of Jagged1 to cells that express Notch2 (Ref. 54). However, in a separate study, no inhibition of Jagged1-Notch binding by L-Fng was detected, even though the same authors observed that L-Fng inhibited Jagged1 signalling to Notch1 (Ref. 53).

Other effects of O-fucose glycans on the Notch pathway? Although the basis for the discrepancies in the mammalian binding studies has yet to be resolved, our present view is that O-fucose glycans do influence Notch-ligand binding, but this does not exclude the possibility that they affect other steps as well. For example, O-fucose might affect both ligand-receptor binding affinities, and interactions with other proteins, such as Scabrous. Scabrous has recently been found to require an endosomal protein Gp150 for its activity, and to co-localize with Gp150 in endosomes, which raises the possibility that O-fucose might also affect intracellular trafficking. The possibility exists that several partners interact with O-fucose glycans that are attached to different EGF repeats of Notch and its ligands to influence different steps in Notch regulation.

How do sugars affect Notch activation?

Regardless of the step or steps that are affected by O-fucose glycans, two basic classes of models can be proposed for how attachment of sugars influences the interactions of Notch with other proteins (Fig. 7).

Conformational change. In 'conformational change' models, O-fucose glycans influence the conformation of the Notch extracellular domain. On the basis of the observation that the structure of a single EGF domain from blood coagulation factor VII is not significantly affected by O-fucosylation, it is unlikely that O-fucose glycans affect the structure of individual EGF domains of Notch. However, it is possible that O-fucose glycans could affect interactions or folding between EGF domains (Fig. 7).

Recognition site. In 'recognition site' models, O-fucose glycans could create or obscure recognition sites for interaction with other proteins. Proteins that bind carbohydrates without modifying them are classified as LECTINS. A large number of animal lectins have been identified, and these participate in processes as diverse as the immune response, cell signalling, cell adhesion and protein folding. The addition of O-fucose might create recognition sites for a fucose-binding lectin. Elongation of O-fucose by Fringe might then modulate Notch signalling, either by covering up binding sites for fucose-binding lectins, or by creating binding sites for lectins that recognize distinct glycans, such as Galβ1,4GlcNAc. In its simplest form, this model implies that the Notch ligands could be lectins, with their binding to Notch influenced by direct recognition of O-fucose glycans (Fig. 7). Alternatively, O-fucose glycans could be recognized by as-yet-unidentified lectins that function as cofactors, which then mediate the interactions of Notch with its ligands.

One appealing feature of models that put forward lectin-based recognition of O-fucose glycans is that they propose an explanation as to why Notch has so many EGF repeats. Biophysical studies of protein-sugar binding have generally found that binding to individual sugars is a low-affinity event, and most lectins achieve a high avidity by forming multivalant attachments (reviewed in Ref. 95). Although only small regions of Notch and its ligands are essential for ligand binding in assays carried out in vitro , achieving the binding that is required to activate Notch in vivo might nonetheless require extensive arrays of O-fucosylated EGF repeats.

Future directions

Although it is now clear that O-fucose glycans are essential for the Notch signalling pathway, we still do not fully understand how they influence Notch activation. Moreover, it remains unclear how the elongated forms of O-fucose influence Notch activation differently from the monosaccharide, and why some Notch signalling processes are sensitive to this Fringe-dependent elongation, whereas others are insensitive. It also remains an open question as to how O-fucose glycans can be dynamically regulated -- are there glycosidases that can trim off sugars on cell-surface Notch, thereby providing another layer of regulation? And what role do the O-glucose glycans have in Notch signalling? In addition to providing a more complete understanding of how glycans regulate Notch signalling, increased understanding of this post-translational modification might also help to identify other examples of protein regulation by glycosylation.

Traditionally, most molecular biologists have ignored glycosylation, or, at best, treated it as a nuisance. However, the demonstration that O-fucosylation is central to the activation of Notch by its ligands, together with the realization that HEPARIN SULPHATE PROTEOGLYCANS also have essential roles in several distinct signalling pathways (reviewed in Ref. 96), has helped to emphasize the important roles of glycosylation in regulating protein function. With whole-genome sequences facilitating the identification of new glycosyltransferases, we anticipate that other examples of important roles for glycosylation will be forthcoming.