Notch Ligands Are Substrates for ProteinO-Fucosyltransferase-1 and Fringe*

O-Fucose has been identified on epidermal growth factor-like (EGF) repeats of Notch, and elongation of O-fucose has been implicated in the modulation of Notch signaling by Fringe. O-Fucose modifications are also predicted to occur on Notch ligands based on the presence of the C2 XXGG(S/T)C3 consensus site (where S/T is the modified amino acid) in a number of the EGF repeats of these proteins. Here we establish that both mammalian andDrosophila Notch ligands are modified withO-fucose glycans, demonstrating that the consensus site was useful for making predictions. The presence of O-fucose on Notch ligands raised the question of whether Fringe, anO-fucose specific β1,3-N-acetylglucosaminyltransferase, was capable of modifying O-fucose on the ligands. Indeed,O-fucose on mammalian Delta1 and Jagged1 can be elongated with Manic Fringe in vivo, and Drosophila Delta and Serrate are substrates for Drosophila Fringe in vitro. These results raise the interesting possibility that alteration of O-fucose glycans on Notch ligands could play a role in the mechanism of Fringe action on Notch signaling. As an initial step to begin addressing the role of the O-fucose glycans on Notch ligands in Notch signaling, a number of mutations in predicted O-fucose glycosylation sites onDrosophila Serrate have been generated. Interestingly, analysis of these mutants has revealed that O-fucose modifications occur on some EGF repeats not predicted by the C2 XXGGS/TC3 consensus site. A revised, broad consensus site, C2 X 3–5S/TC3 (whereX 3–5 are any 3–5 amino acid residues), is proposed.

Notch proteins are single-pass transmembrane receptors that control a broad spectrum of cell fate decisions (1). Deregulation of the Notch signaling pathway has been implicated in various human diseases, including CADASIL, T-cell leukemia, spondylocostal dystosis, and Alagille syndrome (2)(3)(4). Ligands for Notch can be divided into two classes, Delta and Serrate/ Jagged. All Notch ligands are transmembrane proteins that share common structural features including a DSL domain, required for Notch binding, and multiple epidermal growth factor-like (EGF) 1 repeats in their extracellular domains (5). Serrate/Jagged ligands are distinguished structurally from Delta ligands by a greater number of EGF repeats and the presence of a distinct extracellular cysteine-rich region near the transmembrane domain. Ligand binding triggers proteolytic processing events that result in release of the intracellular domain of Notch from the membrane and subsequent translocation to the nucleus, where it participates in a transcriptional activator complex. Drosophila has only one Delta and one Serrate gene, but mammals have three Deltas and two Serrate/Jaggeds.
During Drosophila development, Delta and Serrate are distinguished functionally by their sensitivity to Fringe (reviewed in Ref. 6). Fringe is a ␤1,3-N-acetylglucosaminyltransferase that elongates O-fucose on EGF repeats and acts as a key modulator of Notch signaling (7)(8)(9). Expression of Fringe inhibits the activation of Notch by Serrate and, at the same time, potentiates the activation of Notch by Delta (10,11). Three Fringe genes have been identified in mammals: Lunatic Fringe, Manic Fringe (MFng), and Radical Fringe (12,13). Mammalian Fringe genes have also been shown to be able to influence the activation of Notch by its ligands, and as in Drosophila different mammalian ligands exhibit distinct sensitivities to Fringe (14,15).
Notch contains multiple EGF repeats within its extracellular domain that are modified by O-fucose and are substrates for the glycosyltransferase activity of Fringe (7,8,16). The importance of the glycosyltransferase activity of Fringe in the modulation of Notch signaling has been demonstrated both by genetic studies in Drosophila and by Notch signaling assays in mammalian cells deficient in glycan biosynthesis (7)(8)(9)17). Although Notch is a substrate of Fringe, it has not yet been demonstrated that glycosylation of Notch is actually sufficient to account for the influence of Fringe on Notch signaling.
O-Fucose is a modification of serine or threonine residues that has typically been found to occur on EGF repeats containing the amino acid consensus sequence C 2 XXGG(S/T)C 3 , where C 2 and C 3 are the second and third conserved cysteines of an EGF repeat (18). The enzyme responsible for addition of O-fucose to these sites, protein O-fucosyltransferase-1 (O-FucT-1), has recently been purified and cloned (19,20). O-Fucose in two contexts other than EGF repeats has also been described: one in which O-fucose is attached to a threonine within a thrombospondin repeat (21), and another where a O-fucose is attached to a peptide (PMP-C) isolated from Locusts (22,23). O-Fucose has been thought to be a rare modification, and to date has only been experimentally demonstrated to occur on 11 different proteins, of 6 main types (Table I). However, the C 2 XXGG(S/T)C 3 consensus sequence exists in many other EGF repeat-containing proteins (16,24), including all of the known Notch ligands. This raises the possibility that O-fucosylation occurs on both the Notch receptor and its ligands, and that modification of O-fucose saccharides on Notch ligands may contribute to the modulatory effects of Fringe on Notch signaling.
To begin to investigate these possibilities, we assayed both Drosophila and mammalian Notch ligands for the presence of O-fucose saccharides. Our results confirm that Notch ligands do possess O-fucose, and further show that this O-fucose can be elongated by the glycosyltransferase activity of Fringe. Moreover, we found that O-fucosylation can occur on EGF repeats that do not match the C 2 XXGG(S/T)C 3 consensus sequence for O-fucosylation. These observations suggest that the role of O-fucosylation in the modulation of Notch signaling and other processes is likely to be broader and more complex than suggested previously.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Drosophila Fringe:His 6 protein was expressed in Drosophila S2 cells and purified as described previously (8). Drosophila Delta and Serrate proteins were also expressed in S2 cells using the pRMHA-3 expression vector (25)(26)(27). After 21 h of induction with 0.7 mM CuSO 4 , cells were collected, washed in ice-cold phosphate-buffered saline, and lysed in L buffer (300 mM NaCl, 5 mM KCl, 50 mM Tris-HCl, pH 7.4, 0.1% Triton X-100, 0.05% Tween 20, 1.2 mM EDTA) with protease inhibitors mixture (Roche Molecular Biochemicals) by sonication. The cell lysate was precleared by centrifugation (4°C, 20,000 ϫ g for 30 min) and incubated with rat anti-Serrate (28) or mouse anti-Delta antibodies (Developmental Studies Hybridoma Bank, 1:1000 dilution) for 2 h with gentle agitation at 4°C. Proteinantibody complexes were then mixed with protein A/G beads (Amersham Biosciences) for 2 h with gentle agitation at 4°C. The beads were then precipitated by centrifugation at 2,000 rpm and then extensively washed with L buffer and equilibrated with G buffer (140 mM NaCl, 20 mM HEPES, pH 7.3, 10 mM MnCl 2 , 0.2% Triton X-100) for glycosyltransferase assay.
Notch-EGF25 Construct-EGF repeat 25 of Drosophila Notch was PCR amplified from the FLAG-ECN construct (8) using PCR primers (ccgtctagacctcatcgtttgtctgac and gtctagagagacggacatcaatgagtgcttg) containing an XbaI site. This PCR product was then used to substitute for Notch sequences in FLAG-ECN (8), resulting in the Notch-EGF25 expression construct. FLAG-tagged Notch-EGF25 was purified from Drosophila cell culture after metallothionein induction as described above, using FLAG affinity beads (Sigma). The beads with immobilized Notch-EGF25 repeat were washed as described above and used in in vitro Fringe labeling assays.
In Vitro Labeling of Serrate, Delta, and Notch EGF25 with Fringe-20 l of Protein A/G beads in 60 l of G buffer, with immobilized Serrate, Delta proteins, or FLAG affinity beads with EGF25, were incubated with 2 Ci of UDP-[ 3 H]GlcNAc (60 Ci/mmol, ARC, Inc.) and 0.5 g of purified Fringe:His 6 protein. After a 1-h incubation at 29°C, the beads were washed extensively with ice-cold G buffer and analyzed by scintillation counting, or they were boiled in PAGE protein buffer and analyzed by SDS-PAGE, Western blotting, and fluorography. For fluorography, we used an autoradiography enhancer solution (EN-HANCE, PerkinElmer Life Sciences) and followed the manufacturer's protocol. For Western blotting, we used guinea pig anti-Delta antibody (generously provided by Dr. Mark Muskavitch (29)) and rabbit anti-Serrate antibody (generously provided by Dr. Eli Knust (30)) for detection of Delta and Serrate, respectively.
Chromatographic Analysis of O-Linked Carbohydrates-This was essentially performed as described previously (8,16). Briefly, samples of purified Delta1-Fc or Jagged1-Fc were acetone-precipitated and subjected to alkali-induced ␤-elimination. After desalting, the released O-fucose glycans were subjected to gel filtration analysis on a Pharmacia Superdex peptide column. The column was calibrated with hydrolyzed glucose polymers and authentic tetrasaccharide (Sia␣2, 3Gal␤1,4GlcNAc␤1,3fucitol) and monosaccharide (fucitol). The structure of the tetrasaccharide species was confirmed using exoglycosidase digestions as described previously (16). After digestion with sialidase and ␤-galactosidase, the resulting disaccharide structure was confirmed by high pH anion exchange chromatography (HPAEC) on a Dionex MA1 column as described (16). Elution time of samples was monitored relative to the authentic standards: GlcNAc␤1,2fucitol, GlcNAc␤1,3fucitol, and GlcNAc␤1,4fucitol (standards generously provided by Dr. Khushi Matta, Roswell Park Memorial Cancer Institute, Buffalo, NY).
Site-specific Mutagenesis of Drosophila Serrate-Site-specific mutagenesis was performed by PCR essentially as described in Ref. 31. All mutagenesis was designed to make the most conservative possible amino acid substitutions that would prevent O-fucosylation: either Ser to Ala or Thr to Val. These changes are unlikely to cause gross disruption of EGF structure because there exist endogenous repeats in Notch and its ligands that contain Ala or Val before the third cysteine. Beginning with a plasmid containing full-length, wild-type SER as a template, Ser-mtn (27), 18 thermal cycles were conducted with complementary pairs of oligonucleotides containing single base mismatches that encoded the desired change. The reaction mixture included 0.3 g of template DNA, 50 pmol of each mutagenesis primer, and 5 units of Herculase (Stratagene), in a volume of 50 l. Subsequent rounds of mutagenesis were then conducted using the product of the previous round as a template, in the following order and with the indicated primers: EGF 3 (aagcacggcggcgtctgcgaaaataccgc, gcggtattttcgcagacgccgccgtgctt), EGF 5 (gagcatggtggcgtttgcatcgatctaat, attagatcgatgcaaacgccaccatgctc), EGF 12 (cagaatggtggtgtctgcatgcctggagc, gctccaggcatgcagacaccaccattctg), EGF 13 (cacaatggcggagtctgcgagtcgggagc, gctcccgactcgcagactccgccattgtg), EGF 4 (cgcaacggcggcgtctgcacactcaagac, gtcttgagtgtgcagacgccgccgttgcg), EGF 7 (cggaatggagccgtctgcattgatctggt, accagatcaatgcagacggctccattccg), EGF 14 (cagggcggtgccgtctgcatcgacggaat, attccgtcgatgcagacggcaccgccctg), and EGF 2 (tgcaagcatggtgcctgcaacggcag, ctgccgttgcaggcaccatgcttgca). Each mutation was confirmed by DNA sequencing.

RESULTS
Sequence analysis reveals the presence of multiple C 2 XXGG(S/T)C 3 consensus sites for O-fucosylation in EGF repeats of all Notch ligands ( Fig. 1, red circles). The published consensus sequence for O-fucosylation was derived from sequences of only five EGF repeats in proteins experimentally tested for the presence of O-fucose (18). Nonetheless, this sequence has been used to accurately predict the presence of O-fucose on both Notch (16) and Cripto (32,33). Here we examine whether the predicted O-fucose sites within individual EGF repeats of Notch ligands were also functional. If Notch ligands were modified by O-fucose, then they would also become prospective targets for the ␤1,3-N-acetylglycosaminyltransferase activity of Fringe. However, the C 2 XXGG(S/T)C 3 consensus sequence was developed based on analysis of O-fucosylation, and the extent to which O-fucose monosaccharides on particular EGF repeats were further elongated by Fringe has not yet been determined. Preliminary indications suggest that Fringe elongates O-fucose on some EGF repeats, but not others (15). 2 Interrupted EGF repeats are shown as broken circles. sup5 indicates the location of point mutation (Gly to Arg) in Dl sup5 . Protein sequence accession numbers are as follows: , AAB84215 (human Jagged2), AAC52946 (rat Jagged-2), BAA21713 (chicken Serrate2), AAL08214 (zebrafish Jagged2), AAL08213 (zebrafish Jagged1), AAL08216 (zebrafish Jagged3), BAB59049 (Xenopus Serrate1), AAB06509 (Rat Jagged1), AAF15505 (mouse Jagged1), AAC51731 (Human Jagged1), CAA64604 (chicken Serrate1).
were transiently expressed in Lec1 Chinese hamster ovary cells. Lec1 cells are unable to synthesize complex or hybridtype N-linked glycans and were used in these experiments to facilitate metabolic radiolabeling of O-fucose saccharides (8). Subsequent to metabolic labeling with [ 3 H]fucose, Delta1-Fc and Jagged1-Fc proteins were purified from the cell medium and analyzed by SDS-PAGE and fluorography. Both proteins incorporated [ 3 H]fucose. Because Lec1 cells do not synthesize hybrid or complex-type N-glycans, these results strongly suggest the presence of O-fucose on both of these Notch ligands (Fig. 2, lanes 1 and 3). The presence of O-fucose was confirmed by gel-filtration chromatography of O-linked saccharides released from Delta1-Fc or Jagged1-Fc by alkali-induced ␤-elimination where the radiolabel was almost exclusively in the form of the monosaccharide, fucitol (Fig. 3, A and B).
To evaluate whether MFng was capable of modifying O-fucose residues on Delta1 or Jagged1, the same experiment was performed in Lec1 cells expressing MFng. The labeled, O-fucose saccharides released from Delta1 or Jagged1 isolated from MFng-expressing Lec1 cells (Fig. 2, lanes 2 and 4) contained a significant amount of tetrasaccharide (Fig. 3, C and  D). Some increase in the amount of di-and trisaccharide structures was also detected on Jagged1. The structure of the tetrasaccharide species (as well as the di-and trisaccharides) was confirmed by exoglycosidase sequencing with sialidase and ␤-galactosidase as described previously (16). Analysis of the resulting disaccharide by HPAEC showed it had the structure GlcNAc␤1,3fucitol (Fig. 3, E and F), confirming that MFng was forming the same structure on Delta1 and Jagged1 as was previously reported on Notch1 (8,16). Similar results were obtained when endogenous Jagged1 from Chinese hamster ovary cells was analyzed (data not shown). In summary, these ; Tri, Gal␤1,4GlcNAc-␤1,3Fucitol; Tetra, tetrasaccharide Sia-␣2,3Gal␤1,4GlcNAc␤1,3fucitol. E and F, tetra species from panels C and D were digested with sialidase and ␤-galactosidase as described (16). The resulting disaccharides were analyzed by HPAEC analysis. The elution position of standards is shown: 1, GlcNAc ␤1,2fucitiol; 2, GlcNAc␤1,3fucitiol; 3, GlcNAc␤1,4fucitiol. results indicate that O-fucose was present on the mammalian Notch ligands Delta1 and Jagged1 as a monosaccharide in Chinese hamster ovary cells that can be elongated to a disaccharide by the ␤1,3-N-acetylglucosaminyltransferase activity of Fringe, and then further elongated by galactosyl-and sialyltransferases endogenously expressed in Chinese hamster ovary cells.
Drosophila Delta and Serrate Are Fringe Substrates in Vitro-No analogue of Lec1 cells exists for cultured Drosophila cells. Hence, it is not possible to generate efficient metabolic radiolabeling of O-fucose saccharides. Thus, we took an alternate approach of expressing Notch ligands in cultured Drosophila cells, purifying them, and determining whether they serve as substrates for the glycosyltransferase activity of purified Fringe protein in vitro. In principle, such an experiment could support two important conclusions. First, because Fringe requires an O-fucose in the context of an EGF repeat for effi-cient glycosylation, the ability to serve as a Fringe substrate provides strong evidence that the protein was first modified by O-fucose. Second, in vitro labeling demonstrates that the protein was indeed a substrate for Fringe. To produce Notch ligand substrates for these assays, full-length, wild-type Serrate or Delta were expressed in Drosophila S2 cells under the control of a metallothionein promoter. These proteins were then immunoprecipitated using antibodies against the native proteins and protein A/G-agarose beads. In vitro glycosylation assays of Drosophila Delta and Serrate proteins were then performed using affinity purified Drosophila Fringe and UDP-[ 3 H]GlcNAc as a sugar donor. The amount of Serrate and Delta proteins was monitored by Western blotting (Fig.  4, A and B). The transfer of [ 3 H]GlcNAc was detected by scintillation counting (not shown), as well as by fluorography of SDS-PAGE separated proteins (Fig. 4, A and B). This analysis demonstrated that both of the Drosophila Notch ligands, Delta and Serrate, were substrates for the glycosyltransferase activity of Fringe in vitro. This also provided evidence that both Serrate and Delta were modified with O-fucose in Drosophila S2 cells. Conversely, proteins lacking EGF repeats, such as the IgGs used for the immunoprecipitation and also present in the labeling reaction, were not substrates for Fringe.
To establish the identity of the saccharide product of the Fringe reaction on Notch ligands, O-linked saccharides were released from Fringe-modified Serrate and Delta proteins and then analyzed by HPAEC, revealing a disaccharide product that co-migrated with a GlcNAc␤1,3fucitol standard (Fig. 5). This confirms that the substrate for the GlcNAc transferase activity of Fringe on Notch ligands was an O-fucose, and that Fringe specifically elongates this O-fucose with a ␤1,3-linked N-acetylglucosamine.

Sites of O-Fucose Elongation by Fringe on Drosophila Serrate Protein and EGF Repeat 25 of Notch-
The large number of potential sites for O-fucosylation of Notch and its ligands presents a challenge to the identification of the biologically relevant sites of Fringe action in its modulation of Notch signaling (Fig. 1). To begin to address this question, it was first necessary to identify which O-fucose sites could actually be elongated by Fringe. We therefore conducted site-specific mutagenesis of Serrate, which contains only four intact EGF repeats that match the C 2 XXGG(S/T)C 3 consensus (in EGF repeats 3, 5, 12, and 13) (Figs. 6 and 7A). Four rounds of mutagenesis were employed to make the amino acid substitutions that would prevent O-fucosylation. Ser-4m, the Serrate protein with all four O-fucosylation consensus mutated (EGFs 3, 5, 12, and 13) (Fig. 7A), was expressed in S2 cells and isolated by immunoprecipitation as for wild-type Serrate. Strikingly, this mutant form of Serrate was nonetheless a good substrate for Fringe in vitro (Fig. 7B, lane 2).
Although Ser-4m lacks intact EGF repeats containing the C 2 XXGG(S/T)C 3 consensus, it does contain one more EGF repeat that has this consensus, EGF 4. However, this repeat is interrupted by a 62-amino acid insertion between the third (Cys 3 ) and fourth (Cys 4 ) cysteines (Fig. 6). Thus, it is not clear if EGF 4 folds into a typical EGF structure, thus providing the additional target for O-FucT-1 and Fringe in the Ser-4m mutant protein. Alternatively (or additionally), some nonconsensus sites may be modified on Serrate protein.
Studies of O-FucT-1 activity using site-directed mutants of a human factor IX EGF indicated that the substitution of the conserved glycines in the consensus sequence (C 2 XXGG(S/ T)C 3 ) for alanines (GG to AG, or GA, or AA) preserved the fucosylation in vitro as long as the EGF repeat was properly folded (20). A related sequence motif, C 2 XXGA(S/T)C 3 , was present in seven EGF repeats of Drosophila Notch, one of which, EGF repeat 25, is of particular interest because it is an EGF repeat to which a number of Abruptex mutations of Notch map (34). Abruptex mutations are dominant alleles of Notch that have been suggested to affect its sensitivity to Fringe (35,36). To begin to investigate the possibility that C 2 XXGA(S/T)C 3 motifs could be sites of modification by O-FucT-1 and Fringe, we constructed and expressed in S2 cells a transgene encoding FLAG-tagged EGF repeat 25 of Drosophila Notch polypeptide, which contains the sequence C 2 QNGATC 3 . Notch-EGF25 polypeptide was affinity purified on beads and then incubated with purified Fringe in the presence of UDP-[ 3 H]GlcNAc. Analysis of the reaction products by fluorography revealed a labeled band of the predicted mobility, corresponding to [ 3 H]GlcNAc-modified Notch EGF25-O-fucose (Fig. 8), demonstrating that an EGF repeat containing the sequence C 2 XXGA(S/T)C 3 can be modified with both O-fucose and Fringe.
The modification of O-fucose on EGF25 of Notch suggests that the C 2 XXGA(S/T)C 3 sites in Serrate may also be modified by O-FucT-1 and Fringe. These results open the possibility of modifying any Ser or Thr located amino-terminal to the third cysteine of an EGF repeat with O-fucose. Besides consensus sites in EGF repeats 3, 4 (interrupted repeat), 5, 12, and 13, Serrate has such "broad consensus" sites in repeats 2, 7, and 14 ( Figs. 1 and 6). To assess whether or not the O-fucosylation detected in the Ser-4m mutant was limited to these alternate sites in EGF repeats, we performed four additional rounds of mutagenesis to generate a Ser-8m protein in which all 8 potential sites (EGF 2, 3, 4, 5, 7, 12, 13, and 14) were mutated to Ala or Val (Figs. 6 and 7A). No

FucT-1 and Fringe occurs on consensus as well as nonconsensus EGF repeats.
A, schematic of the Serrate extracellular domain, using the same symbolism as in Fig. 1. X indicates the O-fucose consensus sites subjected to mutagenesis. B, the panel shows Western blot and fluorographic analysis of different forms of Serrate protein after Fringelabeling assay. Lanes 1 correspond to wild-type Serrate protein, lanes 2 correspond to Ser-4m protein, and lanes 3 correspond to Ser-8m protein.

FIG. 8. EGF25 repeat of Drosophila Notch is a substrate for both O-FucT-1 and Fringe.
Notch-EGF25 repeat was purified from Drosophila cell culture using FLAG-affinity beads and assayed for its ability to be a substrate for Fringe in vitro as described under "Experimental Procedures." After this in vitro glycosyltransferase assay, its products were analyzed by SDS-PAGE and fluorography (lane 2). In the control experiment (lane 1), just Drosophila S2 cells without Notch-EGF25 expression were used for the initial purification step. All other steps of the control were performed in parallel and using identical conditions to the experiment with Notch-EGF25. transfer of GlcNAc by Fringe from UDP-GlcNAc onto this mutant form of Ser could be detected, even though efficient labeling of wild-type and Ser-4m mutant occurred in parallel experiments (Fig. 7B). These results indicate that the ability of Fringe to modify Notch ligands was limited to EGF repeats containing Ser or Thr before the third cysteine.

Expansion of the Number of Substrates for O-FucT-1 and
Fringe-In this study, we have experimentally demonstrated that both Drosophila Notch ligands, Serrate and Delta, as well as their mammalian counterparts, Delta1 and Jagged1, are modified with O-fucose and can then be further elongated with GlcNAc by Fringe. The Notch ligands define a new class of substrates for O-fucose modification and are only the third class of proteins identified in which O-fucose is further elongated by GlcNAc. The demonstration that Notch ligands were modified by O-fucose raises the prospect that most or all EGF repeats containing the C 2 XXGG(S/T)C 3 O-fucose consensus site will bear this modification. Moreover, although further work will be required to define the precise structural requirements for O-fucosylation and ␤1,3-GlcNAc elongation by Fringe, the observation that Drosophila Notch-EGF repeat 25 was a Fringe substrate raises the prospect that many other proteins not previously recognized as containing the O-fucose consensus may in fact contain this modification. At the same time, the inability of Ser-8m to act as a Fringe substrate suggests that sites of Fringe modification will be limited to C 2 X 3-5 (S/T)C 3 (where X 3-5 are any 3-5 amino acid residues) sequences within EGF repeats, which we define as broad consensus sites (Fig. 1). If all broad consensus sites within Drosophila Notch were actually utilized, this would increase the number of potential sites for Fringe modification from 11 to 23. However, previous work has demonstrated that not all broad consensus sites are modified with O-fucose. For instance, protein C and protein Z, both EGF repeat-containing serum glycoproteins, are unmodified with O-fucose but contain the sequences C 2 CGHGTC 3 and C 2 LHNGSC 3 , respectively (18). Thus, the consensus site for O-fucose modification is broader than C 2 XXGG(S/T)C 3 but narrower than C 2 X 3-5 (S/T)C 3 . It is also possible that additional features of the three-dimensional structure of an EGF domain may contribute to substrate recognition by O-FucT-1. Further work to define the precise structural requirements for proteins to be in vivo substrates for O-fucosylation and consequent ␤1,3-GlcNAc elongation by Fringe is in progress.
Potential Importance of O-Fucosylation of Notch Ligands for Notch Signaling-A number of observations suggest that modification of ligand by O-FucT-1 and/or Fringe is essential for their proper function. First, the presence of multiple O-fucose sites is highly conserved in all Notch ligands. Moreover, there are distinct patterns of conservation among different classes of ligands (Fig. 1). This conservation makes it likely that there is some functional significance to this modification. Second, genetic studies in Drosophila have identified a mutation in a potential O-fucose site in EGF repeat 3 of Delta. The Delta mutation Dl sup5 is a hypomorphic allele of Delta that was isolated as a dominant suppressor of the Notch split mutant phenotype (small-roughened eyes and bristle defects) (37). Dl sup5 mutation is caused by a Gly to Arg change in the third EGF repeat (Fig. 1, C 2 KNGGTC 3 to C 2 KNGRTC 3 ), destroying the CXXGG(S/T)C consensus, potentially eliminating or reducing O-fucosylation at this site. Third, a missense mutation resulting in the human disorder Alagille syndrome maps to a predicted O-fucosylation site in EGF5 of human Jagged1 (38). This mutation is very similar to the Dl sup5 mutation (C 2 SHGGTC 3 to C 2 SHRGTC 3 in EGF5), again potentially eliminating glycosylation at this site.
If O-fucose sites on the ligands are important for Notch signaling, how might they act? In fibroblast growth factor signaling, heparan sulfate proteoglycans play an essential role by mediating ligand dimerization, which is essential for receptor activation (39). There is some evidence that ligand multimerization may contribute to Notch activation (40), and many lectins are known to act as dimers or multimers (41). Thus, it is possible that a lectin recognizes O-fucose glycans on Notch ligands and contributes to Notch activation by facilitating ligand multimerization. The presence of O-fucose glycans on both Notch and its ligands presents the even more intriguing possibility that an O-fucose binding lectin could potentiate Notch activation by directly facilitating Notch-ligand binding.
Potential Importance of Elongating O-Fucose on Notch Ligands by Fringe for Notch Signaling-Our results have shown that O-fucose on Notch ligands can be further elongated by Fringe. Genetic studies in Drosophila and assays with cultured mammalian cells have largely been interpreted as favoring the hypothesis that the Notch receptor is the key target of Fringe action (11,14). Most notably, the action of Fringe in most contexts appears to be cell autonomous and in the signal receiving cell. Nonetheless, it remains possible that Fringe influences Notch signaling autonomously by modifying Notch ligands, because Notch ligands actually act as both paracrine agonists and autocrine antagonist of Notch receptor activation. The autocrine inhibition of Notch activation by Notch ligands is strictly cell autonomous, and has been termed "autonomous inhibition" or "cis-inactivation" (42)(43)(44)(45)(46). The mechanism of this inhibition is not known, but it has been hypothesized to result from binding between Notch and Notch ligands within the same cell in a fashion that does not activate Notch but precludes it from interacting with and being activated by ligands presented by neighboring cells. Biochemical evidence in favor of such interactions has been obtained recently from overexpression experiments in cultured mammalian cells (47). Thus, one possibility is that glycosylation of Notch ligands by Fringe could affect the ability of cells to respond to ligands expressed by neighboring cells through an influence on autonomous inhibition.
Alternatively, it could be that glycosylation by Fringe affects the ability of Notch ligands to activate the Notch receptor in neighboring cells. Although most studies have emphasized autonomous effects of Fringe on Notch signaling, at least three studies have also described nonautonomous effects of Fringe. In the Drosophila wing, signaling by Serrate to neighboring cells actually appears to be stimulated by co-expression with Fringe (11). During bristle development in Drosophila, Fringe expression appears to decrease the ability of ectopically expressed Serrate or Delta to signal to neighboring cells (45). In the developing thymus, Lunatic Fringe influences Notch signaling in cells adjacent to where it is expressed (48). None of these studies provides strong evidence for a functionally relevant modification of Notch ligands by Fringe, because in all cases the influence of Fringe could be indirect. Nonetheless, they suggest contexts in which a modification of Notch ligands by Fringe may be significant. More importantly, the demonstration that Notch ligands are substrates of Fringe now provides a biochemical basis for further investigating this possibility.
Acknowledgments-We are grateful to Trudy Correia for Drosophila transformation and thank Khushi Matta for generously provided disaccharide standards, Gerry Weinmaster for Delta1-Fc plasmid, Tom Vogt for Jagged1-Fc plasmid, Mark Muskavitch for guinea pig Delta antibodies, Eli Knust for rabbit Serrate antibodies, and the Developmental Studies Hybridoma Bank for mouse Delta antibodies.