The O-fucosyltransferase O-fut1 is an extracellular component that is essential for the constitutive endocytic trafficking of Notch in Drosophila

Notch signaling is an evolutionarily conserved mechanism that regulates a broad spectrum of cell-specification events through local cell-cell communication (Artavanis-Tsakonas et al.,1999). Notch encodes a single-pass transmembrane receptor protein with 36 epidermal growth factor-like (EGF) repeats and three Notch/LIN-12 repeats in its extracellular domain and six CDC10/Ankyrin repeats and a PEST-like sequence in its intracellular domain(Kidd et al., 1983; Wharton et al., 1985). In Drosophila, two ligands for Notch, Delta and Serrate, which are also transmembrane proteins, are known(Fleming, 1998). The binding between Notch and its ligands leads to a proteolytic cleavage within the transmembrane domain that releases the intracellular domain Notch^ICD (Mumm and Kopan,2000). Notch^ICD then translocates to the nucleus and acts as a coactivator for a sequence-specific DNA-binding protein, Suppressor of Hairless, and this complex activates various target genes of the Notch signal (Bray and Furriols,2001).

This main Notch pathway is evolutionarily conserved from nematodes to mammals (Lai, 2004). However,several additional processes, such as the intracellular transportation of Notch and its ligands, tightly regulate Notch signaling(Le Borgne et al., 2005). For example, the activation of Notch requires its endocytic incorporation and/or that of its ligand (Seugnet et al.,1997), and Notch endocytosis is also involved in the downregulation of Notch activity(Jékely and Rørth,2003). Several regulators of the endocytic trafficking of Notch,such as the Nedd4 family proteins, Arrestin, Numb and Deltex, have been identified (Sakata et al.,2004; Wilkin et al.,2004; Mukherjee et al.,2005; Berdnik et al.,2002; Hori et al.,2004). In addition, recent findings show that mutations of genes that are generally involved in endocytosis affect the amount and activity of Notch (Giebel and Wodarz,2006). Although the amount of Notch increased in these mutants,the Notch signal was inactivated in some of them and hyperactivated in others(Lu and Bilder, 2005; Vaccari and Bilder, 2005; Thompson et al., 2005; Moberg et al., 2005; Maitra et al., 2006; Herz et al., 2006). However,it is largely unknown how the endocytic pathway influences the activity of Notch.

In addition to its trafficking, the signaling activity of Notch is also influenced by its glycosylation. Analyses of the Notch signal in a mutant of the UDP-GlcNAc transporter gene, which is probably required for protein N-glycosylation, suggested that this modification is essential for the normal functioning of Notch (Goto et al., 2001; Selva et al.,2001). Furthermore, Notch undergoes O-linked fucosylation, and the functions of this modification have been studied extensively (Haines and Irvine,2003). The EGF-like repeats of the Notch extracellular domain,which contain a consensus sequence, are modified by the O-linked tetrasaccharide Sia-α2,3-Gal-β1,4-GlcNAc-β1,3-Fuc(Moloney et al., 2000). A GDP-fucose protein O-fucosyltransferase1 catalyzes this O-linked fucosylation in mammals and Drosophila(Wang et al., 2001; Okajima and Irvine, 2002). In Drosophila, this enzyme is encoded by O-fut1(Okajima and Irvine, 2002). This O-fucosylation of Notch is essential for Notch signaling and Notch-ligand interactions (Okajima and Irvine, 2002; Sasamura et al.,2003; Okajima et al.,2003; Shi and Stanley,2003). N-acetylglucosamine is subsequently added to this fucose moiety by a fucose-specific β1,3 N-acetylglucosaminyltransferase, Fringe(Moloney et al., 2000). This modification modulates Notch-ligand interactions(Panin et al., 1997; Brückner et al.,2000).

More recently, O-fut1 was shown to act as a chaperon for Notch, a function that does not require the O-fucosyltransferase enzymatic activity of O-fut1 (Okajima et al., 2005). Knocking down O-fut1 by RNA interference (RNAi) in Drosophila cultured cells prevents the Notch extracellular domain polypeptide from being secreted into the medium(Okajima et al., 2005). When this Notch fragment and a mutant O-fut1 that lacks O-fucosyltransferase activity are coexpressed, the mutant O-fut1 still promotes the binding between the Notch polypeptide and its ligands(Okajima et al., 2005). These observations led to the proposal that O-fut1 is required for the proper folding of the Notch extracellular domain, and this function is independent of the enzymatic activity of O-fut1 (Okajima et al., 2005). Here, we demonstrated that O-fut1 has another,distinct, function in Notch signaling. Extracellular O-fut1 was required for the constitutive endocytic transportation of Notch to the early endosome, and this function was also independent of the O-fucosyltransferase activity of O-fut1.

We used Canton-S as the wild-type strain, O-fut1^4R6 as the O-fut1 mutant(Sasamura et al., 2003), Delta^RevF10 Serrate^RX106 as the double mutant of Delta and Serrate(Ligoxygakis et al., 1998) and Notch^55e11 (Kidd et al., 1983) as the Notch mutant. daughterless(da)-GAL4 (GAL4^daG32)(Wodarz et al., 1995), engrailed (en)-GAL4(Johnson et al., 1995) and scalloped (sd)-GAL4(Roy et al., 1997) were used as the GAL4 drivers. We used FLP/FRT (Xu and Rubin, 1993) and the MARCM(Lee and Luo, 1999) system to generate somatic mutant clones. To isolate the Gmd mutant, 171 independent white derivative strains were established from GS13045 and screened using a genomic PCR method(Toba et al., 1999). To generate O-fut1^- clones in the Gmd^H78mutant, hs-flp; Gmd^H78 FRT42D O-fut1^4R6/CyO, Kr-GFP males were crossed to Gmd^H78 FRT42D P{πM}45F/CyO, Kr-GFPfemales.

For the O-fut1-ER^- construct, three tandem oligonucleotides encoding GSEEQKLISEEDLL were inserted into an artificially created BamHI site before the stop codon of O-fut1 cDNA. UAS-O-fut1,UAS-O-fut1-G3, UASO-fut1-ER^- and UAS-ER-CFP(unpublished, provided by A. Sato, Purde University, West Lafayette, IN) were made by inserting cDNAs encoding wild-type O-fut1, Nti-G3, a mutant O-fut1 lacking its O-fucosyltransferase activity(Sasamura et al., 2003),O-fut1-ER^-, a mutant O-fut1 lacking a functional ER-retention signal, and ECFP-ER (Clontech), an ECFP with an ER-retention signal,respectively, into pUAST.

S2 cells were cultured, transfected and stained as described previously(Fehon et al., 1990; Sasamura et al., 2003). To express O-fut1-ER^- in S2 cells, a cDNA encoding O-fut1-ER^- was cloned into pRmHa-3 and introduced into S2 cells. We also used pRmHa-3-Notch (Fehon et al.,1990), pRmHa-3-NotchΔEC(Rebay et al., 1991),pRmHa-3-O-fut1-myc and pRmHa-3-O-fut1-IR(Sasamura et al., 2003). To detect O-fut1 incorporated into Notch-expressing S2 cells, conditioned medium containing O-fut1-ER^- was collected from S2 cells transfected with pRmHa-3-O-fut1-ER^- and added to S2 cells transfected with pRmHa-3-Notch. After a 20-minute incubation, the cells were fixed and stained as described previously (Fehon et al.,1990). To detect the effect of the O-fut1 knockdown on Notch endocytosis, the transfected S2 cells were incubated in Drosophila M3 medium (Sigma) on glass slides coated with concanavalin A (Sigma) for 2 hours at 25°C, anti-Notch extracellular antibody (rat1, 1/500) was added, and the slides were incubated at 4°C for 1 hour. After being washed in M3 medium three times at 4°C, the cells were incubated in M3 medium for 15 minutes at 25°C. In some cases, conditioned medium containing O-fut1-ER^- was used in place of the M3 medium, beginning 20 minutes before the antibody was added and continuing until the end of the culture period. The cells were subsequently fixed and stained.

The primary antibodies used in this study were mouse anti-Notch C17.9C6(1/500) (Fehon et al., 1990),rat anti-Notch rat1 (1/500, a gift from S. Artavanis-Tsakonas), rat anti-DE-cadherin DCAD2 (1/20), guinea pig anti-Hrs (1/1,000)(Lloyd et al., 2002), mouse anti-GAL4 RK5C1 (1/500, Santa Cruz), mouse anti-engrailed (1/1000)(Patel et al., 1989), mouse anti-Wg 4D4 (1/500) (Brook and Cohen,1996), rat anti-GFP GF090R (1/1000, Nacalai) and mouse anti-Myc MC045 (1/500, Nacalai). An O-fut1 guinea pig antibody was raised against O-fut1 (amino acids 27 to 402) that had six histidines added to the C-terminus and was expressed in SF9 insect cells (used in 1/1000 dilution). Immunostaining of wing discs and Garland cells was performed as previously described (Matsuno et al.,2002). To detect cell-surface Notch, dissected wing discs were incubated in 1/100-diluted rat1 in M3 medium at 4°C for 2 hours. They were rinsed four times with M3 medium at 4°C, and then incubated for 20 minutes or 10 hours in M3 medium. At this point the M3 medium was supplemented with 1μl/ml 20-OH ecdysone (Sigma) and 1% fetal calf serum (Gibco). The endocytic tracer uptake assay was performed as described(Entchev et al., 2000). Confocal images were taken with LSM5 PASCAL and LSM510 META. We used Auto Deblur (AutoQuant) as the deconvolution tool.

GDP-L-fucose levels contained in whole-larva homogenates were measured using previously described procedures with minor modifications(Noda et al., 2002). Briefly,larvae were homogenized in a Dounce homogenizer under crushed ice in 250 μl of 0.25 mol/l sucrose buffer containing Protease Inhibitor Mix/DMSO diluted 1/1000 (Wako, Osaka, Japan), 5 mmol/l adenosine-5-monophosphate (AMP) (pH 7.4)(Wako, Osaka, Japan), 10 mmol/l Tris-HCl (pH 7.4), 10 mmol/l KCl and 10 mmol/l MgCl₂. Larva homogenates were spun and the supernatants were subjected to ultracentrifugation at 105,000×g for 1 hour at 4°C to obtain the cytosolic fraction. The protein concentration in these fractions was quantified using a BCA kit (Pierce, IL, USA). In a typical experiment, 120 μg protein from the cytosolic fraction was adjusted to a volume of 20 μl with chilled autoclaved water and then boiled at 100°C for 20 seconds. Then, 8.5 μl ice-cold 200 mmol/l MES-NaOH (pH 7.0) was added, and the samples were spun, mixed with 1 μl 10% Triton X-100 and 0.5μl (36.8 pmol) GnGn-bi-Asn-4-(2-pyridylamino) butylamine (PABA) (Sigma),subjected to a series of enzymatic digestions and coupled with PABA through a peptide bond and 5 μl purified α1-6 FucT (1050 nmol/l). The mixtures were incubated at 37°C for 2 hours and the reaction was terminated by boiling at 100°C for 1 minute. The samples were then spun at 15,000×g for 10 minutes, and 10 μl of the 35 μl of supernatant was subjected to high performance liquid chromatography for the GDP-L-fucose assay as described (Noda et al., 2003).

Whole-cell extract was prepared from S2 cells as previously described(Sasamura et al., 2003). Anti-Notch (C17.9C6) or anti-Myc (9E10) antibodies were added to cell lysates and immunoprecipitated with Protein G Sepharose 4 Fast Flow (Amersham). The beads were washed five times with an extraction buffer and subjected to western blotting as described (Sasamura et al., 2003). The primary antibodies used for blotting were anti-Notch (C17.9C6) and anti-Myc (9E10).

Notch is distributed in a honeycomb pattern, which corresponds to the location of the adherens junctions, as judged by the localization of DE-cadherin, in the apical region of the wild-type wing disc epithelium (Fig. 1A, upper part; Fig. 1E)(Fehon et al., 1991; Oda et al., 1994). However, in O-fut1 homozygous mutant (O-fut1^-) cells, this localization was disrupted, and Notch accumulated in intracellular vesicles,which were distributed mostly in the apical cytoplasm, as reported previously(Fig. 1A-E)(Okajima et al., 2005). This Notch accumulation was cell autonomous in O-fut1^- cells(Fig. 1B,C). It is proposed that this accumulation is caused by the quality control mechanism, which retains mis-folded Notch in the endoplasmic reticulum (ER)(Okajima et al., 2005). However, we previously showed that cell-surface Notch does not decrease significantly when O-fut1 is knocked down, in the Drosophila S2 cell line (Sasamura et al., 2003). Therefore, we decided to determine whether Notch was transported normally to the plasma membrane in O-fut1^- cells in vivo.

To detect the fates of the Notch receptors present at the cell surface, we incubated live wing discs from third-instar larvae in medium containing an antibody against the extracellular domain of Notch (rat1)(Le Borgne and Schweisguth,2003). The antibody was detected in intracellular vesicles in wild-type cells after a 20 minute incubation(Fig. 2C-F, left part of each panel). However, the antibody was not incorporated into these vesicles in live Notch mutant cells, indicating that the antibody specifically labeled Notch under these conditions (Fig. 2A,B). In live O-fut1^- cells under the same conditions, Notch was detected in intracellular vesicles that were mostly located in the apical region, although these Notch-containing vesicles were smaller and the staining was fuzzy compared with the Notch-containing vesicles in wild-type cells (Fig. 2C-F,at right). These results suggest that Notch was delivered to the plasma membrane in these cells.

We next studied the intracellular vesicles in wild-type and O-fut1^- cells in more detail. In wild-type cells, most of the Notch-containing endocytic vesicles, visualized by live Notch staining,were labeled by the early endosome marker Hrs(Lloyd et al., 2002)(Fig. 2G-J, right part). In the O-fut1^- cells, however, live-labeled Notch was not found in the Hrs-positive early endosomes (Fig. 2G-J, left). Furthermore, the live antibody staining for Notch in the intracellular vesicles of the O-fut1^- cells was much stronger than in the wild-type cells after 10 hours, indicating that Notch was indeed incorporated into the cells by endocytosis but failed to be degraded(Fig. 2K). These results suggested that surface Notch was not transported to the early endosome in O-fut1^- cells, thereby preventing the trafficking of Notch to the lysosomes, where Notch is degraded(Lu and Bilder, 2005).

Early and late endosomes in the wing disc epithelium can be visualized by fluorescent dextran added extracellularly to the live wing discs(Entchev et al., 2000). Using this system to follow the localization of Notch, we found that the dextran-positive vesicles also stained for Notch (95% of vesicles, n=77) in permeabilized wild-type cells(Fig. 2L, white arrowheads and upper left inset) (Hori et al.,2004). By contrast, in the O-fut1^- cells, only 6% of the dextran-positive vesicles were also positive for Notch(n=47) (Fig. 2L, open arrowheads and lower right inset). Therefore, Notch failed to be transported to these endosomal compartments. Furthermore, the numbers of dextran-positive endocytic vesicles were equivalent in the wild-type and O-fut1^- cells, indicating that the O-fut1mutation did not affect their formation(Fig. 2L). Together, these observations suggest that O-fut1 is required for the transportation of Notch from early endocytic vesicles to the early endosome. However, we could not identify the early endocytic vesicles in which Notch accumulated in the O-fut1^- cells. None of the available markers for various endocytic compartments, such as Hook (early endosomes)(Chang et al., 2002), Rab11(recycling endosomes) (Ullrich et al.,1996; Dollar et al.,2002), rab7-GFP (late endosomes)(Entchev et al., 2000) or ubiquitinylated proteins (aggresomes)(Kopito, 2000) showed co-labeling with Notch in the O-fut1^- cells (data not shown).

Next, we examined whether Notch co-localized with marker proteins for the ER. Protein disulfide isomerase (PDI) is generally involved in protein folding in the ER lumen and is important for quality control(Ferrari and Söling,1999). A protein trap line, 74-1, expresses PDI-GFP under the control of the original pdi promoter and thus accurately reflects the spatiotemporal expression and localization of the PDI protein(Bobinnec et al., 2003). Under a standard confocal microscope, Notch appeared to mostly co-localize with PDI-GFP in O-fut1^- cells, which was consistent with the previous report (Okajima et al.,2005) (Fig. 3A). However, higher resolution images obtained by deconvolution analysis revealed that this ER marker did not co-localize with the Notch-containing vesicles in O-fut1^- cells (Fig. 3B,C). Similar results were obtained using ER-CFP as the ER marker(Fig. 3D,E). In the positive control, PDI-GFP was detected with the mouse anti-GFP primary antibody and two different fluorescent secondary antibodies(Fig. 3F), and the pattern of the two fluorescent signals was the same, confirming the resolution of our analysis. The Notch-containing vesicles were also negative for cis-Golgi (anti-KDEL receptor) and Golgi (anti-Golgi, Drosophila) markers (data not shown). Therefore, our results indicated that most Notch did not accumulate in the conventional ER. However,these results do not exclude the possibility that Notch accumulated in a specialized ER compartment (Huyer et al.,2004).

O-fut1 has both O-fucosyltransferase-activity-dependent and-independent functions (Okajima et al.,2005). Therefore, we investigated whether the novel function of O-fut1 in Notch endocytic trafficking depends on its enzymatic activity. As reported, an O-fucose-independent function of O-fut1 can be studied using a mutant of the GDP-d-mannose 4,6-dehydratase(Gmd) gene, which encodes the first enzyme of the de novo GDP-fucose synthesis pathway (Roos et al.,2002; Okajima et al.,2005). We generated two putative null-mutant alleles of Gmd (Fig. 4A). The concentration of GDP-fucose in lysates prepared from the third-instar larvae of Gmd mutants was determined biochemically(Fig. 4B). The GDP-fucose level in the Gmd^H78 and Gmd^H43 heterozygotes was approximately half that of wild-type larvae, and it was not detectable in the Gmd^H78 and Gmd^H43 homozygotes(Fig. 4B). These results, which are consistent with these two alleles being null mutations of Gmd,corroborate the belief that GDP-fucose is synthesized exclusively from GDP-mannose via the de novo pathway in Drosophila(Roos et al., 2002). As reported before, using the Gmd¹ allele(Okajima et al., 2005), Notch signaling was abolished in the Gmd^H78 wing discs(Fig. 4C,D). However, Engrailed(En) expression, which is confined to the posterior compartment of the wing disc, was normal in the Gmd^H78 wing disc(Fig. 4E,F), supporting the idea that this defect was specific to the Notch signal.

To test whether the defect in Notch endocytic trafficking in O-fut1^- cells was due to the failure of Notch O-fucosylation, we examined the endocytic transportation of Notch in the Gmd^H78 background. By contrast to the result in O-fut1^- cells (Fig. 2L), we found that Notch was mostly located in the dextran-positive vesicles (98%, n=87) in the Gmd mutant(Fig. 4G). In addition, Notch did not accumulate to abnormally high levels in the Gmd mutant wing discs, although the size of the Notch-containing endocytic vesicles was slightly greater than those observed in wild-type cells(Fig. 4H,I). Nevertheless, this result suggested that the defect in the endocytic trafficking in O-fut1^- cells occurs independently of the O-fucose modification of Notch. To confirm that this defect was not due to a lack of O-fucosylation on Notch, we generated somatic O-fut1^- clones in the Gmd^H78 wing disc. We observed increased Notch protein levels in the O-fut1^-, Gmd^H78double-mutant cells(Fig. 4J, inside the white dotted line). Therefore, the knockout of O-fut1 in the Gmdmutant still induced the abnormal accumulation of Notch(Fig. 4J), again indicating that this effect is independent of the enzymatic function of O-fut1.

Our results suggested that Notch was transported to the plasma membrane and internalized in O-fut1^- cells, although it was not delivered to the early endosomes in these cells. In addition, our live-tissue labeling experiments showed that Notch became stable in endocytic vesicles after its incorporation by endocytosis in O-fut1^- cells(Fig. 2K). Thus, it is likely that Notch failed to be delivered to the lysosomes, where it is normally degraded in these cells (Lu and Bilder,2005). We therefore speculated that the turnover of Notch is reduced in the O-fut1^- cells. To address this possibility,we examined the half-life of Notch in wild-type and O-fut1^- cells. We used a heat-shock-inducible Notch-GAL4 fusion protein, N⁺-GV3, which retains the wild-type function of Notch (Struhl and Adachi,1998). The N⁺-GV3, which is produced for a short period under the control of a heat-shock promoter, can be specifically detected by an anti-GAL4 antibody (Struhl and Adachi,1998). Thus, we could trace the fates of this Notch protein against the background of continuously produced endogenous Notch(Hori et al., 2004). Thirty minutes after heat shock, N⁺-GV3 was expressed uniformly throughout the wing disc (Fig. 5B). Although the N⁺-GV3 was gradually degraded, there was more N⁺-GV3 in the O-fut1^- cells 6 and 12 hours after heat shock than in the surrounding wing disc cells(Fig. 5C,D).

As O-fut1 is thought to act extracellularly(Haines and Irvine, 2003), we next examined the requirement for the Notch extracellular domain for its stabilization in the O-fut1^- clones. N^EGF-GV3,a Drosophila Notch derivative lacking the fifth to 23rd (of 36) Notch EGF-like repeats (Struhl and Adachi,1998; Artavanis-Tsakonas et al., 1999), showed no increased stability in the O-fut1^- clones (Fig. 5E). Therefore, the Notch EGF repeats seem to be essential for the stabilization of Notch in O-fut1^- cells.

Because the half-life of Notch was prolonged in O-fut1^-cells, we tested whether O-fut1 overexpression would promote Notch degradation. O-fut1 overexpression was driven by UAS-O-fut1 under the control of en-GAL4 (Fig. 5F-H). N⁺-GV3 expression was induced by heat shock and detected with anti-GAL4 antibodies, as above. Although the anti-GAL4 antibody also recognized the GAL4 generated from en-GAL4, the GAL4 staining remained at the background level under our conditions(Fig. 5F). We found that N⁺-GV3 was greatly decreased in the O-fut1-overexpressing cells compared with the wild-type cells 24 hours after heat shock(Fig. 5H). Therefore, as opposed to the loss of O-fut1 function, its overexpression promoted the degradation of Notch. O-fut1-G3, which has three amino acid substitutions in an essential motif for glycosyltransferase activity, lacks enzymatic activity[Nti-G3 in Sasamura et al. (Sasamura et al., 2003)]. Significantly, O-fut1-G3 promoted the degradation of Notch as efficiently as did wild-type O-fut1 (compare Fig. 5H,I), further supporting our idea that O-fut1 promotes the transportation of Notch to the lysosomes independent of Notch O-fucosylation.

Because a lack of O-fut1 activity results in the failure of Notch-ligand interactions (Sasamura et al.,2003; Okajima et al.,2003), it is difficult to study the consequences of the disturbance in Notch turnover using loss-of-function mutations of O-fut1. Therefore, to examine the possible role of the O-fut1-regulated Notch turnover in the developmental context, we resorted to overexpression studies. The overexpression of O-fut1 or O-fut1-G3 in wing discs caused wing-nicking and vein-thickening phenotypes(Fig. 5J-L) that were reminiscent of Notch loss-of-function phenotypes(de Celis and Garcia-Bellido,1994). The expression of wingless (wg), a target of Notch signaling in the wing disc, was reduced significantly in the O-fut1-overexpressing wing disc (Fig. 5M,N) (Rulifson and Blair,1995). Additional consistent evidence is the finding that O-fut1 overexpression represses Notch signaling in notal microchaete development(Okajima and Irvine, 2002). These results suggest that O-fut1 downregulates Notch signaling by promoting the degradation of Notch via an O-fucosylation-independent mechanism.

O-fut1 has a KDEL-like motif, a HEEL sequence, at its C-terminus, which probably acts as an ER-retention signal(Teasdale and Jackson, 1996). Indeed, it was reported that O-fut1 mostly localizes to the ER(Okajima et al., 2005). However, given that O-fut1 influences the endocytic trafficking of Notch, it is likely that at least some fraction of O-fut1 is transported to the plasma membrane and then incorporated into the cells. Garland cells are suitable to study the subcellular localization of proteins in vivo because of their large cytoplasm (Culi and Mann,2003). In these cells, we found Myc-tagged O-fut1, which is otherwise wild type [O-fut1-myc; Nti-myc in Sasamura et al.(Sasamura et al., 2003)], in small vesicles, probably exocytic ones, that were not labeled with ER or Golgi markers (Fig. 6A,A′; data not shown). Some of the punctate O-fut1 staining co-localized with N⁺-GV3 staining, which was located in vesicles distant from the peri-nuclear ER (Fig. 6B,B′, arrows). These results suggest that O-fut1 physically interacts with Notch during exocytosis and/or endocytosis in vivo. Indeed,O-fut1 forms a stable complex with Notch, as shown by immunoprecipitation,which has been proposed to account for the enzymatic-activity-independent functions of O-fut1 (Fig. 6C)(Okajima et al., 2005). We also found that a Notch derivative lacking the EGF-like repeats(Notch^ΔEC) did not co-precipitate with O-fut1(Fig. 6C). These results suggest that O-fut1 interacted with the EGF-like repeats of the Notch extracellular domain.

To test our hypothesis that O-fut1 functions during Notch endocytosis, we investigated whether extracellular O-fut1 was incorporated into cells by endocytosis and whether it could rescue the defects in Notch endocytosis associated with the knockdown of O-fut1 function. S2 cells expressing Notch were cultured with conditioned medium containing secreted Myc-tagged O-fut1(O-fut1-ER^-). We detected the O-fut1-ER^-, which was added extracellularly, in vesicles in cells expressing Notch, but not in control cells (Fig. 6D). In the Notch-expressing cells, Notch and O-fut1-ER^- were co-localized in the vesicles (Fig. 6D). In addition, the extracellularly added O-fut1 suppressed the defects in the endocytic transportation of Notch that were induced by the inhibition of O-fut1 function. In the control S2 cells, about 18% of the Notch-containing vesicles co-localized with Hrs (Fig. 6E,H). The co-expression of a double-stranded RNA corresponding to Ofut1 mRNA [O-fut1-IR; nti-IR in Sasamura et al.(Sasamura et al., 2003)]suppressed this co-localization to about 8.5%(Fig. 6F,H). Importantly, the extracellularly added O-fut1 was sufficient to rescue Notch co-localization with Hrs to 16.7% (Fig. 6G,H). Furthermore, we found that O-fut1-ER^- was secreted more efficiently into the medium (data not shown) (Okajima et al., 2005) and accelerated the degradation of N⁺-GV3 more strongly than native O-fut1 in vivo(Fig. 6I,J). In addition, the overexpression of O-fut1-ER^- induced more severe Notch loss-of-function phenotypes than did overexpressed wild-type O-fut1(Fig. 6K,L). Together, these results support the idea that the acceleration of Notch turnover, which is induced by extracellular O-fut1, leads to the downregulation of Notch signaling.

Ligand binding can force a receptor to choose a specific trafficking path(Sorkin and Von Zastrow,2002). Therefore, it is possible that extracellular O-fut1 is required for Notch-ligand interactions, which may trigger the transportation of Notch to the endosome. However, no accumulation of Notch was seen in the somatic clones of cells that were double mutants for Delta and Serrate, as previously reported (data not shown)(Okajima et al., 2005). In addition, Notch co-localized normally with dextran added extracellularly to these double-mutant cells, indicating there was no interference with the endocytic transportation of Notch (Fig. 6M). Therefore, O-fut1 is probably required for the constitutive vesicular transportation of Notch, rather than for its ligand-induced endocytic transportation.

As a novel and enzymatic-activity-independent function of O-fut1, we propose that O-fut1 is required for the constitutive endocytic trafficking of Notch from the plasma membrane to the early endosome. Previously, it was shown that O-fut1 acts as a specific chaperon for Notch in an enzymatic-activity-independent manner(Okajima et al., 2005). Thus,O-fut1 has two distinct roles, neither of which depend on the O-fucosylation of Notch, during either the exocytotic or endocytic transportation of Notch.

As evidence for the chaperon activity of O-fut1, it was reported that Notch accumulates in the ER in O-fut1^- cells(Okajima et al., 2005). The idea that O-fut1 acts as a chaperon was based on the assumption that its absence results in mis-folded Notch, which could be retained in the ER by quality control mechanisms. However, our results suggested that the level of apical surface Notch, which was recognized by live tissue staining with an anti-Notch antibody, was not reduced significantly. Therefore, at least some Notch is delivered to the plasma membrane in O-fut1^-cells. This Notch transportation may be slower or less efficient in these cells, which could account for the previous observation that Notch did not reach the surface of O-fut1-depleted Drosophila S2 cells at a given time point (Okajima et al.,2005). In addition, our high-resolution analysis revealed that Notch did not accumulate in vesicles that were positive for two well-characterized ER markers, although we could not exclude the possibility that Notch accumulates in a specific ER subdomain(Huyer et al., 2004). However,biochemical evidence indicates that O-fut1 promotes the proper folding of Notch (Okajima et al., 2005). Thus, we speculate that the two enzymatic-activity-independent roles of O-fut1 are not mutually exclusive and can take place simultaneously.

Our results showed that Notch failed to be delivered to the Hrs-positive early endosomes in O-fut1^- cells. Therefore, we speculate that O-fut1 may be required for the early endosomal fusion of endocytic vesicles containing Notch, because Notch was internalized in small vesicles in O-fut1^- cells, but was not transported to the early endosomes. Indeed, similar Notch accumulation in small vesicles is observed in cells with defective endosome vesicle fusion caused by a mutation in the avl gene (Lu and Bilder,2005). However, in the present study, we were unable to determine the nature of the vesicles where Notch was accumulated in O-fut1^- cells, because the available markers for ER,Golgi, early endosomes, recycling endosomes, late endosomes and lysosomes did not colocalize with the accumulated Notch.

Degradation of transmembrane proteins in the lysosome requires the proteins to be transported first to the early and then to the late endosome(Babst, 2005). Thus, our model predicts that the half-life of Notch is prolonged in O-fut1^- cells, because Notch was not transported to the early endosome in these cells. Consistent with this model, we found that the half-life of Notch was extended in O-fut1^- cells and reduced in cells overexpressing O-fut1. In addition, the overexpression of O-fut1 suppressed Notch signaling in vivo. Therefore, O-fut1 may play an important role in maintaining the appropriate Notch turnover ratio, which probably functions to downregulate Notch signaling in wild-type cells.

We also found that Notch did not accumulate in the cells that were double mutants for Delta and Serrate, suggesting that interactions between Notch and its ligands are not relevant to the O-fut1-dependent endocytosis of Notch. Therefore, the accumulation of Notch in O-fut1^- cells is due to defects in constitutive endocytosis rather than in ligand-induced endocytosis. Taking these results together, we speculate that O-fut1 may be involved in sweeping unactivated and excess Notch from the plasma membrane under physiological conditions. Therefore, the lack of this O-fut1 function predictably results in the upregulation of Notch signaling. Our overexpression analysis of O-fut1 is consistent with this idea. However, it is difficult to examine this possibility using loss-of-function mutants of O-fut1, because O-fut1 is also essential for the interactions between Notch and its ligands(Sasamura et al., 2003; Okajima et al., 2003).

Although O-fut1 is known to be secreted when it is expressed in S2 cells(Okajima et al., 2005), we found that the function of O-fut1 was required in a cell-autonomous manner in vivo. The accumulation of Notch was observed in all O-fut1^- cells in somatic clones, even in cells surrounded by wild-type cells (arrow in Fig. 1B,C). Therefore, under physiological conditions, O-fut1 is probably transported to the plasma membrane and then endocytosed only into the same cell. In wild-type cells, O-fut1 may continuously interact with Notch as it cycles from exocytic to endocytic pathways. The following three observations support this idea. First, O-fut1 forms a stable complex with the extracellular domain of Notch. Second, O-fut1 and Notch were occasionally found in the same exocytic vesicles. Third, O-fut1 added extracellularly was incorporated into the cells in a Notch-dependent manner. However, we also found that O-fut1 added to the medium was sufficient to restore the endocytosis of Notch in O-fut1 knockdown cells. Therefore, the interaction of O-fut1 and Notch, which occurs after Notch reaches the cell surface, is sufficient for the normal endocytic trafficking of Notch. However, it is presently unknown how extracellular interactions between Notch and O-fut1 affect the endocytic trafficking of Notch. The specific complex of the Notch extracellular domain and O-fut1 may influence the intracellular recognition machinery between Notch-containing early endocytic vesicles and the early endosomes, such as the tethering factors or Rab5(Rodriguez-Boulan et al.,2005). Nevertheless, our results raise the interesting possibility that extracellular modification enzymes may be necessary for, or control, the endocytic transportation path of receptor proteins that are also their substrates.

We thank G. Struhl, A. K. Sato, H. J. Bellen, M. Gonzalez-Gaitan, N. Perrimon, the Bloomington Drosophila Stock Center, and the Developmental Studies Hybridoma Bank for providing fly stocks and antibodies. We also thank S. Artavanis-Tsakonas and A. Mukherjee for critical comments on the manuscript, and T. Okajima, K. D. Irvine, E. Spana, D. Bilder and N. Perrimon for support during the early stages of this work. This work was supported by grants-in-aid from the Japanese Ministry of Education, Culture, Sports and Science (K.M.), and PRESTO, JST (K.M.).