Portal de Periódicos da CAPES

O-Fucosylation of Notch Occurs in the Endoplasmic Reticulum

2005; Elsevier BV; Volume: 280; Issue: 12 Linguagem: Inglês

10.1074/jbc.m414574200

1083-351X

Yi Luo, Robert S. Haltiwanger,

Cellular transport and secretion

LADII (leukocyte adhesion deficiency type II)/CDGIIc (congenital disorder of glycosylation type IIc) is a rare autosomal recessive disease characterized by leukocyte adhesion deficiency as well as severe neurological and developmental abnormalities. It is caused by mutations in the Golgi GDP-fucose transporter, resulting in a reduction of fucosylated antigens on the cell surface. A recent study using fibroblasts from LADII/CDGIIc patients suggested that although terminal fucosylation of N-glycans is reduced severely, protein O-fucosylation is generally unaffected (Sturla, L., Rampal, R., Haltiwanger, R. S., Fruscione, F., Etzioni, A., and Tonetti, M. (2003) J. Biol. Chem. 278, 26727–26733). A potential explanation for this phenomenon is that enzymes adding O-fucose to proteins localize to cell organelles other than the Golgi apparatus. In this study, we investigated the subcellular localization of protein O-fucosyltransferase 1 (O-FucT-1), which is responsible for adding O-fucose to epidermal growth factor-like repeats. Our analysis reveals that, unlike all other known fucosyltransferases, O-FucT-1 is a soluble protein that localizes to the endoplasmic reticulum (ER). In addition, it appears that O-FucT-1 is retained in the ER by a KDEL-like sequence at its C terminus. Our results also suggest that enzymatic addition of O-fucose to proteins occurs in the ER, suggesting that a novel, ER-localized GDP-fucose transporter may exist. The fact that O-FucT-1 recognizes properly folded epidermal growth factor-like repeats, together with this unique localization, suggests that it may play a role in quality control. LADII (leukocyte adhesion deficiency type II)/CDGIIc (congenital disorder of glycosylation type IIc) is a rare autosomal recessive disease characterized by leukocyte adhesion deficiency as well as severe neurological and developmental abnormalities. It is caused by mutations in the Golgi GDP-fucose transporter, resulting in a reduction of fucosylated antigens on the cell surface. A recent study using fibroblasts from LADII/CDGIIc patients suggested that although terminal fucosylation of N-glycans is reduced severely, protein O-fucosylation is generally unaffected (Sturla, L., Rampal, R., Haltiwanger, R. S., Fruscione, F., Etzioni, A., and Tonetti, M. (2003) J. Biol. Chem. 278, 26727–26733). A potential explanation for this phenomenon is that enzymes adding O-fucose to proteins localize to cell organelles other than the Golgi apparatus. In this study, we investigated the subcellular localization of protein O-fucosyltransferase 1 (O-FucT-1), which is responsible for adding O-fucose to epidermal growth factor-like repeats. Our analysis reveals that, unlike all other known fucosyltransferases, O-FucT-1 is a soluble protein that localizes to the endoplasmic reticulum (ER). In addition, it appears that O-FucT-1 is retained in the ER by a KDEL-like sequence at its C terminus. Our results also suggest that enzymatic addition of O-fucose to proteins occurs in the ER, suggesting that a novel, ER-localized GDP-fucose transporter may exist. The fact that O-FucT-1 recognizes properly folded epidermal growth factor-like repeats, together with this unique localization, suggests that it may play a role in quality control. Fucose is a common component of many N-linked and O-linked glycans found in glycoproteins in metazoans (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar). It is involved in a variety of critical physiological processes, such as selectin-mediated leukocyte-endothelial adhesion (2Vestweber D. Blanks J.E. Physiol. Rev. 1999; 79: 181-213Crossref PubMed Scopus (829) Google Scholar), ABO blood type antigen determination (3Lowe J.B. Baillieres Clin. Haematol. 1993; 6: 465-492Abstract Full Text PDF PubMed Scopus (81) Google Scholar), and Notch receptor signaling (4Haltiwanger R.S. Lowe J.B. Annu. Rev. Biochem. 2004; 73: 491-537Crossref PubMed Scopus (649) Google Scholar, 5Haines N. Irvine K.D. Nat. Rev. Mol. Cell Biol. 2003; 4: 786-797Crossref PubMed Scopus (338) Google Scholar). Fucose modification is catalyzed by fucosyltransferases, all of which require GDP-fucose as the donor substrate and a glycan or protein as the acceptor substrate. Currently known fucosyltransferases in mammals can be classified into four groups based on the specific linkages formed and the substrates to which fucose is added (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar). The first group includes the α2-fucosyltransferases that modify the terminal Gal moiety of lactosamine structures with α1,2-linked fucose (FUT1, -2). The second group is the α3/4-fucosyltransferases that add fucose in α1,3/4 linkage to the GlcNAc moiety of lactosamine structures (FUT3–7 and -9). α6 fucosyltransferase, which adds fucose in α1,6 linkage to the first GlcNAc of the N-glycan core, is the unique member of the third group (FUT8). Finally, the fourth group consists of protein O-fucosyltransferases 1 and 2 (O-FucT-1 1The abbreviations used are: O-FucT-1, protein O-fucosyltransferase 1; β4GalT, β-1,4-galactosyltransferase; EGF, epidermal growth factor; LADII/CDGIIc, leukocyte adhesion deficiency type II/congenital disorder of glycosylation type IIc; CHO, Chinese hamster ovary; ER, endoplasmic reticulum; TSRs, thrombospondin type 1 repeats; PNGase F, peptide N-glycosidase F. and O-FucT-2), the most recent additions to the fucosyltransferase family. Sequence alignment reveals that the protein O-fucosyltransferases share a common ancestry with both α2- and α6-fucosyltransferases (6Martinez-Duncker I. Mollicone R. Candelier J.J. Breton C. Oriol R. Glycobiology. 2003; 13: 1C-5CCrossref PubMed Scopus (62) Google Scholar). Nonetheless, O-FucT-1 and O-FucT-2 share several biochemical properties that distinguish them from other known fucosyltransferases (7Wang Y. Lee G.F. Kelley R.F. Spellman M.W. Glycobiology. 1996; 6: 837-842Crossref PubMed Scopus (57) Google Scholar, 8Wang Y. Spellman M.W. J. Biol. Chem. 1998; 273: 8112-8118Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar, 9Wang Y. Shao L. Shi S. Harris R.J. Spellman M.W. Stanley P. Haltiwanger R.S. J. Biol. Chem. 2001; 276: 40338-40345Abstract Full Text Full Text PDF PubMed Scopus (211) Google Scholar). 2Y. Luo and R. S. Haltiwanger, manuscript in preparation. For instance, both enzymes catalyze the addition of fucose directly to a serine or threonine in proteins instead of to another sugar. For both enzymes, the modified serine or threonine occurs in the context of a cysteine knot motif, epidermal growth factor (EGF)-like repeats for O-FucT-1, and thrombospondin type 1 repeats (TSRs) for O-FucT-2. EGF repeats are defined by the presence of six conserved cysteine residues forming three disulfide bonds in a specific pattern, C1–C3, C2–C4, C5–C6 (10Campbell I.D. Bork P. Curr. Opin. Struct. Biol. 1993; 3: 385-392Crossref Scopus (331) Google Scholar). The O-fucose modifies a serine or threonine adjacent to the third conserved cysteine within the consensus sequence C2X4–5(S/T)C3 (11Shao L. Moloney D.J. Haltiwanger R.S. J. Biol. Chem. 2003; 278: 7775-7782Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). O-FucT-1 has been shown to add O-fucose to EGF-like repeats from a variety of proteins including the Notch receptor (12Haltiwanger R.S. Curr. Opin. Struct. Biol. 2002; 12: 593-598Crossref PubMed Scopus (106) Google Scholar). Reduction or elimination of O-FucT-1 levels using gene ablation in mice or interfering RNA or mutants in Drosophila suggests that O-FucT-1 is essential for Notch signaling (13Shi S. Stanley P. Proc. Natl. Acad. Sci. U. S. A. 2003; 100: 5234-5239Crossref PubMed Scopus (326) Google Scholar, 14Okajima T. Irvine K.D. Cell. 2002; 111: 893-904Abstract Full Text Full Text PDF PubMed Scopus (326) Google Scholar, 15Sasamura T. Sasaki N. Miyashita F. Nakao S. Ishikawa H.O. Ito M. Kitagawa M. Harigaya K. Spana E. Bilder D. Perrimon N. Matsuno K. Development (Camb.). 2003; 130: 4785-4795Crossref PubMed Scopus (147) Google Scholar). In vitro binding studies suggest that the O-fucose plays an essential role in ligand binding, offering a potential explanation for the essential nature of O-FucT-1 (16Okajima T. Xu A. Irvine K.D. J. Biol. Chem. 2003; 278: 42340-42345Abstract Full Text Full Text PDF PubMed Scopus (170) Google Scholar). TSRs also contain six conserved cysteines forming three disulfide bonds in a distinct pattern, C1-C5, C2-C6, C3-C4 (17Adams J.C. Tucker R.P. Dev. Dyn. 2000; 218: 280-299Crossref PubMed Scopus (265) Google Scholar). Although a consensus sequence for O-fucose in TSRs has been proposed (WX5C1X2/3S/TC2X2G) (18Gonzalez de Peredo A. Klein D. Macek B. Hess D. Peter-Katalinic J. Hofsteenge J. Mol. Cell. Proteomics. 2002; 1: 11-18Abstract Full Text Full Text PDF PubMed Scopus (103) Google Scholar), little is known about the functional importance of this modification. The biological significance of fucosylation in humans has been made clear by studies of a rare autosomal recessive human disease called LADII (leukocyte adhesion deficiency)/CDGIIc (congenital disorder of glycosylation), which is characterized by leukocyte adhesion deficiency as well as severe neurological and developmental abnormalities (19Etzioni A. Frydman M. Pollack S. Avidor I. Phillips M.L. Paulson J.C. Gershoni-Baruch R. N. Engl. J. Med. 1992; 327: 1789-1792Crossref PubMed Scopus (448) Google Scholar). The molecular defect in LADII/CDGIIc has been identified as mutations in the Golgi GDP-fucose transporter, resulting in decreased GDP-fucose levels in the Golgi lumen and, hence, reduction of fucosylated antigens on the cell surface (20Sturla L. Puglielli L. Tonetti M. Berninsone P. Hirschberg C.B. De Flora A. Etzioni A. Pediatr. Res. 2001; 49: 537-542Crossref PubMed Scopus (51) Google Scholar, 21Etzioni A. Sturla L. Antonellis A. Green E.D. Gershoni-Baruch R. Berninsone P.M. Hirschberg C.B. Tonetti M. Am. J. Med. Genet. 2002; 110: 131-135Crossref PubMed Scopus (66) Google Scholar, 22Lubke T. Marquardt T. von Figura K. Korner C. J. Biol. Chem. 1999; 274: 25986-25989Abstract Full Text Full Text PDF PubMed Scopus (119) Google Scholar, 23Lubke T. Marquardt T. Etzioni A. Hartmann E. von Figura K. Korner C. Nat. Genet. 2001; 28: 73-76Crossref PubMed Scopus (288) Google Scholar). Because Notch signaling is critical to numerous developmental processes in mammals, we had speculated that the neurologic and developmental symptoms of LADII/CDGIIc patients might be linked to impaired Notch signaling because of insufficient GDP-fucose in the Golgi lumen (24Moloney D.J. Shair L. Lu F.M. Xia J. Locke R. Matta K.L. Haltiwanger R.S. J. Biol. Chem. 2000; 275: 9604-9611Abstract Full Text Full Text PDF PubMed Scopus (291) Google Scholar, 25Moloney D.J. Panin V.M. Johnston S.H. Chen J. Shao L. Wilson R. Wang Y. Stanley P. Irvine K.D. Haltiwanger R.S. Vogt T.F. Nature. 2000; 406: 369-375Crossref PubMed Scopus (723) Google Scholar). A recent study using fibroblasts from LADII/CDGIIc patients suggested that although terminal fucosylation of N-glycans is reduced severely, protein O-fucosylation is generally unaffected (26Sturla L. Rampal R. Haltiwanger R.S. Fruscione F. Etzioni A. Tonetti M. J. Biol. Chem. 2003; 278: 26727-26733Abstract Full Text Full Text PDF PubMed Scopus (51) Google Scholar). These authors (26Sturla L. Rampal R. Haltiwanger R.S. Fruscione F. Etzioni A. Tonetti M. J. Biol. Chem. 2003; 278: 26727-26733Abstract Full Text Full Text PDF PubMed Scopus (51) Google Scholar) proposed two potential explanations for this surprising result. One possibility is that O-fucosyltransferases have higher affinity for GDP-fucose than those adding terminal fucose to N-glycans. Alternatively, it is possible that O-fucosyltransferases localize to subcellular compartments other than the Golgi apparatus with a different GDP-fucose transporter. Although O-FucT-1 has been proposed as a Golgi resident protein (9Wang Y. Shao L. Shi S. Harris R.J. Spellman M.W. Stanley P. Haltiwanger R.S. J. Biol. Chem. 2001; 276: 40338-40345Abstract Full Text Full Text PDF PubMed Scopus (211) Google Scholar), no experimental evidence has been put forward to support this idea. Here we have examined the subcellular localization of O-FucT-1 protein and the EGF O-fucosylation reaction. Surprisingly, we found that O-FucT-1 is a soluble protein localized to the ER. Moreover, O-FucT-1 appears to catalyze the O-fucosylation reaction in the ER rather than in the Golgi apparatus. Materials—COS1 cells were from the American Type Culture Collection (Manassas, VA). They were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Fractions from sucrose gradients of microsomes derived from rat liver were a kind gift from Drs. Michael Jadot (Facultés Universitaires Notre-Dame de la Paix) and Peter Lobel (Rutgers University). The procedures for preparation of the sucrose gradients were described previously (27Naureckiene S. Sleat D.E. Lackland H. Fensom A. Vanier M.T. Wattiaux R. Jadot M. Lobel P. Science. 2000; 290: 2298-2301Crossref PubMed Scopus (710) Google Scholar). All other reagents were of the highest quality available. Plasmid Construction—Constructs used for expressing human O-FucT-1 with or without the RDEF sequence were prepared as follows. For pSecTag-HIS6-O-FucT-1, sequences encoding the human O-FucT-1 amino acids 23–388 were amplified by PCR with 5′-primer (5′-CCCAAGCTTCCCATCATCATCATCATCATCCCGCGGGCTCCTGGGACCCG-3′, includes sequences encoding the His6 tag and a Hind III restriction site) and 3′-primer (5′-CGGGGGCCCTTAGAACTCGTCCCGCAGCTTAGG-3′, includes sequences encoding a stop codon and an ApaI site). For pSecTag-HIS6-O-FucT-1ΔRDEF, sequences encoding the O-FucT-1 amino acids 23–384 were amplified by PCR with the same 5′-primer but with the 3′-primer (5′-CGGGGGCCCTTACAGCTTAGGGGGCCTGTCCATG-3′, includes stop codon and ApaI site). The amplified fragments were then subcloned into the pSecTag2c expression vector (Invitrogen) between the Hind III and ApaI sites in-frame with the sequence encoding the signal peptide. All constructs were verified by sequencing analysis. Preparation of Soluble and Membrane Fractions from COS1 Cells— COS1 cells were grown to confluence, collected by scraping, and washed three times with TBS (10 mm Tris-HCl, pH 7.5, 0.15 m NaCl). The washed cells were resuspended in 1 ml of TBS with the protease inhibitor mixture (Roche Applied Science) and sonicated using a probe sonicator (5 × 10-s bursts on ice) to homogenize the cells. The homogenate was subjected to ultracentrifugation at 250,000 × g for 1 h at 4 °C. The supernatant (soluble fraction) was removed, and Nonidet P-40 was added to a final concentration of 1%. The pellet (membrane fraction) was solubilized with an equivalent volume of TBS containing 1% Nonidet P-40 and the protease inhibitor mixture. Immunofluorescent Staining—COS1 cells growing on coverslips in 35-mm dishes were transiently transfected with plasmids using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. After transfection (24 h), cells were fixed in 3% paraformaldehyde in phosphate-buffered saline and permeabilized with 0.1% Triton X-100 in phosphate-buffered saline. Cells were then stained with mixtures of the following primary antibodies: mouse anti-His6 antibody (1:100) (Santa Cruz Biotechnology) and rabbit anti-calreticulin (1:500) (Stressgen); rabbit anti-His6 (1:100) (Santa Cruz Biotechnology) and mouse anti-GM130 (1:100) (BD Biosciences). Alexa fluor 488- and 568-conjugated secondary antibodies (Molecular Probes) were used at 1:100. Fluorescent images were collected using a Leica TCS SP2 confocal microscope. Analysis of O-Fucosylation on Mouse Notch1 EGF29 –36 —Preparation of the plasmid encoding the EGF repeats 29–36 from mouse Notch1 with C-terminal Myc-His6 tags (pSecTag-EGF29–36), transfection, and metabolic radiolabeling with [3H]fucose is described elsewhere (11Shao L. Moloney D.J. Haltiwanger R.S. J. Biol. Chem. 2003; 278: 7775-7782Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). After transfection (24 h), COS1 cells were metabolically radiolabeled for another 24 h. The media were removed and stored. The cells were washed three times with phosphate-buffered saline and lysed in radioimmune precipitation assay buffer (50 mm Tris-HCl, pH8.0, 150 mm NaCl, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS) containing the protease inhibitor mixture. EGF29–36 was purified from the cell lysates by immunoprecipitation using a rabbit anti-Myc antibody (Abcam) and protein A-Sepharose, whereas EGF29–36 in the media was purified by nickel-nitrilotriacetic acid chromatography as described previously (11Shao L. Moloney D.J. Haltiwanger R.S. J. Biol. Chem. 2003; 278: 7775-7782Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). PNGase F digestion was performed on all the samples as described previously (24Moloney D.J. Shair L. Lu F.M. Xia J. Locke R. Matta K.L. Haltiwanger R.S. J. Biol. Chem. 2000; 275: 9604-9611Abstract Full Text Full Text PDF PubMed Scopus (291) Google Scholar) and endoglycosidase H (Roche Applied Science) digestion was carried out following the manufacturer's protocol. Western blots and fluorography were then performed as described (24Moloney D.J. Shair L. Lu F.M. Xia J. Locke R. Matta K.L. Haltiwanger R.S. J. Biol. Chem. 2000; 275: 9604-9611Abstract Full Text Full Text PDF PubMed Scopus (291) Google Scholar). Immunofluorescence of COS1 cells expressing EGF29–36 was performed as described under Immunofluorescent Staining. Other Methods—β4GalT assays and O-fucosyltransferase assays were performed as described previously (9Wang Y. Shao L. Shi S. Harris R.J. Spellman M.W. Stanley P. Haltiwanger R.S. J. Biol. Chem. 2001; 276: 40338-40345Abstract Full Text Full Text PDF PubMed Scopus (211) Google Scholar, 28Moloney D.J. Haltiwanger R.S. Glycobiology. 1999; 9: 679-687Crossref PubMed Scopus (44) Google Scholar). Endogenous O-FucT-1 Appears to Be a Soluble Protein in the ER—In the original report of O-FucT-1 purification, the enzyme was purified from a detergent-free extract of Chinese hamster ovary (CHO) cells (8Wang Y. Spellman M.W. J. Biol. Chem. 1998; 273: 8112-8118Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar). N-terminal sequencing of the purified protein was performed, and the sequence was used to generate probes to isolate the gene encoding O-FucT-1 from a human heart cDNA library (9Wang Y. Shao L. Shi S. Harris R.J. Spellman M.W. Stanley P. Haltiwanger R.S. J. Biol. Chem. 2001; 276: 40338-40345Abstract Full Text Full Text PDF PubMed Scopus (211) Google Scholar). Interestingly, the predicted sequence of the human and mouse enzymes began ∼20 residues N-terminal to the sequence derived from the CHO cell enzyme (Fig. 1A). Originally, we believed that this region encoded a short cytoplasmic domain and a transmembrane sequence, typical of the type II membrane structure found in all known mammalian Golgi glycosyltransferases (29Kleene R. Berger E.G. Biochim. Biophys. Acta. 1993; 1154: 283-325Crossref PubMed Scopus (202) Google Scholar). Proteolytic loss of such sequences during purification has been reported previously (30Kitazume S. Tachida Y. Oka R. Shirotani K. Saido T.C. Hashimoto Y. Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 13554-13559Crossref PubMed Scopus (235) Google Scholar). Re-examination of the N-terminal sequence from mouse and human O-FucT-1 using SMART (smart.emblheidelberg.de/) or other sequence analysis tools suggests that the previously predicted transmembrane sequence is actually a cleavable signal peptide (Fig. 1A). Moreover, the sequenced N terminus of purified O-FucT-1 begins at approximately the predicted cleavage site for the signal peptide (Fig. 1A). To confirm that O-FucT-1 is a soluble protein, COS1 cells were lysed by sonication, and the activity of O-FucT-1 was assayed in high speed supernatant and pellet fractions. The majority of O-FucT-1 activity was found in the supernatant (Fig. 1B). As a control, we also compared the enzyme activity of a known membrane-bound protein, β4GalT, and found the majority of its activity in the pellet fraction (Fig. 1B). These results strongly suggest that endogenous O-FucT-1 is a soluble protein. All known fucosyltransferases responsible for terminal fucose modification in N-linked glycans are believed to be localized to the Golgi, consistent with the localization of the GDP-fucose transporter (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar, 31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar, 32Sousa V.L. Brito C. Costa J. Biochim. Biophys. Acta. 2004; 1675: 95-104Crossref PubMed Scopus (14) Google Scholar, 33Milland J. Taylor S.G. Dodson H.C. McKenzie I.F. Sandrin M.S. J. Biol. Chem. 2001; 276: 12012-12018Abstract Full Text Full Text PDF PubMed Scopus (37) Google Scholar). However, the subcellular localization of O-FucT-1 is still unclear. Interestingly, N-glycans in purified O-FucT-1 from CHO cells are mostly high mannose-type oligosaccharide chains, suggesting that this enzyme may be localized in either the ER or the cis-Golgi region (8Wang Y. Spellman M.W. J. Biol. Chem. 1998; 273: 8112-8118Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar). We sought to determine the localization of endogenous O-FucT-1 by subcellular fractionation of rat liver. Rat liver microsomes were fractionated on a sucrose gradient and assayed for O-FucT-1 activity as well as markers for the ER (glucose-6-phosphatase) and the Golgi (β4GalT). The majority of the O-FucT-1 activity co-fractionates with the ER marker (glucose-6-phosphatase) but not with the Golgi marker (β4GalT1) (Fig. 1C). This result suggests that endogenous O-FucT-1 localizes to the ER. O-FucT-1 Homologues Have a C-terminal KDEL-like Sequence Retaining the Protein in the ER—Comparison of C-terminal amino acids residues from mouse, human, and Drosophila O-FucT-1 reveals the presence of KDEL-like ER retention/retrieval motifs in all three homologues (Fig. 2A) (34Murshid A. Presley J.F. Cell. Mol. Life Sci. 2004; 61: 133-145Crossref PubMed Scopus (81) Google Scholar). We generated the expression plasmids pSecTag-HIS6-O-FucT-1 and pSecTag-HIS6-O-FucT-1ΔRDEF, which encode N-terminal His6-tagged human O-FucT-1 with or without the KDEL-like motif, RDEF (Fig. 2B). These two plasmids as well as the empty pSecTag plasmid were transiently transfected into COS1 cells. Western blot analysis of the media and lysate from the transfected cells using the anti-His6 tag antibody revealed that the O-FucT-1 with the RDEF motif is largely retained in the cells, whereas the O-FucT-1 without the RDEF motif is largely secreted to the media (Fig. 2C). To preclude the possibility that the observed difference in secretion was due to some effect on protein folding by the RDEF sequence, we performed activity assays on the lysates of cells expressing O-FucT-1 and the media of cells expressing O-FucT-1ΔRDEF. Both HIS6-O-FucT-1 and HIS6-O-FucT-1ΔRDEF are active, indicating that O-FucT-1 is folded correctly and active despite the presence or absence of the RDEF motif (Fig. 2C). We then carried out immunofluorescent staining to determine the specific subcellular localization of HIS6-O-FucT-1 and HIS6-O-FucT-1ΔRDEF in COS1 cells. The two expression plasmids were transiently transfected into COS1 cells on coverslips followed by staining using the anti-His6 tag antibody along with either the ER marker anti-calreticulin antibody or the cis-Golgi marker anti-GM130 antibody. The HIS6-O-FucT-1 displays a perinuclear and reticular distribution characteristic of the ER, co-localizing with calreticulin but not with GM130 (Fig. 3, A–F). In contrast, deletion of RDEF results in a dramatic redistribution of cell-associated O-FucT-1 including loss of co-localization with the calreticulin (Fig. 3, G–I) and partial localization with the GM130 (Fig. 3, J–L). The redistribution caused by deletion of the RDEF motif was even more dramatically observed in the Western blot analysis shown in Fig. 2C, demonstrating that the majority of HIS6-O-FucT-1ΔRDEF is secreted. Taken together, the results suggest that the C-terminal KDEL-like sequence functions as an ER retention/retrieval signal for O-FucT-1. Addition of O-Fucose to Proteins Appears to Occur in the ER—Because GDP-fucose transporter activity has only been reported in Golgi fractions (31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar), the localization of O-FucT-1 to the ER raised the question of where the addition of O-fucose to proteins occurs. To examine the subcellular site of O-fucosylation we used a previously reported plasmid encoding the EGF repeats 29–36 from mouse Notch1, with C-terminal Myc and His6 tags, that is secreted to the media (11Shao L. Moloney D.J. Haltiwanger R.S. J. Biol. Chem. 2003; 278: 7775-7782Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). We chose this fragment because it has two predicted N-glycan sites as well as five predicted O-fucose sites (Fig. 4A), and we have previously characterized its O-fucosylation (11Shao L. Moloney D.J. Haltiwanger R.S. J. Biol. Chem. 2003; 278: 7775-7782Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). Although EGF29–36 is secreted to the media, some of the protein can also be isolated from cell lysates. Immunostaining of COS1 cells transiently transfected with the EGF29–36 plasmid demonstrated that cell-associated EGF29–36 co-localizes with the ER marker calreticulin but not with the Golgi marker GM130 (Fig. 4, B–G). This indicates that the vast majority of EGF29–36 within cells is in the ER. Control experiments demonstrated that more than 90% of the cell-associated EGF29–36 could be extracted from the cells using a radioimmune precipitation assay lysis buffer (data not shown). To examine O-fucosylation of EGF29–36, we metabolically radiolabeled COS1 cells expressing EGF29–36 with [3H]fucose. Media and cells were separated, and EGF29–36 was purified from both (see "Experimental Procedures"). Samples were digested with endoglycosidase H and PNGase F to evaluate the processing state of the attached N-glycans. Western blot and fluorographic analysis demonstrated that EGF29–36 retained in the cells is sensitive to endoglycosidase H digestion, whereas secreted EGF29–36 is resistant (Fig. 4H, lanes 1, 2, 5, and 6). This is consistent with the immunofluorescent staining data (Fig. 4, B–D) suggesting that cell-associated EGF29–36 is ER-localized. Surprisingly, the protein from the cells is still modified with [3H]fucose following endoglycosidase H digestion, suggesting that the ER form of the protein is modified with O-fucose (Fig. 4H, lane 2). EGF29–36 from both cells and media was sensitive to PNGase F digestion as expected. Digestion of EGF29–36 from media with PNGase F showed a significant reduction in the amount of [3H]fucose associated with the protein (Fig. 4H, compare lane 7 with lane 8) suggesting the presence of fucose in N-glycans being released by PNGase F. This is consistent with the fucosylation of N-glycans in the Golgi apparatus, which is well known (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar, 31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar, 32Sousa V.L. Brito C. Costa J. Biochim. Biophys. Acta. 2004; 1675: 95-104Crossref PubMed Scopus (14) Google Scholar, 33Milland J. Taylor S.G. Dodson H.C. McKenzie I.F. Sandrin M.S. J. Biol. Chem. 2001; 276: 12012-12018Abstract Full Text Full Text PDF PubMed Scopus (37) Google Scholar). Finally, comparison of the stoichiometry of fucosylation after removal of the N-glycans with PNGase F (compare lane 4 with lane 8, Fig. 4H) shows similar levels of [3H]fucose incorporated (based on fluorography) into similar amounts of protein (based on the Western blot). Because the vast majority of the cell-associated EGF29–36 is in the ER (based on immunofluorescence, Fig. 4, B–D, and endoglycosidase H sensitivity, Fig. 4H, lanes 1 and 2), these results strongly suggest that O-fucosylation occurs in the ER. The recent studies on fucosylation in LADII/CDGIIc patients raised the intriguing question of how fucosylation of N-glycans could be almost completely eliminated whereas O-fucosylation remained unaffected (26Sturla L. Rampal R. Haltiwanger R.S. Fruscione F. Etzioni A. Tonetti M. J. Biol. Chem. 2003; 278: 26727-26733Abstract Full Text Full Text PDF PubMed Scopus (51) Google Scholar). Together with the previous report that O-FucT-1 bears N-glycans sensitive to endoglycosidase H (consistent with an ER or cis-Golgi localization) these results motivated us to examine the localization of O-FucT-1 in more detail. Using subcellular fractionation, we found that endogenous O-FucT-1 is a soluble protein localized in the ER. We next showed that the C-terminal KDEL-like motif (RDEF) of O-FucT-1 is responsible for retaining the protein in the ER. Immunofluorescent staining clearly demonstrated that expressed O-FucT-1 with the RDEF motif co-localizes with the ER marker calreticulin. Although we cannot completely rule out the possibility that some O-FucT-1 might localize within the Golgi, the amount appears to be relatively insignificant. Removal of the RDEF domain causes a dramatic change in O-FucT-1 localization. The majority of O-FucT-1ΔRDEF is secreted, and the small amount still associated with cells no longer co-localizes with calreticulin. These results strongly suggest that RDEF functions as an ER retention/retrieval signal. We next addressed where O-FucT-1 catalyzes the O-fucosylation reaction. It is reasonable to speculate that the reaction occurs in the same compartment as the bulk of the enzyme, but this is in conflict with the fact that GDP-fucose transporter activity has historically been associated with the Golgi apparatus (31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar). Immunofluorescent staining of COS1 cells transfected with a plasmid encoding mouse Notch1 EGF29–36 revealed that the vast majority of cell-associated EGF29–36 is localized to the ER. The N-glycans in this protein were sensitive to endoglycosidase H digestion, consistent with the ER localization. Surprisingly, this protein was also modified with O-fucose. Similar levels of fucosylation in EGF29–36 from cells (ER-localized) and from the media indicated that O-fucosylation occurs in the ER. It is possible that EGF29–36 could be O-fucosylated in the cis-Golgi and cycle back to the ER. This possibility would require that the bulk of EGF29–36 localized to the ER have already cycled through the cis-Golgi. Because EGF29–36 is being secreted and the net flow of the protein is through the Golgi, we think that this possibility is unlikely. Thus, we conclude that O-FucT-1 catalyzes the O-fucosylation reaction in the ER in contrast to all other known forms of fucosylation, which occur in trans-Golgi or the trans-Golgi network (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar). GDP-fucose is the donor substrate for all known fucosyltransferases, including O-FucT-1 (1Becker D.J. Lowe J.B. Glycobiology. 2003; 13: 41R-53RCrossref PubMed Scopus (693) Google Scholar). GDP-fucose is synthesized in the cytoplasm and is known to be transported into the Golgi lumen by a GDP-fucose transporter (31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar). In order for O-FucT-1 to catalyze O-fucosylation in the ER lumen, some mechanism must exist that facilitates transport of GDP-fucose from the Golgi apparatus or cytoplasm. The studies on LADII/CDGIIc fibroblasts suggest that the Golgi of cells from these patients have insufficient GDP-fucose to account for a mechanism involving retrograde transport of GDP-fucose from the Golgi to the ER (14Okajima T. Irvine K.D. Cell. 2002; 111: 893-904Abstract Full Text Full Text PDF PubMed Scopus (326) Google Scholar). This would suggest that there is an unidentified GDP-fucose transporter localized in the ER. Historically, transporter assays using vesicles isolated from the ER or the Golgi have been used to demonstrate the presence of nucleotide sugar transporters (31Hirschberg C.B. Robbins P.W. Abeijon C. Annu. Rev. Biochem. 1998; 67: 49-69Crossref PubMed Scopus (309) Google Scholar). GDP-fucose transporter activity has been detected in Golgi-derived vesicles, but no evidence of transporter activity in ER-derived vesicles has been reported. Our results predict that such a transporter may exist. Nonetheless, further work is necessary to evaluate whether GDP-fucose enters the ER through a novel transporter or a retrograde transport system. The C-terminal KDEL-like sequence, RDEF, of human O-FucT-1 functions as an ER retention/retrieval signal. Similar sequences are found at the C terminus of all known homologues of O-FucT-1 (Fig. 2A and data not shown). Aside from mammalian O-FucT-1, Drosophila O-FucT-1 has also been shown to localize to the ER. 3T. Okajima and K. D. Irvine, personal communication. Interestingly, deletion of the C terminus of O-FucT-1 in Drosophila causes a lethal phenotype, demonstrating the biological importance of ER retention/retrieval (16Okajima T. Xu A. Irvine K.D. J. Biol. Chem. 2003; 278: 42340-42345Abstract Full Text Full Text PDF PubMed Scopus (170) Google Scholar). Thus, the ER localization appears to be critical to the function of O-FucT-1. These results raise the question of why O-FucT-1 localizes to the ER. O-FucT-1 is known to recognize and fucosylate properly folded EGF repeats (8Wang Y. Spellman M.W. J. Biol. Chem. 1998; 273: 8112-8118Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar). Thus, O-FucT-1 has the ability to distinguish between folded and unfolded EGF repeats. This observation, together with the ER localization of O-FucT-1, indicates that O-fucosylation may serve as a quality control signal in the ER. Several possible models exist. For instance unknown escort proteins may recognize O-fucose and facilitate the transportation from the ER to the Golgi in a fashion analogous to recognition of N-glycans by ERGIC53 (34Murshid A. Presley J.F. Cell. Mol. Life Sci. 2004; 61: 133-145Crossref PubMed Scopus (81) Google Scholar). Alternatively, a retention protein may recognize and trap proteins with unfucosylated EGF repeats in the ER. In either case, ER localization of O-FucT-1 would be essential. Another possibility is that the O-fucose could prevent the undesired aggregation of proteins containing EGF repeats, facilitating the exit of such proteins from the ER to the Golgi. Obviously, further experiments are required to better understand the unique localization of this enzyme. It is not yet clear to what extent a folding defect contributes to the embryonic lethality observed in O-FucT-1 knock-outs or mutants (13Shi S. Stanley P. Proc. Natl. Acad. Sci. U. S. A. 2003; 100: 5234-5239Crossref PubMed Scopus (326) Google Scholar, 14Okajima T. Irvine K.D. Cell. 2002; 111: 893-904Abstract Full Text Full Text PDF PubMed Scopus (326) Google Scholar, 15Sasamura T. Sasaki N. Miyashita F. Nakao S. Ishikawa H.O. Ito M. Kitagawa M. Harigaya K. Spana E. Bilder D. Perrimon N. Matsuno K. Development (Camb.). 2003; 130: 4785-4795Crossref PubMed Scopus (147) Google Scholar). Nonetheless, these findings do not alter the fact that the O-fucose modification is required for fringe-mediated modulation of Notch function (4Haltiwanger R.S. Lowe J.B. Annu. Rev. Biochem. 2004; 73: 491-537Crossref PubMed Scopus (649) Google Scholar, 5Haines N. Irvine K.D. Nat. Rev. Mol. Cell Biol. 2003; 4: 786-797Crossref PubMed Scopus (338) Google Scholar) or for the suggested role O-fucose plays in receptor-ligand binding (16Okajima T. Xu A. Irvine K.D. J. Biol. Chem. 2003; 278: 42340-42345Abstract Full Text Full Text PDF PubMed Scopus (170) Google Scholar). Although O-fucosylation of TSRs was not examined here, the studies on fibroblasts from LADII/CDGIIc patients suggest that O-fucosylation of TSRs is also unaffected. We have recently identified the gene encoding O-FucT-2, the enzyme responsible for addition of O-fucose to TSRs.2 Because the O-fucose level in TSRs from LADII/CDGIIc fibroblasts appears to be unaffected, it is likely that O-FucT-2 also localizes to the ER. Studies to examine this possibility are in progress. We thank Drs. Peter Lobel and Michael Jadot for their generous provision of sucrose gradient fractions of rat liver microsomes. We also thank Kelvin Luther and Malgosia Skowron for assistance in generating constructs and Drs. Pamela Stanley, Ken Irvine, and members of the Haltiwanger laboratory for helpful discussions.