Protein O-Fucosyltransferase 2 Adds O-Fucose to Thrombospondin Type 1 Repeats*

O-Fucose is an unusual form of glycosylation found on epidermal growth factor-like (EGF) repeats and thrombospondin type 1 repeats (TSRs) in many secreted and transmembrane proteins. Recently O-fucose on EGF repeats was shown to play important roles in Notch signaling. In contrast, physiological roles for O-fucose on TSRs are unknown. In the accompanying paper (Luo, Y., Nita-Lazar, A., and Haltiwanger, R. S. (2006) J. Biol. Chem. 281, 9385–9392), we demonstrated that an enzyme distinct from protein O-fucosyltransferase 1 adds O-fucose to TSRs. A known homologue of O-fucosyltransferase 1 is putative protein O-fucosyltransferase 2. The cDNA sequence encoding O-fucosyltransferase 2 was originally identified during a data base search for fucosyltransferases in Drosophila. Like O-fucosyltransferase 1, O-fucosyltransferase 2 is conserved from Caenorhabditis elegans to humans. Although O-fucosyltransferase 2 was assumed to be another protein O-fucosyltransferase, no biochemical characterization existed supporting this contention. Here we show that RNAi-mediated reduction of the O-fucosyltransferase 2 message significantly decreased TSR-specific O-fucosyltransferase activity in Drosophila S2 cells. We also found that O-fucosyltransferase 2 is predominantly localized in the endoplasmic reticulum compartment of these cells. Furthermore, we expressed recombinant Drosophila O-fucosyltransferase 2 and showed that it O-fucosylates TSRs but not EGF repeats in vitro. These results demonstrate that O-fucosyltransferase 2 is in fact a TSR-specific O-fucosyltransferase.

Fucose is typically found as a terminal modification in glycoconjugates where it plays a number of important biological roles. Fucose is a component of the A, B, O blood group antigens, as well as the Sialyl Lewis x epitope essential for selectin-mediated extravasation of leukocytes (1). Recent studies have shown that ablation of FUT8, the enzyme responsible for addition of core ␣1,6-linked fucose to N-glycans, results in severe growth retardation and early postnatal death in mice, apparently because of alterations in a number of signaling pathways including those initiated by transforming growth factor ␤1 and epidermal growth factor (2,3).
Fucose also exists in non-terminal positions in glycoconjugates, in direct O-linkage to serine or threonine residues (O-fucose) within two different types of cysteine-knot motifs: epidermal growth factor-like (EGF) 3 repeats (4) and thrombospondin type 1 repeats (TSR) (5,6). The enzyme responsible for adding O-fucose to EGF repeats has been identified as protein O-fucosyltransferase 1 (POFUT1 in humans and OFUT1 in Drosophila) (7,8). Elimination (or reduction) of Pofut1 in mice or Ofut1 in Drosophila results in severe Notch-like phenotypes, indicating that O-fucose modifications are essential for Notch function in both Drosophila and mammalian systems (8 -10).
No clear function for addition of O-fucose to TSR has yet been reported. To begin investigating the biological implications of TSR O-fucosylation, we have initiated studies on the enzymes responsible for addition of O-fucose glycans to TSR. In the accompanying article (42), we demonstrated that POFUT1 is not capable of adding O-fucose to TSR. This finding led us to propose the presence of a novel TSR-specific O-fucosyltransferase. Using the assay we developed for this enzyme, we showed that TSR-specific O-fucosyltransferase is similar to POFUT1 in several respects. Both enzymes utilize GDP-fucose as donor substrate and require properly folded acceptor substrates (either EGF repeats or TSRs) for activity (11,42). In addition, modification of both EGF repeats and TSRs appears to be conserved in vertebrates and invertebrates (5,12). Based on these observations, we reasoned that the TSR-specific O-fucosyltransferase might be a homologue of POFUT1 conserved in invertebrates and vertebrates.
All fucosyltransferases identified or predicted to date can be divided into four groups according to their preference for acceptor substrates: ␣1,2-fucosyltransferases, ␣1,3/4-fucosyltransferases, ␣1,6-fucosyltransferases, and protein O-fucosyltransferases (1,13). The protein O-fucosyltransferase group includes POFUT1 and a closely homologous gene family, POFUT2 (abbreviated Ofut2 in Drosophila), which was originally identified in a data base search for genes involved in fucose metabolism in Drosophila (14). In that study, only human and Drosophila genes (POFUT2 and Ofut2, respectively) were identified. In a subsequent study, POFUT2 orthologs were found in many other metazoan species (13). Although O-fucosyltransferase 2 is assumed to be an O-fucosyltransferase based on sequence similarity, no biochemical demonstration of activity has been reported for this protein family. Here we characterized the biochemical activity of Drosophila OFUT2. We show that OFUT2 is a predominantly ER-localized enzyme that possesses TSR-specific O-fucosyltransferase activity. * This work was supported in part by grants from the Mizutani Foundation for Glyco-

EXPERIMENTAL PROCEDURES
Expression Constructs-A construct for the expression of protein A-tagged Drosophila OFUT2 protein was prepared by in-frame ligation of a PCR-amplified DNA fragment encoding the IgG-binding domain of Staphylococcus aureus protein A from plasmid pRL715 (a gift from Dr. K. Severinov (15)) with the PCR-amplified DNA fragment of cDNA clone GH07929 encoding amino acids 63-490 of Drosophila Ofut2 (gene CG14789, GenBank accession number AY047568). The resulting DNA was inserted into pMT/BiP/V5-HisA vector (Invitrogen) in-frame with the BiP signal sequence (16) using BglII and AgeI restriction sites. The final construct encoded a Drosophila OFUT2 fusion protein in which the first 62 amino acids, including the endogenous signal peptide and part of the N-terminal region, were replaced with BiP signal peptide followed by the IgGbinding domain of S. aureus protein A. A construct for the expression of HA-tagged full-length Drosophila OFUT2 protein was prepared by: (i) introducing an SpeI site immediately after the last codon of the Ofut2 open reading frame by PCR; (ii) in-frame ligation of double-stranded oligonucleotide encoding the HA-His 6 tag and a stop codon (17); and (iii) subcloning the resulting DNA construct into the pRmHA-3 expression vector (18).
Cell Culture and Protein Expression-Drosophila S2 cells were maintained and transfected with the Drosophila Ofut2 expression construct as described (19). For purification of protein A-tagged OFUT2 protein, we collected medium from transfected S2 cells 24 h after the induction of protein expression with 0.7 mM CuSO 4 . The collected medium was incubated with IgG-Sepharose beads (Amersham Biosciences) on a rotator at 4°C overnight. Afterward, the beads were extensively washed with TBS (10 mM Tris-HCl, 0.15 M NaCl, pH 7.5) containing protease inhibitors (Complete mixture, Roche) and analyzed by Western blotting and O-fucosyltransferase assays.
RNAi-mediated Silencing of Ofut1 and Ofut2 in Drosophila S2 Cells-For RNAi experiments, the dsRNA corresponding to the coding region of a specific gene was synthesized using the MEGAscript RNAi kit (Ambion). As templates for dsRNA synthesis, we used ϳ700-bp DNA fragments of Drosophila Ofut2, Ofut1, and CG9220 cDNAs produced by PCR and the following primer pairs that included a T7 RNA polymerase-binding site (designated as {T7}): Ofut2, 5Ј-{T7}CGCCGTGTATA-TCCTTTACGA-3Ј and 5Ј-{T7}GAAGCATCTGTGGTGTCCAG-3Ј; CG9220, 5Ј-{T7}CCAGGCACATGTTGTACTGC-3Ј and 5Ј-{T7}GA-CCTCCTGGTGGTGATCTTC-3Ј; Ofut1, 5Ј-{T7}aggaataccatcgcgtcatc-3Ј and 5Ј-{T7 }cgactaagggccgtgtttag-3Ј; where {T7} is GAATTAAT-ACGACTCACTATAGGGAGA. The CG9220 gene encoding a putative glycosyltransferase of the ␤3GT superfamily (see Flybase and Ref. 20) was used as an irrelevant gene control. Drosophila Ofut1 cDNA was a gift from Dr. Kenneth Irvine (Rutgers). The primers were designed using the E-RNAi program (21). Annealing, nuclease treatment, and purification of dsRNA was performed according to the manufacturer's protocol (Ambion). Drosophila S2 cells were treated with dsRNA as described in Ref. 22 with the following modifications. Culture cells were plated onto a 60-mm culture dish in 6 ml of complete M3 medium (Shield and Sang M3 medium (Sigma) with 5 g/liter of yeast extract, 12.5 g/liter of Bacto-peptone, 12.5% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 250 ng/ml Fungizone) at a final concentration of 1 ϫ 10 6 cells/ml. After cells attached to the dishes, the complete medium was replaced with 2 ml of serum-free M3 medium without antibiotics and 60 g of dsRNA was added to the dish, except for the mock control. As a control, a separate dish of cells was prepared following the same procedures but without addition of dsRNA. The cells were incubated with dsRNA for 6 h followed by replacing the media with 6 ml of complete fetal bovine serum-containing M3 media. Two days after dsRNA transfection, the cells were diluted with 9 ml of complete medium, transferred to 10-cm plates, and then let grow for another day. The cell numbers were counted 3 days after transfection and the cells were harvested immediately by low-speed centrifugation (800 ϫ g, 5 min). The cell pellet was washed 2 times with TBS and kept frozen on dry ice until in vitro assays.
Immunolocalization Experiments-Drosophila S2 cells were transfected or co-transfected with plasmids using the calcium phosphate-DNA coprecipitation protocol (23). As a marker for the ER subcellular compartment, we used a GFP:KDEL construct (24). Immunostaining of OFUT2-expressing S2 cells was performed as described earlier (25). The following primary antibodies and corresponding dilutions were used for immunostaining: rabbit anti-LVA (1:3,000) (a gift from John Sisson, University of Texas, Austin, TX) (26) and rat anti-HA (1:2,000) (Roche). We used the following fluorescent secondary goat antibodies: anti-rat Alexa 488 (1:150) and anti-rabbit Alexa 647 (1:100) (highly cross-absorbed antibodies from Molecular Probes). Digital images were obtained using Zeiss Axioplan 2 fluorescent microscope with the Apo-Tome module for optical sectioning.
Other Methods-The method for making extracts of S2 cells was described previously (27). O-Fucosyltransferase assays were performed as described in the accompanying article (42). (14) were the first to identify OFUT2 as a Drosophila protein with weak sequence similarity to OFUT1. They also reported an orthologue in humans (POFUT2). A more detailed sequence analysis revealed the presence of three conserved peptide motifs in all fucosyltransferases, including O-fucosyltransferase 2 (13). This report also identified O-fucosyltransferase 2 orthologues in many vertebrate and invertebrate species, including human, mouse, chicken, frog, and Caenorhabditis elegans. Fig. 1 shows a Clustal W sequence alignment of the O-fucosyl-transferase 2 orthologues from human, mouse, Drosophila, and C. elegans. The three conserved peptide motifs identified previously are shown. Several features of the sequences were revealed upon analysis using online sequence analysis tools, SMART and SOSUI. All O-fucosyltransferase 2 sequences have an N-terminal signal sequence with no internal transmembrane sequences (Fig. 1), indicating that O-fucosyltransferase 2 is probably a soluble protein like POFUT1. The Clustal W alignment also reveals four cysteines conserved in all four O-fucosyltransferase 2 sequences and three potential N-glycosylation sites conserved in the human and mouse enzymes (Fig. 1). A conserved DXD-like motif (ERE in human, mouse, and Drosophila homologues; DRE in C. elegans homologue) is found near the C terminus (Fig. 1). Frequently seen in many classes of glycosyltransferases including POFUT1, the DXD-like motif is believed to be critical for the proper function of these enzymes (28), although in this case it does not appear to function in metal binding (see Fig. 5).

Sequence Analysis of O-Fucosyltransferase 2-Roos and co-workers
RNAi Knockdown of Ofut2 Reduces TSR O-Fucosyltransferase Activity in S2 Cells-As an initial step to evaluate whether OFUT2 is the TSR O-fucosyltransferase, we used RNAi to knock-down endogenous expression of Ofut2 in Drosophila S2 cells. The effects of the RNAi treatment were evaluated based on the ability of extracts to fucosylate EGF repeats (OFUT1 activity) or TSRs. TSR O-fucosyltransferase activity (using TSP1-TSR3 as substrate, see accompanying paper (42)) was reduced by ϳ90% in cells transfected with the Ofut2 dsRNA compared with S2 cells transfected with an irrelevant control dsRNA (CG9220) and mock transfected S2 cells ( Fig. 2A). In contrast, OFUT1 activity (using factor VII EGF repeat as substrate) was unaffected compared with the irrelevant control. As a control, we also knocked down Ofut1 expression in S2 cells (Fig. 2B). As expected, OFUT1 activity was nearly completely abolished, whereas OFUT2 activity was not affected. These results indicate that OFUT2 is the major TSR O-fucosyltransferase in these cells, and that OFUT1 and OFUT2 catalyze the O-fucosylation of EGF repeats and TSRs, respectively, in a non-redundant manner.
Recombinant Drosophila OFUT2 Has TSR O-Fucosyltransferase Activity-To provide further evidence that OFUT2 is a TSR O-fucosyltransferase, we expressed a portion of the predicted luminal domain of Drosophila OFUT2 in S2 cells. The partial cDNA encoding amino acids 63-490 (lacking the endogenous signal sequence and a stretch of residues not conserved in other OFUT2 homologues, Fig. 1) of Drosophila OFUT2 was cloned into an insect cell expression plasmid in-frame with sequences encoding an N-terminal signal peptide and protein A IgG binding domain (Fig. 3A). S2 cells were transiently transfected with the Ofut2 expression plasmid (or a control empty plasmid), and the fusion protein was isolated from the medium using IgG-Sepharose. O-Fucosyltransferase assays were performed using the protein captured on the IgG-Sepharose with either TSP1-TSR3 (for TSR O-fucosyltransferase activity) or factor VII EGF repeat (for OFUT1 activity) as acceptor substrates. TSR O-fucosyltransferase was found only in samples from cells transfected with the Ofut2 expression plasmid (Fig. 3B). No activity was detected when an EGF repeat was used as acceptor substrate (OFUT1 activity), consistent with the acceptor specificity of the TSR O-fucosyltransferase (42). To rule out the possibility that the TSR O-fucosyltransferase activity is not in OFUT2 but in a different protein that merely associates with or co-purifies with the OFUT2 fusion protein, we generated a mutant form of Drosophila OFUT2 in the putative "DXD" motif ( Fig. 1). Glutamates 442 and 443 were altered to alanines (E442A/ E443A) through site-directed mutagenesis, and the mutant Drosophila OFUT2 was expressed and assayed for TSR O-fucosyltransferase activity. The mutation reduced activity to background levels, indicating that OFUT2 catalyzes the reaction (Fig. 3C). Expression of the wild type and mutant forms of OFUT2 were confirmed by immunoblot analysis with protein A-binding antibody (Fig. 3C, inset). Comparable amounts of a protein species of ϳ60 kDa (consistent with the predicted size of the OFUT2 fusion protein: 59 kDa) were seen from the S2 cells transfected with either Ofut2 expression plasmid (wild type or mutant). Efficient secretion of mutant OFUT2 protein into cell medium suggests proper folding of the protein, which nevertheless, was inactive. In combination with the RNAi data shown in Fig. 2, these results strongly indicate that OFUT2 is a Drosophila form of the TSR O-fucosyltransferase described in the accompanying paper (42).
To confirm that the product of the Drosophila OFUT2 reaction is O-fucosylated TSR, we performed product analysis. Reverse-phase HPLC analysis showed that the radiolabeled assay product co-migrates with TSP1-TSR3, demonstrating that the [ 3 H]fucose is covalently associated with TSP1-TSR3 (Fig. 4A). The [ 3 H]fucose was then released from TSP1-TSR3 by alkali-induced ␤-elimination in the presence of  sodium borohydride. Gel filtration and high-pH anion exchange chromatography analysis showed that the released sugar is the monosaccharide fucitol, demonstrating that the [ 3 H]fucose was directly attached to TSP1-TSR3 in O-linkage (Fig. 4, B and C). These results demonstrate that Drosophila OFUT2 modifies TSP1-TSR3 with O-fucose.
Initial Characterization of Recombinant Drosophila OFUT2-The recombinant Drosophila OFUT2 proved to be fairly heat-labile, so all assays were performed at room temperature (data not shown). In contrast to POFUT1 (7,11), Drosophila OFUT2 activity was not enhanced by Mn 2ϩ , nor impaired by EDTA (Fig. 5A). Several other divalent ions were also tested without effect (data not shown), suggesting that Drosophila OFUT2 does not require divalent metal ions for proper function. The dependence of activity on the concentration of GDP-fucose and TSP1-TSR3 was also examined (Fig. 5, B and C). Lineweaver-Burk plots of the data revealed that the K m for TSP1-TSR3 is ϳ2.5 M, whereas that for GDP-fucose is ϳ4.5 M, comparable with the values obtained for EGF repeats (6 M) and GDP-fucose (4 M), respectively, for POFUT1 (7).
Full-length Drosophila OFUT2 Localizes to the ER in S2 Cells-Initial analysis of subcellular localization was performed by immunolocalization of full-length Drosophila OFUT2 bearing a C-terminal HA tag (OFUT2-HA). The functionality of the OFUT2-HA protein was confirmed by in vitro assays of Drosophila S2 cells transfected with the OFUT2-HA-expressing construct or mock transfected with an empty vector. Assays of lysates from S2 cells expressing OFUT2-HA showed significantly increased OFUT2 activity using TSP1-TSR3 as an acceptor substrate (see "Experimental Procedures"), indicating that the fulllength construct encodes a functional enzyme (Fig. 6A). Immunolocalization of OFUT2-HA showed significant co-localization with KDEL: GFP, an ER marker (24) (Fig. 6, E-G), and a minimal co-localization with LVA, a Golgi marker (26) (Fig. 6, B-D). These results suggest that OFUT2-HA localizes mainly to the ER, although a small amount appears to be in the Golgi.

DISCUSSION
POFUT1 and POFUT2 are two homologous gene families found in a variety of metazoan species that encode POFUT1 and POFUT2, respectively (13). POFUT1 is a well characterized enzyme known to add O-fucose to EGF repeats (7,11). Reduction or elimination of POFUT1 expression (or its homologues) abrogates Notch function in both Drosophila and mice, demonstrating the crucial role it plays in biology. In contrast, POFUT2 was predicted to be an O-fucosyltransferase based on its sequence similarity with POFUT1, although no biochemical evidence showing activity had been reported. In the accompanying paper (42), we demonstrated that an O-fucosyltransferase activity distinct from POFUT1 exists in cells capable of adding O-fucose to TSRs. Here we provide biochemical evidence that Drosophila OFUT2 is a TSR O-fucosyltransferase.
Identification of OFUT2 as a TSR O-fucosyltransferase provides a major tool for evaluation of the physiological function of TSR fucosylation. A recent study of the O-fucosyltransferase 2 homologue in C. elegans provided the first evidence that O-fucosyltransferase 2 may be involved in development (29). This study showed that reduction of C. elegans O-fucosyltransferase 2 by RNAi causes failure of embryos to undergo normal morphogenesis, although this effect was of low penetrance. In contrast, overexpression of O-fucosyltransferase 2 resulted in severe body malformation and abnormal neuronal development in up to 83% of the embryos. To date functional roles of O-fucosyltransferase 2 in Drosophila and mammals remain unknown. Whereas enzymatic activity has not yet been reported for the C. elegans homologue of OFUT2, we have obtained preliminary data showing that mouse POFUT2 is a TSR-specific O-fucosyltransferase (43).
The fact that O-fucose modification of EGF repeats plays an important role in the physical interaction between Notch and its ligands suggests that O-fucose on TSRs may function similarly (30). Like EGF repeats, TSRs are well known to mediate protein-protein interactions (31). Similarly, elongation of O-fucose on TSRs to the disaccharide, Glc-␤1,3-fucose, could modulate these interactions in much the same way that elongation of O-fucose glycans on EGF repeats by Fringe modulates interactions between Notch and its ligands (12,32).
In addition to direct effects on protein-protein interactions, O-fucose modifications have recently been implicated in protein folding and quality control. Both Drosophila OFUT1 and mammalian POFUT1 are localized in the ER (24,33). The fact that POFUT1 has the ability to distinguish between properly folded and unfolded EGF repeats (11) suggests that it could play a role in quality control. In addition, a recent study reported that Drosophila OFUT1 has a chaperone-like activity promoting Notch receptor folding and trafficking from the ER to the cell surface (24). The fact that POFUT2 only modifies properly folded TSRs (42) and that Drosophila OFUT2-HA is mainly localized to the ER (Fig.  6) suggests that POFUT2 may also function in quality control.
Further support for ER localization of POFUT2 enzymes comes from the fact that O-fucosylation of TSRs is unaffected in fibroblasts from patients with leukocyte adhesion deficiency type II/congenital disorder of glycosylation type IIc (LADII/CDGIIc). LADII/CDGIIc is a rare autosomal recessive disease characterized by leukocyte adhesion deficiency as well as severe neurological and developmental abnormalities (34). It is caused by mutations in the Golgi GDP-fucose transporter, resulting in a reduction of fucosylated antigens on the cell surface (35)(36)(37)(38). Analysis of fibroblasts from LADII/CDGIIc patients revealed that terminal fucose modifications on N-glycans are significantly reduced. Surprisingly, O-fucosylation of EGF repeats and TSRs is not affected in cells from these patients (39,40). Specifically, the major form of O-fucose saccharide on TSRs, Glc-␤1,3-fucitol, is unaffected in fibroblasts from LADII/ CDGIIc patients. The localization of POFUT1 and -2 to the ER provides a potential explanation for these results. A separate GDP-fucose transport mechanism, unaffected by the mutations found in the LADII cells, may provide GDP-fucose to the ER, allowing O-fucosylation of EGF repeats and TSRs. Support for such a mechanism has been recently provided by a GDP-fucose transporter study in Drosophila (41).
Examination of the N-terminal sequences from human, mouse, Drosophila, and C. elegans O-fucosyltransferase 2 using online sequence analysis programs suggests that POFUT2 enzymes are soluble proteins with an N-terminal signal peptide (Fig. 1). This is consistent with our observation that the majority of TSR O-fucosyltransferase activity is present in high-speed supernatants of COS1 cells (42). We previously showed that the C-terminal KDEL-like motif RDEF of mammalian POFUT1 is responsible for retaining the protein in the ER (33), whereas the analogous HEEL motif was found to be involved in ER localization of the Drosophila counterpart protein (24). However, these motifs are absent from O-fucosyltransferase 2 protein sequences (Fig. 1), suggesting that an alternate ER retention signal or mechanism exists for O-fucosyltransferase 2. Interestingly, tagged forms of C21orf80 (one potential splice form of human POFUT2) expressed in COS-7 or U2OS cells were found in the Golgi compartment (29); however, functionality of the fusion constructs used in this report remains unknown. Studies to confirm the subcellular localization of POFUT2 in mammalian cells are in progress.