Activity, Splice Variants, Conserved Peptide Motifs, and Phylogeny of Two New α1,3-Fucosyltransferase Families (FUT10 and FUT11)*

We report the cloning of three splice variants of the FUT10 gene, encoding for active α-l-fucosyltransferase-isoforms of 391, 419, and 479 amino acids, and two splice variants of the FUT11 gene, encoding for two related α-l-fucosyltransferases of 476 and 492 amino acids. The FUT10 and FUT11 appeared 830 million years ago, whereas the other α1,3-fucosyltransferases emerged 450 million years ago. FUT10-391 and FUT10-419 were expressed in human embryos, whereas FUT10-479 was cloned from adult brain and was not found in embryos. Recombinant FUT10-419 and FUT10-479 have a type II trans-membrane topology and are retained in the endoplasmic reticulum (ER) by a membrane retention signal at their NH2 termini. The FUT10-479 has, in addition, a COOH-ER membrane retention signal. The FUT10-391 is a soluble protein without a trans-membrane domain or ER retention signal that transiently localizes to the Golgi and then is routed to the lysosome. After transfection in COS7 cells, the three FUT10s and at least one FUT11, link α-l-fucose onto conalbumin glycopeptides and biantennary N-glycan acceptors but not onto short lactosaminyl acceptor substrates as do classical monoexonic α1,3-fucosyltransferases. Modifications of the innermost core GlcNAc of the N-glycan, by substitution with ManNAc or with an opened GlcNAc ring or by the addition of an α1,6-fucose, suggest that the FUT10 transfer is performed on the innermost GlcNAc of the core chitobiose. We can exclude α1,3-fucosylation of the two peripheral GlcNAcs linked to the trimannosyl core of the acceptor, because the FUT10 fucosylated biantennary N-glycan product loses both terminal GlcNAc residues after digestion with human placenta α-N-acetylglucosaminidase.

Fucosyltransferases are globular type II trans-membrane Golgi-resident proteins that catalyze the transfer of ␣-L-fucose from GDP-Fuc onto N-and O-linked glycans, free oligosaccharides, or lipids (1) or directly onto proteins (2). These fucosylations are involved in a variety of biological processes, including selectin-mediated leukocyte-endothelial adhesion, lymphocyte homing, ABO blood group histocompatibility, notch receptor signaling (3), embryo-fetal development, and host-microbe interactions (4). Changes in the glycosylation pattern of proteins may interfere with cellular functions and may thus lead to health disorders, such as cancers or rare autosomal recessive diseases, such as congenital disorders of glycosylation, characterized by glycosylation deficiencies (5).
Human ␣1,3/4-fucosyltransferases and their genetic expression are developmentally regulated (17,18). We have previously shown that FUT4 and FUT9 genes are derepressed early in human embryogenesis, whereas FUT6 and FUT3 appear sequentially after the 8th week of development (19). This suggests that during development, the Le x or CD15 antigen (generated by FUT4 or FUT9) appears earlier than the sialyl-Le x (made by FUT5, FUT6, or FUT7) or the type 1 Lewis structures (Le a and Le b antigens, made by FUT3 or FUT5). Le x has been found mainly in undifferentiated rapidly dividing cells (20), whereas sialyl-Le x is more abundant in differentiated cells (17). This is particularly interesting because these glycotopes are implicated in embryo-fetal development, selectin-dependent leukocyte recruitment, and lymphocyte homing (21).
A few years ago, Renkonen and co-workers (22), using the Drosophila genome-wide bioinformatics approach to identify the proteome involved in ␣-L-fucosylated glycan metabolism, identified a Drosophila fucosyltransferase and two human orthologous genes, encoding for the fucosyltransferases FUT10 and FUT11. Due to the presence of the two main conserved motifs (23), they were assumed to be ␣1,3-fucosyltransferases, but their activity has not been experimentally demonstrated yet in any species, including humans (22), mice (24), 5 flies (26 -28), or honeybees (29). Four insect ␣1,3-fucosyltransferases (Fuc-TA, Fuc-TB, Fuc-TC, and Fuc-TD) were first identified in D. melanogaster (26 -28). The Fuc-TA is a core ␣1,3-fucosyltransferase (28), the Fuc-TB is orthologous to human FUT10 and FUT11, Fuc-TC is probably involved in the synthesis of Le x (29), and no activity has yet been found for the Fuc-TD.
In this work, we cloned three new active splice variants of the human FUT10 gene, two in the embryo and one in the adult. We investigated their subcellular distribution and their fucosyltransferase activity. In addition, we report an ␣1,3-fucosyltransferase activity for FUT11, with an acceptor pattern similar to FUT10.

EXPERIMENTAL PROCEDURES
Fucosyltransferase Assays-Transfected COS7 cells were homogenized on ice in 2% Triton X-100, and protein concentration was measured with the Bio-Rad Bradford protein microassay. Each fucosyltransferase assay was performed, unless otherwise stated, in a total volume of 65 l, containing 15 g of cell protein extract, 65 mM cacodylate buffer (pH 7.25), 10 mM L-fucose, 7 M GDP-[ 14 C]L-fucose (29 ϫ 10 4 dpm/test at 300 mCi/mmol; Amersham Biosciences), and 5 l/test of a 1 mg/ml solution of acceptor substrate (Table 1). For ␣1,3-fucosyltransferase assays, we used the conditions already described for the ␣1,6-fucosyltransferase FUT8 (30), and the activities were compared in the absence or presence of 20 mM MnCl 2 as cofactor. The S.D. values of the mean value activities correspond to 10 independent tests for each enzyme. The mean of three independent enzymatic experiments is shown when there is no S.D. For kinetic studies, we used 25 g of homogenate proteins, 3 M GDP-[ 14 C]L-fucose (125 ϫ 10 3 dpm/test at 300 mCi/mmol) plus 125 M cold GDP-Fuc (kindly provided by Claudine Augé, Paris-XI University of Sciences, Orsay, France).
The reactions were generally continued for 4 h at 37°C, stopped by the addition of 3 ml of cold water, and centrifuged, and the supernatant was applied to a conditioned Sep-Pak C18 reverse-phase chromatography cartridge (Waters, Milford, MA) attached to a 10-ml syringe (19,31). The retained hydrophobic acceptors were separated from the unreacted GDP-[ 14 C]fucose and its hydrolysis products by washing with 25 ml of H 2 O and then eluted with two 5-ml fractions of methanol and counted with 10 ml of Instagel-Plus (Hewlett-Packard, Evry, France) in a liquid scintillation ␤-counter (LS-6500; Beckman). The transfer of [ 14 C]fucose was expressed in pmol/h/mg of protein, unless otherwise stated.
Biantennary or other N-glycan-related acceptors, synthesized as biotinylated monomeric probes (0987-BM, 0988-BM, 0989-BM, 0990-BM, and chitobiose-BM), were bought from Lectinity Corp. (Moscow, Russia). They are synthesized with an elongated spacer linked to biotin (32), giving them the possibility of being adsorbed on Sep-Pak C18 cartridges as described for the above mentioned hydrophobic acceptors.
Preparation of the ␣1,6-Fucosylated 0989-BM Acceptor with the FUT8 Fucosyltransferase-The innermost GlcNAc residue of the 0989-BM acceptor (50 g) was fucosylated in the ␣1,6position with the recombinant FUT8 enzyme. The reaction conditions were as follows: 65 l containing 80 g of protein from the homogenate of COS7 cells transfected with FUT8 cDNA, 65 mM cacodylate buffer (pH 7.25), 10 mM L-fucose, 300 M cold GDP-L-fucose, and 5 g/test of the 0989-BM acceptor substrate (Table 1). After 4 h at 37°C, the reaction is more than 97% complete, and the retained Fuc␣1,6-0989-BM product was purified as described above. The transfer of the [ 14 C]fucose in position 3 of Fuc␣1,6-0989-BM by the different FUT10 isoforms was performed as described above with a 16-h incubation at 37°C. The activity was expressed as dpm/reaction.
Another fucosylation reaction was performed as described above, using 100 g of native chicken egg white conalbumin.
After terminating the fucosyltransferase incubation, Pronase was added, and the reactions were incubated overnight at 37°C. TLC-A fucosyltransferase assay using the homogenate derived from cells transfected with the FUT10-419 construct was performed for 6 or 16 h at 37°C with the BGA-biotin acceptor. The radioactive oligosaccharide product was purified on a Sep-Pak C18 cartridge and eluted with methanol. After lyophilization, two-thirds of the radioactive product was digested overnight at 37°C with 0.6 units of ␤-N-acetylglucosaminidase (NAG) from human placenta (Sigma) in 0.1 M citrate/phosphate buffer, pH 5.5. Subsequent to clean-up using Sep-Pak cartridges as described above, standards and NAG-treated 14 C-fucosylated acceptor molecules were resolved on silica-coated plastic TLC sheets (Merck) devel-  (34) and 14 C-labeled Gal 2 GlcNAc 2 Man 3 GlcNAc 2 , which was generated by incubating GlcNAc 2 Man 3 GlcNAc 2 (Dextra Laboratories) with UDP-[ 14 C]galactose and bovine milk galactosyltransferase (Sigma).
RNA Isolation and Northern Blot Analysis-Embryos aged from 50 to 70 days were obtained from legal abortions and stored at Ϫ80°C as already described (19). Total RNA was extracted with guanidine isothiocyanate and purified by cesium chloride gradient centrifugation. Contaminating DNA was removed by digestion with RNase-free DNase I (10 units/g of RNA from Roche Applied Science) for 15 min at room temperature, followed by 15 min of inactivation at 70°C and purification of the RNA by phenol/chloroform extraction. Embryonic poly(A) ϩ mRNAs were double purified using oligo(dT)-cellulose (type 3; Sigma) chromatography. Poly(A) ϩ RNAs (4 g/lane) were denatured and fractionated with 1.2% phosphate-agarose gel electrophoresis, transferred to Hybond-N membranes (Amersham Biosciences), and immobilized by baking at 80°C for 2 h. Prehybridization and hybridization were performed for 16 h at 42°C in a buffer containing 50% formamide, 5ϫ SSC, 1ϫ PE, 250 g/ml denatured salmon sperm DNA, and 10% dextran sulfate with the cds-FUT10 probe of 350 bp, obtained by PCR using primers sense F10-8s and antisense F10-4as ( Table 2). The blots were first washed at low stringency: 2 ϫ 15 min (in 2ϫ SSC, 0.1% SDS) at 42°C, followed by a single wash of 15 min (2ϫ SSC, 0.1% SDS) at 50°C and autoradiographed. A last 15 min wash in (0.5ϫ SSC, 0.1% SDS) at 60°C was performed, and another autoradiography was made. The films were developed after 3 days at Ϫ80°C.
Construction of cDNA Libraries-Poly(A) ϩ mRNAs (1 g) from a single 50-day embryo and from an adult brain were used to initiate the first strand cDNA synthesis. They were reverse transcribed at 42°C for 90 min, using the oligo(dT)-cDNA synthesis primer (52-mer) and 200 units of the Superscript-II RNase H-reverse transcriptase from the Superscript first strand cDNA synthesis system kit (Invitrogen). The embryonic and adult cDNA libraries were stored at Ϫ20°C until used. The PCRs were carried out with primers specific for FUT10 (Table   2), the Klentaq mixture (Clontech), and 1 l of cDNA templates from the 50-day embryo or from adult brain cDNA libraries for the first PCR and 1 l of the first PCR product diluted 1:10 for the second PCR. The same amplification program with the Advantage cDNA amplification kit mix (BD Clontech, Palo Alto, CA) was used. All PCRs were performed in 50 l, with 1ϫ Klentaq buffer, a 0.2 M concentration of each primer, 1 unit of Klentaq DNA polymerase, and 0.2 mM dNTP with the touchdown-RACE program: initial denaturation 94°C for 90 s, followed by five cycles of 94°C for 30 s and 72°C for 4 min, 5 cycles of 94°C for 30 s and 70°C for 4 min, and 25 cycles of 94°C for 30 s and 68°C for 4 min.
Cloning of the FUT10 Transcripts-Four FUT10 cDNA isoforms (FUT10-357, FUT10-391, FUT10-419, and FUT10-479) were amplified by a double PCR using the embryo and the adult cDNA libraries as templates. The FUT10-357 was a truncated variant of FUT10-391, lacking the conserved motifs I and II. It was devoid of enzyme activity and was not further analyzed. The primer associations F10-1s and F10-13as or F10-1s and F10-as2 were used for the first PCR, and distinct combinations of nested primers as F10-8s and F10-12as or F10-8s and F10-as2 were used for the second PCR (Table 2). With these two nested primer combinations, we obtained a broad PCR product of 1400 bp in the embryo cDNA with F10-8s and F10-12as and a product of 1800 bp from the adult cDNA library, with F10-8s and F10-as2 primers. These final PCR products were gel-purified and cloned into the TA cloning vector PCR3.1 (Eukaryotic TA cloning kit; Invitrogen), and 20 plasmid clones were PCRselected from each positive ligation and sequenced. To generate the FUT10-GFP-tagged transcripts, we amplified the three selected FUT10 constructs: FUT10-391 with primers F10-391s and F10-GFPas, FUT10-419 with primers F10-8s and F10-GFPas, and FUT10-479 with primers F10-479s and F10-K1GFPas, in the presence of the high fidelity Hotstart DNA polymerase (Accuprime-pFx DNA polymerase; Invitrogen). The PCR products were gel-purified and inserted into the mammalian pcDNA3.1-GFP vector (TOPO-CT-GFP-cloning kit; Invitrogen). The resulting GFP-COOH-tagged plasmids were selected by PCR, sequenced, and called FUT10-391-GFP, FUT10-419-GFP, and FUT10-479-GFP.
Isolation of FUT11 cDNA Clones-The IMAGE clone 40005868 (BC100994) in a pCR-Blunt-TOPO vector encoding for a 476-amino acid protein and the IMAGE clone 5271548 Immunofluorescent Localization of the GFP-tagged FUT10 Fusion Proteins-2 ϫ 10 5 cells were seeded on glass coverslips in 35-mm cell culture Petri dishes 24 h before transfection. Five g of the GFP-tagged cDNA constructs were transfected into COS7 cells. After 12, 18, 24, or 48 h of growth, they were washed with PBS and then fixed for 15 min with 2% paraformaldehyde in PBS. To quench residual paraformaldehyde, cells were incubated for 20 min in 50 mM NH 4 Cl in PBS. Thereafter, they were permeabilized with 0.075% saponin, 0.1% bovine serum albumin (BSA) in PBS for 15 min, followed by 1 h of incubation with primary antibodies at room temperature. The GFP-tagged FUT10 recombinant proteins were visualized with a Leica DMR epifluorescence microscope, with a PLAN-APO ϫ63/ 1.32-0.6 oil objective lens and an HC-PLAN ϫ10/25 ocular lens (Leica Microsystems, Wetzlar, Germany). Images were captured with a LEI-750 CE digital camera and LIDA volume 54 Leica software and further processed with Adobe Photoshop 5.0 (Adobe, San Jose, CA).
Double immunofluorescence experiments were conducted using a Golgi-specific mouse monoclonal anti-giantin antibody kindly given by Hans-Peter Hauri (35) (1:1000 in PBS plus 1% BSA) or an endoplasmic reticulum (ER)-specific rabbit polyclonal anti-calnexin antibody (1:200 in PBS plus 1% BSA; StressGen, Assay Designs, Inc., Ann Arbor, MI) or the lysosomal specific mouse monoclonal anti-Lamp-1 (BB6) antibody, kindly given by Sven Carlsson (1:1000 in PBS plus 1% BSA), and the GFP-tagged-FUT10 recombinant protein. After washing the cells three times in PBS plus 0.1% BSA, the anti-giantin and the anti-Lamp-1 antibodies were revealed with conjugated anti-mouse Ig-Cya3 red fluorochrome, diluted 1:200 (Jackson Laboratories, L'Arbresle, France), and the anti-calnexin antibody with anti-rabbit Ig-Cya3, diluted 1:200 (Jackson Laboratories). The secondary labeled antibodies were incubated for 1 h at room temperature, and the reaction was stopped by washing three times in PBS. The coverslips with the labeled cells were mounted on slides with Mowiol and observed using a Leica DMR epifluorescence microscope.
Bioinformatics-Ten ␣1,3-fucosyltransferase-like sequences from different species were retrieved from data banks by Psi-Blast (36) with the FUT10 and FUT11 human sequences. All of these sequences belong to the CAZY glycosyltransferase family 10. The systematic search for new FUT10 and FUT11 sequences was performed by TblastN using the expressed sequence tag and whole genome shotgun data banks, as previously described for sialyltransferases (37). Within each species, the DNA contigs for each new ␣1,3-fucosyltransferase-like gene were made up with Cap3 (38). New complete protein sequences were analyzed for the presence of the NH 2 -terminal transmembrane domain (TMD) by PHD-htm (39) and for the presence of the ␣1,3-fucosyltransferase conserved peptide motifs. The accession numbers of the animal sequences reconstructed in silico by these approaches and the human sequences used in the present paper are as follows: Related amino acids were grouped according to their properties based on a chemical alphabet comprising five groups: acidic or amide (Glu, Asp, Gln, and Asn); hydrophobic (Ile, Leu, Val, and Met); aromatic (Phe, Tyr, and Trp); basic (Arg, His, and Lys); and hydroxyl (Ser and Thr). The remaining four amino acids (Ala, Gly, Pro, and Cys) were analyzed separately. Amino acids of the same group were considered equivalent for the definition of conserved positions, and amino acids conserved at more than 50% in the different ␣1,3-fucosyltransferases families were colored (Fig. 1).
Phylogeny-Protein and DNA alignments were performed by ClustalW (40) and saved in Pir format. The Pir alignment was used to select G-block-informative positions (41). By this computerized method, 206 amino acid positions in 14 G-blocks were selected for the phylogeny analysis. The ClustalW alignment of the G-block-selected positions was also used to count the percentage of FUT10-and FUT11-specific positions. The sequence lines of the ClustalW were then ordered by decreasing percentage (from 100 to 0%) of FUT10-specific positions and increasing percentage (from 0 to 100%) of FUT11-specific positions (Fig. 1). The sequences with more than 80% specific positions for either FUT10 or FUT11 can be ascribed to the corresponding family, whereas the sequences with equivalent proportions of specific positions (50 Ϯ 10%) cannot be ascribed to either of FUT10 or FUT11. This simple method, based on the similarities between sequences is complementary to the more sophisticated phylogeny calculations that are mainly based on the differences between sequences. Phylogeny was carried out with Phylowin (42) (available on the World Wide Web) using BIONJ, Poisson correction, and 500 bootstrap replicates (43).

Conserved Peptide Motifs of FUT10 and FUT11-
The hydrophobic cluster analysis (44,45) of the ␣1,3-fucosyltransferase family (FUT3-FUT7 and FUT9) first revealed two conserved peptide motifs originally called I and II. Then a third conserved motif of six amino acids was called the acceptor motif ((I/V/ F)HH(R/W)(D/E)(I/V/L)), because a single amino acid change (Trp 111 3 Arg) was able to transform the type 1 ␣1,4-fucosylation of FUT3 (leading to the Le a epitope) into the type 2 ␣1,3fucosylation (leading to the Le x epitope) (46,47). A fourth conserved motif, located on the NH 2 -terminal side of the sequence, before the acceptor motif, was later called motif III (12,47), giving the following order for these four conserved motifs: III, acceptor motif, I and II. Finally, while writing the present paper, we found a fifth conserved motif located between the acceptor motif and motif I. In order to avoid confusion, we have here renamed these five peptide motifs from I to V, according to their order in the protein sequence ( Fig. 1).
As expected, the two main conserved motifs (IV and V in Fig.  1) are present in all of the ␣1,3-fucosyltransferases, including the FUT10 and FUT11 isoforms. These two motifs contribute to recognition and binding of the donor substrate GDP-Fuc (12,46), whereas the first three motifs occur toward the NH 2 end of the sequence, in a region that has been shown (by domain swapping experiments) to be involved in the recognition of the acceptor substrate (48). In the middle of this region, the acceptor motif II is present in FUT3-FUT7 and FUT9, but we did not find it in either the FUT10 or FUT11 sequences. However, a different conserved peptide motif (FYGTDF) was present in the equivalent positions of FUT10 and FUT11 (Fig. 1). The classical acceptor motif II shares only one negatively charged amino acid at position 5 (Asp or Glu) with this new acceptor motif II of FUT10 and FUT11 (Fig. 1). The motifs I and III, flanking the acceptor motif II that is involved in acceptor recognition contain six conserved amino acids specific of FUT10 and FUT11.
The intermotif distances between the motifs I, II, III, IV, and V are relatively well preserved in all of the ␣1,3-fucosyltransferase families, including FUT10 and FUT11 (Fig. 1). This plus the very good conservation of motifs IV and V and the less good but significant conservation of motifs I and III suggests that all of the ␣1,3fucosyltransferases derive from a single ancestral gene.
Overall, 24 amino acid positions within the five conserved peptide motifs were specific to both FUT10 and FUT11 and Motifs --- M.musculus absent from the classical monoexonic ␣1,3-fucosyltransferases (Fig. 1). In addition, the existence of other FUT10-and FUT11specific amino acids in the remaining positions of the 14 G-blocks (41), comprising the 206 informative amino acids selected for phylogeny (not shown), allowed us to clearly ascribe all of the vertebrate FUT10 and FUT11 enzymes to one or the other of these two families, because they contain more than 80% of conserved positions specific for either one or the other family (Fig. 1, last two columns). Alternatively, the presence of equivalent proportions of specific positions of FUT10 and FUT11 (50 Ϯ 9%) in the insect FUT10/11 enzymes suggests that they are at the same genetic distance from either FUT10 or FUT11 and therefore can be orthologous to the common ancestor that preceded the duplication event at the origin of the FUT10 and FUT11 families. The same concept applies to the classical monoexonic ␣1,3-fucosyltransferases (FUT3-FUT7 and FUT9) (11) that are also at the same genetic distance, as evaluated by the percentage of specific positions (50 Ϯ 14%) from either the FUT10 or FUT11 family. They have almost no FUT10-or FUT11-specific conserved positions within the five peptide motifs (Fig. 1) and only a small number among the 206 positions selected for phylogeny.

FUT3-7 and FUT9
In addition to the above mentioned FUT3-FUT11 sequences, other ␣1,3/4-fucosyltransferases have been characterized in invertebrates (core insect) and in plants (core plant and Le a ) (not shown). They all had the large donor-related conserved motifs IV and V, but none had significant numbers of FUT10-or FUT11-specific positions. Furthermore, the sequences of motifs I, II, and III of the core invertebrate (28, 49 -51), the core plant (52,53), and the Le a plant enzymes (51,54) are more closely related to the FUT3-FUT7 and FUT9 than to the FUT10 or FUT11 families. Nevertheless, they could not be ascribed to any of the FUT3-FUT7 and FUT9 families of ␣1,3/4fucosyltransferases, suggesting that they are at similar genetic distances from each of the FUT3, FUT4, FUT5, FUT6, FUT7, and FUT9 enzymes. Therefore, they might be orthologous to a common ancestor present before the duplication events at the origin of the paralogous FUT3-FUT7 and FUT9 genes.
Phylogeny of FUT10 and FUT11-The phylogenetic tree of Fig. 2 contains the FUT10 and FUT11 sequences on the first branch and the human FUT3-FUT7 and FUT9 (11) on the second branch. The root is probably between these two branches, since the classical monoexonic ␣1,3-fucosyltransferases appear as an out-group clearly distinct from FUT10 and FUT11. The enzymes of the FUT3-FUT7 and FUT9 branch started to appear early in the vertebrate lineage (about 450 MYA), since they were not found among invertebrates, but they are present in fishes, amphibians, birds, and mammals (1,19,55). They are all monoexonic with the exception of FUT7, whose cds is assembled from two exons (56), whereas all of the genes coding for the enzymes of the FUT10 -FUT11 branch (Fig. 2) plus all of the other invertebrate and plant ␣1,3-fucosyltransferases are polyexonic (1). This suggests that a retrotransposition event of an ancestral rearranged gene occurred in the FUT3-FUT7 and FUT9 branch, before the duplications at the origin of these six paralogous genes (represented by a solid circle in Fig. 2), which is at the origin of the intron loss of the sequences of this branch (57).
Another reference point can be dated at about 35 MYA (47) in this same branch (solid square), since only Old World anthropoid apes and humans have the three Lewis genes (FUT3, FUT5, and FUT6) (58), whereas all of the New World monkeys and lower mammals have a single gene (FUT3/5/6), orthologous to the ancestor present before the duplications at the origin of FUT3, FUT5, and FUT6 genes.  FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7

Expression of Two New FUT Families, FUT10 and FUT11
In the FUT10 -FUT11 branch, a first duplication (solid triangle in Fig. 2) was at the origin of the separation of the insect FUT10/11 group from the ancestor of the two vertebrate FUT10 and FUT11 families, and a second duplication (solid diamond) generated the FUT10 and FUT11 families, which are found in vertebrates and urochordates, since two ciona sequences (C. intestinalis and C. savignyi) (Fig. 2) have 81 and 85% of the FUT11-specific positions (Fig. 1).
The duplication node at the origin of the FUT10 and FUT11 families is flanked by the separation of arthropods 980 MYA (59) and the separation of urochordates 790 MYA (60), suggesting that it occurred about 830 MYA (Fig. 2). It is interesting to note that the insect FUT10/11 are the only sequences of the FUT10 and FUT11 families that lack the last two amino acids (DF) of the acceptor substrate motif II (Fig. 1). Therefore, these insect FUT10/11 sequences, which are the most ancient pro-teins of the FUT10 -FUT11 branch (Fig. 2), might have different acceptor specificity.
Identification of the Embryo and Adult Transcript Splice Variants of FUT10-The presence of the FUT10 splice variants was first verified in our human 50-day embryo cDNA library using the double PCR method reported under "Experimental Procedures." A broad PCR product of ϳ1400 bp was amplified with the primers F10-8s and F10-12as ( Table 2). After cloning in the PCR3.1 mammalian vector, we selected 20 clones for sequencing and found two new  (Fig. 3a). After translation, a soluble protein of 391 amino acids without TMD is expected. The insertion of exon 3 of 60 bp at position 177 of the cDNA induces the use of the ATG2 start codon (position 239 of the cDNA) and the stop codon TGA1. The predicted protein has three putative N-glycosylation sites, located at positions 82, 140, and 290 (Fig. 3a).
The second splice variant of 1373 bp (FUT10-419) has an ORF of 1354 bp, including exons 2, 5, and 6. It skips exons 3, 4, and 7 and uses the ATG1 (position 95 of the cDNA) and the same TGA1 stop codon as the FUT10-391. Post-translational topology suggests a type II trans-membrane protein of 419 amino acids with a short TMD of 16 amino acids and an NH 2terminal ER membrane retention signal (VRIQ), similar to that of the yeast ALG7 enzyme, also localized in the ER. The FUT10-419 has the same three N-glycosylation sites of the soluble isoform FUT10-391 (Fig. 3a). Repetitive PCRs in the different cDNA pools could only amplify these two variants in the embryo library and not in the adult cDNA brain library.
Using the conditions expected to amplify the in silico defined human sequence FUT10-428 (AJ431184 (22)), we were unable to amplify the corresponding transcript from either embryo or adult cDNA libraries, but we did amplify a new splice variant reported as FUT10-479 from the adult library. The complete cDNA of 2312 bp was composed of exons 2, 5, 6a, and 7. We made a double PCR using first primers F10-1s and F10-as2 and then primers F10-8s and F10-as2, generating a fragment of 1866 bp, from the adult brain cDNA library. After cloning this fragment in the mammalian PCR3.1 vector, it has an ORF of 1440 bp, using the ATG1 and a new stop codon TGA2 (Fig. 3a). The FUT10-479 has the same TMD and NH 2 -terminal ER membrane retention signal found in the FUT10-419 splice variant plus a new ER membrane retention signal (LVFK) in the COOH terminus. It has three putative N-glycosylation sites identical to those of the other two FUT10 variants plus a fourth one located at position 466 (Fig. 3a). The corresponding transcript was never amplified from our embryo cDNA library even after repetitive PCR. Compared with the published FUT10-428 sequence, our three FUT10 variants have two amino acid substitutions, Leu to Phe at position 31 and Leu to Asp at position 340 of the FUT10-391, and use different start codons (Fig. 3a).
Tissue Distribution of the Human FUT10 mRNA Transcripts-The size and tissue distribution of the human FUT10 variants were determined by Northern blot analysis with the specific FUT10 cds probe. The poly(A) ϩ RNA from entire 50 -70-dayold embryos show the same transcript profile, with four FUT10 mRNA bands, ranging from 3 to 12 kb (Fig. 4). The 3-kb transcript is a broad band always more abundant than the 5, 8, or 12 kb bands. They are expressed in all of the embryos with similar intensity, but we have no information about their tissue distribution, because they correspond to mRNA from entire embryos.
The fetal and adult tissue distribution profiles were studied with commercial Northern blots. Table 3 summarizes the relative intensity evaluation of the bands of the embryo, fetus, and adult tissues. In all fetal tissues, the FUT10 profile with four bands is found in skin, small intestine, and liver with an intensity similar to that seen in the embryos. In kidney, we found a profile with three weak bands of 3, 5, and 8 kb. Lung and muscle expressed weakly the 3, 5, and 12 kb bands. Heart and brain show only the 3 kb band. The adult tissues also revealed these transcripts but with different patterns of expression. In the lung, the four transcripts were similar to the embryo. Tissues of the digestive tract (jejunum, colon, rectum, and stomach) expressed the 3-and 5-kb transcripts strongly. A similar but weaker pattern was observed with ileum and placenta. The gall bladder reveals strongly expression of the 3 kb band only. This same transcript is also expressed with moderate intensity in kidney and uterus and weakly in brain. Adult spleen, heart, muscle, liver, and pancreas did not express any of these FUT10 transcripts ( Table 3).
Identification of the Transcript Splice Variants of FUT11-The two cDNA splice variants of FUT11 were integrated in the PCR3.1 vector, sequenced, and named FUT11-476 and FUT11-492 (Fig. 3b). The ORF sizes of the selected clones are 1431 and 1479 bp, respectively. A differential splicing event occurred in exon 3, giving the FUT11-476 with the entire exon 3 and the FUT11-492 when it utilizes the exon 3a, starting at position 123 and skipping 15% of this exon. This induces a change of ORF and generates a new stop codon (TAA2). The use of two distinct stop codons predict two proteins with a short TMD of 16 amino acids. The FUT11-476 has no ER membrane retention signal, but the FUT11-492 variant has a KKXX-like motif (KRQH) in its COOH terminus that could retain it in the ER. These pro-

Comparisons of embryo, fetal, and adult expression of FUT10 transcripts after Northern blot analysis
The presence of the transcript bands is shown with the symbols ϩϩϩ, ϩϩ, ϩ, and ϩ/Ϫ, corresponding to the relative intensities of the bands. Absence of transcripts is noted with the negative symbol (Ϫ).   FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 4731
teins have only two of the four potential N-glycosylation sites found in FUT10 (Fig. 3b). Genomic Organization of the FUT10 and FUT11 Genes-The Blast search of the EMBL data bank retrieved AC091144, located on the chromosome band 8p11.23 (61) that contains the complete FUT10 genomic organization spanning 102 kb. The splice intron/exon boundaries followed the AG/GT rule (62) ( Table 4). We also found a microsatellite sequence with a 17-GT repeat localized on the 3Ј-side of exon 4 (about 500 bp downstream; not shown). The gene organization of FUT11 (Fig.  3b and Table 4) is included in the EMBL genomic sequence AC022400 located on the chromosome band 10q22.2 (61), and it spans ϳ4 kb.
The natural biantennary oligosaccharide substrate, BGA-biotin, is a well known and very good acceptor for the FUT8 ␣1,6-fucosyltransferase (1550 pmol/h/mg), because it is the specific substrate of FUT8, and the assay conditions used were optimal for this enzyme (30). These conditions without MnCl 2 are also appropriate for the three FUT10 isoforms (FUT10-391,  FUT10-419, and FUT10-479), giving respective activities of 70, 120, and 105 pmol/h/mg. The FUT11-476 displayed an activity of 93 pmol/h/mg, and a very weak activity of 17 pmol/h/mg for the FUT11-492 isoform was noted. The FUT10 and FUT11 activities are at least 10 times lower than that noted for the FUT8 enzyme but are higher than the activities of the two ␣1,3fucosyltransferases, FUT3 and FUT4 (30 and 55 pmol/h/mg, respectively).
The FUT8, FUT10, and FUT11 activities toward the acceptor 0989-BM were increased by about 50% compared with those observed using BGA-biotin. The 0989-BM acceptor possesses the same N-glycan structure present in the BGA-biotin, but the sugar structure is connected to the biotin via a linking arm NHCOCH 2 NHCO(CH 2 ) 5 instead of the natural asparagine (Table 1). We found activities of 2480 pmol/h/mg for FUT8; 105, 175, and 160 pmol/h/mg for FUT10-391, FUT10-419, and FUT10-479, respectively; and finally an activity of 139 pmol/ h/mg for FUT11-476, and again a weak activity of 25 pmol/ h/mg for the FUT11-492 isoform was observed. By contrast, the FUT3 and FUT4 enzymes revealed similar activities toward both acceptors.
The transfer activity profiles of the FUT10 and FUT11 isoforms are clearly distinguishable from those of the classical monoexonic ␣1,3-fucosyltransferases, and this difference increases in the presence of 20 mM MnCl 2 , which is a cofactor for FUT3 and FUT4, but not for FUT10, FUT11, or FUT8 (see Footnotes b-d, Table 5). In the presence of MnCl 2 , the activity obtained with short lactosaminyl acceptors is doubled for FUT3 and FUT4, reaching 1115 pmol/h/mg for FUT4 with H-type-2-Ogr and 230 pmol/h/mg with LacNAc-type-2-Ogr. For FUT3, the transfer onto H-type-1-Ogr increased to 1280 pmol/h/mg. By contrast, with FUT10 and FUT11, the presence of MnCl 2 reduces by 80% the transfer onto N-glycan acceptors (0989-BM or BGA-biotin) and by 50% the activity of FUT8. From these results, we conclude that FUT10 and FUT11 function well without MnCl 2 , use biantennary N-glycan acceptors, and cannot use short linear acceptors.
As for the natural BGA-biotin acceptors, two types of fucosylation site are possible for these enzymes on the 0989-BM acceptor: either or both of the GlcNAc residues linked ␤1,2 onto the trimannosyl core, and the innermost GlcNAc of the chitobiose unit. For this reason, we prepared new acceptors modified at the innermost GlcNAc of the chitobiose (7OS-ManNAc-BM and 7OS-amino-alditol-GlcNAc-BM) ( Table 1). Data presented in Table 6 indicate that these modified acceptors gave very low rates of transfer with the FUT10 and FUT8 enzymes when compared with the native 0989-BM acceptor. The 2-NAc epimerization of the innermost GlcNAc (7OS-ManNAc-BM) inhibited acceptor potential drastically, show-ing only 8 -10% of that observed for the parent structure. The acceptor potential of 7OS-amino-alditol-GlcNAc-BM was even lower, reaching only 0.1-0.9% for the three FUT10 variants and around 3% for FUT8, compared with that of the nonmodified acceptor 0989-BM (Table 6).
A comparison of the acceptor potential of 0989-BM with its fucosylated counterpart (Fuc␣1,6-0989-BM) revealed an approximately 50% reduction for the FUT10 variant when the acceptor was fucosylated. The FUT8 ␣1,6-fucosyltransferase cannot use this fucosylated acceptor, and the apparent transfer rate was reduced to 2%. Therefore, this acceptor permits discrimination between FUT10-and FUT8-mediated fucosylations (Table 6).
Under the same assay conditions and after 18 or 48 h of transfection, we tested the fucosyltransferase activities expressed in the FUT10-transfected cell supernatants, and no detectable activity was found using the best acceptor for these enzymes (0989-BM) (not shown).
In order to locate more precisely the fucosylation site used by FUT10-419, the 14 C-fucosylated BGA-biotin reaction products generated by this enzyme were analyzed by TLC before and after NAG digestion. The undigested material yielded a predominant species designated Fuc-BGA-biotin (Fig. 5) as well as three minor slower migrating components. After NAG digestion (16 h), Fuc-BGA-biotin is converted into a component whose migration position is compatible with Fuc-BGA-biotin having lost its two terminal GlcNAc residues (see Fig. 5, Ϫ2 GlcNAc). In order to confirm that the product obtained lost both terminal GlcNAc residues, we performed shorter (6-h) NAG incubations. Under these conditions, in addition to the above mentioned digestion product (Fig. 5, Ϫ2 GlcNAc), a second component, whose migration position is compatible with a product having lost only one GlcNAc, was observed (not shown). The ensemble of these results indicates that the FUT10 enzyme predominantly fucosylates the innermost chitobiose core GlcNAc.
Kinetic Studies of FUT10-Apparent K m values were calculated for GDP-Fuc (range of concentration 3-420 M) in the presence of 60 M BGA-biotin and with optimal conditions for the FUT10 enzyme ( Table 8). The FUT10-391 and the FUT10-479 variants had similar apparent affinities (K m ) for GDP-Fuc (between 13 and 15 M). Under these conditions, the maximum velocity obtained for the soluble protein FUT10-391 (V max ϭ 285) is twice that of the ER-resident enzyme FU10-479 (V max ϭ 146) ( Table 8). Apparent K m values were also calculated for the natural BGA-biotin acceptor (3-310 M) at a saturating concentration of GDP-Fuc (125 M). The soluble (FUT10-391) and the ER-resident (FUT10-479) enzymes have higher apparent K m values for this acceptor than those observed for the GDP-Fuc donor, but the K m values for the acceptor are similar for the two variants (154 and 150 M). The V max values for the acceptor were lower than those observed for GDP-Fuc but with the same difference between the soluble FUT10-391 (V max ϭ 119) and the ER-resident FUT10-479 enzyme (V max ϭ 59) ( Table 8).
Subcellular Localization of the GFP-tagged Fusion Proteins-The three-tagged FUT10 isoforms were expressed into COS7 cells and visualized by immunofluorescence after 12, 18, 24, and 48 h of transfection. At 12 h, the soluble FUT10-391-GFP enzyme (green) did not co-localize with the ER marker calnexin (red) (Fig. 6, a-c), nor with the Lamp-1 lysosomal protein marker (not shown). By contrast, FUT10-391 did fully co-localize with giantin (red) in the Golgi apparatus (Fig. 6, d-f), as do the classical monoexonic ␣1,3-fucosyltransferases. Eighteen hours after transfection, the majority of this soluble protein still co-localized with giantin ( Fig. 6, g-i), but some had migrated outside the Golgi apparatus and appeared to be dissociated from giantin. Twenty-four hours after transfection, it increased its migration outside the Golgi apparatus (Fig. 6, j-l), and more than 20% of the FUT10-391-GFP-transfected cells revealed a partition of the GFP fusion protein between the Golgi apparatus and cytoplasmic vacuolated vesicles. At 48 h, all of the FUT10-391-GFP labeling is completely dissociated from giantin (Fig. 6, m-o) and calnexin (Fig. 6, p-r). Strikingly, at 48 h, the staining with Lamp-1 antibody (Fig. 6, s-u) shows full colocalization of FUT10-391-GFP with this lysosome marker. These results demonstrate that the subcellular localization of the soluble FUT10-391-GFP changes with time. It is only transiently present in the Golgi before appearing in the lysosomal compartment associated with Lamp-1.
The two fusion proteins FUT10-419-GFP and FUT10-479-GFP appear to co-localize with the ER marker calnexin (Fig. 7, a-c) at all times of transfection, from 12 to 48 h. Lamp-1 labeling was dissociated from both FUT10-419-GFP (not shown) and FUT10-479 (Fig. 7, j-l), and giantin was also dissociated from both FUT10-419-GFP (Fig. 7, d-f) and FUT10-479-GFP (Fig. 7, g-i). By contrast, transfection with the FUT8-GFP enzyme illustrates co-localization of FUT8 with the Golgi apparatus marker (Fig. 7, m-o), as reported for all of the classical ␣1,3-fucosyltransferases. . The nondigested Fuc-BGA-biotin product (Ϫ) has a main strong spot and three very faint spots. After NAG digestion (ϩ) the main spot migrates faster, at the position expected for the product that has lost the two terminal nonreducing GlcNAc residues linked ␤1,2 to the trimanosyl core of the glycosaparagine N-glycan. The expected migration positions of Fuc-BGA-biotin products having lost one (Ϫ1 GlcNAc) or two GlcNAc (Ϫ2 GlcNAc) and the marker Man 5 GlcNAc 2 (M 5 Gn 2 ) are also indicated. Procedures." Specific fucosyl-glycopeptides activities are expressed as the dpm recovered from the pooled fractions 9 -15 (glycopeptides) of the Biogel P2 columns. b The fucose transfer was performed on native conalbumin glycoprotein, and then the glycoprotein was digested with pronase to obtain radioactive fucosylated glycopeptides. After Pronase digestion, the radioactive products were quantitated after Biogel P2 chromatography as described above. fact, this is a good control for the ␣1,3-fucosylation activity of FUT10. In addition, the FUT10 fucosyltransferase is able to transfer fucose onto native conalbumin glycoprotein or onto its derivative glycopeptides. This FUT10 glycoprotein fucosyltransferase activity profile is similar to the one obtained for the FUT8 enzyme but distinct from the classical ␣3-fucosyltransferase profile. In our conditions, the linear chitobiose-BM disaccharide acceptor is not an acceptor for either FUT10 or FUT11 nor for the classical monoexonic ␣1,3-fucosyltransferases. Although some fucose transfer onto linear acceptors, such as chitobiose, chitotriose, or chitotetraose, has been described for classical ␣1,3-fucosyltransferases, long incubations of 4 days are required. The structures of the final products were well charac-terized by NMR and mass spectrometry and revealed a mixture of chitobiose products fucosylated in ␣1,3and ␣1,6-orientations and chitotriose and chitotetraose products with only weak ␣1,3-fucosylation (63). In good agreement with these results, we also detect a very weak ␣-L-fucosyltransferase activity onto the innermost GlcNAc of BGA-biotin acceptor and onto the chicken egg white conalbumin glycoprotein and their derived glycopeptides with FUT3 and FUT4.
Considering the enzyme activity only, classical monoexonic ␣1,3/4fucosyltransferases prefer short lactosaminyl acceptors, whereas the polyexonic FUT10 enzymes prefer biantennary N-glycans linked to glycopeptides or to biotin aglycone, illustrating that linear short acceptors cannot be used to follow the FUT10 and FUT11 transfer of ␣-L-fucose. This may explain some of the previously reported negative results obtained with these enzymes (24) 5 and the fact that all of the FUT10 fucosyltransferase sequences already reported have differences in motif II implicated in the acceptor substrate recognition.
Regarding the subcellular localization of the FUT10 isoforms, the FUT10-391 variant is a soluble protein showing no TMD and no ER membrane retention signal. Immunofluorescent staining confirms that this protein is transiently co-localized with the Golgi marker giantin during the first 24 h and then migrates to the lysosome, to be probably later degraded in this compartment. In our hands, this protein never co-localized with the ER marker calnexin. Alternatively, the ER-resident FUT10-419 and FUT10-479 variants have a type II transmembrane topology and an ER membrane retention signal at their NH 2 terminus. FUT10-479 has, in addition, a second ER membrane retention signal at its COOH terminus. Immunofluorescent staining confirmed that the recombinant FUT10-419 and FUT10-479 proteins stably reside in the ER, since they are always co-localized with calnexin. This is the first time that active ␣1,3-fucosyltransferases have been found in the ER. Only the soluble protein-O-fucosyltransferase 1 enzyme has previously been shown to be retained, in the ER lumen via a COOH-terminal KDEL-like motif (64). It is not clear why FUT10 and protein-Ofucosyltransferase proteins are located in the ER, because no GDP-Fuc transporter has been characterized in this compartment. However, the addition of O-fucose to epidermal growth factor-like repeats has been demonstrated in the ER, suggesting the existence of a novel GDP-Fuc transporter in this compartment (64) or a retrograde flux of GDP-Fuc from the Golgi apparatus. Protein-O-fucosyltransferase 1, FUT10-419, and FUT10-479 are active enzymes in vitro and perhaps can have other additional functions in vivo, such as the reported chaperone role of protein-O-fucosyltransferase 1, which promotes the correct folding of the epidermal growth factor repeat of the Notch receptor (3).
We hypothesize that the ␣1,3-fucosylation onto the chitobiose unit of biantennary N-glycans can occur in the ER of embryo cells (FUT10-419) or in the ER of adult cells (FUT10-479) and transiently in the Golgi apparatus of embryo cells for the soluble variant FUT10-391. However, in the adult, this core ␣1,3-fucosylation has to be a transient intracellular signal, because secreted outside or expressed at the surface of the cell, the core ␣1,3-fucose on glycoproteins becomes a strong immunogen. It has been shown to be a major cause of allergic reactions, induced by insect and plant allergens containing this ␣1,3-core fucose epitope (25,65,66). Because this type of core ␣1,3-fucosylation is not a regular signal found in mammalian glycoproteins, expressed only in certain cell compartments, it could help to recognize and select the incorrectly glycosylated proteins that escape from the regular quality control mechanism, reorienting them to lysosomes or other compartments able to degrade aberrant glycoproteins.
Finally, enzyme activity was also detected with FUT11, in particular with FUT11-476. The results suggest a substrate acceptor pattern similar to FUT10 variants, but this enzyme activity needs further characterization.