A Single Amino Acid in the Hypervariable Stem Domain of Vertebrate α1,3/1,4-Fucosyltransferases Determines the Type 1/Type 2 Transfer

Alignment of 15 vertebrate α1,3-fucosyltransferases revealed one arginine conserved in all the enzymes employing exclusively type 2 acceptor substrates. At the equivalent position, a tryptophan was found in FUT3-encoded Lewis α1,3/1,4-fucosyltransferase (Fuc-TIII) andFUT5-encoded α1,3/1,4-fucosyltransferase, the only fucosyltransferases that can also transfer fucose in α1,4-linkage. The single amino acid substitution Trp111 → Arg in Fuc-TIII was sufficient to change the specificity of fucose transfer from H-type 1 to H-type 2 acceptors. The additional mutation of Asp112 → Glu increased the type 2 activity of the double mutant Fuc-TIII enzyme, but the single substitution of the acidic residue Asp112 in Fuc-TIII by Glu decreased the activity of the enzyme and did not interfere with H-type 1/H-type 2 specificity. In contrast, substitution of Arg115 in bovinefutb-encoded α1,3-fucosyltransferase (Fuc-Tb) by Trp generated a protein unable to transfer fucose either on H-type 1 or H-type 2 acceptors. However, the double mutation Arg115 → Trp/Glu116 → Asp of Fuc-Tb slightly increased H-type 1 activity. The acidic residue adjacent to the candidate amino acid Trp/Arg seems to modulate the relative type 1/type 2 acceptor specificity, and its presence is necessary for enzyme activity since its substitution by the corresponding amide inactivated both Fuc-TIII and Fuc-Tb enzymes.

Fucosyltransferases are type II transmembrane proteins catalyzing fucose transfer from GDP-fucose to different oligosaccharide acceptors in ␣1,2-, ␣1,3-, ␣1,4-, and ␣1,6-linkages. The fucosylated glycoconjugates they produce are blood group and oncodevelopmental antigens (1) and are involved in tumorigenesis (2), embryogenesis (3), normal leukocyte trafficking (4), and leukocyte extravasation in inflammatory reactions (5)(6)(7). Since all fucosyltransferases utilize GDP-fucose as donor sub-strate, their specificity depends on the recognition of the acceptor substrate and the type of linkage formed. Although the primary sequences of six human ␣1,3-fucosyltransferases are known, the amino acids involved in the recognition of different acceptor substrates as type 1 (Gal␤1,3GlcNAc) and type 2 (Gal␤1,4GlcNAc) disaccharides or blood group H-type 1 (Fuc␣1,2Gal␤1,3GlcNAc) and H-type 2 (Fuc␣1,2Gal␤1, 4GlcNAc) trisaccharides are not yet known. These last trisaccharides are better acceptors than the disaccharides because they give higher values of fucose incorporation with a better K m and they are unambiguous in the sense that C-2 of Gal is already substituted by Fuc, and therefore, they can accept fucose only on C-3 or C-4 of GlcNAc. This is particularly relevant for Fuc-TIII 1 since up to 4% of ␣1,2-fucosyltransferase activity has been found with this enzyme (8,9).
Two enzymes (Fuc-TIII and Fuc-TV) are able to use both H-type 1 and H-type 2 trisaccharide acceptors, and consequently, they have been called ␣1,3/1,4-fucosyltransferases. The Lewis or Fuc-TIII enzyme is ϳ100 times more efficient on H-type 1 compared with H-type 2 acceptor substrates (10). The Fuc-TV enzyme is more efficient on H-type 2 than on H-type 1 substrates, although the relative type 1/type 2 activities are of the same order of magnitude (10). Finally, the remaining four ␣1,3-fucosyltransferases (Fuc-TIV, Fuc-TVI, Fuc-TVII, and Fuc-TIX) are able to use only type 2 acceptor substrates and are consequently called ␣1,3-fucosyltransferases.
The Fuc-TIII, Fuc-TV, and Fuc-TVI enzymes constituting the primate Lewis subfamily of fucosyltransferases have appeared by two successive duplications of an ancestral Lewis gene, which occurred rather late in evolution (Fig. 1), after the great mammalian radiation and before the separation of higher apes and man from their common evolutionary path (10). These three enzymes share ϳ85% sequence identity; the main differences among them are located in the stem amino-terminal region (hypervariable region), whereas their carboxyl-terminal regions are almost identical. Therefore, the differences in type 1/type 2 specificity among these three enzymes are expected to be determined by amino acid differences in their hypervariable regions. Two different strategies have been developed to define the amino acids involved in the relative differences of type 1/type 2 acceptor substrate specificity.
Subdomain swapping of segment 62-110 of Fuc-TIII and the corresponding segment of Fuc-TV increased the type 1 enzyme activity of Fuc-TV (11). Eight amino acids were shown to be Fuc-TV-specific in this area. Site-directed mutagenesis of two of them (Asn 86 and Thr 87 ) by the corresponding His and Ile of Fuc-TIII also increased type 1 enzyme activity (12), whereas site-directed mutagenesis of other amino acids in the central area (13) or in the COOH terminus (14) of Fuc-TIII and Fuc-TV did not modify the relative type 1/type 2 acceptor substrate specificity.
Lowe and co-workers (15) preferred to analyze the more clear-cut difference between Fuc-TIII and Fuc-TVI. They divided the hypervariable region into five subdomains and created enzyme chimeras by subdomain swapping. Functional analysis of the chimeras showed that the type 1/type 2 specificity of human Fuc-TIII and Fuc-TVI depends on amino acids in the segment corresponding to residues 103-153 of Fuc-TIII. Eleven amino acids in this area were specific to Fuc-TVI and were considered as candidates to determine whether or not this ␣1,3-fucosyltransferase can utilize, in addition, type 1 acceptor substrates (15). These 11 amino acids could be reduced to four following the cloning, expression, and peptide sequence analysis of chimpanzee Fuc-TIII, Fuc-TV, and Fuc-TVI (10) and the bovine Fuc-Tb enzyme (16). Furthermore, this number could be further reduced to two amino acids specific to Fuc-TVI (Arg 110 and Glu 111 ) and Fuc-Tb (Arg 115 and Glu 116 ) by addition of the hamster Fuc-Th enzyme 2 to the previous analysis. 3 This study was conducted to define, by site-directed mutagenesis, the contribution to H-type 1/H-type 2 acceptor substrate specificity of the two amino acids Arg 115 and Glu 116 of Fuc-Tb and the corresponding residues Trp 111 and Asp 112 of Fuc-TIII.
Site-directed Mutagenesis-Cloned futb (16) with an additional 3Јuntranslated sequence (800 base pairs) that gives better expression 4 and FUT3 (17,18) 5 were incorporated in the mammalian expression vector pcDNAI/Amp (Invitrogen, Carlsbad, CA) and used for mutagen-esis. Propagation of these vectors was achieved in the Escherichia coli XL1-Blue strain. Polymerase chain reaction-based mutagenesis (ExSite kit, Stratagene, La Jolla, CA) was used for all mutations, except for that which generated the Arg 115 3 Trp/Glu 116 3 Asp changes in Fuc-Tb. The reaction was performed by incubating a vector (0.1 pmol) with 15 pmol of each primer, one harboring the mutation ( Table I). The reaction was cycled for 20 rounds (1 min at 94°C, 2 min at 52-60°C, and 1 min at 72°C) with a highly efficient and reliable Taq DNA polymerase (Stratagene), and then the last extension step was made at 72°C for 5 min. Parental DNA was removed by DpnI restriction endonuclease treatment, and any extended bases produced during the polymerase chain reaction amplification were eliminated by Pfu DNA polymerase. The amplified vector with the expected mutation was used for ligation and quick transformation of the E. coli XL1-Blue strain.
Fucosyltransferase Assay-Fucosyltransferase assays were incubated for 1 h at 37°C in a 60-l volume reaction containing 25 mM sodium cacodylate (pH 6.5), 5 mM ATP, 20 mM MnCl 2 , 10 mM ␣-L-fucose, 3 M GDP-[ 14 C]fucose (310 mCi/mmol; Amersham Pharmacia Biotech), 0.1 mM trisaccharide acceptor, and 50 g of COS-7 extract proteins. The reaction was stopped by addition of 3 ml of cold water and applied to a Waters Sep-Pak C 18 reverse chromatography cartridge. After washing with 15 ml of water, the radiolabeled reaction product was eluted with 2 ϫ 5 ml of ethanol and counted with 2 volumes of biodegradable counting scintillant (Amersham Pharmacia Biotech).
Determination of Kinetic Parameters-The apparent K m values for GDP-fucose of the wild-type enzyme and mutated variants were determined using 10 -250 M GDP-fucose with 8 M GDP-[ 14 C]fucose in each FIG. 1. Phylogenetic tree of the main subfamilies of vertebrate ␣1,3fucosyltransferases. All these enzymes can use type 2 acceptor substrates; but Fuc-TIII and Fuc-TV can use, in addition, type 1 acceptors, and they are the consequence of the latest gene duplication event within the Lewis subfamily of enzymes. Therefore, it is logical to assume that the capacity to use type 1 acceptors had appeared in the common ancestor (q) of these two enzymes. Values of 100 bootstrap replicates are noted at each divergence point. Sequence Analysis-Multiple alignments were performed with Clust-alW version 1.7 (21), and the phylogenetic tree was made by neighbor joining with Seaview and Phylo_win (22). 6

RESULTS
Peptide Alignments of ␣1,3/4-Fucosyltransferases-All previous attempts to define the amino acids involved in the determination of the type 1/type 2 acceptor substrate specificity of Fuc-TIII were performed within the Lewis subfamily of enzymes, comparing either Fuc-TIII and Fuc-TV (11,12) or Fuc-TIII and Fuc-TVI (10,15). These studies led to the conclusion that the acceptor specificity probably resides in the hypervariable stem region.
In this study, we have extended the peptide sequence comparisons of this hypervariable region to all other ␣1,3-fucosyltransferase subfamilies of enzymes: myeloid (Fuc-TIV), leukocyte (Fuc-TVII), and brain (Fuc-TIX). This new multialignment showed that three amino acids are invariant in this region (Val-X-X-His-His), and only one amino acid (boldface Trp in Table II) appears to be specific for enzymes using both type 1 and type 2 acceptors. This amino acid is replaced by Arg (boldface in Table II) in all the enzymes using only type 2 acceptors and is always followed by an acidic residue (boldface Asp or Glu in Table II). The levels of expression of all the recombinant Fuc-TIII enzymes in COS-7 cells were evaluated by Western blotting, and as shown in Fig. 2, some variations occurred. Nevertheless, the enzyme assays were conducted on crude protein extracts from transfected COS-7 cells.

Preparation of Fuc-TIII and Fuc-Tb Mutants-Site-directed
Acceptor Substrate Specificity of Fuc-TIII Mutants-For each Fuc-TIII mutant, fucose transfer reactions were conducted with H-type 1 and H-type 2 acceptors. Under the experimental conditions used, Fuc-TIII had a very weak activity with the H-type 2 acceptor compared with that obtained with the H-type 1 acceptor (Table III).
The recombinant Fuc-TIII enzyme sharing the single mutation Trp 111 3 Arg lost its type 1 acceptor activity and acquired type 2 acceptor activity. The double mutation Trp 111 3 Arg/ Asp 112 3 Glu conferred to the recombinant enzyme higher activity (ϳ2-fold increase with the H-type 2 acceptor), which became equivalent to the activity obtained with the native Fuc-TIII enzyme using the H-type 1 acceptor. The single conservative substitution of Fuc-TIII (Asp 112 3 Glu) decreased by about half the activity of the enzyme without changing its relative type 1/type 2 acceptor substrate specificity, suggesting that the nature of the acidic residue (Asp or Glu) does not by itself determine the H-type 1/H-type 2 transfer activity, although this acidic residue can help to modulate the relative rates of enzyme activities.
Acceptor Substrate Specificity of Fuc-Tb Mutants-The native bovine ␣1,3-fucosyltransferase Fuc-Tb utilizes exclusively type 2 acceptors (16). Any single substitution of the Fuc-Tb enzyme at each of the candidate amino acids (Arg 115 3 Trp or Glu 116 3 Asp) generated an inactive enzyme with both acceptor substrates (Table III). Only a very small increase in activity on the H-type 1 acceptor was observed from 7 Ϯ 0.5 to 11 Ϯ 0.5 pmol/h/mg of protein with the double mutation Arg 115 3 Trp/ Glu 116 3 Asp, which also conserved a weak activity (58 Ϯ 3 pmol/h/mg of protein) on the H-type 2 acceptor. Table II, the candidate amino acid involved in H-type 1/H-type 2 specificity is followed by an acidic residue (Asp or Glu). Aspartic acid is found in enzymes able to transfer fucose either on the H-type 1 or H-type 2 acceptor (Fuc-TIII or Fuc-TV), but also in enzymes acting only on the H-type 2 acceptor (Fuc-TIV, and Fuc-TIX), whereas Glu is present exclusively in ␣1,3-fucosyltransferases such as Fuc-TVI, Fuc-TVII, Fuc-Tb, and Fuc-Th.

Involvement of Acidic Residue Asp 112 of Fuc-TIII and Glu 116 of Fuc-Tb in Enzyme Activity-In all ␣1,3-fucosyltransferases listed in
The presence of Asp or Glu in wild-type Fuc-TIII or its variants seems to modulate enzyme efficiency (Table III). However, in Fuc-Tb, only the presence of Glu, as in the native enzyme, was able to confer activity to the protein. Indeed, the Glu 116 3 Asp change was associated with enzyme activity loss, whatever the substrate acceptor. 6 These programs are available on the server from Cis Infobiogen (E-mail: bioinfo@infobiogen.fr; WEB: http://www.infobiogen.fr/).

321-302
5Ј-GTGCACCACTGGGAAATCATG-3Ј (sense) To define more precisely the requirement of the acidic residue for enzyme activity, Asp 112 of Fuc-TIII was substituted by the corresponding amide Asn. The Fuc-TIII mutant (Asp 112 3 Asn) became inactive with both acceptors (Table III). The same kind of result was obtained with modified Fuc-Tb (Glu 116 3 Gln), which was also unable to transfer fucose on either of the two acceptor substrates (Table III).
Kinetic Parameters of Native and Mutated Enzymes-For Fuc-TIII variants, the apparent K m values for GDP-fucose (ϳ 30 M) were similar to that of the parental enzyme (Table IV). Therefore, amino acid changes in enzyme variants did not modify the affinity of the Fuc-TIII variants for the donor substrate. However, significant changes in the K m values for the H-type 1 and H-type 2 acceptor substrates were observed for native Fuc-TIII compared with the Fuc-TIII mutants. As ex-pected, native Fuc-TIII and Fuc-TIII with the Asp 112 3 Glu mutation had better K m values for H-type 1 than for H-type 2 acceptors, whereas the two Fuc-TIII mutants with the Trp 111 3 Arg change had a good affinity for the H-type 2 acceptor and a poor affinity for the H-type 1 acceptor. It also appeared that the nature of the acidic residue (Asp or Glu) adjacent to the candidate amino acid (Trp or Arg) had a significant effect on the kinetics of the recombinant enzyme. The Trp-Asp association found in wild-type Fuc-TIII is favorable for H-type 1 activity, whereas the Arg-Glu association found in double mutant Fuc-TIII is convenient for H-type 2 activity (Table IV). DISCUSSION Fucosyltransferases have a common domain structure including a short NH 2 -terminal cytoplasmic tail, a signal anchor domain, a stem region, and a globular COOH-terminal catalytic domain. Truncation at the COOH terminus of one or more amino acids induces a dramatic loss of enzyme activity, whereas truncation at the NH 2 terminus of as much as 61 amino acids for Fuc-TIII or 75 amino acids for Fuc-TV does not alter the enzyme activity (11,23). First one (24,25) and then two (26) peptide conserved motifs were described in the COOHterminal catalytic domain of all the ␣1,3-fucosyltransferases, and they are presumed to be involved in GDP-fucose binding (26,27), whereas the acceptor substrate-binding domain has been tentatively ascribed to a portion of the hypervariable region comprised between either positions 62 and 110 (11) or positions 103 and 153 (15) of Fuc-TIII.
Based on our previous results (10), a short peptide segment corresponding to the sequence Pro 101 -Leu 121 of Fuc-TIII was used for peptide sequence alignment of all known vertebrate ␣1,3-fucosyltransferases. A single amino acid, Arg in ␣1,3fucosyltransferases and Trp in ␣1,3/1,4-fucosyltransferases, was expected to contribute to the type 1/type 2 acceptor specificity. By site-directed mutagenesis, we have now demonstrated that the single substitution of Trp 111 by Arg conferred to the recombinant Lewis enzyme the ability to use efficiently H-type 2 instead of H-type 1 acceptor substrate. This newly acquired activity increased when, in addition, Asp 112 was replaced by Glu. The involvement of Arg in fucose transfer was confirmed by the loss of Fuc-Tb activity when Trp substituted for Arg 115 .
Several invertebrate and bacterial ␣1,3-fucosyltransferases have been recently found (reviewed in Ref. 28). Two Helicobacter pylori enzymes have been cloned and expressed, and TABLE II Alignment of the peptide sequences of the known vertebrate ␣1,3 and ␣1,3/1,4-fucosyltransferases around the two candidate amino acids WD and RE or RD for the determination of type 1 and type 2 acceptor substrate specificity Asterisks identify the four previously defined potential candidate amino acids based on the multiple alignment of the Lewis subfamily of enzymes (10). Hyphens indicate amino acids identical to the human Fuc-TIII enzyme.    (24,25) and have the similarly charged amino acid Lys instead of Arg in the candidate position. A Caenorhabditis elegans ␣1,3-fucosyltransferase has also been cloned and expressed (29), and it has Gln in the candidate position, but it can transfer Fuc onto acceptors as GalNAc-␤1,4GlcNAc␤1-R to generate GalNAc␤1,4(Fuc␣1,3)GlcNAc-␤1-R. These nematode oligosaccharides are different from all known type 1 or type 2 oligosaccharides of the lacto or neolacto series. The remaining invertebrate and bacterial putative ␣1,3fucosyltransferases have only been defined by sequence homology to vertebrate enzymes, and nothing is known about their enzyme activity or about the structure of the acceptor substrates used. Nevertheless, some of them did not have either Trp or Arg in the corresponding candidate position of the putative acceptor-binding domain, suggesting that as in the case of C. elegans, other acceptors might be used by these enzymes. However, the overall structure of the acceptor substrate-binding domain is, in general, preserved with a conserved His residue before and a carboxylic group (Asp or Glu) after the candidate position. Previous work by Hindsgaul et al. (30) suggested that oligosaccharide-reactive acceptor hydroxyl groups are involved in a critical hydrogen bond donor interaction with the glycosyltransferases. The main key polar groups on the oligosaccharide acceptors have been identified as the reactive hydroxyls at C-3 or C-4 of GlcNAc and C-6 of Gal (31,32). On the enzyme side, different amino acids can be involved in this hydrogen bond, but typical hydrogen bond acceptors are His and carboxylates (26), as those found in the conserved amino acid positions flanking the candidate Arg or Trp residue for the definition of the type 1/type 2 enzyme activity. Therefore, these His, Glu, and Asp residues can be considered as candidates for the formation of hydrogen bonds with the acceptor oligosaccharide. Based on recent studies, it seems reasonable to suggest that the active-site base in the protein may be a carboxylate anion (33,34). In another study, Britten and Bird (35) investigated the amino acids essential for the activity of Fuc-TVI through chemical modification, and they concluded that the substratebinding site of the enzyme possesses His residue(s) that are essential for enzyme activity.
It has been suggested, in the case of the oligosaccharyltransferase, that the divalent cation cofactor might be in close proximity to the acceptor oligosaccharide substrate (36). A similar mechanism could occur in fucosyltransferases, and it was recently suggested that Mn 2ϩ could interact with a carboxylate anion (Glu or Asp) in the fucosyltransferase on one side and with the oligosaccharide acceptor substrate on the other side (37). The complete loss of enzyme activity by the substitution Asp 112 3 Asn in Fuc-TIII or Glu 116 3 Gln in Fuc-Tb suggests that this carboxylate group is necessary for enzyme activity, and therefore, it constitutes a possible candidate for the activesite carboxylate anion.
The Arg 115 3 Trp substitution in Fuc-Tb was not sufficient to change the specificity of the enzyme toward type 1 acceptors even with the additional mutation Glu 116 3 Asp. Therefore, Trp seems to contribute to a type 1 activity, but it is not sufficient by itself; and even more than the two tested candidates, Trp and Asp, are probably necessary to change the Fuc-Tb specificity. This is in accordance with the fact that only a very small increase in the number of cells producing Le a and sialyl-Le a antigens was observed in cells transfected by a Fuc-TVI chimera containing Fuc-TIII subdomains 4 and 5 (positions 103-153) (15). In another work (12), it was shown that two other amino acid changes in Fuc-TV (Asn 86 3 His and Thr 87 3 Ile) increased the type 1 activity of the recombinant enzyme.
The molecular phylogeny of fucosyltransferase genes suggested that the duplication events at the origin of the bovine futb (16) and primate FUT6 genes occurred before the duplication that produced the primate FUT3 and FUT5 genes (Fig. 1). Hence, it seems that the common ancestor of FUT3 and FUT5 genes has acquired the capacity to use type 1 acceptors without loss of the ability to use type 2 acceptors. This would help to explain the change in enzyme specificity by a single amino acid substitution in Fuc-TIII, whereas more complex changes are expected in Fuc-Tb or Fuc-TVI to generate a strong type 1 acceptor activity because other independent mutations might have occurred in Fuc-TVI and/or Fuc-Tb since their earlier divergence from the common evolutionary path (Fig. 1).