And A3/4-Fucosyltransferases: the A1,4 Fucosylation Requires an Aromatic Residue in the Acceptor-Binding Domain

Glycobiology vol. 14 no. 4 pp. 347±356, 2004 DOI: 10.1093/glycob/cwh053 Advance Access publication on January 12, 2004 Structure/function study of Lewis a3- and a3/4-fucosyltransferases: the a1,4 fucosylation requires an aromatic residue in the acceptor-binding domain

Fabrice Dupuy1, Agnes Germot, Raymond Julien, and the corresponding enzymes transfer fucose in a1,3 link- Abderrahman Maftah age on peripheral GlcNAc of type 2 acceptor substrates (Galb1,4GlcNAc). This is the case for Homo sapiens, Unite de Genetique Moleculaire Animale, UMR 1061 Universite-INRA, GDR-CNRS 2590, Institut des Sciences de la Vie et de la Sante, Faculte where FUT3, 4, 5, 6, 7, and 9 correspond to a3-FUTs des Sciences et Techniques, 87060 Limoges, France (Kaneko et al., 1999; Staudacher et al., 1999). However, these enzymes can be distinguished by their different tissue Received on October 27, 2003; revised on December 30, 2003; accepted on December 30, 2003 distributions and by their distinct affinities for type 2 substrates (Cameron et al., 1995; Kaneko et al., 1999; Mollicone All vertebrate a3- and a3/4-FUTs possess the characteristic et al., 1995). Among a3-FUTs, FUT3 and FUT5, two mem- acceptor-binding motif VxxHH(W/R)(D/E). FUT6 and bers of the Lewis FUT family, also transfer fucose in a1,4 FUTb enzymes, harboring R in the acceptor-binding motif, linkage on type 1 substrates (Galb1,3GlcNAc) and were transfer fucose in a1,3 linkage, whereas FUT3 and FUT5 consequently named a3/4-FUTs. They are involved in enzymes with W at the candidate position can also transfer Lewis blood group antigen synthesis, such as Lea and Leb fucose in a1,4 linkageÐFUT3 being more efficient than antigens (Costache et al., 1997a; Oriol, 1995). FUT3 FUT5. To determine the involvement of the W/R residue enzymes have a predominant a1,4 activity, whereas FUT5 in acceptor recognition, we produced 34 variants of human enzymes add fucose with the same efficiency on type 1 FUT3, FUT5, FUT6, and ox FUTb Lewis enzymes. Among and type 2 acceptors (Costache et al., 1997b). In the animal the FUT3 variants where W111 was replaced by the other kingdom, the ability to link fucose in a1,4 linkage has amino acids, only enzymes with an aromatic residue at the been proved for Primates (Dupuy et al., 2002), whereas candidate position kept about 50% of a1,4 activity and showed this activity is widely spread in the plant kingdom (Costa et al., 2002; Fitchette et al., 1999). a1,4 fucosylation is even no changes in Km values for GDP-Fuc donor and H-type 1 acceptor substrates. All other substitutions produced enzymes detected in the Helicobacter pylori strain UA948 (Rasko with less than 20% of the a1,4 activity. Thus the ability of et al., 2000), but it would rather result from a lateral gene a3/4-FUTs to recognize type 1 substrates involves the aro- transfer. matic character of W in the acceptor-binding domain. The The a3- and a3/4-FUTs have a putative common type II a1,3 activity of FUT6 and FUTb significantly decreased structure (Figure 1). Comparison of their peptide sequences when their R residue was substituted by basic or charged localizes most of the amino acid differences in the amino- residues. Moreover, FUT3 and FUT5 variants with W!R terminal part (variable segment), whereas numerous identi- substitution had a better affinity for H-type 2 substrate and ties are present in the carboxy-terminal portion (constant higher a1,3 activities. Therefore the optimal fucose addition segment). Four highly conserved sequences were described in a1,3 linkage requires the R residue in the acceptor-binding all along the peptide sequence (Figure 1). Motifs I and II motif of Lewis FUTs. are localized in the constant segment and are specific to a3- and a3/4-FUTs (Oriol et al., 1999). In the variable seg- Key words: aromatic amino acid/acceptor substrate ment, motif III contains the first amino acids required for specificity/conserved peptide motifs/Lewis the correct protein folding and catalysis (Dupuy et al., fucosyltransferase/site-directed mutagenesis 2002). Indeed, amino acid deletions occurring in motif III of human FUT3 and FUT5 gave inactive enzymes (Xu et al., 1996). Conversely, deletions upstream of motif III did not Introduction alter the activity (Xu et al., 1996), although mutations in this region disturbed the subcellular localization of the All a3-fucosyltransferases (FUTs) transfer fucose from the enzyme (Sousa et al., 2003). The fourth motif, called the GDP-fucose donor substrate to GlcNAc residues belonging acceptor-binding motif (abm), intercedes in acceptor sub- to the core unit of N-glycans or peripheral lactosamine strate binding and specificity (Dupuy et al., 1999; Sherwood structures of glycoconjugates. Owing to its positions in et al., 2002). glycans, fucose plays a significant role in many biological Without any a3- or a3/4-FUT structure, FUT3-FUT5- processes (for reviews, see Staudacher et al., 1999; FUT6 Lewis enzymes were mainly studied to identify the Becker and Lowe, 2003). In Vertebrates, although several key amino acids involved in enzyme specificity and activity a3-fucosyltransferase genes exist in the same genome, all (for review, see De Vries et al., 2001). Indeed, although these enzymes show more than 85% peptide sequence conser- 1To whom correspondence should be addressed; e-mail: vation,theypossessdifferentacceptorsubstrateaffinitiesand [email protected] specificities. Structure±function studies have demonstrated

Fig. 1. Schematic representation of vertebrate a3- and a3/4-FUTs. These enzymes are type II Golgi-anchored glycoproteins with a short NH2 cytoplasmic tail (cyt. tail), a transmembrane domain (TMD) followed by a luminal COOH segment, composed of a stem and a catalytic domain. In these enzymes, most amino acid differences are localized in the amino terminal part of these enzymes (variable segment), and their catalytic domains are highly conserved (constant segment). The peptide motifs characteristic of a3- and a3/4-FUTs are indicated: the motifs I and II are mark domains (Oriol et al., 1999); motif III could correspond to the beginning of the catalytic domain (Dupuy et al., 2002); amino acids of the acceptor-binding motif (abm) are involved in the acceptor substrate recognition and specificity (Dupuy et al., 1999). The alignment of peptide sequences between motifs III and I of enzymes used in this study is revealed. The amino acids involved in type 1 acceptor substrate specificity are indicated by stars (Legault et al., 1995) and arrows (Nguyen et al., 1998).

that some residues in the constant domain are involved in Results GDP-fucose recognition (Holmes et al., 1995; Sherwood Acceptor-binding motif of vertebrate a3- and a3/4-FUTs et al., 2000), catalysis (Sherwood et al., 1998; Vo et al., 1998), or in disulfide bonding (Holmes et al., 2000). Alter- Acceptor substrate specificities of a3- and a3/4-FUTs natively, residues of the variable segment could determine partly depend on amino acids localized between motifs I a3- and a3/4-FUT acceptor substrate specificity. Eleven and III of these enzymes. The alignment of 15 vertebrate amino acids (Figure 1) were identified as potential candi- a3- and a3/4-FUT sequences helped to define the VxxHH dates for the determination of the type 1 acceptor substrate (W/R)(D/E) consensus acceptor-binding motif (Dupuy specificity of human FUT3 (Legault et al., 1995). Down- et al., 1999). Since then, several new a3- and a3/4-FUTs stream of the motif III, two other residues, H73 and I74 of from a broad panel of species have been identified on the human FUT3, were found to be necessary for fucose trans- basis of their activities and their sequences (Table I). For fer on type 1 acceptors (Nguyen et al., 1998). In a previous this study, we cloned and characterized new simian Lewis- work, we showed that among the residues belonging to the like FUTs (underscored in Table I). In all these new verte- acceptor-binding motif VxxHH(W/R)(D/E), where x corre- brate sequences, the characteristics of the acceptor-binding sponds to a hydrophobic residue, the W residue conserved motif were retrieved that is, a V residue, followed by for a3/4-FUTs and the R residue conserved for a3-FUTs two variable hydrophobic amino acids, two conserved are involved in acceptor substrate specificity (Dupuy H residues, a W or R residue, and an acidic residue. As et al., 1999). The single substitution of W111 by R in the expected, enzymes transferring fucose only in a1,3 linkage VxxHHWD motif of human FUT3 (primarily an a4-FUT) have R in their acceptor-binding motif, whereas enzymes changed the specificity of fucose transfer from H-type 1 to also able to transfer fucose in a1,4 linkage have W at the H-type 2. This change is associated with affinity increase same position. and decrease for H-type 2 and H-type 1 acceptors, respectively (Dupuy et al., 1999). Characterization of new primate Lewis enzymes confirmed the relationship between the Involvement of W residue of the acceptor-binding motif amino acid composition of this motif and the Lewis FUT in a1,3- and a1,4-FUT activities acceptor specificity (Dupuy et al., 2002). Recently, FUT3 enzymes transfer fucose with a1,4 linkage and, less Sherwood et al. (2002) reported that the two conserved H efficiently, with a1,3 linkage. Their higher a1,4 activity residues of this motif are necessary for optimal activity and is explained by their better affinity for type 1 substrates. acceptor binding. To establish the role of W residue of the acceptor-binding The present study focuses on the W/R residue that fol- motif in acceptor substrate specificity, the W111 of human lowed the conserved H doublet in the highly conserved FUT3 was alternately substituted by the 19 other amino acceptor-binding motif of vertebrate a3- and a3/4-FUTs. acids. Activities of wild-type and mutated enzymes were To determine its involvement in a1,3- and a1,4-FUT activ- determined using H-type 1 (Fuca1,2Galb1,3GlcNAc) ities, we produced, by site-directed mutagenesis at this posi- and H-type 2 (Fuca1,2Galb1,4GlcNAc) trisaccharides as tion, 34 mutated forms of human FUT3, FUT5, FUT6, and acceptor substrates, because they produce the best activity ox FUTb. Characterization of these enzymes demonstrated (Costache et al., 1997b; Dupuy et al., 1999) and make it that significant a1,4 fucosylation requires an aromatic resi- possible to analyze only a1,4 and a1,3 fucose addition on due at the position of W/R residue. Conversely, a1,3 fuco- GlcNAc, respectively. Substitutions of W111 by aromatic sylation is optimal only when an R residue is present at the residues (Y or F) decreased the a1,4 activity of the enzyme candidate position. to 58% (W111!Y) and 42% (W111!F) of the wild-type 348 Acceptor substrate specificity of Lewis FUTs

Table I. Peptide alignment of the acceptor-binding domain of vertebrate a3- and a3/4-FUTs

The peptide sequences of new vertebrate a3- and a3/4-FUTs are aligned by reference to human and ox Lewis enzymes, which are used in this study. The W/R residue, involved in the acceptor substrate specificity of vertebrate a3- and a3/4-FUTs (Dupuy et al., 1999), is in a white character on black background. The new simian Lewis-like sequences characterized for this study are underlined. nd, not determined. activity (Figure 2A). Four other mutants (W111!A, blotanalysis using anti-human FUT3antibodies(Figure2B). W111!M, W111!S, and W111!H) conserved 10±20% of The detection of two forms could correspond to two dif- the wild-type a1,4 activity, whereas replacements of W111 ferently glycosylated FUT3 enzymes (Christensen et al., by E, D, or K produced enzymes with undetectable a1,4 2000). activity. Alternatively, all changes generated enzymes with Except for the rhesus macaque FUT5, which has R a weak a1,3 activity except the W111!R substitution, instead of W at the candidate position and only uses which changed type 1 acceptor specificity of human type 2 substrates (Dupuy et al., 2002), FUT5 enzymes use FUT3 toward type 2, giving an enzyme with preferential type 1 and type 2 acceptor substrates. Their substrate spe- a1,3 activity (equivalent to 60% of wild-type a1,4 activity, cificities differ from FUT3 ones in the sense that they more Figure 2A). In all cases, the recombinant enzymes were efficiently transfer fucose on H-type 2 acceptor substrates. efficiently expressed in COS-7 cells because wild-type and Indeed, the a1,3 fucosylation of FUT5 enzymes is three mutated forms of FUT3 were similarly detected by western times higher, compared to a1,4 fucosylation (Table II). 349 F. Dupuy et al.

Fig. 2. Enzyme activities and expressions of wild-type human FUT3 and its variants mutated on W111.(A) Fucose transfer activities were measured in the presence of 0.1 mM H-type 1 (black bars) or H-type 2 (gray bars). Incubation times were 1 h for H-type 1 acceptor substrate and 3 h for H-type 2. 100% activity corresponds to 1667 pmol/h/mg proteins. The order of substituted amino acids was established according to type 1 decreasing activity. (B) Immunodetection of recombinant enzymes expressed in COS-7 cells. The molecular weight deduced from protein sequence of wild-type human FUT3 is 42.1 kDa. Control (Ct) corresponds to protein extract of COS-7 cells transfected by the pcDNA1/Amp vector alone.

Table II. a1,3- and a1,4-FUT activities of human FUT5 and its activity, and V, which nearly cancels both activities. W124 mutated forms substitutions by Y, A, or F residue produced, respectively, enzymes with 52%, 20%, and 17% of the wild-type a1,4 Relative FUT activities (%) activity. Substitutions by R or V produced enzymes with undetectable a1,4 activity (Table II). The W124!Y/A/F a1,4 Fucosylation a1,3 Fucosylation replacements generated enzymes with similar (W!A) or Enzyme (H-type 1 substrate) (H-type 2 substrate) weaker (W!Y/F) a1,3 activity compared to the wild- 124 Control 0 0 type one (Table II). Interestingly, the W !V gave an enzyme with unchanged a1,3 activity, whereas the W124! FUT5 29 100 R substitution produced an enzyme with a 3.8-fold increase 124 FUT5 W !Y 15 59 of a1,3 activity (Table II). In these experiments, the expres- FUT5 W124!A 6 92 sion levels of the modified proteins were similar to the FUT5 W124!F 5 15 wild-type one, as estimated by western blot (data not FUT5 W124!V 0 102 shown). FUT5 W124!R 0 380 Involvement of R residue of the acceptor-binding motif Activities were measured after either 3 h or 1 h incubation with H-type 1 in a1,3-FUT activity or H-type 2 acceptor substrates, respectively. 100% activity corresponds The involvement of R residue in a1,3 fucosylation was to 1382 pmol/h/mg protein and control to crude protein extract from tested on the human a3-FUT FUT6. The R110 of human COS-7 cells transfected by pcDNA1/Amp alone. FUT6 was replaced by only two other basic residues (K and H), two charged amino acids (N and Q), and by W. Sub- stitutions by H, N, K, and Q generated enzymes with less The W124 in the acceptor-binding motif of human FUT5 than 10% of the wild-type a1,3 activity and with undetect- was substituted by amino acids (which were brought out able a1,4 activity (Table III). These decreases in activity by the human FUT3 study) Y, F, and A to test modifica- were not due to different expression levels of mutated tion in a1,4 activity, R to test modification in a1,3 enzymes in COS-7 cells (Figure 3). Replacement of R110 350 Acceptor substrate specificity of Lewis FUTs

Table III. a1,3- and a1,4-FUT activities of human FUT6 and its by W in human FUT6 led to an enzyme without a1,3 and mutated forms a1,4 activities (Table III). The same experiment was performed with FUTb, an a3- Relative FUT activities (%) FUT encoded by the ox futb Lewis gene (Oulmouden et al., 1997). Indeed, futb evolved from an ancestor gene which a1,4 Fucosylation a1,3 Fucosylation duplicates in primate lineage to form the Lewis gene family Enzyme (H-type 1 substrate) (H-type 2 substrate) (Dupuy et al., 2002; Oulmouden et al., 1997). Mutations of 115 Control 0 0 R , in the VxxHHRE acceptor-binding motif, by H, N, K, FUT6 0 100 or Q residue gave recombinant enzymes with less than 10% 110 of wild-type a1,3 activity (data not shown), results quite FUT6 R !H 0 8.8 similar to that obtained for mutated forms of human FUT6. FUT6 R110!N 0 7.2 The substitution by W led to poor a1,3 and a1,4 activities, FUT6 R110!K 0 3.8 as has already been described (Dupuy et al., 1999). There- FUT6 R110!Q 0 3.8 fore, none of the amino acids tested at the position of the FUT6 R110!W 0 0 W/R in the acceptor-binding motif of a3-FUTs generated enzymes that efficiently transfer fucose in a1,3 and/or a1,4 linkages. Activities were measured after either 3 h or 1 h incubation with H-type 1 or H-type 2 acceptor substrates, respectively. 100% activity corresponds to 1215 pmol/h/mg proteins and control to crude protein extract from Kinetic parameters of mutated enzymes COS-7 cells transfected by pcDNA1/Amp alone. The kinetic parameters of human FUT3 and FUT5 wild- type enzymes, and some of their mutated forms that present strong activity, were determined for the nucleotide sugar donor substrate and for the best acceptor substrate (Table IV). The results showed that the apparent Km values for GDP-fucose were similar for wild-type and mutated FUT3andFUT5enzymes.Consequently,aminoacidchanges at the W position in the acceptor-binding motif did not modify the affinity of the mutated enzymes for the donor substrate. Significant changes in the Km values for acceptor substrates were observed compared to the wild-type enzymes for some FUT3 and FUT5 variants. The FUT3 mutated Fig. 3. Expression analyses of recombinant human FUT6 and its 111 variants mutated on R110. Detection of recombinant FUT6 enzymes was enzymes with W !Y or F substitutions had slightly higher carried out with polyclonal anti-human FUT3 antibodies. The molecular Km values for H-type 1 acceptor oligosaccharide, and the 111 weight deduced from protein sequence of wild-type enzyme is 41.8 kDa. W !A change generated a 10-fold increase of the Km Lane Ct, negative control, 50 mg proteins from COS-7 cells transfected value for the same acceptor. In FUT5, the W124!Y muta- with pcDNA1/Amp alone; lane 1, wild-type FUT6 enzyme; lane 2, mutated FUT6 with R110!K substitution; lane 3, mutated FUT6 with tion produced an enzyme with a Km value for H-type 2 R110!H substitution; lane 4, mutated FUT6 with R110!Q substitution; acceptor oligosaccharide close to that of the wild type. lane 5, mutated FUT6 with R110!N substitution. However, the mutated FUT5 enzyme with W124!R

Table IV. Kinetic parameters of human FUT3, FUT5, and some of their mutated variants

Donor substrate (GDP-fucose) Acceptor substrate

Km (mM) Vmax Km (mM) Vmax Enzyme (nmol/h/mg protein) (nmol/h/mg protein)

H-type 1 FUT3 35 21 0.2 52 FUT3 W111!Y 34 10 0.3 34 FUT3 W111!F 37 10 0.3 30 FUT3 W111!A 46 1 2.5 16 H-type 2 FUT5 38 1 1.1 38 FUT5 W124!Y 35 2 1.3 33 FUT5 W124!R 30 7 0.1 28

Kinetic constants were determined as described in Materials and methods. Incubation times were 1 h for wild-type and mutated FUT3 enzymes, 1.5 h for FUT5 and FUT5 W124!Y, and 30 min for FUT5 W124!R. 351 F. Dupuy et al.

Table V. Peptide alignment of the putative acceptor-binding domain of Caenorhabditis elegans and plant a3- and a4-FUTs

The invertebrate and plant sequences are aligned by reference to human Lewis FUT3 enzyme. The putative acceptor-binding motif of C. elegans enzyme was revealed by DIALIGN software, whereas the corresponding motif of plant enzyme was adjusted manually.

substitution had a 10-fold greater affinity for the H-type 2 substrate binding. For example, the V and the first H resi- substrate. Thus compared to wild-type enzymes, the affi- dues of human FUT4 a3-FUT are not directly involved in nities of FUT3 and FUT5 variants toward their best accep- type 2 substrate binding (Sherwood et al., 2002). This sug- tor substrates are in close correlation with their levels of gests flexibility in the amino acid composition of the motif. activity (Figure 2, Table II, and Table IV). Such flexibility is observed for Caenorhabditis elegans lactosamine a3-FUT (DeBose-Boyd et al., 1998) where the potential acceptor-binding motif is VLIAHMD (Table V). Conservation of hydrophobic amino acids underscores Discussion their crucial role, potentially in protein folding. Also, the Numerous species possess several distinct a3- and/or a3/4- presence in C. elegans enzyme of M at the strictly conserved FUTs, which transfer fucose with a1,3 and/or a1,4 linkage position of W/R residue could reflect the ability of this from the same nucleotide sugar donor, the GDP-fucose. enzyme to transfer fucose to a wider variety of lacto- or However, along with other differences, they differ in their neolacto-series structures (DeBose-Boyd et al., 1998), com- acceptor substrate specificities, which mainly depend on pared to vertebrate ones. For bacterial enzymes using type 1 amino acids localized in their amino terminal segment and type 2 acceptor substrates (Ge et al., 1997; Rasko et al., (Legault et al., 1995; Nguyen et al., 1998; Sherwood et al., 2000), no corresponding acceptor-binding motif is found. 2002; Xu et al., 1996). Peptide sequence alignment of these Ma et al. (2003) showed that amino acids involved in accep- domains is a useful way to define conserved residues that tor substrate specificity of Helicobacter pylori enzymes are potentially involved in acceptor specificities. This strat- are located in their carboxy-terminal domain. Moreover, egy was successfully used to define the VxxHH(W/R)(D/E) the bacterial a3/4-FUTs seem unrelated in evolution to the conserved acceptor-binding motif (where x corresponds to a vertebrate ones. a hydrophobic residue) in the NH2 domain of these enzymes In plants, a1,4 fucosylation and Le structures are pre- (Dupuy et al., 1999). We demonstrated that W111!R sub- sent. Several plant a4-FUTs have been cloned and charac- stitution conferred to human FUT3 the ability to use effi- terized (Bakker et al., 2001; Leonard et al., 2002; Wilson, ciently H-type 2 instead of H-type 1 acceptor substrate. This 2001), and their potential acceptor-binding motif identified change in activity was related to affinity modifications (VAYKKWD for A. thaliana, Table V). As in vertebrate toward acceptor substrates (Dupuy et al., 1999). Since a3/4-FUTs, the presence of W and D residues in the plant then, new FUTs that add fucose in a1,3 and/or in a1,4 motif correlates with an a1,4-FUT activity of these linkage on lacto- or neolacto-series structures have been enzymes. Conservative changes are observed for the hydro- isolated from Vertebrates. phobic residuesÐand the conserved H residues are substi- In the present work, we showed that all these vertebrate tuted by the other basic amino acids, R or K. However, this enzymes are characterized by the consensus VxxHH(W/R) motif is located closer to the amino-terminal end compared (D/E) acceptor-binding motif and their a1,3 or a1,3/4 to animal enzymes. The hypothesis of a potential soluble activities are correlated with the presence of R or W, respec- enzyme devoid of transmembrane domain has been pro- tively. Although required for optimal activity, amino posed and could account for this peculiar position in the acids in this motif are not equally involved in the acceptor sequence (LeÂonard et al., 2002). Although the exact role of 352 Acceptor substrate specificity of Lewis FUTs the acceptor-binding motif in activity and substrate-binding residue.Conversely,thepresenceofWresidueintheacceptor- in invertebrate and plant enzymes remains to be established, binding domain of a3/4-FUTs could allow the recogni- its involvement in acceptor substrate recognition by tion of type 1 and type 2 substrates by the stacking of the a3-, a3/4- and a4-FUTs is supported by its lack in core acceptor GlcNAc ring with the cycle of W residue. The a3-FUTs, which use different acceptor substrates (Bakker particular affinities of FUT5 enzymes toward type 1/type 2 et al., 2001; Fabini et al., 2001; Leiter et al., 1999). substrates could be explained by a crucial interaction In this study, we demonstrate that the ability of FUT3 between a residue in FUT5 enzymes, which remains to be enzymestousetype1substrates,andconsequentlytheircapa- identified, and perhaps a reactive group of the acceptor city to transfer fucose with a1,4 linkage, involves the aro- substrate. matic character of the W residue in their acceptor-binding Genes encoding FUTs with a1,4 activity arose by dupli- domain. Indeed, only enzymes with an aromatic residue (W, cations of an a3-FUT ancestor gene (for review, see Javaud Y, or F) at the candidate position had significant a1,4 acti- et al., 2003). Among the different events that conduced to vity and high affinities for GDP-Fuc donor and H-type 1 the appearance of a3/4-FUTs, the switch of the basic R to acceptor substrates. Alternatively, we showed that the opti- the aromatic W residue in the acceptor-binding motif was mal a1,3 activity of Lewis FUTs could depend on the pre- certainly a crucial point. The presence of W residue in a3/4- sence of R residue in the acceptor-binding motif. FUT3 FUTs could be explained by a single point mutation chang- improved its a1,3 activity only when W111 was substituted ing R codon in W codon (CGG!TGG). The new data by R (Figure 2). In the same way, the only increase in a1,3 presented in this study gave support to our model of activity and in affinity for H-type 2 acceptor of FUT5 was Lewis gene evolution in Primates (Dupuy et al., 2002), observed when W124 was replaced by R (Tables II and IV). which proposed that a1,4 fucosylation appeared first in Moreover, a1,3 activities of FUT6 and FUTb decreased FUT3 and latterly in FUT5 independently. The differences whentheirRresiduewassubstitutedbybasicorchargedones. between wild-type and mutated FUT3 and FUT5 enzymes The correlation between acceptor substrate specificity concerning their proper a1,3/4 activities could then be and amino acid composition of the acceptor-binding motif explained by an independent process conferring the a1,4 is less obvious for FUT5 enzymes. They show W in their function. Sequence comparison of a broad panel of Lewis acceptor-binding motif but preferentially use H-type 2 tri- enzymes is now needed to point out the amino acids, other saccharide substrates (Costache et al., 1997b). The mutated than the W in the acceptor-binding motif, conferring the FUT5withanotheraromaticresidue(ForY)atthecandidate original a1,4 fucosylation in the primate lineage. position had a lower a1,3 activity compared with the wild- type enzyme, whereas other substitutions (W124!A/V) allow the conservation of the original a1,3 activity. It is Materials and methods worth noticing that, as for W111!Y/F substitutions in FUT3, the W124!Y substitution in FUT5 gave an enzyme Cloning of simian Lewis-like coding sequences with around 50% of its wild-type a1,4 activity. We con- The gorilla (Gorilla gorilla), orangutan (Pongo pygmaeus), cluded that the type 1/type 2 substrate recognition by FUT5 and gibbon (Hylobates lar) Lewis-like coding sequences enzymes involves amino acids of the acceptor-binding motif were obtained by a one-step polymerase chain reaction and one or several other residues of the catalytic domain. strategy (Dupuy et al., 2002) on genomic DNA. This DNA This hypothesis could also explain the lack of FUT6 a1,3/4 was isolated from blood samples obtained at the Centre de 110 activities, when R was substituted by W. Primatologie(UniversiteLouisPasteur,Strasbourg,France). The type 1 and 2 disaccharides have different molecular The monoexonic open reading frames were amplified using topographies. The minimum energy conformations of type 1 sense 50-CTCTCTTCCCAGCTACTCTGACCCATG-30 and 2 structures invert the relative orientation of Gal and and antisense 50-GCCCAGCCCCATGCCGGCCTCTC-30 GlcNAc residues of disaccharides by approximately 180 . primers. They were inserted into the pcDNA3.1/Amp mam- The key polar groups of lacto- and neolacto-series acceptor malian expression vector (Invitrogen, Carlsbad, CA) and substrates are hydroxyl groups at C-3 or C-4 of GlcNAc sequenced using the dideoxy chain termination method and and C-6 of Gal (De Vries et al., 1997; Du and Hindsgaul, a dye labeling chemistry (kit PRISM Tml Ready Reaction 1996; Gosselin and Palcic, 1996). Because C-6 of Gal is Ampli Taq FS, USA), associated with the ABI Prism 310 necessary for activity whatever the acceptor substrate is, a Genetic Analyser (Perkin Elmer, Norwalk, CT). Nucleotide correctly positioned Gal residue appears to be essential for sequence data are available in GenBank under the accession enzyme activity. The active site of Lewis FUTs could con- numbers AF515436 and AF515437 for G. gorilla FUT5 and secutively form a pocket, which preferentially binds the FUT6 genes, AF515438 and AF515439 for P. pygmaeus GlcNAc residue of type 1 or type 2 acceptor substrates. It FUT3 and FUT5 genes, and AF515440 for H. lar FUT5 has been proposed that recognition of carbohydrate accep- gene. tors by glycosyltransferases occurred through hydrogen bonds and stacking of the sugar ring with aromatic amino acids (Hindsgaul et al., 1991; Matsui et al., 1994; Vyas et al., Peptide alignment of a3-, a3/4-, and a4-FUTs 1991). Although W/R residue could also be involved in the Alignments were performed with DIALIGN (http://bioweb. protein folding, we hypothesize that the binding of type 2 pasteur.fr/seqanal/interfaces/dialign2.html) and refined acceptor substrates by a3-FUTs would partially rely on a with further manual adjustments using the ED program hydrogen bond between the R residue of their acceptor- (Philippe, 1993) of the MUST 2000 package (http:// binding domain and a key polar group of the GlcNAc sorex.snv.jussieu.fr/must2000.html). 353 F. Dupuy et al.

Site-directed mutagenesis trisaccharide acceptor substrates were purchased from Syn- The different FUT coding sequences were inserted into the tesome (Munich, Germany). FUT assays were performed in mammalian expression vector pcDNA1/Amp (Invitrogen) 60 ml volume containing 0.1 mM acceptor substrate, 25 mM and directly used for polymerase chain reaction±based sodium cacodylate (pH 6.5), 5 mM ATP, 20 mM MnCl2, 10 14 mutagenesis. All mutations were performed with the Quick- mM a-L-fucose, 3 mM GDP-[ C]-fucose (310 mCi/mmol; Change Kit (Stratagene, La Jolla, CA). The primers used Amersham Pharmacia Biotech, Little Chalfont, U.K.) and for mutagenesis are summarized in the Table VI. The open 20 mg protein extracts from transfected COS-7 cells. Activ- reading frames containing mutations were completely ities were measured as previously described (Dupuy et al., checked as described. 1999): the reaction was stopped by addition of 3 ml cold water, and the reaction mixture was then applied to a con- Expression of wild-type and mutated enzymes ditioned Sep-Pack C18 reverse chromatography cartridge (Waters Millipore, Bedford MA). Unreacted GDP-[14C]- High pure recombinant plasmids (plasmid midi Kit, fucose was washed off with water. The radiolabeled reaction Qiagen, Hilden, Germany) were used to transiently product was eluted with ethanol and counted with Biode- transfect COS-7 cells during 48 h (SuperFect Transfection gradable Counting Scintillant (Amersham Pharmacia Reagent, Qiagen). These cells lack endogenous lactosamine Biotech) in a liquid scintillation beta counter (Liquid scin- a1,3- and a1,4-FUT activities (Clarke and Watkins, 1999). tillation analyzer, Tri-Carb-2100TR, Packard, USA). COS-7 proteins were extracted in 1% (v/v) Triton-X 100, 10 mM sodium cacodylate (pH 6), 20% (v/v) glycerol, 1 mM Western blot analysis dithiothreitol under shacking for 2 h at 4 C. Cell debris Thirty micrograms of soluble proteins from COS-7 cells was eliminated by centrifugation, and the protein super- were boiled for 3 min after the addition of b-mercapto- natant content was determined (Bradford, BioRad, ethanol (5% v/v) and bromophenol blue (0.02% w/v). Hercules, CA). Sodium dodecyl sulfate±polyacrylamide gel electrophoresis was carried out on Tris/Tricine-10.5% sodium dodecyl sul- FUT assays fate±polyacrylamide gel. Separated proteins were electro- Blood group H-type 1 (Fuca1,2Galb1,3GlcNAc-sp-biotin) transferred onto a nitrocellulose membrane (Schleicher & and H-type 2 (Fuca1,2Galb1,4GlcNAc-sp-biotin) Schuell, Dassel, Germany). The immunoblots were

Table VI. Primers used in polymerase chain reaction for site-directed mutagenesis

Primer bases in bold designate substituted nucleotides. Degenerated bases are annotated according to IUB abbreviations. Positions of primers are given with the adenine residue of the initiation codon of each coding sequence assigned as base 1. s: sense primer; as: antisense primer. 354 Acceptor substrate specificity of Lewis FUTs

processed by chemiluminescence detection (Chemilumines- De Vries, T., Palcic, M.P., Schoenmakers, P.S., Van Den Eijnden, D.H., cence Blotting Substrate [POD], Roche Molecular Bio- and Joziasse, D.H. (1997) Acceptor specificity of GDP-Fuc:Gal chemicals, Mannheim, Germany). Anti-human FUT3 and b1,4GlcNAc-R a3-fucosyltransferase VI (FucT VI) expressed in insect anti-ox FUTb antibodies were produced in rabbit using cells as soluble, secreted enzyme. Glycobiology, 7, 921±927. 45 361 35 De Vries, T., Knegtel, R.M., Holmes, E.H., and Macher, B.A. (2001) truncated proteins of FUT3 (P ±T ) and FUTb (R ± Fucosyltransferases: structure/function studies. Glycobiology, 11, Q365), which were obtained in Escherichia coli BL21 119R±128R. (DE3). The blot was first incubated with rabbit anti- Du, M. and Hindsgaul, O. (1996) Recognition of b-D-Gal p-(1,3)-b-D-Glc human FUT3 antibodies (4 mg/ml) for recombinant pNAc-OR acceptor analogues by the Lewis a-(1,3/4)-fucosyltransfer- FUT3, FUT5, and FUT6 enzymes or rabbit anti-ox ase from human milk. Carbohydr. Res., 286, 87±105. Dupuy, F., Petit, J.M., Mollicone, R., Oriol, R., Julien, R., and Maftah, A. FUTb antibodies (2 mg/ml) for recombinant FUTb, then (1999) A single amino acid in the hypervariable stem domain of with the secondary antibody, a pig anti-rabbit IgG conju- vertebrate a1,3/1,4-fucosyltransferases determines the type 1/type 2 gated to horseradish peroxidase (dilution 1:1000) (Dako, transfer. Characterization of acceptor substrate specificity of the lewis Denmark). enzyme by site-directed mutagenesis. J. Biol. Chem., 274, 12257±12262. Dupuy, F., Germot, A., Marenda, M., Oriol, R., Blancher, A., Julien, R., and Maftah, A. (2002) a1,4-Fucosyltransferase activity: a significant Kinetic constant determination function in the primate lineage has appeared twice independently. Mol. The apparent K values for GDP-fucose were determined Biol. Evol., 19, 815±824. m Fabini, G., Freilinger, A., Altmann, F., and Wilson, I.B. (2001) using 10±200 mM GDP-fucose including 8 mM GDP- Identification of core a3-fucosylated glycans and cloning of the [14C]fucose in each reaction, 20±40 mg proteins, and 2 mM requisite fucosyltransferase cDNA from Drosophila melanogaster. H-type 1 or H-type 2 acceptors. The incubation time Potential basis of the neural anti-horseadish peroxidase epitope. J. Biol. Chem., 276, 28058±28067. was between 30 and 90 min. The apparent Km values for H-type 1 and H-type 2 acceptors were determined with Fitchette, A.C., Cabanes-Macheteau, M., Marvin, L., Martin, B., Satiat- Jeunemaitre, B., Gomord, V., Crooks, K., Lerouge, P., Faye, L., and 0.05±3 mM acceptor and 200 mM GDP-fucose, including Hawes, C. (1999) Biosynthesis and immunolocalization of Lewis a- 8 mM GDP[14C]-fucose and incubation times of 30±90 min. containing N-glycans in the plant cell. Plant Physiol., 121, 333±344. Ge, Z., Palcic, M.M., and Taylor, D.E. (1997) Cloning and heterologous expression of an a1,3-fucosyltransferase gene from the gastric Acknowledgments pathogen Helicobacter pylori. J. Biol. Chem., 272, 21357±21363. Gosselin, S. and Palcic, M.M. (1996) Acceptor hydroxyl group mapping This work was done within the framework of the G3 French for human milk a1-3 and a1-3/4 fucosyltransferases. Bioorg. Med. network and was partially supported by the Conseil Chem., 4, 2023±2028. Regional du Limousin. Hindsgaul, O., Kaur, K.J., Srivastava, G., Blaszczyk-Thurin, M., Crawley, S.C., Heerze, L.D., and Palcic, M.M. (1991) Evaluation of deoxygenated oligosaccharide acceptor analogs as specific inhibitors of glycosyltransferases. J. Biol. Chem., 266, 17858±17862. References Holmes, E.H., Xu, Z., Sherwood, A.L., and Macher, B.A. (1995) Structure-function analysis of human a1,3fucosyltransferases. A Bakker, H., Schijlen, E., de Vries, T., Schiphorst, W.E., Jordi, W., GDP-fucose-protected, N-ethylmaleimide-sensitive site in FucT-III Lommen, A., Bosch, D., and van Die, I. (2001) Plant members of the a1, and FucT-V corresponds to Ser178 in FucT-IV. J. Biol. Chem., 270, 3/4-fucosyltransferase gene family encode an a1,4-fucosyltransferase, 8145±8151. potentially involved in Lewis(a) biosynthesis, and two core a1, Holmes, E.H., Yen, T.Y., Thomas, S., Joshi, R., Nguyen, A., Long, T., 3-fucosyltransferases. FEBS Lett., 507, 307±312. Gallet, F., Maftah, A., Julien, R., and Macher, B.A. (2000) Human Becker, D.J. and Lowe, J.B. (2003) Fucose biosynthesis and biological a1,3/4 fucosyltransferases. Characterization of highly conserved function in mammals. Glycobiology, 13, 41±53. cysteine residues and N-linked glycosylation sites. J. Biol. Chem., Cameron, H.S., Szczepaniak, D., and Weston, B.W. (1995) Expression of 275, 24237±24245. human chromosome 19p a(1,3)-fucosyltransferase genes in normal Javaud, C., Dupuy, F., Maftah, A., Julien, R., and Petit, J.M. (2003) The tissues. Alternative splicing, polyadenylation, and isoforms. J. Biol. fucosyltransferase gene family: an amazing summary of the underlying Chem., 270, 20112±20122. mechanisms of gene evolution. Genetica, 118, 157±170. Christensen, L.L., Jensen, U.B., Bross, P., and Orntoft, T.F. (2000) The Kaneko, M., Kudo, T., Iwasaki, H., Ikehara, Y., Nishihara, S., C-terminal N-glycosylation sites of the human a1,3/4-fucosyltransfer- Nakagawa, S., Sasaki, K., Shiina, T., Inoko, H., Saitou, N., and ase III, -V, and -VI (hFucTIII, -V, and -VI) are necessary for the Narimatsu, H. (1999) a1,3-Fucosyltransferase IX (Fuc-TIX) is very expression of full enzyme activity. Glycobiology, 10, 931±939. highly conserved between human and mouse; molecular cloning, Clarke J.L. and Watkins W.M. (1999) Expression of human a-l- characterization and tissue distribution of human Fuc-TIX. FEBS fucosyltransferase gene homologs in monkey kidney COS cells and Lett., 452, 237±242. modification of potential fucosyltransferase acceptor substrates by an Leonard, R., Costa., G, Darrambide, E., Lhernould, S., Fleurat-Lessard, endogenous glycosidase. Glycobiology, 9, 191±202. P., Carlue, M., Gomord, V., Faye, L., and Maftah, A. (2002) The Costa, G., Lhernould, S., Leonard, R., Dupuy, F., Gastinel, L., and presence of Lewis a epitopes in Arabidopsis thaliana glycoconjugates Maftah A. (2002) Recent progress in plant fucosylation. Curr. Topics depends on an active a4-fucosyltransferase gene. Glycobiology, 12, Plant Biol., 3, 101±111. 299±306. Costache, M., Cailleau, A., Fernandez-Mateos, P., Oriol, R., and Legault, D.J., Kelly, R.J., Natsuka, Y., and Lowe, J.B. (1995) Human Mollicone, R. (1997a) Advances in molecular genetics of a2- and a(1,3/1,4)-fucosyltransferases discriminate between different oligosac- a3/4-fucosyltransferases. Trans. Clin. Biol., 4, 367±382. charide acceptor substrates through a discrete peptide fragment. J. Biol. Costache, M., Apoil, P., Cailleau, A., Elmgren, A., Larson, G., Henry, S., Chem., 270, 20987±20996. Blancher, A., Iordachescu, D., Oriol, R., and Mollicone, R. (1997b) Leiter, H., Mucha, J., Staudacher, E., Grimm, R., Glossl, J., and Evolution of fucosyltransferase genes in vertebrates. J. Biol. Chem., Altmann, F. (1999) Purification, cDNA cloning, and expression of 272, 29721±29728. GDP-L-Fuc:Asn-linked GlcNAc a1,3-fucosyltransferase from mung DeBose-Boyd, R.A., Nyame, A.K., and Cummings, R.D. (1998) beans. J. Biol. Chem., 274, 21830±21839. Molecular cloning and characterization of an a1,3 fucosyltransferase, Ma, B., Wang, G., Palcic, M.M., Hazes, B., and Taylor, D.E. (2003) C- CEFT-1, from Caenorhabditis elegans. Glycobiology, 8, 905±917. terminal amino acids of Helicobacter pylori a1,3/4 fucosyltransferases

355 F. Dupuy et al.

determine type I and type II transfer. J. Biol. Chem., 278, 21893± A Lys/Arg residue located within the a1,3-FucT motif is required 21900. for activity but not substrate binding. J. Biol. Chem., 273, 25256± Matsui, I., Yoneda, S., Ishikawa, K., Miyairi, S., Fukui, S., Umeyama, H., 25260. and Honda, K. (1994) Roles of the aromatic residues conserved in the Sherwood, A.L., Davis, W.C., Ho, S., Macher, B.A., Stroud, M.R., active center of Saccharomycopsis a-amylase for transglycosylation Upchurch, D.A., and Holmes, E.H. (2000) A GDP-fucose-protected, and hydrolysis activity. Biochemistry, 33, 451±458. pyridoxal-50-phosphate/NaBH(4)-sensitive lys residue common to Mollicone, R., Cailleau, A., and Oriol, R. (1995) Molecular genetics of H, human a1,3fucosyltransferases corresponds to Lys(300) in FucT-IV. Se, Lewis and other fucosyltransferase genes. Transfus. Clin. Biol., 2, Biochem. Biophys. Res. Commun., 273, 870±876. 235±242. Sherwood, A.L., Upchurch, D.A., Stroud, M.R., Davis, W.C., and Nguyen, A.T., Holmes, E.H., Whitaker, J.M., Ho, S., Shetterly, S., and Holmes, E.H. (2002) A highly conserved His-His motif present in Macher, B.A. (1998) Human a1,3/4-fucosyltransferases. I. Identifica- a,3/4fucosyltransferases is required for optimal activity and functions tion of amino acids involved in acceptor substrate binding by site- in acceptor binding. Glycobiology, 12, 599±606. directed mutagenesis. J. Biol. Chem., 273, 25244±25249. Sousa, V.L., Brito, C., Costa, T., Lanoix, J., Nilsson, T., and Costa, J. Oriol, R. (1995) ABO, Hh, Lewis and secretion. Serology, genetics and (2003) Importance of Cys, Gln, and Tyr from the transmembrane tissue distribution. In Cartron, J.P. and Rouger, P. (eds.), Molecular domain of human a3/4 fucosyltransferase III for its localization and basis of major blood group antigens. Plenum Press, London, pp. 37±73. sorting in the Golgi of baby hamster kidney cells. J. Biol. Chem., 278, Oriol, R., Mollicone, R., Cailleau, A., Balanzino, L., and Breton, C. 7624±7629. (1999) Divergent evolution of fucosyltransferase genes from verte- Staudacher, E., Altmann, F., Wilson, I.B., and Marz, L. (1999) Fucose in brates, invertebrates, and bacteria. Glycobiology, 9, 323±334. N-glycans: from plant to man. Biochim. Biophys. Acta, 1473, 216±236. Oulmouden, A., Wierinckx, A., Petit, J.M., Costache, M., Palcic, M.M., Vo, L., Lee, S., Marcinko, M.C., Holmes, E.H., and Macher, B.A. (1998) Mollicone, R., Oriol, R., and Julien, R. (1997) Molecular cloning and Human a1,3/4-fucosyltransferases. A single amino acid at the COOH expression of a bovine a(1,3) fucosyltransferase gene homologous to a terminus of FucT III and V alters their kinetic properties. J. Biol. putative ancestor gene of the human FUT3-FUT5-FUT6 cluster. Chem., 273, 25250±25255. J. Biol. Chem., 272, 8764±8773. Vyas, N.K., Vyas, M.N., and Quiocho, F.A. (1991) Comparison of the Philippe, H. (1993) MUST, a computer package of management utilities periplasmic receptors for L-arabinose, D-glucose/D-galactose, and D- for sequences and trees. Nucleic Acids Res., 21, 5264±5272. ribose.Structuralandfunctionalsimilarity.J.Biol.Chem.,266,5226±5237. Rasko, D.A., Wang, G., Palcic, M.M., and Taylor, D.E. (2000) Cloning Wilson, I.B. (2001) Identification of a cDNA encoding a plant Lewis-type and characterization of the a(1,3/4) fucosyltransferase of Helicobacter a1,4-fucosyltransferase. Glycoconj. J., 18, 439±447. pylori. J. Biol. Chem., 275, 4988±4994. Xu, Z., Vo, L., and Macher, B.A. (1996) Structure-function analysis of Sherwood, A.L., Nguyen, A.T., Whitaker, J.M., Macher, B.A., Stroud, human a1,3-fucosyltransferase. Amino acids involved in acceptor M.R., and Holmes, E.H. (1998) Human a1,3/4-fucosyltransferases. III. substrate specificity. J. Biol. Chem., 271, 8818±8823.

356