GAG assembly initiates by the formation of the tetrasaccharide, GlcAβ1,3Galβ1,3Galβ1,4Xylβ, attached to the side-chain hydroxyl group of specific serine residues of core proteins of proteoglycans. This oligosaccharide serves as a primer for chain elongation by the alternate addition of N-acetylglucosamine (GlcNAc) or N-acetylgalactosamine (GalNAc) and GlcA residues to form either heparin/heparan-sulfates or chondroitin/dermatan-sulfates GAG chains, respectively. GlcAT-I catalyzes the final step of the formation of the linkage tetrasaccharide primer, which is essential for the conversion of a core protein into a functional proteoglycan, thus regulating the overall process of proteoglycan synthesis (Kitagawa et al. 1998). Two related β1,3-glucuronosyltransferase enzymes, referred to as GlcAT-P and GlcAT-S, have been identified, which share a high degree of sequence similarity with GlcAT-I and belong to the same gene family (Terayama et al. 1997; Marcos et al. 2002). GlcAT-P and GlcAT-S are required for the addition of GlcA onto the terminal Galβ1,4GlcNAc sequence of glycoproteins and glycolipids, respectively, producing the HNK1 (for Human Natural Killer) epitope precursor sequence GlcAβ1,3Galβ1,4GlcNAc. This trisaccharide sequence then serves as the acceptor site for the 3-O-sulfotransferase that subsequently generates the HNK1 epitope primarily expressed in brain. This carbohydrate epitope that is recognized by HNK1 monoclonal antibodies is found on many neural cell adhesion molecules such as NCAM, myelin-associated glycoprotein (McGarry et al. 1983), and glycolipids (Chou et al. 1986) and is presumed to be involved in cell–cell interactions such as adhesion, migration, and neurite extension.

The presence of GAGs has been demonstrated in the invertebrate organisms Drosophila melanogaster and Caenorhabditis elegans (Toyoda et al. 2000). Furthermore, the existence of a conventional linkage region tetrasaccharide sequence was recently established for these invertebrate GAG chains, suggesting that their fundamental structures and biosynthetic mechanisms are similar to the mammalian GAG chains. Recently, three related β1,3-glucuronosyltransferases have been cloned in D. melanogaster and designated DmGlcAT-I, DmGlcAT-BSI, and DmGlcAT-BSII (where BS stands for broad specificity; Kim et al. 2003). An ortholog gene of GlcAT-I (sqv8) has also been cloned from C. elegans and its defects caused morphological abnormality such as squashed vulva (Bulik et al. 2000).

Among the β1,3-glucuronosyltransferases, human GlcAT-I was the first cloned and has since been extensively studied in our laboratory and others (Kitagawa et al. 1998; Ouzzine et al. 2000a) due to its critical location in the biosynthetic pathway of GAGs and its potential as a pharmacological target (Venkatesan et al. 2004). Biochemical and structural analyses indicated that GlcAT-I is organized as a dimer, each subunit with a Rossman-like fold divided into two regions connected by the so-called DXD motif (D195–D196–D197 in GlcAT-I) (Ouzzine et al. 2000b; Pedersen et al. 2000). The N-terminal region (residues 26–74) comprises the UDP-sugar binding region and is terminated by the DDD sequence involved in the coordination of Mn²⁺ divalent cations essential for GlcAT-I activity (Gulberti et al. 2003). The C-terminal region (75–335) includes the acceptor substrate binding site and is terminated by a C-terminal domain extending to the other molecule in the dimer, that is thought to be important for substrate recognition (Gulberti et al. 2005).

The aim of this study was to identify crucial residues involved in UDP-α-D-glucuronic acid (UDP-GlcA) recognition and β1,3-glucuronosyltransferase activity. A previous study emphasized the key role of H308 in governing the specificity of GlcAT-I toward the nucleotide–sugar (Ouzzine et al. 2002). In order to better understand the recognition process of the donor substrate, we develop here a phylogenetic approach, that allowed us to identify 119 related β1,3-glucuronosyltransferase sequences in vertebrates, invertebrates, and plants. Multiple sequence alignments revealed conserved peptide motifs and amino acids, stimulating the evaluation of the function of these potentially important residues. Systematic site-directed mutagenesis of these residues in the human GlcAT-I led us to delineate their respective importance in UDP-GlcA binding and in β1,3-glucuronosyltransferase activity.

Figure 1. Neighbor-joining phylogenetic tree of β1,3-glucuronosyltransferases. Forty sequences were selected out from the total 119 different β1,3-glucuronosyltransferase-like sequences based on the following criteria: (1) They were the more representative (more ...)

Figure 2. ClustalW sequence alignment of the donor substrate-related conserved peptide motifs 1–4 and 8 of the 40 sequences analyzed in Figure 1. The relative positions of the eight conserved peptide motifs are illustrated in the first line. The positions (more ...)

Analysis of the three-dimensional structure of GlcAT-I–UDP-GlcA complex (Fig. 3) suggests that the position of uracil group is fixed through a ring-stacking interaction with Y84 and a hydrogen bond with D113 (Pedersen et al. 2000, 2002). Both conservative and nonconservative mutations of these residues were carried out and the consequences were evaluated. Upon expression in P. pastoris, immunoblot analysis indicated that the Y84F mutant was produced at higher level than the wild-type protein, while the Y84A mutant was slightly less expressed (Fig. 4A). The activity of the recombinant enzymes was normalized to the amount of expressed GlcAT-I protein and values are presented in Figure 4B. Interestingly, substitution of Y84 to alanine completely abrogated the enzyme activity, whereas the conservative mutation of Y84 to a phenylalanine residue restored enzyme activity up to 60% of the wild-type enzyme (Fig. 4B). Determination of the kinetic parameters of this mutant indicated a moderate increase in K_m value toward UDP-GlcA (threefold; Table 1). The observation that the conserved Y84 residue has diverged to histidine in C. elegans β-1,3-glucuronosyltransferase (Fig. 2) led us to determine the functional consequences of a similar substitution in GlcAT-I. Interestingly, the Y84H mutant has retained significant activity (21%, Fig. 4B) with a K_m value double that of the wild type (Table 1). UDP binding experiments showed that, as expected, the wild-type protein exhibited efficient binding to UDP-beads, which was prevented by addition of UDP in the incubation mixture (Fig. 5A), whereas Y84A exhibited impaired binding to UDP-beads, resulting in the presence of ~50% of the protein in the supernatant. In contrast, Y84F mutant was able to bind efficiently to UDP-beads as the wild-type protein (Fig. 5B). Altogether, these results indicated that the tyrosine residue at position 84 plays a crucial role in GlcAT-I activity and UDP-GlcA binding. This function appeared to be mainly attributable to the phenyl ring, likely via stacking interactions with the nucleotide base, since a phenylalanine or a histidine residue were able to replace, to some extent, the wild-type tyrosine.

Figure 3. Stereo view of the human GlcAT-I donor substrate-binding site. The structure of GlcAT-I is shown in complex with UDP-GlcA (PDB code 1FGG). Conserved amino acid residues predicted to interact with the uridine base (D113 and Y84), with the pyrophosphate (more ...)

Figure 4. Heterologous expression and activity of wild-type human GlcAT-I and mutants expressed in yeast P. pastoris. (A) SDS-PAGE and immunoblot analysis of membrane fraction of recombinant yeast cells expressing wild-type (WT) and GlcAT-I mutants; 2 μg (more ...)

Table 1. Kinetic analysis of wild-type GlcAT-I and mutants toward UDP-GlcA

Figure 5. UDP binding of wild-type human GlcAT-I and mutants expressed in yeast P. pastoris. UDP-beads were incubated in the presence of solubilized membrane proteins from yeast cells expressing wild-type GlcAT-I and mutants. Proteins bound to UDP-beads (B) and (more ...)

Next, we analyzed the role of D113 predicted to interact with N-3 of uracil. Upon expression in P. pastoris, the D113A mutant was produced in a slightly lower amount compared with the wild-type enzyme whereas D113E and D113N mutants were produced in higher or similar amounts, respectively (Fig. 4A). We found that substitution of D113 by alanine and by asparagine strongly decreased the enzyme activity (to 11% and 15% of the wild-type GlcAT-I, respectively; Fig. 4B). The K_m value toward UDP-GlcA for the D113N mutant was increased 13-fold, suggesting that the amide group of asparagine is much less favorable to hydrogen-bond formation with N-3 of uracil compared with the carboxyl group of aspartate. On the other hand, mutation of D113 by glutamic acid affected the enzyme activity to a lesser extent when compared with the D113A or D113N mutants (reduced to 27% of the wild-type enzyme). Determination of the kinetic parameters of the D113E mutant showed a threefold increase in K_m value toward UDP-GlcA. These results suggest that the nucleotide-binding site of GlcAT-I could accomodate a longer carboxyl side chain at this position, although this substitution results in a lower catalytic efficiency of the enzyme. UDP binding experiments showed that the three aspartate-substituted mutants exhibited reduced capacity to bind to the beads, consistent with the increased K_m values of these mutants (Fig. 5B). These results indicate that mutations of D113 impaired, but did not abolish, the UDP binding capacity of GlcAT-I to UDP-beads. These observations are consistent with the important contribution of Y84 residue together with D113 in the positioning of the base of the nucleotide–sugar.

We next examined the role of the basic residues R156 and R310, which are predicted to interact with the pyrophosphate group of UDP (Fig. 3). Atom NH1 of R156 is in hydrogen-bond distance from O2B of the α-phosphate, whereas atom NH1 of R310 is predicted to interact with O1B of the β-phosphate. The role of these residues was investigated by expressing mutants with nonconservative (R to A) and conservative (R to K) substitutions. R156A and R156K mutants were expressed at similar amounts as the wild-type protein (Fig. 4A). These mutations strongly reduced GlcAT-I activity by 90% and 95%, respectively (Fig. 4B), precluding determination of kinetic parameters. However, binding studies revealed that although the mutations were detrimental to the enzyme activity, the mutants R156A and R156K have retained most of their capacity to bind to UDP-beads (Fig. 5C).

On the other hand, mutations of R310 were less deleterious than that of R156 in terms of enzyme activity (Fig. 4B). Substitution of R310 to alanine or lysine reduced the enzyme activity by 70%. K_m value of R310A toward UDP-GlcA was increased about threefold compared with wild-type, whereas that of the R310K mutant was not significantly affected (Table 1). We also substituted R310 by a glutamine residue present at this position in urochordate and arthropod β1,3-glucuronosyltransferase sequences (Fig. 2). Interestingly, this mutation produced about the same effect as the arginine to alanine or lysine substitution (Fig. 4B). Furthermore, the K_m value for the R310Q mutant was unchanged, similarly to the R310K mutant. Binding studies indicated that mutations at position 310 impaired, but did not abrogate, binding of GlcAT-I to UDP-beads (Fig. 5B). Altogether, these data provide evidence for distinguishable roles of the two active site arginine R156 and R310 residues. We show that mutation of R156 produced deleterious effects on GlcAT-I activity, which may not primarily result from impaired interactions of this residue with the UDP part of the nucleotide–sugar, whereas substitutions of R310 affected, to a lesser extent, either binding of the donor substrate or enzyme activity.

Finally, we determined the impact of mutation of R161, which was invariant in all species considered in this study. The R161A and R161K mutants were expressed in higher and slightly lower levels than the wild-type protein, respectively (Fig. 4A). Substitution of the positively charged arginine residue by the neutral alanine amino acid completely inactivated GlcAT-I, precluding determination of kinetic parameters for this mutant. Replacement of R161 by the positively charged lysine residue severely compromised GlcAT-I activity (reduced to 5% of wild type) together with a 20-fold increase in K_m, consistent with the strong conservation of this amino acid (Fig. 4B; Table 1). On the other hand, none of the mutations modified binding to UDP-beads (Fig. 5D). This was in agreement with the predicted role of the guanidino group of R161 in hydrogen-bonding interactions with the 3-OH group of glucuronic acid and not with the UDP part of the donor substrate. Altogether, these data emphasized the critical role of R161 in the function of GlcAT-I, consistent with the strong conservation of this amino acid residue (Fig. 2).

The implication of GAGs in a wide range of biological processes has led to a considerable interest in the understanding of the glycosyltransferases involved in their biosynthetic pathway. GlcAT-I, which catalyzes a key step of GAG synthesis, belongs to the β1,3-glucuronosyltransferase family. The aim of this study was to acquire a better knowledge of the organization of these glycosyltransferases by phylogenetic analyses. The sequence similarities observed between the different members of this family suggest that they are dictated by the need for UDP-GlcA recognition, an assumption supported by structural analyses. In the present study, we were able to define the importance of conserved amino acids and to delineate their respective contribution in the recognition of the different parts of the donor substrate by conducting systematic mutagenesis analyses, affinity binding experiments, and kinetic studies.

Phylogeny suggests that the GTA–β1,3-glucuronosyltransferases (for review, see Unligil and Rini 2000) are rather old enzymes that have been well conserved through evolution, since β1,3-glucuronosyltransferase-like sequences were found in all the eukaryotes tested from vertebrates to invertebrates and plants, with the exception of the most primitive eukaryotes (fungi, protozoa, and algae). However, the two gene duplication events at the origin of the three vertebrate subfamilies (GlcAT-I, GlcAT-P, and GlcAT-S) are relatively recent and have probably occurred early in the vertebrate lineage, since only vertebrates express these three subfamilies. Recent studies have demonstrated that GAGs of the chondroitin- and heparan-sulfate types are present in D. melanogaster and in C. elegans and bear structural features common to vertebrate polysaccharides (Toyoda et al. 2000). A β1,3-glucuronosyltransferase from D. melanogaster referred to as DmGlcAT-I has been shown to present a restricted specificity toward Galβ1,3Galβ-O- derivatives, similarly to its vertebrate GlcAT-I counterpart (Kim et al. 2003). Two other drosophila β1,3-glucuronosyltransferases exhibited a broader specificity, and their role in the synthesis of glycosaminoglycans, glycolipids, and glycoproteins has been proposed, although the function of these enzymes was less clearly defined than that of DmGlcAT-I (Kim et al. 2003). Our phylogenetic analysis and functional studies suggest that the implication of the β1,3-glucuronosyltransferases in the completion of glycosaminoglycan linkage region and in the synthesis of the sulfoglucuronyl HNK1 epitope represents an example of convergent evolution, since similar acceptor substrate specificities have evolved in a parallel manner in vertebrates and invertebrates. However, the separation in distinct subfamilies of invertebrates is less apparent than in the vertebrates. In the nematode C. elegans, functional studies of the glycosyltransferase encoded by sqv8 have provided convincing evidence for the involvement of this enzyme in the synthesis of the GAG–protein linkage region (Bulik et al. 2000). In addition, our phylogenetic analysis predicts the presence of other related β1,3-glucuronosyltransferases in this invertebrate organism. It may be hypothesized that these enzymes are involved in the synthesis of HNK1 epitopes as these glycoepitopes have also been described in C. elegans (Haltiwanger and Lowe 2004). However, this hypothesis awaits further investigation.

Furthermore, the present phylogenetic analysis led us to include several putative plant β1,3-glucuronosyltransferases in the same gene family. Although a large number of genes encoding UDP-glycosyltransferases have been recently identified in plant genomes, to our knowledge none has been ascribed to this β1,3-glucuronosyltransferase family (Ross et al. 2001). Interestingly, a glucuronosyltransferase has been shown recently to be a major actor of pectin biosynthesis, and thus is being implicated in plant adhesion mechanisms (Iwai et al. 2002). This glucuronosyltransferase is related to the vertebrate EXT2 involved in heparan-sulfate polymerization. As yet, no physiological function has been assigned to the β1,3-glucuronosyltransferases identified in our study. This would be of special interest with regard to plant cell adhesion events.

Multiple alignments of the 119 β1,3-glucuronosyltransferase sequences from distant species highlighted conserved peptide motifs that contain highly conserved amino acids. Among the most conserved from plants to vertebrates, is a tyrosine residue in motif 1 (at position 84 in GlcAT-I). Comparison of the three-dimensional structure of the crystallized glycosyltransferases has provided evidence that the human GlcAT-I and GlcAT-P display a GTA fold with a structurally conserved N-terminal nucleotide–sugar binding domain (Coutinho et al. 2003). This domain consists of a general central parallel β-strand arrangement surrounded by a variable number of helices on both sides, reminiscent of the Rossman fold (Breton et al. 2006). Crystal analysis of GlcAT-I indicated that the side chain of Y84 maintains the uridine base in a parallel ring-stacking arrangement (Pedersen et al. 2000). We document in this study a complete loss of activity resulting from tyrosine to alanine substitution at position 84, substantiating the critical function of Y84. Interestingly, the replacement of Y84 by a phenylalanine significantly restored GlcAT-I activity and binding properties of the enzyme. The presence of a histidine residue in place of tyrosine in the sqv8-encoded C. elegans glucuronosyltransferase led us to analyze the effect of this substitution in GlcAT-I. Interestingly, the Y84Q mutant retains significant activity with a K_m value toward UDP-GlcA slightly increased compared to that of the wild type. Altogether, these data provide evidence for the major contribution of the phenyl group of Y84 in π-stacking interactions with the uracil base, and that this structural element represents a key feature of the β1,3-glucuronosyltransferase family. Consistently, the side chain of Y93 in GlcAT-P (corresponding to Y84 in GlcAT-I) has been shown to be involved in stacking interaction with the uridine base ring of the donor substrate (Kakuda et al. 2004). The contribution of aromatic stacking interactions in binding the uracil base has also been highlighted in the case of several other glycosyltransferases, including the bovine β4-galactosyltransferase (Gastinel et al. 1999).

Beyond structural similarities, a conserved feature among GTA enzymes consists of the presence of an aspartate residue, which coordinates N-3 of the uracil base (Coutinho et al. 2003). Indeed, a conserved aspartate in motif 2 (corresponding to D113 in GlcAT-I) was identified among the β1,3-glucuronosyltransferases, although this residue was not strictly invariant, especially in plants. Analysis of the X-ray crystal structure of the human GlcAT-P in complex with UDP indicated that D122 is in hydrogen-bond distance to N-3 of uridine (Kakuda et al. 2004). We show here that replacement of D113 of GlcAT-I by the longer carboxylate chain of glutamic acid or by the amide group of asparagine strongly reduced the enzyme activity together with an increase in K_m value toward UDP-GlcA. A mutational analysis of GlcAT-P fused to glutathione-S-transferase (GST) showed that substitution of D122 to alanine decreased the transferase activity to <5% of the wild-type fusion protein (Ohtsubo et al. 2000). However, we observed that none of the conservative or nonconservative mutations of D113 in GlcAT-I resulted in the suppression of the enzyme activity or UDP binding. Altogether, these results indicate that this position may tolerate some amino acid changes, without completely compromising the function of the enzyme.

The GTA glycosyltransferases share a common, though not sequence-invariant ribose/Mn²⁺ coordinating carboxylates, generally termed DXD motif that is found in the conserved motif 4 of the β1,3-glucuronosyltransferase family. We have previously highlighted that the organization of this motif is very similar between SpsA from Bacillus subtilis and GlcAT-I with the two last adjacent aspartates of a XDD sequence playing a major role in ribose/Mn²⁺ interactions (Tarbouriech et al. 2001; Gulberti et al. 2003). In plants, D194 and D195 are the best conserved. The biological significance of this finding remains to be investigated.

In the present study, we investigated the role of the two active site arginine residues R156 and R310 located in the vicinity of the pyrophosphate bridge of UDP-GlcA. Crystal structure analysis of GlcAT-I suggested that the linkage region between uridine and GlcA plays a key role in the glycosyltransferase catalytic mechanism (Pedersen et al. 2002; Negishi et al. 2003). In addition to coordinating the Mn²⁺ ions in a conformation typical of bidendate divalent metal-bound nucleotides, the phosphates directly interact with the protein. In UDP–N-acetylglucosaminyltransferase I, the α-phosphate makes a salt bridge with R117 and a hydrogen bond to the amide nitrogen of V321, and the β-phosphate hydrogen bonds to the hydroxyl group of S322 (Unligil et al. 2000). In GlcAT-I, atom NH1 of R156 is in hydrogen-bond distance from O2β of the α-phosphate, whereas atom NH1 of R310 may interact with O1β of the β-phosphate (Pedersen et al. 2002). Similarly, in GlcAT-P, phosphoric acid moieties of UDP are recognized by R165 and R313 through hydrogen bonds (Kakuda et al. 2004). The functional importance of R156 was supported by the strong conservation of this residue among the β1,3-glucuronosyltransferase family members and by mutational analysis of this residue in GlcAT-I. Our results indicated that replacement of R156 by alanine or lysine strongly impairs the enzyme activity. Nevertheless, we show that all the mutants have retained most of their ability to bind to the UDP-affinity gel, leading to the suggestion that the interactions between R156 and the pyrophosphate group may not be essential in this respect. Recent crystal analysis of the GlcAT-I–UDP-GlcA complex has highlighted potential interactions between R156 and the GlcA part of the donor substrate (Pedersen et al. 2002). Our results support the idea that the importance of the interactions of R156 with GlcA may predominate over those with the α-phosphate in the functional role of this residue.

On the other hand, we found that R310 partially contributes to the donor substrate interaction and activity. This assertion is substantiated by the observation that the mutations reduced, but did not abrogate, enzyme activity, nor obviate binding of the protein to UDP-beads. These results are consistent with studies performed on GlcAT-P fused to GST (Ohtsubo et al. 2000). In the fusion protein, mutation of R313 (corresponding to R310 in GlcAT-I) did not abolish enzyme activity, leading the authors to the conclusion that the phosphate moiety partially contributes to the donor substrate interaction but is not essential. Supporting this idea, the arginine residue at position 310 in GlcAT-I or 313 in GlcAT-P is replaced by a glutamine in the D. melanogaster orthologs forms (DmGlcAT-BSI and BSII; Kim et al. 2003), whereas R156 is invariant in the β1,3-glucuronosyltransferase family. Furthermore, the R310Q GlcAT-I mutant constructed in the present work retains significant activity, indicating that amino acid substitution at this position does not dramatically impair the function of the enzyme.

The position of GlcA in GlcAT-I appears to be determined through a hydrogen-bond network with several conserved residues (Pedersen et al. 2002). We previously emphasized the critical role of the conserved H308 in UDP-GlcA recognition (Ouzzine et al. 2002). Here, we provide evidence for the strict conservation of R161 in all species considered, including plants, and we show that it is essential for enzyme function, likely through hydrogen-bond formation with the 4-OH group of GlcA (Pedersen et al. 2002). To our knowledge, no structural or functional information on the role of this residue is available in other related β1,3-glucuronosyltransferase members.

In conclusion, we describe the phylogenetic organization of the β1,3-glucuronosyltransferase family and suggest that the function of these enzymes in the synthesis of glycosaminoglycan and glucuronosyl epitopes is probably conserved throughout eukaryotic evolution, since conserved β3-glucuronosyltransferase peptide motifs and conserved inter-motif distances were found in most eukaryote species analyzed. The physiological function of the plant β1,3-glucuronosyltransferase-like proteins remains to be investigated. In addition, we have demonstrated that the amino acid residues belonging to the UDP-GlcA binding site perform distinguishable roles and exhibit different functional importance. We found that D113 and R310 residues are promiscuous to amino acid substitutions. However, R156 and R161 residues, which are strictly conserved among β1,3-glucuronosyltransferases, are critical for activity. In addition, our results emphasize the major contribution of Y84 in UDP binding and activity. Altogether, this work suggests that the conservation of specific residues through evolution may ensure UDP-GlcA donor substrate specificity and the β1,3-glucuronosyltransferase activity of these enzymes.

A systematic search for new β1,3-glucuronosyltransferase sequences was performed by TblastN using EST and WGS data banks as previously described for sialyltransferases (Harduin-Lepers et al. 2005). DNA contigs were made by Cap3 (Huang and Madan 1999) for each new β1,3-glucuronosyltransferase-like gene, within each species. New complete protein sequences were analyzed for the presence of the N-terminal transmembrane domain by PHD-htm (Combet et al. 2000) and for the presence of the eight β1,3-glucuronosyltransferase conserved peptide motifs. Eighty-seven new β1,3-glucurunosyltransferase-like sequences were found by these approaches and were submitted to EMBL: AJ557581-2; AJ567371-5; AJ577238-9; AJ577360-2; AJ577569-70; AJ577582-3; AJ577827-8; AJ888971-82; AJ889606-17; AJ890836-45; AJ969044-65; AJ971040-3; AJ971295, and AJ971871-5 (see online supplementary table).

Protein and DNA alignments were performed by ClustalW (Thompson et al. 1994) and saved in Pir format. The Pir alignment was used for the selection of informative positions by G-Blocks (Castresana 2000). Phylogeny analysis was carried out with Phylowin (Galtier et al. 1996) using neighbor joining, observed distances, and 500 bootstrap replicates (Felsenstein 1985). The length of the horizontal branches are a measure of the genetic distance between nodes or between the peripheral side of each branch and a node, based on the number of substitutions per site for a unit branch length (see scale bar in Fig. 1). The genetic distance between two subfamilies is the sum of the branch distances from the periphery of the first subfamily to the first common node plus the distance from this node to the periphery of the second subfamily.

Table 2. Sequence of the primers used for site-directed mutagenesis

Kinetic parameters were evaluated from initial velocity values of the reaction performed as described above, using varying concentrations of UDP-GlcA (0–20 mM) at a constant concentration of Galβ1,3Gal (20 mM). K_m and V_max values were determined by nonlinear least square regression analysis of the data fitted to Michaelis-Menten rate equation (v = V_max × [S] / K_m + [S]) using the curvefitter program of Sigmaplot 9.0 (Segel 1975).

Supplementary material for Electronic Edition: 1, EMBL acc number.doc.

This work was supported by grants from “Fonds National pour la Science” (ACI no. 0693 “Biologie Cellulaire, Moléculaire et Structurale” BMCS152 2004), from Association Nationale pour la Recherche (GT-GAG NT05-3_42251/2005), from IT2B “Centre National de la Recherche Scientifique” (CNRS)-“Institut National de la Santé et de la Recherche Médicale” (INSERM), and from PRO-A INSERM and PHRC Régional programs. This work was also funded partially by CNRS (GDR 2590) and by “Association pour la Recherche sur le Cancer” (ARC, 3611).