Human alpha(1,3/1,4)-fucosyltransferases discriminate between different oligosaccharide acceptor substrates through a discrete peptide fragment.

Five different human α(1,3)-fucosyltransferase (α(1,3)-Fuc-T) genes have been cloned. Their corresponding enzymes catalyze the formation of various α(1,3)- and α(1,4)-fucosylated cell surface oligosaccharides, including several that mediate leukocyte-endothelial cell adhesion during inflammation. Inhibitors of such enzymes are predicted to operate as anti-inflammatory agents; in principle, the isolation or design of such agents may be facilitated by identifying peptide segment(s) within these enzymes that interact with their oligosaccharide acceptor substrates. Little is known, however, about the structural features of α(1,3)-Fuc-Ts that dictate acceptor substrate specificity. To begin to address this problem, we have created and functionally characterized a series of 21 recombinant α(1,3)-Fuc-T chimeras derived from three human α(1,3)-Fuc-Ts (Fuc-TIII, Fuc-TV, and Fuc-TVI) that maintain shared and distinct polypeptide domains and that exhibit common as well as idiosyncratic acceptor substrate specificities. The in vivo acceptor substrate specificities of these α(1,3)-Fuc-T chimeras, and of their wild type progenitors, were determined by characterizing the cell surface glycosylation phenotype determined by these enzymes, after expressing them in a mammalian cell line informative for the synthesis of four distinct α(1,3)- and α(1,4)-fucosylated cell surface oligosaccharides (Lewis x, sialyl Lewis x, Lewis a, and sialyl Lewis a). Our results indicate that as few as 11 nonidentical amino acids, found within a “hypervariable” peptide segment positioned at the NH2 terminus of the enzymes' sequence-constant COOH-terminal domains, determines whether or not these α(1,3)-Fuc-T can utilize type I acceptor substrates to form Lewis a and sialyl Lewis a moieties.

Our results indicate that as few as 11 nonidentical amino acids, found within a "hypervariable" peptide segment positioned at the NH 2 terminus of the enzymes' sequence-constant COOH-terminal domains, determines whether or not these ␣(1,3)-Fuc-T can utilize type I acceptor substrates to form Lewis a and sialyl Lewis a moieties.
Cell surface ␣(1,3)and ␣(1,4)-fucosylated oligosaccharides have received a substantial amount of attention because some are thought to be essential to the initiation of immune cell adhesion to vascular endothelium during the inflammatory process (reviewed in Refs. 1-5 and 6 -15). This is perhaps best characterized by the "rolling" type of adhesion mediated by interactions between sialyl Lewis x (sLe x )-bearing glycoconjugates on leukocytes and P-and E-selectin expressed by activated vascular endothelium (5). Animal intervention studies indicate that such fucosylated oligosaccharide molecules can act as anti-inflammatory agents by inhibiting leukocyte-endothelial cell interactions (15)(16)(17)(18). Furthermore, absence of such molecules on the leukocytes of individuals with the rare human leukocyte adhesion II syndrome is associated with profound defects in leukocyte-endothelial cell adhesion and with nonpyogenic infections (19,20). These observations suggest that compounds capable of specifically inhibiting leukocyte sialyl Lewis x expression might represent candidates for anti-inflammatory pharmacologic agents.
Although the primary sequences are known for several ␣(1,3)-Fuc-Ts (9,(21)(22)(23)(24)(25)(26)(27)(28), the structural determinants within these enzymes that dictate their different substrate specificities remain undefined. The purpose of this work is to identify such protein sequence(s) and to provide a conceptual background for work to design or identify molecules that might inhibit ␣(1,3)-Fuc-Ts by interacting with acceptor substrate binding sites.
In this study, we chose to explore three human ␣(1,3)-Fuc-Ts (Fuc-TIII, Fuc-TV, and Fuc-TVI) with informative structural and catalytic properties. These enzymes share approximately 85% overall amino acid sequence identity (24,25). The COOH termini of these enzymes maintain nearly identical amino acid sequences, whereas their NH 2 -terminal regions are punctuated by foci of nonidentical amino acids; we term these regions "hypervariable" segments. These three enzymes maintain shared and distinct acceptor substrate specificities. In partic-ular, each of the three can effectively utilize neutral (Gal␤13 4GlcNAc-) and ␣(2,3)-sialylated (NeuNAc␣233Gal␤134Glc-NAc-) type II acceptors to create cell surface Lewis x (Le x ) and sLe x determinants, respectively, when tested by transfection in cultured cells (24,25), and when tested with low molecular weight acceptor substrates in in vitro fucosyltransferase assays. Fuc-TIII can also efficiently utilize neutral (Gal␤13 3GlcNAc) and sialylated (NeuNAc␣233Gal␤133GlcNAc-) type I acceptor substrates to generate Lewis a (Le a ) and sialyl Lewis a (sLe a ) molecules (6,21) in transfected cells and in vitro. By contrast, neither Fuc-TV nor Fuc-TVI directs formation of fucosylated type I antigens (i.e. Le a and sLe a ) when tested in transfected cells, and neither can efficiently utilize low molecular weight acceptor substrates in vitro (24,25). Thus, the three enzymes' shared, and distinct, catalytic specificities, along with similarities in, and differences between, their primary sequences, provide an opportunity to identify amino acid residues present in Fuc-TIII and/or absent from Fuc-TV and Fuc-TVI that determine type I oligosaccharide acceptor substrate utilization.
To identify such residues, we studied the pair of enzymes that differ most dramatically in their respective abilities to utilize type I acceptors (Fuc-TIII versus Fuc-TVI; Fig. 1A). We divided the amino-terminal hypervariable regions of each enzyme into five subdomains, each consisting of a distinct region of nonidentical peptide sequence, separated by a region of identity (Fig. 1B). We then created a series of recombinant ␣(1,3)-Fuc-T chimeras in which these domains, or other domains derived from the enzymesЈ COOH-terminal regions, were transposed by themselves, or in combination, between the three enzymes, at corresponding positions. Functional analysis of these Fuc-TIII:Fuc-TVI chimeras (Fuc-TCs), and their wild type progenitors, via expression in a transfected mammalian cell line, indicates that the acceptor substrate specificity of these enzymes is determined by peptide sequences in the hypervariable NH 2 -terminal segments of these enzymes. When considered together with results obtained with a Fuc-TIII:Fuc-TV chimera, these analyses indicate that type I acceptor substrate utilization is determined by as few as eleven residues within a 61-amino acid-long segment immediately proximal to the enzymesЈ "sequence-constant" COOH-terminal domains.
Site-directed Mutagenesis-Site-directed mutagenesis procedures were used to install novel restriction sites in the wild type Fuc-TIII and Fuc-TVI coding sequences to facilitate excision and replacement of DNA segments corresponding to the various Fuc-TIII and Fuc-TVI subdomains. DNA fragments corresponding to the coding regions of Fuc-TIII and Fuc-TVI were first generated by the polymerase chain reaction (32). Fifty nanograms of plasmid template (pCDM7-␣(1,3/1,4)-FT (21), or pcDNAI-Fuc-TVI (25) were used in reactions with the Gene Amp kit (Perkin-Elmer). Primers were used that flank the coding region of each gene (Fuc-TIII sense primer is GCGCGAATTCATGGATCCCCTGG-GTGCAGCCAAGCCACAAT, Fuc-TIII antisense primer is GCGCTC-TAGAGGCAGATGAGGTTCCCGGCAGCCCAGGCAC; Fuc-TVI: sense primer is GCGCGAATTCTCCTCTCTCCCCACTTCCCAGAGACT, Fuc-TVI antisense primer is CGCGTCTAGAGGTGAAGCTTCAGGCAA-ACGAGTCCTTAGGT). Primers were designed to yield PCR products with an EcoRI site at the 5Ј end, and an XbaI site at the 3Ј end. Twenty cycles of amplification were performed, consisting of 1.5 min at 94°C, followed by annealing and extension for 3.5 min at 72°C. The PCR products were digested with EcoRI and XbaI and were cloned between these sites in the vector pcDNAI (Invitrogen), using standard molecular biology procedures (33). The DNA sequence of the insert in each vector was determined using the dideoxy chain termination method (34) and T7 DNA polymerase (Sequenase, U. S. Biochemical Corp.). A representative error-free clone was chosen for each gene (termed pcDNAI-Fuc-TIII and pcDNAI-Fuc-TVI) and was used for expression and mutagenesis procedures.
The coding regions of Fuc-TIII and Fuc-TVI were mutated to create unique intersubdomain restriction endonuclease sites. Restriction sites were chosen to correspond to positions within the conserved amino acid sequence between subdomains, to maintain the wild type protein sequence, and to be unique within the plasmid vector. For Fuc-TIII, this involved installation of five new intersubdomain restriction sites and elimination of an MscI site. Creation of six new intersubdomain restriction sites was required for Fuc-VI. Mutagenesis was completed using the p-Alter Mutagenesis kit (Promega), after subcloning the inserts in pcDNAI-Fuc-TIII and pcDNAI-Fuc-TVI between the EcoRI and XbaI sites of the plasmid p-Alter (Promega). Simultaneous multiple oligonucleotide site-directed mutagenesis was performed on single strand templates, using procedures recommended by the manufacturer. Sequences of the mutagenic oligonucleotides are shown in Fig. 1B. Candidate mutant clones were screened for the desired restriction sites by restriction enzyme digestion. Mutated clones containing each of the desired mutations were sequenced (34) in their entirety to identify ones that contained no undesired additional mutations. Single, appropriately mutated, but fully correct clones for p-Alter-Fuc-TIII and p-Alter-Fuc-TVI, were digested with EcoRI and XbaI, and the released mutated inserts were recloned between the EcoRI and XbaI sites of pcDNAI.
Finally, chimera Fuc-TC21 was constructed by replacing subdomains 4 and 5 of Fuc-TIII with a region corresponding to subdomains 4 and 5 of Fuc-TV. The PCR and oligonucleotide primers containing a 5Ј Bst11071 site and a 3Ј BglII site (sense primer: CCGACTCCAGTGTAT-ACCCACAGGCAGAC; antisense primer: GTCGGAGTCAGATCTG-TAGGACAT) were used to generate this segment. PCR amplification conditions were chosen to minimize Taq polymerase-induced errors (annealing at 94°C for 3 min; 20 cycles consisting of 1 min at 94°C, 45 s at 65°C, and 1 min at 72°C; a final 7-min extension at 72°C). The PCR product was digested with Bst1107I and BglII, and the resulting fragment was used to replace the corresponding Bst1107I and BglII fragment (subdomains 4 and 5) of the Fuc-TIII coding region. An errorfree clone identified by sequence analysis was termed Fuc-TC21 and was used for transfection analyses.
Transfection of COS-7 Cells-COS-7 cells were cultured and transfected (DEAE-dextran method) as described previously (21,28). Cell were transfected with 20 g of negative control vector plasmid DNA (pcDNAI) or with the plasmid DNAs pcDNAI-Fuc-TIII, pcDNAI-Fuc-TVI, or each of the various pcDNAI-Fuc-TCs. Cells were also cotransfected with 1 g of the plasmid pCDM8-CAT (35) to allow for normalization of flow cytometry and Western blot data to transfection efficiency.
Antibodies-Anti-Lewis x (anti-SSEA-1, IgM, as ascites; Ref. 36 Flow Cytometry Analysis-Transfected COS-7 cells were harvested 72 h after transfection and were stained with antibodies diluted in staining media (41), using procedures described previously (6,21,29). Anti-Lewis a and anti-H ascites were used at a dilution of 1:500 and 1:100, respectively. Anti-SSEA-1 ascites was used at a dilution of 1:1000. Anti-sialyl Lewis a was used at a concentration of 10.8 g/ml. Anti-sialyl Lewis x was used at a dilution of 10 g/ml. Cells were subsequently washed and then stained with either a fluorescein isothiocyanate-conjugated goat anti-mouse IgM (Vector) used at a dilution of 1:200 or with a fluorescein isothiocyanate-conjugated goat antimouse IgG (Sigma) used at a dilution of 40 g/ml. Stained cells were then analyzed on a Becton-Dickinson FACScan (6, 21, 29). Thresholds for antigen positivity were set at a fluorescence intensity level that excludes 99% of COS-7 cells that had been transfected with the negative control vector and stained with the test antibody. These thresholds were set individually for each test antibody, on the day of each experiment, before the experimental transfectants were analyzed.
Fucosyltransferase Assays-Extracts for ␣(1,3)-Fuc-T assays were prepared by harvesting COS-7 cells 72 h after transfection and solubilizing the cell pellets in 1% Triton X-100 (21,22,24,25). Cell extract protein contents were determined with a micro BCA protein assay reagent (Pierce). Fucosyltransferase assays were performed in a final volume of 20 l and contained 25 mM sodium cacodylate (pH 6.2), 5 mM ATP, 10 mM L-fucose, 20 mM MnCl 2 , and 3 mM GDP-[ 14 C]fucose. Cell extract protein concentration was varied to approximate linear reaction rate conditions (net [ 14 C]fucose incorporation rate of a less than 20% of total input GDP-[ 14 C]fucose radioactivity). Concentrations of low molecular weight acceptor substrates (Gal␤134GlcNAc,N-acetyllactosamine or Gal␤133GlcNAc lacto-N-biose I) were varied between 0.25 and 167 mM for apparent K m determinations. Reactions were incubated at 37°C for 1 h and were terminated by the addition of 20 l of ethanol, followed by dilution with 560 l of water. The total radioactivity in each reaction was determined by subjecting an aliquot to scintillation counting. To assess the amount of radioactive fucose incorporated into product, an additional aliquot was applied to a Dowex 1ϫ2-400 column in the formate form (27)(28)(29)(30). The flow-through fraction plus an additional 2 ml of a water column "wash" was collected, pooled, and subjected to scintillation counting. Control assays containing no added acceptor substrate were also completed in parallel. Radioactivity in the flowthrough fraction, obtained in the absence of added acceptor, was no more than 1% of the total added radiolabel.
In some instances, a fucosyltransferase was found to maintain a very low inherent specific activity toward one or both low molecular weight acceptor substrates, such that the rate of transfer was found to be directly proportional to acceptor substrate concentrations between 0.25 mM and at least 167 mM. In these instances, data in Table I indicates that the apparent Michaelis-Menten constant exceeds 50 mM. As a measure of the relative efficiencies with which the enzymes utilize N-acetyllactosamine and lacto-N-biose I, the specific activity of each enzyme was determined for each of these two acceptor substrates (20 mM concentration). The ratio of the specific activity was calculated and displayed in Table I.
Chloramphenicol Acetyltransferase Assays-Transfected COS-7 cells were harvested 72 h after transfection, and extracts were prepared from them by sonicating the cells in 0.25 M Tris-HCl (pH 8.0). Chloramphenicol acetyltransferase assays were performed as described previously (35,42). Total cell extract protein in each reaction was adjusted to produce less than 15% incorporation of [ 14 C]chloramphenicol during a 1-h incubation at 37°C.
Western Blot Analysis-Cell extracts were prepared from transfected COS-7 cells 72 h after transfection. Extracts contained 50 mM Tris-HCl (pH 6.8), 10% SDS, and 10% glycerol. Extracts were boiled for 3 min immediately after preparation and were stored frozen until use. Protein content was determined using the BCA reagent procedure. Extracts were prepared for SDS-polyacrylamide gel electrophoresis by adding dithiothreitol to a final concentration of 0.1 M and bromphenol blue to a final concentration of 0.05%. Samples were then boiled and fractionated by electrophoresis through a 10% SDS-polyacrylamide gel. After electrophoresis, the proteins were electrotransferred to a polyvinylidene difluoride membrane (Bio-Rad). The membrane was rinsed and then blocked for 12-14 h at 4°C in phosphate-buffered saline (pH 7.4) containing 10% bovine serum albumin and 0.2% Tween 20. The blot was washed at room temperature in phosphate-buffered saline (pH 7.4), 0.2% Tween 20 and was probed with a 1:1000 dilution of a polyclonal rabbit antiserum prepared against a fragment of the human Fuc-TIII catalytic domain expressed in Escherichia coli. 2 Alternatively, an identically prepared blot was probed with a 1:1000 dilution of pre-immune rabbit serum. The blot was then washed and then probed with a 1:5000 dilution of a horseradish peroxidase-conjugated anti-rabbit immunoglobulin (Sigma). The blot was then rinsed, exposed to ECL reagent (Amersham Corp.) and subjected to autoradiography.

Analysis of COOH-terminal Subdomain Exchanges; Type I Substrate Utilization Is Determined by Sequences in the NH 2 -
terminal Hypervariable Segment-The amino acid sequences of human Fuc-TIII and Fuc-TVI differ from each other at only seven corresponding positions, COOH-terminal to residue 152/ 151 in Fuc-TIII/Fuc-TVI (Fig. 1A). By contrast, their amino acid sequences differ substantially in a hypervariable region bounded by their transmembrane segments (residue 37) and the beginning of their "sequence-constant" COOH-terminal domains (residues 152 and 151). It therefore seemed most likely that peptide sequence(s) within their respective hypervariable segments determine whether each enzyme can (Fuc-TIII; Refs. 6 and 21) or cannot (Fuc-TVI; Ref. 25) (Fig. 1A) efficiently utilize type I acceptor substrates.
This hypothesis was tested by first characterizing the substrate utilization patterns of recombinant Fuc-T chimeras in which the nearly identical COOH termini of Fuc-TIII and Fuc-TVI were exchanged with each other (Figs. 1A and 2). Substrate utilization was determined by expressing recombinant chimeras in COS-7 cells, which maintain acceptor substrate precursors for surface-localized Le x , sLe x , Le a , and sLe a epitopes (6,21) and characterizing the chimera-determined cell surface glycosylation patterns of the transfected cells using flow cytometry.
Four chimeras informative for the COOH termini of these enzymes were evaluated. In two of these chimeras (Fuc-TC1 and Fuc-TC3), the exchanged segments were between amino acid residue 301 of Fuc-TIII or 300 of Fuc-TVI (at positions corresponding to a BstXI site in both genes; Figs. 1A and 2) and the COOH terminus of each enzyme. This represents a 60amino acid-long peptide segment in Fuc-TIII and a 59-residuelong peptide in Fuc-TVI. These segments differ from each other at three amino acid positions and by the presence (Fuc-TIII) or absence (Fuc-TVI) of a valine residue at position 353 (in Fuc-TIII) (Fig. 1A). In the other two chimeras (Fuc-TC2 and Fuc-TC4; Fig. 2), the entire COOH-terminal regions were exchanged. These transposed segments encompass all seven amino acid sequence differences distal to the BglII site corresponding to amino acid 160 in Fuc-TIII and 159 in Fuc-TVI (Figs. 1 and 2). COS-7 cells transfected with chimeras Fuc-TC3 and Fuc-TC4 (which maintain the NH 2 -terminal segment of Fuc-TVI and portions of the COOH terminus of Fuc-TIII) express the Le x and sLe x antigens, but not the type I-based Le a or sLe a epitopes (Fig. 2). By contrast, COS-7 cells transfected with chimeras Fuc-TC1 and Fuc-TC2 (which maintain the NH 2terminal segment of Fuc-TIII and portions of the COOH terminus of Fuc-TVI) express the Le x , Le a , and sLe a epitopes; however, neither chimera determines expression of the sLe x antigen (Fig. 2). The reason that these latter two chimeras do not yield sLe x expression is not known. Nonetheless, these results clearly demonstrate that the NH 2 -terminal hypervariable segments of Fuc-TIII and Fuc-TVI determine their respective abilities to utilize or disregard the type I acceptor substrates expressed by COS-7 cells.
This conclusion is reinforced by in vitro determinations of the relative efficiencies with which the fucosyltransferases utilize a pair of representative, neutral low molecular weight type I and type II acceptor substrates, N-acetyllactosamine and lacto-Nbiose I (Table I). Wild type Fuc-TIII exhibits similar Michaelis-Menten constants for these type II and type I substrates (8.1 and 12.7 mM, respectively), although the enzyme exhibits a substantially higher relative specific activity with the type I substrate. By contrast, Fuc-TVI maintains a relatively low apparent K m for the type II acceptor (4.9 mM), a relatively high apparent Michealis-Menten constant for the type I acceptor (greater than 50 mM), and exhibits a substantially higher relative specific activity toward the type II substrate (Table I).
The similarity between Fuc-TC4 and Fuc-TVI observed by flow cytometry analysis (Fig. 2) is also evident from the enzymatic analyses; Fuc-TC4 exhibits apparent Michealis-Menten con-  striped box (o). Vertical lines within the rectangles denote the positions of amino acid residues that differ between the two enzymes. These differences are concentrated in the hypervariable segment between the AflI I and BglII restriction sites shown under the rectangles. The carboxyl-terminal regions of these enzymes, distal to the BglII site, differ at seven positions. The relative positions of these differences, and their nature (six nonidentical amino acids and one amino acid/insertion (identified in single letter amino acid code)) are also indicated in the figure. Sites of potential asparagine-linked glycosylation are indicated by ( ). The disaccharide portions of the potential oligosaccharide acceptors are also shown in the figure. "LacNAc" denotes the type II precursor Nacetyllactosamine. LNB-I denotes the type I precursor lacto-N-biose I. Arrows near the disaccharide moieties point to the hydroxyl group that will be modified by fucosylation. R denotes the underlying glycoconjugate that displays the disaccharide. RЈ represents either a hydroxyl group in the neutral precursors for Lewis x or Lewis a products or ␣(2,3)-linked sialic acid in the sialylated precursor for the sialyl Lewis x or sialyl Lewis a products. Fuc-TIII (top) can utilize GDP-fucose to form products from both type II and type I precursors, whereas Fuc-TVI (bottom; shaded to match stants for the two acceptors, and a specific activity ratio, that do not differ substantially from those obtained with wild type Fuc-TVI (Table I). Fuc-TC2, the chimeric counterpart to Fuc-TC4 exhibits an inherently low specific activity toward both substrates, as reflected by the high apparent K m 's for the two substrates (Table I). Nonetheless, this enzyme, like wild type Fuc-TIII, utilizes the type I acceptor substrate much more efficiently than it does the type II acceptor (Table I). This result is also consistent with the observation that Fuc-TC2 directs type I product expression in COS-7 cells with an efficiency that equals or exceeds its ability to determine type II product expression (Fig. 2).

Analysis of Single Subdomain Exchanges Points to Subdomains 4 and 5 in
Type I Substrate Utilization-To further localize peptide sequence(s) in the NH 2 -terminal hypervariable segments that are important to type I and type II acceptor utilization, we arbitrarily divided these segments into five distinct subdomains. The subdomains represent short peptide segments whose sequences are different in Fuc-TIII and Fuc-TVI. Subdomains are separated from each other by short peptide sequences that are identical in the two enzymes. We used site-directed mutagenesis procedures ("Experimental Procedures") to install restriction sites in the DNA sequences encoding Fuc-TIII and Fuc-TVI at positions corresponding to the intersubdomain conserved sequence regions. These sites were located at the same relative positions in these genes and were designed to be unique and to conserve the wild type intersubdomain amino acid sequences of the two enzymes (Fig. 1). These sites allow the use of simple restriction fragment interchange procedures to create ␣(1,3)-Fuc-T chimeras in which one (or more) subdomain(s) from one enzyme are interchanged with the corresponding subdomain(s) in the other enzyme.
We used this approach to construct a series of 17 additional recombinant ␣(1,3)-Fuc-T chimeras derived from Fuc-TIII and Fuc-TVI. Each chimera was evaluated for its ability to determine expression of cell surface Le a , sLe a , Le x , and sLe x epitopes in transfected COS-7 cells, using the flow cytometry procedures described above for chimeras Fuc-TC1, Fuc-TC2, Fuc-TC3, Figs. 2-4) can utilize GDP-fucose to form products only from the type II precursors. B, subdomains within the hypervariable segments of Fuc-TIII and Fuc-TVI. The DNA and derived protein sequences of the two enzymes are aligned. The upper DNA and protein sequences correspond to Fuc-TIII. The lower DNA and protein sequences correspond to Fuc-TVI. Dotted lines indicate positions in one enzyme that do not have a counterpart in the other. Residues are numbered beginning with the initiator methionine codon of Fuc-TIII (residue number 1). Nonidentical amino acid residues are shaded. Underlined DNA sequences correspond to the synthetic oligonucleotides used to create or destroy restriction sites (positions and identity indicated above, for Fuc-TIII, or below, for Fuc-TVI) that define the boundaries of each subdomain. The nature of the DNA sequence changes needed to create these sites are denoted above (Fuc-TIII) or below (Fuc-TVI) the wild type DNA sequence. As noted in the text, these changes maintain the wild type protein sequence of each enzyme.  and Fuc-TC4. A series of single subdomain transpositions were used to identify either a Fuc-TVI subdomain that inactivates Fuc-TIII type I substrate utilization or a Fuc-TIII subdomain that confers type I acceptor utilization when transposed into Fuc-TVI (Fig. 3, A and B). Each of the five Fuc-TIII chimeras containing a single Fuc-TVI subdomain expresses each of the four cell surface antigens determined by wild type Fuc-TIII (Fig. 3A). Nonetheless, two of these chimeras (Fuc-TC8 and Fuc-TC9), representing substitutions in subdomain 4 or 5, respectively, yields a cell surface antigen profile that differs substantially from the Fuc-TIII profile. These two chimeras yield populations of transfectants that exhibit a relative reduction in the number of cells expressing the type I-based Le a and sLe a antigens and a relative increase in the fraction of cells expressing the type II-based epitopes Le x and sLe x . Specifically, the ratio of the fraction of Le a positive cells (29.8%) to the fraction of Le x positive cells (37.2%) for wild type Fuc-TIII in this experiment is 0.80. By contrast, for Fuc-TC8, this ratio is 0.19 (9.1% Le a positive versus 47.9% Le x positive), and the ratio is 0.29 for Fuc-TC9 (10.2% Le a positive versus 35.4% Le x positive). Similar ratios are observed using results obtained with antibodies against the sialylated forms of these antigens (Fig. 3A). These two chimeras thus yield a glycosylation phenotype that is roughly intermediate between the phenotypes observed for Fuc-TIII and Fuc-TVI. In contrast to the results shown in Fig.  3A, substitution of single Fuc-TIII subdomains at corresponding positions in Fuc-TVI yields a set of chimeras that each determine a glycosylation phenotype virtually identical to that determined by wild type Fuc-TVI (Fig. 3B). Taken together, these results indicate that sequences within subdomain four and five of Fuc-TIII, by themselves, can strongly, if not completely, influence the type I acceptor substrate utilization efficiency of this enzyme.

Analysis of Multiple Subdomain Exchanges; Type I Substrate Utilization Is Determined by Composites of Sequences in Subdomains 3, 4, or 5-
To determine if peptide sequence within two or more adjacent domains can more fully direct type I acceptor substrate discrimination, we constructed and tested a series of chimeras assembled via multiple subdomain exchanges. Chimera Fuc-TC15, containing subdomains 1, 2, and 3 from Fuc-TVI transposed into the wild type Fuc-TIII sequence, generates levels of cell surface Le x , sLe x , Le a , and sLe a determinants similar to that generated by wild type Fuc-TIII (Fig. 4A). Likewise, the complementary chimera Fuc-TC17 (Fig. 4B) utilizes only type II acceptor substrates to generate Le x and sLe x , in a manner virtually identical to wild type Fuc-TVI. These results indicate that domains 1, 2, and 3 together contribute little, if anything, to type I acceptor substrate utilization.
By contrast, results obtained with an additional set of chimeras indicate that subdomains 4 and 5, when transposed together (with or without subdomain 3), strongly influence type I substrate usage. Replacement of subdomains 3, 4, and 5 in Fuc-TIII with the corresponding subdomains from Fuc-TVI yields a fucosyltransferase chimera (Fuc-TC16; Fig. 4A) that utilizes type II substrates exclusively to generate cell surface Le x or sLe x determinants. This phenotype is virtually identical to that generated by wild type Fuc-TVI (Fig. 4A).
In a similar manner, replacement of subdomains 3, 4, and 5 in Fuc-TVI with the corresponding subdomains from Fuc-TIII creates a fucosyltransferase chimera (Fuc-TC18; Fig. 4B) with a phenotype qualitatively similar to that of wild type Fuc-TIII. Specifically, COS-7 cells expressing this chimera express all four epitopes analyzed (Fig. 4B), although at reduced absolute levels (Le x , 8.1% positive; sLe x , 3.0% positive; Le a , 4.8% positive; and sLe a , 4.0% positive) relative to levels generated by wild type Fuc-TIII. The diminished amount of cell surface antigen expression yielded by Fuc-TC18 is apparently not a consequence of decreased expression of the chimeric enzyme, since Western blot analysis of Fuc-TC18 transfectants identifies amounts of immunoreactive Fuc-TC18 in these cells that are approximately equivalent to the amounts of wild type Fuc-TVI and nearly equivalent to wild type Fuc-TIII accumulations (data not shown). Enzyme activity assays suggest that the decreased cell surface fucosylated antigen expression we observe with this chimeric fucosyltransferase is due to the enzyme's inherently low specific activity and not because it is mislocalized or otherwise not able to effectively interact with its intracellular target oligosaccharides. ␣(1,3)-Fuc-T activity assays indicate that the specific activity of this chimera is substantially lower than either Fuc-TIII or Fuc-TVI when assayed with the low molecular weight type I and type II acceptor substrates lacto-N-biose I and N-acetyllactosamine (apparent K m values exceeding 50 mM for these substrates; Table I). Nonetheless, this chimeric enzyme exhibits the same in vitro "preference" for the type I acceptor substrate observe for wild type Fuc-TIII (Table I). Thus, even though the inherent efficiency of this chimeric fucosyltransferase is apparently substantially less than that of the wild type enzyme, the presence in this chimera of subdomains 3, 4, and 5 from Fuc-TIII is associated with retention of a phenotype that is qualitatively similar to wild type Fuc-TIII, in vitro (Table I) and in vivo (Fig. 4B).
Subdomain 3 in Fuc-TIII and Fuc-TVI varies only by the presence (Fuc-TVI) or absence (Fuc-TIII) of a consensus sequence for asparagine-linked glycosylation (Fig. 1B). This sequence, by itself, is without apparent effect on acceptor substrate utilization as judged from transfection analyses, since exchange of this subdomain alone does not alter the glycosylation phenotype determined by the corresponding mutant Fuc-TIII and Fuc-TVI chimeras (Fig. 3A, Fuc-TC7, and Fig. 3B,  Fuc-TC12). Thus, when this observation is considered together with other data in Figs. 2 and 3, and with the results obtained with chimeras Fuc-TC15, Fuc-TC16, Fuc-TC17, and Fuc-TC18, it can be supposed that type I acceptor substrate discrimination (i.e. ␣(1,4)-Fuc-T activity) can be conferred upon a "generic" ␣(1,3)-Fuc-T scaffold, derived from either Fuc-TIII or Fuc-TVI, by a set of sequences between the NH 2 -terminal boundary of subdomain 4 and the COOH-terminal boundary of subdomain 5. This hypothesis is confirmed by analysis of chimeras in which these two subdomains have been transposed together. These chimeric ␣(1,3)-Fuc-Ts (Fuc-T19 and Fuc-T20) generate glycosylation phenotypes (Fig. 4) that correspond nearly exactly to the phenotypes generated by chimeras constructed by exchange of subdomains 3, 4, and 5 (Fig. 4). Specifically, Fuc-TC19 (subdomains 4 and 5 from Fuc-TVI installed into Fuc-TIII) utilizes type II acceptor substrates with an efficiency approximately equivalent to wild type Fuc-TVI, but does not determine expression of type I molecules (Fig. 4A). The in vitro properties of this chimera are consistent with these flow cytometry results. Chimera Fuc-TC19 maintains a relatively low apparent K m for the type II acceptor (1.1 mM), a relatively high apparent Michealis-Menten constant for the type I acceptor (greater than 50 mM) and displays a high relative specific activity toward the type II substrate (Table I). The chimeric enzyme Fuc-TC19 is therefore similar to wild type Fuc-TVI and to the other "Fuc-TVI-like" chimera Fuc-TC4 analyzed in vitro.
The complementary chimera Fuc-TC20 (subdomains 4 and 5 from Fuc-TIII transposed into the corresponding position in Fuc-TVI) utilizes neutral and sialylated type I molecules, as well as neutral type II molecules, in vivo, to generate Le a (6.3% positive cells), sLe a (3.3% positive cells), and Le x (5.4% positive cells) antigens. Western blot analyses indicate that Fuc-TC20 accumulates in COS-7 cells to levels nearly equivalent to those observed when wild type Fuc-TVI is expressed in this cell line (data not shown). In vitro, fucosyltransferase assays demonstrate that this chimeric enzyme maintains an inherently low specific activity toward the type I and type II acceptor substrates examined (apparent K m values exceeding 50 mM; Table  I). However, Fuc-TC20 utilizes the type I acceptor substrate much more efficiently than it does the type II acceptor (Table I) and thus maintains the in vitro type I acceptor substrate preference exhibited by wild type Fuc-TIII. These results are analogous to those obtained with Fuc-TC18, above. They indicate that the low level of surface antigen expression determined by Fuc-TC20 is due at least in part to an inherently diminished general specific activity. However, they also indicate that this enzyme, like Fuc-TC18 above, has acquired a Fuc-TIII-like phenotype by virtue of subdomains 4 and 5 derived from Fuc-TIII.
Fuc-TC20 is unusual in that it does not yield cell surface sialyl Lewis x expression, but does direct expression of the other three fucosylated cell surface antigens (Fig. 4B). Given that this chimera differs from Fuc-TC18 only by the absence of a potential asparagine-linked glycosylation site within subdomain 3, the observations derived from these two chimeras suggested the possibility that display of an asparagine-linked oligosaccharide by subdomain 3, at least in the context of an otherwise intact Fuc-TVI scaffold, enables catalysis of sialyl Lewis x formation, whereas its absence disables sialyl Lewis x formation. This possibility is supported by the results obtained with chimeras Fuc-TC1 and Fuc-TC2 (Fig. 2), which determine a glycosylation phenotype analogous to that of Fuc-TC20 (Fig.  4B). In chimeras Fuc-TC1 and Fuc-TC2, absence of an asparagine-linked oligosaccharide modification site in subdomain 3, in the context of a full (Fuc-TC2) or partial (Fuc-TC1) COOHterminal Fuc-TVI scaffold, also correlates with an inability to determine cell surface sialyl Lewis x formation. However, we note that Fuc-TC12 (Fig. 3B), representing a fully wild type Fuc-TVI sequence, except for absence of the asparagine-linked glycosylation site in subdomain 3, determines robust expression of the cell surface sialyl Lewis x antigen (Fig. 3B). Clearly then, the unusual phenotype exhibited by chimeras Fuc-TC20, Fuc-TC1, and Fuc-TC2 cannot be explained by the simple absence of a single asparagine-linked glycosylation site. Naturally, additional experiments will be required to determine if this site is actually modified by glycosylation in the transfected cells and how it may participate in this unusual phenotype.
In aggregate, the results obtained with the recombinant fucosyltransferase chimeras derived from Fuc-TIII and Fuc-TVI indicate that some complement of the 21 nonidentical amino acid residues encompassed within subdomains 4 and 5 of Fuc-TIII and Fuc-TVI (Fig. 1B) can determine whether or not these enzymes efficiently utilize type I oligosaccharide acceptor substrates in COS-7 cells and strongly influence their ability to discriminate between type I and type II acceptors in vitro.
Subdomains 4 and 5 from Fuc-TV Leave Intact the Type I Substrate Utilization Activity of Fuc-TIII-A third human ␣(1,3)-Fuc-T, termed Fuc-TV (24), shares enzymatic properties with both Fuc-TIII and Fuc-TVI and is very similar in its primary sequence to these enzymes. Specifically, Fuc-TV determines a cell surface glycosylation phenotype similar to that generated by Fuc-TVI when examined in transfected cell lines (24), but can also utilize type I acceptor substrates at detectable efficiencies, when tested in vitro with low molecular weight acceptor substrates (24,25). Its primary sequence across subdomains 4 and 5 is equally similar to both Fuc-TIII and Fuc-TVI (34 identical residues across spans of 47; Fig. 5). This enzyme and its properties, therefore, present an opportunity to further define type I substrate utilization-determining amino acid residues within subdomains 4 and 5 by creating and testing a chimera in which subdomains 4 and 5 from Fuc-TV are exchanged with the corresponding domains in Fuc-TIII. The results of these manipulations indicate that robust type I substrate utilization is maintained when subdomains 4 and 5 within Fuc-TIII are substituted with the corresponding subdomains from Fuc-TV (Fig. 4A). These results, when considered together with those obtained with the chimeras derived from Fuc-TIII and Fuc-TVI, indicate that the type I acceptor substrate utilization properties of these enzymes can be controlled by a quite limited subset of the 21 nonidentical amino acid residues suggested to be important by the Fuc-TIII:Fuc-TVI chimera experiments (Fig. 5). DISCUSSION These experiments demonstrate that utilization of type I oligosaccharide substrates by the human ␣(1,3)-fucosyltransferases Fuc-TIII, Fuc-TV, and Fuc-TVI is determined in large measure by a relatively short peptide segment, encompassed by subdomains 4 and 5, and localized to the NH 2 terminus of each enzyme's Golgi lumenal domain. Examination of the residues within this segment, in each enzyme, suggests the simple possibility that, in the context of a "generic" ␣(1,3)-fucosyltransferase background, efficient type I substrate utilization is allowed by residues shared by Fuc-TIII and Fuc-TV and found within subdomains 4 and 5, whereas type I substrate utilization is "disabled" by residues in this segment that are unique to Fuc-TVI.
In the context of this simple model, the more obvious "Fuc-TVI-unique" amino acid residues (Fig. 5) that might operate singly, or in concert with other neighboring residues, to disable type I acceptor substrate utilization, include the nonconservative differences represented by (i) the arginine residue at position 110 in Fuc-TVI that corresponds to a tryptophan residue that occupies the analogous position in both Fuc-TIII and Fuc-TV, (ii) a pair of arginine residues in Fuc-TVI at positions (122 and 126) that correspond to a pair of proline residues in Fuc-TIII and Fuc-TV, (iii) a tryptophan residue (143) in Fuc-TVI that corresponds to glutamine or arginine in Fuc-TIII or Fuc-TV, respectively, and (iv) a lysine residue at position 146 in Fuc-TVI that corresponds to a glutamic acid residue in Fuc-TIII and Fuc-TV. It is also possible that type I acceptor substrate utilization may be influenced by the more conservative amino acid sequence differences identified in Fig. 5.
Although these studies have identified candidate residues for acceptor substrate discrimination, they do not allow us to discern the mechanism(s) through which these residues operate in this process. However, previous work by others in this area suggests a role for hydrogen bonding interactions in ac-ceptor substrate binding and utilization. For example, in an evaluation of deoxygenated oligosaccharide acceptor substrates for two different human fucosyltransferases (corresponding to Fuc-TIII and Fuc-TVI), Hindsgaul et al. (43) report that acceptor substrates deoxygenated at the reactive hydroxyl position do not act as inhibitors of these enzymes. Hindsgaul et al. (43) interpret these results to mean that these fucosyltransferases, as well as other glycosyltransferases, bind to their acceptor substrates, and subsequently effect catalysis, through hydrogen bonding interactions between the reactive hydroxyl group on the acceptor and one or more basic groups in the enzyme. In the context of this hypothesis, we note that, in aggregate, the "Fuc-TVI-specific" residues shown in Fig. 5 are largely basic in character, relative to the corresponding residues in Fuc-TIII and Fuc-TV. The nonconserved tryptophan residues in subdomains 4 and 5, and other such polar residues, may also mediate fucosyltransferase acceptor substrate recognition, again presumably through hydrogen bonding interactions, as suggested from studies of ␤(1,4)-galactosyltransferase (44). Whether or not these residues operate to determine binding to, and fucosylation of, type II acceptor substrates exclusively (and perhaps at the expense of type I acceptor substrate binding) remains to be experimentally determined.
On a more global scale, it seems probable that fucosyltransferase-specific substrate recognition and modification properties will also be found to be more than simply a function of the ability of the relevant subdomains to "bind" to the oligosaccharide acceptor. For instance, we can anticipate that some of the residues in these subdomains may interact in space with amino acids far away along the linear sequence of these enzyme, to alter acceptor substrate utilization. Alternatively, residues in these subdomains may influence acceptor substrate recognition by residues elsewhere in the molecule, via transmitted conformational change dictated by the specific residues within these subdomains. Finally, it is possible that subdomain-dependent proteolytic events indirectly determine acceptor substrate specificity by controlling the liberation of distinct soluble fucosyltransferase fragments with different acceptor substrate specificities. The relevance of these possibilities to our observations will await an exploration of the biosynthesis and in vitro catalytic properties of the recombinant wild type and chimeric enzymes and the availability of the tertiary structures of these enzymes.
We emphasize that our studies have focused on detecting the products of these enzymes at a relatively low level of resolution (i.e. at the level of cell surface antigen expression). This approach is advantageous in that it measures product(s) formed by an enzyme within its correct cellular environment. Furthermore, since the fraction of antigen-positive cells we observe correlates very well with catalytic activities measured in vitro, this parameter seems to accurately reflect the biochemical Dashed lines represent positions in Fuc-TV and Fuc-TVI that are identical to corresponding residues in Fuc-TIII. Nonidentical residues are explicitly indicated by the single letter amino acid designation. Residues present in Fuc-TVI, but not found in either Fuc-TV or in Fuc-TIII ("unique" to Fuc-TVI) are displayed below, among residues (indicated by an asterisk) that are common to at least two of the three enzymes. The asterisks indicate residues in Fuc-TVI that are also in either Fuc-TV or Fuc-TIII or both. nature of the enzyme under study. By contrast, the mean fluorescent intensities of the antigen-positive cells fall within a similar range, regardless of the enzyme analyzed, and do not correlate with in vitro activity. We do not understand this phenomenon, although we speculate that this can be accounted for by the possibility that acceptor substrate concentration and/or accessibility within COS cells is limiting to the point that utilization of the small amounts of the available acceptor substrate molecules yields a virtually maximal, though easily detectable, signal. The broad spectrum of enzyme protein expression level obtained within a population of transiently transfected cells may play upon such limiting substrate levels to yield an off-on or digital signal response that is a function of the efficiency of the enzyme under study. For example, in a population of cells expressing an enzyme with robust inherent activity, many cells, including those that express a relatively small amount of enzyme protein, will contain sufficient intracellular enzyme activity to convert them to antigen positivity (hence the relatively larger fraction of antigen-positive cells seen with efficient enzymes), even though the limiting amounts of substrate operate to place a strict upper boundary on the "brightness" of such cells. By contrast, in a population of cells expressing an enzyme with a low inherent activity, only the relatively small fraction of cells that have accumulated relatively large amounts of enzyme protein will contain sufficient enzyme activity to exceed the threshold for substrate conversion (hence the relatively small fraction of antigen positive cells with inefficient enzymes), and will become antigen-positive, with a mean fluorescence intensity that can be roughly maximal. Additional work will be required to confirm or refute this notion.
Our approach cannot tell us much about the nature of the glycoconjugate(s) that displays the fucosylated antigens or about the structures of the glycoconjugate precursors that serve as substrates for these enzymes within these cells (aside from the fact that some of these moieties terminate with neutral and ␣233-sialylated type I and type II structures). It will be important to eventually know this information, however, since it is possible that some of the apparent subdomain-dependent acceptor substrate specificity we observe is also a function of the type of glycoconjugate that displays the acceptor di-or trisaccharide moieties with which we have concerned ourselves here. We note that there are well known precedents for this type of cis-modulation of glycosyltransferase substrate specificity by the underlying glycoprotein sequence (45,46). Whether or not this is actually the case for the fucosyltransferases studied here will first require a biochemical characterization of the cellular products formed by the transfected enzymes.