Influence of Lewis (cid:97) 1–3/4- L -Fucosyltransferase ( FUT3 ) Gene Mutations on Enzyme Activity, Erythrocyte Phenotyping, and Circulating Tumor Marker Sialyl-Lewis a Levels*

Fucosylated glycoproteins carrying (cid:97) 1–4 fucose resi-dues are of importance for cell adhesion and as tumor markers. The Lewis gene, FUT3 , encodes the only known (cid:97) 1–4-fucosyltransferase (FucT), and individuals who are deficient in this enzyme type as Lewis-negative on erythrocytes. We examined the mutational spectrum of the Lewis gene in Denmark and found 6 different muta- tions. Five, T59G, T202C, C314T, G508A, and T1067A, were frequent, and one, C445A, was only detected in one out of 40 individuals. Allele-specific polymerase chain reaction as well as cloning of FUT3 alleles showed that the 202 and 314 mutations were co-located on the same allele.

Fucosylated cell surface glycoconjugates belonging to the Lewis blood group system have recently been shown to be physiologically important. Not only do they present a challenge to transfusion medicine, but they are 1) involved in cellular development, 2) serve as tumormarkers, and 3) are believed to be essential for the adhesion between leukocytes and vascular endothelium during the inflammatory reaction.
The fucosylated structures are formed by fucosyltransferases in the Golgi apparatus that act by transferring fucose from a nucleotide sugar donor to one or more acceptor substrates. The fucosyltransferases in mammals have been named according to the bond they create, ␣1-2 FucT, 1 ␣1-3FucT, and ␣1-4FucT. These three distinct bonds can be catalyzed by a large repertoire of transferases, of which some are confined to certain tissues, and subjected to developmental regulation. Several genes encoding distinct fucosyltransferases have been cloned. Two of them encode ␣1-2-fucosyltransferases (4 -7). FUT4, FUT5, FUT6, and FUT7 encode ␣1-3-fucosyltransferases (8 -12), and FUT3 encodes an enzyme with both ␣1-3 and ␣1-4fucosyltransferase activity (13). The predicted aminoacid sequences of FUT3, -5, and -6 have a high degree of similarity, and may be the result of gene duplication (14,15).
Mutations in these genes have been the subject of several recent studies. Thus, the lack of ␣1-3FucT activity in plasma in 9% of Indonesian individuals was shown to be caused by a missense mutation (G247K) in the catalytic domain of FUT6, which was inherited in the deficient families (16). Different inactivating mutations in the FUT3 gene have been reported from Japan and Indonesia. In Indonesia, the two mutations L20R and I356K have been seen to occur individually as well as together (17). The L20R mutation is located in the transmembrane segment and has been suggested to lead to improper membrane insertion without severely affecting the catalytic activity. The I356K mutation is located to the catalytic domain and is deleterious for the catalytic activity when it is alone, as well as when it is accompanied by L20R. In Japan, two types of coupled mutations are found, the L20R combined with either G170S or I356K (18), of which the G170S mutation is deleterious to enzyme function.
The amount of circulating sialyl-Lewis a structure is widely used as a tumor marker for colorectal, pancreatic, and gastric cancer (19) and is named the Ca19 -9 cancer antigen (20,21).
Unfortunately, this tumor marker cannot be used in individuals who are Lewis-negative because of inactivating mutations in both FUT3 alleles. To use the tumor marker it is therefore important to establish which mutations inactivate the FUT3 alleles. In the present study of a Danish population, we identified two new FUT3 alleles with inactivating mutations, in addition to the four alleles already known. Furthermore, we have used FUT3 allelotyping to examine whether FUT3 heterozygous individuals have a lower amount of ␣1-4FucT activity in saliva and of circulating sialyl-Lewis a structure in serum than in individuals who are homozygous for the wild-type allele. This could be of clinical importance as the serum cut-off level between normal individuals and cancer patients would be lower in heterozygous individuals than in homozygous wildtype individuals. We found a significantly (p Ͻ 0.04) lower amount of serum sialyl-Lewis a in heterozygous individuals compared with homozygous wild-type individuals, and we suggest that allele-corrected cut-off levels should be used to improve the clinical performance of the Ca19 -9 tumor marker in the future.
In transfusion medicine, it has been reported that some individuals who type as Lewis-positive on erythrocytes can change their erythrocyte phenotype to Lewis-negative during diseases or during pregnancy (22,23). They have been named non-genuine Lewis-negative individuals as they have ␣1-4FucT activity in saliva (24). Due to this phenomenon, the Lewis-negative phenotype is more common among cancer patients (approximatley 20%) than among healthy individuals (approximately 8%).
One aim of this study has been to investigate the mechanism that leads to the non-genuine phenotype by examining the FUT3 alleles in a number of non-genuine and genuine Lewisnegative cancer patients and Lewis-positive healthy individuals. In the majority of cases, we found the change from Lewispositive to Lewis-negative to occur in FUT3 heterozygous individuals, and it is our hypothesis that FUT3 heterozygosity predisposes to the non-genuine Lewis-negative phenotype, due to the lower enzyme activity and lower production of Lewis structures.

EXPERIMENTAL PROCEDURES
Patients and Samples-Delipidated blood clots, urine deposits, and paraffin-embedded tissue samples from colonic tumors, as well as normal colon far from the tumor, were obtained from a previously published patient material (25). Saliva, peripheral leukocytes for DNA extraction, and plasma for Ca19 -9 measurements were obtained from 32 Lewis-negative and 20 Lewis-positive patients with either colon cancer or bladder cancer. 7 Lewis-negative and 59 Lewis-positive healthy volunteers were included as controls.
Preparation of Genomic DNA-DNA was extracted using a standard proteinase K-based method (26). DNA was extracted from paraffinembedded tissue samples using a boiling method. 20-m sections were deparaffinized by boiling for 5 min, proteinase K was added, and the samples were incubated overnight at 50°C. Nucleic acids were extracted once with Tris-buffered phenol and then five times with a mixture of phenol:chloroform:isoamyl alcohol (25:24:1) and precipitated with 3 M sodium acetate and 80% ethanol on ice.
Erythrocyte Phenotypes-Erythrocyte phenotypes (ABO, including A1-subgroups and Lewis) and secretor status were determined on fresh blood samples by standard agglutination methods in a specialized blood grouping laboratory at Aarhus University Hospital, Skejby.
FT Assays-The saliva was centrifuged at 12,000 ϫ g for 2 ϫ 10 min, and the supernatant was used for the assays. Supernatant protein content was determined with a BCA protein assay reagent (Pierce). COS cells were homogenized in 10 volumes of 50 mmol/liter HEPES buffer (pH 7.2), 1% Triton X-100 by three strokes of a Potter-Elvehjem homogenizer, followed by centrifugation and protein determination as described for saliva.
For transfected COS cells, a modification of a previously published (27) procedure with oligosaccharide acceptors was used: for the ␣1-4-L-fucosyltransferase assays, 20 l of the enzyme source was added to GDP-[ 14 C]fucose (0.25 nmol, 72,000 cpm) (Amersham Corp., Birkeroed, DK), MnCl 2 (1 mol), acceptor Lacto-N-biose I (0.5 mol) (Sigma), ATP (0.51 mol), Triton X-100 (500 g), and 0.1 M Tris-Hcl (pH 7.2), up to a total volume of 100 l. The reaction mixture was incubated at 37°C for 2 h. To assess the amount of radioactive fucose incorporated into product, the mixture was applied to a 2-ml Econo-column (Bio-Rad, Glostrup, DK)) packed with anion exchange resin, AG 4-X4 resin. The column was washed with water, and the radioactivity of the flowthrough fractions was counted. Control assays containing no added acceptor substrate were also completed in parallel. Radioactivity in the flow-through fraction, obtained in the absence of added acceptor, was no more than 2% of the total added radiolabel. All assays were performed at least twice.
DNA Sequencing-PCR was used to amplify the open reading frame (ORF) and the sequence flanking the start codon of FUT3. For amplifying ORF and the flanking region, the sequences of the sense and antisense oligonucleotide primers, the sizes of the product obtained, and the annealing temperature of the PCR reaction, respectively, were as follows: ORF, 5Ј-ATG GAT CCC CTG GGT GCA GCC AAG CCA CAAT-3Ј, 5Ј-GGC AGA TGA GGT TCC CGG CAG CCC AGC CAC-3Ј, 1125 bp, 72°C; flanking region, 5Ј-TCA GGA CTC ATG GCC CGG AGC TTT GGT AAG-3Ј, 5Ј-GGG AGT GGT GTC CTG TCG GGA GGA CCC ACT-3Ј, 264 bp, 70°C. The PCR products were purified with Wizard PCR preps DNA purification system (Promega, Madison, Wi) according to the manufacturer protocol. Nucleotide sequencing was performed on the entire ORF on 6 individuals, p115, p216, p215, p214, p138, and p246 (see Table II), by cycle sequencing using a direct blotting system (DBEsystem GATC 1500, MWG-Biotech, Munich, Germany), Taq DNA polymerase, and several digoxigenin-labeled oligonucleotides corresponding to internal sequences of the wild-type FUT3 gene.
Determination of Ca19 -9 -Serum and plasma concentrations of tumor-associated carbohydrate antigen Ca19 -9 was performed by a solidphase, two-site chemiluminescent enzyme immunometric assay for use with the Immulite automated analyzer according to the manufacturer instructions (Diagnostic Products Corp., Los Angeles, CA) The upper limit of normal serum level was defined as 37 units/ml.
Transfection of COS7 Cells-FUT3 ORF was amplified from the individuals p216, who was wild-type, and p214, who was homozygous mutated at 202/314, by the use of Expand high fidelity polymerase (Boehringer Mannheim). PCR was performed according to the manufacturer protocol with the same primers used for sequencing ORF and the annealing temperature at 68°C. The PCR fragments were ligated into a pCR II vector by the use of the TA cloning system from Invitrogen and sequenced to exclude PCR errors using the ABI 373A DNA sequencer (Perkin-Elmer) and the ABI PRISM dye terminator cycle sequencing ready reaction kit with AmpliTaq polymerase, FS (Perkin-Elmer).
The inserts were cut out with NaeI and EcoRI and blunt end ligated (26) into pMP6, an AAV-based expression vector, containing a CMV promotor (a generous gift from Dr. Ramila Philip, Applied Immunoscience, Santa Clara, CA). The resulting plasmids were transfected into COS7 cells using calcium phosphate-DNA co-precipitation (29).
Flow Cytometry Analysis-Transfected COS7 cells were harvested 65 h after transfection and labeled with mouse monoclonal anti-Lea (BioClone, Ortho Diagnostic Systems, Dortmund, FRG) and mouse monoclonal anti-H (Dakopatts, Copenhagen, Denmark). After incubation with primary antibody, the cells were washed and subsequently stained with fluorescein isothiocyanate-conjugated rabbit anti-mouse IgG (DAKO). Flow cytometry analysis was performed with a Becton Dickinson FACSort.
Northern Blot Analysis-Total RNA was isolated from transfected COS7 cells using Purescript RNA isolation kit (Gentra Systems Inc., Los Angeles, CA) according to the manufacturer protocol. Two to three g were electrophoresed in a denaturing formaldehyde gel, 2.2 M in running buffer (0.04 M MOPS, 0.01 M sodium acetate, 0.001 M EDTA) according to standard procedures (26), and transferred to a nylon membrane (Zeta-Probe; Bio-Rad). An 1125-bp DIG-labeled probe was amplified with the same primers and annealing temperature as for sequencing ORF using the PCR DIG probe synthesis kit (Boehringer Mannheim) according to the manufacturer protocol. Hybridization was carried out at 42°C for 16 -20 h. Blots were subsequently rinsed in 0.2 ϫ SSC, 0.1% SDS at 62°C. DIG-labeled probes were detected by an enzyme-linked immunoassay using the DIG nucleic acid detection kit (Boehringer Mannheim) according to the manufacturer protocol.

RESULTS
Lewis Gene Alleles in the Danish Population-Sequencing of Lewis-positive and Lewis-negative individuals in the FUT3 ORF was carried out on two PCR templates, one of 1125 bp covering the coding sequence and 39 bp into the 3Ј-flanking region and one of 264 bp covering the 5Ј-untranslated region and the first 174 bp of the ORF. With this approach, we identified three mutations, T202C, C314T, C445A, in Danish Lewis-negative Caucasians that differ from those previously shown to be frequent in Lewis-negative Asian people (17,18). Other mutations found were identical with those found in Asian people, T59G, G508A, and T1067A (Fig. 2, and Table II).
A set of PCR-cleavage assays that detected all 6 mutations was established by the use of natural restriction enzyme cleaving sites or by introducing restriction sites with mismatch primers (Table I and Fig. 3). In the group of genuine Lewisnegative individuals, several had four different mutations, two on each allele, and it seemed that 202/314 mutations were almost always coupled, similar to 59/508 and 59/1067 in previous publications (17,18).
To verify the co-localization of the 202/314 mutations, we made allele-specific PCRs with sense primers that either matched the T or the G at position 59 at the 3Јprime-end (Fig.  4). In order only to amplify the mutated or wild-type alleles at position 59, a mismatch was also introduced 3 bases from the 3Ј-end, as this led to a complete separation of T-59 and G-59.
The PCR data showed that the position 59 mutation never occurred on the same allele as the 202 and 314 mutations (Fig.  4), and the 202 and 314 mutations always occurred together in 38 individuals, indicating that they exist on the same allele. Cloning of an allele with both the 202 and 314 mutations supported this result (see below). Patient 123 was an exception, as 59 and 1067 occurred on an allele that also had the 202 mutation (see le6 in Fig. 2).
The combination of 59/1067 or 59/508 mutations in one allele leads to the production of an enzymatically inactive protein (17,18). The same seemed to be the case with the 202/314 mutations as 11 individuals, homozygous for both these mutations, were Lewis-negative on erythrocytes and had no saliva enzyme activity (Table II). To support this finding, an allele with the mutation couple of 202/314 as well as the wild-type allele were cloned and transfected into COS7 cells. The cloned genes were sequenced to verify the correct orientation and the fact that no PCR artifacts had been introduced.
COS7 cells transfected with the wild-type FUT3 gene (the enzymatically active allele Le in Fig. 2) expressed Lewis a structure, as evidenced by flow cytometry (Fig. 5), and showed ␣1-3/4-fucosyltransferase activity with a number of acceptors, as expected from the literature (Table III) (30). COS7 cells transfected with a FUT3 gene harboring the 202/314 mutation showed absence of both Lewis a structure and fucosyltransferase activity (Fig. 5 and Table III) although the transgene was actively transcribed, as evidenced by the presence of large amounts of mRNA in Northern hybidization with an FUT3 probe (Fig. 6).
An allele with the 59/202/1067 mutations was cloned from a Lewis-negative individual (patient 123) who was 202/314 on the other allele. Sequencing confirmed the existence of the mutations 59 and 202 on the same allele (data not shown). It is already known that the 1067 mutation is deleterious for the enzyme function (17). The C445A mutation was cloned from a non-genuine Lewis-negative cancer patient, and sequencing showed the existence of this mutation and the 59 mutation on the same allele (data not shown). The C445A mutation occurred in only one non-genuine Lewis-negative individual (patient 115) and was not found in cancer patients nor in the healthy controls examined. In one individual, patient 173, the 59 mutation was present on one allele and 202/314 on the other. This patient had saliva ␣1-4FucT activity, probably due  4 Cer. The reaction was conducted as described under "Experimental Procedures," and the labeled glycolipids were separated on high performance thin layer chromatography plates (Baker) using a solvent system composed of CHCl 3 : CH 3 OH:H 2 0, 60:40:9, prior to autoradiography with Kodak X-OMAT x-ray film. The identities of each band, which were confirmed in separate experiments by thin layer chromatography immunostaining with specific antibodies after transfer of unlabeled fucose to the indicated acceptors, is shown in the margins.
to the allele with the 59 mutation which is enzymatically active, and also demonstrated Lewis antigens on immunostaining of colorectal tissue (data not shown).
Based on our observations, we propose that there are at least seven different FUT3 alleles (see Le, le1, and le3-7 in Fig. 2) in Denmark, containing six single base mutations. These are located at positions 59, 202, 314, 445, 508, and 1067 (Fig. 2). The allele named le2 in Fig. 2 was first demonstrated in Indonesians (17) and has not been detected in Denmark.
Heterozygosity of the Lewis Gene Determines Levels of Lewis Enzyme Activity in Saliva and Sialyl-Lewis a Structure in Plasma-The saliva ␣1-4-fucosyltransferase activity was measured in 37 healthy Lewis-positive controls by the use of lactotetra-and lactoneotetraosylceramide glycolipid acceptors. The mobility on thin layer chromatography plates leads to immediate product identification (Fig. 1). Synthesis of ␣1-2, ␣1-3, and ␣1-4 products (H, Lex, Lea (28)) were all measured, as the ubiquitously present ␣1-3 activity served as control of enzyme preservation in the samples.
The activity of the ␣1-4-fucosyltransferase in saliva was correlated to the allelic status of the individuals (Fig. 7A), as the level in the heterozygous individuals was lower than that in the homozygous wild-type individuals. The level of sialyl-Lewis a structures circulating in plasma was measured by immuno-  assay and correlated to the allelic status of the Lewis and the secretor loci. Those who were heterozygous in the Lewis locus showed a significantly (p Ͻ 0.04) lower level than those who had two wild-type alleles (Fig. 7B). Based on the FUT3 genotype and the secretor status, individuals can be arranged in four groups showing a distinctly increased level of sialyl-Lewis a in serum from each group to the next. The FUT3 heterozygous secretors belong to group 1, the FUT3 wild-type secretors belong to group 2, the FUT3 heterozygous nonsecretors belong to group 3, and the FUT3 wild-type nonsecretors belong to group 4 (Fig. 7B). When using sialyl-Lewis a as tumor marker, these findings have the clinical impact that cut-off levels between normal healthy individuals and those suspected of having cancer should be defined on the basis of fucosyltransferase genotyping.
Identification of Individuals Who Lose Lewis Antigen Reactivity in Erythrocyte Membranes during Disease and Become "Non-genuine" Lewis-negative Individuals-The frequency of the different Lewis genotypes was estimated by applying a set of PCR assays (Fig. 3). In Lewis-positive individuals with saliva ␣1-4-fucosyltransferase activity and Lewis-positive erythrocytes, the frequency of heterozygous individuals was 40%, and the rest were homozygous wild type. Among genuine Lewisnegative individuals, without saliva enzyme activity and with Lewis-negative erythrocytes, some were homozygous for the 202/314 or 59/1067 mutations, and some were compound heterozygous with a combination of the 59/1067 and 202/314 alleles. In 30 Lewis-negative individuals with cancer, the saliva enzyme activity, plasma sialyl-Lewis a structures, and FUT3 genotype were determined (Table II). 10 of the 30 were identified as non-genuine Lewis-negative individuals.
The allele distribution in the non-genuine Lewis-negative group was significantly different from that found in the Lewispositive groups (Fig. 8). The number of heterozygous individuals with the mutations 59 and 202/314 was significantly increased (p Ͻ 0.02, Fisher exact test) in the group of nongenuine individuals, and also, the blood group A1 phenotype was significantly (p Ͻ 0.05, Fisher) more common in the nongenuine group. Two of the three non-genuine Lewis-negative individuals, who were not FUT3 heterozygous, were blood group A1.
Especially, the 59 mutation was over-represented in the nongenuine Lewis-negative cancer patients. Perhaps some dominant negative activity specific for alleles with this mutation is operating. Also, since Lea levels vary so widely within both the heterozygote and homozygote groups, it seems possible that more than one contributing factor is operating. Perhaps a threshold of FUT3 activity exists, below which individuals (heterozygous or not) tend to convert to Lewis-negative phenotype.
These data indicate that individuals who are FUT3 heterozygous, A1, or both are the ones most liable to develop the nongenuine phenotype in relation to a biological burden like cancer or pregnancy. Combined with the data above, this could very well be explained by the lower enzyme activity and the lower synthesis of Lewis structures in heterozygous individuals, which may, consequently, have a weaker phenotype with a lower density of Lewis epitopes in the erythrocyte membrane. The increased presence of A1 in the non-genuine group is not unexpected as it is well known that A1 individuals are more difficult to Lewis type than other ABO subgroups. DISCUSSION We have identified five common and one rare single base missense mutation within the coding region of the FUT3 gene in Denmark. Two of the common mutations, located at nucleotide position 59 and 1067, correspond to mutations frequent for Lewis-negative individuals in Indonesia. The 1067 mutation is in the catalytic domaine where it inactivates the enzyme, whereas the 59 mutation leads to a reduced enzymatic activity with type 1 as well as type 2 chain acceptors in vitro. But it has the same affinity (K m ) as the wild-type enzyme (17). However, COS cells transfected with the gene carrying the 59 FIG. 3. PCR-cleavage assays for detection of mutations in  FUT3. A, T59G mutation. FUT3-specific PCR products were generated with mismatch primers VE1mms and EL3as and digested with restriction enzyme MspI that cleaved the mutated fragment into two fragments of 24 and 116 bp. B, T202C mutation. PCR products generated with mismatch primers VE2mms and VE3as were digested with restriction enzyme RsaI that cleaved the mutated fragment into three fragments of 209, 100, and 23 bp. C, C314T mutation. PCR products generated with the primers EL3s and VE4as were cleaved with NlaIII that cleaved the mutated fragment into three fragments of 105, 34, and 65 bp. D, C445A mutation. PCR products were generated with primers VE5s and VE4as and digested with NlaIII that cleaved the mutated fragment into three fragments of 101, 36, and 25 bp. M, marker; N, no digestion.
FIG. 4. Allele-specific PCR followed by cleavage assays to determine the allelic distribution of the mutations T59G and T202C/C314T. A, AS-PCR products of different Le/le allele combinations following amplification with the primer specific for the normal allele at bp 59. B, corresponding AS-PCR products following amplification with the primer specific for the mutated allele at bp 59. Note the presence and absence of AS-PCR-products for Le/Le and le3/le3, respectively, in A and the opposite situation in B. C, cleavage of AS-PCR products from A with NlaIII. D, cleavage of AS-PCR products from B, also with NlaIII. Cleavage at an NlaIII site in the AS-PCR products from le5 alleles separates the 133-bp fragment into two fragments of 106 and 27 bp. Note, only AS-PCR products amplified with the primer specific for the normal allele at bp 59 yields cleavage with NlaIII. Numbers in parentheses indicate small fragments cleaved by the enzyme. M, marker; N, no digestion; O, negative PCR control reactions. mutation have no expression of Lewis structures. This might be explained by an inappropriate membrane insertion of the enzyme as the 59 mutation is located in the trans-membrane domain, and the transfected cells seem to secrete their enzyme, at least partly (17). The Japanese population has a 508 mutation in addition to the 59 and 1067 mutations (18,31). This mutation completely abolishes enzyme function when expressed in COS cells (18). The mutations 202 and 314 have been associated with the Lewis-negative phenotype (32,33) in Sweden, and this paper reports them to be frequent in Denmark and co-localized to the same allele. We examined the importance of these co-localized mutations for enzymatic activity by transfecting them into COS cells. No enzymatic activity could be found with types 1 and 2 chain acceptors, and the transfected cells lost the ability to express Lewis structures compared with the cells transfected with a wild-type allele. This could be explained by the localization of the mutation 314 since this particular mutation substitutes an amino acid that, as one out of eleven amino acid residues, is believed to be essential for the determination of types 1 and 2 chain substrate specificity (34).
Northern blots of the transfected COS cells showed large amounts of mRNA to be present in the cells transcribing the 202/314 mutated inserts. Based on this, it seems that the functional absence of fucosyltransferase activity must be caused by alterations occurring further downstream, such as interference with translation, folding, and trafficking of the mutant protein, or simply a properly inserted but non-functional protein.
On the basis of previously published data as well as data in

and Lewis genotype of all cancer patients and representative healthy controls
Pat is patient identification number. The pathologist scored the tumors as invasive, inv., or noninvasive, nonin. Eryt. indicates the erythrocyte blood group. Plasma sialyl-Lewis a structure was determined immunologically and expressed as units/ml. Saliva enzyme activity was measured as incorporated 14 C-fucose into glycolipid acceptors, expressed as pmol/h/ml of saliva. Mutations are presented as nucleotides at the two alleles at positions 59, 202, 314, 508, and 1067. If no base pairs are shown the nucleotides are wild-type. Individuals within each group are sorted according to mutations. ND, not determined. Sq, Verified by sequencing of ORF.   Fig. 1). b 59 was present on one allele and 202/314 on the other, as there was saliva enzyme activity and Le a and Le b antigens on tissue sections by immunostaining, the patient was included as non-genuine. this paper, it is possible to list those alleles known to be able to lead to alteration of FUT3 activity and altered Lewis phenotype (Fig. 2). The mutations 59 and 1067 seem to be widespread, all over the world. They frequently occur as single mutations in Indonesia, whereas in Japan and Europe, they are almost always co-localized. A similar finding is the colocalization of 59/508, which is found in both Japanese and European people but not in Indonesian.
In this paper, we demonstrate that the fucosyltransferase  activity in saliva is lower in heterozygous individuals than it is in homozygous wild-type individuals. As the enzyme is the primary gene product, the explanation for this is unambiguous and is in accord with data published on Indonesians (17). A statistical recalculation of the plasma ␣1-3-fucosyltransferase activity published (17) gives a mean value in six homozygous wild-type individuals of 7573 dpm and in 17 heterozygous individuals (all having a 59/1067 allele) of 3549 dpm, which is significantly different (p ϭ 0.04; Student's t test).
It is surprising that the difference in ␣1-4-fucosyltransferase activity level leads to a different level of circulating sialyl-Lewis a structure in plasma. This indicates that the assembly line of fucosylated structures in the Golgi is not oversaturated with fucosyltransferase activity but rather that the amount of fucosylated structures produced by a given cell is sensitive to just a two-fold increase in fucosyltransferase activity. Furthermore, these data suggest that the transcription and translation of the FUT3 gene is relatively stable among different individuals, and if not so, the gene dosage effect would have been completely obscured by individual regulations of transcription and translation. The variation we found among wildtype individuals agreed well with previous publications (17), and it seemed to be specific for each individual as repeated testing of individuals with either high or low enzyme activity showed the level to be highly reproducible (data not shown).
With the knowledge of the gene dosage effect in mind, we examined if this effect could explain the long standing mystery of erythrocyte membranes losing their Lewis antigens during pregnancy and during diseases such as cancer. Under circumstances such as those, we, and others, have identified nongenuine Lewis-negative individuals who change from Lewispositive to Lewis-negative on erythrocytes although they persistently express saliva Lewis enzyme activity (22-25, 35, 36). The reason for this change has been attributed to an increased level of circulating lipoproteins during the burden of disease or pregnancy, which alter the balance between production of Lewis glycolipids, transport in lipoproteins, and incorporation into erythrocyte membranes (23). In this paper, we are able to identify most of the non-genuine individuals as, first of all, heterozygous in the FUT3 gene and, due to the gene dosage effect, to be low producers of Lewis structures.
Another well known factor that influences the frequency of Lewis-negative individuals is the technical difficulty of Lewis phenotyping. It is difficult to obtain good and reliable anti-Lewis antibodies for Lewis blood grouping, especially in blood group A1 secretor individuals as almost all Lewis b structures are replaced by A Lewis b structures in secretors. The long repetitive type 2 chain A structure in A 1 individuals (37) may also structurally mask the shorter Lewis a structures. Our data on the lower level of circulating sialyl-Lewis a in FUT3 heterozygous individuals has implications for the use of this structure as a tumor marker, the Ca19 -9 antigen. It is very important to define a cut-off level for a tumor marker, above which cancer may be suspected. Our data clearly indicate that FUT3 genotyping and ␣1-2FucT genotyping/phenotyping could be important in this aspect as the two gene systems define four groups of individuals with different levels of sialyl-Lewis a in plasma. Those who are homozygous wild type for the FUT3 gene produce a larger amount of Lewis a than the heterozygous individuals; and if they are mutated in the secretor gene (nonsecretors), they do not fucosylate this Lewis a structure to H and the Lewis b, in competition with a sialyltransferase. Accordingly, they end up with a large amount of circulating sialyl-Lewis a. The second highest level of circulating sialyl-Lewis a is found among individuals who are heterozygous in the FUT3 gene, thus producing a comparatively lower amount of Lewis a. They are, however, also mutated in the secretor gene so no Lewis b is formed. Although they are homozygous wild-type in FUT3, the third group has a low amount of circulating sialyl-Lewis a. The amount of Lewis a produced is partly fucosylated to Lewis b as they are wild-type in the secretor gene. The fourth group, has a barely detectable amount of circulating Lewis a as it is heterozygous in FUT3, and the Lewis a formed is partly fucosylated to Lewis b.
In conclusion, we have identified four common and three rare FUT3 alleles in a Danish population. The relatively common allele with the 202/314 mutations was shown by expression studies to be enzymatically inactive. Furthermore, we have shown that the effect of heterozygosity for these FUT3 alleles is a relatively low ␣1-4-fucosyltransferase activity in saliva and a relatively low plasma level of the sialyl-Lewis a structure. The lower enzymatic activity in heterozygous individuals seemed to be causally involved, together with the A1 phenotype, in the conversion of genuine Lewis-positive individuals to non-genuine Lewis-negative individuals during disease periods or during pregnancy.