The E-selectin ligand-1 is selectively activated in Chinese hamster ovary cells by the alpha(1,3)-fucosyltransferases IV and VII.

The E-selectin ligand-1 (ESL-1) has recently been identified as the major ligand on mouse neutrophils using a recombinant antibody-like form of E-selectin as affinity probe. The remarkable selectivity with which ESL-1 can be affinity-isolated is unexplained. Since ESL-1 is endogenously expressed in Chinese hamster ovary (CHO) cells in a non-E-selectin binding form, which can become activated upon transfection of a fucosyltransferase (FucT), we analyzed various CHO cell clones, each overexpressing one of seven different fucosyltransferases, by affinity isolation experiments with E-selectin-IgG. Two of the cell lines were the regulatory CHO mutants LEC11 and LEC12, each overexpressing a different hamster FucT, while the five other clones were stably transfected with human FucTIII to -VII. A large panel of glycoproteins was affinity-isolated with E-selectin-IgG from LEC11 cells and FucTIII transfectants, demonstrating that many different glycoproteins can acquire ligand activity upon α(1,3)-fucosylation. In contrast, ESL-1 was almost exclusively isolated as the dominant glycoprotein ligand from LEC12 cells as well as from FucTIV and FucTVII transfectants and less selectively from FucTV and FucTVI transfectants. The selective generation of ligand activity correlated with the selective generation of the HECA452-reactive carbohydrate epitope, which is known to bind to E-selectin. These data suggest that, dependent on the type of fucosyltransferase, ESL-1 is a strongly preferred target molecule for the generation of E-selectin-binding carbohydrate modifications.

The entry of leukocytes into inflamed tissue is controlled by cell adhesion events between leukocytes and the endothelial cells of the blood vessel wall. The selectins initiate the first transient adhesive events in this process followed by leukocyte activation, firm adhesion to the apical endothelial cell surface, and finally transendothelial migration. Each of the three known selectins (L-, E-, and P-selectin) have been shown to mediate leukocyte rolling along the blood vessel wall and to be important for leukocyte extravasation in vivo (1)(2)(3)(4)(5)(6)(7). E-selectin is specifically expressed by cytokine-activated vascular endothelial cells and mediates the binding of neutrophils, monocytes, eosinophils, basophils, and a small subset of CD4 ϩ T-lymphocytes. It recognizes sialylated and fucosylated carbohydrate structures resembling the tetrasaccharide sialyl Lewis x (sLe x ) 1 (NeuAc␣2,3Gal␤1,4(Fuc␣1,3)-GlcNAc) or its stereo-isomer sialyl Lewis a (8). Another closely related carbohydrate epitope is defined by the antibody HECA452 and is probably involved in the binding of certain T-cell subsets to E-selectin (9).
The expression of these carbohydrate epitopes is dependent on the activity of ␣(1,3)-fucosyltransferases, which catalyze the last step in the synthesis pathway. Five human fucosyltransferases have been cloned in recent years (10 -14). Four of them, namely FucTIII, -V, -VI, and -VII can use ␣(2,3)-sialyl-Nacetyllactosamine to generate the sLe x moiety in vitro (15). Consequently transfection of each of these fucosyltransferases into mammalian cells gives rise to sLe x structures on the cell surface. In contrast, transfection with FucTIV generates different results, depending on the type of mammalian cells used (16 -20). In COS cells and in CHO cells derived from Pro Ϫ 5 parental strains no sLe x is generated, while in CHO cells derived from DHFR Ϫ parental strains FucTIV gives rise to sLe x on the cell surface (20). The generation of sLe x structures in different non-leukocyte, mammalian cell lines correlated with the binding of the transfected cells to E-selectin. This led to the early hypothesis that many different scaffold molecules could possibly serve as functionally relevant presenters of sLe x or sLe x -like structures that bind to E-selectin (21).
Despite these findings and despite the lectin character of Eselectin, very few distinct glycoprotein ligands could directly be affinity-isolated with E-selectin probes analyzing various leukocyte cell lysates. The E-selectin ligand-1 (ESL-1) was isolated by this approach as the major 150-kDa glycoprotein ligand on mouse neutrophils (22,23). This protein was found to be expressed on many different cell types, including CHO cells, but a glycoform of ESL-1 that was capable of binding to E-selectin was only found on myeloid cells. Modification of ESL-1 with sialic acid and fucose was found to be essential for ligand activity, and antibodies against ESL-1 could partially inhibit the binding of neutrophils to E-selectin in nonstatic (rotation) adhesion assays. The P-selectin glycoprotein ligand-1 (PSGL-1) which can be affinity-isolated with a P-selectin probe (24,25) can also be affinity-isolated with an E-selectin probe from human (26) as well as from mouse neutrophils (27). 2 Additional glycoprotein ligands that can be isolated with an E-selectin probe but have not yet been cloned are a 250-kDa glycoprotein (reduced 230 kDa) on mouse neutrophils (27) and a 250-kDa (reduced 280 kDa) glycoprotein on bovine ␥/␦ T-cells (28). Thus, very few distinct glycoproteins can be affinity-isolated from E-selectin-binding leukocytes. This raises the question of which mechanism determines this surprising ligand selectivity. One possibility would be that only these few ligands acquire certain carbohydrate modifications that enable them to bind to E-selectin. Alternatively, such ligands could carry additional sites, which could be necessary for high affinity interactions with E-selectin.
We have examined this question by analyzing CHO cell clones, each transfected with one of the five known human fucosyltransferases. While FucTIII expression generated a large panel of glycoprotein ligands that could be affinity-isolated with an E-selectin-Ig fusion protein, the other fucosyltransferases and most strikingly FucTIV and FucTVII selectively generated ESL-1 as the major ligand that could be isolated with an E-selectin-IgG affinity matrix. This selective activation of ESL-1 as an E-selectin binding ligand correlated with the selective expression of the HECA452-defined carbohydrate epitope on ESL-1.

EXPERIMENTAL PROCEDURES
Cell Culture-CHO cells DUKX B1, obtained from Dr. Pierre Vassalli (University of Geneva, Switzerland) and CHO mutants Lec11 and Lec12 (29 -32), obtained from Dr. Pamela Stanley (Albert Einstein College, New York) were grown in ␣-minimal essential medium supplemented with 10% fetal calf serum. Medium for transfected CHO cells was supplemented with 800 g/ml G418. HECA452 and CSLEX-1 producing hybridomas (American Type Culture Collection, Rockville, MD) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
E-selectin-Immunoglobulin Chimeric Protein-Construction of the mouse E-selectin-IgG chimeric protein has been described (38). The selectin part includes the signal sequence, the lectin domain, the EGFlike repeat, and the first two consensus repeats fused to the hinge region followed by the CH2 and CH3 domain of human IgG 1 .
Expression Plasmids-Full-length cDNAs for the human fucosyltransferases FucTIV and FucTVII (kindly provided by Dr. John Lowe, University of Michigan, Ann Arbor) and FucTIII, FucTV, and FucTVI (kindly provided by Dr. Mark Edbrooke, Glaxo, Greenford, UK) were cloned in sense orientation into pcDNA3 (InVitrogen), a plasmid carrying the gentamycin resistance gene.
Transfection and Stable Expression of Human Fucosyltransferases-CHO DUKX B1 cells were electroporated with pcDNA3 containing either no insert or the cDNA for one of the five different human fucosyltransferases. In brief, 1 ϫ 10 7 CHO cells resuspended in 600 l of phosphate-buffered saline containing 5 g of plasmid DNA in a 0.4-cm cuvette were electroporated at 950 microfarads and 0.25 kV. Transfected cells were selected in the presence of 800 g/ml of G418 (Sigma), and 10 individual cell clones of each transfectant were selected.
Flow Cytometry-CHO transfectants were analyzed for the expression of sLe x and the HECA452 epitope by flow cytometry as described (22). First, antibodies were detected with fluorescein isothiocyanatelabeled second stage antibodies from Dianova (Hamburg, Germany) Affinity Isolation Procedures-4 h after seeding, 3 ϫ 10 6 cells were washed twice with phosphate-buffered saline and labeled for 8 h with 286 Ci of Promix (200 Ci of [ 35 S]methionine and 86 Ci of [ 35 S]cysteine) in 1.5 ml of minimal essential medium without methionine and cysteine supplemented with 10% fetal calf serum dialyzed against phosphate-buffered saline. Labeled cells were lysed at a density of 2 ϫ 10 6 cells/ml in lysis buffer (1% Triton X-100, 20 mM Tris-HCl (pH 8.4), 160 mM NaCl, 1 mM CaCl 2 , 1 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride, 0.1 unit/ml ␣ 2 -macroglobulin) for 12 min on ice. Insoluble material was pelleted at 10,000 ϫ g for 15 min. Lysates were preincubated for 30 min with 50 l of packed protein A-Sepharose to remove unspecifically binding proteins. Beads were pelleted, and aliquots of supernatant (corresponding to 2 ϫ 10 6 cells for precipitation with Eselectin-IgG and corresponding to 5 ϫ 10 5 cells for precipitation with affinity-purified antibodies) were incubated with 20 l of protein A-Sepharose preincubated for 3 h with 5 g of human IgG 1 , E-selectin-IgG chimeric protein, total rabbit IgG, or affinity-purified anti ESL-1 antibodies. After incubation for 4 h at 4°C, the resin was washed five times with washing buffer 1 (0.05% Triton X-100, 50 mM Tris-HCl (pH 8.4), 400 mM NaCl, 1 mM CaCl 2 , 1 mg/ml ovalbumin) and once with washing buffer 2 (0.05% Triton X-100, 50 mM Tris-HCl (pH 8.4), 150 mM NaCl, 1 mg/ml ovalbumin). Proteins specifically bound to E-selectin-IgG were eluted twice with 20 l of elution buffer (3 mM EDTA, 50 mM ammonium acetate (pH 7.0), 0.05% Triton X-100). Antibody-bearing resin was eluted with 60 l of SDS-PAGE sample buffer containing 50 mM dithiothreitol. After SDS-PAGE, proteins were detected by fluorography using Kodak Biomax x-ray films (proteins derived from metabolically labeled cells) or were blotted on nitrocellulose for Western blot analysis (proteins from unlabeled cells).
For reprecipitation experiments of EDTA-eluted material from the E-selectin-IgG affinity matrix (see above), eluted aliquots were diluted into 0.5 ml of lysis buffer supplemented with 4 mM CaCl 2 and 20 g/ml bovine serum albumin and subjected to immunoprecipitations with affinity-purified ESL-1 rabbit antibodies. Precipitated material was eluted as described above.
Western Blot Analysis-Cells were lysed at a density of 2 ϫ 10 8 cells/ml in lysis buffer (see above), and insoluble material was pelleted at 10,000 ϫ g for 15 min. Cell extracts corresponding to 5 ϫ 10 6 cells/lane were electrophoresed and transferred to nitrocellulose filters (Schleicher & Schuell). The membranes were incubated for 2 h in blocking buffer (Tris-buffered saline, pH 7.4, containing 4% low fat milk powder) and probed with 10 g/ml affinity-purified anti-ESL-1 rabbit antibody in blocking buffer or with hybridoma supernatant of mAb HECA452 or mAb CSLEX-1. After washing with blocking buffer, the blot was probed with horseradish peroxidase-conjugated anti-rabbit IgG, anti-rat IgM, or anti-mouse IgM in blocking buffer, washed, and visualized by chemoluminescence using the ECL kit from Amersham (Braunschweig, Germany).
Fucosyltransferase Assays-Cells were extracted at a density of 2 ϫ 10 8 cells/ml in lysis buffer (see above), and protein concentration was determined by the BCA assay (Pierce, Köln, Germany). 10 g of protein extract were assayed for fucosyltransferase activity in a reaction volume of 40 l containing 25 mM cacodylate (pH 6.2), 0.25% Triton X-100, 10 mM MnCl 2 , 5 mM GDP-fucose, 0.07 Ci of GDP-[ 3 H]fucose, and 10 mM N-acetyllactosamine (NAL) or 4 mM 3Ј-sialyllactosamine (3Ј-SLN) as acceptor oligosaccharide (purchased from Dextra Laboratories). Added acceptor substrate was omitted from the controls to allow for the possible fucosylation of endogenous substrates. After incubation at 37°C for 2 h, 1 ml of a Dowex 1-X8 slurry (1:4 (w/v) in water) was added to the reaction and vortexed. 500 l of the supernatant was counted in 5 ml of scintillant (Ultima Gold XR, Packard). To obtain values solely due to fucosylation of acceptor substrate, total counts of the control (without acceptor substrate) were subtracted from total counts of samples with acceptor. The activity of the fucosyltransferase containing cell extract was calculated as pmol/min/mg.

Different Patterns of Glycoprotein Ligands for E-selectin in the CHO Mutant Cell Lines LEC11 and LEC12-It has been
shown previously that the selectin ligands ESL-1 and PSGL-1 both need to be fucosylated for ligand activity (23,24). Transfection of their cDNAs into CHO cells gives rise to selectin binding ligands only if they are co-transfected with fucosyltransferase III. We wanted to examine whether the two CHO glycosylation mutants LEC11 and LEC12, which carry regulatory mutations causing up-regulation of two distinct ␣(1,3)fucosyltransferases (29 -32), would express glycoprotein ligands that could be affinity-isolated with an E-selectin-Ig fusion protein. To this end, cells were metabolically labeled with [ 35 S]methionine/[ 35 S]cysteine, detergent extracts were incubated with E-selectin-IgG affinity matrix, and specifically bound proteins were eluted with EDTA.
Surprisingly, the repertoire of specifically isolated glycoprotein ligands was very different for the two different mutant cell lines. While a large number of proteins was isolated from LEC11 cells, only one major glycoprotein ligand of 150 kDa and two more weakly detectable species at around 220 and 280 kDa were specifically purified from LEC12 cells (Fig. 1A). Identical numbers of cells were analyzed. As expected, no proteins could be isolated from normal CHO cells (Fig. 1A). A similar broad range of glycoprotein ligands as in LEC11 cells was also found when we analyzed various colon carcinoma cell lines. 2 In contrast, the result with LEC12 cells resembled that with mouse neutrophils, from which we have previously isolated ESL-1 as the major 150-kDa ligand besides another specific ligand at 250 kDa (22) and two more weakly detectable bands at 230/130 kDa, which we have identified as mouse PSGL-1 (27). 2 Indeed, the major 150-kDa ligand that was isolated from LEC12 cells with E-selectin-IgG could be reprecipitated with affinity-purified antibodies against mouse ESL-1 (Fig. 1B).
These results suggest that, dependent on the type of upregulated ␣(1,3)-fucosyltransferase, different repertoires of high affinity glycoprotein ligands are generated. We have pre-viously shown that CHO cells express the hamster homolog of the ESL-1 protein, although not as an E-selectin-binding glycoform (23). The results with LEC12 cells suggest that among all cellular proteins of this cell line, ESL-1 seems to be a preferred target molecule for post-translational modifications, which transform it into a high affinity ligand for E-selectin.
Affinity Isolation Experiments with Fucosyltransferasetransfected CHO Cells-Since the two hamster fucosyltransferases in the two mutant cell lines have not yet been cloned and the molecular details of the mutations have not yet been fully revealed, we repeated the affinity isolation experiments as described above, using different CHO cell clones stably transfected with one of the five known human fucosyltransferases, numbered from III to VII. Cells were transfected with expression plasmids carrying the DNA sequence for the respective fucosyltransferase together with the neomycin resistance gene as selection marker. Clones resistant to G418 were analyzed in affinity isolation experiments with E-selectin-IgG as affinity probe. More than 80% of the G418-resistant clones were able to produce E-selectin-IgG precipitable glycoproteins. As shown in Fig. 2A, a large panel of glycoproteins was affinityisolated with E-selectin-IgG from FucTIII transfected CHO cells very similar to the result with LEC11 cells (Fig. 1A). In contrast, a major ligand of 150 kDa was precipitated with E-selectin-IgG from FucTIV-and FucTVII-transfected cells, just as observed with LEC12 cells. This ligand, precipitated by E-selectin-IgG from FucTVII-transfected cells could be reprecipitated with affinity-purified antibodies against mouse ESL-1 (Fig. 2B). In addition to ESL-1, two minor higher molecular mass species of 220 and 280 kDa were found as well as a 120-kDa protein.
Affinity isolation analysis of FucTV-and FucTVI-transfected CHO cell clones revealed comparable results, although the selectivity with which the 150-kDa ligand was precipitated by E-selectin-IgG was not as distinctive as that observed with FucTIV-and FucTVII-transfected cells (Fig. 2A). The affinity isolation results depicted in Fig. 2 for each fucosyltransferase transfection are representative in each case for four or five independent clones.
The expression levels of ESL-1 in the various CHO mutants and CHO transfectants were examined by immunoprecipitations with anti-ESL-1 antibodies analyzing equal amounts of metabolically labeled cells. As shown in Fig. 3, the amounts of precipitated ESL-1 were similar for all tested clones.
Expression Levels of Fucosyltransferase Activity Are Not Responsible for the Generation of Different Patterns of Glycoprotein Ligands-Since the expression of FucTIII led to the synthesis of numerous E-selectin binding glycoproteins while FucTIV and FucTVII generated almost exclusively ESL-1 as a ligand for E-selectin, we examined whether these differences would be caused by different expression levels of the enzymes. Detection of the enzymes in immunoblots using specific rabbit antisera (obtained from John Lowe) was unsuccessful for each of the stably transfected CHO clones. This was most likely due to insufficient sensitivity, since the enzymes were only detectable when they were transiently overexpressed in COS cells (data not shown). As an alternative method, we analyzed the expression levels of the fucosyltransferases by in vitro fucosyltransferase enzyme assays using detergent extracts of the transfected cells. Extract aliquots of identical protein content were analyzed, revealing comparable levels of enzyme activity for each enzyme (Fig. 4). In agreement with published data (13,39), FucTIII containing cell lysates could utilize NAL as well as 3Ј-SLN as acceptor substrate, while FucTIV samples showed a clear preference for the nonsialylated neutral acceptor NAL (the precursor of Le x ) and FucTVII preferentially fucosylated ]cysteine-labeled CHO cells, and LEC11 and LEC12 mutants were incubated with protein A-Sepharose loaded with human IgG (IgG) or E-selectin-IgG (E-Sel-IgG). Beads were washed, and specifically bound proteins were eluted with EDTA, electrophoresed on a 6% polyacrylamide gel under reducing conditions, and detected by fluorography. B, LEC12 cells were labeled as in A, and detergent extracts were incubated with protein A-Sepharose loaded with human IgG (IgG) or E-selectin-IgG (E-Sel.-IgG). Proteins specifically eluted with EDTA from the E-selectin-IgG matrix were directly electrophoresed, or aliquots were subjected to reprecipitations either with IgG of a rabbit nonimmune serum (re. control) or with affinity-purified IgG of the rabbit anti-ESL-1 serum 65 (re. ␣-ESL-1). Specifically bound proteins were eluted with SDS-PAGE loading buffer and electrophoresed and visualized as in A. Molecular mass markers (in kDa) are indicated on the left. the acidic acceptor 3Ј-SLN (the precursor of sLe x ). Based on these measurements, the FucTIII-transfected cells did not express higher specific activities for the transfer of fucose to NAL or to 3Ј-SLN than the FucTIV or FucTVII transfectants, respectively. The results depicted in Fig. 4 were obtained with the same clones that were analyzed by affinity isolation with Eselectin-IgG (Fig. 2). Thus, differences in the expression level of the various fucosyltransferases, as far as they can be determined in in vitro enzyme assays, are unlikely to be the reason for the different patterns of E-selectin-binding glycoproteins in the various transfectants. This was also verified by analyzing different CHO clones that were transfected with the same fucosyltransferase but expressed different levels of enzyme activity. As determined by in vitro enzyme assays, FucTIV-transfected CHO clone 9 expressed 6 times higher levels of FucTIV than clone 7. However, as illustrated in Fig. 5, there was no difference in the glycoprotein pattern that could be isolated from both clones with E-selectin-IgG. In both clones, ESL-1 was the major isolated ligand. The only observed difference was the increased overall level of isolated material in clone 9 versus clone 7.
ESL-1 Is a Major Carrier for HECA452 Epitopes in FucTIVor FucTVII-expressing CHO Cells-We wanted to examine whether the surprising selectivity with which ESL-1 was generated as an E-selectin ligand in FucTIV and FucTVII expressing CHO cells was indeed based on the selective generation of certain carbohydrate modifications on this protein. The monoclonal antibody HECA452 was described as recognizing carbohydrate epitopes that can be recognized by E-selectin and that are related to but not identical with sLe x (8,(33)(34)(35)(36). All FucTtransfected CHO clones were positive for HECA452 as well as for the anti-sLe x antibody CSLEX-1, as was analyzed by flow cytometry (Fig. 6). Probing immunoprecipitates of ESL-1 from CHO cells transfected with FucTIII, FucTIV, or FucTVII in immunoblots with the HECA452 antibody revealed that ESL-1 was positive for the HECA452 carbohydrate epitope in each of

FIG. 2. Affinity isolation of E-selectin-binding proteins from fucosyltransferase-transfected CHO cells with E-selectin-IgG.
A, equal numbers of CHO cells transfected with FucTIII (CHO-III), FucTVII (CHO-VII), FucTVI (CHO-VI), FucTV (CHO-V) and FucTIV (CHO-IV) were labeled as in Fig. 1, and detergent extracts were incubated with protein A-Sepharose loaded either with human IgG (IgG) or with E-selectin-IgG (E-Sel-IgG). Beads were washed, and specifically bound proteins were eluted with EDTA, electrophoresed on a 6% polyacrylamide gel under reducing conditions, and detected by fluorography. Samples of the FucTIV-transfected cells were electrophoresed on a separate gel. For each transfection, four to five independent clones were analyzed, giving similar results as the one which is shown. B, FucTVIItransfected CHO cells were subjected to affinity isolation with human IgG (IgG) or E-selectin-IgG (E-Sel.-IgG), and EDTA-eluted proteins from the E-selectin-IgG matrix were either directly electrophoresed or subjected to reprecipitations either with IgG of a rabbit nonimmune serum (re. control) or with affinity-purified IgG of the rabbit anti-ESL-1 serum 65 (re. ␣-ESL-1). Specifically bound proteins were eluted with SDS-PAGE loading buffer, electrophoresed, and visualized as in A. Molecular mass markers (in kDa) are indicated on the left.  Fig. 2. For each transfection, three independent clones were analyzed giving similar results as the one depicted here. Data shown correspond to fucosyltransferase activities (expressed as pmol/min/mg) measured with samples containing acceptor substrate minus activities measured without acceptor substrate and represent the mean and standard deviation from three replicate assays. the transfectants, while ESL-1 in mock-transfected CHO cells was negative (Fig. 7). Immunoblot analysis with CSLEX-1 did not generate conclusive results, since signals were almost undetectable even in cell extracts of the sLe x -expressing human monocytic cell line HL60 (not shown).
When total cell extracts of the same CHO clones were analyzed in immunoblots, FucTIII transfectants were found to express a large panel of glycoproteins that were positive for HECA452 (Fig. 8). In contrast, in FucTIV-and FucTVII-expressing CHO cells, a 150-kDa glycoprotein was detected as the major carrier for HECA452-reactive epitopes (Fig. 8).
In order to examine whether this 150-kDa protein would be reactive with anti ESL-1 antibodies, we depleted cell extracts of the FucTIII and FucTVII transfectants for the ESL-1 glycoprotein by two rounds of incubations with protein A-Sepharose beads bearing anti ESL-1 antibodies. Mock depletions were performed with nonimmune IgG. Depleted cell extracts were analyzed in immunoblots with HECA452 antibody. As shown in Fig. 9, the 150-kDa glycoprotein reactive for HECA452 was specifically removed by anti-ESL-1 antibodies. In FucTIII transfectants, only this 150-kDa protein and no other HECA452-reactive glycoprotein was removed by anti-ESL-1 antibodies (Fig. 9). We conclude that it is ESL-1 that is selec-tively decorated with HECA452 epitopes in FucTVII-transfected CHO cells. DISCUSSION In this study we show that different ␣(1,3)-fucosyltransferases, when expressed in CHO cells, generate different repertoires of E-selectin-binding glycoprotein ligands. While Fuc-TIII generates a large panel of glycoprotein ligands, among which the hamster equivalent of mouse ESL-1 is just one, FucTIV and FucTVII almost exclusively modify ESL-1 in a way that allows it to bind to E-selectin. The same selectivity was seen for the generation of the carbohydrate epitope HECA452. These data demonstrate that FucTIV and FucTVII, known to be expressed in myeloid cells, are indeed very selective in choosing acceptor glycoproteins that they can transform into E-selectin ligands.
Our analysis of FucTIII-expressing CHO cells shows that simply expressing this enzyme is sufficient to transform numerous glycoproteins into E-selectin "ligands," which bind with sufficient strength to allow affinity isolation. This suggests that ␣(1,3)-fucosylation of carbohydrate side chains that are found on many different proteins is sufficient to generate high affinity recognition epitopes for E-selectin. Although possible, it is unlikely that all of these proteins share additional structural elements that are necessary for E-selectin binding. This would argue for the hypothesis that high affinity binding does not necessarily require additional structural elements besides certain carbohydrate components. This is in agreement with the work of Patel et al. (40), who showed that certain tetraantennary carbohydrate compounds that contained a di-sLe x structure on one branch could be affinity-isolated by an Eselectin affinity matrix out of a mixture of all carbohydrate side chains that had been released by hydrazinolysis from plasma membrane glycoproteins of human myeloid cells. These results indicate that carbohydrate moieties can bind with high affinity to E-selectin in the absence of any other structural element.
Expression of each of the five fucosyltransferases in CHO cells gave rise to the expression of CSLEX-1-and HECA452reactive epitopes on the cell surface. The FucTIV was originally characterized as a fucosyltransferase that generates E-selectin ligand activity in transfected cells and was therefore named ELFT for E-selectin ligand fucosyl transferase (16). However, it was found later that FucTIV-transfected COS and CHO cells did not bind to E-selectin and did not display CSLEX-1-reactive epitopes on the cell surface and that detergent extracts of such transfectants were not able to metabolize sialylated type II acceptor carbohydrate structures (17,18). These contradictory results were then partly explained when Goelz et al. (20) found that FucTIV can generate CSLEX-1-reactive epitopes only in DHFR Ϫ CHO DUKX B1 cells and not in CHO cell clones derived from Pro Ϫ 5 parental strains. Since we have used CHO FIG. 5. Different levels of in vitro fucosyltransferase activity do not alter the substrate specificity of transfected fucosyltransferase. Detergent extracts of two FucTIV-transfected CHO cell clones (CHO-IV, clone 7 and clone 9) were analyzed and compared for their in vitro fucosyltransferase activity (FT-activity) as well as for their content of E-selectin IgG-precipitable glycoproteins. A value of 100% was assigned to the fucosyltransferase activity that was measured for clone 9, using the acceptor NAL. E-selectin-binding proteins isolated from identical numbers of cells were electrophoresed on a 6% polyacrylamide gel and visualized by fluorometry. Note that the relative enzyme activities of clones 7 and 9 correlated with the amount of precipitable proteins from the respective cell lines but did not influence the pattern of E-selectin-binding glycoproteins. DUKX B1 cells, our data are in agreement with this latter report. Despite the FucTIV-driven expression of sLe x on the surface of these cells, we, like them, found that detergent extracts of FucTIV transfectant CHO DUKX B1 were unable to fucosylate sialylated type II acceptor molecules in enzyme assays. The molecular basis for these discrepancies between the two subclones of CHO cells cannot yet be explained. It may be possible that DHFR Ϫ CHO cells contain a sialyltransferase that is able to modify ␣(1,3)-fucosylated lactosamines, although such an enzyme has not yet been described.
The selectivity with which FucTIV and FucTVII almost exclusively transform the ESL-1 protein backbone into an Eselectin binding form while almost all other proteins do not acquire ligand activity is surprising. This raises the question of what structural element on ESL-1 is recognized by E-selectin. If the recognition motif is exclusively based on carbohydrate structures, either FucTIV and FucTVII or any other glycosyl transferase that acts prior to these enzymes would have to exclusively modify ESL-1 and not the majority of all other cellular proteins. Alternatively, ESL-1 could carry additional structural elements (within its protein backbone or as further post-translational modifications) that would be lacking on other proteins and that would be directly involved in E-selectin binding. While this latter possibility cannot be ruled out, our results obtained with the mAb HECA452 demonstrate that ESL-1 indeed acquires selectively carbohydrate modifications that are only found on ESL-1 and on very few other glycoproteins, and not on the majority of all other cellular glycoproteins. This suggests that ESL-1 is indeed selectively and specifically glycosylated. Since the HECA452 antibody can block the binding of certain T-cell populations to E-selectin (9), it is conceivable that the HECA452 carbohydrate epitope that is selectively generated on ESL-1 is involved in the binding to E-selectin.
What determines the selectivity with which FucTIV or FucT-VII generate the HECA452 carbohydrate epitope on ESL-1 but not on most other glycoproteins although many of them are able to acquire this epitope upon expression of FucTIII? Two possibilities exist. The first is that FucTIV and FucTVII could fucosylate many different cellular proteins but only ESL-1 could carry acceptor structures that allow the generation of the HECA452 epitope. Indeed, FucTIV and FucTVII only accept type II substrates where they fucosylate GlcNAc in the ␣(1-3)position, while FucTIII accepts type II and type I substrates, i.e. can add fucose in position 3 as well as in position 4. The selective generation of HECA452 epitopes could thus be based on the selective appearance of type II acceptor-like carbohydrate modifications on ESL-1. However, such a selective expression pattern of type II acceptor structures on only very few proteins has not been described. Alternatively, it could be Fuc-TIV and FucTVII that selectively modify ESL-1 and very few other glycoproteins. This selective interaction could be based on structural elements on ESL-1 that are favorable for the interaction with FucTIV or FucTVII, or it could be based on the specific distribution of fucosyltransferases in subdomains of the Golgi membranes. FucTIII might be more broadly distributed, while FucTIV and FucTVII could be preferentially located in subdomains of the Golgi membranes that are passed by only a few glycoproteins and preferentially by ESL-1. Thus the specificity of fucosylation would be based on a "targeting" mechanism. Such a hypothesis is testable now.
Like ESL-1, PSGL-1 can be selectively precipitated with E-selectin-Ig from detergent extracts of myeloid cells (24,27). In analogy to our results for ESL-1 it might be possible that this ligand is also selectively glycosylated. Indeed, Wilkins et FIG. 7. Immunoprecipitated ESL-1 of fucosyltransferasetransfected CHO cells is recognized by mAb HECA452 in immunoblots. ESL-1 was immunoprecipitated from mock-transfected CHO cells (CHO) or CHO cells transfected with FucTIII (CHO-III), FucTIV (CHO-IV), or FucTVII (CHO-VII), using affinity-purified antibodies from the anti-ESL-1 rabbit antiserum 65, electrophoresed on 6% polyacrylamide gels, transferred onto nitrocellulose filters, and analyzed in immunoblots either with mAb HECA452 or with affinity-purified antibodies against ESL-1 (␣-ESL-1) from antiserum 89060, as indicated. The additional band at 120 kDa, which reacts with anti-ESL-1 antibodies, most likely is a breakdown product of ESL-1. Molecular mass markers (in kDa) are indicated on the left. al. (41) have shown recently that PSGL-1 from HL60 cells carries two species of fucosylated O-glycans that were not found on the sialomucin CD43 purified from the same cells.
In summary, our data show that the remarkable selectivity with which E-selectin binds to ESL-1 from mouse neutrophils can be mimicked in CHO cells transfected with fucosyltransferases from myeloid cells. While our data do not rule out that ESL-1, besides specific carbohydrate modifications, might also carry other structural elements that are directly involved in the recognition by E-selectin, they suggest that ESL-1 is a selective target for the generation of carbohydrate structures that are necessary for the binding to E-selectin.