Expression of α-1,3-Galactose and Other Type 2 Oligosaccharide Structures in a Porcine Endothelial Cell Line Transfected with Human α-1,2-Fucosyltransferase cDNA*

The binding of xenoreactive natural antibodies to the Galα1–3Galβ1–4GlcNAc (α-galactose) oligosaccharide epitope on pig cells activates the recipient’s complement system in pig to primate xenotransplantation. Expression of human α-1,2-fucosyltransferase in pigs has been proposed as a strategy for reducing the expression level of the α-galactose epitope, thereby rendering the pig organs more suitable for transplantation into humans. The aim of this study was to examine how the cell surface expression of α-galactose, H, and related fucosylated and sialylated structures on a pig liver endothelial cell line is affected by transfection of human α-1,2-fucosyltransferase cDNA. Nontransfected and mock-transfected cells expressed α-galactose, α-2,3-sialylated, and α-2,6-sialylated epitopes strongly, with low level expression of type 2 H and LewisX. By contrast, expression of the H epitope was increased 5–8-fold in transfected cells with a 40% reduction in the expression of α-galactose epitope and a 50% decrease in sialylation, as measured by binding of Maackia amurensis and Sambuccus nigra agglutinins. LewisX expression was reduced to background levels, while the LewisY neoepitope was induced in human α-1,2-fucosyltransferase-expressing pig cells. The activities of endogenous α-1,3-galactosyltransferase, α-1,3-fucosyltransferases, and α-2,3- and α-2,6-sialyltransferases acting on lactosamine were unaffected. Our results show that a reduction in α-galactose epitope expression in porcine endothelial cells transfected with human α-1,2-fucosyltransferase cDNA may be achieved but at the expense of considerable distortion of the overall cell surface glycosylation profile, including the appearance of carbohydrate epitopes that are absent from the parent cells.

Hyperacute rejection of an organ in pig to primate xenotransplantation is triggered by the binding of the recipient's natural preformed antibodies to the antigens expressed on the endothelium of the graft (1). Within minutes, this results in the activation of endothelial cells, fixation of complement (2,3), and blood coagulation (4). The main epitope for the human natural preformed IgM and IgG antibodies on the pig endothelium has been identified as a linear type 2 oligosaccharide Gal␣1-3Gal␤1-4GlcNAc1-R (␣-Gal) 1 (5)(6)(7)(8). This ␣-Gal epitope is synthesized by UDP-Gal:Gal␤1-4GlcNac-R (Gal to Gal) ␣-1,3-galactosyltransferase (EC 2.4.1.124/151), an enzyme that is expressed in all mammals except Old World primates (9 -12). In pigs, the ␣-Gal epitope is expressed in a variety of tissues including endothelium (13), where it can be present on Nlinked (14) and O-linked oligosaccharides (15) as well as on glycolipids (16).
Several human complement control proteins, such as CD46, CD54, and CD59, have been found to protect porcine cells against human complement in vitro (17)(18)(19)(20) and in vivo (7,(21)(22)(23). Nonetheless, inhibition of the expression of the ␣-Gal epitope is a highly desirable goal in that it would greatly reduce preformed human natural antibody binding and consequently reduce the risk of endothelial cell activation (2,24,25) and antibody-dependent cell-mediated cytotoxicity (26,27). Given that genetic knockout is not currently possible in pigs, due to the absence of suitable embryonic stem cells, several strategies have been considered for preventing the binding of human preformed IgM antibodies to the ␣-Gal epitopes expressed on pig endothelial cells. Besides IgM antibody depletion (28 -30), oligosaccharides such as Gal␣1-3Gal␤, have been found to block the binding of human and baboon preformed natural xenoreactive antibodies to the pig kidney cell line PK15 in vitro (31) and to delay hyperacute rejection in pig to baboon cardiac transplantation (32).
To investigate the consequences of human ␣-1,2FT expression on the overall glycosylation profile of porcine cells, we have transfected a pig liver endothelial cell line (PLECT) with a cDNA encoding human ␣-1,2FT. We studied cell surface expression of H, ␣-Gal, and several fucosylated and sialylated oligosaccharide epitopes.
Cell Lines, Transfection, and Selection-The pig liver cell line used in the expression studies was established from the liver of a male inbred blood group H SLAb/b pig developed at the Babraham Institute (Cambridge, UK). Briefly, 2 g of freshly collected tissue were minced with a scalpel and then pushed gently through 250-m stainless steel mesh. The disrupted tissue was collected and washed four times with PBS containing 10 mM glucose, 100 units/ml penicillin, and 100 units/ml streptomycin (Sigma). Washed cells were digested with 0.1% collagenase type V (Sigma) in PBS/glucose/penicillin/streptomycin for 1 h at 37°C. Finally, the sieved and collagenase-digested cells were fractionated using Dolichos biflorus agglutinin-coated Microcellactor flasks (Applied Immune Services, Inc.), since D. biflorus agglutinin is known to be an endothelial cell-specific marker in mice (38 -41) and pigs (42,43). Purified cells were maintained in EC-SFM (Life Technologies, Inc.) containing 5% heat-inactivated (30 min at 56°C) FCS (GlobePharm, UK) and penicillin/streptomycin as above. Immortalized cell lines were established by transfection with pSV3neo using Lipofectin (Life Technologies, Inc.). Selection for transformants was performed by adding G418 (Boehringer Mannheim) at 200 g/ml to the growth medium. Cells were plated into 96-well plates (Nunc, Denmark) at 200 cells/well 72 h after transfection to isolate individual clones. Transformed cells were found to take up 3,3Ј-dioctadecylindocarbocyanine-labeled low density lipoprotein (Biogenesis), bound FITC-conjugated D. biflorus agglutinin, and secreted factor VIIIa-like activity.
Transformation of PLECT cells with ␣-1,2FT cDNA was carried out using the calcium phosphate precipitation method of Chen and Okayama (44). Transformed cells were plated out 72 h after transfection into 96-well plates (Nunc, Denmark) at 200 cells/well and selected in the presence of 200 g/ml hygromycin (Boehringer Mannheim). Clones were screened for the expression of blood group H antigen using flow cytometry. Plasmid DNA for mammalian cell transfections was purified using Qiagen Midi columns.
Antibodies, Immunoglobulins, and Lectins-Anti-blood group A, B, and H mAbs were from Dako (Denmark), anti-Lewis X (CD15) was from Sigma, and anti-Lewis Y was from Serotec. D. biflorus agglutinin, FITC-DBA, and FITC-Bandeireia simplicifolia isolectin B 4 (I/B 4 ) lectins were from Vector Laboratories. Reagent grade human IgM was from Sigma.
Second-stage Reagents-FITC-conjugated sheep anti-mouse IgG F(abЈ) 2 fragment was from Sigma. FITC-conjugated goat anti-human IgM and goat anti-mouse IgG were from Vector Laboratories, UK.
Oligosaccharides were from Dextra Laboratories, except sucrose, which was from Sigma.
Flow Cytometry-Cells were detached from the tissue culture flasks by brief treatment with 5 mM EDTA in PBS, centrifuged for 5 min at 256 ϫ g in a Heraeus 2.0R Megafuge in the presence of 10% FCS, and resuspended at 10 6 cells/ml in cold (4°C) PBS, 0.1% bovine serum albumin, 0.1% NaN 3 (Sigma, UK). 50 l of cells in triplets or quadru-plets were incubated in U-well Falcon microtiter plates with the primary antibody added at 10 g/ml, or at the dilutions recommended by the manufacturer, for 1 h with shaking at 4°C. Cells were washed three times with cold PBS/FCS/NaN 3 before being resuspended in the same buffer containing second-stage reagents at the dilutions recommended by the manufacturer. Following a 1-h incubation with shaking at 4°C in the dark, the samples were again washed three times with cold PBS/ FCS/NaN 3 and resuspended in 50 l of ice-cold PBS, 1% formaldehyde. FITC-conjugated Ulex europaeus, D. biflorus agglutinin, and B. simplicifolia I/B 4 lectins were used at 5 g/ml in PBS/FCS/NaN 3 . Cells were labeled for 1 h with shaking at 4°C and then washed and fixed as above. Each individually labeled sample was analyzed separately by flow cytometry using a Coulter XL cytometer. The errors reported are S.D. values.
Labeling of PLECT cells with 3,3Ј-dioctadecylindocarbocyanine-conjugated low density lipoprotein and flow cytometric analysis was performed as recommended by Biogenesis.
Cell membrane extracts were prepared using n-octyl glucopyranoside (Sigma). 2 million cells were detached from the plastic using 5 mM EDTA in PBS. After two washes with PBS, the cells were resuspended in 200 l of ice-cold PBS containing 50 mM n-octyl glucopyranoside, phenylmethylsulfonyl fluoride, pepstatin, leupeptin, and cystatine (all inhibitors from Sigma) and incubated on ice for 1 h with occasional gentle mixing. At the end of the solubilization step, the cells were spun down at 500 ϫ g, and the supernatant was used for electrophoresis and Western blots.
Papain Digests-2-3 million cells were removed from the plastic by brief treatment with cell culture 5 mM EDTA in PBS and washed into 200 l of 150 mM NaCl solution. Papain digestion was carried out at 37°C for 2 h after addition of an equal volume of Papinex reagent to the cell suspension. Cells were resuspended from time to time and, at the end of the incubation, were washed several times with cold PBS before lectin binding experiments for flow cytometry or preparation of cell membrane extracts as described above.
Western Blots-The Mini-protean II Western blotting system was used to transfer 10 l of 10% SDS-polyacrylamide gel electrophoresisseparated PLECT cell membrane-extracted proteins onto 0.45-m nitrocellulose membrane (Bio-Rad) at 100 V for 1 h or at 20 V overnight. Background binding was blocked using 1% bovine serum albumin in Tris-buffered saline for 2 h at room temperature. Biotinylated lectin was applied at 5 g/ml in Tris-buffered saline, 9 mM CaCl 2 , 4 mM MgCl 2 and left to bind at room temperature for 1 h. After several washes, bound lectin was detected using Streptavidin-alkaline phosphatase conjugate (Sigma). Alkaline phosphatase activity was detected using nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate substrate as recommended by the manufacturer (Promega).
Glycosyltransferase Assays-Aliquots of cell suspensions (2 ϫ 10 4 in 150 mM NaCl) were stored at Ϫ70°C prior to assay. Cells were solubilized in 0.2% Triton X-100 for 1 h at 4°C. ␣-1,3-Fucosyltransferase was assayed with the following 8-methoxycarbonyloctyl (R) acceptors: Incubation mixtures contained [ 3 H]GDP-fucose (50,000 dpm; NEN Life Science Products; specific activity, 6.95 Ci/mmol), 400 M acceptor, sodium cacodylate buffer 0.1 M, pH 7.0, 0.2% bovine serum albumin, and 10 mM MnCl 2 . The reaction volume was 20 l. Assays were incubated for 15 min or 3 h (to obtain adequate incorporation of radioactive sugar) at 37°C and stopped by the addition of 0.5 ml of water. Radioactive product was separated from the nucleotide sugar donor using Whatman Sep-Pak C18 cartridges and eluted in 3.5 ml of methanol for scintillation counting (46).
␣-1,2-Fucosyltransferase was assayed with 5 mM phenyl-␤-D-galactoside as the acceptor substrate. To compare the levels of enzyme activity in transfected cells, saturating concentrations (50 M) of GDPfucose were achieved by adding cold GDP-fucose (Sigma).

Cell Surface Expression of H, ␣-Gal, and Related Structures-Untransfected PLECT cells and cells transfected with
the pREP10 vector (mock-transfected) expressed low levels of H and Lewis X epitopes. No blood group A, B, VIM-2, Lewis a , sialyl Lewis X or sialyl Lewis a antigens were present on the parent cell line or on any of the human ␣-1,2FT-expressing clones. Transfection of PLECT cells with human ␣-1,2FT cDNA resulted in the isolation of several stable clones displaying increased levels of type 2 H epitope. Clones D and E, displaying the highest staining with anti-H mAb and U. europaeus lectin using flow cytometry (Fig. 2, A and B), were chosen for further study. Increased staining with anti-H reagents correlated with increased ␣-1,2FT activity as measured in vitro using ␤-phenylgalactoside as acceptor substrate (Table I). However, there was no significant difference between the level of ␣-1,2FT activity in clones D and E despite measurable two-fold difference in U. europaeus lectin or anti-H mAb binding (Fig. 3). Increased expression of H was accompanied by a decrease in the expression of the ␣-Gal epitope as detected by the binding of B. simplicifolia I/B 4 (Fig. 3). There was no change in the ␣-1,3GT activity (data not shown).
The low levels of cell surface expression of Lewis X antigen detected on untransfected and mock-transfected PLECT cells were reduced to background on the cells transfected with human ␣-1,2FT cDNA. By contrast, the Lewis Y antigen, which  was not previously detectable, appeared following transfection (Fig. 4A). ␣-1,3FT activity, which is necessary for the synthesis of Lewis X and Lewis Y structures on PLECT cells, was present with a substrate specificity similar to that of human Fuc-TIV (Table II). There were no differences in activity measured with H type 2-R between the control and ␣-1,2FT-transfected cells (data not shown). These results suggest that down-regulation of Lewis X expression in transfected cells was a result of direct competition between human ␣-1,2and pig ␣-1,3-fucosyltransferases occurring in the ␣-1,2FT-transfected cells.
The Binding of Human IgM to Pig Endothelial Cell Transfectants-Human IgM binding was studied by flow cytometry using total purified human IgM. At concentrations up to 700 g/ml the mean fluorescence intensity of IgM bound to the PLECT cells was proportional to the IgM concentration in the incubation buffer (data not shown). It was not possible to obtain saturating conditions with the IgM preparation used (sup-plied at 1 mg/ml). Therefore, at the concentrations used (100 and 200 g/ml), the mean fluorescence intensity measures the level of xenoreactive natural antibody present in the IgM preparation rather than the level of xenoreactive epitope expression on the PLECT cells. Contrary to expectation, the human ␣-1,2FT-expressing clones D and E bound more human IgM than the mock-transfected PLECT cells (Fig. 6).
Oligosaccharide inhibition studies of human IgM binding to the PLECT cells by sucrose, type 2 H trisaccharide, Lewis Y tetrasaccharide, and ␣-Gal trisaccharide were carried out to characterize the xenoreactive epitopes on mock-transfected and human ␣-1,2FT-expressing PLECT cells. Only ␣-Gal trisaccharide inhibited human IgM binding to PLECT cells, by 65% in mock transfected cells and by 80% in the ␣-1,2FTexpressing PLECT cells (Fig. 7). This suggests that the increased binding of human IgM to the ␣-1,2FT-expressing PLECT cells was still largely due to the ␣-Gal-related epitopes. This is further supported by a similar pattern of inhibition achieved by ␣-Gal trisaccharide for the mock-transfected and the ␣-1,2FT-expressing PLECT cells (Fig. 7).
Western Blot Analysis of Changes in the ␣-1,3-Galactosylation and ␣-1,2-Fucosylation of Cell Membrane Glycoproteins-Western blots of cell membrane extracts from human ␣-1,2FTexpressing and control cell lines were probed with biotinylated U. europaeus and B. simplicifolia I/B 4 lectins to identify the glycoprotein substrates of ␣-1,3GT and ␣-1,2FT. Although lactosamine is a substrate for both ␣-1,3GT and ␣-1,2FT in vitro, it can be seen that the glycoprotein substrate repertoires of these two porcine enzymes in the PLECT cells are not identical (Fig. 7, lanes 1 and 3). It was also noted that expression of human ␣-1,2FT in PLECT cells did not affect either overall intensity or distribution of H or ␣-Gal epitopes on the cell surface proteins (Fig. 8, lanes 2 and 4). To further distinguish between H and ␣-Gal epitopes on glycoproteins and glycolipids synthesized by PLECT cells, the cells were subjected to papain digestion. Preincubation of mock-transfected and ␣-1,2FT-expressing PLECT cells with papain was found to decrease sig-   nificantly the amount of U. europaeus lectin-and B. simplicifolia I/B 4 -reactive glycoproteins in cell membrane extracts (Fig.  8, lanes 5-8). Although there was a marked decrease in the lectin binding to blotted proteins from papain-treated cells, binding to the cell surface, as measured by flow cytometry, was virtually unchanged (Fig. 9). This difference may be explained by the fact that flow cytometric analysis measures antigen expression on both glycoprotein and glycolipid components of the cell surface, whereas Western blotting characterizes primarily glycoproteins. These data, together with the lack of increase in H staining and decrease in ␣-Gal staining on West-ern blots from transfected cells suggests that a significant fraction of the increased expression of H may have been on glycolipids.

DISCUSSION
In the present study, we have established an immortalized pig liver endothelial cell line (PLECT) and stably transfected it with human ␣-1,2FT, an intracellular class II transmembrane Golgi-located glycosyltransferase that directs the expression of type 2 H epitopes (47). Increased expression levels of ␣-1,2FT activity in PLECT cells led to increased H epitope expression and a decrease in the expression of ␣-Gal, Lewis X , and ␣-2,3and ␣-2,6-linked sialic acid. Our results suggest that new type 2 H epitopes may be associated predominantly with glycolipids rather than glycoproteins. Additionally, overexpression of human ␣-1,2FT in the presence of endogenous pig ␣-1,3FT activity resulted in the formation of cell surface Lewis Y .
The observed effects of human ␣-1,2FT expression on the PLECT cells are best described within the framework of the terminal glycosylation pathways in these cells (Fig. 1) Effective competition between ␣-1,2FT and ␣-1,3GT requires that they both glycosylate the same acceptor substrates, be they glycolipids or glycoproteins. Both utilize low molecular weight type 2 substrates in vitro, but in vivo their activity may be determined by much finer substrate specificity. For instance, on human erythrocyte cell membrane proteins, the H epitope is not uniformly expressed on all membrane proteins but is predominantly associated with polylactosamine residues on band 3 and 4.5 membrane glycoproteins (48 -50). Similarly, the distribution of the ␣-Gal epitope on porcine platelets and endothelial cell glycoproteins is not uniform but appears to be associated primarily with integrins and von Willebrand factor (14,51). Consistent with this is our observation that Western blots of mock-transfected and human ␣-1,2FT-expressing PLECT cell membrane extracts probed with ␣-Gal-specific B. simplicifolia I/B 4 and H-specific U. europaeus I lectins are not identical (Fig. 8), suggesting that the glycoprotein substrates of these two enzymes may be different. The absence of any change in the glycoprotein profiles introduced by overexpression of ␣-1,2FT in PLECT cells was a surprising finding in view of the results obtained by Sharma et al. (60), who transfected Chinese hamster ovary cells with pig ␣-1,3GT and human ␣-1,2FT cDNAs and demonstrated effective competition between these two enzymes for the same glycoprotein substrate. It is of interest, however, that the other B. simplicifolia I/B 4 -reactive glycoproteins on their Western blots did not show such an effect. Whether those bands represented genuine ␣-Gal epitopes synthesized by endogenous Chinese hamster ovary cell ␣-1,3GT or merely nonspecific binding of the lectin is unclear, since there are reports both ruling out (61) and supporting (62) the expression of endogenous ␣-1,3GT in Chinese hamster ovary cells. Our data suggest that the natural acceptor substrates for ␣-1,3GT in porcine endothelial cells may not function as substrates for human ␣-1,2FT. Therefore, any reduction of ␣-Gal expression that occurs through competition with ␣-1,2FT is likely to be cell type-specific and determined by the individual protein repertoires of cells in different tissues. This may also explain why expression of human ␣-1,2FT resulted in an approximately 70% decrease in the ␣-Gal expression in pig epithelial PLL-K1 cells observed by Sandrin et al. (33) but only a 40% decrease in the PLECT cells in this study. A greater than 200-fold increase in ␣-1,2FT activity in PLECT cells resulted in at most a 9-fold increase in cell surface expression of H and only a 40 -50% decrease in the expression of ␣-Gal. Alternatively, these differences could be explained by a higher level of endogenous ␣-1,3GT activity in PLECT cells compared with PLL-K1 cells.
Sequential activity of endogenous ␣-1,3FT on H structures in transfected cells resulted in the formation of Lewis Y neoepitopes. Lewis Y is an oncodevelopmental antigen expressed on the surface of cancer (52)(53)(54)(55) and embryonic cells (56 -58). It is a marker for apoptosis (59). More research will be necessary to establish the significance of high level expression of Lewis Y epitope in pig to primate xenotransplantation.
In addition to substitution with ␣-galactose and fucose, lactosamine structures are also sialylated in PLECT cells. Following transfection with human ␣-1,2FT cDNA, there was significant down-regulation of ␣-2,3and ␣-2,6-sialic acid expression as assessed by M. amurensis agglutinin and S. nigra agglutinin binding. PLECT cells have detectable levels of ␣-2,3-sialyltransferase activity measured with LNB1-R; therefore, the binding of M. amurensis agglutinin was likely to be at least partly directed to ␣-2,3-sialyllactosamine. The complete loss of Lewis X from the cell surface of human ␣-1,2FT-expressing PLECT cells suggests that, unlike ␣-1,3GT, endogenous porcine Fuc-TIV has a glycoprotein substrate repertoire that comprises molecules that are utilized by ␣-1,2FT as well.
The level of human anti-porcine IgM varies from 5 to 100 g/ml (63). Approximately 80 -90% of this activity is specific for ␣-Gal (6, 63), a figure that was confirmed in the present study. Dose-dependent binding of human IgM both to the control transfected and ␣-1,2FT-transfected PLECT cells indicates that 40% decrease in ␣-Gal epitope expression as detected by B. simplicifolia I/B 4 lectin binding was not sufficient to bring the number of IgM binding sites per cell down to the level at which they would all be saturated. Rather, ␣-Gal trisaccharide inhibition of the small increase in IgM binding to the ␣-1,2FTexpressing pig cells suggests that a relatively small number of new ␣-Gal-related epitopes were induced on these cells. We could not draw conclusions about the absolute level of expression of ␣-Gal and other xenoreactive epitopes that determine human IgM binding, but, given that the total level of terminal ␣-Gal was significantly reduced, there is the possibility that new and previously unexpressed xenoreactive epitopes appeared on the ␣-1,2FT-expressing PLECT cells.
In conclusion, we have shown first that human ␣-1,2FT expressed in pig endothelial cells can compete with endogenous pig ␣-1,3GT and results in cell surface down-regulation of ␣-Gal. Second, in transfected cells other glycosylation pathways are also affected. Lewis X disappears, and Lewis Y appears together with changes in sialylation. Third, although substantial reduction in ␣-Gal expression can be achieved by ␣-1,2FT overexpression, it may not be sufficient to inhibit the binding of human xenoreactive IgM to all cell types. Future studies should address the relative importance of the glycolipid component. The slow growth characteristics of ␣-1,2FT-transfected PLECT cells did not allow cells to be harvested in sufficient numbers for glycolipid analysis. This must be an important consideration in selecting model systems for future studies.
Finally, the modification of cell surface oligosaccharides such as sialyl Lewis X in response to cytokine stimulation is an important characteristic of endothelial cells. The results of our experiments suggest that this may be altered in ␣-1,2FT-expressing porcine endothelial cells. Substantially decreased levels of ␣-2,3-sialylated structures may affect the synthesis of sialyl Lewis X and hence, for example, normal lymphocyte adhesion and trafficking. Therefore, alternative, more targeted approaches, such as gene knockout to inhibit expression of the ␣-Gal epitope, may be more effective in the context of pig to human xenotransplantation.