Target Cell Susceptibility to Lysis by Human Natural Killer Cells Is Augmented by α(1,3)-Galactosyltransferase and Reduced by α(1,2)-Fucosyltransferase*

Susceptibility of porcine endothelial cells to human natural killer (NK) cell lysis was found to reflect surface expression of ligands containing Gal α(1,3)GlcNAc, the principal antigen on porcine endothelium recognized by xenoreactive human antibodies. Genetically modifying expression of this epitope on porcine endothelium by transfection with the α(1,2)-fucosyltransferase gene reduced susceptibility to human NK lysis. These results indicate that surface carbohydrate remodeling profoundly affects target cell susceptibility to NK lysis, and suggest that successful transgenic strategies to limit xenograft rejection by NK cells and xenoreactive antibodies will need to incorporate carbohydrate remodeling.

The severe shortage of human organs has focused recent investigation into cross-species transplantation. Pigs are an appropriate donor source, because their organs have similar physiology and size to human organs, they can be bred in large numbers, and they are relatively free of pathogens capable of causing infection in humans. However, porcine xenografts transplanted into primate recipients undergo hyperacute rejection within minutes to hours of engraftment. The process is mediated by host complement and preformed IgM antibodies directed against Gal ␣(1,3)Gal epitopes present in various cell surface structures on porcine endothelium (1)(2)(3)(4)(5)(6). In contrast to pigs, humans and Old World monkeys do not express Gal ␣(1,3)Gal in their tissues, because the gene encoding the ␣(1,3)galactosyltransferase, which links a terminal galactose residue to Gal ␤(1,4)GlcNAc oligosaccharide backbone structures, is inactive in these species (7,8).
Anti-Gal ␣(1,3)Gal antibodies develop in humans and higher primates within the first months of life, in parallel with the colonization of the gastrointestinal tract with bacteria containing ␣(1,3)-linked galactose residues in their cell walls (9,10). Consequently, there exists a window period in which these IgM antibodies are not present in neonatal primates (11). The absence of preformed IgM anti-Gal ␣(1,3)Gal antibodies in neonatal primates enables porcine cardiac xenografts transplanted heterotopically into unmedicated newborn baboons to survive beyond the hyperacute period (12); making this an appropriate model for studying the subsequent immunological barriers to xenotransplantation. In these recipients, a second primate anti-pig immunological response occurs after 3-4 days, resulting in graft loss accompanied by dense xenograft infiltration with natural killer (NK) 1 cells, macrophages, and deposition of induced IgG antibodies (13)(14)(15)(16)(17). Because similar findings have been demonstrated in guinea pig-to-rat cardiac xenotransplantation in which the recipients were treated with cobra venom factor to inactivate the host complement system (18), these observations suggest that a T cell-independent delayed rejection process, mediated largely by NK cells, occurs in widely disparate transplant combinations, including pig to primate.
NK cell lysis is regulated by a balance of intracellular signals transmitted via stimulatory and inhibitory cell surface receptors after specific binding to their respective target cell ligands (19,20). Inhibitory receptors on NK cells have carbohydrate binding domains with specificity for target cell glycoprotein ligands encoded by certain major histocompatibility complex (MHC) class I genes (21,22). Stimulatory receptors on NK cells also have carbohydrate binding domains within C-type lectin structures; however, their target cell glycoprotein ligands have not been well-defined (23,24). Recent evidence suggests that NK cells and a subset of B cells may belong to an innate immunological system designed to combat frequently encountered carbohydrate antigens, such as those in the cell walls of bacterial pathogens (25)(26)(27). Carbohydrate antigens can induce T cell-independent B cell antibody responses and can directly stimulate NK cells, without previous antigen sensitization or MHC restriction, to initiate lysis and to produce IFN-␥. Costimulatory signals provided by the NK cells, together with the effects of NK cell-derived IFN-␥ on B cell differentiation, isotype switching, and immunoglobulin secretion, ultimately result in augmentation of the IgG humoral response against the T cell-independent antigen (28 -30). Because the T cell-independent process of delayed xenograft rejection involves NK cells and IgG antibodies, and the principal antigen on porcine endothelium recognized by xenoreactive human antibodies is the carbohydrate epitope Gal ␣(1,3)Gal, we addressed the possibility that receptors on human NK cells may also react with ligands containing terminal Gal ␣(1,3)Gal residues, leading to augmented natural cytotoxicity as well as IgG humoral activity against porcine endothelium.

Preparation of Target Cells
Pig Aortic Endothelial Cells (PAECs)-Fresh pig aortas were treated for 1 h with 0.5% collagenase (Type IV, Sigma), lightly washed with Hanks' solution, and gently raked with a plastic cell scraper. The liberated endothelial cells were added to tissue culture vessels in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Life Technologies, Inc.) and 1% penicillin-streptomycin (Life Technologies). The cells were grown to confluence and then transferred to T25 flasks (Becton Dickinson, Franklin Lakes, NJ) in fresh medium.
Human Umbilical Vein Endothelial Cells (HUVECs)-HUVECs were purchased from the American Type Culture Collection (Rockville, MD; cell line CRL-1730), transferred to T25 flasks with fresh medium, and grown to confluence.
COS Cells-COS-7 cells were purchased from the American Type Culture Collection (cell line CRL-1651), transferred to T25 flasks, and grown to confluence. All cells were used between the third and seventh passages.

Preparation of Effector Cells
Peripheral Blood Mononuclear Cells (PBMCs)-Human PBMCs were isolated from heparinized whole blood using Isopaque-Ficoll (Gallard Schlesinger Co., Carle Place, NY) and suspended at a concentration of 2.5 ϫ 10 6 cells/ml in augmented RPMI 1640 medium. Depending on the assay condition, the cells were cultured for 12-14 h with or without the addition of 1000 units/ml recombinant human interleukin 2 (IL-2; Peprotech, Rocky Hill, NJ) before being used in functional assays.
Purified NK Cells-PBMCs were suspended at a concentration of 2.0 ϫ 10 6 cells/ml in phosphate-buffered saline (Life Technologies) with 1% bovine serum albumin (Life Technologies). A mixture of magnetic beads conjugated with antibody directed against T cells (CD3), B cells (CD19), and monocytes (CD14) (Dynal, Inc., Lake Success, NY) was added to the cell suspension (at a ratio of 10 beads/cell) and electronically stirred for 60 min at 4°C. The Magnet Particle Concentrator (MPC-1, Dynal) was used to isolate the beads containing CD3-, CD19-, and CD14-positive cells. The suspension was collected, washed, and stained for the presence of CD56-and CD16-positive cells. This technique reliably isolated a population of cells that was Ͼ80 -85% NK cells (CD56 ϩ , CD16 ϩ ) with Ͻ1-3% contamination with T cells (CD3 ϩ ). NK cells were resuspended at a concentration of 2.5 ϫ 10 6 cells/ml in RPMI 1640 medium and cultured for 12-14 h with or without the addition of 1000 units/ml recombinant human IL-2 before being used in functional assays.

Cytotoxicity Assay
Details of the cytotoxicity assay have been extensively described elsewhere (31). Briefly, target cells (2 ϫ 10 4 cells/well) were seeded in flat bottom 96-well plates (Becton Dickinson) and grown to confluence at 37°C and 5% CO 2 overnight. The monolayers were washed with Hanks' solution (Life Technologies, Inc.), labeled with 51 Cr (2-4 Ci/ well, Amersham Pharmacia Biotech) for 60 min, and then extensively washed before being used in functional assays. Effector cells were added at the desired effector:target ratio, typically 20:1 for assays using PBMCs and 10:1 for assays using purified NK cells, and brought to a final volume of 200 l/well. After incubation for 4 h, 100 l of supernatant was collected, and 51 Cr was measured in a gamma counter (Clinigamma 1272; Wallac Inc., Gaithersburg, MD). Data are presented as percent specific lysis or percent inhibition of lysis. Percent specific lysis was calculated using the formula (experimental cpm Ϫ spontaneous cpm)/(maximum cpm Ϫ spontaneous cpm) ϫ 100; where maximum cpm was determined by adding 10% Triton X-100 (Sigma). Percent inhibition of lysis was calculated using the formula 100 ϫ (% lysis observed with control condition Ϫ % lysis observed for each experimental condition)/% lysis observed with control condition. All assays were done in triplicate with a minimum of three donors. Results are presented as the mean Ϯ S.E.

Treatment of PAECs with IB 4
51 Cr-Labeled PAEC monolayers were treated with the plant isolectin b4 (IB 4 Sigma) isolated from Bandeiraea (Griffonia) simplicifolia at concentrations of 2.0, 20, and 200 g/ml for 60 min at 37°C and 5% CO 2 . The monolayers were washed in Hanks' solution and used in a standard lytic assay.

Enzymatic Treatment of PAECs
Porcine endothelium was treated with ␣-galactosidase at concentrations shown to reduce antibody-directed complement lysis of porcine endothelium (32). Briefly, PAEC monolayers were treated with either ␣-galactosidase isolated from the Green coffee been (Sigma) or ␤-galactosidase isolated from Escherichia coli bacteria (Sigma) for 4 h at pH 6.0 or 7.3, respectively. The monolayers were then extensively washed with Hanks' solution and labeled with 51 Cr and used in standard NK lytic assays. Treatment of the porcine endothelium with ␣-galactosidase was specific for Gal ␣(1,3)Gal epitopes and did not affect other carbohydrate epitopes (data not shown). Furthermore, treatment with ␤-galactosidase did not remove Gal ␣(1,3)Gal epitopes or other carbohydrate epitopes serving as an appropriate negative control.

Inhibition of NK Lysis by Soluble Oligosaccharides
All oligosaccharide derivatives were obtained from Dextra Laboratories (Reading, UK). Purified NK cells were incubated with a 10 Ϫ5 M concentration of each oligosaccharide derivative at 37°C for 60 min before being used in standard lytic assays.

Transfection of COS-7 Cells
COS-7 cells were grown to confluence in RPMI 1640 medium. Details of the CDM8 plasmid containing the murine ␣(1,3)-galactosyltransferase gene have been previously described (4). Briefly, vector DNA (7-10 g/1 ϫ 10 6 targeted cells) was added to Optimem culture medium (Life Technologies) with 5% LipofectAMINE (Life Technologies) and brought to a final concentration of 5-10 g/ml. The DNA was incubated for 20 min at room temperature and then diluted to a final volume of 7-10 ml. The transfection media were added to 1 ϫ 10 6 COS cells and incubated overnight at 37°C with 5% CO 2 . The following morning the transfection media were aspirated, and the cells were cultured for 48 h before use.

Transfection of Porcine Endothelial Cells
The porcine endothelial cell line was cotransfected with the pig ␣(1,2)-fucosyltransferase cDNA in the expression vector pCDNA-1 (pHT plasmid; Ref. 33) and the pcDNA-1-neo plasmid (Invitrogen, Carlsbad, CA) using a standard calcium phosphate technique with 20 g of pHT plasmid and 1 g of pcDNA-1-neo plasmid. Cells were selected for stable integration of transfected DNA by selection in media containing G418, cloned using limiting dilution, and maintained in media containing G418.
Flow Cytometry 5 ϫ 10 5 of appropriate target cells were washed and resuspended in phosphate-buffered saline with 0.1% bovine serum albumin. Fluorescein isothiocyanate (FITC)-conjugated lectin (2 g/ml), IB 4 , or Ulex europaeus agglutinin, type I (Sigma) was added to each cell suspension and incubated for 45 min at 4°C. The cells were washed and fixed in 1% paraformaldehyde (Sigma). Mean channel fluorescence was measured in a FACScan flow cytometer (Becton Dickinson).

Susceptibility of Porcine Endothelium to Human NK Lysis
Correlates with Expression of Gal ␣(1,3)-Gal-Two endothelial cell lines with opposing ␣(1,3)-galactosyltransferase activities and surface expression of terminal Gal ␣(1,3)Gal residues were selected as initial targets for lysis by human NK cells: PAECs and HUVECs (Fig. 1a). Human NK lysis of xenogeneic porcine endothelium was Ͼ2-fold greater than that of allogeneic human endothelium (Fig. 1b), consistent with the possibility that expression of Gal ␣(1,3)Gal increases susceptibility of xenogeneic endothelium to lysis by human NK cells. To more directly examine the role of the terminal Gal ␣(1,3)Gal structure in the heightened susceptibility of xenogeneic porcine endothelium to human NK lysis, inhibition experiments were performed using the plant lectin IB 4 , which specifically binds to this structure (34). NK lysis of porcine endothelium was markedly reduced in the presence of IB 4 in a concentration-dependent manner (Fig.  1c). The next set of experiments sought to identify the terminal ␣(1,3)-linked galactose residue within the Gal ␣(1,3)Gal structure as an essential component of porcine ligands involved in triggering human NK cell lysis. Enzymatic treatment of porcine endothelium with ␣-galactosidase reduced NK lysis in a concentration-dependent manner, which correlated with the level of reduced Gal ␣(1,3)Gal expression (Fig. 1, d and e). At the highest concentration of ␣-galactosidase used, 20 units/ml, NK lysis was inhibited by a mean of 35% accompanying a 44% reduction in cell surface expression of Gal ␣(1,3)Gal. This inhibition of NK lysis was specific to cleavage of terminal ␣(1,3)linked galactose residues, because enzymatic treatment with ␤-galactosidase had no effect (Fig. 1f).
Expression of Gal ␣(1,3)-Gal in COS Cells Increases Susceptibility to NK Lysis-To directly demonstrate the effect of surface expression of ␣(1,3)-linked galactose residues on susceptibility to NK lysis, COS cells were transfected with the murine ␣(1,3)-galactosyltransferase gene (Fig. 2a). These cells do not normally express Gal ␣(1,3)Gal epitopes and acquire susceptibility to complement-mediated lysis in the presence of human serum after transfection with ␣(1,3)-galactosyltransferase (4). In the present study, COS cells transfected with ␣(1,3)-galactosyltransferase, but not the vector alone, showed increased susceptibility to lysis by human NK cells at every effector: target ratio tested (Fig. 2b). A recent study using similarly transfected COS cells demonstrated enhanced adhesion of human NK cells to COS cells expressing Gal ␣(1,3)Gal (35). Our results extend these observations and show that the increased binding of NK cells to terminal Gal ␣(1,3)Gal residues expressed by ligands on cells transfected with ␣(1,3)-galactosyltransferase leads to activation of NK cell stimulatory receptors and causes increased target cell lysis. Moreover, because augmented NK lysis of ␣(1,3)-galactosyltransferase transfected cells was observed for both NK cells at rest and after cytokine activation (Fig. 2c), these findings suggest that the stimulatory NK cell receptors that bind ligands containing Gal ␣(1,3)Gal are constitutively expressed. Gal residues on NK lysis of porcine endothelium, human NK cells were incubated with two pairs of soluble oligosaccharides, each pair consisting of the tetrasaccharide backbone and its appropriate derivative after glycosyltransferase catalysis (Fig. 3,  a-d). The type I tetrasaccharide lacto-N-tetra inhibited NK lysis by 2.1-fold higher levels than the type II tetrasaccharide lacto-N-neo-tetra (Fig. 3e), suggesting that carbohydrate binding structures on human NK cells may have a preference for ligands containing type I structures. Addition of a terminal Gal ␣(1,3)Gal residue inhibited specific NK lysis of porcine endothelium by 3.3-fold higher levels than the lacto-N-neo-tetra backbone structure (Fig. 3e), consistent with our previous data that ligands containing Gal ␣(1,3)Gal are bound by receptors on human NK cells. The addition of a terminal Fuc ␣(1,2)Gal residue also increased inhibition of NK lysis of porcine endothelium by levels 2.5-fold higher than the lacto-N-tetra backbone structure (Fig. 3e). Thus, human NK cells can bind both Gal ␣(1,3)Gal and Fuc ␣(1,2)Gal residues.

Inhibition of Human NK Cell Lysis of Porcine Endothelium by Soluble Gal
Expression of H Substance in Porcine Endothelium Reduces Target Cell Susceptibility to NK Lysis-To investigate the effect of cell surface carbohydrate remodeling on susceptibility to human NK lysis, porcine endothelial cells were transfected with ␣(1,2)-fucosyltransferase cDNA, and lines were derived that demonstrated stable expression but widely divergent levels of the H substance (Fig. 4, a and b). Surface expression of Gal ␣(1,3)Gal ␤(1,4)Glc was inversely proportional to that of Fuc ␣(1,2)Gal ␤(1,4)Glc, reflecting the degree of competition for Gal ␤(1,4)Glc substrate by the glycosyltransferases. Reduction in surface expression of Gal ␣(1,3)Gal significantly reduced susceptibility of porcine endothelial cells to lysis by human NK cells (Fig. 4c). Although lytic susceptibility decreased in direct parallel with reduction in surface levels of Gal ␣(1,3)Gal, human NK lysis could not be reduced by Ͼ55% even with Ͼ80% reduction of Gal ␣(1,3)Gal expression. This level of human NK lysis approaches that seen with allogeneic endothelium. thelium against the effects of human complement by expression of human complement inhibitory proteins (37)(38)(39) and 2) reduction in the level of Gal ␣(1,3)Gal expression on porcine endothelium by high level expression of the enzyme ␣(1,2)fucosyltransferase (36). The second strategy is predicated on the knowledge that both ␣(1,3)-galactosyltransferase and ␣(1,2)fucosyltransferase use the same acceptor substrate, Gal ␤(1,4)-GlcNAc, to direct synthesis of Gal ␣(1,3)Gal ␤(1,4)GlcNAc and Fuc ␣(1,2)Gal ␤(1,4)GlcNAc (H substance or blood type O phenotype), respectively. When both enzymes are cotransfected into COS cells, ␣(1,2)-fucosyltransferase dominates over ␣(1,3)galactosyl transferase so that Gal ␣(1,3)Gal expression is almost completely suppressed in the presence of Fuc ␣(1,2)Gal (36). This effect is the result of the temporal order of action of these enzymes, with ␣(1,2)-fucosyltransferase having preferential access to the Gal ␤(1,4)GlcNAc acceptor substance because of specific amino acid sequences in its cytoplasmic domain, which target its localization to particular compartments within the Golgi apparatus (40).
Using soluble oligosaccharide derivatives to competitively inhibit human NK cell lysis, we found that the addition of a terminal Gal ␣(1,3)Gal or terminal Fuc ␣(1,2)Gal residue to their respective carbohydrate derivative further inhibited human NK lysis of porcine endothelium. These results suggest that both Gal ␣(1,3)Gal and Fuc ␣(1,2)Gal residues are able to bind human NK cells and raise the possibility that substitution of a terminal ␣(1,2)-linked fucosyl residue for a terminal ␣(1,3)linked galactosyl residue to the Gal ␤(1,4)Glc backbone structure would not reduce target cell susceptibility to human NK cell lysis. We directly investigated this possibility using porcine endothelial cells that were transfected with ␣(1,2)-fucosyltransferase cDNA. Reduction in surface expression of Gal ␣(1,3)Gal significantly reduced susceptibility of porcine endothelial cells to lysis by human NK cells in direct parallel with reduction in surface levels of Gal ␣(1,3)Gal. However, NK lysis could not be fully eliminated even with almost complete reduction of Gal ␣(1,3)Gal expression, suggesting that additional factors may contribute to the human NK cell response to porcine endothelium. Possible additional mechanisms include interactions between noncarbohydrate ligands on porcine endothelium and stimulatory receptors on human NK cells and/or incompatibility between swine MHC class I molecules and inhibitory receptors on human NK cells.
These data suggest that carbohydrate remodeling of porcine endothelium by high level expression of ␣(1,2)-fucosyltransferase decreases susceptibility to human NK lysis by two possible mechanisms: 1) a reduction of Gal ␣(1,3)Gal residues within porcine endothelial cell ligands, which bind to stimulatory receptors on human NK cells; and 2) an increase of Fuc ␣(1,2)Gal residues within porcine endothelial cell ligands, which bind nonactivating or inhibitory receptors on human NK cells. In both human and murine MHC class I structures, a conserved N-linked glycosylation site is located at Asn-86 (41), adjacent to residues 74 -83 of the ␣-chain, which encode the polymorphic epitopes recognized by inhibitory NK cell receptors (42)(43)(44). In humans, the oligosaccharide structures at this site are remarkably uniform among various MHC class I allotypes and generally contain terminal sialic acid residues (45). Because high level endothelial cell expression of ␣(1,2)-fucosyltransferase reduces terminal sialylation (46,47), presumably because of competition with sialytransferases for the lactosamine substrate, it is possible that the substitution of a terminal ␣(1,2)-linked fucose residue to the oligosaccharide chain at Asn-86 of the swine MHC class I structure may enhance binding of the MHC molecule to human NK cell inhibitory receptors. This possibility is currently the subject of inves-tigation in our laboratory.
With the development of transgenic pig organs resistant to complement-mediated hyperacute rejection, the subsequent immunological barrier confronted by these genetically modified xenografts on transplantation into primate recipients will be that comprising NK cells and macrophages. In this report, we have shown that primate NK cells react prominently with the same principal xenoantigen on porcine endothelium that is recognized by naturally occurring xenoreactive antibodies, confirming the relationship between NK cells and B cells within an innate compartment of the immune system that is T cell-independent. High level expression of ␣(1,2)-fucosyltransferase, which reduces binding of xenoreactive antibodies, protected porcine endothelium against lysis by human NK cells. Because the alternative transgenic strategy for overcoming complement-mediated hyperacute rejection is to induce expression of human complement inhibitory proteins to protect porcine endothelium against the effects of human complement, organs modified in this manner will continue to be susceptible to a process of delayed xenograft rejection mediated by NK cells and induced IgG antibodies reactive with ligands expressing Gal ␣(1,3)Gal epitopes. Our study suggests that successful transgenic strategies for pig-to-primate xenotransplantation will need to incorporate carbohydrate remodeling to limit xenograft rejection by a T cell-independent cellular and humoral process.