Lack of Galactose-α-1,3-Galactose Expression on Porcine Endothelial Cells Prevents Complement-Induced Lysis but Not Direct Xenogeneic NK Cytotoxicity1

The galactose-α-1,3-galactose (αGal) carbohydrate epitope is expressed on porcine, but not human cells, and therefore represents a major target for preformed human anti-pig natural Abs (NAb). Based on results from pig-to-primate animal models, NAb binding to porcine endothelial cells will likely induce complement activation, lysis, and hyperacute rejection in pig-to-human xenotransplantation. Human NK cells may also contribute to innate immune responses against xenografts, either by direct recognition of activating molecules on target cells or by FcγRIII-mediated xenogeneic Ab-dependent cellular cytotoxicity (ADCC). The present study addressed the question as to whether the lack of αGal protects porcine endothelial cells from NAb/complement-induced lysis, direct xenogeneic NK lysis, NAb-dependent ADCC, and adhesion of human NK cells under shear stress. Homologous recombination, panning, and limiting dilution cloning were used to generate an αGal-negative porcine endothelial cell line, PED2*3.51. NAb/complement-induced xenogeneic lysis of PED2*3.51 was reduced by an average of 86% compared with the αGal-positive phenotype. PED2*3.51 resisted NK cell-mediated ADCC with a reduction of lysis ranging from 30 to 70%. However, direct xenogeneic lysis of PED2*3.51, mediated either by freshly isolated or IL-2-activated human NK cells or the NK cell line NK92, was not reduced. Furthermore, adhesion of IL-2-activated human NK cells did not rely on αGal expression. In conclusion, removal of αGal leads to a clear reduction in complement-induced lysis and ADCC, but does not resolve adhesion of NK cells and direct anti-porcine NK cytotoxicity, indicating that αGal is not a dominant target for direct human NK cytotoxicity against porcine cells.

P ig-to-human xenotransplantation may resolve the severe shortage of human organs for transplantation (1). However, differences in cell surface glycosylation between humans and pigs leading to immunological responses constitute major hurdles for xenotransplantation (2,3). Vascularized organs transplanted from pigs to nonhuman primates undergo hyperacute rejection (HAR), 3 which is caused by the binding of preformed xenoreactive natural Abs (NAb) to porcine endothelial cells (pEC) (4 -6). Besides mediating complement and endothelial cell activation, noncomplement-fixing NAb may play an important role by initiating tissue damage in xenotransplants through Ab-dependent cell-mediated cytotoxicity (ADCC) (4,7,8). The majority of these NAb interact specifically with the galactose-␣-1,3-galactose (␣Gal) carbohydrate structure, an epitope that is synthesized by the ␣-1,3-galactosyltransferase (␣1,3GT) and abundantly expressed on the cell surface of all mammals, except humans, apes, and Old World monkeys. Several approaches may prevent HAR, including the temporary removal of xenoreactive Abs by plasmapheresis and specific extracorporeal immunoadsorption (9) and the inhibition of complement by the expression of human complement regulatory proteins in transgenic pigs (10). However, xenotransplantation performed after removal of NAb induces the production of even larger amounts of Abs against ␣Gal (11). These induced anti-␣Gal Abs contribute to a delayed xenograft rejection, also referred to as acute vascular rejection, which severely limits organ survival notwithstanding the presence of human complement regulatory proteins. Immunosuppressive drugs that efficiently suppress allograft rejection fail to prevent the production of anti-␣Gal Abs (12). Alternatively, the administration of soluble ␣Gal glycoconjugates reduces the levels of circulating anti-␣Gal Abs and the number of cells that secrete anti-␣Gal Abs. Some studies report a diminished Ab production for several weeks after such treatments (13,14). However, to date, no therapy has prevented the return of xenoreactive Abs in pig-to-primate models. Therefore, several attempts have been made to modify ␣Gal expression on porcine cells. Competitive inhibition of the ␣1,3GT in ␣-1,2-fucosyltransferase and N-acetylglucosaminyltransferase-III transgenic pigs or the intracellular expression of single chain Fv Abs against ␣1,3GT all have resulted in only partial reduction in epitope numbers and failed to substantially prolong graft survival in primates (15)(16)(17). The yet most promising strategy to overcome xenograft rejection concentrates on the elimination of the ␣Gal epitope by knocking out the ␣1,3GT locus in pigs. Whereas for many years embryonic stem cell technology essential for the generation of gene knockout (KO) animals was restricted to mice, the advent of new cloning and nuclear transfer technologies (18,19) led to the recent production of several pigs lacking one allele of the ␣1,3GT locus (20,21) or being completely knocked out for this enzyme (22). Currently, preclinical pig-to-nonhuman primate trials are being conducted to test whether organs derived from ␣Gal KO animals confer permanent and complete protection from Ab-mediated rejection. The availability of ␣Gal KO cells, however, should also enable a more systematic and rapid investigation of other antigenic targets involved in immune rejection of porcine xenografts.
Ab-independent mechanisms also contribute to the high susceptibility of porcine cells to damage inflicted by the human immune system. The finding that human NK cells are able to lyse pEC in vitro suggested that they may participate in the direct rejection of pig-to-human xenografts (23,24). This hypothesis was supported by NK cell infiltrates found in pig organs perfused with human blood ex vivo (25,26) as well as in pig-to-nonhuman primate xenografts (27) and in small animal models of both concordant and discordant xenotransplantation (28,29). In agreement with these observations, several adhesion receptors are functional across the human-porcine species barrier, providing the basic requirements for human leukocytes to interact with pEC (30,31).
The function of NK cells is regulated by a concert of activating and inhibitory signals (32,33). After the binding of NK cells to potential target cells, several receptor-ligand interactions take place that determine the fate of the latter. Recognition of self MHC class I molecules by inhibitory NK receptors represents a dominant-negative signal for NK cytotoxicity (34). However, in the absence of sufficient inhibitory signals, ligation of activating receptors induces NK cytotoxicity. The reported susceptibility of pEC to human NK cytotoxicity may be explained by the failure of human NK inhibitory receptors to recognize xenogeneic porcine MHC class I molecules. This hypothesis is supported by the fact that expression of human MHC class I molecules on pEC provides partial protection from human NK cytotoxicity (35,36). The nature of the potential activating NK ligands on porcine cells is largely unknown at present; however, some groups reported a direct recognition of ␣Gal by human NK cells (37)(38)(39). Apart from this direct NK cytotoxicity, the expression of Fc␥RIII (CD16) on human NK cells mediates ADCC against porcine target cells (4,7,8). In this process, NK recognition of xenoreactive human IgG Abs bound to ␣Gal and other to date unknown epitopes on pEC leads to a rapid lysis of these pig cells.
The aim of the present study was to examine whether the lack of ␣Gal protects pEC from human NAb/complement-induced lysis, direct human NK-mediated xenogeneic lysis, ADCC, and adhesion of human NK cells under physiological shear stress. We demonstrate that the lack of ␣Gal protects pEC from complement-induced cytotoxicity, from ADCC, but not from adhesion and direct lysis mediated by human NK cells.

Cells
The SV40-immortalized aortic pEC line PEDSV.15 was established and characterized in our laboratory (40), and the ␣Gal-negative cell line PED2*3.51 was generated, as described below. Both porcine cell lines were cultured in DMEM (Invitrogen AG, Basel, Switzerland) supplemented with 10% FCS (PAA Laboratories, Luzern, Switzerland), 1 mM sodium pyruvate, 2 mM L-glutamine, nonessential amino acids (ϫ100), and 20 mM HEPES (all Invitrogen); PED2*3.51 were repeatedly exposed to 400 g/ml G418 (Life Technologies, Gaithersburg, MD). For induction of VCAM-1 expression, pEC were stimulated with porcine IFN-␥ (Innogenetics, Zwijndrecht, Belgium) for 24 h (100 ng/ml). The isolation of human PBMC, the purification of NK cells, and the generation of polyclonal NK populations have been described previously (35). After isolation, NK cells were either used directly or activated by culture in AIM-V medium (Invitrogen AG, Basel, Switzerland) containing 100 U/ml human IL-2 (Chiron, Palo Alto, CA). The human clonal NK cell line NK92 (41) was obtained from the American Type Culture Collection (Manassas, VA) and maintained in MyeloCult H5100 medium (StemCell Technologies, Vancouver, Canada) containing 500 U/ml human IL-2.

Targeting vector construction
A genomic library was constructed from DNA of the same d/d haplotype partially inbred miniature swine line used for isolation of the pEC line PEDSV.15 (40,42). DNA from overlapping genomic clones was assembled into a contiguous 23-kbp segment of the ␣1,3GT gene locus (GGTA1), beginning downstream of exon 7 and continuing 10.5 kbp beyond exon 9. In-frame termination codons were introduced near the beginning of exon 9, which should result in translational termination N-terminal to the catalytic domain of the enzyme. This mutated segment was cloned into a pOCUS II-based vector containing a G418 resistance gene under the control of the murine phosphoglycerol kinase promoter (Novagen, San Diego, CA). The resulting construct, pGallaway, was linearized at a unique XhoI site near exon 8 for use as an insertion-type targeting vector.

Gene-targeted immortalized endothelial cell line isolation
PEDSV.15 cells were electroporated at 260 V and 960 FD in 0.8 ml of HEPES buffered saline containing 0.5 pmol/ml XhoI-linearized pGallaway DNA assuring isogenicity of vector and target locus. Thirteen pools of transfected PEDSV.15 cells were selected for 2 wk in 400 g/ml G418. Based on analysis of smaller pools, these 13 pools were expected to contain ϳ7500 G418-resistant clones, 1% of which (75) would be targeted at the ␣1,3GT gene locus. To select for spontaneous null mutant cells arising from heterozygously targeted PEDSV.15 cells, each pool was subjected to multiple rounds of lysis with affinity-purified baboon Abs against ␣Gal and complement. 4 Following five rounds of selection, 1 of the 13 pools gave rise to a cell line that was totally resistant to subsequent selection with Ab and complement. This line was designated PED2*3. RNA from this line was analyzed by RT-PCR using a forward primer upstream of the 5Ј end of the targeting vector and a downstream primer within exon 9 downstream of the introduced termination codons. Only RNA containing the artificial stop codons was readily detected in the PED2*3 line. Further limiting dilution cloning by standard methods resulted in the stable ␣Gal-negative cell line PED2*3.51, as demonstrated by flow cytometry.

Human serum samples
Human sera were obtained from healthy adult volunteers. Decomplementation by heat inactivation was conducted at 56°C for 30 min. Samples were stored at 4°C for short periods or aliquoted and stored at Ϫ20°C. Binding of human Igs to pEC was analyzed by incubation with different dilutions of human serum for 30 min at 4°C and subsequent flow cytometry, as described below.

Cell staining and flow cytometry
Surface expression of ␣Gal molecules on pEC was analyzed on a FACScan (BD Biosciences, Basel, Switzerland) by direct and indirect immunofluorescence using either the FITC-labeled isolectin B4 from Bandeiraea simplicifolia (44) (BS-IB4; Sigma-Aldrich, Buchs, Switzerland) or the mouse anti-␣Gal IgM mAb M86 (45) (Alexis, Lausen, Switzerland). For the latter, a secondary FITC-conjugated goat anti-mouse Ig Ab (Boehringer Mannheim, Rotkreuz, Switzerland) was used. pEC cells were resuspended after trypsinization (0.25% trypsin; Life Technologies) at 5 ϫ 10 5 cells per tube in staining buffer (PBS, 0.1% BSA) and incubated for 30 min at 4°C with saturating concentrations of IB4 lectin or the mAb M86. For the expression of adhesion molecules on pEC, the following mAb (all IgG1) were used: 3F4 (anti-VCAM-1; kindly provided by Alexion, New Haven, CT), 1.2B6 (anti-E-selectin; Serotec, Kidlington-Oxford, U.K.), and 12C5 (anti-P-selectin; kindly provided by D. Haskard, Hammersmith Hospital, London, U.K.). Binding of human Igs to pEC was analyzed by incubating pEC with different dilutions of human serum, anti-␣Gal NAb-depleted serum, or purified anti-␣Gal NAb for 30 min at 4°C. For detection, a secondary FITC-conjugated goat anti-human IgM (Sigma-Aldrich) or a mouse antihuman IgG (Zymed, Basel, Switzerland) Ab was used. Phenotypic analysis of NK cells was conducted by direct immunofluorescence using FITC-UCHT1 (anti-CD3), PE-B73.1 (anti-CD16), and PE-B159 (anti-CD56) mAb (all from BD PharMingen, San Diego, CA). Irrelevant, isotype-matched mAb or FACS buffer alone (for BS-IB4) was used as controls and propidium iodide gating to exclude dead cells in all experiments. To compare the levels of surface expression, the geometric mean fluorescence intensity ratios (MFIR) were calculated by dividing the mean fluorescence intensity of staining with the mAb of interest with the mean fluorescence intensity of the control mAb.

Cytotoxicity assays
The cytotoxic activity of polyclonal human NK populations and the NK line NK92 was tested in 4-h 51 Cr release assays in serum-free AIM-V medium, as described previously (8). Briefly, target cells were added to triplicate samples of serial 2-fold dilutions of freshly isolated or IL-2activated NK cells in round-bottom 96-well plates at E:T ratios ranging from 40:1 to 5:1. To determine the effects of complement-induced xenogeneic lysis and ADCC, human serum containing NAb and complement, decomplemented human serum, decomplemented anti-␣Gal NAb-depleted serum, or purified ␣Gal-reactive NAb was added to 51 Cr-labeled target cells. Concentrations ranging from 25 to 3% serum or 50 to 6 g/ml NAb were used in the absence or presence of freshly isolated NK cells at an E:T ratio of 40:1. After incubation for 4 h at 37°C, the release of radioactive 51 Cr was analyzed on a gamma counter, and the percentage of direct specific lysis was calculated. For ADCC, the percentage of relative lysis was calculated as the difference in NK cytotoxicity against pEC in the presence or absence of anti-porcine NAb.

Adhesion assay
Adhesion of human NK cells on pEC was analyzed under shear stress, as previously described (46). Briefly, pEC were grown to confluency in 30-mm culture petri dishes and cultured for 24 h in the presence or absence of 100 ng/ml porcine IFN-␥. The resulting monolayers were washed and overlayed with 10 6 purified IL-2-activated NK cells in a volume of 100 l. The dishes were then rotated at 64 rpm in a prewarmed (37°C) horizontal shaker-incubator (Infors AG, Bottmingen, Switzerland). After 10 min, the assay was stopped by rapidly placing the dishes on ice and by prefixing the cells for 2 min with 1% paraformaldehyde in PBS. The monolayers were then gently washed, fixed, and finally protected with a glass coverslip. For quantification, four fields of 0.16 mm 2 were defined at a distance of 0.6 cm from the center of rotation, and the number of adhering NK cells was counted by light microscopy.

Generation of the ␣Gal KO porcine endothelial cell line PED2*3.51
A combined approach of gene targeting and panning was used to generate the ␣Gal-negative PED2*3 line. RNA from the PED2*3 line was analyzed by RT-PCR, and only RNA containing the artificial stop codons was detected (data not shown). This apparent lack of transcription of the wild-type ␣1,3GT genes in the PED2*3 line indicated that one allele was successfully targeted and the second allele silenced. Explanations could include a chromosome loss, promoter deletion, gene rearrangement, splicing mutation, epigenetic down-regulation, or other unknown mechanisms. The fast growing and stable PED2*3.51 subline was established by limiting dilution cloning. The lack of ␣Gal surface expression on PED2*3.51 was demonstrated by staining with the specific mAb M86 (45). Flow cytometry revealed an MFIR of 20.4 on the paternal PEDSV.15 line and 1.1 on PED2*3.51 (Fig. 1a). In addition, the BS-IB4 lectin stained PEDSV.15 and PED2*3.51 with an MFIR of 541.1 and 3.8, respectively (Fig. 1b). A possible explanation for the faint positive staining of PED2*3.51 with BS-IB4 lies within the cross-reactivity of this lectin to other oligosaccha-rides expressed on porcine cells including Gal␣1,6Gal (47). As a negative control for ␣Gal expression, primary human aortic endothelial cells were found to be completely negative for staining with M86 ( Fig. 1c) or BS-IB4 (Fig. 1d). Repetitive analysis of M86 binding throughout the study confirmed that the PED2*3.51 line does not express an active ␣1,3GT. Additional phenotyping of PED2*3.51 cells with Abs directed against CD14, CD31, CD44, CD49e, CD62E, CD62P, CD86, and swine leukocyte Ag class I d/d revealed an expression pattern (data not shown) similar as for the paternal PEDSV.15 line (40).

Binding of human NAb to porcine endothelium is strongly reduced by the lack of ␣Gal
To determine the influence of ␣Gal expression on the binding of NAb, pEC were incubated with different concentrations of human serum, anti-␣Gal-depleted serum, and purified anti-␣Gal NAb. As detected by flow cytometry, the binding of human IgM and IgG NAb in nonabsorbed serum was ϳ10 times lower on ␣Gal KO PED2*3.51 cells than on the paternal line PEDSV.15 (Fig. 2, a and  d). In contrast, when anti-␣Gal-depleted human serum was used, binding of IgM and IgG molecules was similar to both cell lines (Fig. 2, b and e). Moreover, the level of binding observed in the latter experiments was similar to the level of binding of nonabsorbed serum to the ␣Gal KO PED2*3.51 cells. These results suggested, first, that the higher amount of NAb binding to PEDSV.15 using nonabsorbed serum is due to ␣Gal-specific Abs and, second, that the remaining Abs binding to PED2*3.51 are of non␣Galspecific nature. Flow cytometry after incubation of PED2*3.51 cells with affinity-purified anti-␣Gal Abs revealed virtually no binding, thus confirming our previous negative staining using the anti-␣Gal mAb M86 (Fig. 2, c and f). However, there was a weak staining with IgG at very high Ab concentrations (Ն25 g/ml) probably due to small amounts of contaminating non-␣Gal-spe-cific IgG NAb or unspecific binding due to supersaturating conditions.

Strong reduction of xenogeneic NAb/complement lysis of PED2*3.51
It is well known that binding of human xenoreactive NAb to pEC activates the complement cascade, resulting in lysis of the cells. These data were confirmed in our study by the incubation of PEDSV.15 cells with human serum. A strong dose-dependent lysis of PEDSV.15 cells was observed with small amounts of serum (6%) derived from the donor MS, inducing a lysis over 40% (Fig.  3). In contrast, ␣Gal KO PED2*3.51 cells were highly resistant to NAb/complement-mediated lysis even in the presence of up to 25% of MS serum. The degree of complement-induced xenogeneic lysis of PED2*3.51 and PEDSV.15 was donor dependent probably due to different levels and reactivities of the xenoreactive NAb present in the respective sera. However, when sera from five different donors were analyzed, lysis of PED2*3.51 was reduced by an average of 86% as compared with PEDSV.15, as shown in Table I. A similar effect was seen with human plasma (data not shown).

Direct cytotoxicity of human NK cells and NK92 is independent of ␣Gal expression
pEC are susceptible to human NK cytotoxicity, and there have been controversial reports about a direct involvement of ␣Gal as a target molecule. To test the role of ␣Gal in xenogeneic NK cytotoxicity, PEDSV.15 and ␣Gal KO PED2*3.51 cells were used as targets for freshly isolated or IL-2-activated human NK cells in    (Fig. 4a). In addition, the human NK line NK92 also exhibited the same level of cytotoxicity against both targets (Fig. 4b). These findings indicate that ␣Gal is either not directly recognized by human NK cells or does not play a dominant role in triggering xenogeneic cytotoxicity in NK cells.

Reduction of xenogeneic ADCC against ␣Gal KO PED2*3.51
As previously shown in vitro, NAb-dependent ADCC of pEC mediated by human NK cells is a very efficient mechanism of lysis and may also represent an important in vivo mechanism leading to graft rejection. In this study, we investigated whether the lack of ␣Gal on the surface of pEC protects against human NK cell-mediated ADCC. Therefore, freshly isolated NK cells were added to PEDSV.15 and ␣Gal KO PED2*3.51 target cells at an E:T ratio of 40:1 in the presence of decomplemented human serum, anti-␣Galdepleted serum, and purified ␣Gal-NAb. In general, the level of xenogeneic ADCC correlated with the amount of anti-pig NAb present in the serum of the three donors tested. As shown in Fig.  5, ADCC of PED2*3.51 cells was strongly reduced as compared with PEDSV.15. The level of ADCC of PED2*3.51 was similar when using nondepleted or anti-␣Gal-depleted serum, suggesting that the remaining non-␣Gal-specific NAb are responsible for this lysis. However, using anti-␣Gal-depleted serum, the level of ADCC was still clearly higher against PEDSV.15 than against PED2*3.51 cells possibly due to small amounts of remaining anti-␣Gal Abs after depletion, as detected by ELISA (data not shown). When using purified anti-␣Gal NAb, ADCC against PED2*3.51 cells was almost completely abrogated. These findings indicate an important role for ␣Gal-specific Abs in xenogeneic ADCC, whereas non-␣Gal-specific Abs seem to be of minor importance.

Adhesion of human NK cells to PED2*3.51 cells under physiological shear stress is not dependent on ␣Gal expression
Before target cell lysis in vivo, NK cells have to adhere to pEC, a process that includes rolling and, subsequently, firm adhesion. Because ␣Gal expression was reported to be involved in adhesion of human NK cells to pEC in an earlier report (37), this question was also addressed in the present study. Adhesion of IL-2-activated human NK cells derived from different healthy donors was almost absent on PED2*3.51 cells. However, VCAM-1, which is crucial for the adhesion of human NK cells to pEC (46), was only detected on PEDSV.15, but not on PED2*3.51, unless stimulated with porcine IFN-␥ (Fig. 6, a and b). When IFN-␥-treated PED2*3.51 cells with up-regulated VCAM-1 expression were used in dynamic adhesion assays, human NK cells adhered to PED2*3.51 cells (Fig.  6d). However, the expression pattern of VCAM-1 on IFN-␥stimulated PED2*3.51 cells was heterogenous compared with PEDSV.15, a likely explanation for the lower numbers of adherent cells. Of note, the expression of ␣Gal on PED2*3.51 cells was not affected by treatment with porcine IFN-␥ remaining completely negative (Fig. 6c). We therefore conclude that ␣Gal is not crucial for the adhesion of human NK cells to porcine endothelial cells.

Discussion
The vigorous Ab response to porcine carbohydrate Ags, mainly to the ␣Gal epitope, has to be conquered before significant progress in clinical xenotransplantation can be made. It has been estimated that anti-␣Gal NAb comprise ϳ80% of all xenoreactive NAb in human serum (48), 1.0 -2.4% of total IgG, and 1-8.0% of total IgM (49 -51). Because a key role of anti-␣Gal Abs in xenograft rejection has been clearly demonstrated, great expectations are associated with the generation of pig organ donors that do not express ␣Gal. First results of pig-to-baboon kidney or heart transplantation using organs of ␣1,3GT KO pigs suggest that HAR can be overcome even without extracorporeal immunoabsorption and/or treatment with cobra venom factor (52,53). Consequently, the contribution of NAb with alternative non-␣Gal specificities and the study of subsequent cellular immune responses to xenografts such as acute vascular rejection are of growing importance.
In this study, we established an ␣Gal-negative porcine endothelial cell line, PED2*3.51, to investigate the influence of ␣Gal on the cytotoxicity mediated by xenoreactive human NAb and NK cells. Our approach included targeting of one allele of the ␣1,3GT gene and a subsequent panning procedure. Recent reports demonstrated that besides ␣1,3GT, another enzyme, isoglobotriaosylceramide-3 synthase, is able to synthesize ␣Gal in rats (54,55). Theoretically, a similar enzyme may also exist in pigs. However, because we could not detect any binding on PED2*3.51 cells using either the anti-␣Gal mAb M86 or purified anti-␣Gal NAb, we conclude that no other functional enzymes synthesizing ␣Gal are present. There is only one other description of porcine cells, primary fetal fibroblasts, lacking the ␣1,3GT gene (56). In extensive FACS analysis, the ␣1,3GT KO fetal pig fibroblasts were negative for staining with BS-IB4 and several different anti-␣Gal Abs. However, surprisingly, the cells still gave a positive FACS signal using two other mouse anti-␣Gal mAb, including the mAb that was used for the panning procedure to eliminate ␣Gal expression. Therefore, it remains to be investigated whether this positive signal was due to a lack of Ab specificity, technical problems such as supersaturating conditions, or whether pig cells express ␣Gal residues in the absence of ␣1,3GT, which are only detected by a subset of Abs.
The removal of ␣Gal led to a strong reduction of complementinduced xenogeneic lysis. Moreover, NK cell-mediated xenogeneic ADCC was reduced, but not totally absent when ␣Gal-negative pEC were used as targets. In line with these results, binding of human serum IgM and IgG molecules was strongly reduced, but not absent on ␣Gal-negative pEC. This strong reduction of binding was observed for all donors tested; however, the magnitude of reduction varied considerably. The remaining NAb binding on ␣Gal-negative pEC, most probably due to non-␣Gal-reactive Abs, including those reactive with N-glycosylneuraminic acid (57), represents an additional hurdle to xenotransplantation (48). Moreover, it has been suggested that the elimination of ␣Gal from porcine cells could lead to the generation of new epitopes such as terminal ␤-galactosyl residues or the N-acetyllactosamin disaccharide, which are recognized by preformed human xenoreactive Abs (58). Surprisingly, the levels of ADCC of ␣Gal-positive PEDSV.15 and ␣Gal KO PED2*3.51 cells using anti-␣Gal-depleted serum differed more than expected. We therefore assume that non-␣Gal Ags recognized by NAb inducing ADCC are possibly expressed to a lesser extent on ␣Gal KO PED2*3.51 as compared with PEDSV.15 cells. A possible explanation for this difference might be the extensive panning procedure using anti-␣Gal Abs affinity purified from baboon serum, which might have led to the elimination of other non-␣Gal Ags on the ␣Gal KO line by contaminating non-␣Gal Abs. In contrast to ADCC, non-␣Gal-specific NAb were not very effective in inducing xenogeneic complement lysis. The amount of bound non-␣Gal Ab required to induce complement lysis was rather high compared with ADCC. This finding correlates with an earlier study indicating that the threshold for NAb/complement-induced cytotoxicity may be higher than the threshold to induce ADCC (59).
An important, but still controversial question is whether carbohydrate determinants, in addition to proteins, can act as target molecules for NK cells. Although ␣Gal is likely to be the most important carbohydrate in xenotransplantation, other carbohydrates could also have an impact on NK cell biology as in response to tumors or viral infections. To date, only little is known about ligands for lectin-like receptors on NK cells and their effects on NK cell functions (60). Whereas direct recognition of ␣Gal by human NK cells leading to adhesion and cytotoxicity was reported by earlier studies (37)(38)(39)61), our results suggest that elimination of ␣Gal on pEC does not resolve NK adhesion and direct anti-porcine NK cytotoxicity. pEC lacking ␣Gal were equally lysed by human NK cells as pEC expressing high ␣Gal levels, and the adhesion of NK cells to PED2*3.51 expressing the important adhesion molecule VCAM-1 (46), but still lacking ␣Gal, was detected on these cells. Therefore, we conclude that there is no direct correlation between the absence of ␣Gal and VCAM-1 and that ␣Gal does not play a crucial role in the adhesion process. In line with our findings, naive human NK cells adhered, activated, and lysed pEC independently of ␣Gal (62), and no differences regarding NK adhesion and lysis were observed upon introduction of ␣Gal in human aortic endothelial cells (63). In contrast, using ␣1,3GT-transfected COS-7 cells, soluble carbohydrates, and F(abЈ) 2 of xenoreactive NAb, Inverardi et al. (37) showed a role for ␣Gal in Ab-independent NK adhesion to and destruction of pEC. In addition, a partial decrease in NK cell-mediated direct cytotoxicity was reported when ␣Gal epitopes were blocked by the mAb M86, the Griffonia simplicifolia lectin, or after treatment of pEC with ␣-galactosidase (39). Nevertheless, the level of NK killing was very low and no ADCC was demonstrated in this study. ␣-galactosidase treatment of pEC had not reduced NK killing in an earlier study (58). Finally, genetical modification of the ␣Gal epitope by expressing ␣-1,2-fucosyltransferase also resulted in a decreased susceptibility to human NK lysis against porcine endothelial cells (38) and fibroblasts (61). The apparent lack of congruence in analyzing the studies on the role of ␣Gal in xenogeneic NK responses is hard to explain. Differences in the experimental protocols, including the grade of activation of the NK effector cells and the source of porcine target cells, may in part be responsible. The present study is the first to test NK cytotoxicity against pEC in the absence of ␣1,3GT gene expression and without the need to treat the cells with either ␣-galactosidase or blocking reagents. We therefore conclude that ␣Gal does not act as a dominant cytotoxicity-inducing NK target molecule. Our results do not exclude a role for ␣Gal in NK recognition; it may be necessary, but not sufficient to interfere with ␣Gal to overcome xenogeneic NK responses.
In summary, the results presented in this work show that a modification of glycosyl epitopes such as the lack of ␣Gal expression on porcine cells provides an effective protection from ␣Gal Abmediated rejection, including complement lysis and ADCC. Non-␣Gal Ab-mediated damage of and direct NK cytotoxicity against porcine targets, however, are not resolved by knocking out the ␣1,3GT gene. These data extend the current knowledge of the mechanisms leading to xenograft rejection; may prompt further preventive and therapeutic strategies; and underline that only a combination of approaches addressing both Ab-and cell-mediated mechanisms of xenograft rejection may facilitate clinical pig-tohuman xenotransplantation.