B4GALNT2 and xenotransplantation: A newly appreciated xenogeneic antigen.

Abstract Analysis of non‐Gal antibody induced after pig‐to‐baboon cardiac xenotransplantation identified the glycan produced by porcine beta‐1,4‐N‐acetyl‐galactosaminyltransferase 2 (B4GALNT2) as an immunogenic xenotransplantation antigen. The porcine B4GALNT2 enzyme is homologous to the human enzyme, which synthesizes the human SDa blood group antigen. Most humans produce low levels of anti‐SDa IgM which polyagglutinates red blood cells from rare individuals with high levels of SDa expression. The SDa glycan is also present on GM2 gangliosides. Clinical GM2 vaccination studies for melanoma patients suggest that a human antibody response to SDa can be induced. Expression of porcine B4GALNT2 in human HEK293 cells results in increased binding of anti‐SDa antibody and increased binding of Dolichos biflorus agglutinin (DBA), a lectin commonly used to detect SDa. In pigs, B4GALNT2 is expressed by vascular endothelial cells and endothelial cells from a wide variety of pig backgrounds stain with DBA, suggesting that porcine vascular expression of B4GALNT2 is not polymorphic. Mutations in B4GALNT2 have been engineered in mice and pigs. In both species, the B4GALNT2‐KO animals are apparently normal and no longer show evidence of SDa antigen expression. Pig tissues with a mutation in B4GALNT2, added to a background of alpha‐1,3‐galactosyltransferase deficient (GGTA1‐KO) and cytidine monophosphate‐N‐acetylneuraminic acid hydroxylase deficient (CMAH‐KO), show reduced antibody binding, confirming the presence of B4GALNT2‐dependent antibodies in both humans and non‐human primates. Preclinical xenotransplantation using B4GALNT2‐deficient donors has recently been reported. Elimination of this source of immunogenic pig antigen should minimize acute injury by preformed anti‐pig antibody and eliminate an induced clinical immune response to this newly appreciated xenotransplantation antigen.


| INTRODUC TI ON
Xenotransplantation (XTx) is limited by recalcitrant antibodymediated rejection occurring either hyperacutely immediately after transplant (HAR) or at later times, referred to as delayed xenograft rejection. [1][2][3] These rejection mechanisms result from the abundance of human and non-human primate (NHP) antibodies directed to the classic xenogeneic antigen galactose alpha 1,3 galactose (Gal) which leads to chronic or induced antibody-mediated vascular endothelial cell (EC) injury or activation. 4,5 Gal is not expressed in humans or Old World NHPs but is expressed at high levels in porcine tissues. Genetically modified pigs, with a mutation in the GGTA-1 locus (GTKO), do not express the Gal antigen. 6 The introduction of GTKO donor organs eliminated anti-Gal-mediated xenograft rejection, but did not eliminate antibody-mediated rejection and instead highlighted the importance of antibody directed to non-Gal pig antigens. 7,8 Non-Gal antibody in human and NHP serum is reactive to both protein and carbohydrate antigens. [9][10][11][12][13] The currently identified immunogenic EC carbohydrate antigens include Gal, glycans modified with N-glycolylneuraminic acid (Neu5Gc), and a carbohydrate antigen (SDa) produced by the porcine β1,4 Nacetylgalactosaminyltransferase-2 (B4GALNT2). The human blood group A antigen is also potentially immunogenic 14 ; however, high rates of A-type blood group polymorphism in the pig permit exclusion of this antigen by selective breeding. 15,16 Collectively, antibody reactivity to the 3 major xenogeneic glycans, Gal, Neu5Gc-modified glycans, and SDa, accounts for the majority of preformed human anti-pig antibody reactivity. 17,18 Humans, but not NHPs, make an array of antibodies to Neu5Gc-modified glycans. 19,20 This antibody reactivity is expected to contribute to clinical xenograft rejection, but determining its impact remains difficult due to the absence of anti-Neu5Gc antibody in experimental NHP models. There have been excellent recent reviews on Neu5Gc and the potential immunogenicity of Neu5Gc-modified glycans in XTx. [21][22][23] Less is known about the expression and immunogenicity of the glycan produced by porcine B4GALNT2. The clinical contribution of B4GALNT2 and SDa, the glycan it synthesizes, has been recently reviewed. 24 The purpose of this review is to summarize the current experimental and clinical information on B4GALNT2 gene expression and the SDa antigen, with an emphasis of its potential impact on future clinical use of porcine organs and XTx.

| THE S DA H UMAN B LOOD G ROUP
The SDa blood group, synthesized by the human B4GALNT2 glycosyltransferase, was independently identified 50 years ago by Renton et al 25 and by Macvie et al. 26 The initial identification was based on a small set of unrelated serum samples which agglutinated the majority of Caucasian red blood cell (RBC) samples. The degree of RBC agglutination varied widely, presenting a mixed field reaction with variably sized agglutinates against a large background of free RBCs. Further analysis, using agglutination inhibition, identified SDa+ expression in human saliva, milk, feces, and urine with more than 50% of people with SDa− RBCs being SDa+ in saliva or urine. 26 The initial descriptions of anti-SDa antibody, based on RBC agglutination studies, indicated only 1%-2% of individuals produced anti-SDa IgM which weakly agglutinated SDa+ RBCs, preferentially at low temperature. 25,26 The apparent frequency of anti-SDa antibody increased, however, when cells with stronger SDa expression are used. Clear evidence supporting the classification of SDa as a high-frequency polyagglutinable antigen 28  O-linked glycans, and paraglobosides. 24 For each of these, the SDa trisaccharide (Table 1) present on human red blood cells (RBCs), rare CAD RBC protein, and glycolipids, 41,42 and the major urinary mucin protein Tamm-Horsfall glycoprotein (THGP), 43 is the terminal B4GALNT2-dependent structure. The SDa epitope is commonly detected by reactivity to DBA 30,44 or by anti-SDa monoclonal antibodies. 40,45,46 The SDa trisaccharide present on the GM2 ganglioside (Table 1) is synthesized by the distinct B4GALNT1 transferase which shows only limited homology with B4GALNT2. 47 This is analogous to the relationship between rat GGTA-1, which adds Gal antigen to glycoproteins, and iGB3 synthase, which adds the Gal antigen to iGb3, but not glycoproteins. 48 Excessive accumulation of GM2 ganglioside, due to a deficiency in β-N-acetylhexosaminidase, contributes to ongoing cellular damage associated with Tay-Sachs disease. Mutations in B4GALNT1 have been identified, which lead to deficiency in GM2 synthesis and are associated with a hereditary spastic paraplegia. 49 This suggests that targeted mutation of the porcine B4GALNT1 gene, to eliminate SDa antigen on GM2 gangliosides, may have deleterious effects on donor animals. Mutation of B4GALNT1 would not affect SDa modification of glycoproteins or of other glycolipids that are present in CAD RBCs. For xenotransplantation, the utility of mutating B4GALNT1 remains uncertain as the primary immune stimulation, at least in NHPs, appears to be from B4GALNT2-dependent antigens.

| INDUC TI ON OF ANTI -S DA ANTIBODY
The GM2 ganglioside is a tumor-associated carbohydrate antigen (TACA) detected at higher levels in malignant melanoma and other forms of cancer. 50

| E XPRE SS I ON OF B G ALNT2 AND S DA
Human B4GALNT2 is predominantly expressed in gastrointestinal (GI) epithelium of the colon with lower levels of expression, evident by Northern blot, in the kidney, ileum, stomach, and rectum. 47,60,61 These sites also show reactivity to DBA or anti-SDa antibody. 62 Consistent with this analysis, B4GALNT2 enzyme has been purified from human and guinea pig kidney 36,38 and human and porcine large intestine. 35,39 There is also low expression of B4GALNT2, detected by reverse transcription PCR, in most other human tissues. 47 This very low expression does not always correlate with detectable SDa staining; however, this does not mean that the antigen is not present as human RBCs exhibit widely variant levels of SDa, but only the highest levels of expression react with DBA. 29,30 Human mast cells, present in connective tissues, may also express B4GALNT2 as they show strong, blood group-independent reactivity with DBA. 63 There is a significant loss of B4GALNT2 expression in colonic tumors compared to normal tissue. 64 The reduction of B2GALNT2 expression often correlates with a change in promoter methylation 65,66 and with an increase in the synthesis of sialyl Lewis antigens. 60 The increase in sialyl Lewis antigens appears to result from the reduction in B4GALNT2 expression, which alters the balance of competition between B4GALNT2 and fucosyltransferases in colonic tumors for the same acceptor substrate. 24 Expression of B4GALNT2 and SDa in other species is chiefly characterized in mice and as a whole has not been systematically studied in other mammals.

| THE B 4 G ALNT2 G ENE IN H UMAN S AND NHPS
The murine B4GALNT2 gene was isolated from a cDNA expression library made from a cytotoxic T lymphocyte cell line. 40 This expression library was screened for reactivity to hybridoma IgM antibodies (CT1 and CT2) which bound activated cytolytic T lymphocytes, blocked target cell lysis, and reacted with the human SDa blood group antigen. 77 The human B4GALNT2 gene was identified based on homology with the mouse gene. 47 Both genes are composed of 11 coding exons. The murine B4GALNT2 gene produces a single transcript, whereas the human B4GALNT2 gene uniquely produces alternative splice variants resulting in a short and long form of the human protein. 47 The long form of human B4GALNT2 includes an unusual 67 amino acid cytoplasmic domain which may affect cellular localization of this enzyme. The significance of this long isoform and its prevalence in non-human primates remains under investigation.
In both humans and mice, the encoded B4GALNT2 protein is a type II transmembrane protein containing an acidic DXD amino acid motif conserved in glycosyltransferase using UDP sugar as a donor substrate. The human protein shows 74% amino acid identity to mouse polypeptide.
Although human B4GALNT2 expression is essential for synthesis of the SDa antigen, the molecular mechanism, which affects polymorphic SDa expression in human RBCs, has not been identi-

| P ORCINE B 4 G ALNT2
The porcine B4GALNT2 gene was originally identified by screening a porcine EC cDNA library for non-Gal antigens expressed in human HEK293 cells. 10 Purified IgG enriched in non-Gal reactivity, derived from pig-to-baboon cardiac XTx recipients, was used to identify human HEK293 cell-expressing library-encoded non-Gal porcine antigens on the cell surface. Six porcine gene products were identified. Most of these non-Gal porcine target genes corresponded to well-known EC membrane proteins (PROC, CD9, CD46, CD59 and ANXA2). HEK293 expression of these porcine gene products is subject to HEK-derived glycosylation, suggesting that non-Gal baboon IgG reactivity to these cell lines is directed to porcine peptide epitopes and not carbohydrates, which would be present on all HEK293 cells. In contrast, the porcine B4GALNT2 gene is a Golgi expressed glycosyltransferase not typically expressed on the cell surface. This indicates that B4GALNT2 expression in HEK293 cells altered the glycosylation of HEK293 cell membrane proteins to create a non-Gal carbohydrate antigen.
The porcine B4GALNT2 cDNA is derived from an 11-exon gene and encodes a protein with 76% amino acid identity to the human protein and less than 50% amino acid identity to human B4GALNT1.
Expression of porcine B4GALNT2 in HEK293 cells (HEK-B4T) results in increased reactivity to anti-B4GALNT2 antibody, anti-SDa (KM694), and DBA reactivity. 9 There is also a 20-fold enhancement of HEK-B4T cell, compared to HEK293 cell, sensitivity to antibodydependent complement-mediated lysis using serum from cardiac XTx recipients.
In agriculture-based strains of pigs, mRNA for B4GALNT2 is expressed in ECs, peripheral blood mononuclear cells (PBMCs), in the GI system and major vascularized organs. 9 Porcine EC expression of B4GALNT2 is independent of AO blood group or GGTA-1 (GTKO) status. Pig ECs are strongly agglutinated by DBA lectin, and cultured ECs from agricultural strains, Yucatan hairless minipigs, 87 and Panepinto micropigs 88 have been reported to bind DBA, independent of AO blood group status. In vivo vascular B4GALNT2 EC expression, measured by DBA stain, may not be uniform in all vascular beds but is prominent in cardiac capillaries, large renal blood vessels, and glomerular ECs, in reticuloendothelial cells of the liver 9 and ECs of the femoral artery. 89 Cardiac capillary staining with DBA and EC expression of B4GALNT2 is also evident in Gottingen minipigs ( Figure 1A,B). This is consistent with previous studies of ECs from variant porcine backgrounds and suggests that porcine EC expression of B4GALNT2 is not polymorphic and will be commonly present in most pig strains.
Porcine B4GALNT2 expression is not confined to ECs and is present in PBMCs, spleen, 9 and epithelial cells such as PK15 ( Figure 1B).
Porcine primordial germ cells are also reported to express SDa and stain with anti-SDa antibody 90 . In contrast, cultured bovine ECs do not express B4GALNT2 ( Figure 1C), which is comparable to human and most murine strains, suggesting that clinical products produced from bovine tissues, such as heart valves, may be less likely to express B4GALNT2-dependent glycans.
Porcine B4GALNT2 expression in HEK-B4T cells results in increased anti-SDa (KM694) and DBA reactivity; however, porcine ECs, which are strongly agglutinated by DBA, do not bind this anti-SDa monoclonal antibody. 9 A similar disparity between lectin and antibody binding was reported for ovine glycoproteins secreted by granulosa cells. 75 The strict specificity of KM694, compared to DBA, suggests that on porcine cells, variation in the structure or presentation of the glycan(s) produced by porcine B4GALNT2 may occur. What is clear is that NHP non-Gal antibody reactivity to HEK-B4T cells can be eliminated by pre-absorption with porcine ECs but not by human ECs, which do not express B4GALNT2 or the SDa  antigen. 9 This indicates that the porcine EC and HEK-B4T cells share a common, B4GALNT2-dependent, non-Gal antigen. In NHPs after cardiac XTx with minimal immune suppression, non-Gal antibody with preferential binding to HEK-B4T cells, compared to control HEK cells, is consistently induced. 10 This immune response is also evident in GTKO pig-to-baboon cardiac XTx recipients with full immune suppression, when a non-Gal antibody response, determined by antibody reactivity to GTKO pig ECs, is also present.

| MUTATI ON S IN B 4 G ALNT2
Mutations in B4GALNT2 have been produced in mice (John Lowe, Consortium of Functional Glycomics, 69 ) and pigs. 18 The mutant mice are generally healthy and show no histological changes to the major organs but do show variations in the level of neutrophils, plasma cells, and some T-cell subsets. 24 Mutant mice lack DBA staining in GI epithelia and exhibit changes in microfloral composition. 73 The loss of B4GALNT2 expression in the GI is associated with greater resistance to Salmonella infection. 74  This combination of porcine mutations, however, may have unexpected effects on the glycome which could affect human antibody reactivity in unexpected ways. This may contribute to the rather large apparent reduction in human anti-SDa IgG binding to these cells, as IgG is not a frequent isotype for human anti-SDa antibody.
The triple knockout pigs are apparently healthy but, to our knowledge, have only been produced through nuclear transfer technology.

| CON CLUS IONS
The SDa blood group, produced by the B4GALNT2 enzyme, is a recently appreciated xenogeneic antigen. The SDa glycan is commonly expressed in human GI epithelial cells and at widely variant levels in human RBCs and other tissues and fluids. Despite this antigen expression, most humans make low levels of cold reactive anti-SDa IgM making SDa a polyagglutinable red cell antigen.
Recent genetic engineering of the porcine B4GALNT2 locus confirms the presence of preformed NHP and human antibody to B4GALNT2-dependent antigens. Although the SDa blood group is not a significant transfusion risk, the expression of B4GALNT2 in porcine ECs, the induced antibody response seen in NHPs and the results of cancer vaccination studies suggest that B4GALNT2dependent pig glycans may be immunogenic in humans. The major known xenogeneic glycans (Gal, Neu5Gc-modified glycans, and SDa) account for the majority of preformed human anti-pig antibody reactivity, with residual antibody reactivity apparently binding to a restricted set of SLA antigens. 17 There remains, however, the possibility of additional immunogenic glycans and proteins.
Identifying the B4GALNT2-dependent non-Gal antigens using serum from XTx recipients with minimal immune suppression underscores the utility of such transplants and suggests that future similar studies using "triple knockout" glycan-depleted donor organs will be useful for further identification of residual preformed and induced non-Gal antibody specificities.