THE α-1,3-GALACTOSYLTRANSFERASE KNOCKOUT MOUSE

Implications for Xenotransplantation¹

Abstract

Organ xenografts in discordant combinations such as pig-to-man undergo hyperacute rejection due to the presence of naturally occurring human anti-pig xenoantibodies. The galactose α(1,3)-galactose epitope on glycolipids and glycoproteins is the major porcine xenoantigen recognized by these xenoantibodies. This epitope is formed byα(1,3)-galactosyltransferase, which is present in all mammals except man, apes, and Old World monkeys. We have generated mice lacking this major xenoantigen by inactivating the α(1,3)-galactosyltransferase gene. These mice are viable and have normal organs but develop cataracts. Substantially less xenoantibody from human serum binds to cells and tissues of these mice compared with normal mice. Similarly, there is less activation of human complement on cells from mice lacking the galactose α(1,3)-galactose epitope. These mice confirm the importance of the galactoseα(1,3)-galactose epitope in human xenoreactivity and the logic of continuing efforts to generate pigs that lack this epitope as a source of donor organs.

Xenotransplantation offers a potential solution to the worldwide shortage of organs for transplantation. While non-human primate to human transplants have been performed ⁽¹⁾, ethical objections as well as practical and economic considerations preclude the widespread use of primates as organ donors. Most are endangered species which breed poorly in captivity and they may harbor infectious agents which are pathogenic in man. Consequently, interest has focused on the pig as an organ donor. Its erythrocyte diameter and, by implication, its capillary size are similar. Furthermore, pigs have large litters and a short gestation time. However, the immunological consequence of transplantation at this phylogenetic distance is hyperacute rejection due to the presence in human plasma of preformed“naturally occurring” xenoantibodies, and the relative inability of porcine cell surface complement regulators to limit activation of human complement on the surface of porcine endothelial cells^{(2, 3)}. Ensuing endothelial cell damage and loss of the natural anticoagulant surface of the endothelium give rise to local, and then generalized, intravascular coagulation within the graft^{(2, 4)}.

The major xenoantigen on pig cells and tissues which is recognized by xenoantibody in human plasma is galactose α(1,3)-galactose (Gal epitope), a terminal disaccharide on glycoproteins and glycolipids^(5-8). Theα(1,3)-galactosyl-transferase (GalT^*) enzyme that forms the Gal linkage is present in all mammals with the exception of humans, apes, and Old World monkeys (catarrhines) ^{(5, 9)}, where the gene has been inactivated by single-basepair mutations⁽¹⁰⁾. As a result, humans do not express the Gal epitope. Instead, human serum contains high titer antibodies (both IgG and IgM) against the Gal epitope, probably due to continual exposure to intestinal bacterial flora that have the Gal epitope ⁽¹¹⁾. The importance of anti-Gal xenoantibodies was demonstrated initially by competition studies, in which α-galactosyl sugars inhibited binding of human xenoantibodies to porcine endothelial cells ^{(6, 7, 12, 13)}. In other experiments, COS cells, which are derived from Old World monkeys and therefore lack the Gal epitope, were lysed by human serum only after transfection with a construct containing the cDNA for the GalT enzyme⁽¹³⁾. Most significantly, human serum that had been absorbed with COS cells expressing the GalT gene no longer reacted with porcine endothelial cells.

A variety of strategies to eliminate the contribution of anti-Gal antibodies to hyperacute rejection have been proposed. Some are designed to deplete the xenoantibody either nonspecifically by plasmapheresis or specifically by immunoabsorption. Other strategies depend on depleting or eliminating the Gal epitope from the donor. Inactivation of the GalT gene has the potential advantage of eliminating involvement of anti-Gal xenoantibody in hyperacute rejection permanently and completely. The validity of this approach has not been tested. At present, it is not possible to generate pigs lacking the Gal epitope because porcine pluripotent cells, which are essential for gene inactivation, are not yet available. In addition, it has been suggested that deletion of the Gal epitope may be lethal ^{(14, 15)}. As a first step, we have generated mice that do not synthesize the Gal epitope and have used them to examine the consequences of inactivating the GalT gene and to study the role of the Gal epitope in hyperacute rejection.

MATERIALS AND METHODS

Construction of targeting vector. To make the targeting construct, 9.5 kb of 129 mouse genomic DNA containing exons 8 and 9 of the GalT gene were used. A 1.4-kb Sal fragment containing the neo^R resistance gene flanked by yeast FLP recombinase target sites⁽¹⁶⁾ was inserted into a unique SalI site within exon 9. FLP recombinase target sites were included in this targeting construct to investigate whether the neo^R gene could be excised from the genome of GalT^-/- mice using FLP recombinase. Excision of the neo^R gene has not be tested to date. In addition, a 30-basepair linker containing a BglII restriction site and termination condons in all three reading frames was inserted just downstream of the neo^R gene. TheBgl II site provides a convenient restriction site for delineating wild-type and targeted GalT alleles (Fig. 1), and the stop codons ensure that no functional GalT protein would be synthesized after excision of the neo^R gene.

Targeting. One hundred twenty-nine embryonic stem (ES) cells(3×10⁷) were electroporated with XhoI-linearized targeting vector (Fig. 1). After culture for 24 hr, neo^R colonies were selected with G418 (200 μg/ml) for 14 days. Homologous recombinants were detected by Southern analysis of BglII digested genomic DNA using a 0.7-kb EcoRI/XmnI probe, which detects bands of 6.4 and 8.3 kb in the targeted and wild-type alleles, respectively (Fig. 1). Targeting at the 5′ end was verified by polymerase chain reaction (PCR) amplification of a 5.5-kb fragment using a 5′ primer directed to exon 8 and a 3′ primer homologous to sequence near the 3′ end of the neo^R gene (data not shown; the positions of the primers are indicated in Fig. 1A). Targeted cells were microinjected into CBA×C57B16 F2 blastocysts and transferred to pseudo-pregnant mice. ES cell contribution to the coat color was assessed visually, and ES cell-derived sperm contribution was determined by isoenzyme analysis ⁽¹⁷⁾. Progeny were genotyped by PCR using a 3′ primer of the SalI site in exon 9 in combination with either a neo^R-specific primer to generate a 364-bp product specific for the targeted allele, or an exon 9-specific primer 5′ of theSal I site, to generate a product specific for the wild-type allele(data not shown).

Flow cytometric analysis (FACS). Peripheral blood leukocytes(PBLs) and splenocytes were analyzed by FACS. Before staining, red cells were lysed by treatment with NH₄Cl buffer (0.168 M). To detect expression of the Gal epitope, cells were fixed in 4% paraformaldehyde and then incubated with fluoresceinated IB₄ lectin (FITC-IB₄, Sigma, St Louis, MO) at 20 μg/ml on ice for 30 min. To determine reactivity with human serum, unfixed cells were incubated sequentially with 50 μl of normal human serum at the dilutions indicated, and fluoresceinated sheep anti-human IgG or IgM F(ab)₂ fragments (Silenus, Australia). Stained cells were analyzed using a FACScan (Becton Dickinson, CA).

Immunohistology. Tissues were surgically removed and snap-frozen in OCT using isopentane cooled in liquid nitrogen. Sections (4 μm) were air dried onto gelatin-coated slides for 30 min and background staining was minimized by preincubation with blocking buffer containing 10% sheep serum and 2% bovine serum albumin in Tris-buffered saline. Sections were then incubated with either normal human serum or FITC-IB₄ (20 μg/ml), followed by peroxidase-conjugated anti-human IgG and IgM (Dako, Denmark) or anti-FITC Fab fragments (Boehringer Mannheim, Germany), respectively. Bound peroxidase was detected using 3,3′-diaminobenzidine tetra-hydrochloride as the chromogen (DAB; Sigma). All sections were counterstained with Harris's hematoxylin.

C3c binding assay. Splenocytes were prepared by sieving, pelleted(500×g, 5 min, 4°C), and resuspended in 0.168 M NH₄Cl to lyse red blood cells. Cells were washed twice and incubated in 200 μl of normal human serum in concentrations varying from 2.5% to 10% for 10 min at 37°C, washed again, and then incubated with FITC-conjugated anti-human C3c (Dakopatts, Denmark) at 1:50 dilution at 4°C for 30 min. Controls consisted of cells incubated with 10% normal human serum treated with 10 mM EDTA and cells incubated with C3c antibody alone. Stained cells were analyzed using a FACScan.

RESULTS

Production of mice homozygous for an inactivated GalT gene. The GalT targeting construct was prepared by interrupting exon 9 of the mouse GalT gene (coding for the catalytic domain of the enzyme) with the neomycin resistance gene (neo^R) (Fig. 1A). The construct was electroporated into murine ES cells. After selection for G418 resistance, 26 of 227 (11%) resistant colonies were identified by Southern analysis to contain a 6.4-kb BglII fragment diagnostic for targeting at the 3′ end (Fig. 1B). Precise recombination at the 5′ end of the construct was confirmed by long-range PCR (data not shown). Cells from 16 independently targeted clones were injected into blastocysts to generate chimeric mice, and mice from 3 of these transmitted the interrupted GalT allele through the germ line. Chimeras from two of these ES cell lines were used to generate mice homozygous for the inactive GalT gene(GalT^-/- mice) for investigation.

Phenotypic effects of deleting the GalT gene. The GalT^-/- mice were born healthy and were fertile. However, two phenotypic effects of eliminating GalT activity were apparent. First, matings between mice heterozygous for the inactivated GalT gene produced genotype ratios that deviated significantly from the predicted Mendelian 1:2:1 (wild type:heterozygote:homozygote) ratio (Table 1). Second, cortical cataracts developed in all GalT^-/- mice at 4-6 weeks of age. These were the only abnormalities detected. Histological examination of organs, including heart, lung, liver, kidney, brain, pancreas, and aorta, with both hematoxylin and eosin and periodic acid-Schiff stain, failed to identify any other abnormalities. The oldest GalT^-/- mice, now 9 months old, appear normal except for cataract formation.

GalT^-/- mice do not express the Gal epitope. GalT mRNA was not detected by reverse transcriptase PCR analysis in GalT^-/- mice, but was found in wild-type and heterozygous mice (data not shown). Organs examined included heart, lung, liver, and kidney. Similarly, the Gal epitope was not detected in GalT^-/- mice. Using FACS, PBL and splenocytes from GalT^-/- mice did not stain with the lectin IB₄, specific for terminal α-galactosyl residues. In contrast, cells from wild-type mice reacted strongly with the lectin (Fig. 2). Histological examination showed that IB₄ lectin reacted strongly with the endothelial cell surfaces in heart, lung, kidney, and liver from wild-type and heterozygous mice (Fig. 3). There was no staining of any tissues from the GalT^-/- mice. Similarly, affinity-purified human anti-Gal antibodies bound to wild-type and heterozygous tissue, but not to tissue sections from GalT^-/- mice (results not shown).

Effect of GalT gene inactivation on reactivity with human serum. Differences in the ability of human serum to react with splenocytes from wild-type and GalT^-/- mice were investigated by FACS. Cells were incubated with human serum (0.5-20%) and bound IgG and IgM were detected with FITC-labeled anti-human IgG and IgM. Median channel fluorescence was used to compare levels of antibody binding. Both IgG and IgM fluorescence values for GalT^-/- cells were reduced by approximately 60% compared with wild-type cells over a range of serum concentrations (Fig. 4). Maximum IgM and IgG binding occurred at 5% and 1.25% serum, respectively. At very low concentrations (below 0.25%), there was minimal binding of antibody of either class to GalT^-/- cells, but there was still strong binding of both antibody classes to wild-type splenocytes. At high serum concentrations, a reduction in the degree of staining was observed, characteristic of a prozone.

To determine whether the reduced binding of xenoantibody to splenocytes of GalT^-/- mice was reflected in reduced activation of complement, splenocytes from wild-type and GalT^-/- mice were incubated with human serum (2.5-10%) and deposition of C3c, a component of activated complement, was determined by FACS analysis. Figure 5 shows that significantly less C3c (30-50%) was deposited on splenocytes from GalT^-/- mice than cells from wild-type mice.

DISCUSSION

The first barrier to successful discordant xenotransplantation, such as pig to human, is hyperacute rejection initiated by naturally occurring xenoantibody and complement. Production of transgenic pigs expressing the genes for the human complement regulatory molecules (CD46, CD55, and CD59) offers a partial solution to hyperacute rejection^{(3, 18-20)}. For example, rejection was delayed in baboons when pig heart grafts expressing CD59 were used⁽³⁾. The possibility of using a genetic approach to provide a permanent solution to the xenoantibody component of hyperacute rejection is the focus of this study.

The major target of human xenoantibodies in pigs is the Gal epitope, the product of the GalT enzyme, which is not expressed in humans. Inactivating the GalT gene by gene targeting has been proposed as a means of totally and permanently eliminating the Gal epitope ^{(15, 21, 22)}, thereby markedly reducing the involvement of xenoantibodies. Although this technology is not yet possible in pigs, we have inactivated the GalT gene in mice by gene targeting to investigate the xenoresponse of discordant organs that lack the Gal epitope.

A GalT-targeting construct was used to inactivate one of the GalT alleles in murine ES cells by homologous recombination. Mice chimeric for the ES cell genotype were generated and bred to obtain mice homozygous for the inactivated GalT gene. Homozygous mice obtained from two independently targeted ES cell colonies were investigated.

Mice homozygous for the inactivated GalT gene were viable and fertile, refuting the previous speculation that deletion of the GalT gene would be lethal ^{(14, 15)}. However, matings between mice heterozygous for the inactivated GalT gene resulted in a disturbance of the expected 1:2:1 ratio of wild type, heterozygote, and homozygote with reduction in the transmission of the targeted allele, suggesting that selective pressure operates against these mice during development. It is unlikely that the Gal^- phenotype alone was responsible for this selective disadvantage, since a large number of liveborn GalT^-/- mice were produced.

Development of cortical cataracts within 4-6 weeks of birth was the only pathological abnormality detected in GalT^-/- mice. These animals, some of which are now 9 months old, remain otherwise as healthy and fecund as wild-type littermates. Reports of numerous other single-gene defects causing cataracts in mice ⁽²³⁾ suggest that the lens, at least in mice, may be highly sensitive to biochemical changes. It is possible that this would also occur in pigs, as Galili et al. ⁽⁵⁾ have demonstrated that the epithelial cells of the pig lens express the Gal epitope. The production of blind pigs may present ethical and animal husbandry problems depending on the time of onset. Consequently, we are exploring genetic approaches to correct the defect.

Failure to detect IB₄ binding to cells and tissues of mice homozygous for the interrupted GalT gene demonstrates that these mice do not express the Gal epitope and indicates further that there are no other enzymes that can perform the same or similar function. This is an important observation, as gene targeting does not always result in the expected phenotype, and Samuelsson et al. ⁽²⁴⁾ postulated that redundancy of the glycosyltransferases may circumvent GalT inactivation.

In FACS analysis, the binding of naturally occurring IgG and IgM xenoantibodies present in human serum to PBL of GalT^-/- mice was reduced by approximately 60% compared with cells from wild-type mice. This suggests that approximately 60% of human anti-mouse xenoantibodies are directed against the Gal epitope. Eliminating the Gal epitope from mice also reduced the activation of complement (assessed by deposition of C3c) by 30-50%. Together, these studies indicate that anti-Gal antibodies are the dominant anti-mouse xenoantibodies, and that a substantial reduction in complement activation is achieved when anti-Gal antibodies are not involved in the xenoresponse.

The specificities of the residual (approximately 40%) non-Gal human anti-mouse xenoantibodies are unknown and the GalT^-/- mice will be a valuable tool in their identification. While anti-Gal antibodies represent at least 60% of human anti-mouse xenoantibodies, they are estimated to account for 70% to 90% of human anti-pig xenoantibodies ^{(6, 25)}. Whether this difference in relative anti-Gal antibody proportions is real, or due to the different approaches used to estimate them, has not been established. Similarly, it is not yet known whether the non-Gal xenoantigens in mice are the same as those in the pig. However, preliminary indications suggest that the xenoresponse to pig organs lacking the Gal epitope will be reduced, since no reactivity to pig endothelial cells was detected in human serum depleted of anti-Gal antibodies ⁽¹³⁾.

In summary, these results indicate that xenoantibody binding and complement activation is reduced when normal human serum is reacted with discordant cells and tissues lacking the Gal epitope, and justify further investigation of strategies for eliminating the Gal epitope in the pig.

Acknowledgments. The technical assistance of Steve McIlfatrick and Andrew Cowan is gratefully acknowledged.

Figure 1: Disruption of the mouse GalT gene by gene targeting. (A) A targeting construct was generated from fragments of genomic DNA from the exon 8 and 9 region of the mouse GalT gene. The diagram shows the wild-type GalT gene region, the targeting construct, and the disrupted allele generated by homologous recombination with the targeting construct. The neo^R and the flanking FLP recombinase target sites (F) are indicated. Also shown are the positions of the probe used for Southern analysis, and primers used for PCR analysis. (B) A phosphorimager representation of a Southern blot of ES cell genomic DNA hybridized with the probe shown in A. The 8.3-kb band represents the wild-type GalT allele, and the 6.4-kb band is diagnostic of the disrupted allele.

Figure 2: FACS profiles of splenocytes of GalT^-/- and wild-type mice stained with fluoresceinated IB₄ lectin.

Figure 3: Histological sections of kidney (A and B), liver (C and D), lung(E and F), and heart (G and H) from wild-type (left-hand column) and GalT^-/- mice (right-hand column) stained with IB₄ lectin by the avidin-biotin-peroxidase technique.

Figure 4: FACS binding assays of human serum IgM and IgG with splenocytes of GalT^-/- and wild-type mice. Results are mean ± SEM of 3 separate experiments.

Figure 5: FACS binding assay of C3c deposition on splenocytes of GalT^-/- (broken line) and wild-type (solid line) mice after reaction with human serum at the concentrations indicated. Results are mean ± SEM of 3 separate experiments.