Increased Immunogenicity of Tumor Vaccines Complexed with Anti-Gal: Studies in Knockout Mice for α1,3Galactosyltransferase¹

Studies on the immune response to tumor vaccines have indicated that, to achieve effective activation of tumor-specific naive T cells, the vaccinating antigens must be effectively taken up by APCs.³ These cells migrate thereafter from the vaccination site to the draining lymph node, process the TAAs, and present the TAA peptides in association with MHC molecules. Together with costimulatory signals provided by the APCs, the presented TAA peptides activate naive tumor-specific T cells within the lymph nodes. Only after they are activated, tumor-specific T cells migrate from the lymph nodes into the circulation to seek and destroy metastatic tumor cells expressing the TAAs (1, 2, 3, 4). In contrast, naive T cells exposed to TAAs on the tumor cells in the absence of costimuli, are anergized, thus enabling the tumor to escape immune surveillance (5, 6). Furthermore, no immune response occurs at the vaccination site, and the transport of vaccines to the draining lymph nodes is imperative for the activation of tumor-specific T cells (1, 2, 3, 4). All these studies underscore the major significance of effective uptake of tumor vaccines by APCs.

Developing methods for increasing the uptake of tumor vaccines by APCs is of particular significance when autologous tumor cells are to be used as vaccines. Many human TAA molecules that may serve as vaccines have not been identified or isolated. Moreover, the combination of the various TAAs varies in different tumors of the same type (7, 8, 9). Therefore, autologous tumor cells have been considered as potentially one of the best sources for vaccines directed against the metastasizing tumors (8). These considerations have been supported by studies showing that dendritic cells pulsed with acid-eluted autologous TAA peptides served as a vaccine that effectively protected against challenge with live tumor cells (10, 11). Similarly, autologous dendritic cells transfected with tumor mRNA served as effective vaccines (12). The use of such vaccines requires the in vitro growth of autologous dendritic cells for the subsequent pulsing or transfection. We have developed a simple alternative method for targeting autologous tumor vaccines to APCs in vivo by the in situ complexing of tumor vaccines with the natural anti-Gal antibody (13, 14).

Previous studies have shown that immunogenicity of bacterial and viral vaccines could be increased by the formation of immune complexes with IgG antibodies that target the vaccine to APCs. The Fc portion of the complexed IgG binds to Fcγ receptors on APCs and induces effective uptake of the vaccine by the APCs (15, 16, 17, 18, 19). On the basis of this principle, we have proposed that human autologous and allogeneic tumor vaccines may effectively be targeted to APCs by the in situ formation of immune complexes with the natural anti-Gal antibody (13, 14). This antibody, which constitutes 1% of circulating IgG in humans (20), interacts specifically with the carbohydrate epitope Galα1,3Galβ1,4GlcNAc-R (termed the α-gal epitope; Refs. 21 and 22). The α-gal epitope is produced in large amounts on cells of nonprimate mammals and New World monkeys (monkeys of South America) by the glycosylation enzyme α1,3GT (23, 24). This epitope is completely absent in humans, apes, and Old World monkeys (monkeys of Asia and Africa) because these species lack α1,3GT (23, 24).

In vivo binding of anti-Gal to α-gal epitopes on xenografts, such as pig organs, induces their rejection in humans and monkeys (25, 26, 27). Similarly, in vivo binding of anti-Gal to α-gal epitopes on envelope glycoproteins of retroviral vectors used for gene therapy induces their destruction (28, 29, 30). This immunological potential of anti-Gal can be exploited for the in situ formation of immune complexes with autologous tumor vaccines engineered to express α-gal epitopes. We have previously shown that binding of anti-Gal to human tumor cells engineered to express these epitopes, indeed, resulted in excessive uptake of the cells by macrophages (13, 14).

Until recently, the efficacy of tumor vaccines expressing α-gal epitopes could not be tested in an experimental small animal model because mice or rats synthesize α-gal epitopes, thus they do not produce anti-Gal. In contrast, Old World monkeys as rhesus monkey and baboon are capable of producing anti-Gal (24), but there are no syngeneic tumor models in primates. The cloning of the mouse α1,3GT gene (31, 32) has enabled the generation of KO mice for this gene (33). These mice lack α-gal epitopes and are not immunotolerant to it (33, 34). Recently, we have shown that by immunizing with RRBC membranes, these mice can produce anti-Gal IgG molecules in titers similar to those observed in humans (35). Here, we have used these mice as an experimental animal model for testing the hypothesis of increased immunogenicity of tumor vaccines engineered to express α-gal epitopes. The tumor model used in this study is a highly tumorigenic and very poorly immunogenic subclone of the mouse B16 melanoma, termed B16-BL6 or BL6 (36, 37). We found this tumor to lack α-gal epitopes because of the suppression of α1,3GT gene expression (38). In contrast, most other mouse tumors express an abundance of α-gal epitopes (39). Thus, the carbohydrate make-up of BL6 cells simulates that of human tumor cells, which also lack α-gal epitopes.

We found that immunization of KO mice with irradiated BL6 cells engineered to express α-gal epitopes results in the generation of a protective cellular immune response, which prevents tumor development in a significant proportion of the mice challenged with live BL6 cells. In contrast, immunization with parental BL6 cells that lack α-gal epitopes does not elicit any protective cellular immune response.

KO mice for α1,3GT and WT mice [C57BL/6 × DBA/2J × 129sv (H-2^b × H-2^d); Ref. 33] were received from Drs. J. B. Lowe and A. Thall at the University of Michigan (Ann Arbor, MI) and were bred in our animal facility. All mice were maintained under standard conditions according to institutional guidelines.

The BL6 cell line is a poorly immunogenic subclone of the B16 murine melanoma cell line and was originally isolated by its repeated invasion of the bladder wall of C57BL/6 mice (40). We previously performed stable transfection of BL6 cells with murine α1,3GT cDNA and isolated a subclone that continuously expressed α-gal epitopes (41). The transfected BL6 cells (designated BL6_αGT) display cell surface characteristics that are similar to those of BL6 cells, except for the expression of α-gal epitopes on the former cells (41). We further found BL6_aGT cells to express ≈2 × 10⁶ α-gal epitopes per cell (42). The tumor cells were grown in tissue culture medium consisting of DMEM supplemented with 10% fetal bovine serum (Sigma Chemical Co., St. Louis, MO), 50 units/ml penicillin, 50 mg/ml streptomycin, and 2 mm l-glutamine (Life Technologies, Inc., Grand Island, NY).

Prior to experimental procedures, 4-week-old KO mice received i.p. injections of 3 × 10⁸ RRBC membranes every 2 weeks for 6 weeks to boost anti-Gal IgG titers to levels similar to those of humans. RRBCs were used for this purpose because the α-gal epitope is the major carbohydrate on the these red cells (43, 44). RRBC membranes were prepared by lysing packed RRBCs (Hemostat, Dixon, CA) with sterile water. After centrifugation at 15,000 rpm, membranes were washed thoroughly and resuspended in an equal volume of sterile saline for a final 50% (v/v) stock suspension.

Two weeks after the third RRBC membrane injection, sera were tested for anti-Gal IgG by ELISA using synthetic α-gal epitopes linked to BSA (α-gal-BSA; Dextra Laboratories, Reading, United Kingdom) as a solid-phase antigen as described previously (42). Serial 2-fold dilutions of mouse serum were made (starting at 1:50) and were added to 96-well microtiter plates precoated with 10 mg/ml α-gal-BSA in carbonate buffer (pH 9.5) and blocked with 1% BSA in carbonate buffer. The plates were incubated for 2 h at room temperature and then washed five times with PBS/0.05% Tween 20 goat antimouse IgG-horseradish peroxidase (1:500; Accurate Chemical, Westbury, NY) was added as a secondary antibody. Plates were incubated for 1 h at room temperature then washed with PBS/Tween 20. Orthophenylenediamine (1 mg/ml; Sigma) was added, and color development was measured on an ELISA reader at 492 nm. Only mice displaying anti-Gal IgG titers similar to those found in humans (i.e., titers of >1:400) were used further for the analysis of efficacy of tumor vaccines expressing α-gal epitopes.

Anti-Gal from 2 ml of pooled sera from KO mice immunized with RRBC membranes was purified by affinity chromatography on a column of synthetic α-gal epitopes linked to silica beads (Synsorb 115; Chembiomed Ltd., Edmonton, Alberta, Canada; Refs. 22 and 23). Anti-Gal was eluted from the column using glycine-HCl (pH 3.0), brought to a volume of 2 ml, and neutralized with 0.1 n NaOH. The concentration of the antibody was found to be 170 μg/ml. The restricted specificity of the purified anti-Gal for α-gal epitopes was confirmed by its binding to these epitopes on fetuin (referred to here as α-gal-fetuin) and the inability of this antibody to bind to fetuin with SA epitopes (referred to here as SA-fetuin; Ref. 14). Production of α-gal-fetuin is described below. The binding of anti-Gal was determined by ELISA as described above.

KO mice received an i.p. injection of a 10% Brewer’s thioglycolate solution (Difco Laboratories, Detroit, MI). Five days after injection, peritoneal macrophages were isolated by peritoneal lavage using PBS containing 2 units/ml heparin. Macrophages were washed three times in PBS and were incubated on glass coverslips (12 mm) in a 24-well tissue culture plate in complete medium at 2 × 10⁵ cells/well. The macrophages were incubated with RRBCs (1% v/v), with or without mouse anti-Gal at a 1:50 dilution, in a final volume of 0.5 ml of complete medium for 3 h at 37°C and 5% CO₂. After incubation, noninternalized RRBCs were removed by washing the coverslips three times with PBS, and adherent RRBCs were lysed by hypotonic shock. Coverslips were stained with Giemsa/Wright solution (Leukostat; Fisher Scientific, Pittsburgh, PA) and observed under light microscopy for phagocytosis of RRBCs by macrophages.

Substitution of the carbohydrate epitopes SA-Galβ1–4GlcNAc-R on the bovine fetal serum protein fetuin with α-gal epitopes was carried out according to a previously described procedure (14). Native fetuin with SA (referred to here as SA-fetuin; Sigma) was desialylated by incubation for 2 h at 80°C with 0.05 m sulfuric acid. The desialylated fetuin was dialyzed and then incubated with recombinant α1,3GT (10 μg/ml) and 2 mm UDP galactose, as sugar donor. Subsequently, fetuin with several α-gal epitopes (referred to here as α-gal-fetuin) was isolated by affinity chromatography on Sepharose coupled with Bandeiraea simplicifolia I lectin (Vector Laboratories, Burlingame, CA) and eluted with 50 mm melibiose, with subsequent dialysis.

Differential interaction of anti-Gal with BL6_αGT cells but not with BL6 cells was demonstrated by two independent methods.

BL6 or BL6_αGT cells were incubated with anti-Gal purified from serum of KO mice (10 μg/ml) for 2 h at 4°C. Subsequently, the cells were washed and incubated with phycoerythrin-coupled goat antimouse IgG (PharMingen, San Diego, CA) for 1 h at 4°C. After additional washing the cells were fixed and analyzed in FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA).

Suspensions of 1 × 10⁶ BL6 or BL6_αGT cells/ml were incubated with a 1:50 dilution of serum from KO mouse or WT mouse, in PBS containing 1% BSA for 1 h at 4°C, with occasional stirring. After incubation, cells were washed three times in PBS, resuspended in PBS/1% BSA containing 2 × 10⁵ cpm of ¹²⁵I-protein A (ICN, Irvine, CA), and incubated for 1 h at 4°C. Cells were washed three times in PBS, and cpm of ¹²⁵I-protein A bound to the IgG molecules on the cells were counted in a gamma counter.

KO mice were immunized with RRBC membranes, and anti-Gal presence in their serum was confirmed by ELISA. Subsequently, the mice were vaccinated s.c. in the abdomen with 2 × 10⁶ irradiated (4000 rad) BL6 or BL6_αGT tumor cells. Two weeks after vaccination all mice were challenged s.c. in the back with 0.5 × 10⁶ or with 0.2 × 10⁶ live parental BL6 cells, as indicated in the individual experiments. Subsequently, mice were examined for tumor growth every day for 60 days at the challenge site. A positive score for tumor growth was defined as a tumor of ≈5 mm in size. Assessment of the extent of protection against tumor growth was determined by the percentage of mice remaining tumor free.

Developing tumors (<5 mm) were removed from the vaccinated KO mice 2 weeks postchallenge with live BL6 cells. Tumors were immediately fixed in formalin, sectioned, and stained with H&E or with T and B lymphocyte-specific biotinylated monoclonal antibodies (PharMingen, San Diego, CA). Tumor-infiltrating macrophages were identified by staining for peroxidase using 3-amino-9-ethylcarbazole, a peroxidase chromogen (Biomeda, Foster City, CA).

Data from the tumor protection studies were evaluated by the Fisher’s exact test. Statistical significance was regarded as P ≤ 0.05.

The main characteristic of human anti-Gal that enables its exploitation as an augmenter of vaccine immunogenicity is its ability to induce uptake (i.e., phagocytosis) of cells or cell membranes expressing α-gal epitopes (13, 14). To demonstrate the suitability of the KO mouse system for studying the efficacy of tumor vaccines expressing α-gal epitopes, we first analyzed the ability of anti-Gal, produced in these mice, to induce phagocytosis of cells by APCs. For this purpose, anti-Gal IgG was isolated from sera of KO mice by affinity chromatography on columns expressing α-gal epitopes. The restricted specificity of anti-Gal for α-gal epitopes was confirmed by its binding to synthetic α-gal epitopes on BSA and not to N-acetyl-lactosamine linked to BSA (Fig. 1). Moreover, the isolated anti-Gal IgG molecules bound to the natural glycoprotein fetuin that has α-gal epitopes (i.e., α-gal-fetuin) but not to the same protein with terminal SA units (i.e., SA-fetuin; Fig. 1).

The ability of this antibody to induce phagocytosis of cells expressing α-gal epitopes was studied with RRBCs. These cells are small enough to be visualized within a macrophage, and they express α-gal epitopes (24, 43, 44). Incubation of RRBCs with 10 μg/ml purified mouse anti-Gal and with adherent macrophages from these mice resulted in extensive phagocytosis of the red cells by the macrophages (Fig. 2). Each macrophage contained 7–22 phagocytosed RRBCs. However, no phagocytosis of RRBCs was observed in the absence of anti-Gal (data not shown). These data indicate that anti-Gal IgG produced by KO mice displays characteristics similar to those of human anti-Gal, in that it interacts specifically with α-gal epitopes and that it can induce the uptake of cells expressing such epitopes by APCs.

We have previously shown that BL6 melanoma cells completely lack α-gal epitopes and that stable transfection of these cells with murine α1,3GT cDNA results in expression of ≈2 × 10⁶ of these epitopes per cell without altering other cell surface characteristics (41, 42). The use of these cells for studying tumor vaccine immunogenicity required the demonstration of specific binding of mouse anti-Gal to α-gal epitopes on BL6_αGT cells. This was first demonstrated by fluorescence-activated cell sorting analysis with purified anti-Gal. As shown in Fig. 3, the isolated mouse anti-Gal readily bound to BL6_αGT cells; however, only marginal nonspecific binding of the antibody was observed with the parental BL6 cells.

We further demonstrated the ability of nonpurified anti-Gal IgG within the serum of KO mice to interact with α-gal epitopes on BL6_αGT cells. This was achieved by the use of ¹²⁵I-protein A, which interacts with the Fc portion of cell bound IgG molecules. BL6_αGT and BL6 cells in suspension were incubated with KO or WT mouse serum diluted 1:50, washed, and assayed for ¹²⁵I-protein A binding to Fc portion of the IgG molecules complexed to the tumor cells. Incubation of BL6_αGT cells with KO mouse serum resulted in a 12-fold higher binding of ¹²⁵I- protein A to the cells than the binding to BL6 incubated under similar conditions (Fig. 4). In contrast, only minimal ¹²⁵I-protein A binding was detected when both cell lines were incubated with serum from WT mouse immunized with RRBCs. This binding, which is not significantly different from the binding of KO serum IgG molecules to BL6 cells, reflects a background level of antibody adhesion to the tumor cells. These data further imply that similar immune complexes between circulating anti-Gal IgG molecules and vaccinating BL6_αGT cells would be formed in situ in KO mice.

The studies above indicated that the KO mouse and the tumor cells BL6 and BL6_αGT comprise an experimental model that simulates the human parameters that are necessary for measuring the efficacy of α-gal epitope-expressing vaccines in cancer immunotherapy. The efficacy of such vaccines was assessed by measuring the ability of irradiated BL6_αGT cell vaccines to induce a protective immune response against challenge with live BL6 cells. KO mice with documented anti-Gal production, following immunization with RRBC membranes, were vaccinated with 2 × 10⁶ irradiated BL6_αGT or BL6 cells. All mice were challenged 2 weeks later with 0.5 × 10⁶ live BL6 cells and monitored for tumor development for 60 days. As many as a third of the mice receiving the BL6_αGT vaccination were protected from live BL6 challenge in two separate experiments (Fig. 5, A and B). In the first experiment, all nine mice vaccinated with BL6 cells developed tumors 10–21 days postchallenge (Fig. 5,A). In contrast, three of nine KO mice receiving BL6_αGT vaccines remained tumor free (Fig. 5,A). In the second experiment (Fig. 5,B), 6 of 18 KO mice vaccinated with BL6_αGT cells did not develop tumors for 60 days postchallenge. In contrast, all mice vaccinated with BL6 cells developed tumor within 10–26 days postchallenge (Fig. 5 B).

To further determine whether mice immunized with BL6_αGT cells developed tumors because of a large tumor burden in the challenge and insufficient vaccination, the vaccination and challenge experiment was repeated with mice that were immunized twice with irradiated BL6_αGT cells and challenged with only 0.2 × 10⁶ live BL6 cells. As shown in Fig. 6, 60% of these mice were tumor free for 60 days, whereas only 20% of the mice in the control group immunized with BL6 cells remained tumor free. Thus, repeating the immunization and decreasing tumor burden in the challenge, indeed, increased the extent of protection. However, a small proportion of the mice in the control group also developed immune resistance to the tumor.

To establish whether vaccination with BL6_αGT cells also elicited cellular immune response against BL6 cells in mice that developed tumors, KO mice were vaccinated with either irradiated BL6_αGT or BL6 cells as described above (five mice per group), challenged with live BL6 cells, and developing tumors were removed at day 14 postchallenge (tumor size of ≈2–4 mm). The tumors were fixed, sectioned, stained with H&E, and examined for mononuclear cell infiltrates. Tumors from mice vaccinated with BL6 cells displayed no lymphoid infiltrates (Fig. 7,A). The tumors from these mice were surrounded by a thin layer of connective tissue. The melanoma cells were actively proliferating, as suggested by the relatively small size of the tumor cells, some of which contained melanin granules. Furthermore, the basophilic staining of the cytoplasm of the melanoma cells implied extensive protein synthesis, characteristic of actively dividing cells (Fig. 7,A). In contrast, the tumors from mice vaccinated with BL6_αGT cells exhibited distinct mononuclear cell infiltrates that surrounded the developing tumors (Fig. 7 B). Immunostaining analysis (data not shown) indicated that as many as 70% of the cells in these infiltrates were T cells stained by anti-CD3 antibodies, and 30% were macrophages as indicated by peroxidase staining. No B cells were found in these infiltrates.

At the areas adjacent to the mononuclear cells, the tumor cells were found to be large, with vacuolated cytoplasm, which may suggest ongoing cytolysis (Fig. 7,B). The large melanin content of the tumor cells presented in Fig. 7 B, was not apparent in all regions near the mononuclear infiltrate. However, the tumors in mice vaccinated with BL6_αGT cells usually displayed higher level of melanin production than tumors in mice vaccinated with BL6 cells. This suggests that tumor cells adjacent to the infiltrating lymphocytes stopped their proliferation, possibly under influence of cytokines secreted by the lymphocytes, and became more differentiated as indicated by the excessive melanin production.

Overall, these histological studies demonstrated no cellular immune response against the developing tumor in mice vaccinated with BL6 cells. However, mice vaccinated with BL6_αGT cells developed a cellular immune response, implied by the mononuclear cell infiltrates, which prevented tumor growth in many of the mice. In the rest of these animals, the tumor load in the challenge was too large, and proliferating tumor cells could not be completely eliminated as a result of the immune response elicited by the BL6_αGT cell vaccine. It should be stressed that this cellular immune response was not strong enough to be detected in the spleens of the immunized mice. This is suggested by the finding that in vitro cytotoxicity assays with spleen cells from the immunized mice, as effector cells, revealed no specific killing of BL6 target cells (data not shown). Thus, the antitumor immune response could be demonstrated only in vivo.

The data in this study support our hypothesis on the augmentation of tumor vaccine immunogenicity by the in situ formation of immune complexes between the vaccinating tumor membranes and anti-Gal IgG molecules. The unique experimental tumor model of BL6 melanoma cells in KO mice used in this study addresses the question of whether, in an animal producing anti-Gal, a tumor vaccine that expresses α-gal epitopes can confer immune protection against the tumor lacking this epitope. We first showed that anti-Gal IgG produced in KO mice shares similar characteristics with human anti-Gal, in that it is highly specific for α-gal epitopes and can induce phagocytosis of cells expressing these epitopes by APCs. Subsequently, we showed that irradiated tumor vaccines expressing α-gal epitopes (i.e., BL6_αGT cells) protect many KO mice from challenge with the poorly immunogenic parental BL6 cells. Moreover, challenged mice that develop tumors subsequent to vaccination with BL6_αGT cells displayed extensive mononuclear cell infiltrates around the developing tumor. In contrast, mice challenged following immunization with BL6 cells developed tumors that lack any indication of lymphoid infiltrates. These observations imply that vaccination with BL6_αGT cells elicits a cellular immune response against TAAs on the melanoma cells. This immune response can prevent tumor development in a significant proportion of the mice.

The observed specific binding of the mouse anti-Gal IgG to α-gal epitopes on BL6_αGT cells and the ability of this antibody to induce phagocytosis by macrophages strongly imply that the immune protection conferred by the BL6_αGT vaccine is associated with the in situ uptake of the opsonized vaccinating membranes by APCs. This uptake is important both for processing and presentation of TAAs and for the transport of the TAAs to the adjacent draining lymph nodes. Recently, it has become apparent that activation of tumor-specific naive T cells can occur only in the lymph nodes or spleen and that these T cells can migrate to the periphery and seek and destroy metastases expressing TAAs only after they are activated (1, 3, 4).

Our study supports the hypothesis that successful vaccination against weak TAAs may be achieved by effective uptake of the vaccine by APCs. The use of adjuvant or cytokines such as GM-CSF seems to effectively recruit APCs to the vaccination site (2, 45, 46, 47). However, without specific “label” on the vaccinating tumor membranes, which would direct their uptake by APCs, the amount of weak TAAs transported to the draining lymph nodes may be insufficient for eliciting an effective immune response. These considerations may explain the success of GM-CSF-secreting B16 melanoma vaccines to protect against challenge (46) versus the failure of GM-CSF-secreting BL6 vaccines (36). The latter tumor cell is much less immunogenic than the parental B16 melanoma (36, 40), and thus far, there have been no reports on induction of a cellular immune response against BL6 cells. It is likely that in situ immune complexing between anti-Gal and α-gal epitopes on the BL6_αGT vaccinating membranes enables the subsequent extensive uptake and processing of these vaccines by APCs. Without this critical step, the amount of processed TAAs in the BL6 vaccine might be insufficient for the effective activation of tumor-specific T lymphocytes, which are capable of destroying the tumor cells in the challenge. In view of the potent ability of GM-CSF to recruit APCs to the vaccination site, it would be of interest to determine whether GM-CSF secretion by vaccinating cells can further increase the efficacy of tumor vaccines expressing α-gal epitopes.

Tumor cells engineered to express α-gal epitopes were obtained in this study by stable transfection of tumor cells with the α1,3GT cDNA in a plasmid containing neomycin resistance gene as a selectable marker (39). Such an approach is impractical with freshly obtained tumor cells because, in most cases, these cells do not proliferate in vitro. Freshly isolated cells may be engineered to express α-gal epitopes by transduction with viral vectors (e.g., replication-defective adenovirus or herpes virus), which introduce multiple copies of the α1,3GT cDNA into the transduced cells. Transduction by virus may further elicit a “xenogenization” process, in which viral antigens expressed on the vaccinating tumor cells may enhance T-cell response to TAAs, as suggested by Kobayashi and colleagues (48, 49). Moreover, it may be possible that α-gal epitopes on tumor vaccines partly mimic this xenogenization effect of viral antigens on the vaccinating tumor cell. It remains to be determined whether the viral antigens and α-gal epitopes on the vaccinating tumor cells have a synergistic effect on the immunogenicity of tumor vaccines. In addition to the method of viral transduction, vaccinating tumor cells or cell membranes can be engineered to express α-gal epitopes by incubation with recombinant α1,3GT and UDP-galactose, as we have demonstrated previously (13, 14). This method results in the expression of ≈1 × 10⁶ α-gal epitopes per cell.

It could be argued that expression of α-gal epitopes on vaccinating tumor cells is similar to previous methods of “haptenization” of tumor vaccines (e.g., linking dinitrophenol hapten to the membrane of the tumor cell; Ref. 50). Our method is superior to linking dinitrophenol to tumors because, in haptenization, the immune response to the hapten does not preexist in the patients and must be induced. The extent of this response and the subsequent opsonization of the vaccine are likely to greatly vary from one patient to the other. In contrast, anti-Gal is present in large amount in all patients, unless they are severely immunocompromised (20). Therefore, targeting of tumor vaccines to APCs by anti-Gal is likely to occur in almost any treated patient within a short period of time postvaccination, and it does not require the generation of an additional immune response to the hapten. In addition, haptenization may result in covalent linking of the hapten to TAA, thus destroying the antigenicity of TAA peptides. In contrast, α-gal epitope expression has no such effect on TAA because the modification is limited to carbohydrate chains and does not affect protein molecules.

The mechanism of tumor destruction by the infiltration of mononuclear cells is not yet clear. Previous observations demonstrated the role of cytotoxic T cells in the destruction of the parental B16 melanoma in immunized mice (2, 46, 47). It is possible that a similar mechanism facilitates BL6 tumor destruction. Immunostaining studies, indeed, demonstrated that the mononuclear infiltrates observed within the BL6 tumors primarily comprise T cells and macrophages. The exact TAAs recognized by these T cells also require characterization. It is probable that these TAAs are peptides and not carbohydrate epitopes, such as the core structure of the α-gal epitope (i.e., Galβ1,4GlcNAc-R). This is because this core structure is also present on normal cells, and thus, the mouse is immunotolerant to it.

Overall, our studies suggest that expression of α-gal epitopes on tumor vaccines increases the immunogenicity of weak TAAs. It is probable that results of the similar use of autologous tumor vaccines, engineered to express α-gal epitopes in humans, would greatly vary from one patient to the other and, in many patients, would be insufficient for inducing an effective antitumor immune response. Nevertheless, this type of vaccine, used as adjuvant immunotherapy, will provide the immune system with an additional opportunity to be effectively exposed to autologous TAA peptides that are processed and presented by APCs. In some of the immunized patients, this exposure may be sufficient for the induction of an anti-TAA immune response that is effective enough to destroy metastases.

We thank Dr. E. Gorelik for helpful discussions and Drs. J. B. Lowe and A. Thall for providing the KO mice used in this study.