Immunity to the α(1,3)Galactosyl Epitope Provides Protection in Mice Challenged with Colon Cancer Cells Expressing α(1,3)Galactosyl-transferase

INTRODUCTION

Cancer gene therapy offers a potential replacement or augmentation of traditional cancer treatments, which use invasive or toxic protocols. Suicide genes that encode an enzyme that activates a prodrug into a toxic molecule, or genes that induce apoptosis, have been or are currently being tested in clinical trials for their efficacy in cancer therapy (1, 2, 3) . Gene therapy vaccine technology is under development for several malignancies. Melanoma has been the favored target because it is a very immunogenic tumor. However, the more common and important forms of cancer have minimal or no immunogenicity. We are developing a novel approach to cancer gene therapy that uses a patient’s innate immunity against xenoantigens to identify and destroy tumor cells. The key aspect of this work is to determine whether colon cancer cells that express a xenoantigen can be rejected by an immune response to the antigen. Colon cancer is the third most common form of cancer and third leading cause of death (4) and provides a good target for a successful cancer gene therapy.

Humans possess specific humoral immunity to αGal, 3 a major xenotransplant antigen. Although human cells do not carry a functional enzyme for the expression of αGal epitopes because of a 2-base frameshift gene mutation (5) , there is evidence that suggests that high titer natural antibody to αGal is produced in humans because of continuous antigenic stimulation by gastrointestinal bacteria (5, 6, 7) . Clonal B-cell analyses estimated that ∼1% of circulating B cells produce anti-αGal antibody (8) . The αGT catalyzes the transfer of galactose from UDP galactose to the N-acetyl-lactosamine acceptors on carbohydrate side chains of glycoproteins and glycolipids to create the αGal moiety. The anti-αGal immune response is responsible for initiating hyperacute rejection of vascularized xenotransplants, a severe immunological reaction observed in primates. When αGal and specific antibody form immune complexes, complement is activated via the classical pathway (9, 10, 11, 12, 13) .

Our interest in αGal-mediated destruction of tumor cells was inspired by studies describing lysis of murine retroviral VPCs after exposure to human peritoneal fluid. VPCs have been used for in vivo gene delivery in several cancer gene therapy studies (14 , 15) . Our laboratory and others have demonstrated that antibody and complement in human serum binds αGal within 30 min of exposure and induces complement-mediated lysis of VPCs and the viral vectors they produce (16, 17, 18, 19, 20) . Additionally, Collins et al. (21) showed that human fibroblast cells expressing porcine αGT were destroyed by antibody and complement. To test whether this gene could be used to induce destruction of tumor cells, a truncated version of the murine αGT was cloned into a retroviral vector backbone and used to transduce human A375 melanoma cells (22) . During in vitro experiments, >90% of transduced A375 cells expressing αGal were killed after exposure to human serum. αGal-expressing A375 cells were treated for 30 min with human serum and then injected in vivo into athymic nude mice. All experimental mice remained tumor free, whereas control groups developed tumors (22) . Lysis of αGal-expressing murine cells by human serum can be blocked by the addition of complement inhibitors (heparin, enoxaparin) or soluble complement receptor 1 (20) . These data demonstrate the key role that complement has in destruction of targets expressing the αGal xenoantigen.

Transgenic knockout mice that lack the αGT gene (αGT KO) have been produced (23 , 24) and provide an ideal small animal model to study the in vivo immune response against αGal epitopes. While not expressing detectable cell surface αGal epitopes, these mice can produce low detectable titers of natural anti-αGal, possibly from bacterial stimulation (25 , 26) . Immunization with RRBCs results in the production of anti-αGal antibody with titers and specificity similar to those observed in humans (27) . In this report we present in vivo data that shows clear protective benefits of an anti-αGal immune response, when RRBC-immunized αGT KO mice are challenged with αGal⁺ tumor cells. These findings have implications for generation of a system to deliver the αGal suicide gene to human tumor cells, and making them susceptible to destruction by natural human immunity to αGal.

MATERIALS AND METHODS

Cells and Media.

MC38 colon carcinoma cells are syngeneic for C57BL/6 mice and express the αGal antigen on their cellular surface. B16.BL6-2 melanoma cells, a metastatic nonimmunogenic derivative of the B16.F10 cell line, are also syngeneic for C57BL/6 mice and are αGal-negative. All cells were maintained at 37°C in a 5% CO₂ incubator. The growth medium consisted of DMEM (Invitrogen-Life Technologies, Inc., Carlsbad, CA) supplemented with 10% FBS (d-10; Invitrogen-Life Technologies, Inc.).

Animal Model.

Knockout mice for αGT (αGT KO) were received for establishing a breeding colony from Dr. John B. Lowe of the University of Michigan (23) . C57BL/6 mice were obtained from Harlan Sprague Dawley, Inc. (Indianapolis, IN), and were used as control mice during tumor implantation studies. All animals were cared for under Institutional Animal Care and Use Committee-approved protocol and housed in a contained facility.

Lectin Staining for αGal Epitopes.

MC38 and B16.BL6-2 cells were seeded into 30-mm dishes in d-10, incubated at 37°C in 5% CO₂, and grown to confluent monolayers. Cell monolayers were washed twice with HBSS and incubated for 15 min at room temperature with a 1:50 dilution in Opti-MEM (Invitrogen-Life Technologies, Inc.) of FITC-labeled Griffonia simplicifolia IB4 (Vector Laboratories, Inc., Burlingame, CA). This lectin has previously been shown to bind specifically to αGal epitopes (28 , 29) . The lectin solution was removed, monolayers were washed twice with HBSS, and fresh Opti-MEM was added to cells. Monolayers were observed for lectin binding using a Nikon Diaphot 300 fluorescent microscope (Nikon, Inc., Melville, NY) and photographed using Fuji 1600 Provia color film (Fuji Photo Film Co., Tokyo, Japan).

Complement-mediated Cell Death.

MC38 cells in culture were trypsinized for 2–3 min at 37°C. The trypsin was inactivated with complete culture media (d-10), and cells were collected by centrifugation at 3000 rpm for 5 min at 4°C. Cells were suspended in 200 μl of one of three possible treatment solutions: 50% DMEM and 15% FBS (d-15) in Opti-MEM; 50% fresh human serum in Opti-MEM; or 50% heat-inactivated human serum in Opti-MEM. In addition, cells were incubated at 37°C for 1 h. Treated cells were collected, resuspended in 100 μl of FITC-labeled IB4 lectin (diluted 1:100 in Opti-MEM), and incubated at room temperature for 10 min. One hundred μl of a PI solution (25 μg/ml) diluted in HBSS were added to the cells and incubated for 5 min at room temperature. Cells were collected, resuspended in Opti-MEM, and analyzed by flow cytometry (Coulter Epics Altra Flow Cytometer, Miami, FL).

αGal Antigen Immunization.

Eight to 12-week-old αGT KO mice were used in this study and cared for under an approved animal protocol using American Association of Laboratory Animal Care guidelines. Mice were immunized i.p., with 10⁷ αGal⁺ female NZW RRBCs (Cocalico Biologicals, Inc., Reamstown, PA) suspended in 100 μl of HBSS. Immunizations were given either a single time 14 days before MC38 and B16.BL6-2 tumor cell challenge or twice at 28 days and 14 days before MC38 tumor cell challenge. Control αGT KO and control C57BL/6 mice were mock immunized with 100 μl of HBSS i.p.

Antibody Titration and Subtyping.

Antibody specific for the αGal epitope was detected, and end point was titrated using an indirect ELISA. αGal antigen (αGal-BSA; V-Labs, Inc., Covington, LA) was diluted to 5 μg/ml in carbonate buffer (pH 9.5) and coated onto polyvinyl chloride (PVC) ELISA plates (Falcon 3912; Becton Dickinson Labware, Franklin Lakes, NJ) overnight at 37°C in a humidified chamber. Nonspecific binding sites in assay wells were blocked for 2 h with a solution of 1% BSA (Fraction V; Sigma) in carbonate buffer. Two-fold serial dilutions of primary sera were made in wash buffer [1× PBS (pH 7.4), 0.05% Tween 20), added to antigen-coated wells, and incubated for 1 h at 37°C. Wells were washed five times, and a secondary antibody, horseradish peroxidase-labeled goat antimouse IgG H+L diluted 1:5000 in wash buffer (Pierce Chemical Co., Rockford, IL), was added to assay wells and incubated 1 h at room temperature. Wells were washed five times, and 100 μl of 3,3′,5′,5-tetramethylbenzidine liquid substrate (Sigma) was added. After a 15-min incubation at room temperature, the substrate reaction was stopped with 0.5 n H₂SO₄, and the absorbance at 450 nm for each well was determined using a Molecular Dynamics SpectraMax 250 Plate reader (Sunnyvale, CA).

The murine immune response to αGal was measured over time. Serum from immunized mice was diluted 1:100 in wash buffer before detection by ELISA as before. Anti-αGal antibody class and subtype was determined from serum collected 14 days after the last RRBC immunization, using a Zymed MonoAb ID ELISA kit (Zymed Laboratories, Inc., San Francisco, CA). αGal-BSA antigen was coated onto PVC ELISA plate wells as before, and nonspecific binding sites were blocked using 1% BSA. Diluted primary sera were added and incubated with antigen. Secondary rabbit antimouse isotype horseradish peroxidase-labeled antibodies and 3,3′,5′,5-tetramethylbenzidine substrate were used for detection.

Tumor Cell Challenge.

In the first experiment, 15 αGT KO mice were immunized i.p. a single time with 10⁷ RRBC in 100 μl of HBSS 14 days before tumor cell challenge. As experimental controls, 15 αGT KO mice and 15 syngeneic C57BL/6 mice were mock immunized with HBSS. In a blinded experiment, all mice were divided into three sets, with each set comprised of 5 immunized αGT KO mice, and 5 mice each of the two control groups. Each set of mice was challenged i.p. with a different dilution of MC38 colon carcinoma cells suspended in Plasma-Lite (Baxter Healthcare Corp., Deerfield, IL). Mice in set A were challenged with 2.5 × 10⁴ MC38 cells. Mice in sets B and C were challenged with 5.0 × 10⁴ and 1.0 × 10⁵ MC38 cells, respectively. Separately, 8 αGT KO mice were immunized a single time as before with 10⁷ RRBC in HBSS 14 days before i.p. challenge with 1.0 × 10⁵ B16.BL6-2 melanoma cells suspended in Plasma-Lite. A second experiment was designed in which 15 αGT KO mice were immunized twice with 10⁷ RRBC 28 and 14 days before MC38 tumor cell challenge. Fifteen αGT KO mice and 15 syngeneic C57BL/6 mice were mock immunized with HBSS and served as experimental controls. All RRBC-immunized and mock-immunized control mice were challenged i.p. with 2.5 × 10⁴ MC38 colon carcinoma cells. In all experiments, mice were observed daily for animal morbidity and palpated for tumor growth.

RESULTS

FITC-Lectin Staining.

MC38 colon carcinoma and B16.BL6-2 melanoma cells were incubated with a solution of FITC-labeled IB4 lectin to identify the differences between the two cell lines in expression of the αGal epitope. FITC staining is prominent along the outer surface membrane of cultured MC38 cells (Fig. 1) ⇓ because this cell line has a functional αGT gene and expresses the surface αGal moiety that is detected by IB4 lectin binding. B16.BL6-2 melanoma cells do not express the αGal moiety because they lack a functional αGT gene and are not stained with IB4 lectin-FITC.

Complement-mediated Cell Death.

αGal-positive MC38 cells were incubated with culture medium containing no human serum, 50% normal human serum, or 50% heat-inactivated human serum. Flow cytometry of FITC-lectin-stained cells provided total cell counts. In addition, PI uptake by dead cells as a percentage of total cell numbers was used to measure cell death after incubation with the test media. Flow cytometry of cells treated with the three test media demonstrated killing of 98% of αGal-positive MC38 cells by media containing human serum with active complement (Fig. 2) ⇓ . Cells that were incubated with media containing heat-inactivated human serum exhibited a 30% cell death by PI uptake, whereas background uptake by untreated cells was 22%.

Antibody Titers and Subtyping.

Antibody titers to the αGal epitope from immunized αGT KO mice and mock-immunized control αGT KO and C57BL/6 mice were determined by indirect ELISA. Fig. 3 ⇓ shows the results of antibody titration in serum from mice that were immunized either a single time 14 days before serum collection (Fig. 3A) ⇓ , or mice that were immunized twice, 28 and 14 days, before serum collection (Fig. 3B) ⇓ . Serum from mock-immunized αGT KO and C57BL/6 mice exhibited an average A⁴⁵⁰ absorbance of <0.05 at a 1:50 dilution and was considered negative. This serum was pooled and used to determine background antibody binding. Immune sera with an A⁴⁵⁰ absorbance of >0.1 above background were considered positive. The average titer of anti-αGal antibody in mice immunized one time with RRBC was 1:1600 (Fig. 3A) ⇓ . Mice that were immunized twice with RRBC had an average titer of 1:8000 (Fig. 3B) ⇓ . The immune response in αGT KO mice to αGal was determined by collecting serum from mice at various times after RRBC immunization on days 14 and 28. An anamnestic IgG immune response to αGal that peaks at 7 days after the second immunization was observed (Fig. 4A) ⇓ .

Serum samples collected on day 0 from αGT KO mice immunized once on day −14 (14 mice) or twice on days −28 and −14 (8 mice) were assayed for their antibody isotype. An A⁴⁵⁰ absorbency > 0.2 for an individual antibody isotype was considered positive. Serum was negative for an antibody isotype and subclass if the A⁴⁵⁰ absorbance was <0.2. RRBC immunization stimulated the production of IgM, IgG1, and IgG3 antibodies (Fig. 4B) ⇓ . The predominant subclass was IgG3, which developed in 50% (11 of 22 immunized mice). The IgG1 subclass was observed with higher frequency (36%) then either IgG2a or IgG2b (5 and 13%). All mice immunized regardless of schedule demonstrated high levels of IgM antibody, and no immunized mice demonstrated any IgA antibody isotype. Interestingly, all immunized mice developed antibody with κ light chain, and no λ light chain was found in these mice.

Tumor Cell Challenge.

Fourteen days after RRBC immunization of αGT KO mice and mock immunization of αGT KO and C57BL/6 controls, the mice were divided into three groups and implanted i.p. with different dilutions of MC38 colon carcinoma cells in a blinded experiment. Tumor cell dilutions were made, coded, and randomized before injection into mice. All mice were observed daily for tumor development and were euthanized when tumors reached ∼1200 mm³, exhibited ascites fluid production, or when the mice were moribund. Survival curves of RRBC-immunized and mock-immunized control mice challenged with the three doses of MC38 tumor cells (2.5 × 10⁴, 5.0 × 10⁴, or 1.0 × 10⁵) are presented (Fig. 5) ⇓ . Immunized αGT KO mice were observed to develop tumors more slowly regardless of the amount of MC38 cells used for challenge. Immunized and mock-immunized mice challenged with the three dilutions of MC38 cells were pooled and analyzed as a group. The percentage of mock-immunized control αGT KO mice (47%) that survived challenge with the three dilutions of MC38 cells was lower than the percentage of RRBC-immunized αGT KO mice (87%) that were challenged (P = 0.031). All three dilutions of MC38 cells were able to rapidly establish tumors in mock-immunized C57BL/6 control mice, and all C57BL/6 control mice were euthanized by 18 days after tumor cell. In a separate experiment, the survival curve for RRBC-immunized αGT KO mice challenged with 1 × 10⁵ B16.BL6-2 cells demonstrates that immunity to αGal does not protect against challenge with αGal⁻ melanoma cancer cells (Fig. 6) ⇓ . In this experiment, all 8 immunized αGT KO mice developed tumors rapidly and were euthanized by day 25.

The challenge experiment was repeated using a single dilution (2.5 × 10⁴) of MC38 tumor cells to challenge αGT KO mice immunized twice with RRBC to increase their antibody titers and mock-immunized control mice. After tumor cell challenge, all mice were closely monitored and palpated for tumor growth (Fig. 7) ⇓ . None of the mock-immunized C57BL/6 control mice (0%) survived beyond 17 days after tumor cell challenge. Only 9 of 15 (60%) mock-immunized αGT KO mice survived beyond 32 days after challenge. In contrast, all of the immunized αGT KO mice (100%) survived MC38 challenge (P = 0.0069), and palpated tumors were smaller and developed much more slowly when compared with mock-immunized αGT KO mice.

DISCUSSION

The role of αGal epitopes in xenograft rejection has been well studied (9, 10, 11, 12, 13 , 30) . Rapid immune recognition of αGal epitopes on the surface of foreign cells results in antibody-mediated cell cytotoxicity and complement-mediated lysis of targeted cells. This immunity is a barrier to the use of nonhuman organs and tissue for transplant purposes. Recent advances in transplant science have resulted in the development of αGT KO pigs that may overcome immune obstacles to successful xenotransplantation in humans (31) . Cancer gene therapy protocols that rely upon murine VPC for delivery of genes can also be limited by the potent human immune response to αGal epitopes. Both VPC and the vectors they produce are destroyed by anti-αGal immune mechanisms (16, 17, 18, 19, 20) . Clonal B-cell analysis estimated that ∼1% of circulating B cells produce anti-αGal antibody (8) , and an estimated 1–2.4% of circulating IgG and 3.9–8% of IgM are specific anti-αGal antibodies (32 , 33) . We hypothesize that innate anti-αGal immunity that is disadvantageous for xenograft transplantation and VPC therapies could be used as an advantageous method to induce the destruction of cancer cells. Human cancer cells expressing αGal epitopes on their cell surface would appear as xenoantigens and induce a strong immune response that destroys them. A murine colon cancer cell line was chosen for this study. Colon cancer is the third most common form of cancer and third leading cause of death (4) .

The αGT KO mouse is an ideal small animal model to study our hypothesis. The loss of the αGT gene by these mice mimics the evolutionary loss of this gene by ancestral Old World primates and humans (5) . αGT KO mice produce little or no αGal-specific antibody (25 , 26) but are able to develop an immune response to αGal when immunized with RRBCs. These mice can produce αGal-specific antibody with high titers and specificity in some animals similar to those observed in humans (27) . The RRBC immunization protocol we used with the αGT KO mice resulted in high titers of anti-αGal antisera that could be detected by indirect ELISA at a 1:16,000 dilution. Previously, LaTemple et al. (34) demonstrated a partially protective immune response when αGT KO mice are vaccinated with αGal-expressing B16 cells (after stable transfection of αGal-negative B16 cells with αGT cDNA) and challenged with parental B16 cells.

MC38 murine colon carcinoma cells have a functional αGT enzyme and express αGal on their cell surface glycoproteins. Fig. 1 ⇓ shows the difference in αGal expression between MC38 colon carcinoma and B16.BL6-2 melanoma cells used in this study. The αGal moiety expressed on the surface of MC38 cells is labeled brightly with IB4 lectin-FITC conjugates, whereas B16.BL6-2 melanoma cells lack αGal expression on their surface and do not bind the IB4 lectin. A serum exposure assay showed the complement-mediated destruction of αGal-positive MC38 colon carcinoma cells. Cells were incubated with media that contained 50% human serum with active complement or media that contained 50% heat-inactivated human serum (Fig. 2) ⇓ . After incubation with the test media, PI uptake measured by flow cytometry was used to estimate the percentage of cells killed. A total of 98% of MC38 cells was killed when exposed to untreated human serum. Control cells that were untreated or treated with heat-inactivated serum (devoid of active complement) showed PI uptake of 22 and 30%, respectively). Therefore, the presence of active human complement induced dramatically higher killing as expected (22) . Takeuchi et al. (18) demonstrated similar results using human cells. Transfected cells that express porcine αGT are lysed by human serum with complement (18) .

The titer of anti-αGal antibody in immunized αGT KO mice was measured by indirect ELISA. Mice immunized twice with RRBC developed an average titer of 1:8000 (Fig. 3B) ⇓ that is comparable with measured serum titers from patients receiving VPC treatment (data not shown). Mice that were immunized twice with RRBC developed an anamnestic immune response to the αGal antigen (Fig. 4A) ⇓ . The titer of anti-αGal peaked 7 days after the second immunization, and IgG1, IgG3, and IgM are the dominant anti-αGal heavy chain isotypes (Fig. 4B) ⇓ .

Tumor challenge studies were designed to determine whether immunity to αGal epitopes could provide protection from challenge with αGal-expressing MC38 murine colon carcinoma cells. In the first experiment, αGT KO mice were immunized a single time with 10⁷ RRBC and challenged with different dilutions (2.5 × 10⁴, 5.0 × 10⁴, or 1.0 × 10⁵) of MC38 cells. A total of 13 of 15 αGT KO mice survived tumor challenge, whereas only 7 of 15 mock-immunized αGT KO mice survived challenge with the same dilutions of MC38 cells (P = 0.031). Despite our evidence that mice immunized a single time with RRBC do not develop high titers of anti-αgal antibody, 87% of immunized mice were protected and survived. None of the mock-immunized control C57BL/6 mice survived the tumor challenge, and all were euthanized by day 18. Although anti-αGal antibody could not be detected by ELISA in sera from mock-immunized control mice, others have suggested that αGT KO mice have a low natural titer of anti-αgal antibody (25 , 26) . This natural antibody in the transgenic knockout mouse may, as hypothesized in humans, arise from stimulation of environmental antigens. The combination of preexisting low antibody titers and stimulation by αGal-positive MC38 cells may have allowed 47% of the mock-immunized αGT KO mice to survive challenge with the MC38 cell dilutions. In a separate experiment, αGT KO mice were immunized a single time with 10⁷ RRBC and challenged with 1.0 × 10⁵ αGal-negative B16.BL6-2 murine melanoma cells. None of these mice survived beyond day 25 of the tumor challenge (Fig. 6) ⇓ , and results of this experiment provide evidence that anti-αGal antibodies do not provide protection against tumors that do not express the αGT gene. The potential cellular differences (besides the expression of αGal epitopes) between MC38 and B16.BL-2 cells have been addressed by additional experiments in which immunized mice were challenged with B16.BL-2 cells transduced with a vector carrying the αGT gene. Results show a significant difference in tumor development and kinetics of tumor growth between immunized mice challenged with αGal-expressing B16.BL-2 cells and mice challenged with αGal-negative B16.BL-2 cells. 4 These findings are similar to those reported by LaTemple (34) . A second MC38 challenge experiment incorporated two RRBC immunizations of αGT KO mice to generate higher titers of anti-Gal antibodies. Immunized and mock-immunized control mice were challenged with 2.5 × 10⁴ of MC38 tumor cells. All 15 immunized αGT KO mice survived MC38 challenge, whereas only 9 of 15 mock-immunized αGT KO mice survived (P = 0.0069 by ANOVA). Again, all mock-immunized C57BL/6 control mice developed tumors rapidly and were euthanized by day 17. After the challenge, mice were observed and palpated daily. Observations included slower development of tumors and smaller tumors in immunized αGT KO mice compared with mock-immunized mice (data not shown). The survival of 60% of the control mock-immunized αGT KO mice may again be attributable to a combination of low-titer natural antibody and immune stimulation by the MC38 cells. These data provide a first step toward the development of αGal-based colon cancer vaccines for humans.

Colorectal cancer vaccines are as yet in the experimental stage of development. Potential vaccines based upon 17-1A, 791Tgp, carcinoembryonic antigen (35) , and the SART3 peptide antigens (36) are currently being tested with marginal but encouraging results. Colon cancer cells expressing a functional αGT gene will be readily identified by the high percentage of innate circulating human antibodies. This, in turn, would lead to hyperacute rejection of such genetically modified cells. Hyperacute rejection of porcine xenotransplants occurs after recognition and binding of multiple glycoprotein epitopes by human serum (13 , 37) . At least five major cell surface glycoprotein groups on porcine cells express αGal epitopes and are detected by human IgM and IgG antibodies and also bind IB4 lectin (38) . Similarly, the MC38 colon carcinoma cells used in this study can also be assumed to present multiple epitopes to the αGT KO murine immune system.

We have developed an in vivo animal model to demonstrate that immunity to αGal can protect αGT KO mice against challenge with colon cancer cells that express αGal. These results demonstrate the potential for a cancer gene therapy that uses the innate immunity to αGal antibody in humans. Direct gene transfer of αGT to in vivo tumors will be the key next step in determining whether αGT gene therapy for colon cancer will be successful. Gene delivery to tumor cells and expression of the αGT gene will present multiple targets for the immune system. Because both opsonization and complement fixation are dependent upon epitope density, the potential for a protective immune response directed against human tumor cells that express αGal is great.