Expression of α-gal epitopes on ovarian carcinoma membranes to be used as a novel autologous tumor vaccine

Available online at www.sciencedirect.com R

Gynecologic Oncology 90 (2003) 100 –108

www.elsevier.com/locate/ygyno

Expression of ␣-gal epitopes on ovarian carcinoma membranes to be used as a novel autologous tumor vaccine Uri Galili,a,b Zhao-chun Chen,a and Koen DeGeestc,* a

Department of Cardiovascular Thoracic Surgery, Rush University, Chicago, IL 60612, USA b Department of Immunology, Rush University, Chicago, IL 60612, USA c Department of Obstretrics and Gynecology, Division of Gynecology/Oncology, Rush University, Chicago, IL 60612, USA Received 20 September 2002

Abstract Objective. Poor presentation of tumor-associated antigens (TAA) to the immune system remains a major obstacle to effective anti-tumor vaccine therapy. The aim of this study is to demonstrate the feasibility of producing a novel autologous tumor vaccine from ovarian carcinoma that is expected to have increased immunogenicity. The strategy is based on the ability of the anti-Gal IgG antibody (a natural antibody comprising 1% of IgG in humans) to target tumor membranes expressing ␣-gal epitopes (Gal␣1-3Gal␤1-4GlcNAc-R) to antigen-presenting cells (APC). Study design. Freshly obtained ovarian carcinoma tumors are homogenized, washed, and incubated with a mixture of neuraminidase, recombinant ␣1,3galactosyltransferase (r␣1,3GT) and uridine diphosphate galactose (UDP-Gal) to synthesize ␣-gal epitopes on carbohydrate chains of glycoproteins of these membranes. Subsequently, the processed membranes are analyzed for expression of ␣-gal epitopes and for the binding of anti-Gal. Results. Incubation of 3 g of ovarian carcinoma membranes, from five different patients, at 100 mg/ml, mixed together with r␣1,3GT (50 ␮g/ml), neuraminidase (1 mU/ml), and UDP-Gal (2 mM), resulted in the effective synthesis of ␣-gal epitopes to the extent of ⬃2 ⫻ 1011 epitopes/mg of tumor membranes. As a result of this de novo expression of ␣-gal epitopes, the tumor membranes readily bound purified anti-Gal antibody, as well as anti-Gal in autologous serum. Conclusions. The method described in this study is very effective in the synthesis of many ␣-gal epitopes on tumor membranes obtained from ovarian carcinoma. These novel epitopes readily bind the naturally occurring anti-Gal antibody. This technique of opsonization of ␣-gal-modified autologous tumor membranes carrying TAA is expected to increase effective uptake of the vaccine by APC, which is key to successful anti-tumor vaccination. © 2003 Elsevier Science (USA). All rights reserved.

Introduction Tumor-associated antigens (TAA)1 are expressed on ovarian carcinoma cells, as on other tumors [1–3]. Most * Corresponding author. Department of Obstetrics and Gynecology, Rush University, 1653 West Congress Parkway, Chicago, IL 60612, USA. Fax: ⫹1-312-942-4043. E-mail address: [email protected] (K. DeGeest). 1 Abbreviations: ␣-gal epitope, Gal␣1-3Gal␤1-4GlcNAc-R epitope; ␣1,3GT, ␣1,3galactosyltransferase; r␣1,3GT, recombinant ␣1,3GT; APC, antigen presenting cells; BSA, bovine serum albumin; ELISA, enzymelinked immunosorbent assay; Fc␥R, Fc␥ receptor; Gal, galactose; HRP, horse radish peroxidase; PBS, phosphate-buffered saline; SA, sialic acid; TAA, tumor-associated antigens.

TAA have not yet been characterized, and it is believed that their expression varies from one type of tumor to the other, and in different patients that have the same type of tumor [1–5]. Therefore, autologous tumor vaccines have been considered for use in immunotherapy as a source of TAA. However, TAA on autologous tumor cells or cell membranes cannot induce an effective anti-tumor immune response by themselves, since tumor cells lack costimulatory molecules [6,7]. A prerequisite for the induction of an anti-tumor immune response by an autologous tumor vaccine is the effective uptake of the vaccinating tumor cells, or tumor cell membranes, by professional antigen-presenting cells (APC) such as dendritic cells, skin Langerhans cells, and macrophages. Subsequent processing of the internalized

0090-8258/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0090-8258(03)00148-3

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

TAA in the vaccine, and presentation of antigenic TAA peptides on APC in association with class I and class II MHC molecules, enables the activation of anti-tumor cytotoxic T cells and helper T cells [6 –11]. In addition, activation of tumor-specific T cells requires the delivery of a costimulatory nonspecific signal that can be provided only by professional APC [8 –11]. Another reason for the absolute need for effective uptake of tumor vaccines by APC is that activation of TAA-specific T cells takes place within the draining lymph nodes of the vaccination site. Therefore, the tumor vaccine must be transported by APC to adjacent lymph nodes, or to the spleen [11–13]. Such transportation of vaccines can occur after effective uptake of the vaccine by APC at the vaccination site [11–13]. Only after they are activated, tumor-specific T cells can leave the lymph nodes to seek and destroy tumor cells expressing these TAA. However, most tumor cells express no marker that identifies them as particles to be taken up by APC. Therefore, the immune system in most patients is “indifferent” to the developing tumor. The immunogenicity of autologous tumor vaccines may be increased by making such vaccines “palatable” to APC [14], i.e., by targeting the vaccine to APC. Such targeting can be achieved if IgG molecules can bind to the vaccinating tumor membranes (i.e., opsonize the membranes). These bound IgG molecules interact via their Fc portion with Fc␥ receptors (Fc␥R) on APC [15–17], thereby inducing the uptake (i.e., phagocytosis) of the vaccinating membranes by the APC. The significance of this mechanism was demonstrated in Fc␥R knockout mice that lack Fc␥R on their APC. These mice fail to develop immune protection against a tumor challenge, following immunization with melanoma vaccine, that is normally effective in inducing a protective response in wild-type mice [18]. We hypothesized that this principle of IgG-mediated targeting of vaccines to APC can be used for similar enhancement of autologous TAA immunogenicity by exploiting the natural anti-Gal antibody in humans [19,20]. Anti-Gal is the most abundant naturally occurring antibody in humans, constituting ⬃1% of serum IgG (20 –100 ␮g/ml serum) [21]. Anti-Gal interacts specifically with the ␣-gal epitope (Gal␣1-3Gal␤1-4GlcNAc-R) on glycolipids and glycoproteins [22–24] and is produced throughout life as a result of antigenic stimulation by bacteria of the gastrointestinal flora [25]. Whereas the ␣-gal epitope is absent in humans, it is abundantly expressed on cells of nonprimate mammals (e.g., mice rabbits, cows, or pigs), and in New World monkeys [26,27], where it is synthesized by the glycosylation enzyme ␣1,3galactosyltransferase (␣1,3GT) within the Golgi apparatus of the cells. Humans, apes, and Old World monkeys lack active ␣1,3GT gene, but all produce the anti-Gal antibody in large amounts [26,27]. AntiGal binds avidly in vivo to cells expressing ␣-gal epitopes. This is indicated in xenotransplantation where the in vivo binding of anti-Gal to ␣-gal epitopes on pig xenografts, such

101

as heart or kidney, results in rejection of xenografts in humans and in Old World monkeys [28,29]. We hypothesize that vaccination of ovarian carcinoma patients with autologous tumor cell membranes, modified to express ␣-gal epitopes, may result in the in situ binding of the patient’s anti-Gal IgG molecules to ␣-gal epitopes on the vaccinating tumor membranes. This, in turn, is likely to target the vaccine to APC by interaction of anti-Gal on the vaccinating tumor cell membrane with Fc␥R on APC. This interaction induces the uptake of the vaccinating membranes by APC, which subsequently transport these tumor membranes to draining lymph nodes. These APC further process and present TAA peptides for the effective activation of TAA-specific helper and cytotoxic T cells, within the lymph nodes. Once the TAA-specific T cells are activated, they can leave the lymph nodes, circulate in the body, and seek metastatic malignant cells expressing the TAA, in order to destroy them. In previous studies, we demonstrated the highly effective anti-Gal-mediated phagocytosis of human leukemia and lymphoma cells, processed to express ␣-gal epitopes, by human macrophages and dendritic cells [19,20,30]. We further demonstrated the efficacy of this type of vaccine in ␣1,3GT knockout mice, which lack the ␣-gal epitope and are capable of producing anti-Gal [31]. Immunization of these mice with irradiated melanoma cells, manipulated to express the ␣-gal epitope, elicited an immune response that protected a significant proportion of the mice against challenge with the same melanoma cells, lacking the ␣-gal epitope. In contrast, immunization with melanoma cells that lack ␣-gal epitopes conferred no protection. In humans with solid tumors, such as ovarian carcinoma, a big obstacle to the preparation of a tumor vaccine is the isolation of live tumor cells in sufficient numbers. Tumor membranes prepared from homogenates of freshly obtained tumors may provide an alternative source for autologous TAA. The present study demonstrates a novel method for expressing ␣-gal epitopes on freshly obtained ovarian carcinoma tissue homogenates, so that upon the use of such membranes as vaccine, they may be targeted in situ to APC, by the patient’s natural anti-Gal antibody.

Materials and methods Patients Tumor specimens were obtained from five patients at the time of surgical exploration for primary epithelial carcinoma of the ovary. The study was approved by the IRB and all patients provided informed consent for studying their tumor tissue and serum. Patient 1 had stage IIIc endometrioid adenocarcinoma; patient 2 had stage IV serous carcinoma; patient 3 had IIIc serous carcinoma; patient 4 had stage IIIc serous carcinoma; patient 4 had stage IV serous carcinoma; and patient 5 had stage IV serous carcinoma.

102

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

found it to be 2-fold higher than the saturating concentration of r␣1,3GT for synthesis of ␣-gal epitopes on cell membranes (not shown). At the end of incubation with the enzymes the membranes were washed twice with saline containing 1 mM EDTA and twice with saline. Subsequently, the processed tumor membranes were frozen at 200 mg/ml in saline. Fig. 1. Synthesis of ␣-gal epitopes on carbohydrate chains of glycoproteins in two-step reaction: (1) removal of sialic acid (SA in the left structure) to expose N-acetyllactosamine residues (Gal␤1-4GlcNAc-R) (center structure), and (2) linking galactose, contributed by uridine diphosphate galactose (UDP-Gal), to the exposed N-acetyllactosamine residues to generate ␣-gal epitopes (right structure).

Each tumor was brought under sterile conditions from the operation room to the laboratory, where it was immediately frozen at ⫺70°C. The tumor size ranged between 3 and 12 g. A 5-ml blood sample was obtained prior to surgery. Preparation of tumor vaccines expressing ␣-gal epitopes To synthesize ␣-gal epitopes on tumor membranes, we cloned the ␣1,3GT cDNA from a New World monkey [32] and expressed it in a soluble form in the yeast expression system of Pichia pastoris [33]. In this expression system the enzyme is secreted by the yeast into the culture medium, isolated by the (His)6 tag on nickel-Sepharose column, and eluted from the column by imidazole [33]. Synthesis of ␣-gal epitopes on these carbohydrate chains by recombinant (r)␣1,3GT is performed in a two-step reaction illustrated in Fig. 1. Sialic acid (SA) is removed by neuraminidase to expose the penultimate N-acetyllactosamine (Gal␤1-4GlcNAc-R) residues. Subsequently, r␣1,3GT synthesizes ␣-gal epitopes on these N-acetyllactosamine residues by linking galactose (Gal), which contributed by the sugar donor uridine diphosphate-galactose (UDP-Gal). This second step is similar to the natural synthesis of ␣-gal epitopes within the Golgi apparatus in cells of nonprimate mammals and New World monkeys. For synthesis of ␣-gal epitope on ovarian carcinoma membranes, tumor tissues were thawed, homogenized under sterile conditions, and washed four times in 200 ml of cold saline by centrifugation at 30,000 g. The membranes were resuspended at 100 mg/ml in enzyme buffer containing 0.1 M Methylethylmorpholino sulfate (MES) (pH 6.2), 25 mM MnCL2, and 1 mM UDP-Gal (Boehringer-Mannheim, Germany, received as a generous gift from Neose Inc. Horsham, PA). Subsequently, the membrane suspensions were incubated with constant rotation for 2 h at 37°C with neuraminidase (1 mU/ml; Sigma, St. Louis, MO), with r␣1,3GT (50 ␮g/ml) [33], or with both neuraminidase and r␣1,3GT at these concentrations. Membranes incubated in the enzyme buffer in the absence of the two enzymes served as controls. The r␣1,3GT concentration of 50 ␮g/ml was determined to be the working concentration, since in separate studies we

Analysis of ␣-gal epitope expression on tumor membranes by ELISA Suspensions of membrane homogenates in phosphatebuffered saline (PBS) were placed as 50-␮l aliquots in enzyme-linked immunosorbent assay (ELISA) wells (Falcon 3219), and dried overnight. This results in firm adherence of the membranes to the ELISA wells. The plates were blocked with 1% bovine serum albumin (BSA) in PBS. Subsequently, the ELISA plates were washed, incubated for 2 h with the monoclonal anti-Gal IgM antibody M86, at various dilutions of the antibody [34]. In parallel, plates were incubated with human anti-Gal, isolated from human AB serum, on affinity columns of ␣-gal epitopes linked to silica beads [23,24]. At the end of incubation, the antibodies were removed, the plates washed and incubated with horseradish peroxidase (HRP)-conjugated anti-mouse IgM (Accurate Labs, Westbury, NY) as secondary antibody for the monoclonal M86, or with HRP-conjugated anti-human IgG (Dako, Copenhagen, Denmark), as secondary antibody for the human anti- Gal. After 1-h incubation at room temperature, plates were washed, the color reaction developed with O-phenylenediamine (OPD), and absorbance measured at 492 nm. An independent measure of ␣-gal epitopes on the tumor membranes was the binding of the HRP-coupled lectin Bandeiraea (Griffonia) simplicifolia IB4 (BS lectin) (Vector Laboratories, Burlingame, CA), at a concentration of 10 ␮g/ml, or lower. This lectin binds specifically to ␣-gal epitopes [35]. The color development was similar to that used for anti-Gal. Analysis of anti-Gal activity by ELISA Anti-Gal IgG activity in ovarian carcinoma patients was determined in serum samples obtained prior to surgery, as previously described [36]. ELISA plates were coated with synthetic ␣-gal epitopes linked to BSA (␣-gal-BSA, Dextra, Reading, UK) as solid-phase antigen. Serum samples at various dilutions were incubated in the ELISA wells for 2 h. Subsequently the plates were washed, incubated with HRPcoupled anti-human IgG, and color development performed as above. In a parallel assay, anti-Gal binding was studied with the various tumor membranes as solid-phase antigen. For this purpose, the tumor membranes at 2 mg/ml in PBS were dried in ELISA wells, as 50-␮l aliquots. Once dried, the membrane fragments adhere firmly to the wells and can serve as solid-phase antigen. The specific anti-Gal binding

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

to the tumor membranes was measured and calculated by subtracting the optical density (O.D.) in wells containing the original unprocessed membranes. Western blots for analysis of ␣-gal epitope expression on glycoproteins Twenty micrograms of processed or control membranes were electrophoresed in 12% SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis), then transferred to a polyvinylidene diflouride (PVDF) membrane by semidry electroblotting, as we have previously described [33]. The proteins in this amount of tumor membranes can be readily detected in the gels by nonspecific Coomassie staining (not shown). The electroblotted membrane was blocked overnight at 4°C in PBS with 3% defatted milk. The membrane then incubated with human anti-Gal (5 ␮g/ ml) in PBS containing 1% BSA, for 2 h at room temp. Subsequently, the membrane was washed and incubated for 1 h with HRP-conjugated anti-human IgG. Color reaction was developed with diaminobenzidine (DAB; Sigma) substrate. ELISA inhibition assay for quantification of ␣-gal epitopes on processed tumor membranes The quantification of ␣-gal epitope expression on the processed membranes was performed as we have previously described [34,37,38]. Tumor membranes at an initial concentration of 200 mg/ml were subjected to serial 2-fold dilutions in 100-␮l aliquots of PBS containing 1% BSA. The membranes in each dilution were mixed with an equal volume of monoclonal anti-Gal M86 at the final dilution of 1:100 of the antibody (a concentration of the antibody that yields in ELISA O.D. absorption value at the slope of the binding curve). The mixture was incubated overnight at 4°C with constant rotation to enable maximum binding of the antibody to ␣-gal epitopes. Subsequently, the membranes and bound antibodies were removed by centrifugation and the activity of free M86 antibody remaining in the supernatant was determined by ELISA with ␣-gal BSA as solidphase antigen, using HRP-conjugated goat anti-mouse IgM antibody as secondary antibody. The amount of free antiGal M86 antibody remaining in the supernatant, after the removal of tumor membranes and anti-Gal M86 antibody bound to them, is reciprocally proportional to the number of epitopes expressed on the tumor membranes. By comparing the binding of anti-Gal M86 to membranes with known concentration of ␣-gal epitopes, with that of the antibody binding to the processed tumor membranes, it is possible to determine the concentration of ␣-gal epitopes synthesized de novo on tumor membranes [37,38]. Rabbit red cell membranes were used as the standard cells because they were previously shown to express 2 ⫻ 1013 ␣-gal epitopes per milligram of cell membranes [34].

103

Results Synthesis of ␣-gal epitopes on ovarian carcinoma tumor membranes Tumor membranes from five patients were studied for synthesis of ␣-gal epitopes. Representative data with the tumor membranes from various patients (two representative patients in each group) are shown in Fig. 2. Ovarian carcinoma membranes expressed an abundance of ␣-gal epitopes after incubation with both neuraminidase (1 mU/ml) and r␣1, 3GT (50 ␮g/ml), as indicated by the extensive binding of the monoclonal anti-Gal M86 (Fig. 2A and B), human anti-Gal (Fig. 2C and D), and of BS lectin (Fig. 2E and F). In contrast, control membranes, or membranes incubated only with neuraminidase, completely lacked ␣-gal epitopes and thus, did not bind the antibodies, or lectin. Membranes incubated with r␣1, 3GT, in the absence of neuraminidase, also expressed ␣-gal epitopes. However, the binding of antibodies and lectin suggest a lower number of ␣-gal epitopes than those on membranes incubated with both neuraminidase and r␣1,3GT. Distribution of ␣-gal epitopes on tumor membrane glycoproteins Western blot analysis of tumor membranes immunostained with human anti-Gal is shown in Fig. 3. Glycoproteins of untreated ovarian carcinoma membranes of two patients (lanes 1 and 3) did not bind the anti-Gal antibody because they completely lack ␣-gal epitopes. However, membranes incubated with neuraminidase and r␣1,3GT, in the presence of UDP-Gal, displayed multiple band staining in the form of a smear along the full length of the lanes (lanes 2 and 4). This pattern of staining implies that a very large number of glycoprotein molecules with different sizes, on the ovarian carcinoma membranes, were subjected to the enzymatic activity of r␣1,3GT, resulting in multiple expression of ␣-gal epitopes on their carbohydrate chains. Quantification of ␣-gal epitopes on processed ovarian carcinoma membranes The number of synthesized epitopes on ovarian carcinoma can be derived from the data on the ELISA inhibition assay presented in Fig. 4. Untreated membranes, or membranes incubated only with neuraminidase, did not bind the M86 antibody even at the high concentration of 200 mg/ml (i.e., 0% inhibition of M86 binding), implying the complete absence of ␣-gal epitopes on these membranes. Tumor membranes treated only with r␣1,3GT, or with neuraminidase and r␣1,3GT, both bound the antibody; however, the latter membranes were twice as effective as those treated only with r␣1,3GT. The tumor membranes incubated with the two enzymes displayed an ⬃100-fold lower inhibitory activity than that of rabbit red cells (i.e., 50% inhibition

104

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

Fig. 2. Binding of the monoclonal anti-Gal M86 (A and B), the natural human anti-Gal (C and D), and of BS lectin (E and F) to ovarian carcinoma membranes processed to express ␣-gal epitopes. The binding was assayed by enzyme-linked immunosorbent assay with processed membranes as solid-phase antigen. Control membranes (䡬); membranes incubated only with neuraminidase (1 mU/ml) (䊐); membranes incubated with r␣1,3GT (50 ␮g/ml) in the absence of neuraminidase (‚); membranes incubated with both neuraminidase (1 mU/ml) and r␣1,3GT (50 ␮g/ml) (●). Data from two representative ovarian carcinoma specimen of five with similar results. The patient’s number is included in the title of each figure.

were observed at 25 and 0.25 mg/ml, respectively). Since 1 mg of rabbit red cells has 2 ⫻ 1013 ␣-gal epitopes/mg of ␣-gal epitopes [34], ovarian carcinoma membranes incu-

bated with r␣1,3GT and neuraminidase expressed ⬃2 ⫻ 1011 ␣-gal epitopes/mg and membranes incubated only with r␣1,3GT expressed a two-fold lower concentration, i.e., ⬃1

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

105

Fig. 3. Western blot staining of ovarian carcinoma membrane glycoproteins with the human natural anti-Gal antibody. Lane 1, control tumor membranes from patient 1; lane 2, tumor membranes from the same patient, incubated with neuraminidase, r␣1,3GT, and UDP-Gal; lane 3, control tumor membranes from patient 3; lane 4, tumor membranes from patient 3 incubated with neuraminidase, r␣1,3GT, and UDP-Gal. Note that many of the proteins on processed tumor membranes have ␣-gal epitopes and, thus, are stained by anti-Gal as a smear throughout the lane.

⫻ 1011 ␣-gal epitopes/mg. Membranes from the other ovarian carcinoma specimens, which were incubated with neuraminidase and r␣1,3GT, displayed 5 ⫻ 1010 to 4 ⫻ 1011

Fig. 5. Demonstration of anti-Gal IgG activity in the serum of four ovarian carcinoma patients (䡬, ‚, 䉫, 䊐) as measured by (ELISA) with ␣-gal BSA as solid-phase antigen. Anti-Gal activity in a representative healthy individual (●). (B) Binding of anti-Gal in serum of three patients (diluted 1:10) to ␣-gal epitopes expressed on autologous processed ovarian carcinoma membranes, as measured by ELISA. Open columns, neuraminidase-treated membranes; hatched columns, r␣1,3GT-treated membranes; closed columns, membranes treated with neuraminidase and r␣1,3GT.

Fig. 4. Quantification of ␣-gal epitope expression on untreated ovarian carcinoma membranes (䡬); ovarian carcinoma membranes incubated only with neuraminidase (䊐); ovarian carcinoma membranes incubated only with r␣1,3GT (‚), or with both r␣1,3GT and neuraminidase (●). Rabbit red cells membranes served as standard control which express 2 ⫻ 1013 ␣-gal epitopes per milligram of red cells (䉬). The data are of a representative tumor obtained from on out of 5 patients with similar results. Percentage (%) of inhibition is determined as the reciprocal proportion of monoclonal anti-Gal M86 antibody remaining in the supernatant after overnight incubation with the assayed tumor membranes. The remaining anti-Gal that is not bound to the tumor membranes is determined by enzyme-linked immunosorbent assay. Note that the original tumor membranes, or membranes treated only with neuraminidase, bound no monoclonal anti-Gal (i.e., 0% inhibition) because they lack ␣-gal epitopes. In contrast, tumor membranes express ␣-gal epitopes that bind anti-Gal after treatment with r␣1,3GT and to a larger extent after treatment with both neuraminidase and r␣1,3GT, as indicated by the percentage of inhibition at the various membrane concentrations.

␣-gal epitopes/mg, whereas untreated membranes completely lacked this epitope (not shown). Anti-Gal activity in sera of ovarian carcinoma patients The presence of circulating anti-Gal IgG in the serum of ovarian carcinoma patients, was demonstrated by ELISA with ␣-gal BSA as solid-phase antigen. As shown in Fig. 5A, anti-Gal activity was found in the serum of all ovarian carcinoma patients studied. Variations in anti-Gal in different individuals are within the range observed in the healthy population [21,24]. We could further show that these antiGal molecules can bind to the de novo synthesized ␣-gal epitopes on the autologous tumor membranes. The membranes were attached to ELISA wells and binding of IgG

106

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

antibodies was measured following incubation with autologous serum diluted 1:10. To prevent measuring of nonspecific binding of IgG to the ELISA wells, absorbance (O.D.; optical density units) measured following binding to untreated tumor membranes was subtracted from O.D. in each of the wells containing treated membranes. As expected, only minimal binding of autologous IgG was observed in wells containing neuraminidase-treated membranes (open columns in Fig. 5B). However, IgG antibodies readily bound to the autologous tumor membranes treated with r␣1, 3GT, implying that the binding is that of anti-Gal (hatched columns). In accord with the observations on the higher number of ␣-gal epitopes in membranes treated with both neuraminidase and r␣1, 3GT (Figs. 2 and 4), binding of anti-Gal to these membranes (filled columns) was higher than that to membranes treated only with r␣1,3GT. The binding of anti-Gal within the patients’ sera to the enzymatically treated autologous tumor membranes was lower than that observed in ELISA wells coated with ␣-gal BSA. This is because the number of ␣-gal epitopes on the tumor membranes is significantly lower than that on ␣-gal BSA used as solid-phase antigen. This binding to tumor membranes was also lower than that observed with isolated human anti-Gal in Fig. 2C and D, because the concentration and affinity of the isolated anti-Gal are higher than those in the patients’ sera, as a result of the affinity column isolation process for obtaining purified anti-Gal from pooled normal human sera [22–24].

Discussion The present study describes a novel method for preparing autologous ovarian carcinoma vaccine by processing ovarian carcinoma membranes in such a way that they can be opsonized by an abundant natural human IgG antibody. Such in situ binding can target the vaccinating membranes to APC, thus increasing the uptake, processing, and presentation of the autologous TAA to the immune system. This method is based on the synthesis of ␣-gal epitopes on the tumor membranes by r␣1,3GT and on the presence of the natural anti-Gal antibody in all humans, including ovarian carcinoma patients. Our data imply that there are many terminal N-acetyllactosamine groups on carbohydrate chains of the glycoproteins of ovarian carcinoma membranes on which r␣1,3GT can synthesize ␣-gal epitopes. However, the number of these epitopes can be doubled to ⬃2 ⫻ 1011/mg of membrane, if additional N-acetyllactosamine groups are exposed by removal of sialic acid with neuraminidase. The binding of human anti-Gal from normal serum, or of anti-Gal in the autologous serum, to these ␣-gal epitopes was observed both in Western blots and in ELISA. These findings strongly suggests that a similar binding of the patient’s anti-Gal will occur in situ, at the vaccination site, thus targeting the vaccinating membranes for effective uptake by APC. Because the ␣-gal epitope is covalently

linked to the membranes, it is probable that it would be highly stable in vivo, until the membranes are taken up be APC that bind to the opsonizing anti-Gal. Since anti-Gal IgG is capable of diffusing out of the blood vessels, it is likely that the local trauma to capillaries, at the vaccination site, will suffice for accumulation of extra vascular anti-Gal IgG, in amounts that will enable effective opsonization of the tumor membranes. It could be argued that some of the TAA are within the cytoplasm of the cancer cells, rather than on their cell membranes. Therefore, cytoplasmic TAA may not be targeted to APC when the autologous tumor vaccine consists of tumor membranes. It is currently difficult to obtain large enough numbers of intact cells as vaccines in patients with solid tumors because of the paucity of live cells that can be isolated from such tumors. Based on the information on the well-characterized TAA in melanoma [3], it is probable that many of the uncharacterized TAA are expressed on the cell membranes. In addition, some of the cytoplasmic TAA are likely to be attached to the inner part of the cell membrane, and thus will be internalized by APC that phagocytoze vaccinating tumor membranes opsonized by anti-Gal. We propose to study this autologous tumor vaccine processed to express ␣-gal epitopes, in ovarian carcinoma patients. The vaccinating membranes will be irradiated with 50 Gy to ensure the killing of any live tumor cells that accidentally may survive the homogenization of the tumor tissue and the freezing and thawing of the processed membranes. The dose of 50 Gy was found to be 4-fold higher than that required to kill the ovarian carcinoma cell line OVCAR 3 in vitro (not shown). Neither irradiation, nor freezing and thawing, affect ␣-gal epitope expression on the processed membranes. Because of the sterile processing and the irradiation, the processed membranes were found to be sterile. Moreover, the vaccine preparations were found to lack any endotoxin activity, thus they are suitable for clinical use as vaccines. Patients will receive six vaccinations of up to 80 mg of tumor membranes per vaccine and anti-tumor immune response tested both in vivo as delayed type hypersensitivity reaction in the skin, and in vitro by analyzing TH1 and TH2 activation by ELISPOT with the tumor membranes as the stimulatory antigens. The use of this vaccine has been approved by the FDA for a phase I study (BB-IND 9685). Various studies suggest that no major toxicity is to be anticipated with the ␣-gal-modified ovarian carcinoma vaccine. Studies on immunization of melanoma patients with their autologous tumor expressing the hapten dinitrophenol (DNP) demonstrated no toxicity in a very large number of patients [39,40]. Similarly, vaccination of leukemia patients with autologous leukemia cells that were treated with neuraminidase resulted in no adverse effects [41,42]. These studies in melanoma and leukemia patients further suggest that the vaccination with autologous tumor vaccines is unlikely to cause breakdown of tolerance to self-antigens and may not induce autoimmunity. Nevertheless, careful immu-

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

nological monitoring in a phase I study of ovarian carcinoma patients will help to determine whether autoimmunity is a risk in patients treated with autologous tumor vaccines. The presence of the de novo synthesized ␣-gal epitopes on the vaccinating membranes is not expected to be harmful. This can be inferred from previous studies on transplantation of pig cells into humans [43]. In these studies, 10 diabetic patients were transplanted with 3– 6 g of fetal pig islet cells, either into the portal vein or under the kidney capsule. These pig islet cells express ␣-gal epitopes and interact with anti-Gal. Nevertheless, such in vivo interaction resulted in no adverse effects to the patients [43], suggesting that a similar interaction between anti-Gal and ␣-gal epitopes on vaccinating tumor membranes are of no risk to treated patients. Overall, the anti-Gal mediated in situ targeting of immunizing tumor membranes to APC may provide the immune system with an opportunity to be activated effectively by TAA in the autologous tumor. In some of the patients this improved immune response to TAA may be potent enough to destroy micrometastases expressing these TAA. This strategy may contribute to the active immunotherapy of patients with ovarian carcinoma and deserves further study.

Acknowledgments This study was supported by NIH grant CA85868.

[13]

[14] [15] [16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

References [24] [1] Livingston P. Active specific immunotherapy in the treatment of patients with cancer. Immunol Allerg Clin North Am 1991;11:401– 10. [2] Pardoll DM. Spinning molecular immunology into successful immunotherapy. Nat Rev Immunol 2002;2:227–38. [3] Hainaut P, Weynants P, Coulie PG, Boon T. Antitumor T-lymphocyte response. Med Clin North Am 1990;10:639 –53. [4] Old LJ. Immunotherapy for cancer. Sci Am 1996;Sept:136 – 43. [5] Pardoll DM. Cancer vaccines. Immunol Today 1993;14:310 – 6. [6] Guinan EC, Gribben JG, Boussiotis VA, Freeman GJ, Nadler LM. Pivotal role of the B7: CD28 pathway in transplantation, tolerance and tumor immunity. Blood 1994;84:3261– 82. [7] Chen L, Ashe S, Brady WA, Hellstro¨ m I, Hellstro¨ m KE, Ledbetter JA, et al. Costimulation of anti-tumor immunity by the B7 counterreceptor for the T-lymphocyte molecule CD-28 and CTLA-4. Cell 1992;71:1093–102. [8] Townsend SE, Allison JP. Tumor rejection after direct costimulation of CD8⫹ T cells by B7-transfected melanoma cells. Science 1993; 259:368 –70. [9] Lanzavecchia A. Mechanism of antigen uptake for presentation. Curr Opin Immunol 1996;8:348 –54. [10] Grabbe S, Beisserts S, Schwartz T, Granstein RD. Dendritic cells as initiators of tumor immune response: a possible strategy for tumor immunotherapy. Immunol Today 1995;16:117–21. [11] Schwartz RH. Costimulation of T lymphocytes: the role of CD28, CTLA-4 and B7/BB1 in interleukin-2 production and immunotherapy. Cell 1992;71:1065– 8. [12] Maass G, Schmidt W, Berger M, Schilcher F, Koszik F, Schneeberger A, et al. Primary of tumor-specific T cells in the draining lymph nodes

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

107

after immunization with interleukin-2 secreting tumor cells: three consecutive stages may be required for successful tumor vaccination. Proc Natl Acad Sci USA 1995;92:5540 – 4. Zinkernagel RM, Ehl E, Aichele P, Ku¨ ndig T, Hengartner H. Antigen localization regulates immune responses in a dose and time dependent fashion: a geographical view of immune reactivity. Immunol Rev 1997;156:199 –209. Nossal DJV. Tolerance and ways to break it. Ann NY Acad Sci 1993;690:34 – 41. Unkeless JC. Functional heterogeneity of human Fc receptors for immunoglobulin G. J Clin Invest 1989;83:355–561. Fanger NA, Wardwell K, Shen L, Tedder TF, Guyre PM. Type I (CD64) and type II (CD32) Fc receptor-mediated phagocytosis by human blood dendritic cells. J Immunol 1996;157:541– 8. Schmitt DA, Hanan D, Bieber T, Dezutter-Dambuyant C, Schmitt D, Fabre M, et al. Human epidermal Langerhans cells express only trhe 40-kilodalton Fc␥ receptor (FcRII). J Immunol 1990;144:4284 –90. Clynes R, Takechi Y, Moroi Y, Houghton A, Ravetch JV. Fc receptors are required in passive and active immunity to melanoma. Proc Natl Acad Sci USA 1998;95:652– 656. LaTemple DC, Henion TR, Anaraki F, Galili U. Synthesis of ␣-galactosyl epitopes by recombinant ␣1,3galactosyl transferase for opsonization of human tumor cell vaccines by anti-galactose. Cancer Res 1996;56:3069 –74. Galili U, LaTemple DC. The natural anti-Gal antibody as a universal augmenter of autologous vaccine immunogenicity. Immunol Today 1997;18:281–5. Galili U, Rachmilewitz EA, Peleg A, Flechner I. A unique natural human IgG antibody with anti-␣-galactosyl specificity. J Exp Med 1984;160:1519 –31. Galili U, Macher BA, Buehler J, Shohet SB. Human natural anti-␣galactosyl IgG. II. The specific recognition of ␣(1–3)-linked galactose residues. J Exp Med 1985;162:573– 82. Galili U, Buehler J, Shohet SB, Macher BA. The human natural anti-Gal IgG. III. The subtlety of immune tolerance in man as demonstrated by crossreactivity between natural anti-Gal and anti-B antibodies. J Exp Med 1987;165:693–704. Galili U. Evolution and pathophysiology of the human natural antiGal antibody. Springer Semin Immunopathol 1993;15:155–71. Galili U, Mandrell RE, Hamadeh RM, Shohet SB, Griffis JM. Interaction between human natural anti-␣-galactosyl immunoglobulin G and bacteria of the human flora. Infect Immun 1988;56:1730 –7. Galili U, Clark MR, Shohet SB, Buehler J, Macher BA. Evolutionary relationship between the anti-Gal antibody and the Gal␣1-3Gal epitope in primates. Proc Natl Acad Sci USA 1987;84:1369 –73. Galili U, Shohet SB, Kobrin E, Stults CLM, Macher BA. Man, apes, and Old World monkeys differ from other mammals in the expression of ␣-galactosyl epitopes on nucleated cells. J Biol Chem 1988;263: 17755– 67. Galili U. Interaction of the natural anti-Gal antibody with ␣-galactosyl epitopes: a major obstacle for xenotranplantation in humans. Immunol Today 1993;14:480 –2. Collins BH, Cotterell AH, McCurry KR, Alvarado CG, Magee JC, Parker W, et al. Cardiac xenografts between primate species provide evidence for the importance of ␣-galactosyl determinant in hyperacute rejection. J Immunol 1995;154:5500 –10. Galili U, Chen ZC, Manches O, Plumas J, Preisler H. Preparation of autologous leukemia and lymphoma vaccines expressing ␣-gal epitopes. J Hematother Stem Cell Res 2010;10:501–11. LaTemple DC, Abrams JT, Zhang SU, Galili U. Increased immunogenicity of tumor vaccines complexed with anti-Gal: studies in knock out mice for ␣1,3galactosyltranferase. Cancer Res 1999;59: 3417–23. Henion TR, Macher BA, Anaraki F, Galili U. Defining the minimal size of catalytically active primate ␣1,3galactosyltransferase: structure function studies on the recombinant truncated enzyme. Glycobiology 1994;4:192–210.

108

U. Galili et al. / Gynecologic Oncology 90 (2003) 100 –108

[33] Chen ZC, Tanemura M, Galili U. Synthesis of ␣-gal epitopes on human tumor cells by recombinant ␣1,3galactosyltransferase produced in Pichia pastoris. Glycobiology 2001;11:577– 86. [34] Galili U, LaTemple DC, Radic MZ. A sensitive assay for measuring ␣-Gal epitope expression on cells by a monoclonal anti-Gal antibody. Transplantation 1998;65:1129 –32. [35] Wood C, Kabat EA, Murphy LA, Goldstein IJ. Immunochemical studies on the combining sites of two isolectins A4 and B4 isolated from Bandeiraea simplicifolia. Arch Biochem Biophys 1979;198:1–11. [36] Galili U, Tibell A, Samuelsson B, Rydberg L, Groth CG. Increased anti-Gal activity in diabetic patients transplanted with fetal porcine islet cell clusters. Transplantation 1995;59:1549 –56. [37] Stone KR, Ayala G, Goldstein J, Hurst R, Walgenbach A, Galili U. Porcine cartilage transplants in cynomolgus monkey: III. Transplantation of ␣-galactosidase treated porcine cartilage. Transplantation 1998;65:1577– 83. [38] Tanemura M, Maruyama S, Galili U. Differential expression of ␣-gAL epitopes (Gal␣1-3Gal␤1-4GlcNAc-R) on pig and mouse organs. Transplantation 2000;69:187–90.

[39] Berd D, Maguire Jr HC, Schuchter LM, Hamilton R, Hauch WW, Soto T, et al. Autologous, hapten-modified melanoma vaccine as post-surgical adjuvant treatment after resection of nodal metastases. J Clin Oncol 1997;15:2359 –70. [40] Berd D, Sato T, Cohn H, Maguire Jr HC, Mastrangelo MJ. Treatment of metastatic melanoma with autologous, hapten-modified melanoma vaccine: regression of pulmonary metastases. Int J Cancer 2001;94:531–9. [41] Bekeshi JG, Holland JF, Roboz JP. Specific immunotherapy with neuraminidase-modified leukemic cells: experimental and clinical trials. Med Clin North Am 1977;61:1083–100. [42] Bekesi JG, Roboz JP, Holland JF. Therapeutic effectiveness of neuraminidase treated tumor cells as an immunogen in man and experimental animals with leukemia. Ann NY Acad Sci 1976;277: 313–9. [43] Groth CG, Korsgren O, Tibel A, Tollerman J, Muller E, Bolinder J, et al. Transplantation of fetal porcine pancreas to diabetic patients: biochemical and histological evidence for graft survival. Lancet 1994; 344:1402– 4.