Bioengineering of Surface GD3 Ganglioside for Immunotargeting Human Melanoma Cells*

N -Propionyl, N -butyryl ( N -Bu), and N -benzoyl mannosamine, as precursors of sialic acid biosynthesis, were incubated with human melanoma SK-MEL-28 cells and resulted in the replacement of N -acetyl groups on the cell surface sialic acid residues, including those associ-ated with GD3. Meanwhile, vaccines containing GD3 and modified GD3 tetrasaccharide-keyhole limpet hemocyanin conjugates were synthesized, and BALB/c mice were immunized with them together with mono-phosphoryl lipid A adjuvant. The GD3Bu-keyhole limpet hemocyanin conjugate raised the highest IgG titers without any cross-reactivity to unmodified GD3. Expression of GD3Bu epitopes on the surface of SK-MEL-28 cells was confirmed in vitro and in vivo by the binding of a polyclonal antiserum and monoclonal antibody (mAb) 2A, both of which specifically recognize GD3Bu, and by mass spectroscopic analysis of glycolipids extracted from cells. Following expression of GD3Bu on the surface of SK-MEL-28 cells, the cells could be lysed by mAb 2A and GD3Bu antiserum in the presence of complement. Although less effective in the control of existing large size tumors ( (cid:1) 10 mm inner diameter) on BALB/c nu / nu mice, mAb 2A in combination with ManNBu effec-tively conjugates. Disialolac- toside is the carbohydrate epitope of GD3. This enzymatically synthesized GD3 tetrasaccharide was chemically modified on the N -acyl sub- stitution of both sialic acids by base treatment to remove N -acetyl groups and followed by N -acylation using propionyl anhydride, butyryl anhydride, and benzoyl chloride, respectively. The azido group was quantitatively reduced to amine by catalytic hydrogenation, which was then coupled to a maleimide-containing spacer. The final conjugates were formed by reaction of sulfhydryl-KLH with the maleimide group in GD3 tetrasaccharide derivatives.

Another therapeutic strategy is to use GD3 as a potential target for passive immunotherapy. A murine mAb (R24) recognizing GD3 has been tested in patients in various clinical trials. Regression of melanoma metastasis after treatment has been documented repeatedly (22)(23)(24)(25)(26). More recently, a chimeric antibody KM871 showed its efficacy in the treatment of melanomas in a nude mice model by slowing the tumor growth (27). The humanized antibody tested in a phase I clinical trial is not immunogenic in man (28) and is directed to the tumor site in mice (29). Although these approaches are promising, neither active vaccination nor antibody administration are completely successful (30,31), and new and improved treatments are obviously needed.
We have reported recently (32) that poorly immunogenic PSA on the surface of RMA leukemia cells can be biochemically engineered to express N-propionyl PSA by using ManNPr as a precursor, and that the resultant cells became susceptible to the treatment with an N-propionyl PSA-specific monoclonal antibody in vitro and in vivo. In this work we have extended the same strategy to another poorly immunogenic GD3 antigen. As a result of biochemically engineering the sialic acid residues of the GD3 on the surface of SK-MEL-28 melanoma cells, the neoantigen (modified GD3) formed on the cell surface serves as a target for cytolytic antibodies. Since the specific antibodies are not cross-reactive with unmodified GD3 on the cell surface, this approach, in principle, may enable us to turn on and turn off immune response to this neoantigen by adding or withdrawing N-acylmannosamine precursors during treatment. Consequently, this approach may avoid the risk of a possible autoimmune response (33).

Animals and Cell Lines
BALB/c mice (female, 6 -8 weeks), BALB/c nu/nu, and CgFoxNude nude mice (female 6 weeks old) were purchased from Charles River Breeding Laboratories (Montreal, Quebec, Canada) and The Jackson Laboratory (Bar Harbor, ME) and were maintained in our Institutional Animal Facility following the animal care guidelines. SK-MEL-28 and G361 human melanoma cell lines and mouse monoclonal antibody R24 (IgG3 specific to GD3) were purchased from the American Type Culture Collection (Manassas, VA).

Chemicals and Reagents
Mannosamine, N-acetylmannosamine, and other chemicals were purchased from Aldrich unless stated otherwise. KLH and cross-linking * This is National Research Council of Canada Publication Number 42490. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
N-Deacetylation of GD3 Tetrasaccharide-A solution of GD3Ac tetrasaccharide (50 mg) in 2 N NaOH (10 mg/ml) was heated at 100°C for 4 h. After cooling, the solution was carefully neutralized by the addition of 2 N HCl and purified by passage through a Bio-Gel P-6 column, using 0.03 M NH 4 HCO 3 as eluent. The product obtained after lyophilization was an amorphous solid in almost quantitative yield.
N-Acylation of N-Deacetylated GD3 Tetrasaccharide-To a solution of N-deacetylated GD3 tetrasaccharide (5 mg) in 5% Na 2 CO 3 (2.5 ml), propionic anhydride at room temperature (10 l, three times with 10-min intervals) was added with vigorous stirring. After 30 min, the mixture was adjusted to pH 11.0 by the addition of 2 N NaOH and kept for 1 h. The solution was then adjusted to pH 8.0 by the addition of 0.5 N HCl. Purification on a Sephadex G-10 column using water as eluent afforded, after lyophilization, GD3Pr tetrasaccharide as an amorphous solid in almost quantitative yield.
To a solution of N-deacetylated GD3 tetrasaccharide (5 mg) in a mixture of 5% Na 2 CO 3 (2.5 ml) and diethyl ether (2.5 ml), butyric anhydride (30 l) or benzoyl chloride (30 l) at room temperature was added with vigorous stirring. After 30 min the organic layer was removed, and the aqueous solution was adjusted to pH 11.0 by the addition of 2 N NaOH and kept for 1 h. The solution was then adjusted to pH 8.0 by 0.5 N HCl and passed through a Sephadex G-10 column, using water as eluent. The respective products, GD3Bu and GD3Bz tetrasaccharides, were obtained after lyophilization as amorphous solids in almost quantitative yields.
Reduction of Azido Group to Amine and Introduction of Maleimide-A solution of the above tetrasaccharides (5 mg each) in water (0.5 ml) was subjected to catalytic (Pd/C) hydrogenation (30 pounds/square inch) for 2 h, respectively. The filtrate was passed through a Sephadex G-10 column, using water as eluent. The lyophilized products (amines) were dissolved in 20 mM phosphate buffer (2 ml, pH 7.2) and mixed with N-(␥-maleimidobutyryloxy)sulfosuccinimide (5 mg each). The solution was kept at room temperature for 0.5 h, when TLC (CHCl 3 /MeOH/H 2 O 9:9:1) indicated the formation of a faster moving product. Purification was performed on a Sephadex G-10 column with water as eluent. Four maleimide-containing products were obtained as an amorphous solid after lyophilization in almost quantitative yields.

Determination of Antibody Titers by ELISA
Serum samples were titered against BSA conjugates in 96-well plates (ICN Biomedicals Inc, Aurora, OH). Wells were coated overnight at 20°C with 100 l of BSA conjugates (5 g/ml) in PBS per well. The wells were blocked by the addition of 100 l of 1% BSA in PBS for 2 h at 20°C. The plates were washed three times with a washing solution of 0.05% Tween 20 in PBS. Antisera serially diluted in 1% BSA, 0.05% Tween 20, PBS was added (100 l per well) and incubated for 3 h at 20°C. This was followed by washing three times as before followed by the addition of 100 l per well of alkaline phosphatase-labeled goat anti-mouse IgG or IgM (Caltag Laboratories, San Francisco, CA) diluted 1:2000 in 1% BSA, 0.05% Tween 20, PBS. The plates were incubated overnight at 20°C and again washed three times. 100 l of substrate solution containing p-nitrophenyl phosphate substrate (1 mg/ml, Kirkegaard & Perry lab, Gaithersburg, MD) was added to the wells. After 30 min at 20°C, the plates were scanned at 405 nm in an EL 800 (Biotek Instruments, Inc., Winooski, VE) microplate reader. The titer was defined as the highest dilution yielding an absorbance of Ն0.25.

Cell Culture Conditions
SK-MEL-28 and G361 human melanoma cells were cultured as adherent monolayers in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal calf serum inactivated at 56°C for 30 min (Invitrogen), 1 mM sodium pyruvate, 4 mM L-glutamine, and 100 units of penicillin/ streptomycin (Invitrogen) as the standard medium. Cells were cultured under a humidified atmosphere containing 5% CO 2 at 37°C. For flow cytometry experiments, the cells were trypsinized with trypsin-EDTA (Invitrogen) and suspended in PBS prior to the antiserum treatment. Cells were cultured in the presence of ManNAc, ManNPr, ManNBu, or ManNBz by the addition of the appropriate amount of the precursor to the culture medium.

GD3 Extraction and Mass Spectroscopic (MS) Analysis
The gangliosides were extracted from SK-MEL-28 cells according to the method of Svennerholm and Fredman (40) with simplification. Briefly, precursor-treated and untreated cells (ϳ1.0 ϫ 10 7 ) suspended in PBS were precipitated by centrifugation to pellets that were extracted twice with chloroform/methanol/water (4:8:3). The solvents were evaporated to residues, which were dissolved in methanol. Insoluble materials were removed by centrifugation, and the compounds in methanol were analyzed by capillary electrophoresis-MS (CRYSTAL CE System, Thermo Bioanalysis).

Monoclonal Antibody Production
Murine mAbs to GD3Bu-KLH were prepared by standard methods according to Plested et al. (41). Briefly the mice were immunized four times intraperitoneally, followed by one intravenous injection without adjuvant. Hybridomas were prepared by the fusion of spleen cells with Sp2/O-Ag 14 as described (42). Putative hybridomas secreting GD3Buspecific antibodies were selected by ELISA using synthetic GD3Bu-BSA as a coating antigen. Immunoglobulin class and subclass were also determined by ELISA. Two clones, mAbs 1 and 2A, were grown in BALB/c mice after treatment with pristane to generate ascitic fluid, which was purified through a protein A column (ImmunoPure Plus Immobilized Protein A IgG Purification Kit, Pierce).

Inhibition ELISA
For inhibition studies, mAb 1 or 2A was incubated with synthetic GD3Ac, GD3Pr, GD3Bu, and GD3Bz (from 0.1 to 100 g/ml) prior to addition to GD3Bu-BSA-coated plates and then assayed as described above.

Expression of Modified GD3 on SK-MEL-28 in Vivo
Three nude mice (CgFoxNude) were subcutaneously transplanted with 1 ϫ 10 7 SK-MEL-28 cells/mouse. When the tumors were about 10 mm in width, ManNBu was injected intraperitoneally daily (5 mg/ mouse) for 2 weeks (5 days/week). Cells from dissected tumors were analyzed by flow cytometry with mAb 2A and mAb R24 to detect the GD3Bu and GD3Ac expression, respectively.

CDC-mediated Antibodies
The assays were performed as described previously (27,43). Briefly, SK-MEL-28 and modified Bu-SK-MEL-28 melanoma cells (1 ϫ 10 6 cells) were labeled with 3.7 MBq Na 2 51 CrO 4 (Amersham Biosciences) for 1 h at 37°C and then washed twice with PBS. Samples of the labeled cells (50 l) were added into 96-well microtiter plates (1 ϫ 10 5 cells/ well) and incubated with 100, 50, 10, 5, 1, 0.5, 0.1, and 0.05 g of mAb 2A in PBS (50 l) or mouse serum (50 l, 400-fold dilutions in PBS) for 30 min at 4°C. Then 100 l/well of 10% rabbit complement (Cedarlane) were added to wells and incubated for 2 h at 37°C. After centrifugation 51 Cr release was counted in a 1450 Microbeta Trilux liquid scintillation and luminescence counter (Wallac, Helsinki, Finland). The percentage of specific lysis was calculated based on the experimental 51 Cr release, the total release, and the spontaneous release.

Evaluation of mAb against Established Tumors
Approximately 1.0 ϫ 10 7 SK-MEL-28 cells in 0.3 ml of RPMI medium were subcutaneously transplanted into the right flank of 10 BALB/c nu/nu mice. Tumors started to develop 4 -5 weeks after transplantation, and mice were used when tumors were ϳ10 mm in width. Five tumor-grafted nude mice were injected intraperitoneally with ManNBu in RPMI medium (5 mg/mouse) for 2 days prior to intraperitoneal injection of mAb 2A (200 g/mouse) and ManNBu (5 mg/mouse) for 2 weeks (5 days/week). Another five mice in a control group received only intraperitoneal injection of mAb 2A (200 g/mouse) for 2 weeks (5 days/week). The tumor size (volume) was calculated by the following formula: tumor size V (mm 3 ) ϭ 0.4 ϫ (major axis) ϫ (minor axis) 2 .

Evaluation of mAb against Tumor Grafting
Approximately 1.0 ϫ 10 7 SK-MEL-28 cells in 0.3 ml of RPMI medium were subcutaneously transplanted into the right flank of 48 BALB/c nu/nu mice. Administration of mAb 2A and/or ManNBu was started 3 days later. Three groups of 10 mice each were injected subcutaneously with ManNBu (5 mg/mouse), mAb 2A (200 g/mouse), or both ManNBu (5 mg/mouse) and mAb 2A (200 g/mouse), respectively; 8 mice in another group were injected intraperitoneally with both ManNBu (5 mg/mouse) and mAb 2A (200 g/mouse) for 2 weeks (5 days/week), using one group of 10 mice as control. Tumor incidence was observed within 30 days and was confirmed 45 days after tumor transplantation.

Synthesis of GD3 Tetrasaccharide-KLH Glycoconjugates-
Chemical syntheses of GD3 have been reported previously (44 -47). However, large scale synthesis of the GD3 tetrasaccharide is still difficult to achieve. Recently, a bi-functional sialyltransferase from Campylobacter jejuni, which transfers sialic acid to  1. Synthesis of disialoctoside-KLH conjugates. Disialolactoside is the carbohydrate epitope of GD3. This enzymatically synthesized GD3 tetrasaccharide was chemically modified on the N-acyl substitution of both sialic acids by base treatment to remove N-acetyl groups and followed by N-acylation using propionyl anhydride, butyryl anhydride, and benzoyl chloride, respectively. The azido group was quantitatively reduced to amine by catalytic hydrogenation, which was then coupled to a maleimide-containing spacer. The final conjugates were formed by reaction of sulfhydryl-KLH with the maleimide group in GD3 tetrasaccharide derivatives. both 3-O-Gal and 8-O-NeuAc, was cloned and expressed (35). We used this enzyme successfully for the synthesis of disialolactoside (GD3 tetrasaccharide) on a 200-mg scale. The main by-product was sialolactoside (GM3 trisaccharide), which was separated from disialolactoside by a Bio-Gel P-6 column eluted with 0.03 M NH 4 HCO 3 . Modified disialolactosides in which the N-acetyl groups of sialic acids were replaced by respective N-propionyl, N-butyryl, and N-benzoyl groups were derived in two steps, namely removal of N-Ac by base treatment and quantitative re-N-acylation with either propionyl and butyryl anhydride or benzoyl chloride. Consequently, four disialolactosides, GD3Ac, GD3Pr, GD3Bu, and GD3Bz tetrasaccharides, were obtained with a ␤-linked 3-azidopropyl spacer. The spectroscopic data of GD3Ac tetrasaccharide are listed in Table I. Catalytic reduction of the azido groups generated their respective amines quantitatively, which were subsequently converted to maleimide-containing disialolactosides as indicated by a singlet resonance at ␦ H 6.8 ppm in their 1 H NMR spectra. After reaction of the latter with thiolated KLH, four glycoconjugates, GD3Ac-KLH, GD3Pr-KLH, GD3Bu-KLH, and GD3Bz-KLH, were obtained (see Fig. 1). The ratios of disialolactoside to KLH were between 300 and 460 based on sialic acid analysis (39) and protein assay (see Table II). By using the same procedures, four respective BSA conjugates were also prepared with 5-11 disialolactosides attached to each BSA molecule.
Immunizations-BALB/c mice were immunized and boosted three times weekly with GD3Ac-KLH, GD3Pr-KLH, GD3Bu-KLH, and GD3Bz-KLH conjugates. Ten days after the last boost (day 31), whole serum was analyzed by ELISA using GD3Ac-BSA, GD3Pr-BSA, GD3Bu-BSA, and GD3Bz-BSA as coating antigens. The antibody levels, both for IgG and IgM, were determined (see Table III). All four conjugates gave high titers of antibodies. In addition, subtyping analysis revealed that a majority of the specific antibodies was IgG rather than IgM. The GD3Bu-KLH conjugate was the most immunogenic, followed by GD3Pr-KLH, GD3Ac-KLH, and GD3Bz-KLH. The extension of the N-acyl chain seemed to correlate with the increased immunogenicity, except that the N-benzoylated de-rivative was the least immunogenic. Since the BSA conjugates used in ELISA shared the same linkage structure with the KLH conjugates, there was a possibility that antibodies to the linkage structure might also be raised, which could contribute in part to the total antibody titers. Therefore, we also investigated whether these antisera recognize GD3Ac, GD3Pr, GD3Bu, and GD3Bz expressed on the SK-MEL-28 cell surface as well as their cross-reactivity with native cell surface GD3.
Biochemical Engineering of SK-MEL-28 Melanoma Cells in Vitro-SK-MEL-28 and G361 cells were cultured for 1-3, 6, or 10 days in standard medium supplemented with three different concentrations (1, 3, and 5 mg/ml) of ManNAc, ManNPr, Man-NBu, and ManNBz. The expression of modified GD3 was evaluated by flow cytometric analysis using diluted sera raised against the GD3 tetrasaccharide-KLH conjugates. SK-MEL-28 cells cultured in the presence of the precursors exhibited no detectable reactivity with pre-immunization sera from mice. When grown in standard medium without precursors, SK-MEL-28 cells exhibited only background fluorescence staining after treatment with dilutions of anti-GD3Pr (Bu and Bz) sera (1:100). However, cells cultured in the presence of ManNPr, ManNBu, or ManNBz exhibited a marked increase in the immunoreactivity with their respective anti-GD3Pr, anti-GD3Bu, and anti-GD3Bz sera (Fig. 2a). No increase in reactivity to the anti-GD3Ac serum was observed when ManNAc was added to the medium. Precursors at a concentration of 1 mg/ml resulted in the expression of modified GD3 within 24 h, and increased precursor concentration (3 and 5 mg/ml) did not advance further expression (Fig. 2a). In addition, incubations of SK-MEL-28 cells with precursors (ManNBu and ManNBz) for 1-3, 6, or even 10 days did not significantly change the expression of GD3Bu and GD3Bz (Fig. 2b). After the removal of precursors from the growth medium, the modified GD3 was slowly replaced by GD3, yet the cells were still found to express GD3Bu after 5 days without ManNBu (Fig. 2c). Similar results were also observed with G361 cells (data not shown).  a Data are based on resorcinol assay of sialic acid (39) and BCA protein assay (Pierce) using sialic acid and BSA as the standards. The molecular weight of KLH was Ϸ 8.6 ϫ 10 6 .
Cross-reactivity between the above antisera and the cell surface GD3Bu and GD3Ac of SK-MEL-28 cells was also analyzed. Cell surface GD3Bu and anti-GD3Pr sera were crossreactive as well as GD3Bu and anti-GD3Bz sera (Fig. 3a), but no cross-reactivity was observed between the cell surface GD3Ac and anti-GD3Pr, anti-GD3Bu, or anti-GD3Bz sera (Fig.  3b). In these experiments mAb R24, a murine IgG3 antibody specific for GD3 with the terminal N-Ac disialyl residues (48), was used to assess the expression of surface GD3 on SK-MEL-28 cells incubated with or without precursors. The surface GD3 expression was significantly suppressed and replaced by modified GD3 when the cells were treated with precursors (Fig. 3).
Spectroscopic Confirmation of Modified GD3 Expression-The biochemical engineering of GD3Bu on the SK-MEL-28 cell surface in the presence of ManNBu was further confirmed by MS spectroscopic analysis. The glycolipids extracted from SK-MEL-28 cells were subjected to capillary electrophoresis, and the negatively charged glycolipids were analyzed by electrospray-MS. By using the fragment ion (m/z 290 for NeuAc and 318 for NeuBu) scan technique, GD3 was found to be a dominant ganglioside on the SK-MEL-28 cells with a molecular weight of 1556 based on the observation of a double negative charged ion at m/z 777 (Fig. 4a). The fragmentation derived from MS-MS analysis confirmed the GD3 structure and the ceramide composition (m/z 647), which were in agreement with previous reports (49,50). When ManNBu was added to the growth medium at the concentrations of 1, 3, and 5 mg/ml, GD3Bu was spontaneously expressed, which was detected in the MS analysis as a major component (m/z 805, a double negative charged ion); an increase of m/z 28 from 777 or a 56-Da difference in molecular mass represents the difference in molecular mass between GD3Bu and GD3Ac (Fig. 4b).
Unmodified GD3 and mono-N-Bu GD3 (m/z 777 or 791) were not detected in the glycolipid extraction under the above ManNBu concentrations by MS analysis (data not shown). Therefore, we assume that the modified GD3 molecules are highly expressed on the cell surface.
Production of Monoclonal Antibodies-Immunization of mice with glycoconjugated GD3Bu-KLH, fusion, and initial screening by ELISA against GD3Bu-BSA gave hybridomas 2A (IgG2a) and 1 (IgG1). After re-cloning by limiting dilution, each culture supernatant was tested by ELISA against the following glycoconjugates as antigens: GM3-BSA, GD3Ac-BSA, GD3Pr-BSA, GD3Bu-BSA, and GD3Bz-BSA. mAb 1 only reacted with GD3Bu-BSA, whereas mAb 2A bound to both GD3Pr-BSA and GD3Bu-BSA (Fig. 5c). Neither mAbs were reactive with SK-MEL-28 or Bz-SK-MEL-28 cells by flow cytometry, but both reacted with Bu-SK-MEL-28 cells and Pr-SK-MEL-28 cells (Fig. 5, a and b). To confirm further the epitope specificity of mAbs 2A and 1, ELISA inhibition was performed with GD3Bu-BSA glycoconjugate as coating antigen by using four disialolactosides as inhibitors (GD3Ac, GD3Pr, GD3Bu, and GD3Bz). Only GD3Bu disialolactoside was able to inhibit the binding of both mAbs to the antigen. This result indicates that N-butyryl group(s) are involved in the binding, but detailed structural parameters of the epitope have yet to be established. Because of its better ability to bind complement, mAb 2A (IgG2a) was selected for the further experiments.
In Vitro Anti-tumor Effects (CDC Mediated by Monoclonal Antibody 2A)-We investigated whether the mAb 2A and the anti-GD3Bu serum were capable of mediating CDC on SK-MEL-28 cells. Thus, CDC of mAb 2A and anti-GD3Bu serum on SK-MEL-28 cells with and without treatment of ManNBu (1 mg/ml) was examined in the presence of rabbit complement. mAb 2A killed only Bu-SK-MEL-28 cells that were treated with ManNBu in a dose-dependent manner in the range of 0.05-100 g/ml (Fig. 6a). Similar results were also obtained with the anti-GD3Bu serum at various dilutions (1: 400 -1:1600, Fig. 6b). Neither antibody showed any killing effect on SK-MEL-28 cells without precursor treatment.
Expression of Modified GD3 on SK-MEL-28 in Vivo-After administration of ManNBu to nude mice (CgFoxNude) for 2 weeks (5 days/week, 5 mg/mouse), SK-MEL-28 cells from tumors were examined by flow cytometry, and both GD3 and GD3Bu epitopes were detected by mAbs R24 and 2A (Fig. 7). Compared with the expression of modified GD3 in vitro (Fig. 2), the incorporation of GD3Bu epitope in vivo was less effective, but nevertheless it was still sufficient to provide a specific immunotarget.
Anti-tumor Effects in Vivo-After tumors grafted on nude mice had grown to about 10 mm in diameter, mAb 2A in combination with ManNBu was able to suppress tumor growth during the 2 weeks of treatment (average V/V 0 Ϸ1.0), but tumors started to grow rapidly after treatment was withdrawn (data not shown). In the control group, tumors in mice treated with mAb 2A alone grew faster (average V/V 0 Ϸ1.4). Flow cytometry analysis of cells from dissected tumors after the treatment of mAb2A/ManNBu showed reactivity only to R24 but not mAb 2A (data not shown). This observation was in direct contrast to the prolonged expression of modified GD3 observed in SK-MEL-28 cells in vitro. Thus this mAb/precursor combination had a limited effect in controlling tumor growth, but the antibodies were obviously not effective in removing solid tumors. Further experiments designed for early intervention were more successful. Without any treatment, all 10 nude mice after transplantation of SK-MEL-28 cells developed solid tumors within 30 days that grew to about 4 -7 mm in diameter after 45 days. Whereas subcutaneous injection, 3 days after tumor cell transplantation for 2 weeks (5 times/week), of either mAb 2A or ManNBu alone provided some protection (about 30% of mice did not develop solid tumor), the best results were obtained with the combination of both, delivered by either subcutaneous or intraperitoneal routes, that resulted in 90% of mice being free of tumors after 45 days (see Table IV). DISCUSSION Immunotherapy based on carbohydrate antigens for the treatment of cancers has been intensively examined. Although they are self-antigens and the antibodies raised against these antigens may react to normal tissues, such therapy is still a viable option. However, the drawbacks of self-antigen-based vaccines are obvious; poor immunogenicity and potential autoimmune responses are two major concerns (33). By biochemically engineering cell surface antigens, specifically the sialic acid-containing antigens, we are able to temporarily remodel the cell surface and render it susceptible to targeted antibody responses. Such antibodies can be elicited by using a glycoconjugate vaccine in which a specifically modified antigen is covalently linked to a protein carrier.
In this study, we first synthesized four disialolactoside-KLH conjugates as vaccines in order to raise antibodies against the GD3 and its analogs on the cell surface. Two N-Ac groups of disialolactoside (GD3 tetrasaccharide) were substituted by N-Pr, N-Bu, and N-Bz groups, respectively, in order to optimize their immunogenicities and specificities. In contrast to its human response, GD3 is immunogenic in mice. However, it has been demonstrated that replacement of N-Ac groups of PSA by N-Pr groups overcomes the immune tolerance in animals (51,52), indicating that a similar modification on GD3 may achieve similar effects in man. As expected GD3Pr-KLH and GD3Bu-KLH were good immunogens, and only a weak crossreactivity between GD3Pr sera and GD3Bu was observed. The fact that neither of these antisera was cross-reactive with native GD3 confirms our supposition that it is possible to raise antibodies by using structurally modified immunogens to avoid generating autoantibodies. Sub-typing analysis revealed that the majority of the antigen-specific antibodies was of the IgG rather than the IgM subclass, suggesting the effective recruitment of T-helper cells in the induction of the immune response.
In order to use these specific antibodies to immunotarget SK-MEL-28 cells, one has to assume that modified GD3 is expressed on the cell surface, and many previous studies have succeeded in modifying cell surface sialic acids by the introduction of chemically modified precursors in the biosynthetic pathway (53). Thus, SK-MEL-28 cells treated with exogenous precursors, such as ManNPr, ManNBu, and ManNBz, express the corresponding N-acyl-modified sialic acid residues on the cell surfaces. Specific modifications of GD3 on the cells were identified by strong binding to their homologously modified N-acyl GD3-conjugate antisera.
Two parameters were considered in the remodeling of cell surface antigens by using metabolic precursors: 1) the incorporation efficiency, and 2) the metabolic rate. The effective expression of modified GD3 occurred at 1 mg/ml of precursor, and the fact that increasing precursor concentration did not improve the expression suggests that N-acyl mannosamines in the N-acyl sialic acid biosynthesis can be transferred efficiently by the sialyltransferases to lactosyl ceramide, forming modified GD3. The low concentration of precursor required and the relatively fast expression of modified GD3 (less than 24 h) in comparison with previous results on the sialic acid engineering of PSA (32) may add therapeutic advantages in clinical settings. Moreover, in comparison to the cell surface PSA antigen, the turnover rate of cell surface GD3 was slow. GD3Bu expression was still detectable on most cells 5 days after removal of ManNBu in vitro, indicating that modified GD3 could be a favorable immunotarget. Although Bertozzi and co-workers (54,55) observed that ManNBu is able to block the biosynthesis of PSA of neural cell adhesion molecule, its impact on the biosynthesis of GD3 seems less significant. The identification of GD3Bu on Bu-SK-MEL-28 cell surface by flow cytometric and mass spectroscopic (electro-spray-MS) analyses suggests NeuNBu is a good substrate for ␣2-3siaT (SAT-I) and ␣2-8 siaT (SAT-II), both involved in the GD3 biosynthesis (56).
Although SK-MEL-28 cells expressed the modified GD3 epitope well in vitro, the expression of such an epitope on the tumor surface in vivo is a prerequisite for immunotargeting. Clearly, SK-MEL-28 cells grown in nude mice accepted Man-NBu as a metabolic precursor and incorporated it into GD3 molecules as evidenced by flow cytometric analysis on cells obtained from tumors grafted on the nude mice. But although FIG. 7. Expression of GD3Bu in vivo. Nude mice (CgFoxNude) were subcutaneously injected with 1 ϫ 10 7 SK-MEL-28 cells in the rear right flank. When the tumors were about 10 mm in width, mice were injected daily with ManNBu (5 mg/mouse, intraperitoneal) in RPMI medium for 2 weeks (5 days/week). Tumors were dissected, and tumor cells were analyzed by flow cytometry using mAbs R24 and 2A. The expressions of both GD3 and GD3Bu on cell surface were detected. the tumor cells did express GD3Bu, significant amounts of GD3 still remained on the cell surface. These results could be attributed to an insufficient concentration of precursor and/or that the biosynthesis in vivo favors using ManNAc as substrate. Nevertheless, the successful bioengineering of modified GD3 in vivo does provide a means to target SK-MEL-28 cells for immune destruction. Similar to the function of mAb KM871 reported previously (27), both mAb 2A and GD3Bu antisera were capable of stimulating the complement-mediated lysis of Bu-SK-MEL-28 cells that express GD3Bu, and such CDC activity depended on the mAb and antibody concentrations. Without modification, SK-MEL-28 cells were affected by neither mAb 2A nor GD3Bu antiserum, which further indicates that expression of GD3Bu may serve as a trigger (switch) for the immune response. This result is significant because we may be able to achieve augmentation of immunogenicity without risk of evoking the autoantibodies, which has always been a dilemma in the use of common cancer vaccines.
Unfortunately, there are other limitations to the in vivo application of this technology when it was shown that the treatment of mice having established tumors, with mAb 2A and ManNBu, could arrest further tumor growth but not eliminate the tumor. Although disappointing, our results are not surprising because Livingston and co-workers (9, 12) also obtained similar results with GD2 antibodies, whereby the homologous antibodies were only able to remove microtumors and prevent the metastasis but not remove solid tumors. However, when the above treatment was applied to mice soon after tumor grafting, it was able to prevent establishment of the tumor, providing evidence that an active vaccination strategy using the GD3Bu-KLH conjugate in conjunction with ManNBu might be applicable to the prevention of tumor metastasis. Unfortunately, the nude mouse model that we employed was not a good model to further investigate the potential of this strategy.
Examination of tumor tissues from large size tumors after the above treatment failed to identify GD3Bu-expressing tumor cells by flow cytometric analysis. This observation indicates the limitation of immunotargeting a single epitope, because it is possible that sub-populations of tumor cells without this epitope could overgrow those with it and thus render the treatment ineffective. Similarly, it is known that tumors can down-regulate specific antigens such as sialyl Lewis x (57), and it is of interest to note that the down-regulation of GD3 expression has also been implicated in the reduction of both cellular proliferation and metastasis of neuroblastoma F-11 cells in animals (58 -60).
In conclusion, we have demonstrated that synthetic disialolactoside conjugates can accurately imitate homologous epitopes expressed on the cell surface. The fact that no crossreactivity was observed between SK-MEL-28 cells and GD3Bu-KLH antiserum indicates that the immunodominant epitope should minimize the risk of unwanted autoimmune responses. We have also demonstrated that GD3Bu-specific antibodies became potent cytotoxic reagents against Bu-SK-MEL-28 cells when GD3Bu molecules were incorporated. Studies in vivo indicated that the treatment using a combination of precursor and specific antibodies can prevent mice from tumor grafting but was ineffective for eliminating solid tumors.