|Home | About | Journals | Submit | Contact Us | Français|
The carbohydrate antigen sialyl-Lewis A (sLea), also known as CA19.9, is widely expressed on epithelial tumors of the gastrointestinal tract and breast, and on small cell lung cancers. Since over-expression of sLea appears to be a key event in invasion and metastasis of many tumors and results in susceptibility to antibody mediated lysis, sLea is an attractive molecular target for tumor therapy.
We generated and characterized fully human monoclonal antibodies (mAbs) from blood lymphocytes from individuals immunized with a sLea –KLH vaccine.
Several mAbs were selected based on ELISA and FACS including two mAbs with high affinity for sLea (5B1 and 7E3, binding affinities 0.14 nM and 0.04 nM, respectively) and further characterized. Both antibodies were specific for Neu5Acα2-3Galβ1-3(Fucα1-4)GlcNAcβ as determined by glycan array analysis. Complement dependent cytotoxicity against DMS-79 cells was higher for r7E3 (IgM) compared to r5B1 (IgG1), (EC50 0.1 μg/ml vs 1.7 μg/ml). In addition, r5B1 antibodies showed high level ADCC activity on DMS-79 cells with human NK cells or peripheral blood mononuclear cells. To evaluate in vivo efficacy, the antibodies were tested in a xenograft model with Colo205 tumor cells engrafted into SCID mice. Treatment during the first 21 days with 4 doses r5B1 (100 μg/dose) doubled the median survival time to 207 days, and 3/5 animals survived with 6 doses.
Based on the potential of sLea as a target for immune attack and their affinity, specificity and effector functions, 5B1and 7E3 may have clinical utility.
Many tumor restricted monoclonal antibodies resulting from immunization of mice with human cancer cells were shown to be directed against carbohydrate antigens expressed at the cell surface as glycolipids or glycoproteins (1, 2), but carbohydrate chemistry has been rather challenging and the clinical development of such antibodies has been slow for various reasons.
Sialyl Lewis A (sLea), also known as CA 19.9, is widely expressed on tumors of the gastrointestinal tract and is used as a tumor marker in pancreatic and colon cancer (3, 4). sLea is a known ligand for endothelial leukocyte adhesion molecule and expression of sLea was found to have an impact on metastatic potential (5, 6) and to correlate with increased metastatic potential in human colon cancer (7, 8) and pancreatic adenocarcinoma (9). In a study of 43 breast cancer patients with infiltrating ductal carcinoma it was found in about 79% of samples (10) and there was higher expression in patients with greater node involvement. Interestingly, sLea is predominantly expressed in cancers, while the natural counterpart, disialyl Lewis A, is found in nonmalignant epithelial cells (4). sLea is expressed primarily as a glycolipid, and glycolipids are proven targets for immune attack against cancer cells (1-3). Thus, passive administration of antibodies could potentially eradicate free tumor cells and early metastases in a minimal disease setting and have a significant impact on cancer recurrence.
Here we report the discovery and initial characterization of two fully human antibodies that were generated from blood lymphocytes from individuals immunized with sLea-KLH vaccine (11). Two promising candidate antibodies with high affinity for sLea (5B1 and 7E3) were expressed as recombinant antibodies and further characterized in vitro. Both antibodies were very potent in complement dependent cytotoxicity assays, and furthermore, the 5B1 IgG1 antibody was also highly active in our antibody dependent cytotoxicity assays. To evaluate the in vivo efficacy, the 5B1 antibodies were tested in a xenograft model of Colo205 tumor cells engrafted into SCID mice. Treatment with 5B1 antibodies cured 40-60% of the mice depending on dose, while 5/5 untreated animals died within 155 days.
DMS79 (12), SW626, EL4, HT29, BxPC3, SK-MEL28, and P3X63Ag8.653 cell lines were purchased from ATCC (Manassas, VA). Colo205-luc cells (Bioware® ultra) were obtained from Caliper Life Science (Hopkinton, MA). The murine control mAb 121SLE (IgM) was purchased from GeneTex (Irvine, CA). Sialyl Lewis A tetrasaccharide (Cat # S2279) was purchased from Sigma-Aldrich (St. Louis, MO). sLea-HSA conjugate (Cat # 07-011), monovalent biotinylated-sLea (sLea-sp-biotin; Cat # 02-044), polyvalent biotinylated sLea-PAA (Cat # 01-044), biotin-labeled Lea-PAA (Cat # 01-035) and sLex-PAA-biotin (Cat # 01-045) were purchased from GlycoTech (Gaithersburg, MD). In the polyvalent presentation, the tetrasaccharide is incorporated into a polyacrylamide matrix (PAA) thereby creating a 30kd multivalent polymer with approximately every 5th amide group of the polymer chain N- substituted with biotin in a 4:1 ratio and approximately 20% carbohydrate content. Other HSA or BSA glycoconjugates used in this study were prepared in house using sLea pentenyl glycoside as described (11). GD3, fucosyl-GM1, GM2 and GM3 were purchased from Matreya (Pleasant Gap, PA) and GD2 was purchased from Advanced ImmunoChemical (Long Beach, CA).
Blood samples were obtained from 3 patients in an ongoing trial with sLea-KLH conjugate vaccine in patients with breast cancer initiated at MSKCC under an MSKCC and FDA approved IRB protocol and IND. Blood specimens were selected from 2 patients after 3 or 4 vaccinations, which showed antibody titers against sLea of 1/160 and 1/320, respectively. These sera (and murine mAb 19.9) react well with sLea positive cell lines in FACS assays and mediate, potent CDC (11). PBMCs were isolated from approximately 80-90 ml of blood by gradient centrifugation on Histopaque-1077 (Sigma-Aldrich).
PBMCs were cultured in RPMI-1640 medium (Mediatech, Manassas, VA) supplemented with L-Glutamine, non-essential amino acids, sodium pyruvate, vitamin, penicillin/streptomycin, 10%FBS (Omega Scientific, Tarzana, CA), 10 ng/ml IL-21 (Biosource, Camarillo, CA), and 1μg/ml anti-CD40 mAb (G28-5 hybridoma supernatant, ATCC). Cells were fused by electrofusion to P3X63Ag8.653 myeloma cells.
For the sLea ELISA, plates were coated either with 1 μg/ml of sLea-HSA conjugate, monovalent biotinylated-sLea, or with polyvalent biotinylated sLea-PAA captured on Neutravidin (Pierce, Rockford, IL) coated plates. Uncoated wells (PBS) and human serum albumin (HSA) coated wells were used as controls. Bound antibodies were initially detected with HRP-labeled goat anti-human IgA+G+M (Jackson ImmunoReseach, West Grove, PA) and positive wells were subsequently probed with IgG-Fc or IgM specific secondary antibodies to determine isotypes.
Cross-reactivity against the closely related antigens, Lea and sLex, was evaluated by Surface Plasmon Resonance (SPR) and confirmed by ELISA using biotin-labeled Lea-PAA and biotin-sLex-PAA. Binding to gangliosides GD2, GD3, fucosyl-GM1, GM2, and GM3 was tested by ELISA. A competition ELISA was used to evaluate the specificity of the mAbs against several other related carbohydrate moieties. In brief, 2 μg/ml sLea-HSA conjugate was coated onto plates followed by blocking with 3%BSA in PBS. Next, 30 μl of different carbohydrate moieties (40 μg/ml in PBS prepared from 1 mg/ml stock solutions) either unconjugated or conjugated to HSA or BSA was mixed each separately with 30 μl of test antibody and incubated at room temperature in a sample plate. After 30 minutes 50 μl of the mixture was transferred to the coated assay plate and incubated for 1 hr, followed by incubation with HRP-labeled goat anti-human IgA+G+M, washing and colorimetric detection of bound antibody using a Versamax spectrofluorometer (all steps are performed at room temperature). The tested carbohydrate moieties included globo H, Lewis Y, Lewis X, sialyl-Thomson-nouveaux (sTn), clustered sTn, Thomson Friedenreich (TF), Tighe Leb/LeY mucin, porcine submaxillary mucin (PSM), as well as sLea tetrasaccharide and sLea-HSA conjugate. To determine the fine specificity of the antibodies, glycan array analysis was performed by the Consortium for Functional Glycomics Core H group. 5B1 and 7E3 antibodies were tested at 10 μg/ml using version 4.1 of the printed array consisting of 465 glycans in replicates of 6.
Variable region of human mAb heavy and light chain cDNA was recovered by RT-PCR from the individual hybridoma cell line and subcloned into IgG1 or IgM heavy chain, or IgK or IgL light chain expression vector as described before (13). Ig heavy chain or light chain expression vector were double digested with Not I and Sal I, and then both fragments were ligated to form a dual gene expression vector. CHO cells in 6 well-plates were transfected with the dual gene expression vector using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). After 24 hrs, transfected cells were transferred to 10 cm dish with selection medium [D. MEM supplemented with 10% dialyzed FBS (Invitrogen), 50 μ M L-methionine sulphoximine (MSX), GS supplement (Sigma-Aldrich) and penicillin/streptomycin (Omega Scientific, Tarzana, CA)]. Two weeks later MSX resistant transfectants were isolated and expanded. High anti- sLea antibody producing clones were selected by measuring the antibody levels in supernatants in a sLea specific ELISA assay and expanded for large scale mAb production.
Antibodies were purified using the Äkta Explorer (GE Healthcare, Piscataway, NJ) system running Unicorn 5.0 software. In brief, stable clones of 5B1 or 7E3, respectively, were grown in serum-free culture medium in a Wave™ bioreactor and the harvested supernatant was clarified by centrifugation and filtration and stored refrigerated until used. Human IgG antibodies were purified on appropriate sized Protein A columns using 10 mM PBS, 150 mM NaCl running buffer. Human IgM antibodies were purified on a hydroxyapatite column, and IgM was eluted with a gradient of 500 mM phosphate. The antibody concentrations were determined by OD280 using an E1% of 1.4 and 1.18 for IgG and IgM, respectively, to calculate the concentration. The purity of each preparation was evaluated by SDS-PAGE analysis (1-5 μg/lane) under reducing conditions and the purity was estimated > 90% based on the sum of heavy and light chains.
sLea positive or negative tumor cell lines (0.5×106 cells per condition) were washed in PBS/2% FBS (PBSF). Test or control human mAb was then added (1-2μg/ml in complete medium) and incubated on ice for 30 minutes (14, 15). After washing in PBSF, the cells were incubated with Alexa-488 anti-human IgG (Fcγ) or anti-human IgM (μ) (Invitrogen) for 30 minutes on ice. Cells were washed twice in PBSF and analyzed by flow cytometry using the Guava Personal Cell Analysis-96 (PCA-96) System (Millipore, Billerica, MA). Colo205-luc cells were incubated with 2 μg/ml primary antibody followed by staining with secondary antibodies from SouthernBiotech (Birmingham, AL), and analyzed on a Becton Dickinson FACS Advantage IV instrument using FlowJo 7.2.4 software.
Affinity constants were determined using the principal of Surface Plasmon Resonance (SPR) with a BiaCore 3000 (GE Healthcare, Piscataway, NJ). Biotin-labeled univalent sLea (Cat # 02-044) or polyvalent sLea-PAA-biotin (Cat # 01-044) were coupled to separate flow cells of an SPA biosensor chip according to the manufacturer’s instructions. A flow cell blocked with HSA and culture medium containing free biotin was used as a reference cell. The binding kinetic parameters were determined from several known concentrations of antibody diluted in HBS-EP buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant P20) using the sLea-PAA-biotin coated flow cell. The curve-fitting software provided by the BiaCore instrument was used to generate estimates of the association and dissociation rates from which affinities are calculated.
sLea antigen positive and negative cell lines were used with a 90 minute cytotoxicity assay (Guava PCA-96 CellToxicity™ kit, Millipore, Billerica, MA, Cat # 4500-0200) using human complement (Quidel, San Diego, CA, Cat # A113) and purified human mAb at various dilutions (0.1 – 25 μg/ml) or with positive control mAbs as previously described (15-18). In brief, 2.5×106 target cells were painted with carboxyfluorescein diacetate succinymyl ester (CSFE) to yield green/yellow fluorescent target cells. The painted cells (1×105/50 μl sample) were incubated for 40 minutes with 100 μl of antibodies on ice. Next, 50 μl of human complement diluted 1:2 in complete medium (RPMI-1640, 10% FCS) or medium alone was added to triplicate samples and incubated for 90 minutes at 37°C. Thus, the final complement dilution in the assay was 1:8. Cells that were killed during this incubation time were labeled by adding the membrane impermeable dye 7-aminoactinomycin D (7-AAD) and samples were analyzed by dual color immunofluorescence utilizing the Guava CellToxicity™ software module. Control samples that receive NP40 were used to determine maximal killing and samples receiving complement alone served as baseline. The percentage of killed cells was determined by appropriate gating and calculated according to the following formula: % killed = (% sample – % complement alone) / (%NP40 - % complement alone)x100.
PBMC effector cells were isolated from blood samples obtained under an MSKCC IRB approved protocol by Ficoll-Hypaque density centrifugation. The target cells were incubated at 5×106 cells/ml in complete growth media with 15 μl of 0.1% calcein-AM solution (Sigma-Aldrich) for 30 minutes at 37°C, in the presence of 5% CO2. The cells were washed twice with 15 ml of PBS-0.02% EDTA and resuspended in 1ml complete growth medium. Fifty μl (10,000 cells) of labeled target cells were plated into a 96 well plate in the presence or absence of antibodies at the concentrations described in Figure 3, and incubated with 50 μl of freshly isolated peripheral blood mononuclear cells (effector cells, at 100:1 E/T ratio) accordingly. After 2 hours incubation, the plate was centrifuged at 300xg for 10 minutes and 75 μl of supernatant was transferred into a new flat bottomed 96 well plate. The fluorescence in the supernatant was measured at 485nm excitation and 535nm emission in Fluoroskan Ascent (Thermo Scientific). Spontaneous release was determined from target cells in RPMI-1640 medium with 30% FBS without effector cells and Maximum release was determined from target cells in RPMI-1640 medium with 30% FBS and 6% Triton X-100 without effector cells. Percent cytotoxicity was calculated as [counts in sample-spontaneous release]/[maximum counts-spontaneous release]x100.
Internalization of 5B1 antibody was evaluated by measuring the cytotoxic activity of r5B1 and Hum-ZAP secondary conjugate (Advanced Targeting Systems, San Diego, CA) complex against sLea expressing BxPC3 cells. BxPC3 cells were plated into a 96 well plate (2,000 cells/90μl/well) and incubated overnight in duplicates. Various concentrations of 5B1 antibody were incubated with Hum-ZAP secondary conjugates at RT according to the manufacturer’s instruction. Next, 10 μl/well of r5B1 and Hum-ZAP complex was added to the cells and incubated for 3 days. Twenty five μl of Thiazolyl Blue Tetrazolium Bromide (Sigma-Aldrich) solution (5mg/ml in PBS) was added to each well and incubated at 37°C. After a 2hr incubation 100μl/well of solubilization solution (20% SDS/50% N,N-Dimethylformamide) was added to each well and incubated for another 16 hrs at 37°C. The OD was measured at 570/690nm and values obtained with medium alone were used for plate background subtraction. Eight parallel cultures without antibody were used to normalize the sample values (Sample/Mean Untreated*100).
Female CB17 SCID mice (5-8 weeks old) were purchased from Taconic (Germantown, NY). Colo205-luc cells (0.5 × 106) in 0.1 ml complete growth media were injected via the tail vein on Day 0 using a BD insulin syringe with 28G needle (BD, Franklin Lakes, NJ). One hundred μg of mAb 5B1 was injected intraperitoneally on days 1, 7, 14 and 21 (experiment 1), or on days 1, 4, 7, 10, 14 and 21 (experiment 2). Mice were monitored for tumor development. All procedures were performed under a protocol approved by the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee. Kaplan-Meier survival curves were generated using GraphPad Prism 5.1 (GraphPad Software, San Diego, CA) and analyzed using the Mantel-Haenszel log-rank test.
Blood samples from three vaccinated patients were used for hybridoma generation efforts and many positive wells were detected in the antigen-specific ELISA assays. Extensive screening was used to eliminate antibodies that showed inferior or non-specific binding. Eight human antibody expressing hybridoma cells (one IgM and 7 IgG) with strong reactivity against sLea were initially selected, expanded and subcloned for further characterization. Two antibodies (9H1 and 9H3) showed strong binding to sLea-HSA conjugates, but not to sLea-PAA coated plates. Three antibodies (5B1, 5H11, 7E3) showed strong binding to monovalent- and polyvalent sLea, and sLea-HSA conjugates (Supplement Fig.1).
The heavy and light chain variable regions from four selected antibodies were recovered by RT-PCR and cloned into our full length IgG1 or IgM expression vectors. Molecular sequence analysis using IMGT/V-Quest (19) revealed that the three selected IgG antibodies 5B1 (IgG/λ), 9H3 (IgG/λ), and 5H11 (IgG/λ) were derived from the same VH family and all used lambda light chains. These IgG1 antibodies showed different CDR sequences with 16, 5 or 3 mutations deviating from germline, respectively (Supplement Table1). The IgM antibody (7E3) utilizes the kappa light chain and has 6 heavy chain mutations. The increased mutations in 5B1 are indicative of affinity maturation. Recombinant antibodies were produced in CHO cell lines in a Wave bioreactor system and purified using Protein A or Hydroxyapatite chromatography for IgG and IgM, respectively. The purified recombinant antibodies retained the properties of the original hybridoma derived antibodies with respect to ELISA binding and specificity.
Cell surface binding is crucial for cytotoxic activity and was therefore tested next. Flow cytometry showed strong binding of 5B1, 9H3, 5H11 and 7E3 recombinant antibodies to DMS-79 cells, a small cell lung cancer suspension cell line (Figure 1A). Binding of r5B1 and r7E3 was also confirmed on HT29 colon cancer cells (B), BxPC3 pancreatic cancer cells (C), SW626 ovarian cancer cells (D), and Colo205-luc colon cancer cells (F). These antibodies failed to bind to sLea negative (SLE121 negative) SK-MEL28 melanoma cells (Figure 1E) or EL4 mouse lymphoma cells (data not shown).
The relative affinity/avidity of the binding to sLea was probed by SPR using a streptavidin coated biosensor chip to capture biotinylated sLea-PPA. As shown in Table 1, r5B1 and r7E3 bind rapidly to sLea-PPA and show a significantly slower off-rate compared to 121SLE, a commercial available murine IgM anti-sLea antibody that was used for comparison. The affinity of 5B1 was measured at 0.14 nM and the apparent affinity/avidity of 7E3 was approximately 4 times lower (Table 1). Determination of 9H3 affinity was hampered since 9H3 antibodies (native and recombinant) failed to bind to the sLea-PAA coated biosensor chip.
Preliminary assays to probe carbohydrate specificity showed that. 5B1, 9H3 and 7E3 did not bind to the closely related sLeX, Lea, or LeY antigens or the gangliosides GD2, GD3, fucosyl-GM1, GM2, and GM3 as measured by ELISA or SPR. The binding of 5B1 to sLea-PAA was also inhibited by sLea-tetrasaccharide in a dose dependent manner in a BiaCore concentration analysis series (data not shown). These results are consistent with previous observations that sera with high anti-sLea antibody titer were found to be specific for sLea, i.e. not reactive with gangliosides GM2, GD2, GD3, fucosyl GM1 or the neutral glycolipids globo H and Ley by ELISA (11). In a competition assay with nine distinct related carbohydrate moieties in various presentations (e.g. as ceramide, or conjugated to BSA or HSA) only sLea-tetrasaccharide and sLea-HSA conjugate were able to inhibit binding to sLea-HSA conjugate (data not shown).
To examine the carbohydrate specificity in further detail, 5B1 and 7E3 antibodies were also tested by glycan array analysis performed by the Consortium for Functional Glycomics Core H group. Both antibodies were tested at 10 μg/ml on printed arrays consisting of 465 glycans in 6 replicates. The results confirmed the high specificity of both antibodies with selective recognition of the sLea tetrasaccharide, Neu5Acα2-3Galβ1-3(Fucα1-4)GlcNAcβ and virtual absence of binding to closely related antigens that were present in the array, including sLex, Lea, Lex, and Ley. The results are summarized in Supplement Table 2, and the raw data file is available at http://www.functionalglycomics.org/glycomics/HServlet?operation=view&sideMenu=no&psId=primscreen_3421.
To evaluate the functional activity of 5B1 and 7E3, we tested the cytotoxic activity with DMS-79 cells in the presence of human serum as a source of complement. Both antibodies showed in some assays close to 100% killing activity at 10 μg/ml while a control antibody with different specificity (1B7, anti-GD2 IgG1 mAb) had no effect at the same concentrations (data not shown). The CDC activity is concentration dependent and 7E3 was significantly more active than 5B1 in this assay (Figure 2), which is expected since IgM antibodies are known to be more effective in complement mediated cytotoxicity assays. The EC50 (50% cytotoxicity) was 1.7 μg/ml for 5B1 and 0.1 μg/ml for 7E3, which translates to roughly 85-fold higher potency for 7E3 on a molar basis.
While 7E3 is significantly more potent in the CDC assay, IgG antibodies are known to have ADCC activity, which is thought to be important for tumor killing in vivo. High levels of cytotoxicity were measured using 5B1 antibody with human PBMC and DMS79 target cells at various E:T ratios (Figure 3A). Similar levels of cytotoxicity were observed at lower E:T ratios with primary NK cells (data not shown). A dose response experiment with PBMC from 2 donors measured at an E/T ratio of 100:1 showed similar efficacy and >85% cytotoxicity was reached at concentrations ≥0.5 μg/ml of 5B1 (Figure 3B). The cytotoxicity mediated by 5B1 requires FcγRIII receptors since it can be blocked with 3G8 anti-CD16 antibodies (data not shown). The ADCC activity achieved with 1 μg/ml of 5B1 antibodies was superior to the activity observed with antibodies to GM2, Fucosyl-GM1, globo H, or polysialic acid. As expected, 7E3 and murine 121SLE (both are IgM) were inactive in this assay (data not shown).
Antibody conjugates directed at antigen “closely related to” Lewis Y were previously shown to be rapidly internalized and very effective in animal models (20, 21). To examine whether sLea is internalized, we incubated the pancreatic cell line, BxPC3 with 5B1, and then added Hum-ZAP, an anti-human-IgG conjugated to the ribosome-inactivation protein, saporin (22). Cells that internalize the saporin containing complex will die, while non-internalized saporin leaves the cells unharmed. As shown in Figure 4, BxPC3 cells were effectively killed in presence of increasing doses of 5B1, while the presence of an isotype matched IgG1 antibody directed against GD2, which is not expressed on these cells, does not kill the cells.
To evaluate the activity of 5B1 in vivo, the antibodies were tested in a xenograft model using Colo205-luc tumor cells in SCID mice. Five mice per group were injected with 0.5x106 cells into the tail vein on day zero, and successful injection of the cells was verified by imaging the animals using the IVIS 200 in vivo imaging system (Caliper Life Science). One day later, animals were treated with 5B1 antibodies given intraperitoneal or PBS mock injection. In experiment 1, 100 μg of 5B1 was given on days 1, 7, 14 and 21 (400 μg total dose), and in experiment 2 the animals received 100 μg 5B1 on days 1, 4, 7, 10, 14 and 21 (600 μg total dose). The average median survival of untreated animals was 102 days in the 2 experiments and all untreated animals died within 155 days (Figure 5). Treatment of animals improved survival significantly: the median survival was doubled to 207 days in the group that received 4 doses of 5B1 and 2/5 animals survived until termination of the experiment after 301 days (logrank test, p=0.0499; Hazard ratio = 3.46). The proportion of survivors further increased to 3/5 mice when 6 doses were administered (logrank test, p=0.0064; Hazard ratio = 6.375). The second study was terminated after 308 days and the surviving animals failed to reveal Colo205-luc tumors at the highest sensitivity of the imaging system (data not shown). These data demonstrate a significant survival benefit with 5B1 treatment. PK studies will be required to establish exposure levels and to further refine the dosing schedule.
Altered carbohydrate expression is a hallmark of tumor cells that could be an ideal target for active or passive immunotherapy. Unfortunately, the complexity of carbohydrate chemistry and biology has hampered the development of effective therapeutics targeting carbohydrate antigens. We have recently demonstrated that our sLea-KLH conjugate vaccine could induce high titer of both IgG and IgM antibodies against sLea in mice and humans without cross-reactivity to other similar blood group carbohydrate antigens (11).
To further characterize the humoral immune response to the vaccine, we generated human monoclonal antibodies from blood samples of vaccinated patients. A number of antibodies were recovered that specifically bind sLea in an ELISA assay and on the surface of tumor cell lines. Molecular cloning of some of the antibodies showed evidence of affinity maturation as indicated by the number of CDR mutations deviating from germline in the cDNA sequence. Two high affinity antibodies (5B1 and 7E3) were further characterized in vitro, and cell surface binding was demonstrated for CA19.9 positive colon cancer (HT29 and Colo205), ovarian cancer (SW626), small cell lung cancer (DMS-79), and pancreatic cancer (PxPC3) cell lines, but not to CA19.9 negative melanoma cell line (SK-MEL28). Both antibodies failed to bind to sialyl-Lewisx, Lewisa, and other related carbohydrates when tested by ELISA and Surface Plasmon Resonance. In addition, binding analysis on a glycan array with 465 distinct carbohydrates revealed that both antibodies have exquisite specificity for Neu5Acα2-3Galβ1-3(Fucα1-4) GlcNAcβ. The high specificity of 7E3 measured by glycan array analysis is remarkable since IgM antibodies tend to have lower affinities and show less specificity compared to affinity matured IgG antibodies. Both antibodies were very potent in complement dependent cytotoxicity assays albeit the 7E3 IgM antibody is considerably more potent in this assay, which was expected for a high affinity/avidity IgM antibody. The 5B1 IgG1 antibody has the added benefit of being highly active in ADCC assays, which is thought to be a major activity contributing to tumor regression. The results obtained so far show that both antibodies have significantly improved affinity/avidity and effector function profile compared to SLE121, a mouse monoclonal antibody with specificity for sLea.
To evaluate the in vivo efficacy, the 5B1 antibodies were tested in a xenograft model of Colo205 tumor cells engrafted into SCID mice. Treatment with 5B1 antibodies rendered 40-60% of the mice disease free for more than 300 days depending on dose, while 100% (5/5) untreated animals died within 155 days under these experimental conditions. Since all animals were injected with Colo205-luc cells that carry the luciferase gene, we could verify that all animals received similar amounts of tumor cells on day zero. The surviving animals failed to show tumors at the highest sensitivity of the imaging system at the end of the study. There could be several reasons why two animals died despite 5B1 treatment: Variation of the response could be due to natural variations in the animals’ pharmacokinetic (e.g. lower trough levels, faster elimination), variances in effector cell function, differences in complement activation and any combination of these. Another possibility could be selection of escape mutants in the host that lack or have reduced levels of sLea. If escape mutants were generated, one would expect that the percentage of surviving animals would remain similar, independent of r5B1 dose. We are poised to address this issue in ongoing experiments by evaluating higher doses and, if applicable, further examining residual tumors. Further animal studies are needed to determine pharmacokinetic parameters and also to evaluate and compare the efficacy of 7E3.
sLea is expressed as glycolipid as well as glycoprotein on leukosialin (CD43) and the latter plays a dominant role in adhesion to E-selectin, which is thought to be important for hematogenous metastasis (4, 23). A number of mouse antibodies against sLea have been discovered and some (e.g. CA19.9) are routinely used in diagnostic assays, while many others lack specificity and potency. Most of those antibodies were obtained following immunization with cancer cells and subsequent characterization revealed the antigen’s identity. In contrast, the antibodies described here are derived from humans following vaccination with a well defined carbohydrate (sLea) antigen. The results demonstrate that functional, high affinity antibodies with exquisite specificity and cytotoxic potency can be recovered, and the resulting recombinant antibodies are suitable for clinical testing without the need to initiate time consuming humanization efforts. We have not observed any toxicity in mice so far, but further studies will be required to evaluate human tissue cross-reactivity, potential cell aggregation in Lewis a positive individuals, and hemolytic potential.
Since sLea is widely and selectively expressed on human cancers, while disialyl-Lewis A and other related blood group antigens are predominantly expressed on normal cells, the antibodies described here might have clinical utility and further studies are warranted to explore their activity in various animal models. To our knowledge, there are currently no other fully human anti-sLea antibodies in development.
Carbohydrates expressed on the tumor cell surface are ideal targets for passive immunotherapy. This paper describes the discovery of potent fully human monoclonal antibodies established from blood samples of patients that were immunized with sialyl-Lewis A-KLH conjugate vaccine. The translational relevance is twofold: First, the approach serves to validate the antibody response that is elicited by the vaccine on the cellular level and ultimately this information could be used to improve the vaccine itself. Second, the most potent antibodies that are generated by only a few patients in a clinical trial can be preserved and could ultimately be developed as therapeutics for the target cancer population. The high affinity of mAb 5B1 and potent effector functions support this translational potential.
This work was supported by grant CA-128362 (to POL and WS) from the NCI, National Institutes of Health, and by MabVax Therapeutics. Glycan array analysis was performed by Core H of the Consortium for Functional Glycomics funded by the National Institute of General Sciences grant GM62116. We also thank David Smith (Core H, Emory University, Atlanta, GA) for helpful advice and assistance.
Potential Conflict of Interest: RS, SMS and WWS are full time employees of MabVax. GR and POL are paid consultants and shareholders of MabVax.