Antitumor antibodies with the same specificity as cytotoxic T lymphocytes that recognize antigenic peptides encoded by tumor-associated genes and presented by MHC class I molecules would be valuable tools to analyze the antigenicity or target tumor cellsin vivo. To obtain a human antibody directed against a peptide encoded by gene melanoma-associated antigen (MAGE)-A1 and presented by HLA-A1 molecules, we selected a large phage Fab antibody repertoire on a recombinant version of the complex HLA-A1–MAGE-A1 produced byin vitrorefolding. One of the selected phage antibodies shows binding to HLA-A1 complexed with the MAGE-A1 peptide, but does not show binding to HLA-A1 complexed with a peptide encoded by gene MAGE-A3 and differing from the MAGE-A1 peptide by only three residues. Phages carrying this recombinant antibody bind to HLA-A1+cells only afterin vitroloading with MAGE-A1 peptide. These results indicate that nonimmunized phage Fab libraries are a source of antibodies with a T cell antigen receptor-like specificity. The human anti-HLA-A1–MAGE-A1 antibody described here may prove very useful for monitoring the cell surface expression of these complexes, and eventually, as a targeting reagent for the specific immunotherapy of HLA-A1 patients bearing a MAGE-A1-positive tumor.
To generate antibodies to defined cell‐surface antigens, we used a large phage antibody fragment library to select on cell transfectants expressing one of three chosen receptors. First,in vitropanning procedures and phage antibody screening ELISAs were developed using whole live cells stably expressing the antigen of interest. When these methodologies were applied to Chinese hamster ovary (CHO) cells expressing one of the receptors for a neuropeptide, somatostatin, using either direct cell panning or a strategy of depletion or ligand‐directed elution, many different pan‐CHO‐cell binders were selected, but none was receptor specific. However, when using direct panning on CHO‐cells expressing the human membrane protein CD36, an extraordinary high frequency of antigen‐specific phage antibodies was found. Panning on myoblasts expressing the rat homologue of CD36 revealed a similar selection dominance for anti‐(CD36). Binding of all selected 20 different anti‐(CD36) phage was surprisingly inhibited by one anti‐(CD36) mAb CLB‐IVC7, which recognizes a functional epitope that is also immunodominantin vivo. Similar inhibition was found for seven anti‐(rat) CD36 that cross‐reacted with human CD36. Our results show that, although cells can be used as antigen carriers to select and screen phage antibodies, the nature of the antigen target has a profound effect on the outcome of the selection.
Many patients with colorectal carcinoma (CRC) mount a cellular as well as a humoral immune response to the tumor. To investigate the nature and specificity of the humoral immune response in a CRC patient, lymphocytes infiltrating the primary colorectal tumor and lymph nodes draining the tumor were used as antibody variable (V)‐gene pools for the construction of phage antibody repertoires. These libraries were first validated by selection on the antigen tetanus toxoid and shown to contain antibodies that were probably derived from both naive and memory B cells. The repertoires were then screened for the presence of antibodies directed to CRC by selection on the cell line CaCo2. For comparison, the same selections were performed with a phage antibody repertoire made from B cells of healthy donors. Striking differences were observed in the panel of specificities selected from these different repertoires: although a large panel of antibodies reactive with patient‐derived primary tumors was obtained from the immune repertoires, none of these discriminated between normal colonic epithelium and colon cancer and none were reactive with cell‐surface antigens. However, selections using the non‐immune library did result in numerous antibodies that recognized cell surface markers on CaCo2. These data suggest a bias in the local humoral immune response in this CRC patient, directed primarily toward intracellular epithelial‐cell specific target antigens.
The Ras proteins cycle in the cell between an inactive state and an active state. In the active state, Ras signals via the switch I region to effectors like c‐Raf kinase, leading to cell growth. Since Ras mutations in cancer are often associated with the presence of permanently active Ras, molecules that prevent downstream signaling may be of interest. Here, we show that by selection on the active conformation of Ras, using a recently described large phage antibody repertoire [de Haard et al. (1999) J. Biol. Chem. 274, 18218–18230], a Fab antibody (Fab H2) was identified that exclusively binds to active Ras, and not to inactive Ras. Using surface plasmon resonance (SPR) analysis, the interaction was demonstrated to be of high affinity (7.2 nM). In addition, the interaction with Ras is specific, since binding to the homologous Rap1A protein in BIAcore analysis is at least three orders of magnitude lower, and undetectable in an enzyme‐linked immunosorbent assay. The antibody fragment prevents the binding of active Ras to the immobilized Ras‐binding domain of c‐Raf kinase (Raf‐RBD) at an IC50 value of 135 nM. This value compares well to the KD of active Ras‐binding to immobilized Raf‐RBD using SPR, suggesting identical binding sites. Like the IgG Y13‐259, which does not demonstrate preferential binding to either inactive or active Ras, Fab H2 inhibits intrinsic GTPase activity of Ras in vitro. Mapping studies using SPR analysis demonstrate that the binding sites for the antibodies are non‐identical. This antibody could be used for dissecting functional differences between Ras effectors. Due to its specificity for active Ras, Fab H2 may well be more selective than previously used anti‐Ras antibodies, and thus could be used for gene therapy of cancer with intracellular antibodies.