Single-domain antibody-based and linker-free bispecific antibodies targeting FcγRIII induce potent antitumor activity without recruiting regulatory T cells.

Antibody-dependent cell-mediated cytotoxicity, one of the most prominent modes of action of antitumor antibodies, suffers from important limitations due to the need for optimal interactions with Fcγ receptors. In this work, we report the design of a new bispecific antibody format, compact and linker-free, based on the use of llama single-domain antibodies that are capable of circumventing most of these limitations. This bispecific antibody format was created by fusing single-domain antibodies directed against the carcinoembryonic antigen and the activating FcγRIIIa receptor to human Cκ and CH1 immunoglobulin G1 domains, acting as a natural dimerization motif. In vitro and in vivo characterization of these Fab-like bispecific molecules revealed favorable features for further development as a therapeutic molecule. They are easy to produce in Escherichia coli, very stable, and elicit potent lysis of tumor cells by human natural killer cells at picomolar concentrations. Unlike conventional antibodies, they do not engage inhibitory FcγRIIb receptor, do not compete with serum immunoglobulins G for receptor binding, and their cytotoxic activity is independent of Fc glycosylation and FcγRIIIa polymorphism. As opposed to anti-CD3 bispecific antitumor antibodies, they do not engage regulatory T cells as these latter cells do not express FcγRIII. Studies in nonobese diabetic/severe combined immunodeficient gamma mice xenografted with carcinoembryonic antigen-positive tumor cells showed that Fab-like bispecific molecules in the presence of human peripheral blood mononuclear cells significantly slow down tumor growth. This new compact, linker-free bispecific antibody format offers a promising approach for optimizing antibody-based therapies.

Functional characterization of an anti-estradiol antibody by site-directed mutagenesis and molecular modelling: modulation of binding properties and prominent role of the V(L) domain in estradiol recognition.

The high-affinity monoclonal anti-estradiol antibody 9D3 presents a specificity defect towards estradiol-3-sulphate and 3-glucuronide conjugates incompatible with use in direct immunoassays. The corresponding single-chain variable fragment (scFv), cloned and produced in E. coli, exhibited a 10-fold lower affinity for estradiol (K(a)=1.2 x 10(9) M (-1)) and a slightly increased specificity defect for the 3-position. Site-directed mutagenesis revealed critical residues involved in estradiol recognition and produced mutants exhibiting up to a 3-fold increase of the binding affinity for estradiol and up to a 2-fold decrease of the cross-reactivity with estradiol-3-sulphate. A comparative model of the antibody 9D3-estradiol complex was built in which the estradiol D-ring is buried into the binding pocket while the 3-, 6- and 7-positions are solvent exposed, agreeing with the lack of specificity for these three positions. Two potential alternative orientations of the A-ring, one close to CDR H3 and L2 loops, and the other one close to CDR H2 and L3 loops, have been considered for the docking of estradiol, none of which could be unambiguously privileged taking into account data from cross-reactivity measurements, photolabelling and mutagenesis studies. For both orientations, estradiol is stabilized by hydrogen bonding of the 17beta-OH group with TyrL36, His89 and GlnH35 in the first case, or TyrL36, only, in the second case and by van der Waals contacts from TyrL91 with alpha- or beta-face of estradiol, respectively, and from ValH95 and GlyH97 with the opposite face. To elucidate the molecular basis of antibody 9D3 specificity, as compared with that of another anti-estradiol antibody 15H11, single variable domains (V(H) and V(L)) and scFv hybrids have been constructed. The binding activity of V(L)9D3 as well as the specificity of the V(L)9D3/V(H)15H11 hybrid, both similar to antibody 9D3, revealed a prominent role of V(L) in estradiol recognition. These findings establish premises for antibody engineering to reduce cross-reactivity, especially with estradiol-3-conjugates.