Selection of antibodies that regulate phenotype from intracellular combinatorial antibody libraries

Abstract

A method is presented that uses combinatorial antibody libraries to endow cells with new binding energy landscapes for the purpose of regulating their phenotypes. Antibodies that are expressed in cells infected with a lentiviral combinatorial antibody library are selected directly for function rather than only for binding. The potential diversity space can be very large because more than one lentivirus can infect a single cell. Thus, the initial combinatorial diversity of ~1.0 × 10(11) members generated by the random association of antibody heavy and light chains is greatly increased by the reassortment of the antibody Fv domains themselves inside cells. The power of the system is illustrated by its ability to select unusual antibodies. Here, the selected antibodies are potent erythropoietin agonists whose ontogeny depends on recombination at the protein level of pairs of antibodies expressed in the same cell to generate heterodimeric bispecific antibodies. The obligate synergy between the different binding specificities of the antibody's monomeric subunits appears to replicate the asymmetric binding mechanism of authentic erythropoietin.

Fig. 1.

Scheme for selection of antibody agonists from combinatorial libraries. Antibodies that bound EpoR were selected from a combinatorial antibody library displayed in phage by affinity-based selection. The antibody genes from the selected phage were cloned into lentiviral vectors to allow phenotypic selections after infection of eukaryotic cells and integration of the antibody genes into the genome. The transduced cells were plated in methylcellulose agar such that the secreted antibodies were trapped around the cells producing them. The colonies that formed were harvested using a micromanipulator, and the antibody genes were recovered by PCR. The PCR products were cloned and sequenced, and the respective antibodies or antibody combinations were tested for their activity. The active antibodies were expressed in mammalian cells and purified for further characterization.

Fig. 2.

Selection of agonist antibodies. (A) Western blot analysis showing that EF1α is a much stronger promoter for antibody synthesis than the UbC promoter. For all further experiments we used lentiviral vectors with EF1α promoters. The TF-1 cells cotransduced with hEpoR-T2A-GFP and antibody libraries were cultured in cytokine-free methylcellulose agar for 2–3 wk. (B) Colonies are shown where either bright-field microscopy to study the morphology and the color of the colonies or fluorescence microscopy to monitor the expression of the EpoR were used. TF-1 cells transduced with the EPO gene were used as positive controls. The colonies whose growth was independent of EPO were harvested with a micromanipulator. To produce antibodies, single antibody genes from the selected cell colonies were transfected into HEK293T cells.

Fig. 3.

Antibody synergy from transfections. (A) The ability of the individual antibodies to induce TF-1 cell proliferation was tested using the conditioned medium from HEK293T cells 48 h posttransfection of individual antibodies. The TF-1 cells were mixed with an equal volume of conditioned medium in microtiter wells and cultured in the absence of EPO for 72 h. The number of viable cells was determined by an MTS assay. Antibody E-3 was the strongest agonist, with ∼60% of the activity of authentic EPO. (B) To study synergy, all combinations of two antibody genes isolated from the same colonies were transfected into HEK293T cells. TF-1 cell proliferation was tested using conditioned medium obtained 48 h posttransfection. The TF-1 cells were mixed with an equal volume of conditioned medium and cultured without EPO for 72 h. The number of viable cells was determined by an MTS assay. Cotransfection with the V-1/V-2 genes gave the strongest activity that was equal to that of EPO. All unique antibody sequences and the combination of all possible antibody pairs from each colony (A–AG) were tested. Some antibodies with the same sequences were recovered from different colonies and are, thus, repeated in the graph. Approximately the same amount of antibody as determined by Western blot was used for each study.

Fig. 4.

Synergy in antibody protein constructs. The single antibodies V-1 and V-2 and the BsAb V-1/V-2 that were generated using knobs-into-hole technology were purified from HEK293F cells. (A) The purified proteins were analyzed using 7% Tris-acetate gels with Tris-acetate running buffer. (B) Antibody binding to the EpoR was tested by a pull-down experiment. Different antibodies were mixed with the extracellular domain of hEpoR His-tag. The complexes were captured with His-tag Dynabeads, and bound antibodies were detected with an anti-Fc:HRP antibody. (C) The dose–response curve for the abilities of the heterodimeric BsAb and the homodimeric antibody E-3, the most potent single antibody (Fig. 3A) to stimulate the proliferation of the EPO-dependent TF-1 human erythroleukemic cells. The maximum response of the BsAb is equal to EPO showing that it is a full agonist, whereas the strongest homodimeric antibody E-3 showed only 60% of the maximum response generated by authentic EPO. (D) To show that the purified EpoR ectodomain Fc chimera inhibits the antibody agonist activity, TF-1 cells were treated with antibodies or untreated (Ctrl) at a concentration of 10 nM in the presence of increased concentrations of the EpoR ectodomain Fc fusion protein.

Fig. 5.

Activation of the EpoR signaling pathway. (A) TF-1 cells were maintained in suboptimal concentration of GM-CSF (0.1 ng/mL) plus EPO or the BsAb at various concentrations for 1 wk. Cells were pelleted and lysates were analyzed for the production of hemoglobin by observation of the color of the pellets and Western blots that were probed with anti-hemoglobin antibodies. (B) To study the induced phosphorylation of JAK2 and Stat5, cytokine-depleted TF-1 cells were treated with 2 ng/mL GM-CSF, 4 IU/mL EPO, or the BsAb for 30 min at 37 °C. Unstimulated cells were used as a control. Half of the cell lysate was subjected to immunoprecipitation (IP) with anti-JAK2 antibody, followed by Western blot analysis (WB) with anti-phosphotyrosine antibody. After stripping the film, the total amount of JAK2 in the gel bands was detected with anti-JAK2 antibody. Western blot analysis of the other half of the cell lysates was carried out using antibodies against phospho-Stat5 (Tyr694) and Stat5 antibodies. The experiments were carried out on both TF-1 cells (Left) or engineered TF-1 cells complemented with WT hEpoR (Right). (C) The BsAb-induced erythroid differentiation of human stem cells. Human CD34+ hematopoietic stem cells freshly isolated from bone marrow were incubated for 14 d with the bsAB or EPO incorporated in soft agar. Typical red CFU-E colonies are shown.

Fig. 6.

Potential mechanisms of activation of EpoR by agonist antibodies. In the model, potent agonist activity is likely to be associated with asymmetric binding of the heterodimeric bispecific antibodies as for EPO (14, 15). Asymmetric binding reorients the D1 and D2 ectodomains for optimal signaling via the JAK2/Stat5 pathway. One homodimeric antibody is only a partial agonist, which suggests a more symmetric interaction with less activation as for a peptide agonist (16). Antibodies that bind EpoR, but do not activate, likely bind to only one arm of the EpoR unliganded dimer (–19) in an orientation that precludes bivalent association of the antibody.