Identification of a human anti-CD55 single-chain Fv by subtractive panning of a phage library using tumor and nontumor cell lines.

A large naive human single-chain (sc) Fv phage library was used to search for tumor-associated antigens by panning with a lung adenocarcinoma cell line, 1264, and counter-selecting with a nontumor bronchial epithelial cell line, BEAS-2B. After three rounds of subtractive panning, 239 of 673 clones analyzed bound selectively to 1264 tumor cells in a phage ELISA. Diversity analysis of these tumor-selective clones by BstNI fingerprinting and nucleotide sequencing revealed 14 distinct scFv fragments. Four clones bound selectively to 1264 over BEAS-2B cells when analyzed by a more discriminating flow cytometric assay using scFv. Moreover, these clones showed only limited cross-reactivity to several primary human cell lines. One clone, LU30, also cross-reacted strongly with the lung adenocarcinoma line, A549. The LU30 antigen was identified as decay-accelerating factor (CD55) by expression cloning from a 1264 cDNA library. The mean number of decay-accelerating factor molecules on the surface of 1264 and BEAS cells used for panning and counter-selection was estimated as 75,000 +/- 5,000 and 13,000 +/- 10,000, respectively. Thus, phage library panning combined with expression cloning permits identification of antibodies and their cognate antigens for proteins that are differentially expressed on the surface of distinct cell populations.

Stable heterodimers from remodeling the domain interface of a homodimer using a phage display library.

Structure-guided phage display was used to select for combinations of interface residues for antibody C(H)3 domains that promote the formation of stable heterodimers. A C(H)3 "knob" mutant was made by replacement of a small residue, threonine, with a larger one, tryptophan: T366W. A library of C(H)3 "hole" mutants was then created by randomizing residues 366, 368 and 407, which are in proximity to the knob on the partner C(H)3 domain. The C(H)3 knob mutant was fused to a peptide flag and the C(H)3 hole library was fused to M13 gene III. Phage displaying stable C(H)3 heterodimers were recovered by panning using an anti-flag antibody. Phage-selected C(H)3 heterodimers differed in sequence from the previously designed heterodimer T366W-Y407'A, and most clones tested were more stable to guanidine hydrochloride denaturation. The thermal stability of individual C(H)3 domains secreted from Escherichia coli was analyzed by differential scanning calorimetry. One heterodimer, T366W-T366'S:L368'A:Y407'V, had a t(m) of 69.4 degrees C, which is 4.0 deg.C higher than that for the designed heterodimer and 11.0 deg.C lower than that for the wild-type homodimer. The phage-selected C(H)3 mutant maintained the preference for forming heterodimers over homodimers as judged by near-quantitative formation of an antibody/immunoadhesin hybrid in a cotransfection assay. Phage optimization provides a complementary and more comprehensive strategy to rational design for engineering homodimers for heterodimerization.

'Knobs-into-holes' engineering of antibody CH3 domains for heavy chain heterodimerization.

'Knobs-into-holes' was originally proposed by Crick in 1952 as a model for the packing of amino acid side chains between adjacent alpha-helices. 'Knobs-into-holes' is demonstrated here as a novel and effective design strategy for engineering antibody heavy chain homodimers for heterodimerization. In this approach a 'knob' variant was first obtained by replacement of a small amino acid with a larger one in the CH3 domain of a CD4-IgG immunoadhesin: T366Y. The knob was designed to insert into a 'hole' in the CH3 domain of a humanized anti-CD3 antibody created by judicious replacement of a large residue with a smaller one: Y407T. The anti-CD3/CD4-IgG hybrid represents up to 92% of the protein A purified protein pool following co-expression of these two different heavy chains together with the anti-CD3 light chain. In contrast, only up to 57% of the anti-CD3/CD4-IgG hybrid is recovered following co-expression in which heavy chains contained wild-type CH3 domains. Thus knobs-into-holes engineering facilitates the construction of an antibody/immunoadehsin hybrid and likely other Fc-containing bifunctional therapeutics including bispecific immunoadhesins and bispecific antibodies.

Engineering linear F(ab')2 fragments for efficient production in Escherichia coli and enhanced antiproliferative activity.

We developed a novel bivalent antibody fragment, the linear (L-) F(ab')2, comprising tandem repeats of a heavy chain fragment VH-CH1-VH-CH1 cosecreted with a light chain. Functional humanized L-F(ab')2 directed against p185HER2 was secreted from Escherichia coli at high titer (> or = 100 mg/l) and purified to homogeneity. The L-F(ab')2 binds two equivalents of antigen with an apparent affinity (Kd = 0.46 nM) that is within 3-fold of the corresponding thioether-linked F(ab')2 fragment. The N-terminal site binds antigen with an affinity (Kd = 1.2 nM) that is approximately 4-fold greater than that for the C-terminal site, as shown by the comparison of L-F(ab')2 variants containing a single functional binding site. L-F(ab')2 has greater antiproliferative activity than the thioether-linked F(ab')2 against the p185HER2-overexpressing tumor cell line BT474. Linear and thioether-linked F(ab')2 have very similar pharmacokinetic properties in normal mice, and their serum permanence times are respectively 7- and 8-fold longer than the corresponding Fab fragment. L-F(ab')2 offers a facile route to bivalent antibody fragments that are potentially suitable for clinical applications, and that may have improved biological activity compared with thioether-linked F(ab')2 fragments.

Humanization of an anti-p185HER2 antibody for human cancer therapy.

The murine monoclonal antibody mumAb4D5, directed against human epidermal growth factor receptor 2 (p185HER2), specifically inhibits proliferation of human tumor cells overexpressing p185HER2. However, the efficacy of mumAb4D5 in human cancer therapy is likely to be limited by a human anti-mouse antibody response and lack of effector functions. A "humanized" antibody, humAb4D5-1, containing only the antigen binding loops from mumAb4D5 and human variable region framework residues plus IgG1 constant domains was constructed. Light- and heavy-chain variable regions were simultaneously humanized in one step by "gene conversion mutagenesis" using 311-mer and 361-mer preassembled oligonucleotides, respectively. The humAb4D5-1 variant does not block the proliferation of human breast carcinoma SK-BR-3 cells, which overexpress p185HER2, despite tight antigen binding (Kd = 25 nM). One of seven additional humanized variants designed by molecular modeling (humAb4D5-8) binds the p185HER2 antigen 250-fold and 3-fold more tightly than humAb4D5-1 and mumAb4D5, respectively. In addition, humAb4D5-8 has potency comparable to the murine antibody in blocking SK-BR-3 cell proliferation. Furthermore, humAb4D5-8 is much more efficient in supporting antibody-dependent cellular cytotoxicity against SK-BR-3 cells than mumAb4D5, but it does not efficiently kill WI-38 cells, which express p185HER2 at lower levels.