The VH framework region 1 as a target of efficient mutagenesis for generating a variety of affinity-matured scFv mutants

Abstract

In vitro affinity-maturation potentially generates antibody fragments with enhanced antigen-binding affinities that allow for developing more sensitive diagnostic systems and more effective therapeutic agents. Site-directed mutagenesis targeting "hot regions," i.e., amino acid substitutions therein frequently increase the affinities, is desirable for straightforward discovery of valuable mutants. We here report two "designed" site-directed mutagenesis (A and B) targeted the N-terminal 1-10 positions of the V_H framework region 1 that successfully improved an anti-cortisol single-chain Fv fragment (K_a, 3.6 × 10⁸ M^-1). Mutagenesis A substituted the amino acids at the position 1-3, 5-7, 9 and 10 with a limited set of substitutions to generate only 1,536 different members, while mutagenesis B inserted 1-6 random residues between the positions 6 and 7. Screening the resulting bacterial libraries as scFv-phage clones with a clonal array profiling system provided 21 genetically unique scFv mutants showing 17-31-fold increased affinity with > 10⁹ M^-1 K_a values. Among the mutants selected from the library A and B, scFv mA#18 (with five-residue substitutions) and mB_1-3#130 (with a single residue insertion) showed the greatest K_a value, 1.1 × 10¹⁰ M^-1.

Figure 1

Backgrounds of the designed mutagenesis targeting the V_H-FR1 performed in this study. (a) Selected data of our previous CAP-based antibody-breeding experiments with scFvs against cortisol. A schematic illustration of the primary structures and K_a values of wt-scFv and improved scFvs having only a single amino acid substitution or insertion in the V_H-FR1. In wt-scFv, the V_H and V_L domains were combined via a linker sequence VSS(GGGGS)₃T. (b) The most common amino acids found at the positions 1–30 in V_H (i.e., the V_H-FR1) in different subgroups as defined by Kabat et al.. The frequency of appearance of each residue is shown with different colors: magenta, invariant (> 95%) and common in all subgroups; green, “subgroup-specific residues” that are invariant within the relevant subgroup(s). The V_H-FR1 amino acid sequence of wt-scFv is also shown for comparison. (c) The frequency of amino acids at the positions 1–10, 21, 23, 28, and 29 in V_H, compiled for 1,820 antibodies that were available in the Kabat database^,. The detailed data listing for the positions 1–30 is available in Supplementary Table S1.

Figure 2

Workflow of the present study. (a) Design of the libraries A (left) and B (right), based on the V_H-FR1-targeted site-directed randomization and insertion, respectively. The library A was composed of scFv sequences, whose codons for amino acids at the positions 1–3, 5–7, 9, and 10 were degenerated to encode 2–6 kinds of predefined residues as indicated. The library B involved the scFv sequences, in which extra 1–6 amino acid residues were inserted between the positions 6 and 7 using the (NNS)_n degenerated codons (n = 1–6): this was divided into three sublibraries slB-1–3, slB-4/5, and slB-6. (b) Screening of the libraries with CAP or CAP/ORD system. E. coli TG1 cells were transformed with one of the scFv libraries (A, slB-1–3, slB-4/5, or slB-6) to generate bacterial libraries with similar transformant numbers (0.38–5.1 × 10⁶ cfu), which were each grown on agar plates. Among the resulting colonies, 4,700 were randomly selected and subjected to CAP using 50 microplates: for each plate, 94 microwells were used for inoculating colonies and 2 for background without colonies but with KM13 helper phage. Top 40 scFv-phage clones that showed higher RLU (> 100,000 RLU) were analyzed for binding to cortisol with competitive ELISA. For slB-1–3, top 188 clones were also subjected to ORD. Then, 40 (after CAP) or 24 (after CAP/ORD) scFv-phages that afforded higher sensitivity (as the midpoint) than wt-scFv were converted to the soluble-form scFvs for examining the affinities. The pie charts show the distribution of luminescent signal (RLU) detected for each single microwell in CAP screening. The background RLU (mean ± SD; n = 100) of each library was varied between microplates as follows: 2758 ± 1909 (A), 3329 ± 3211 (slB-1–3), 507 ± 928 (slB-4/5), 679 ± 2500 (slB-6). (c) Summary of CAP system. Bacterial clones are individually cultured in microwells containing the helper phage. Therein, each clone propagates and produces scFv-phages without competing with different clones. Antigen-specific scFv-phages bind the pre-immobilized antigen and are detected with a bioluminescence assay using an in-house-prepared fusion protein combining anti-M13-phage scFv and Gaussia luciferase.

Figure 3

Structures and affinities of scFvs. (a) Amino acid sequences (the positions 1–10 of the V_H) of the improved scFvs obtained from the library A with the K_a (M⁻¹) values. (b) Amino acid insertions of the improved scFvs obtained from library B with the K_a (M⁻¹) values. No other substitution was found on the rest part of scFv sequences for all mutants. For comparison, the sequence and K_a (M⁻¹) value of wt-scFv were shown at the top of this figure. The K_a values of wt-scFv and the mutants with > 10¹⁰ M⁻¹ were determined in triplicate, and mean ± SD was shown. The K_a of wt-scFv was determined anew here for strict comparison, which differed slightly from the previous value (reported to be 3.8 × 10⁸ M⁻¹)^,. Amino acid sequences were deduced from the nucleotide sequences and the numbering was based on the definition of Kabat et al.. We have partly examined the slB-1–3 previously, and have found the scFv clones mB_1–3#4, 17, 138, 163, and 172, which were reported as preliminary results (with the name of scFv#em1-4, -17, em2-138, -163, and -172, respectively). (c) SDS-PAGE analysis (Coomassie brilliant blue staining) of the selected mutant scFvs, which were affinity-purified with an anti-FLAG M2 agarose (Sigma–Aldrich)^–: lane 1, M_r marker; 2, scFv mA#5; 3, scFv mA#18; 4, scFv mB_1–3#130; and 5, scFv mB_4/5#1.

Figure 4

Protein ribbon structures of wt-scFv, mutant scFvs mA#18 and mB_1-3#130 were constructed using the SWISS-MODEL Protein Modelling Server, and their conformations when docked to cortisol were predicted using SwissDock. The V_H-CDR1 (yellow), V_H-CDR2 (orange), V_H-CDR3 (magenta), V_L-CDR1 (dark blue), V_L-CDR2 (light green), and V_L-CDR3 (light blue) are shown with β-sheet structures (bold gray arrows). The amino acid residues that are close to cortisol (< 4 Å), and the substituted and inserted amino acids in mA#18 and mB_1-3#130, respectively (indicated in red), are shown with wireframe and specified with one letter code.

Figure 5

Typical dose–response curves for cortisol in competitive ELISAs using wt-scFv and the improved scFvs. The vertical bars indicate the SD for intra-assay variance (n = 4). The midpoint values (pg/assay) are listed together. In these assays, the scFv concentrations were adjusted to give bound enzyme activities at B₀ (the reaction without cortisol standard) of approximately 1.0–1.5 absorbance after a 30-min enzyme reaction. The background absorbance (observed without addition of scFvs) was lower than 5.0% of the B₀ absorbance.