2025 Volume 48 Issue 11 Pages 1753-1761
“Antibody engineering” is a promising strategy for generating high-affinity antibodies required for developing sensitive immunoassays. Therein, the variable domains (VH and VL) of the parental antibody are genetically randomized and combined to produce diverse single-chain Fv fragment (scFv) molecules. Subsequently, high-affinity scFv mutants are selectively isolated. In the randomization process, mutations have conventionally been targeted to the complementarity-determining regions (CDRs) in the variable domains, which often interact directly with antigens. However, we previously discovered that, pinpoint insertion of only a single amino acid (leucine, asparagine, aspartic acid, proline, glutamine, arginine, or histidine) between positions 6 and 7 in the framework region 1 (FR1) of the VH, which is unlikely to interact with antigens, enhanced the affinity of an anti-cortisol scFv (original Ka, 3.6 × 108 M−1) up to 17–61-fold. These findings prompted us to conduct a comprehensive study of this affinity-enhancement phenomenon involving the remaining amino acids. Thus, we generated the necessary 13 scFv mutants and compared their Ka values. Remarkably, all mutants showed enhanced affinities, similar to those of the previous 7 mutants. Among the 20 mutants, the leucine-inserted scFv showed the largest Ka (2.2 × 1010 M−1) and consequently enabled a 75-fold more sensitive enzyme-linked immunosorbent assay (midpoint, 9.86 pg/assay) compared to the assay using the parental scFv (midpoint, 744 pg/assay). In silico modeling suggested that, regardless of the amino acid inserted, elongated FR1 can alter the conformation of the CDR3 in VH to facilitate a favorable interaction with cortisol.
Antibodies that capture target antigens with high affinity and specificity are required for developing diagnostic methods as well as therapeutic agents.1–3) In diagnostic applications, antibodies function as analytical reagents in immunoassays that allow the quantification of trace amounts of various biomarkers. However, the conventional hybridoma method4) does not always generate high-affinity antibodies suitable for practical use.5) Therefore, we have studied “antibody engineering,” i.e., in vitro genetic manipulation of antibody molecules as a novel strategy that potentially generates artificial antibodies with higher antigen-binding performances compared with the conventional native ones.6–13) In typical procedures, a native antibody selected as the prototype is first converted into a single-chain Fv fragment (scFv) by genetically linking its heavy- and light-chain variable domains (VH and VL, respectively). The resulting wild-type scFv (wt-scFv) is then randomized to construct a library containing various mutant scFvs.14,15) Finally, the rarely generated mutants with enhanced affinities are selected and isolated from the library.
To increase the chances of acquiring the desired mutants, we should construct vast libraries composed of diverse scFv mutants. Conventionally, random point mutations were introduced into entire scFv sequences through the error-prone PCR16) or related methods.17,18) Alternatively, amino acids in reasonably selected regions of scFv molecules were altered or randomized by introducing missense or degenerate codons using synthetic oligonucleotides.19,20) The latter strategy, called site-directed mutagenesis, is considered more efficient, particularly when multiple residues are targeted in the complementarity-determining regions (CDRs), which usually play important roles in antigen binding. However, this strategy was not always successful in improving the affinity for the original antigens, although it often generated useful scFv species showing unexpected specificity for different antigens.21,22)
In our previous studies exploring a library derived from a cortisol-specific scFv (equilibrium affinity constant Ka, 3.6 × 108 M−1), we isolated mutants with distinctive structures that showed 17–61-fold improved affinities against cortisol compared with the wt-scFv.23–26) These mutants were generated through the insertion of 1–6 consecutive amino acid(s) between positions 6 and 7 of the VH domain, which are present in the framework region 1 (VH-FR1): VH-FR1 is a partial structure located at the N-terminus of the VH domain27) (Fig. 1A). However, FRs have been considered to be unlikely to interact with antigens because they form a β-sheet sandwich that supports CDR loops.19,28–30) Therefore, this region was not mutagenized with the aim of affinity maturation.
(A) Structure of the wt-scFv used as the prototype for mutagenesis and its Ka. The amino acid sequence of positions 1–10 in the VH domain is indicated along with the site of insertion. The 13 mutants with a single amino acid insertion in VH-FR1 newly prepared in this study are shown in light purple. (B) Previously generated scFv mutants with VH-FR1 insertions are classified according to the number of inserted residues. The ranges of their Ka values are shown with the inserted amino acid sequences, which are listed in descending order of Ka.
We expected that the VH-FR1 is a newly discovered “hot region” whose mutagenesis generates desirably affinity-matured mutants with high probability. The mutants referred to above contained 7 species with the insertion of a single amino acid, i.e., leucine (L), asparagine (N), aspartic acid (D), proline (P), glutamine (Q), arginine (R), and histidine (H).23,24,26) Among them, the L-inserted scFv showed the highest affinity with a Ka of 2.2 × 1010 M−1 (Fig. 1B). A set of the 20 mutants with a single insertion is a suitable system for investigating whether the affinity-enhancing phenomenon depends on the inserted amino acid species. Therefore, we generated 13 scFv mutants, each having the insertion of one of the remaining 13 proteinogenic amino acids, and compared their affinity as well as their utility in an enzyme-linked immunosorbent assay (ELISA). A possible mechanism for the affinity enhancement due to the insertion was also proposed based on in silico modeling of scFv–cortisol complexes.
PB: 0.050 M sodium phosphate buffer (pH 7.3); PBS: PB containing 9.0 g/L NaCl; G-PBS: PBS containing 1.0 g/L gelatin; and T-PBS: PBS containing 0.050% (v/v) Tween 20.6–13,23–26)
Preparation of 13 scFv Mutants with a Single Amino Acid InsertionPCR was performed with the anti-cortisol wt-scFv gene (25 ng) as the template in a buffer solution (100 µL) containing the reverse (5′-ATTGTTATTACTCGCGGCCCAACCGGCCATGGCCCAGGTCCAACTGCAGCAGNNSCCTGGGGCTGAACTTGTGAAGC-3′) and forward (5′-GCCATTTGGGAATTAGAGCCA-3′) primers (50 pmol each), KOD Fx DNA polymerase (TOYOBO, Osaka, Japan) (5 U), and 40 nmol of each dNTP with the following cycling conditions: 94°C (2 min), then 35 cycles of 98°C (10 s), 55°C (30 s), and 68°C (1 min), followed by a hold step at 68°C (5 min). The amplified scFv genes were ligated into the pEXmide 7′ vector,11) and a portion of the resulting plasmid (approximately 0.4 µg) was used for transformation of XL1-Blue cells (Agilent Technologies; Hachioji, Tokyo, Japan) by the heat-shock method. Thus, a suspension (100 µL) of the competent cells mixed with the plasmid was cooled on ice for 30 min, heated at 42°C for 45 s, and then immediately cooled on ice for 2 min. The mixture was suspended in SOC medium (1.0 mL) and incubated at 37°C for 60 min under continuous shaking (200 rpm). An aliquot of the resulting cell suspension (100 µL) was spread onto 90ϕ plates containing 2 × YT agar supplemented with 100 µg/mL ampicillin, 10 µg/mL tetracycline, and 1.0% glucose. After overnight incubation at 37°C, each colony grown on the plate was cultured in 2 × YT liquid medium, from which the plasmid was extracted. The nucleotide sequence of the scFv gene therein was determined, and the relevant amino acid sequence was inferred. The transformants were grown in 2 × YT medium until the OD600 nm reached approximately 0.8 and induced with 100 µM isopropyl β-d-thiogalactopyranoside and 400 mM sucrose. Periplasmic extracts containing the scFv proteins were obtained through treatment with Tris–HCl buffer (pH 8.0) containing 20% sucrose and 1 mM ethylenediaminetetraacetic acid.31)
Western Blot and Scatchard Analyses of scFv MutantsPeriplasmic extracts were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and subsequently transferred onto a polyvinylidene difluoride membrane. The membrane was blocked overnight in PBS containing 2% skim milk and then incubated with peroxidase (POD)-labeled anti-FLAG M2 antibody (Sigma-Aldrich, St. Louis, MO, U.S.A.) at 37°C for 30 min. After washing, 1-Step™ Ultra TMB-Blotting Solution (Thermo Fisher Scientific, Waltham, MA, U.S.A.), diluted 10-fold with water, was added to the membrane and incubated at room temperature for 20 min. The Ka values of scFv mutants were determined through the Scatchard analysis.32) Reaction mixtures containing [1,2,6,7-3H]-cortisol (3.19 TBq/mmol; PerkinElmer Inc., Waltham, MA, U.S.A.) (approximately 300 Bq), various concentrations of standard cortisol, and a fixed amount of scFv protein (i.e., adjusted to bind approximately 50% of the labeled cortisol when no standard cortisol was present) were incubated in G-PBS at 4°C for approximately 16 h. The bound and free fractions were separated using the dextran-coated charcoal method, and the radioactivity of the bound fraction was quantified.23–26)
Functional Characterization of scFv Mutants in ELISAELISAs were carried out using 96-well microplates (#3590; Corning, Corning, NY, U.S.A.) coated with a cortisol–bovine serum albumin (BSA) conjugate (0.5 µg/mL).11) Standard cortisol or related steroids (50.0 µL/well) and the scFv proteins (100 µL/well), each dissolved in G-PBS, were added to the wells and incubated at 4°C for 120 min. The plates were washed 3 times with T-PBS, and the bound scFvs were detected with POD-labeled anti-FLAG M2 antibody diluted in G-PBS (0.20 µg/mL; 100 µL/well). The plates were further incubated at 37°C for 30 min, washed again, and the retained POD activity was measured at 490 nm with o-phenylenediamine as a chromogenic substrate.33) ELISA dose–response curves were generated using ImageJ software (NIH, Bethesda, MD, U.S.A.), which was used to fit the curves and calculate assay parameters. The midpoint values were derived from a 4-parameter logistic equation [log(analyte dose) vs. B/B0 (%)] as the IC50 value. The abscissa unit “X g/assay” indicates the total mass (X g) of the analyte or cross-reactive analogs added to the assay chamber (microwells) for the competitive antigen–antibody reactions.23–26)
Construction of in Silico Modeling of scFvs Docked with CortisolThe scFv protein 3D structures were constructed using the AlphaFold server,34) and their conformations when docked to cortisol were predicted using AutoDock Vina in the SwissDock server.35–38) Structural visualization and analysis of the models were performed using the PyMOL Molecular Graphics System, Version 3.0 (Schrödinger, New York, NY, U.S.A.). The most reasonable structure was selected from among several outputs, considering that the carbonyl group at position 3, used for coupling with BSA, should not be buried in the paratope because antibodies raised against hapten–BSA conjugates hardly recognize partial structures on the hapten molecule near the BSA-coupling site due to a steric hindrance with bulky BSA molecules. Conversely, the hydroxy groups at positions C17 and C21 were estimated to interact with the paratope based on low cross-reactivities with corticosterone and 17α-hydroxyprogesterone.
To investigate the potential of the VH-FR1 as a novel region for affinity maturation, we prepared a complete set of 20 scFv mutants, each with an insertion of a proteinogenic amino acid at a fixed site (i.e., between positions 6 and 7). Among the 20 mutants, 7 species with additional L, N, D, P, Q, R, and H residues were generated in our previous studies. Thus, the remaining 13 species were newly prepared from an scFv gene mini-library (approximately 7.3 × 103 bacterial transformants), which was generated by site-directed introduction of an NNS-degenerated codon, [i.e., (A/C/G/T)(A/C/G/T)(C/G) codon], in the wt-scFv gene. The nucleotide sequencing of the scFv genes in 64 randomly selected transformant clones allowed us to identify the 13 mutants necessary for this study. Then, the 20 scFv proteins were prepared via the periplasmic expression of these genes in each transformant. Periplasmic extracts containing the 13 scFv mutants were subjected to western blot analysis. We confirmed that these scFvs showed single bands corresponding to the expected molecular weight (approximately 28 kDa) (Fig. 2A).
(A) Western blot analysis of periplasmic extracts including scFvs. Lanes 1 and 9, Mr marker; lanes 2–8 and 10–15, V-, T-, F-, C-, E-, K-, M-, and S-, G-, W-, A-, Y-, and I-inserted scFv mutants, respectively. (B) Scatchard analysis to determine the affinity of the scFvs against cortisol. Scatchard plots for the V-, F-, and K-inserted scFv mutants are shown. Vertical bars indicate standard deviation (n = 4). The Ka values are listed from the largest to the smallest. The values marked with an asterisk (*) have been reported previously.23,24,26)
The equilibrium binding affinities of the 20 scFv mutants for cortisol were evaluated based on Scatchard analysis with a tritium-labeled tracer. Although the newly generated 13 scFvs were ranked below the L-inserted scFv, which was previously determined as the highest-affinity mutant, all of them exhibited significantly enhanced affinity (Fig. 2B). Consequently, the Ka values of the 20 mutants ranged from 0.62 to 2.2 × 1010 M−1, exceeding the wt-scFv by 17–61-fold. Following the L-inserted scFv, mutants with insertion of similarly hydrophobic [valine (V) and phenylalanine (F)] and amidated (N) amino acids showed sufficiently improved affinity with the Ka of 1.6–2.0 × 1010 M−1. However, the insertion of isoleucine (I; the isomer of L) and Q (the homolog of N) was obviously less effective (Ka < 1.0 × 1010 M−1), whereas insertion of acidic, basic, and alcoholic amino acids, i.e., glutamic acid (E), lysine (K), and threonine (T) resulted in comparably enhanced affinities (Ka, 1.4–1.7 × 1010 M−1). Thus, we did not find a clear explanation for the present affinity enhancements based on the physicochemical properties of the inserted amino acid species, such as polarity, charge, or bulkiness. Amino acid residues at positions 6 and 7 are part of the unordered structure between the 2 β-strands (at positions 3–5 and 9–12) in the VH-FR1, and are approximately 26.5 Å away from the cortisol molecule (Fig. 3A). Thus, it is unlikely that the affinity enhancement was caused by a specific interaction between the functional groups of the inserted amino acids and the antigen or amino acids in the scFv paratope. Instead, it may be attributable to conformational changes around the paratope that are triggered by the extension of the peptide backbone in the VH-FR1, as mentioned below based on in silico modeling.
(A) Overall view of the wt-scFv docked with cortisol, highlighting the VH-FR1 (orange) and VH-CDR3 (magenta) along with the entire backbone structure of the scFv (purple). The side chains of VH2, QH6, PH7, RH94, and RH97 are shown as stick models. Cortisol molecules are shown in light blue. The yellow dashed line represents the distance between the middle of the QH6 and PH7 residues and the B-ring of cortisol (approximately 26.5 Å). (B) Superimposed overall structures of the L-inserted scFv (green) and wt-scFv (purple). The inserted LH6a residue is shown in red along with its side chain structure. (C) Enlarged view of the superimposed structures of the L-inserted and wt-scFv, focusing on the region around VH2, RH94, RH97, and cortisol. (D) For wt-scFv (left) and L-inserted scFv (right), distances between carbon atoms of the RH97 side chain and those of cortisol (positions C3–C6), as well as the distances between hydrogen atoms (HNα and Hε in the α-amino and imino groups, respectively) and oxygen atoms of cortisol (at the positions C3 and C17), are shown. The N–H–O bond angles are indicated in brown. (E) Superimposed structures of V-, N-, T-, F-, and K-inserted scFvs (V-wt, N-wt, T-wt, F-wt, and K-wt, respectively; green) on wt-scFv (purple). Alterations of the VH-CDR3 loops induced by the single amino acid insertion are shown with yellow arrows.
To elucidate the structural basis underlying the affinity enhancements, we performed in silico modeling of the scFv–cortisol complex, followed by molecular docking simulations. Comparative analysis between the wt-scFv and the L-inserted scFv revealed notable structural alterations in the VH-CDR3 (i.e., the positions 95–102)9) (Figs. 1A and 3A–C), which might be critical for enhancing antigen binding. Thus, in the L-inserted scFv, the side chain of the V residue at position 2 in the VH domain (VH2), located upstream of the inserted residue, moved upwards approximately 0.7 Å. This pushed the amino group of the RH94 residue adjacent to the VH-CDR3 loop downwards approximately 0.7 Å, thereby altering the angle of this loop. Consequently, the α-carbon of the RH97 residue, positioned in the middle of the VH-CDR3, shifted approximately 1.5 Å closer to cortisol. As a result, the amide hydrogen in the peptide bond formed with the α-amino group of the RH97 residue approached the oxygen atom of the C17 hydroxy group of cortisol (4.0 Å → 3.6 Å), changing the angle between them favorably for hydrogen bond formation (97.1° → 120.2°), although the imino hydrogen of RH97 moved away from the C3 carbonyl group in cortisol (Fig. 3D). In addition, the α- and β-carbon of the RH97 side chain moved closer toward the C6 and C5 carbons of cortisol (4.2 → 4.0 and 4.6 → 4.3 Å, respectively) (Fig. 3D), which might have resulted in enhanced hydrophobic attractions between the scFv paratope and cortisol.
Similar structural transformations were observed in other mutants with V, N, T, F, and K insertions, irrespective of the chemical nature of their side chains. This suggests a common mechanism in which a single amino acid insertion into VH-FR1 alters the spatial orientation of VH-CDR3 (Fig. 3E) via the VH2–RH94–RH97 interaction cascade, which is mainly caused by steric adjustments without polar interactions. No other structural changes were observed in these scFv molecules. Particularly, the conformation of the VL domain remained almost constant before and after insertion (Fig. 3B).
Functional Evaluation of the scFv Mutants in Competitive ELISAsCortisol is used as an index of the function of the hypothalamic–pituitary–adrenal axis; thus, practical anti-cortisol antibodies are in great demand.39) To evaluate the diagnostic utility of the scFv mutants, ELISAs were performed using cortisol–BSA immobilized microplates. To allow accurate comparison of assay sensitivities, the midpoint values in dose–response curves were newly determined for all 20 mutants (including the 7 previously obtained species) along with the wt-scFv (Fig. 4). All mutants exhibited 22–75-fold smaller midpoints (9.86–34.0 pg/assay) than the wt-scFv (744 pg/assay), showing significant improvements in the assay sensitivity (Fig. 4). The L-inserted scFv, which showed the largest Ka, reasonably generated the most sensitive dose–response curve with the midpoint of 9.86 pg/assay. The limit of detection (LOD) was 2.45 pg/assay, which was defined as the amount of cortisol yielding a signal 3 standard deviations below the mean signal obtained without cortisol (n = 10). This LOD value corresponds to a cortisol concentration of 0.49 ng/mL, assuming direct application of 10-fold diluted clinical specimens. Given the typical reference range for serum cortisol40) and salivary cortisol41) (10–250 and 0.7–11 ng/mL, respectively), the ELISA using this mutant is sufficiently sensitive for diagnostic applications. Other high-affinity mutants, such as the N- and F-inserted scFvs, also exhibited similarly improved sensitivity (midpoint values were 16.7 and 13.5 pg/assay, respectively) (Fig. 4). However, some discrepancies were observed between the Ka values and the sensitivities. For instance, the V-inserted scFv, despite having the second-largest Ka, was ranked 17th in terms of ELISA sensitivity (midpoint was 23.6 pg/assay). This might be attributable to the bridging phenomenon,42) namely, the V-inserted scFv might have shown stronger affinity against the cortisol–BSA conjugate coated on the ELISA microplates than for unmodified cortisol because of the recognition of the bridge structure linking cortisol and BSA.
Vertical bars indicate standard deviation (n = 4). The midpoints of the dose–response curves are listed.
Specificity of 6 selected scFv mutants (L-, V-, N-, T-, F-, and K-inserted), chosen based on their Ka and/or ELISA sensitivity, was assessed by determining their cross-reactivity with 11 cortisol analogs (Fig. 5). Compared to the wt-scFv, all mutants tested here showed markedly reduced cross-reactivity with cortisone and 11-deoxycortisol, whereas slightly increased cross-reactivity was observed with 21-hydroxyprogesterone. Among these, the L-inserted mutant, the species with the highest affinity, showed the most suitable recognition pattern, exemplified by the lowest reactivity with 17α-hydroxyprogesterone, prednisolone, and prednisone. Although 11-deoxycortisol showed 103% cross-reactivity, this analog circulates in the blood at much lower concentrations than cortisol (serum/plasma reference values for adults are <10 ng/mL),40) and thus does not cause serious overestimation in the usual diagnostic use.
In the field of antibody engineering, a few studies have reported affinity enhancement through amino acid insertions into CDRs.43,44) For instance, Lamminmäki et al. successfully improved both the affinity and specificity of an anti-estradiol Fab fragment by the insertion of 2–4 residues into VH-CDR2.43) In contrast, FR-targeted insertions have rarely been studied so far. Recently, Skamaki et al. reported a rare case in which an scFv against interleukin-13 exhibited a 14-fold increase in Ka after a single amino acid insertion into VL-FR3.45) However, VL-FR3 is conformationally located in close proximity to the paratope (thus sometimes referred to as CDR4), and the inserted residue was estimated to interact directly with the antigen. It should be noted that this insertion into the FR occurred accidentally, contrary to the original purpose of their research.
In contrast, we focused on VH-FR1 to identify a novel “hot region” for the affinity maturation, although it is considerably apart from the paratope (Fig. 3A) and thus has long been neglected or overlooked. This project began with our previous findings, in which both insertional and substitutional mutations targeted to an N-terminal portion of VH-FR1 in an anti-cortisol scFv significantly enhanced the antigen-binding affinity. In the present study, we compared the efficacy of 20 proteinogenic amino acids for enhancing the affinity when singly inserted at a fixed site (between positions 6 and 7) in VH-FR1. Unexpectedly, all the 20 amino acids yielded substantially improved affinity regardless of the physicochemical nature of the inserted residue, as shown by the Kas, which increased more than 17-fold (Fig. 2B). To the best of our knowledge, such robust affinity improvement through a pinpoint site, but with a wide range of modifications, has not been reported to date. The in silico modeling suggested a general mechanism for these affinity enhancements: the inserted residue might trigger a cascade of conformational changes that finally result in migration of the RH97 residue to enable interactions with cortisol (Fig. 3). A comparison with a new mutant in which RH97 is replaced with a small and neutral residue (e.g., alanine)46) or X-ray crystallography of the cortisol-complexed scFv mutants might allow for a more decisive understanding of the mechanism.
Considering that conventional hybridoma methods rarely generated anti-cortisol antibodies with Ka greater than 1 × 1010 M−1,9,23,24) the present study is remarkable because it provides an efficient strategy for generating multiple antibodies with even higher affinities. Owing to such enhanced affinities, these scFv mutants improved ELISA sensitivity by 22–75-fold to enable the direct measurements of serum/plasma and salivary cortisol levels, the latter of which has recently been required for stress monitoring (Fig. 4). In particular, the L-inserted scFv, which showed the highest sensitivity among the 20 mutants, is promising because of its reduced cross-reactivity with cortisone (Fig. 5), a major cortisol analog abundantly present in saliva. The validation of the assay values for clinical specimens is currently ongoing in our laboratory.
In the future, we plan to conduct systematic studies on the effect of inserting 2 or more residues. The universality of this insertion-based strategy for various antibodies needs to be investigated. We expect that these studies will expand the methods for generating practically engineered antibodies, which will contribute to the progress of clinical and analytical chemistry based on antigen–antibody reactions.
This work was supported by JSPS KAKENHI (Grant Number: JP24K18265). We thank Dr. Eskil Söderlind and Professor Carl A. K. Borrebaeck (Lund University, Sweden) for permitting us to use the pEXmide 7′ vectors.
The authors declare no conflict of interest.
