More than 370-Fold Increase in Antibody Affinity to Estradiol-17β by Exploring Substitutions in the V_H-CDR3

Abstract

Antibodies that specifically target biomarkers are essential in clinical diagnosis. Genetic engineering has assisted in designing novel antibodies that offer greater antigen-binding affinities, thus providing more sensitive immunoassays. We have succeeded in generating a single-chain Fv fragment (scFv) targeted estradiol-17β (E₂) with more than 370-fold improved affinity, based on a strategy focusing the complementarity-determining region 3 in the V_H domain (V_H-CDR3). Systematic exploration of amino acid substitutions therein, using a clonal array profiling, revealed a cluster of four substitutions, containing H99P and a serial substitution E100eN–I100fA–L100gQ that lead to a 90-fold increase in E₂-binding affinity. This substitution quartet in the V_H-CDR3, combined with the substitution cluster I29V/L36M/S77G in the V_L domain, resulted in a scFv fragment with a further increase in the affinity (K_a, 3.2 × 10¹⁰ M⁻¹). This enabled a highly sensitive enzyme-linked immunosorbent assay capable of detecting up to 0.78 pg/assay. The current study has, thus, focused on the significance of reevaluating the potential of mutagenesis targeting the V_H-CDR3, and encouraging the production and use of engineered antibodies that enable enhanced sensitivities as next-generation diagnostic tools.

INTRODUCTION

Antibodies are widely used as essential tools in clinical diagnosis.¹⁾ More recently, “Antibody engineering”²⁾ has emerged as a revolutionary strategy for generating artificially-modified antibodies that have higher antigen-binding affinities than that of conventional native antibodies and which enable immunoassays with higher sensitivities.³⁾ Standard strategies²⁾ include introducing various mutations that lead to amino acid substitutions in the genes encoding heavy and light chain variable (V_H and V_L) domains of a parent antibody. The resulting libraries containing diverse mutated antibody fragments [typically, single-chain Fv fragments (scFvs)]^2,3) are then explored to discover rare species with desired characteristics with the aid of genotype–phenotype linking technology, e.g., phage display.^2–4)

Although, a variety of engineered antibodies have been designed, only few display an actual practical application in clinical diagnosis. In this study, we have performed an in vitro affinity-maturation of an antibody against estradiol-17β (E₂; an essential diagnostic marker) based on a systematic exploration of “decisive substitutions” focusing on narrow but crucial area, the complementarity-determining region (CDR) 3 of the V_H domain (V_H-CDR3).

We have summarized the background of the study in Fig. 1. Three-step random mutagenesis was previously performed based on error-prone-PCR, targeting the entire molecule of a prototype scFv#WT containing the V_H and V_L domains of a mouse anti-E₂ antibody Ab#E4-4 (full sequences of which were reported previously).^5–8) The final product scFv#M3rd with 11 substitutions displayed a 10¹⁰-range affinity constant (K_a) to E₂ that showed a 140-fold improvement compared with that of scFv#WT.⁸⁾ However, a slightly higher affinity was achieved with only four substitutions selected from the initial 11, as shown with scFv#4mut.⁸⁾ Among the four substitutions, L100gQ (see Supplementary Materials)⁹⁾ in the V_H-CDR3 contributed the most. This is indicated by the fact that the L100gQ alone achieved greater enhancement (17-fold higher K_a; shown with scFv#R5-1) than that after each of the other substitutions I29V, L36M, and S77G in the V_L domain. Moreover, even when these three substitutions were made simultaneously, the affinity improved by only 3.8-fold (shown with scFv#R3-1). However, the V_L substitution cluster in cooperation with the major substitution V_H-L100gQ achieved a 174-fold enhancement to generate the scFv#4mut.

Fig. 1. Summary of the in Vitro Affinity Maturation of Anti-E₂ Antibody That We Previously Reported^5–8)

In the scFvs, the V_H and V_L domains were linked with a common linker sequence (GGGGS)₃. Full sequence of the prototype scFv#WT was reported previously⁵⁾ Stars represent amino acid substitutions, details of which in scFv#M3rd are as follows: (V_H) K19R, Y56F, S84P, E85G, L100gQ (V_L) Q27R, I29V, L36M, H50N, S63G, S77G; scFv#4mut, (V_H) L100gQ (V_L) I29V, L36M, S77G. In this study, we used the amino acid numbering and CDR definition proposed by Kabat et al.⁹⁾ The orange and blue arrows indicate increases in the K_a, and the magnitudes are shown beside the arrows.

These results indicated that the L100gQ in the V_H-CDR3 may function as a decisive substitution that can be fine-tuned by other substitutions.⁸⁾ This may be owing to the fact that the common recognition for the V_H-CDR3 sequences often play a significant role in antigen-binding interactions because of their most diversified structures compared to other CDRs.^9–11) Interestingly, the V_H-L100gQ substitution in scFv#M3rd was introduced by chance, independently of our intent.^7,8) This prompted us to analyze other substitutions in the V_H-CDR3 that might have an even greater effect. We, therefore, undertook affinity maturation of scFv#WT anew based on the systematic exploration of the V_H-CDR3 (Fig. 2A).

Fig. 2. Affinity Maturation Based on Exploration of the Decisive Substitutions in the V_H-CDR3 as the Initial Step

The concept (A) and summary of exploration of the decisive single substitutions (B). The 13 mini-libraries (lib-s1–s13) were constructed by randomizing one of the 13 residues in the V_H-CDR3: highly conserved residues at the 101- and 102-positions were excluded from the targets. “X” means that any of the 20 proteinogenic amino acids can appear. Sensitivity-improving factor (SIF) was calculated as the ratio of the midpoints in the phage ELISAs obtained with phages displaying scFv#WT and a scFv mutant: [midpoint with scFv#WT (pg)/midpoint with a scFv mutant (pg)]. SIF values larger than one suggest increase in the E₂-binding affinity of the tested scFv mutant.

MATERIALS AND METHODS

Mutagenesis Targeting a Single Amino Acid

The region spanning the V_H of scFv#WT gene was PCR-amplified using a E2#4-VH5 primer (constant) and each of the 13-sets of primers E2#4-HCDR3-95–100 g (Supplementary Table S1) comprising a single NNS [i.e., (A/C/G/T)(A/C/G/T)(C/G)] degenerated codon (Supplementary Fig. S1). The 13 amplified fragments of the V_H region were each combined with the fragments of the V_L, similarly amplified using E2#4-VL5-1 (or -2) and E2#4-VL3 primers, via overlap-extension PCR.^5–7) The resulting 13 scFv genes were separately subcloned into pEXmide5 phagemid vector¹²⁾ and transformed into Escherichia coli TG1 cells to generate 13 mini-libraries (lib-s1–s13) (Fig. 2B). Transformant clones (approx. 282) were randomly selected from each library and subjected to the clonal array profiling (CAP) system^13,14) (Supplementary Fig. S2). Five recombinant clones were selected that generated scFv-displaying phages (scFv-Phs) with higher bindings in the phage enzyme-linked immunosorbent assay (ELISA) (described below). The recombinants that had “sensitivity-improving factor” (SIF; Fig. 2B) exceeding one were converted to soluble (i.e., non-phage-linked) scFvs, by inserting stop codons between the genes encoding scFv and the phage minor coat protein III in the recombinant plasmids, for determining the K_as at 4 °C via the Scatchard analysis¹⁵⁾ using a [³H]-labeled E₂.^5–8)

Mutagenesis Targeting Serial Triple Amino Acids

The sequence covering the V_H of the scFv#WT gene was amplified similarly, but using E2-H3mut-T11 primer (Supplementary Table S1) having a serial triple NNS codons.¹⁶⁾ The product was used for assembling scFv genes, which were introduced to TG1 cells. The resulting bacterial library was subjected to the CAP system and the top five transformant clones were evaluated as described above.

Phage ELISA^5–8)

The 96-well microplates (Costar#3590), coated with a conjugate of E₂ and bovine serum albumin (E₂–BSA) and blocked with BlockAce, were incubated at 37 °C for 60 min with a mixture of E₂ standard (50.0 µL) and a scFv-Ph (approx. 2 × 10⁹ colony forming unit, 100 µL), both diluted in PVG-PBS. The microwells were washed, probed with anti-M13 antibody labeled with peroxidase (POD), and bound POD activity was determined colorimetrically.

ELISA Determining E₂ with Soluble scFvs

(a) Equilibrium Method^5–8)

The aforementioned set of microplates were incubated at 4 °C for 240 min with a mixture of E₂ standard (50.0 µL) and soluble scFv (100 µL), both diluted in G-PBS. After washing, the microwells were probed with POD-labeled anti-FLAG M2 antibody and the captured POD activity was determined similarly.

(b) Sequential Saturation Method¹⁷⁾

An E₂ standard (60.0 µL) and a soluble scFv (120 µL), both diluted in G-PBS were mixed and incubated at 4 °C for 240 min in a small tube. An aliquot (150 µL) was transferred to a microwell of the aforementioned microplates and further incubated at 37 °C for 15 min. The microwells were washed and probed as described above.

RESULTS

Exploration of Decisive Single Substitutions

The prototype scFv#WT contained a V_H-CDR3 composed of 15 amino acids, seven of which (100a–g) are regarded as inserted residues^8,9) (Fig. 2B). We first developed decisive single substitutions by exploring 13 mini-libraries, each of which had a randomized amino acid residue at one of the 13 positions (i.e., 95–100 g) (Fig. 2B). These randomizations were performed by inserting the NNS codon providing any of the 20 proteinogenic amino acids. Top five scFv-Ph mutants, selected from each mini-library, were assessed for their E₂-binding in the phage ELISA with the aid of an anti-M13 antibody specific to M13-phage envelope. Unexpectedly, only two mini-libraries randomizing the 99 and 100 g positions afforded scFv-Phs with satisfactory SIF (Fig. 2B). These scFv-Phs were further evaluated after converting into the soluble scFvs to avoid aggregation between phage envelopes that might interfere with adequate characterization of scFvs. Consequently, we found the scFv with a L100gQ substitution (scFv#m1Q) that displayed a 17-fold higher affinity than that of scFv#WT, as the most improved mutant (Figs. 2B, 3). The mutant with the second-highest affinity scFv#m1P had H99P substitution and showed 6.9-fold higher affinity (Figs. 2B, 3). Remarkably, the L100gQ and H99P functioned cooperatively, as observed in scFv#m2 that had both the substitutions and displayed a 65-fold increase in affinity.

Fig. 3. Illustration of the Step-Wise Improvement of the Anti-E₂ scFvs in Terms of the E₂-Binding Affinity (K_a)

Red stars indicate the decisive substitutions initially found in the V_H-CDR3, and purple stars indicate the substitution clusters generated in the V_H-CDR3, by combining of the initial substitutions. Blue stars indicate the substitution cluster in the V_L domain that cooperatively functioned with the L100gQ substitution.⁸⁾ The one-letter codes denoted over or under the scFv sequences mean the introduced amino acids. The orange and blue arrows indicate increases in the K_a, and the magnitudes are shown beside the arrows. The K_a values (M⁻¹) of these scFvs are as follows: scFv#WT, 8.6 × 10⁷; scFv#m1P, 5.9 × 10⁸, scFv#m1Q, 1.5 × 10⁹, scFv#m3, 3.0 × 10⁹, scFv#m2, 5.6 × 10⁹, scFv#m4, 7.7 × 10⁹, scFv#4mut, 1.5 × 10¹⁰, scFv#m5, 2.0 × 10¹⁰, scFv#m6, 2.1 × 10¹⁰, scFv#m7, 3.2 × 10¹⁰.

Exploration of Decisive Serial Triple Substitutions

To investigate the potential of a substitution at the 100 g-position under cooperation with different substitutions in the V_H-CDR3, we explored a library generated by randomizing serial triple amino acids including the 100 g position (i.e., 100e–g residues) by the NNS codons,¹⁶⁾ which potentially contained 20³ scFv members. Comprehensive searching with the CAP system disclosed the serial substitutions E100eN–I100fA–L100gQ expressed in scFv#m3, which caused a 35-fold enhanced affinity (Fig. 3). Presumably, in this substitution triplet, the L100gQ functioned dominantly with the assistance of E100eN and I100fA that magnified the potential of L100gQ by 2.0-fold. Interestingly, the effect of this triplet was magnified by 2.6-fold by cooperation of the H99P, as shown with scFv#m4 (Fig. 3).

Synergistic Effect of the V_L Substitutions

The three substitutions in the V_L (I29V, L36M, and S77G)⁸⁾ (Fig. 1) were shown to act in a cooperative manner with the substitutions(s) currently reported in the V_H-CDR3: i.e., H99P/L100gQ, E100eN/I100fA/L100gQ, and H99P/E100eN/I100fA/L100gQ (contained in scFv#m2, #m3, and #m4, respectively). In fact, the combination with the “V_L-substitution cluster” further improved the affinity by 3.6-, 7.0-, and 4.2-fold in scFv#m5, #m6, and #m7, respectively (Fig. 3). In particular, scFv#m7 displayed the highest affinity (K_a = 3.2 × 10¹⁰ M⁻¹) with 372-fold increase in affinity relative to scFv#WT.

Performance of the Improved scFvs in ELISA

It should be noted that the improved scFv mutants (scFv#4mut, m5–7), as well as scFv#WT, migrated as single bands on sodium dodecyl sulfate-polyacrylamide electrophoresis (SDS-PAGE) with nearly the expected relative molecular masses (M_r)^18,19) (Supplementary Fig. S3). Analytical performances of the affinity-enhanced scFvs were evaluated in competitive ELISAs, in which the bound soluble scFvs (the FLAG tag was added at their C-terminus) were detected with the aid of anti-FLAG M2 antibody. In the equilibrium ELISA, scFv#m7 afforded a dose–response curve, which was approx. 6-fold more sensitive (by comparison of midpoints) than that of scFv#WT (Fig. 4). In the sequential saturation format,¹⁷⁾ often employed in automated immunoassay analyzers, further increased sensitivity was achieved (Fig. 4). Dose–response curves covered approximately 1.0–50 pg/assay with midpoint and limit of detection^7,8) of 4.46 and 0.78 pg/assay, respectively. Cross-reactivity with estrone, estriol, ethynylestradiol, progesterone, and cortisol were 0.53, 13, 3.8, 0.001, and <0.001%, respectively, which were satisfactorily low for clinical use.

Fig. 4. Dose–Response Curves for E₂ ELISAs with the Equilibrium Format Using scFv#WT (●) and scFv#m7 (▲), and with the Sequential Saturation Format Using scFv#m7 (▲)

The ratio of B₀ absorbance against the absorbance of non-specific binding was >30. The unit “X g/assay” was used in the abscissa, which refers to the total mass (X g) of analyte that was added to each microwell for the antigen–antibody reactions. The midpoints of the dose–response curves (pg/assay) are shown over the abscissa. The vertical bars indicate the standard deviation (n = 4).

DISCUSSION

The V_H-CDR3 sequences are known to play a crucial role in antigen recognition^10,11) and have been targeted for mutagenesis to study antigen–antibody interactions. In the context of antibody engineering, thorough randomization using the NNS or NNK(G/T) codons has often been performed targeting serial and multiple amino acid residues therein, generating vast libraries of mutants (often exceeding 10⁸ members)^8,20) The strategy has proven to be effective in generating antibodies with different specificities.²⁰⁾ Unfortunately, however, these approaches may not always be effective for improving the affinity while maintaining the original specificity.

We here devised a systematic approach composed of two stages: each deals with much smaller libraries (Fig. 2A). The first stage is systematic exploration of amino-acid substitutions in the V_H-CDR3 that decisively increase the affinity, and the second stage is for further enhancing the affinity of the first-stage products by adding extra substitutions randomly in the entire scFv sequence. It was unexpected that, in the first step, only 99 (proline) and 100 g (glutamine) positions provided an effective single substitution. By combining these substitutions together and with the separately found serial triple substitution, however, we observed four decisive substitution patterns in the V_H-CDR3 (i.e., L100gQ, H99P/L100gQ, E100eN/I100fA/L100gQ, and H99P/E100eN/I100fA/L100gQ) (Fig. 3). Furthermore, an additional three V_L substitutions generated three novel scFvs (scFv#m5, #m6, and #m7) displaying an exponential rise in antibody affinity.

In the second step of our strategy for “overwriting” fine-tunable substitutions in entire scFv molecules (Fig. 2A), the error-prone PCR would be available as a general method. Instead, we here tested cooperativity of the I29V/L36M/S77G substitutions in the V_L that we found previously (Fig. 1). As expected, this synergistically functioned and generated scFv#m5–m7 with further enhanced affinity. The scFv#m7 in particular displayed the highest K_a (3.2 × 10¹⁰ M⁻¹), 372-fold higher than that of the prototype scFv#WT. To the best of our knowledge, this improvement in antibody affinity is the greatest reported for mutagenesis targeting anti-steroid antibodies, so far.^3,8,13) Moreover, the K_a value is the highest among the anti-E₂ antibodies reported, as well.⁷⁾

The current study will encourage researchers to reevaluate the potential of mutagenesis targeting the V_H-CDR3, and in production and use of the engineered antibodies as next-generation diagnostic tools. We should mention that the current success is largely due to the potential of CAP system (Supplementary Fig. S2)^13,14) for discovering improved mutants. The scFv#m7 enabled highly sensitive ELISA with a promising specificity, clinical application of which is now under progress.

Acknowledgments

This work was supported by JSPS KAKENHI Grant No. 19K07021. We would like to thank Dr. Eskil Söderlind (Avena Partners AB, Sweden) and Professor Carl A. K. Borrebaeck (Lund University, Sweden) for providing the pEXmide 5 vector.

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

This article contains supplementary materials.

REFERENCES

• 1) The Immunoassay Handbook: Theory and applications of ligand binding, ELISA and related techniques, 4th Edition. (Wild D ed.) Elsevier, the Netherlands (2013).
• 2) Introduction to Antibody Engineering. (Rüker F, Wozniac-Knopp G eds.) Springer, Switzerland (2021).
• 3) Kobayashi N, Oyama H. Antibody engineering toward high-sensitivity high-throughput immunosensing of small molecules. Analyst, 136, 642–651 (2011).
• 4) Smith GP, Petrenko VA. Phage display. Chem. Rev., 97, 391–410 (1997).
• 5) Kobayashi N, Kato Y, Oyama H, Taga S, Niwa T, Sun P, Ohtoyo M, Goto J. Anti-estradiol-17β single-chain Fv fragments: generation, characterization, gene randomization, and optimized phage display. Steroids, 73, 1485–1499 (2008).
• 6) Kobayashi N, Oyama H, Kato Y, Goto J, Söderlind E, Borrebaeck CAK. Two-step in vitro antibody affinity maturation enables estradiol-17β assays with more than 10-fold higher sensitivity. Anal. Chem., 82, 1027–1038 (2010).
• 7) Oyama H, Yamaguchi S, Nakata S, Niwa T, Kobayashi N. “Breeding” diagnostic antibodies for higher assay performance: A 250-fold affinity-matured antibody mutant targeting a small biomarker. Anal. Chem., 85, 4930–4937 (2013).
• 8) Oyama H, Kiguchi Y, Morita I, Yamamoto C, Higashi Y, Taguchi M, Tagawa T, Enami Y, Takamine Y, Hasegawa H, Takeuchi A, Kobayashi N. Seeking high-priority mutations enabling successful antibody-breeding: systematic analysis of a mutant that gained over 100-fold enhanced affinity. Sci. Rep., 10, 4807 (2020).
• 9) Kabat EA, Wu TT, Perry HM, Gottesman KS, Foeller C. Sequences of Proteins of Immunological Interest; U.S. Department of Health and Human Services, National Institutes of Health, U.S. Government Printing Office, Washington, DC (1991).
• 10) Oliva B, Bates PA, Querol E, Avilés FX, Sternberg JE. Automated classification of antibody complementarity determining region 3 of the heavy chain (H3) loops into canonical forms and its application to protein structure prediction. J. Mol. Biol., 279, 1193–1210 (1998).
• 11) D’Angelo S, Ferrara F, Naranjo L, Erasmus MF, Hraber P, Bradbury ARM. Many routes to an antibody heavy-chain CDR3: necessary, yet insufficient, for specific binding. Front. Immunol., 9, 395 (2018).
• 12) Jirholt P, Ohlin M, Borrebaeck CAK, Söderlind E. Exploiting sequence space: shuffling in vivo formed complementarity determining regions into a master framework. Gene, 215, 471–476 (1998).
• 13) Kiguchi Y, Oyama H, Morita I, Morikawa M, Nakano A, Fujihara W, Inoue Y, Sasaki M, Saijo Y, Kanemoto Y, Murayama K, Baba Y, Takeuchi A, Kobayashi N. Clonal array profiling of scFv-displaying phages for high-throughput discovery of affinity-matured antibody mutants. Sci. Rep., 10, 14103 (2020).
• 14) Kiguchi Y, Oyama H, Morita I, Nagata Y, Umezawa N, Kobayashi N. The V_H framework region 1 as a target of efficient mutagenesis for generating a variety of affinity-matured scFv mutants. Sci. Rep., 11, 8201 (2021).
• 15) Scatchard G. The attractions of proteins for small molecules and ions. Ann. N. Y. Acad. Sci., 51, 660–672 (1949).
• 16) Siegel RW, Baugher W, Rahn T, Drengler S, Tyner J. Affinity maturation of tacrolimus antibody for improved immunoassay performance. Clin. Chem., 54, 1008–1017 (2008).
• 17) Zettner A, Duly PE. Principles of competitive binding assays (saturation analyses). II. Sequential saturation. Clin. Chem., 20, 5–14 (1974).
• 18) Morita I, Oyama H, Yasuo M, Matsuda K, Katagi K, Ito A, Tatsuda H, Tanaka H, Morimoto S, Kobayashi N. Antibody fragments for on-site testing of cannabinoids generated via in vitro affinity maturation. Biol. Pharm. Bull., 40, 174–181 (2017).
• 19) Oyama H, Morita I, Kiguchi Y, Morishita T, Fukushima S, Nishimori Y, Niwa T, Kobayashi N. A single-step “breeding” generated a diagnostic anti-cortisol antibody fragment with over 30-fold enhanced affinity. Biol. Pharm. Bull., 40, 2191–2198 (2017).
• 20) Barbas CF III, Bain JD, Hoekstra DM, Lerner RA. Semisynthetic combinatorial antibody libraries: a chemical solution to the diversity problem. Proc. Natl. Acad. Sci. U.S.A., 89, 4457–4461 (1992).