Journal of Molecular Biology
Evolution of an Interloop Disulfide Bond in High-Affinity Antibody Mimics Based on Fibronectin Type III Domain and Selected by Yeast Surface Display: Molecular Convergence with Single-Domain Camelid and Shark Antibodies
Introduction
The 10th human fibronectin type III domain (10Fn3) is one of the protein scaffolds used in recent years to design and select, in vitro, families of proteins that bind with high affinity and specificity to a variety of macromolecular targets.1., 2., 3. Such protein scaffolds, also known as antibody mimics, are designed to play many of the roles of antibodies and their fragments in research, diagnostics, and therapy; in addition, they often have the advantages of small size, inexpensive microbial production, and extreme stability. Wild-type 10Fn3, for example, is only 10 kDa in size, is expressed at a high level in Escherichia coli, and has a melting temperature of 82 °C.4 Despite its lack of sequence homology with antibody domains, 10Fn3 has an immunoglobulin-like fold;5,6 its solvent-accessible loops BC, DE, and FG are reminiscent of complementarity-determining regions (CDR) 1, 2, and 3, respectively. Wild-type 10Fn3 contains no disulfide bonds, which allows it to preserve its stability under reducing conditions.
Variants of 10Fn3 that bind protein targets with high affinity and specificity have been selected using phage display,7., 8., 9. yeast two-hybrid,10 and mRNA display.4., 11. In addition to binding their targets in solution and when immobilized on a solid surface,11 10Fn3-based antibody mimics have been shown to work as Western-blot detection reagents9 and affinity-purification reagents,12 to block the interaction between a ligand and its natural receptor,4 and to recognize specific conformations of their target.10
Dissociation constants of target-binding 10Fn3 variants selected to date have varied from picomolar to micromolar. For example, 10Fn3 variants with 20 pM affinity for TNF-α11 and with 340 pM affinity for vascular endothelial growth factor receptor 24 were selected by mRNA display from libraries of 1012 unique sequences, which had been constructed by randomizing 21 residues in three loops (7 randomized residues in loop BC, 4 randomized residues in loop DE, and 10 randomized residues in loop FG). In contrast, 10Fn3 variants with 1.3 μM and 250 nM affinity for the Src SH3 domain were selected by phage display from a library of 2 × 109 unique sequences, which had been constructed by randomizing 10 residues in two loops (5 randomized residues in loop BC and 5 randomized residues in loop FG, which also contained a 3-residue deletion). Whereas it appears likely that the 500-fold lower complexity of the phage-display library precluded the sampling of highest affinity 10Fn3 variants,9 an alternative explanation is that the randomization of two loops did not create a large enough target-binding surface for high-affinity binding. In this study, we set out to determine whether 10Fn3 variants with sub-nanomolar affinity could be selected from libraries with only one or two randomized loops. This question is salient because a larger number of mutations from wild-type 10Fn3 is associated with lowered stability and solubility,4 as well as, presumably, a higher risk of immunogenicity. Stability, solubility, and immunogenicity of 10Fn3-based antibody mimics will greatly affect their suitability for therapeutic applications.
We constructed 10Fn3-based libraries of 107–108 unique, full-length sequences by randomizing 7 residues in loop BC (library BC7), 7 residues in loop FG (library FG7), or the same 14 residues in both loops BC and FG (libraries 2L14 and BF14, which differ only in the complexity of the library). We then used yeast surface display13 to select 10Fn3 variants that bind hen egg white lysozyme, an antigen used extensively in structural studies of antibody–antigen binding.14., 15., 16., 17. As expected, we found that antibody mimics selected from libraries with two randomized loops (libraries BF14 and 2L14) had higher affinity for lysozyme than those selected from libraries with a single randomized loop (libraries BC7 and FG7) and that antibody mimics selected from a library with a higher complexity (library BF14) had higher affinity for lysozyme than those selected from a library of the same design with a lower complexity (library 2L14). A striking feature of the highest affinity antibody mimics selected from a naive library is a pair of cysteines in structurally adjacent positions on the two loops, which appear to form a disulfide bond reminiscent of those found in shark17,18 and camel16,19 antibody variable domains. The antibody mimic with the lowest Kd, 350 pM, was obtained by an affinity maturation method similar to CDR walking originally reported for antibodies.20
Section snippets
Library construction
We constructed four different libraries based on the wild-type human 10Fn3 sequence by replacing codons in DNA stretches that encode 7 residues in each of two CDR-like loops, BC and FG, with mixtures of codons that encode all 20 amino acids. Library BC7 was constructed by randomizing the 7 residues (23–29) in loop BC; library FG7 was constructed by randomizing 7 residues (77–83) in loop FG; and libraries 2L14 and BF14 were constructed by simultaneously randomizing the 14 residues in loops BC
Discussion
The comparative success of the four naive, 10Fn3-based libraries in the selection of lysozyme-binding antibody mimics provides circumstantial evidence of the relative importance of thorough sequence-space sampling versus paratope size. Compared to the library made by randomizing 7 FG residues, the libraries made by randomizing 14 residues sample a much lower fraction of their theoretical sequence space: 2 × 10−11 and 2 × 10−10 for libraries 2L14 and BF14, respectively, in contrast with 2 × 10−2 for
Library and vector construction
Oligonucleotides used in the construction of 10Fn3-based libraries are listed in Table 4. Those oligos that contained randomized regions were synthesized by TriLink (San Diego, CA) from mixtures of triphosphoramidites designed to equalize the probability of incorporating a codon for each of the 20 amino acid residues (Glen Research, Sterling, VA). The remaining oligonucleotides were purchased from ThermoElectron (Ulm, Germany) or MWG Biotech (High Point, NC). The strategy used to construct the
Acknowledgements
This project was funded by the DuPont-MIT Alliance (DL), CA96504, CA101830, a National Science Foundation Graduate Fellowship (SML), and a National Defense Science and Engineering Graduate Fellowship (BJH). We thank the MIT Flow Cytometry Core Facility for technical support and Mike Schmidt for permission to include his unpublished results in Figure 8.
References (41)
- et al.
Binding proteins from alternative scaffolds
J. Immunol. Methods
(2004) - et al.
Engineered proteins as specific binding reagents
Curr. Opin. Biotechnol.
(2005) - et al.
The three-dimensional structure of the tenth type III module of fibronectin: an insight into RGD-mediated interactions
Cell
(1992) - et al.
Crystal structure of the tenth type III cell adhesion module of human fibronectin
J. Mol. Biol.
(1994) - et al.
The fibronectin type III domain as a scaffold for novel binding proteins
J. Mol. Biol.
(1998) - et al.
Engineered fibronectin type III domain with a RGDWXE sequence binds with enhanced affinity and specificity to human alphavbeta3 integrin
J. Mol. Biol.
(2003) - et al.
Molecular recognition properties of FN3 monobodies that bind the Src SH3 domain
Chem. Biol.
(2004) - et al.
Directed evolution of high-affinity antibody mimics using mRNA display
Chem. Biol.
(2002) - et al.
Engineering antibody affinity by yeast surface display
Methods Enzymol.
(2004) - et al.
The contribution of contact and non-contact residues of antibody in the affinity of binding to antigen. The interaction of mutant D1.3 antibodies with lysozyme
J. Mol. Biol.
(1993)
Molecular evolution of antibody affinity for sensitive detection of botulinum neurotoxin type A
J. Mol. Biol.
High affinity T cell receptors from yeast display libraries block T cell activation by superantigens
J. Mol. Biol.
Characterization of T cell receptors engineered for high affinity against toxic shock syndrome toxin-1
J. Mol. Biol.
Selection by phage display of llama conventional V(H) fragments with heavy chain antibody V(H)H properties
J. Immunol. Methods
Comparison of llama VH sequences from conventional and heavy chain antibodies
Mol. Immunol.
Multivalent avimer proteins evolved by exon shuffling of a family of human receptor domains
Nature Biotechnol.
Antibody mimics based on human fibronectin type three domain engineered for thermostability and high-affinity binding to vascular endothelial growth factor receptor two
Protein Eng. Des. Sel.
Probing protein conformational changes in living cells by using designer binding proteins: application to the estrogen receptor
Proc. Natl Acad. Sci. USA
Conformation-specific affinity purification of proteins using engineered binding proteins: application to the estrogen receptor
Protein Expr. Purif
Yeast surface display for screening combinatorial polypeptide libraries
Nature Biotechnol.
Cited by (86)
Engineered protein-small molecule conjugates empower selective enzyme inhibition
2022, Cell Chemical BiologyDesigner proteins that competitively inhibit Gαq by targeting its effector site
2021, Journal of Biological ChemistryA 45-Amino-Acid Scaffold Mined from the PDB for High-Affinity Ligand Engineering
2015, Chemistry and BiologyA Pronectin™ AXL-targeted first-in-class bispecific T cell engager (pAXLxCD3ε) for ovarian cancer
2023, Journal of Translational MedicineNon-Immunoglobulin Synthetic Binding Proteins for Oncology
2023, Biochemistry (Moscow)