Many large synthetic antibody libraries have been designed and constructed to successfully generate high-quality antibodies suitable for various demanding applications. An advantage of synthetic antibody libraries is the ability to exert fine control over factors like framework sequences, amino acid and codon usage, and complementary determining region (CDR) diversity. While synthetic antibody libraries have many advantages such as optimized framework sequences and a broader sequence landscape than natural antibodies, their sequence diversities are typically generated by random combinatorial synthetic processes. However, random combinatorial synthesis of oligonucleotides for CDR sequence diversity also produces many clones with unnatural sequences and/or undesirable modification motifs.
To alleviate these issues, researchers from Ewha Womans University describe the construction and validation of a human single-chain variable fragment (scFv) library with a novel design approach to synthetic CDR diversification.
The SCIEN (Simulation of CDRs Inspired by and Emulating Nature) principle is based on the simulation of natural rearranged and hypermutated CDRs and the parallel synthesis of thousands of oligonucleotides encoding the predefined CDR sequences. Because the CDR sequence diversity is predefined and synthesized without relying on random combinatorial events, the incorporation of undesired sequences to the library can be prevented in the design stage.
The researchers designed and constructed a novel semi-synthetic human scFv library with non-combinatorial, pre-designed CDR diversity and a single native human framework each for heavy, kappa, and lambda chain variable domains.
The design process
CDR sequences of thousands of natural human antibodies were compared with the human germline CDR sequences and their somatic hypermutation patterns/frequency, germline sequence usage, and length distribution were analyzed. Thousands of CDR sequences were simulated based on this analysis and sequences with undesirable post-translational modification motifs were removed from the repertoire. The resulting CDR repertoires contained the sequences that dutifully mimic the naturally produced human CDRs, but without many of the deleterious sequence motifs.
Construction of scFv library with pre-defined CDR sequences via OligoMix® synthesis
OligoMix® pools of oligonucleotides encoding the designed CDR sequences, with 5’- and 3’- flanking sequences from the framework regions of human variable heavy chain gene DP-47, kappa light chain gene DPK-22, or lambda light chain gene DPL3, were prepared by array synthesis. A maximum of 3,918 oligonucleotides could be synthesized per pool. Therefore, CDR-H3 (7,836 sequences) were prepared in two separate pools to increase the diversity of this critical region. Other CDRs (a total of 11,000 sequences) were synthesized in three additional pools.
Pools of oligonucleotides with the designed CDR sequences were array-synthesized and amplified by PCR. These CDR-encoding oligonucleotides were inserted into the template scFv sequence by PCR. A single-CDR library (scFv library with only one of the six CDRs are diversified) was constructed for each CDR using the amplified CDR oligonucleotide mixtures.
Next-generation sequencing analysis indicated that the library consists of antibody clones with highly nature-like CDR sequences
Next generation sequencing (NGS) of the library was performed to assess the fidelity with which the CDR design was reflected in the constructed library. Millions of CDR sequences were analyzed and compared with the designed sequences. The designed sequences were nearly completely covered in the constructed library, and the frequency of occurrence of each designed CDR sequence also roughly represented in the actual library. Overall, next-generation sequencing analysis indicated that the library consisted of antibody clones with highly nature-like CDR sequences and the occurrence of the post-translational modification motifs was minimized.
Multiple unique clones with nanomolar affinity could be isolated from the library against a number of target antigens, validating the library design strategy
The constructed library was panned against four protein antigens to validate its functionality. Each single-CDR library was panned against an anti-HA antibody that binds to an HA-tag at the C-terminus of scFv in order to proofread the CDR repertoires for in-frame sequences. Multiple target-binding scFv clones were isolated from the library after four rounds of panning on antigens. Output colonies from the third or fourth round of panning were screened by ELISA.
Binding kinetics of some of the ELISA-positive scFv clones isolated from the library were analyzed by SPR. Dissociation constants (Kd) ranging from 10−9 to 10−7 M were obtained for 10 scFvs against three different antigens. These values were comparable to those from other previously reported antibody libraries, as well as typical hybridoma-derived antibodies, supporting the validity of the non-combinatorial CDR design approach.
Although the diversity of each of the six CDRs is low because of the non-combinatorial synthetic approach (500 ~ 8000 unique sequences per CDR in the current example), the combination of six such regions provides a total diversity that is large enough to construct a highly functional antibody library.
These results demonstrate that it is possible to construct a functional antibody library using low, non-combinatorial synthetic CDR diversity, and provides a new strategy for the design of antibody libraries suitable for demanding applications.