Next-generation antibody discovery platforms

Antibody discovery platforms have emerged in recent years as an important source of both therapeutic molecules and research reagents. In PNAS, Larman et al. (1) provide the foundation for the next generation of antibody production using a rationally designed, fully defined scFv library and analysis pipeline, which is optimized for analysis with short-read deep-sequencing technologies. DNA sequencing has played a pivotal role in our understanding of antibody structure function and diversification by the immune system (2). It was instrumental in the identification of antibody variable and hypervariable regions and to our understanding of the manner in which the Ig repertoire is generated (3). Remarkably, although massively parallel sequencing instrumentation was not even commercially available a few years ago, we have seen a rapid transition in high-throughput sequencing (HTS) reflecting a dynamic technology transition in this field (4). This instrumentation has been transformative for genomic research and has major consequences for the future development of antibody platforms, as evidenced by its recent application by Larman et al. (1). Integrating HTS with protein or peptide display technologies allows many of the arduous and lengthy upfront screening procedures to be by-passed or complemented. The sequence information provides valuable insights into the selection process and thereby has the potential to greatly improve library design and quality of antibody repertoires (5).

Antibody Libraries

Antibodies are reagents valued for their ability to bind molecular surfaces with high-affinity and specificity. Antibodies and related products represent the fastest growing class of therapeutic agents (6, 7). A variety of in vitro approaches have been developed to generate libraries for identification of antibodies against numerous targets (8). In vitro display and selection approaches involve three key stages. The first is the generation of a large collection of variants (the so-called library). The second is several rounds of enrichment for variants that possess the desired properties via the genotype–phenotype linkage provided by the display system used. The final stage involves functional screening and characterization of selected variants (5).

Historically libraries were graded according to their size and diversity because this correlated with the number and affinity of target-specific antibodies that could be isolated. Sanger DNA sequencing was used to monitor the process at each of the stages and identify sequences of interest. This posed major limitations because only a minute fraction of the library was actually sampled, and its content therefore received only a superficial evaluation. HTS sequencing approaches are much better suited to library

Larman et al. have developed an antibody production platform that has seamless integration with short-read DNA sequencing technology.

evaluation. The Illumina HiSeq2500 instrument currently yields up to 200 million (2 × 10⁸) mapped reads per lane, with read lengths of up to 150 bp, and is therefore ideally suited for providing a comprehensive evaluation of library diversity and quality.

Strategies for Antibody Library Diversification

Antibody library diversification strategies have been based on the amplification and random association of naturally rearranged heavy and light chain variable genes. An alternative approach has been the introduction of synthetic diversity into the complementarity-determining regions (CDRs) of selected antibody frameworks (2, 8, 9, 10). Larman et al. (1) have developed an antibody production platform that has seamless integration with short-read DNA sequencing technology.

Amino acid residues within CDRs can contribute to antigen binding directly, via contribution of a side group that makes contact(s) with the antigen. Additionally, the amino acids can have an indirect effect, affecting the conformation of the peptide backbone in a manner that facilitates direct interaction of neighboring amino acid side groups. A CDR sequence design that uses a two-state hidden Markov model (HMM) trained on all antibody–antigen cocrystal structures present in the Protein Data Bank was implemented by Larman et al. (1). This enabled the behavior of CDR amino acid sequences to be captured with the “contact” state enriched for amino acids capable of sharing/exchanging electrons or burying hydrophobic surfaces, and the “noncontact” state enriched for amino acids capable of constraining or relaxing the CDR polypeptide backbone as appropriate.

An underlying feature of HMMs is the requirement that the state of each position is dependent upon its nearest neighbor. This obligation would pose a challenge to traditional synthetic antibody library approaches in which degenerate nucleotides or codons are used. Only the composition of one amino acid position at a time could be controlled using this strategy. Larman et al. (1) circumvent this problem with an approach that synthesizes complete HMM-derived CDR sequences as releasable oligonucleotides on a programmable microarray. This elegant approach affords the ability to weed out undesirable sequences that could introduce unwanted peptide motifs or restriction sites that would lead to deleterious effects.

A ribosome display strategy was chosen to implement the antibody production pipeline (1). This technique is widely used to create proteins that can bind a desired ligand with picomolar affinities (11). Ribosome display yields translated proteins that remain associated with their cognate mRNAs. The mRNA–protein complex is used to bind to an immobilized ligand via a selection step. The mRNA–protein hybrids that bind with affinities are subsequently selected, reverse transcribed to cDNA, and their sequence amplified via PCR, yielding a nucleotide sequence that encodes strongly binding proteins. A major advantage of ribosome display libraries is that they can be constructed and transcribed entirely in vitro, thus bypassing the many limitations with bacterial transformation or phage-based infections. These antibody libraries can be readily subjected to repeated rounds of selection and optionally mutagenic reamplification to enrich for scFv variants not present in the original library.

As HTS platforms evolve and mature further this approach will undoubtedly become widely integrated into synthetic antibody research and discovery programs. This will enable low-cost production of high-quality synthetic antibodies. In the short term the ability of these reagents to bind select ligands will likely have to be validated by FACS analysis to rule out false positives or inaccurate quantification of scFv clone enrichment. Finally, the continued availability and steady increase in antibody–antigen cocrystal structures will lead to greater improvements in HMM models and further advance the development of high-affinity antibodies.