Engineering CDR Loops for High-Affinity Therapeutic Antibodies

Therapeutic antibodies are one of the fastest growing areas of the pharmaceutical industry. However, one challenge to their development is the optimization of antibody CDR (complementarity-determining region) loops, which govern the specificity and strength of antigen binding. 

The Significance of CDR-Grafting in Antibody Engineering

CDR-grafting is a technique for generating humanized antibodies with high affinity and low immunogenicity. This method involves transferring the murine antibody's CDRs to a human antibody framework, thus reducing the risk of immune rejection. However, simple CDR-grafting often fails to maintain the original antigen-binding affinity. To address this, scientists also transfer certain key framework residues from the murine antibody, which stabilize the CDRs in the required conformation, effectively preserving the antibody's binding properties.

Since the first humanized antibody, Zenapax (daclizumab), hit the market in 1997, CDR-grafting has been central to therapeutic antibody development.¹ Approaches have evolved to ensure that CDRs maintain their functional integrity within human frameworks, making CDR-grafting a standard in producing effective and safe antibody therapeutics.

Related: What are CDRs?

Optimizing Antibody Affinity through CDR and Framework Shuffling

Antibody optimization often requires reshuffling CDRs and frameworks to find combinations that support high-affinity binding. Framework shuffling involves combining murine CDRs with various human frameworks to identify those that best support the binding characteristics of the original murine antibody. In a study by Hwang et al., human templates were chosen based on their structural similarity to the murine CDR, allowing the engineered antibody to retain effective binding without sacrificing human compatibility.² Another approach by Dall’Acqua et al. combines murine CDRs with pools of synthetic human germline frameworks, creating a diverse library for screening optimized antibodies.³

Recent advancements enable the use of computational models to facilitate shuffling processes, significantly reducing time and resources. For instance, scientists created an optimized anti-SARS-CoV-2 antibody by computationally identifying CDR-framework combinations, achieving a sevenfold improvement in affinity and a 75-fold increase in virus neutralization, demonstrating the effectiveness of this computationally enhanced shuffling approach in developing high-affinity antibodies with therapeutic potential.⁴

Increasing Affinity through Random Mutagenesis and Hot Spot Targeting

Two primary approaches to optimizing antibody affinity are: (1) generating large libraries of randomly mutated CDRs or entire variable domains, and (2) using focused mutagenesis to target specific CDR sites, often referred to as "hot spots."

In the random approach, large libraries are produced by random mutagenesis of CDRs or variable domains, followed by screening for high-affinity variants. Techniques such as chain shuffling and error-prone PCR allow researchers to create vast pools of potential antibodies, from which the most effective ones are selected. This process often reveals mutations that, when combined, synergistically improve binding strength.

The focused mutagenesis approach, in contrast, narrows in on specific CDR regions known to influence affinity significantly. By randomizing amino acids at these hot spots, scientists create smaller libraries with higher-quality candidates. This strategy is particularly useful for antibodies that already exhibit moderate affinity, as targeted mutations can fine-tune interactions without compromising structural integrity. For example, hot spot mutagenesis has successfully generated antibodies with significantly improved antigen-binding affinity by optimizing discrete positions within CDR3 and other variable domain regions.¹

Related: Affinity Maturation

Randomized CDR Libraries and the Importance of CDR3

Among the CDR loops, CDR3 is recognized as the primary contributor to antigen binding due to its high variability and structural complexity. Therefore, it is a main target for randomization when developing antibody libraries. In early experiments, semi-synthetic libraries generated by randomizing CDR3 sequences yielded antibodies with relatively low binding affinities, generally around 100 nM. However, by introducing variations in both the heavy chain CDR3 and light chain CDRs, scientists achieved a wider diversity of binding affinities, resulting in antibodies with enhanced functional efficacy. 

Computational Advances: Structure Prediction and Machine Learning

Recent developments in computational modeling have improved CDR loop engineering. For example, H3-OPT, a toolkit integrating AlphaFold2 and protein language models, can accurately predict the structure of the CDR-H3 loop, which is highly variable and difficult to model. H3-OPT’s predictions have shown an impressive alignment with experimental structures, achieving an RMSD (root mean square deviation) of 2.24 Å on CDR-H3 loops. This tool allows scientists to predict which modifications to the CDR-H3 loop might improve affinity without needing exhaustive experimental validation.⁵

ABlooper, another deep learning-based tool, rapidly predicts CDR structures and provides confidence estimates, which helps researchers quickly identify viable candidates for further testing. It uses E(n)-Equivariant Graph Neural Networks (E(n)-EGNNs to predict each loop multiple times. Tools like these have transformed CDR engineering from an experimental-heavy approach to one that leverages large datasets to speed up the pipeline from investigational antibody to clinical candidate.⁶

References:

1. Kim, S. J., Park, Y., & Hong, H. J. (2005). Antibody Engineering for the Development of Therapeutic Antibodies. Molecules and Cells, 20(1), 17-29. https://doi.org/10.1016/S1016-8478(23)25245-0

2. Khee Hwang, W. Y., Almagro, J. C., Buss, T. N., Tan, P., & Foote, J. (2005). Use of human germline genes in a CDR homology-based approach to antibody humanization. Methods, 36(1), 35-42. https://doi.org/10.1016/j.ymeth.2005.01.004

3. Dall’Acqua, W. F., Damschroder, M. M., Zhang, J., Woods, R. M., Widjaja, L., Yu, J., & Wu, H. (2005). Antibody humanization by framework shuffling. Methods, 36(1), 43-60. https://doi.org/10.1016/j.ymeth.2005.01.005

4. Gopal, R., Fitzpatrick, E., Pentakota, N., Jayaraman, A., Tharakaraman, K., & Capila, I. (2022). Optimizing Antibody Affinity and Developability Using a Framework–CDR Shuffling Approach—Application to an Anti-SARS-CoV-2 Antibody. Viruses, 14(12), 2694. https://doi.org/10.3390/v14122694

5. Chen, H., Fan, X., Zhu, S., Pei, Y., Zhang, X., Zhang, X., Liu, L., Qian, F., & Tian, B. (2024). H3-OPT: Accurate prediction of CDR-H3 loop structures of antibodies with deep learning. BioRxiv (Cold Spring Harbor Laboratory). https://doi.org/10.7554/elife.91512.3

6. Abanades, B., Georges, G., Bujotzek, A., & Deane, C. M. (2022). ABlooper: Fast accurate antibody CDR loop structure prediction with accuracy estimation. Bioinformatics, 38(7), 1877-1880. https://doi.org/10.1093/bioinformatics/btac016