Antibody Sequence and Structure Analyses Using IMGT®: 30 Years of Immunoinformatics

Molecular convergence of neutralizing antibodies in human revealed by repeated rabies vaccination

Rabies vaccines require repeated immunization to robustly elicit neutralizing antibodies that prevent fatal diseases. Here, we analyzed rabies glycoprotein antibody repertoires at both polyclonal and monoclonal levels following repeated vaccination. Booster vaccination dramatically elevated the neutralizing activity of recalled antibodies, primarily targeting an immunodominant site III epitope with hydrophilic and rugged structures. Strikingly, the majority of site III-directed antibodies in the recall response used a convergent VH gene (IGHV3-30), and they exhibited more hydrophilic and shorter paratopes than non-site III antibodies, providing physicochemical advantages for binding to site III. Additionally, several amino acids on heavy chain CDR3 were identified as key sites for acquiring an ultrapotent neutralizing activity through site III binding. Our in-depth analysis of antibody repertoires revealed the molecular signatures of neutralizing antibodies generated by repeated rabies vaccination, possibly as a result of adaptive convergence.

The function of antibodies

Antibodies have multiple biological activities. They can both recognize and act on specific antigens. They can protect against and cause serious diseases, enhance and inhibit antibody responses, enable survival, and threaten life. Which among their many, often antagonistic properties explains that antibodies were selected half a billion years ago and transmitted to mammals across millions of generations? In other words, what is the function of antibodies? Here I examine how their structure endows antibodies with unique cognitive and effector properties that contribute to their multiple biological activities. I show that rather than specific properties, antibodies have large functional repertoires. They have a cognitive repertoire and an effector repertoire that are selected from larger available repertoires, themselves drawn at random from even larger virtual repertoires. These virtual repertoires provide the adaptive immune system with immense, constantly renewed, reservoirs of cognitive and effector functions that can be actualized at any time according to the context. I propose that such a flexibility, which enables living individuals to adapt to a rapidly changing environment, and even deal with an unknown future, may provide a better selective advantage than any particular function.

Using IMGT unique numbering for IG allotypes and Fc-engineered variants of effector properties and half-life of therapeutic antibodies

Therapeutic monoclonal antibodies (mAb) are usually of the IgG1, IgG2, and IgG4 classes, and their heavy chains may be modified by amino acid (aa) changes involved in antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), complement-dependent cytotoxicity (CDC), and/or half-life. Allotypes and Fc-engineered variants are classified using IMGT/HGNC gene nomenclature (e.g., Homo sapiens IGHG1). Allotype names follow the WHO/IMGT nomenclature. IMGT-engineered variant names use the IMGT nomenclature (e.g., Homsap G1v1), which comprises species and gene name (both abbreviated) followed by the letter v (for variant) and a number. Both allotypes and engineered variants are defined by their aa changes and positions, based on the IMGT unique numbering for C domain, identified in sequence motifs, referred to as IMGT topological motifs, as their limits and length are standardized and correspond to a structural feature (e.g., strand or loop). One hundred twenty-six variants are displayed with their type, IMGT numbering, Eu-IMGT positions, motifs before and after changes, and their property and function (effector and half-life). Three motifs characterize effector variants, CH2 1.6-3, 23-BC-41, and the FG loop, whereas three different motifs characterize half-life variants, two on CH2 13-AB-18 and 89-96 with H93, and one on CH3 the FG loop with H115.

Investigating the ability of deep learning-based structure prediction to extrapolate and/or enrich the set of antibody CDR canonical forms

Deep learning models have been shown to accurately predict protein structure from sequence, allowing researchers to explore protein space from the structural viewpoint. In this paper we explore whether "novel" features, such as distinct loop conformations can arise from these predictions despite not being present in the training data. Here we have used ABodyBuilder2, a deep learning antibody structure predictor, to predict the structures of ~1.5M paired antibody sequences. We examined the predicted structures of the canonical CDR loops and found that most of these predictions fall into the already described CDR canonical form structural space. We also found a small number of "new" canonical clusters composed of heterogeneous sequences united by a common sequence motif and loop conformation. Analysis of these novel clusters showed their origins to be either shapes seen in the training data at very low frequency or shapes seen at high frequency but at a shorter sequence length. To evaluate explicitly the ability of ABodyBuilder2 to extrapolate, we retrained several models whilst withholding all antibody structures of a specific CDR loop length or canonical form. These "starved" models showed evidence of generalisation across CDRs of different lengths, but they did not extrapolate to loop conformations which were highly distinct from those present in the training data. However, the models were able to accurately predict a canonical form even if only a very small number of examples of that shape were in the training data. Our results suggest that deep learning protein structure prediction methods are unable to make completely out-of-domain predictions for CDR loops. However, in our analysis we also found that even minimal amounts of data of a structural shape allow the method to recover its original predictive abilities. We have made the ~1.5 M predicted structures used in this study available to download at https://doi.org/10.5281/zenodo.10280181.