Design of Antigen-Specific Antibody CDRH3 Sequences Using AI and Germline-Based Templates.

Antibody-antigen specificity is engendered and refined through a number of complex B cell processes, including germline gene recombination and somatic hypermutation. Here, we present an AI-based technology for de novo generation of antigen-specific antibody CDRH3 sequences using germline-based templates, and validate this technology through the generation of antibodies against SARS-CoV-2. AI-based processes that mimic the outcome, but bypass the complexity of natural antibody generation, can be efficient and effective alternatives to traditional experimental approaches for antibody discovery.

Potent HPIV3-neutralizing IGHV5-51 Antibodies Identified from Multiple Individuals Show L Chain and CDRH3 Promiscuity.

Human parainfluenza virus 3 (HPIV3) is a widespread pathogen causing severe and lethal respiratory illness in at-risk populations. Effective countermeasures are in various stages of development; however, licensed therapeutic and prophylactic options are not available. The fusion glycoprotein (HPIV3 F), responsible for facilitating viral entry into host cells, is a major target of neutralizing Abs that inhibit infection. Although several neutralizing Abs against a small number of HPIV3 F epitopes have been identified to date, relatively little is known about the Ab response to HPIV3 compared with other pathogens, such as influenza virus and SARS-CoV-2. In this study, we aimed to characterize a set of HPIV3-specific Abs identified in multiple individuals for genetic signatures, epitope specificity, neutralization potential, and publicness. We identified 12 potently neutralizing Abs targeting three nonoverlapping epitopes on HPIV3 F. Among these, six Abs identified from two different individuals used Ig heavy variable gene IGHV 5-51, with five of the six Abs targeting the same epitope. However, despite the use of the same H chain variable (VH) gene, these Abs used multiple different L chain variable genes (VL) and diverse H chain CDR 3 (CDRH3) sequences. Together, these results provide further information about the genetic and functional characteristics of HPIV3-neutralizing Abs and suggest the existence of a reproducible VH-dependent Ab response associated with VL and CDRH3 promiscuity. Understanding sites of HPIV3 F vulnerability and the genetic and molecular characteristics of Abs targeting these sites will help guide efforts for effective vaccine and therapeutic development.

Development of LIBRA-seq for the guinea pig model system as a tool for the evaluation of antibody responses to multivalent HIV-1 vaccines.

Consistent elicitation of serum antibody responses that neutralize diverse clades of HIV-1 remains a primary goal of HIV-1 vaccine research. Prior work has defined key features of soluble HIV-1 Envelope (Env) immunogen cocktails that influence the neutralization breadth and potency of multivalent vaccine-elicited antibody responses including the number of Env strains in the regimen. We designed immunization groups that consisted of different numbers of SOSIP Env strains to be used in a cocktail immunization strategy: the smallest cocktail (group 2) consisted of a set of two Env strains, which were a subset of the three Env strains that made up group 3, which, in turn, were a subset of the six Env strains that made up group 4. Serum neutralizing titers were modestly broader in guinea pigs that were immunized with a cocktail of three Envs compared to cocktails of two and six, suggesting that multivalent Env immunization could provide a benefit but may be detrimental when the cocktail size is too large. We then adapted the LIBRA-seq platform for antibody discovery to be compatible with guinea pigs, and isolated several tier 2 neutralizing monoclonal antibodies. Three antibodies isolated from two separate guinea pigs were similar in their gene usage and CDR3s, establishing evidence for a guinea pig public clonotype elicited through vaccination. Taken together, this work investigated multivalent HIV-1 Env immunization strategies and provides a novel methodology for screening guinea pig B cell receptor antigen specificity at a high-throughput level using LIBRA-seq.IMPORTANCEMultivalent vaccination with soluble Env immunogens is at the forefront of HIV-1 vaccination strategies but little is known about the influence of the number of Env strains included in vaccine cocktails. Our results suggest that adding more strains is sometimes beneficial but may be detrimental when the number of strains is too high. In addition, we adapted the LIBRA-seq platform to be compatible with guinea pig samples and isolated several tier 2 neutralizing monoclonal antibodies, some of which share V and J gene usage and >70% CDR3 identity, thus establishing the existence of public clonotypes in guinea pigs elicited through vaccination.

Negative Binomial Mixture Model for Identification of Noise in Antigen-Specificity Predictions by LIBRA-seq.

Motivation: LIBRA-seq (linking B cell receptor to antigen specificity by sequencing) provides a powerful tool for interrogating the antigen-specific B cell compartment and identifying antibodies against antigen targets of interest. Identification of noise in LIBRA-seq antigen count data is critical for improving antigen binding predictions for downstream applications including antibody discovery and machine learning technologies. Results: In this study, we present a method for denoising LIBRA-seq data by clustering antigen counts into signal and noise components with a negative binomial mixture model. This approach leverages the VRC01 negative control cells included in a recent LIBRA-seq study(Abu-Shmais et al.) to provide a data-driven means for identification of technical noise. We apply this method to a dataset of nine donors representing separate LIBRA-seq experiments and show that our approach provides improved predictions for in vitro antibody-antigen binding when compared to the standard scoring method used in LIBRA-seq, despite variance in data size and noise structure across samples. This development will improve the ability of LIBRA-seq to identify antigen-specific B cells and contribute to providing more reliable datasets for future machine learning based approaches to predicting antibody-antigen binding as the corpus of LIBRA-seq data continues to grow.