An in silico method to assess antibody fragment polyreactivity

Antibodies are essential biological research tools and important therapeutic agents, but some exhibit non-specific binding to off-target proteins and other biomolecules. Such polyreactive antibodies compromise screening pipelines, lead to incorrect and irreproducible experimental results, and are generally intractable for clinical development. We designed a set of experiments using a diverse naive synthetic camelid antibody fragment ('nanobody') library to enable machine learning models to accurately assess polyreactivity from protein sequence (AUC > 0.8). Moreover, our models provide quantitative scoring metrics that predict the effect of amino acid substitutions on polyreactivity. We experimentally tested our model's performance on three independent nanobody scaffolds, where over 90% of predicted substitutions successfully reduced polyreactivity. Importantly, the model allowed us to diminish the polyreactivity of an angiotensin II type I receptor antagonist nanobody, without compromising its pharmacological properties. We provide a companion web-server that provides a straightforward means of predicting polyreactivity and polyreactivity-reducing mutations for any given nanobody sequence.

Biochemical patterns of antibody polyreactivity revealed through a bioinformatics-based analysis of CDR loops

Antibodies are critical components of adaptive immunity, binding with high affinity to pathogenic epitopes. Antibodies undergo rigorous selection to achieve this high affinity, yet some maintain an additional basal level of low affinity, broad reactivity to diverse epitopes, a phenomenon termed ‘polyreactivity’. While polyreactivity has been observed in antibodies isolated from various immunological niches, the biophysical properties that allow for promiscuity in a protein selected for high-affinity binding to a single target remain unclear. Using a database of over 1000 polyreactive and non-polyreactive antibody sequences, we created a bioinformatic pipeline to isolate key determinants of polyreactivity. These determinants, which include an increase in inter-loop crosstalk and a propensity for a neutral binding surface, are sufficient to generate a classifier able to identify polyreactive antibodies with over 75% accuracy. The framework from which this classifier was built is generalizable, and represents a powerful, automated pipeline for future immune repertoire analysis.

Polyreactivity and Somatic Hypermutation Analysis Reveals the Innate B Cell Origin of Human PF4/Heparin Reactive Antibodies

Heparin-induced thrombocytopenia (HIT) is a serious reaction to heparin treatment characterized by antibodies that recognize a complex formed between heparin and platelet factor 4 (PF4/H) and are capable of activating platelets and inducing a pro-thrombotic state. Although a high percentage of heparin-treated patients produce antibodies to PF4/H, only a subset of these antibodies are platelet-activating (pathogenic) and capable of causing HIT. We previously reported that we cloned B cells from six patients experiencing HIT and identified two types of PF4/H-binding antibodies: seven platelet-activating (PA) and 48 non-activating (NA). Comparison of the structural features in the PA, NA, and non PF4/H-binding (NB) clones showed that the length and the number of basic amino acid and tyrosine residue in the heavy chain complementarity determining region 3 (HCDR3) were significantly different, and was in the order of PA>NA>NB. Most significantly, the seven platelet-activating antibodies each have one of the two pathogenic motifs: RX1-2 R/KX1-2 R/H and YYYYY in an unusually long HCDR3 (≥ 20 residues). In the current study, we attempt to understand the origin of the B cells that produce the PA and NA antibodies and the nature of the immune response in HIT through analyzing somatic hypermutation and biological property of such antibodies. Longer HCDR3 and more basic Aas and Tyr residues in the HCDR3 are features of autoreactive and polyreactive antibodies. With this in mind, we tested PA and NA clones in a standard antinuclear antibody (ANA) assay and found that these clones were significantly more reactive than NB antibodies, and the plasma of HIT patients were significantly more reactive than normal plasma (Figure1). We then compared reactions of PA, NA and NB clones against a group of self and foreign antigens commonly used in polyreactivity assays: dsDNA, ssDNA, LPS, insulin, and keyhole limpet hemocyanin (KLH). About 90% of PA and NA clones were reactive to at least two antigens, this was true of only 20% of the NB clones, and the latter is consistent with the frequency of polyreactive clones in the IgG+ B cells (Figure2). Taken together, these data indicate that PA and NA antibodies are largely polyreactive. We then investigated the development of the PA and NA B cells through analyzing somatic hypermutation in the antibodies. Through analyzing the HCDR3 nucleotide insertion, trimming and VDJ segment usage, we found that longer HCDR3 typical of PF4/H-binding clones and the RKH and Y5 motifs identified in PA clones were the result of original recombination not somatic hypermutation. Consistently, the average number of nucleotide mutations in the VH genes of the binding clones was lower (PA and NA, 9.4 ± 9.5) compared to that of peripheral blood IgG+ memory B cells in healthy subjects (~18) (Figure3). Total mutation frequency in the VH and Vk CDRs of the PF4/H-binding PA and NA clones was comparable to that of the framework regions. This finding contrasts with findings made in peripheral blood IgG+ memory B cells of healthy subjects showing that the mutation frequencies are much higher in the CDRs than in the FRs of VH. Taken together, these findings suggest that affinity maturation plays a limited role in the evolution PF4/H-binding antibodies during the immune response that leads to HIT. In this study, we showed thay PF4/H-binding PA and NA IgGs are largely polyreactive antibodies and contain lower levels of mutations compared to IgG+ memory B cells. B1 and MZ B cells are innate B cells that are main producers of polyreactive natural antibodies and can respond to toll-like receptor signaling, quickly differentiate into antibody-secreting cells, and undergo IgG class switch extrafollicularly. Polyreactivity identified in the PF4/H-binding PA and NA IgGs supports the possibility that human B cells producing PF4/H-binding antibodies are innate B cells akin to MZ B cells shown to be a source of PF4/H antibodies in mice. A mutation rate lower than that of IgG+ memory cells in the PF4/H-binding IgGs is also consistent with an extrafollicular response typical of innate B cells. These observations would help to improve our understanding of the immunological responses and B cell origin in HIT patients. Disclosures Padmanabhan: Retham Technologies: Current equity holder in private company; Veralox Therapeutics: Membership on an entity's Board of Directors or advisory committees; Versiti Blood Research Institute: Patents & Royalties.

Biochemical Patterns of Antibody Polyreactivity Revealed Through a Bioinformatics-Based Analysis of CDR Loops

Antibodies are critical components of adaptive immunity, binding with high affinity to pathogenic epitopes. Antibodies undergo rigorous selection to achieve this high affinity, yet some maintain an additional basal level of low affinity, broad reactivity to diverse epitopes, a phenomenon termed “polyreactivity”. While polyreactivity has been observed in antibodies isolated from various immunological niches, the biophysical properties that allow for promiscuity in a protein selected for high affinity binding to a single target remain unclear. Using a database of nearly 1,500 polyreactive and non-polyreactive antibody sequences, we created a bioinformatic pipeline to isolate key determinants of polyreactivity. These determinants, which include an increase in inter-loop crosstalk and a propensity for an “inoffensive” binding surface, are sufficient to generate a classifier able to identify polyreactive antibodies with over 75% accuracy. The framework from which this classifier was built is generalizable, and represents a powerful, automated pipeline for future immune repertoire analysis.