Antibody-specific model of amino acid substitution for immunological inferences from alignments of antibody sequences.

Alexander Mirsky,Maria Anisimova,Linda Kazandjian

doi:10.1093/molbev/msu340

Abstract

Antibodies are glycoproteins produced by the immune system as a dynamically adaptive line of defense against invading pathogens. Very elegant and specific mutational mechanisms allow B lymphocytes to produce a large and diversified repertoire of antibodies, which is modified and enhanced throughout all adulthood. One of these mechanisms is somatic hypermutation, which stochastically mutates nucleotides in the antibody genes, forming new sequences with different properties and, eventually, higher affinity and selectivity to the pathogenic target. As somatic hypermutation involves fast mutation of antibody sequences, this process can be described using a Markov substitution model of molecular evolution. Here, using large sets of antibody sequences from mice and humans, we infer an empirical amino acid substitution model AB, which is specific to antibody sequences. Compared with existing general amino acid models, we show that the AB model provides significantly better description for the somatic evolution of mice and human antibody sequences, as demonstrated on large next generation sequencing (NGS) antibody data. General amino acid models are reflective of conservation at the protein level due to functional constraints, with most frequent amino acids exchanges taking place between residues with the same or similar physicochemical properties. In contrast, within the variable part of antibody sequences we observed an elevated frequency of exchanges between amino acids with distinct physicochemical properties. This is indicative of a sui generis mutational mechanism, specific to antibody somatic hypermutation. We illustrate this property of antibody sequences by a comparative analysis of the network modularity implied by the AB model and general amino acid substitution models. We recommend using the new model for computational studies of antibody sequence maturation, including inference of alignments and phylogenetic trees describing antibody somatic hypermutation in large NGS data sets. The AB model is implemented in the open-source software CodonPhyML (http://sourceforge.net/projects/codonphyml) and can be downloaded and supplied by the user to ProGraphMSA (http://sourceforge.net/projects/prographmsa) or other alignment and phylogeny reconstruction programs that allow for user-defined substitution models.

Highlights

Antibodies are glycoproteins that constitute a fundamental part of the humoral adaptive immune response and protect all jawed vertebrates from invading pathogens, such as bacteria, viruses, and parasitic eukaryotes (Das et al 2012).Studying and modeling antibody biology and functionality has important influences in several fields: understanding the tightly regulated mechanisms that govern B lymphopoiesis and antibody maturation is important for understanding the pathogenesis of diseases where these mechanisms are deregulated, such as certain types of autoimmunity, immunodeficiency, and lymphomas
Using large sequence data from different species and antibody classes, we inferred 12 candidate models describing antibody somatic evolution among which one best-fitting model was selected for its better description of antibody sequence data
We demonstrate that the AB model outperforms general amino acid substitution models such as LG and WAG, as assessed by the gains in the optimized log-likelihood values when applied to antibody sequence data

Summary

Introduction

Antibodies are glycoproteins that constitute a fundamental part of the humoral adaptive immune response and protect all jawed vertebrates (elasmobranches, teleosts, amphibians, reptiles, birds, and mammals) from invading pathogens, such as bacteria, viruses, and parasitic eukaryotes (Das et al 2012).Studying and modeling antibody biology and functionality has important influences in several fields: understanding the tightly regulated mechanisms that govern B lymphopoiesis and antibody maturation is important for understanding the pathogenesis of diseases where these mechanisms are deregulated, such as certain types of autoimmunity, immunodeficiency, and lymphomas. In any given alignment of antibody sequences, the mutational history of antibody maturation was modeled independently for each homologous residue as a stochastic Markov process of amino acid substitutions along a tree structure (or genealogy) describing the history of sampled sequences as they evolved from their respective germline sequences.

Results

Conclusion