Bias and Error Correction in Antibody Repertoire Sequencing for More Accurate Vaccine Profiling

Recent advances in Next-generation DNA sequencing (NGS) technology have enabled high-throughput systems analysis of immunological responses. NGS of antibody repertoires has emerged as a method to support or augment existing immunological technologies such as screening (e.g. hybridoma, recombinant surface display, etc), antibody measurements (e.g. titers), and cellular characterization (e.g. ELISPOT, cell culture supernatant cytokine quantification, and flow cytometry).  One potential pitfall of antibody NGS is the lack of uncertainty when quantifying the relative abundance between antibody clones, which compromises the quantitative accuracy of repertoire data. This uncertainty stems from systematic and stochastic biases introduced during sample preparation (i.e. PCR). This is especially problematic for samples derived from sources that currently rely on primer sets, where slight differences in individual primer annealing temperatures manifest into large systematic biases in clonal relative abundance. These primer sets used for NGS of antibody variable regions from mice and humans have relied on multiplex primers composing up to 150 different individual primers to compensate for the high diversity of V genes. Therefore, we have developed an experimental-bioinformatic approach to ensure accurate quantitative assessment of clonal frequencies based on molecular transcript barcoding (MTB). MTB relies on generating cDNA with unique identifiers, incorporated by gene specific primers in the form of oligonucleotides with specified degenerate regions, to enable bias calibration data. During method development, we employed qPCR assessment to minimize bias effects due to over amplification. After PCR amplification, we then collect MTB NGS sequencing data on the Illumina MiSeq platform enabling read lengths of 2x300. Identified biases are then able to be corrected using bioinformatics approaches to produce more accurate information about clonal relative abundance. In addition to correction of clonal frequencies,  MTB enables removal of sequencing error and PCR error, thus ensuring high quality data and distinction of synthetic mutations from true biological mutations (somatic hypermutations). Furthermore, by introducing unique identifiers for each transcript, we are also developing a process to recover bioinformatically relevant genes using primers targeting those unique identifiers (dial-out PCR). By providing more accurate antibody amplicon NGS information, we are using various statistical approaches to characterize the degree of skewed clonal distributions post-immunization in mice (e.g. Shannon entropy). These types of analyses will lead to improved strategies for creating better quantitative vaccine response characterization methods. We believe these strategies will be able to support quantitative serum based epitope mapping characterization, and as a single method could be used to evaluate responses to all antigens and pathogens. This work will lead to increased speed and efficiency for profiling of vaccine-induced immune responses.