Longitudinal Analysis of Influenza Vaccination

All study participants gave informed consent, and protocols were approved by the Stanford Institutional Review Board. Five healthy humans aged 18–28 were vaccinated with the 2011–2012 seasonal trivalent inactivated influenza vaccine and gave blood 3 and 5 days before vaccination (D3 and D5), immediately before vaccination (D0), and 1, 4, 7, 9, and 11 days afterward (D1, D4, D7, D9, D11). Peripheral blood mononuclear cells were isolated, total RNA was extracted, and sequencing libraries were prepared from 500 ng of total RNA using isotype-specific IGH constant region primers for reverse transcription and IGH variable region primers for second-strand cDNA synthesis followed by PCR, following Vollmers et al. (17 (link)) and Horns et al. (18). Sequencing was performed for all libraries using the Illumina HiSeq 2500 or MiSeq platform with paired-end reads. Sequences were preprocessed using a custom informatics pipeline to perform consensus unique molecular identifier (UMI)-based error correction, annotation of V and J gene use and CDR3 length using IgBLAST (42 (link)), and isotype determination using BLASTN. Clonal lineages were identified by grouping sequences sharing the same V and J germline genes and CDR3 length, and then performing single-linkage clustering with a cutoff of 90% nucleotide identity across both the CDR3 and the rest of the variable region (18 (link)). SFSs were constructed based on somatic mutations relative to the germline V and J genes (excluding CDR3 polymorphisms because the ancestral state may not be known with high confidence in the CDR3) and then compared with simulations of evolutionary models using betatree (43 (link)) or custom software. Multiple sequence alignment was performed using a custom fast heuristic algorithm based on MUSCLE (44 (link)), and phylogenetic reconstruction was performed using FastTree (45 (link)). Selection on subclones was detected using a custom algorithm that performs greedy breadth-first search based on Fay and Wu’s H statistic (27 (link)) of subtrees. Fitness inference based on the local branching rate of a phylogeny was performed following Neher et al. (34 ).