Dual randomly barcoded transposon sequencing (Dual Tn-seq) data for <em>Streptococcus pneumoniae</em> D39

Gene redundancy complicates systematic characterization of gene function as single-gene deletions may not produce discernible phenotypes. Dual-TnSeq (dual transposon sequencing) couples random barcode transposon site sequencing with the Cre-lox system to enable the characterization of >1 billion double mutant strains. This data set reports dual-tnseq data for Streptococcus pneumoniae D39 in rich media. Large libraries of mutants with two different transposons, carrying two different resistance markers and each with their own barcodes, were constructed and sequenced. Then, two libraries of mutants were combined by using transformation and selected on rich media; the Cre-lox system was induced to place the two barcodes into proximity; and the pairs of barcodes were sequenced. Pairs of genes with low rates of dual insertions often indicate synthetic lethality or a milder genetic interaction. The genetic interactions identified span a wide range of biochemical processes and reveal new factors in well-studied pathways, including a novel cytidine triphosphate synthase and an activator of cell wall biosynthesis.

Methods
First, three RB-TnSeq (randomly barcoded transposon sequencing) libraries were constructed (named ML1, ML2, and ML3), and in each library of single mutants, the random 20-nucleotide barcodes were mapped to insert locations. Then, five runs of large collections of double mutants were made by transformation (either ML1 x ML2 or ML3 x ML2) and selected on blood agar plates with both antibiotics. The Cre-lox system was induced to move the two barcodes into proximity, and then the pairs of barcodes were amplified and sequenced with Illumina. Barcode pairs were counted, and very rare pairs of barcodes were discarded, as these may be chimeras. Strains where both insertions were within the central 10-90% of a gene were tabulated to give the number of strains and the number of reads per gene pair. These were combined across runs (weighting the #reads lower for the first run, which had fewer strains). We computed the expected #strains and #genes for each gene pair, based on the relative abundance of each gene in each library (i.e., expected is proportionate to count for gene 1 in library 1 x count for gene 2 in library 2). We then combined counts and expected for the two "directions" for each gene pair. Then, we examined the bias due to the chromosomal position of the two genes: we divided the genome into 30 bins and computed, for each pair of bins, the median #reads/expected and #strains/expected. These final expected #reads and #strains were scaled by these biases. Finally, for each pair of genes with sufficient coverage, we computed the read ratio (#reads / expected) and a z score for the number of strains, (#strains - expected)/sqrt(expected). A similar analysis was also conducted using all insertions in each gene (0-100%) instead of 10-90%. The latter analysis increases the number of usable insertions but may include some non-disrupting insertions.

We also generated test sets of the same kind, using a collection of 96 double mutants. After inducing Cre-lox and extracting DNA, we conducted PCRs with varying levels of template (300 - 1000 ng). These tests confirmed that rare strains were chimeras.