RNA-seq procedure and RNA-seq analysis pipeline. cDNA libraries were produced for each sample from poly(A) enriched RNA. These were sequenced to three distinct read depths (~17, ~25 and ~50 M reads/sample). Reads not passing the Illumina quality filter were discarded. A) Filtered reads were mapped to the genome, allowing a maximum of one mismatch between the sequence and the reference genome (Rn5); reads that could not be mapped to the reference genome or that could be mapped to more than one genomic location (ambiguous reads) were discarded. B) The remaining reads were mapped onto the genome and classified as exonic, intronic or intergenic as described in the Methods section. C) Stacked bar charts, showing the proportions of exonic, intronic and intergenic reads, at a 50 M read depth (the same pattern was obtained at 17 M and 25 M read depths). The unstacked barcharts show there is a significantly higher proportion of reads that align to intronic regions in SNT samples than in naive samples, and that the proportion of reads mapping to exonic regions is significantly higher in naive samples than in SNT samples. P-values were calculated using the overdispersed logistic regression test described in the Methods section. Evidence of a difference between SNT and naive was found for intergenic reads, however this did not retain significance following the Bonferroni correction for multiple testing. Following alignment, gene expression was quantified by counting the number of reads mapping to each gene. Read counts were normalised by, and differential gene expression analysis was performed using DESeq. The effect of reads mapping to intronic regions on differential gene expression was assessed by comparing exonic expression to exonic and intronic expression.