Abstract: To better understand Streptococcus pneumoniae pathogenesis we performed RNA sequencing on cerebrospinal fluid (CSF) from meningitis patients to identify bacterial genes expressed during invasion of the central nervous system. Comparison to transcriptome data for serotype 1 S. pneumoniae cultured in ex vivo human CSF defined a subset of 57 genes with high expression during human meningitis. Deletion of two of the most highly expressed genetic loci, bgaA (encodes for a ß-galactosidase) or the SP_1801-5 putative stress response operon, resulted in S. pneumoniae strains still able to transmigrate the blood brain barrier but which were more susceptible to complement opsonisation and unable to maintain brain infection in a murine meningitis model. In 1144 meningitis patients, infection with bgaA containing S. pneumoniae strains was associated with a higher mortality (22% versus 14% p=0.02). These data demonstrate that direct bacterial RNAseq from CSF can identify previously undescribed S. pneumoniae virulence factors required for meningitis pathogenesis.Data available: Tab 1: Summary transcripts/million reads (TPM) per human meningitis sample, annotated by gene name and gene number (TIGR4 annotation) across the entire mapped pneumococcal transcriptome. Tab 2: Summary differential gene expression of S. pneumoniae in vitro between THY, CSF and CSF + neutrophils. Median TPM for each sample type from 8 replicates are given, with the fold change and statistical significance. Gene numbers are given for the original human mapped S. pneumoniae serotype 1 strain gamPNI0373 (genbank CP001845.1) for ease of comparison with the human meningitis data. Tab 3: Mapped alignment of the meningitis reads from human samples against a range of pneumococcal serotypes. Heatmap colours indicate higher alighment rates.