Joint sequencing of human and pathogen genomes reveals the genetics of pneumococcal meningitis

Abstract

Streptococcus pneumoniae is a common nasopharyngeal colonizer, but can also cause life-threatening invasive diseases such as empyema, bacteremia and meningitis. Genetic variation of host and pathogen is known to play a role in invasive pneumococcal disease, though to what extent is unknown. In a genome-wide association study of human and pathogen we show that human variation explains almost half of variation in susceptibility to pneumococcal meningitis and one-third of variation in severity, identifying variants in CCDC33 associated with susceptibility. Pneumococcal genetic variation explains a large amount of invasive potential (70%), but has no effect on severity. Serotype alone is insufficient to explain invasiveness, suggesting other pneumococcal factors are involved in progression to invasive disease. We identify pneumococcal genes involved in invasiveness including pspC and zmpD, and perform a human-bacteria interaction analysis. These genes are potential candidates for the development of more broadly-acting pneumococcal vaccines.

Fig. 1

Overview of cohorts sequenced and associations performed. Left, host data; right, bacterial data; the centre represents samples with both host and pathogen sequence data. Samples in green are those collected from our MeninGene cohort that form the centre of this work. Owing to unbalanced case–control ratios, we show the effective sample size, specific numbers of cases and controls of human genotypes in Supplementary table 2

Fig. 2

Burden of rare variation between invasive and carriage isolates, based on mapping and calling short variants against a single reference genome. Loss-of-function (LoF) are frameshift or nonsense mutations. a The site frequency spectrum (SFS) stratified by niche and by predicted consequence. Frequency has been normalized with respect to the number of samples in each population. b Histogram of Tajima’s D for all coding sequences in the genome, stratified by niche. c Boxplot of the number of rare variants per sample, stratified by niche and predicted consequence. Damaging mutations are LoF mutations and missense mutations predicted damaging by PROVEAN. Centre line is the median, box spans lower to upper quartiles. Whiskers show the outlier range, defined as being >1.5× the interquartile range above or below the lower and upper quartiles

Fig. 3

Phylogenetic tree of all samples included in the pathogen genome-wide association study. Rings show metadata about samples, from inside to outside: phenotype (carriage or invasive); cohort (Netherlands or South Africa); common serotypes; patient age on a continuous scale from younger (white) to older (blue). Scale bar: 0.01 substitutions per site. An interactive version is available at https://microreact.org/project/Spn_GWAS/9eb0bd5d (project link https://microreact.org/project/Spn_GWAS)