Genome-wide identification of lineage and locus specific variation associated with pneumococcal carriage duration

Abstract

Streptococcus pneumoniae is a leading cause of invasive disease in infants, especially in low-income settings. Asymptomatic carriage in the nasopharynx is a prerequisite for disease, and the duration of carriage is an important consideration in modelling transmission dynamics and vaccine response. Existing studies of carriage duration variability are based at the serotype level only, and do not probe variation within lineages or fully quantify interactions with other environmental factors.Here we developed a model to calculate the duration of carriage episodes from longitudinal swab data. By combining these results with whole genome sequence data we estimate that pneumococcal genomic variation accounted for 63% of the phenotype variation, whereas host traits accounted for less than 5%. We further partitioned this heritability into both lineage and locus effects, and quantified the amount attributable to the largest sources of variation in carriage duration: serotype (17%), drug-resistance (9%) and other significant locus effects (7%). For the locus effects, a genome-wide association study identified 16 loci which may have an effect on carriage duration independent of serotype. Hits at a genome-wide level of significance were to prophage sequences, suggesting infection by such viruses substantially affects carriage duration. These results show that both serotype and non-serotype specific effects alter carriage duration in infants and young children and are more important than other environmental factors such as host genetics. This has implications for models of pneumococcal competition and antibiotic resistance, and leads the way for the analysis of heritability of complex bacterial traits. Significance statementOther than serotype, the genetic determinants of pneumococcal carriage duration are unknown. In this study we used longitudinal sampling to measure the duration of carriage in infants, and searched for any associated variation in the pan-genome. While we found that the pathogen genome explains most of the variability in duration, serotype did not fully account for this. Recent theoretical work has proposed the existence of alleles which alter carriage duration to explain the puzzle of continued coexistence of antibiotic-resistant and sensitive strains. Here we have shown that these alleles do exist in a natural population, and also identified candidates for the loci which fulfil this role. Together these findings have implications for future modelling of pneumococcal epidemiology and resistance.

Genome wide association with quantitative resistance phenotypes in Mycobacterium tuberculosis reveals novel resistance genes and regulatory regions

“…For example resistance to any of the second line agents, fluoroquinolones like MXF, SLI's like AMI, or to first line agents like PZA and EMB, nearly always co-exists with resistance to INH and RIF, and it is thus not possible to perform association conditioning on the absence of resistance to those agents. Further, the performance of linear mixed models for performing GWAS in bacteria has not been systematically studied, although applied recently to MTB and other bacteria with demonstrated success 13,[55][56][57] . We acknowledge that we cannot be certain that these models adequately control for population structure in clonal bacteria and because of this we performed validation in an independent dataset with a different lineage distribution.…”

Improved inference and prediction of bacterial genotype-phenotype associations using pangenome-spanning regressions

“…Most work on prediction of bacterial phenotypes have focused on antibiotic resistance, but many more complex phenotypes relating to bacterial virulence are now available. For these phenotypes, which are under weaker or no selection, instead of a few strong effects, multiple smaller effects are expected in the genome (28,30,54) . Therefore, a model which may include more of these effects, which would be missed with a p-value threshold, may be expected to perform well.…”

“…Supplementary figures 1-6) .By virtue of the fact that all pan-genomic variation enters this model, if a linear additive model of heritability is assumed, which it typically is in bacterial GWAS(28)(29)(30) , the value of R 2 calculated from the fit of the elastic net: also serves as an estimate of the narrow-sense heritability h 2 . As R 2 measures the variance explained by the model's predictors, in this case all genetic features, this is equivalent to the proportion of phenotypic variance explained by genetic variation , the definition of h 2 .…”

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

“…We then analysed meningitis versus carriage isolates. We first performed this analysis in the Dutch cohort (supplementary tables 5-7) revealing that many of our results involved rare variation, which has largely been ignored in previous bacterial GWAS studies 16 . This variation may be more important in a disease like meningitis where there is no pervasive selection for the phenotype.…”

“…Bacterial genome wide association studies (GWAS) provide a way to identify pneumococcal sequence variation associated with meningitis, independent of genetic background in an unbiased manner. While GWAS is more challenging in bacteria than humans due to strong population structure and high levels of pan-genomic variation, recent methodological advances have helped overcome these issues [14][15][16] .…”