Diverse evolutionary patterns of pneumococcal antigens identified by pangenome-wide immunological screening

Abstract

Characterizing the immune response to pneumococcal proteins is critical in understanding this bacterium's epidemiology and vaccinology. Probing a custom-designed proteome microarray with sera from 35 healthy US adults revealed a continuous distribution of IgG affinities for 2,190 potential antigens from the species-wide pangenome. Reproducibly elevated IgG binding was elicited by 208 "antibody binding targets" (ABTs), which included 109 variants of the diverse pneumococcal surface proteins A and C (PspA and PspC) and zinc metalloprotease A and B (ZmpA and ZmpB) proteins. Functional analysis found ABTs were enriched in motifs for secretion and cell surface association, with extensive representation of cell wall synthesis machinery, adhesins, transporter solute-binding proteins, and degradative enzymes. ABTs were associated with stronger evidence for evolving under positive selection, although this varied between functional categories, as did rates of diversification through recombination. Particularly rapid variation was observed at some immunogenic accessory loci, including a phage protein and a phase-variable glycosyltransferase ubiquitous among the diverse set of genomic islands encoding the serine-rich PsrP glycoprotein. Nevertheless, many antigens were conserved in the core genome, and strains' antigenic profiles were generally stable. No strong evidence was found for any epistasis between antigens driving population dynamics, or redundancy between functionally similar accessory ABTs, or age stratification of antigen profiles. These results highlight the paradox of why substantial variation is observed in only a subset of epitopes. This result may indicate only some interactions between immunoglobulins and ABTs clear pneumococcal colonization or that acquired immunity to pneumococci is an accumulation of individually weak responses to ABTs evolving under different levels of functional constraint.genomics | pathogens | evolution | immunology | epidemiology

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

“…Table 3 shows the genes which were significant in this combined analysis using any of our association methods. The genes noted in bold font in table 3 are all immunogenic 32 , and have all previously been associated with pneumococcal virulence in animal models [33][34][35][36] , but this is the first time an association has been shown in patients with invasive disease. Previous conclusions, drawn from protein binding to the Laminin receptor, have suggested that pspC ( cbpA ) is necessary for meningitis 37 .…”

“…We then used GWAS to find locus effects on invasive disease, independent of serotype. As pneumococcal genomes vary extensively in their pan-genome 30 , as well as having phenotypically important mosaic structural variants 31 and antigens alleles 32 , we used a combination of methods to catalogue the population level variation then tested for associations while adjusting for population structure. We first performed this GWAS with disease severity, which was only measured in the Dutch cohort.…”

“…dacB is involved in preserving cell wall shape, and has shown to attenuate virulence in a mouse model of lung infection. zmpD is homologous to IgA1 protease ( zmpA ) 39 , and while it is immunogenic 32 , its function is unknown -these results suggest a role in human cases of disease. The other genes in table 3 have not previously been directly associated with virulence or invasive disease in S. pneumoniae .…”

“…We produced a core gene alignment using roary 71 , where we used reciprocal best BLAST 72 to choose a cut-off of 95% BLAST ID that maximised the balanced accuracy. To produce a presence/absence matrix of accessory elements, to be used as variants in association testing, we mapped the annotated genes in each isolate to a manually curated reference set 32,73 . As a first pass we used cd-hit 74 , and then used blastp against the genes which were unclustered after the first pass.…”

“…For pspC , we used an existing classification scheme of 11 alleles from 48 sequences 76 (supplementary figure 17). For zmpA and pspA we built trees from previously characterised alleles 32 (supplementary figures 18 & 19). Finding that the ancestral branches in the phylogenies were poorly supported, while the topology of clades were well resolved, we took a cut through the deep branches of the phylogeny to define four allele groups for pspA and three for zmpA .…”