Refining and characterizing signals of selection in bacterial pangenomes

Abstract

Bacteria maintain more genetic variation across their populations than any individual member can carry. This collective gene pool (the pangenome) is continuously reshaped by gene gain, loss, and selection. Understanding the forces that maintain its diversity remains a central question in microbial evolutionary biology. One compelling mechanism that could underlie this diversity is negative frequency-dependent selection (NFDS), in which rare variants are favored and common ones disfavored, generating stable intermediate frequencies that are maintained across populations and restored following perturbation. Signatures of NFDS have been consistently documented in the pneumococcal pangenome, yet the biological functions driving them have remained unidentified. In this dissertation, we sought to identify the biological functions generating this signal. First, we develop CLARC (Connected Linkage and Alignment Redefinition of COGs), a bioinformatics tool that corrects a systematic error in pangenome analyses: the over-splitting of alleles of the same gene into separate orthologous clusters. We show that widely used tools inflate the pneumococcal accessory genome by over 30% through this misclassification, obscuring downstream evolutionary signals. By combining sequence identity with functional annotation and population-level co-occurrence, CLARC generates refined gene definitions that improve core and accessory genome classification across multiple bacterial species and substantially strengthen the detectable NFDS signal in pneumococcus. Second, we investigate whether phage defense systems are direct targets of NFDS in natural pneumococcal populations. Using longitudinal carriage data from three independent cohorts spanning different geographic settings and vaccine histories, we show that intermediate-frequency defense systems are conserved in prevalence across populations, shift predictably following vaccine introduction, and re-equilibrate toward pre-vaccine frequencies more strongly than can be explained by genomic linkage to other selected loci. These findings identify phage defense systems as direct targets of frequency-dependent selection rather than passive passengers and establish them as one functional class contributing to the broader NFDS signal in the accessory genome. Third, we ask how much of this NFDS signal is attributable to defense systems, and whether other functions contribute independently. We find that NFDS acting on defense systems explains a large portion of the observed re-equilibration signal across the accessory genome, but not all. The residual signal (reproducible across two clonally independent populations) pointed to a set of candidate genes encoding a variety of functions including components of the type 1 pilus, capsule biosynthesis clusters, antigenic surface proteins, among others. Incorporating these candidates into an NFDS-based predictive model significantly improved post-vaccine strain frequency predictions in both populations independently. Together, these studies refine how signals of selection are detected in bacterial pangenomes and characterize one of the forces responsible for generating them.