Search

The spread of antimicrobial resistance (AMR) among pathogenic bacterial species threatens to undercut much of the progress made in treating infectious diseases. AMR genes can disseminate between and within populations via horizontal gene transfer (HGT). Selfish mobile genetic elements (MGEs) can encode resistance and spread between host cells. Homologous recombination can alter the core genes of pathogens with resistant donors via HGT too. MGEs may be cured from host genomes through transformation. Hence, MGEs may be able to avoid deletion by disrupting transformation. This work aims to understand how the dynamics of these processes affect the epidemiology of AMR pathogens. To understand these dynamics, I co-developed a new version of the popular recombination detection tool Gubbins. Through simulation studies, I find this new version to be both accurate in reconstructing the relationships between isolates, and efficient in terms of its use of computational resources. I then apply Gubbins to both AMR lineages and species-wide datasets of the pathogen Streptococcus pneumoniae. I find that recombination frequently occurs around core genes involved in both drug resistance and the host immune response. Additionally, an MGE was able to successfully spread within a population by disrupting the transformation machinery, preventing its loss from the host. Finally, I investigate two recent examples of MGEs disrupting transformation in the gram-negative species Acinetobacter baumannii and Legionella pneumophila. I find that while these insertions may decrease the efficiency of transformations within cells, the observed recombination rates largely reflect the selection pressures on isolates. With MGEs only partially able to inhibit these observable transformation events. These results show how selection pressures from clinical interventions shape pathogen genomes through diverse, often interspecies, recombination events. The spread of MGEs can also be favoured by both these selection pressures, and their ability to disrupt host cell machinery.

Of over 100 serotypes of Streptococcus pneumoniae, only 7 were included in the first pneumo- coccal conjugate vaccine (PCV). While PCV reduced the disease incidence, in part because of a herd immunity effect, a replacement effect was observed whereby disease was increasingly caused by serotypes not included in the vaccine. Dynamic transmission models can account for these effects to describe post-vaccination scenarios, whereas economic evaluations can enable decision-makers to compare vaccines of increasing valency for implementation. This thesis has four aims. First, to explore the limitations and assumptions of published pneu- mococcal models and the implications for future vaccine formulation and policy. Second, to conduct a trend analysis assembling all the available evidence for serotype replacement in Europe, North America and Australia to characterise invasive pneumococcal disease (IPD) caused by vaccine-type (VT) and non-vaccine-types (NVT) serotypes. The motivation behind this is to assess the patterns of relative abundance in IPD cases pre- and post-vaccination, to examine country-level differences in relation to the vaccines employed over time since introduction, and to assess the growth of the replacement serotypes in comparison with the serotypes targeted by the vaccine. The third aim is to use a Bayesian framework to estimate serotype-specific invasiveness, i.e. the rate of invasive disease given carriage. This is useful for dynamic transmission modelling, as transmission is through carriage but a majority of serotype-specific pneumococcal data lies in active disease surveillance. This is also helpful to address whether serotype replacement reflects serotypes that are more invasive or whether serotypes in a specific location are equally more invasive than in other locations. Finally, the last aim of this thesis is to estimate the epidemiological and economic impact of increas- ing serotype coverage in PCVs using a dynamic transmission model. Together, the results highlight that though there are key parameter uncertainties that merit further exploration, divergence in serotype replacement and inconsistencies in invasiveness on a country-level may make a universal PCV suboptimal.