Antimicrobial resistance on the move: Computational methods to identify and reconstruct mobile genetic elements contributing to AMR dissemination

Antimicrobial resistance (AMR) is a growing challenge for public health. In 2019 alone, 1.27 million people have died of causes directly linked to infections caused by resistant bacteria. This problem does not only focus on the clinical health of humans, but also causes issues in other aspects of public health of humans and animals. Resistant bacteria, but also genes conferring AMR, can be transferred to humans via direct animal contact, the food chain or the environment. AMR genes can arise through mutations, but can also be transferred between cells through means of horizontal gene transfer by mobile genetic elements (MGEs). An important factor in the dissemination of AMR genes is dissemination by plasmids. These extrachromosomal DNA molecules can between bacterial cells, not necessarily following species barriers. This capability makes that plasmids are an important contributing factor to the spread of AMR genes. Therefore, to investigate the spread of AMR genes, it is essential to investigate the spread of plasmids. Two techniques to investigate the DNA of plasmids are whole genome sequencing (WGS) and metagenomics. In WGS, the genome of a cultured bacterial colony is read, whereas in metagenomics, a large part of the genetic material in environmental samples is sequenced, without prior filtering or culturing of bacteria. This makes the resulting data much more complex and generates much more raw data compared to WGS experiments. In Chapter 2 and 3, I focus on plasmids in WGS data. I describe a collection of software tools: plasmidEC, plasmidCC, and gplas2 versions Bilbao and Flevo. PlasmidEC and plasmidCC are capable of identifying sequences originating from plasmids, based on six species-specific plasmid databases and one species-agnostic plasmid database. Gplas2 is an extension of the gplas algorithm that uses the assembly graph to reconstruct plasmid fragments into plasmid ‘bins’ that are likely to originate from the same plasmid. In Chapter 4 and 5, I focus on metagenomics data to assess the spread of mobile AMR genes. In Chapter 4, I present a software pipeline called MetaMobilePicker, that uses existing tools to assemble metagenomic reads, identify MGEs and annotate AMR genes. By validating this pipeline using simulated metagenomics data with known MGEs, I show that the metagenomics assembly step, is the bottleneck for the identification of MGEs like plasmids. I show that not all reads originating from MGEs are assembled correctly, or are assembled without enough context to be correctly identified as an MGE. In Chapter 5, I focus on the composition of the microbiome, the collection of AMR genes (the resistome), and the collection of MGEs (the mobilome) in the caecum and the faeces of broiler chickens, and compare these between two interventions to prevent coccidiosis. I show that there are measurable differences in the microbiome and resistome between these intervention methods. Additionally, I identify AMR genes located on plasmids that are present in both the caecum as well as the faeces on the farmhouse floor, which makes them interesting starting points for further research into the way mobile AMR genes can spread between environments.