A Reference Pan-Genome Approach to Comparative Bacterial Genomics: Identification of Novel Epidemiological Markers in Pathogenic <i>Campylobacter</i>

<div><p>The increasing availability of hundreds of whole bacterial genomes provides opportunities for enhanced understanding of the genes and alleles responsible for clinically important phenotypes and how they evolved. However, it is a significant challenge to develop easy-to-use and scalable methods for characterizing these large and complex data and relating it to disease epidemiology. Existing approaches typically focus on either homologous sequence variation in genes that are shared by all isolates, or non-homologous sequence variation - focusing on genes that are differentially present in the population. Here we present a comparative genomics approach that simultaneously approximates core and accessory genome variation in pathogen populations and apply it to pathogenic species in the genus <i>Campylobacter</i>. A total of 7 published <i>Campylobacter jejuni</i> and <i>Campylobacter coli</i> genomes were selected to represent diversity across these species, and a list of all loci that were present at least once was compiled. After filtering duplicates a 7-isolate reference pan-genome, of 3,933 loci, was defined. A core genome of 1,035 genes was ubiquitous in the sample accounting for 59% of the genes in each isolate (average genome size of 1.68 Mb). The accessory genome contained 2,792 genes. A <i>Campylobacter</i> population sample of 192 genomes was screened for the presence of reference pan-genome loci with gene presence defined as a BLAST match of ≥70% identity over ≥50% of the locus length - aligned using MUSCLE on a gene-by-gene basis. A total of 21 genes were present only in <i>C. coli</i> and 27 only in <i>C. jejuni</i>, providing information about functional differences associated with species and novel epidemiological markers for population genomic analyses. Homologs of these genes were found in several of the genomes used to define the pan-genome and, therefore, would not have been identified using a single reference strain approach.</p>