Gene prediction was broadly consistent between three different tools; PROKKA and RAST. PROKKA predicted 4341 gene features across all contigs, whereas RAST was slightly higher with 4494 gene features. Of the genes predicated by RAST, 757 (16.8%) are ’hypothetical’ proteins. These numbers are consistent with Y.enterocolitica annotation by Batzilla et al, who predicted 4478 and 4439 genes for two 1A biovars, of which 756 and 865 were hypothetical \cite{21798805}. As a first pass analysis, assembled contigs were passed through the PathogenFinder database (https://cge.cbs.dtu.dk/services/PathogenFinder/). PathogenFindr uses a curated list of known pathogenic and non-pathogenic features to predict the likleyhood of pathogenicity. Our Y.enterocolitica query sequence matched 105 pathogenic features and 8 non-pathogenic features and, overall, the sequence was predicted to be pathogenic to humans. However, PathogenFinder is not tailored to assess gene features which are known to be highly relevant to Yersinia pathogenicity.
The presence of the pYV plasmid is regarded as one of the key elements separating pathogenic (high and low) from non-pathogenic Yersinia; the former includes y.enterocolitica 1B and 2-5, whereas the latter includes y.enterocolitica 1A. pYV is understood to be required for inflammation of the intestines following infection, and carries the T3SS-Yop secretion system that is instrumental in biovar virulence \cite{11418330}. A nucleotide blast of all assembled contigs against; i) NCBI plasmid library (taxid:36549) and ii) pYV reference sequence (GenBank:NZ_CP009845.1), failed to find any evidence for pYV in the query Yersinia. As an alternative approach cBAR (http://csbl.bmb.uga.edu/ ffzhou/cBar/) was used isolate contigs likely to be extrachromosomal \cite{20538725}. However, this approach also failed to support the presence of pYV in this sample.
In addition to pYV, the query Yersinia lacks the ail invasin gene, which is known to be a marker of pathogenicity \cite{24753568}. Ail is required for host attachment and also to evade the hosts innate immune response \cite{26377177}. Like other non-pathogenic Y.enterocolitica biovars, other invasin genes inv, ifp are present.
Another genomic feature of Yersinia that differentiates pathogenic from non-pathogenic biovars is the ’high-pathogenicity island’ (HPI) \cite{11418330}. The HPI is present in highly pathogenic Yersinia (pestis, pseudotuberculosis and enterocolitica 1B) and absent in mildly- and non-pathogenic biovars \cite{11418330}. Although the architecture of the HPI differs between Yersinia species, there is a common repertoise of yersiniabactin genes. Yersiniabactin is a sidophore pivitol for the harnessing and trafficking iron, an essential cofactor in many enzymatic reactions. The HPI found in Y.enterocolitica, HPI_Yen, is approximatly 44kbp, larger than the HPI_Yps (36.8kbp) found in other pathogenic Yersinia \cite{15493818}. HPI_Yen comprises of irp1-9, textitYbtA and fyuA, in addition to insertion sequences; IS1328 and IS1400. Genes known to flank the HPI, glycosyl hydrolases locus on one flank, and the citrate synthase locus on the other, were present within a single 157kbp contig in our query sequence. However none of the yersiniabactin machinery is found in our sample (Figure 3). Compared to a reference Y.enterocolitica biovar 1B (GenBank: AM286415.1), which does contain the HPI, the distance between flanking loci was 32kbp shorter in our query sequence. Furthermore, interrogating the ISfinder database did not reveal any of the repetative sequences (IS1328, IS1329 and IS1400) charateristic of the HPI island.