Investigation of horizontal gene transfer of pathogenicity islands in Escherichia coli using next-generation sequencing

Abstract

Horizontal gene transfer (HGT) contributes to the evolution of bacteria. All extraintestinal pathogenic Escherichia coli (ExPEC) harbour pathogenicity islands (PAIs), however relatively little is known about the acquisition of these PAIs. Due to these islands, ExPEC have properties to colonize and invade its hosts efficiently. Even though these PAIs are known to be acquired by HGT, only very few PAIs do carry mobilization and transfer genes required for the transmission by HGT. In this study, we apply for the first time next-generation sequencing (NGS) and in silico analyses in combination with in vitro experiments to decipher the mechanisms of PAI acquisition in ExPEC. For this, we investigated three neighbouring E. coli PAIs, namely the high-pathogenicity island (HPI), the pks and the serU island. As these PAIs contain no mobilization and transfer genes, they are immobile and dependent on transfer vehicles. By whole genome sequencing of the entire E. coli reference (ECOR) collection and by applying a phylogenetic approach we could unambiguously demonstrate that these PAIs are transmitted not only vertically, but also horizontally. Furthermore, we could prove in silico that distinct groups of PAIs were transferred "en bloc" in conjunction with the neighbouring chromosomal backbone. We traced this PAI transfer in vitro using an F' plasmid. Different lengths of transferred DNA were exactly detectable in the sequenced transconjugants indicating NGS as a powerful tool for determination of PAI transfer.

Fig 1. Radial tree of the six housekeeping gene fragments.

The radial tree of the six housekeeping gene fragments (trpA, trpB, pabB, putP, icd and polB) from the ECOR collection and strains S107, S108 and 536 performed by PhyML using the Maximum Likelihood algorithm with Bayesian branch support. The scale bar represents the number of SNPs per nucleotide. The node colour represents the distribution of the PAIs. The node shapes show the phylogenetic group according to the triplex PCR [2].

Fig 2. Radial tree of the E. coli core genome.

The radial tree of the core genome was generated by Parsnp. Strain MG1655 was set as reference. The total coverage among all sequences was 40.9%. The phylogenetic groups are highlighted. The scale bar represents the number of SNPs per nucleotide. The node colour represents the distribution of the PAIs. The node shapes show the phylogenetic group according to the triplex PCR [2].

Fig 3. The phylogenetic tree of the entire HPI.

All strains are at least HPI-positive. The text and dot colour represents the PAI-group and the dot shape the phylogenetic group. Except strain ECOR65 (asterisk) from PAI-group 2a, all members of PAI-groups 2a (blue), 2b (green) and 3 (red) showed a HPI subtype specific for their group. The utilized algorithm was Maximum Likelihood with Bayesian branch support performed by PhyML. The scale bar represents the percentage of SNPs per nucleotide. The length of the HPI sequence is about 31.5 kb. The average homology and SNPs within the PAI-groups: 2a 99.99% (4.5), 2b 99.98% (7), 3 99.99% (2.4). The average homology and SNPs between the PAI-groups: 2b-2a 99.63% (116.8), 2b-3 99.54% (144.2), 2a-3 99.59% (129.9). The average homology and SNPs between ECOR65 and the PAI-groups: EC65-2a 99.66% (107.3), EC65-2b 99.67% (104.5), EC65-3 99.53% (149.3).

Fig 4. The phylogenetic tree of the entire serU island.

All strains are at least HPI- and serU island-positive. The text and dot colour represents the PAI-group and the dot shape the phylogenetic group. The members of PAI-groups 2b (green) and 3 (red) showed a serU island subtype specific for their group. The algorithm which was used by PhyML was Maximum Likelihood with Bayesian branch support. The scale bar represents the percentage of SNPs per nucleotide. The size of the serU island is about 27 kb. The average homology and number of SNPs were similar within the PAI-groups for 2b and 3 with 99.93% (18.0) and 99.94% (16.6), respectively. Between these two PAI-groups, the homology was 99.27% with 197.3 SNPs on average.

Fig 5. The phylogenetic tree of the entire pks island.

All strains are at least HPI- and pks island-positive. The text and dot colour represents the PAI-group and the dot shape the phylogenetic group. Except strain ECOR65 (asterisk) from PAI-group 2a, all members of PAI-groups 2a (blue) and 3 (red) showed a pks island subtype specific for their group. The algorithm we used was Maximum Likelihood with Bayesian branch support performed by PhyML. The scale bar represents the number of SNPs per nucleotide. The sequence length of the pks island is about 54.5 kb. Within PAI-group 2a the homology and the number of SNPS on average are 99.99% and 5.7 respectively, within PAI-group 3 99.97% and 18.3. The average homology and SNPs between ECOR65 and the PAI-groups: EC65-2a 99.89% (61.8), EC65-3 99.89% (59.3).

Fig 6. The arrangement of the PAIs on the chromosome.

Each island is inserted in a tRNA (serU island: serU tRNA; HPI: asnT tRNA; pks island: asnW tRNA). The size is given in kilobases (kb). The regions between the PAIs are called inter-PAI regions (between the serU island and the HPI (region A): about 1 kb; between the HPI and the pks island (region B): about 12 kb).

Fig 7. Phylogenetic tree of region A.

The inter-PAI region between the serU island and the HPI (region A) is about 1 kb and is shown as phylogenetic tree. The text and dot colour represents the PAI-group and the dot shape the phylogenetic group. The algorithm which was used by PhyML was Maximum Likelihood with Bayesian branch support. The scale bar represents the percentage of SNPs per nucleotide. Within PAI-groups 2a and 2b, the percental homology is 100% without any SNP. Within PAI-group 3, the homology is 99.97% and the number of SNPs 0.3 on average.

Fig 8. Phylogenetic tree of region B.

The dendrogram of the inter-PAI region between the HPI and the pks island (region B). The size of the sequence is about 12 kb. The text and dot colour represents the PAI-group and the dot shape the phylogenetic group. The algorithm which was used by PhyML was Maximum Likelihood with Bayesian branch support. The scale bar represents the number of SNPs per nucleotide. For the PAI groups 2a, 2b and 3, the average homology was 99.99%,99.98% and 99.98% respectively and the number of SNPs were 1.3, 3 and 2.3 on average.