No evidence for a bovine mastitis Escherichia coli pathotype

Abstract

Background: Escherichia coli bovine mastitis is a disease of significant economic importance in the dairy industry. Molecular characterization of mastitis-associated E. coli (MAEC) did not result in the identification of common traits. Nevertheless, a mammary pathogenic E. coli (MPEC) pathotype has been proposed suggesting virulence traits that differentiate MAEC from commensal E. coli. The present study was designed to investigate the MPEC pathotype hypothesis by comparing the genomes of MAEC and commensal bovine E. coli.

Results: We sequenced the genomes of eight E. coli isolated from bovine mastitis cases and six fecal commensal isolates from udder-healthy cows. We analyzed the phylogenetic history of bovine E. coli genomes by supplementing this strain panel with eleven bovine-associated E. coli from public databases. The majority of the isolates originate from phylogroups A and B1, but neither MAEC nor commensal strains could be unambiguously distinguished by phylogenetic lineage. The gene content of both MAEC and commensal strains is highly diverse and dominated by their phylogenetic background. Although individual strains carry some typical E. coli virulence-associated genes, no traits important for pathogenicity could be specifically attributed to MAEC. Instead, both commensal strains and MAEC have very few gene families enriched in either pathotype. Only the aerobactin siderophore gene cluster was enriched in commensal E. coli within our strain panel.

Conclusions: This is the first characterization of a phylogenetically diverse strain panel including several MAEC and commensal isolates. With our comparative genomics approach we could not confirm previous studies that argue for a positive selection of specific traits enabling MAEC to elicit bovine mastitis. Instead, MAEC are facultative and opportunistic pathogens recruited from the highly diverse bovine gastrointestinal microbiota. Virulence-associated genes implicated in mastitis are a by-product of commensalism with the primary function to enhance fitness in the bovine gastrointestinal tract. Therefore, we put the definition of the MPEC pathotype into question and suggest to designate corresponding isolates as MAEC.

Keywords: Bovine mastitis; Commensals; Comparative genomics; E. coli; Genomic diversity; Pathotype; Phylogeny; Virulence.

Fig. 1

Whole genome alignment phylogeny of bovine-associated and reference E. coli strains. The phylogeny is based on a whole core genome alignment of 2,272,130 bp. The best scoring maximum likelihood (ML) tree was inferred with RAxML’s GTRGAMMA model with 1000 bootstrap resamplings. The tree was visualized with Dendroscope and bootstrap values below 50 removed. E. fergusonii serves as an outgroup and the corresponding branch is not to scale. Bovine-associated E. coli are indicated by colored cows, and both E. coli pathotypes and phylogroups are designated with a color code. ST numbers from the MLST analysis for each strain are given in parentheses. E. coli isolated from cows are widely distributed in the phylogroups and both commensal and MAEC strains are interspersed in the phylogenetic groups with a polyphyletic history

Fig. 2

Gene content clustering tree of the bovine-associated E. coli. The gene content best scoring ML dendrogram is based upon the presence or absence of orthologous groups (OGs) with 1000 resamplings for bootstrap support values. The tree was visualized midpoint rooted with FigTree and bootstrap values below 50 removed. The distance between the genomes is proportional to the OGs present or absent. The tree topology of the gene content tree follows closely the core genome WGA phylogeny. There is no functional convergence between MAEC or commensal strains, rather a highly diverse gene content

Fig. 3

Venn diagrams for gene family enrichment in pathotypes or phylogroups. a Enrichment of OGs in pathotypes (MAEC or commensal) was determined statistically (numbers in parentheses; Fisher’s exact test, p < 0.05) after applying 70% inclusion and 30% exclusion group cutoffs (numbers without parentheses). Numbers with a single asterisk correspond to OGs with a statistically significant association while two asterisks indicate remaining significant associations after a Bonferroni correction. b Enrichment of OGs in phylogenetic groups (A, B1, B2, or E) was determined based on 70% inclusion and 30% exclusion group cutoffs. Statistic testing for OG association (Fisher’s exact test, p < 0.05) was performed only for the multi-genome phylogroups A versus B1. Only very few OGs could be detected as pathotype enriched. Instead, OG distribution is strongly affected by phylogenetic background

Fig. 4

Heatmap indicating presence or absence of virulence factors. Each row of the binary matrix indicates the presence or absence of a virulence-associated gene (a BLASTP+ hit). VF classes are indicated at the side in black and grey. Strain names are color-coded for MAEC (red) or commensal (green) pathotype affiliation, columns for strain phylogroup affiliation (green: A, blue: B1, orange: B2, red: E). The clustering dendrogram attached to the heatmaps is based upon the whole binary dataset (not for each heatmap separately) of a best scoring ML tree with 1000 bootstrap resamplings (a more detailed representation of the cladogram can be found in Additional file 3: Figure S2C). Bootstrap support values are arbitrarily indicated at the bifurcations of the cladogram. Statistically significant pathotype-enriched VF genes are indicated for MAEC and commensal isolates by cows in red or green, respectively. Only the aerobactin biosynthesis cluster (Aer) plus transport protein ShiF is significantly commensal-enriched and not associated with a phylogroup (indicated by a black-rimmed and opaque green cow). All other pathotype-enriched virulence-associated genes also have a significant phylogroup association. The genes of well-known and important E. coli VFs are highlighted in alternating red and brown squares: Curli = curli fibres, AFA-VIII = aggregative adherence fimbriae AFA-VIII, Auf = fimbrial adhesin, CS31A = CS31A capsule-like antigen (K88 family adhesin), Lpf = long polar fimbriae, F17b = F17b fimbriae, Pap = P/Pap pilus, Pix = Pix fimbriae, Flag-1 = E. coli peritrichous flagella 1 gene cluster, Flk = alternative flk flagellin islet, Flag-2 = E. coli lateral flagella 2 gene cluster, Heme = chu heme transport system, Enterobactin = enterobactin biosynthesis/transport gene cluster, Fec = ferric iron(III)-dicitrate uptake system, Fit = ferrichrome iron transport system, Aer = aerobactin biosynthesis cluster with iutA receptor, Ybt = yersiniabactin iron transport system, G4C = group 4 capsule, K5 = K5 capsule, T2SS-1 = gsp general secretion pathway 1, ETT2 = E. coli type three secretion system 2, T6SS/1_ECC-1470 = MAEC ECC-1470 subtype i1 T6SS/1, T6SS/2_536 = UPEC 536 subtype i2 T6SS 2, AAI/SCI-II = EAEC 042 subtype i4b T6SS 3, SCI-I = EAEC 042 subtype i2 T6SS 2, Cdt = cytolethal distending toxins, Hly = alpha-hemolysin, Mch_H47 = microcin H47. Clustering of the strains according to virulence-associated gene presence/absence also follows mostly the phylogenetic history of the strains, no clustering of pathotypes was detected. Both MAEC and commensal isolates are distinguished by the lack of classical pathogenic E. coli VFs. The same heatmap, but including gene names/locus tags, can be found in Additional file 3: Figure S6

Fig. 5

Venn diagrams of virulence-associated gene enrichment in pathotypes and phylogroups. Enriched virulence-associated genes (numbers without parentheses) were identified with 70% inclusion and 30% exclusion group cutoffs for the bovine-associated E. coli classified either by a pathotype (MAEC or commensal) or b phylogenetic groups (A, B1, B2, or E). Statistical significance of VF association was tested with Fisher’s exact test (p < 0.05, numbers with a single asterisk) and Bonferroni corrected (numbers with two asterisks). In the phylogroups association was only tested for the multi-genome phylogroups A versus B1. As with the OG enrichment analysis, phylogenetic lineage of the strains dominates VF content and only very few virulence-associated genes were enriched in the pathotypes

Fig. 6

Gene organization of the ETT2 gene cluster in the bovine-associated E. coli genomes. Comparison of the ETT2 gene cluster in the E. coli of the strain panel based on BLASTN+. Homologous regions are connected via grey vertices and colored by nucleotide identity. The genomes are ordered according to the WGA core genome phylogeny (Additional file 3: Figure S2A), which is attached on the left side (bootstrap support values below 50 were removed). Phylogroups are indicated correspondingly. MAEC strain names are colored in light red and commensals in green. Gene names are indicated above genomes encoding for these. The respective contigs of the draft genomes containing the gene cluster were concatenated (contig boundaries are indicated by red vertical lines) and CDS spanning contig borders reannotated if needed (indicated by asterisks). ETT2 contigs of genome D6-117.29 were difficult to concatenate, because of its high fragmentation. Backbone genes not belonging to ETT2 are colored black. Genes within the ETT2 region have different colors (see the legend) to be able to evaluate their presence. Pseudogenes have a lighter color fill. ETT2 shows a large number of different mutational isoforms. Nevertheless, ETT2 composition follows phylogenetic history rather than pathotype affiliation