Specific strains of Escherichia coli are pathogenic for the endometrium of cattle and cause pelvic inflammatory disease in cattle and mice

Abstract

Background: Escherichia coli are widespread in the environment and pathogenic strains cause diseases of mucosal surfaces including the female genital tract. Pelvic inflammatory disease (PID; metritis) or endometritis affects approximately 40% of cattle after parturition. We tested the expectation that multiple genetically diverse E. coli from the environment opportunistically contaminate the uterine lumen after parturition to establish PID.

Methodology/principal findings: Distinct clonal groups of E. coli were identified by Random Amplification of Polymorphic DNA (RAPD) and Multilocus sequence typing (MLST) from animals with uterine disease and these differed from known diarrhoeic or extra-intestinal pathogenic E. coli. The endometrial pathogenic E. coli (EnPEC) were more adherent and invasive for endometrial epithelial and stromal cells, compared with E. coli isolated from the uterus of clinically unaffected animals. The endometrial epithelial and stromal cells produced more prostaglandin E(2) and interleukin-8 in response to lipopolysaccharide (LPS) purified from EnPEC compared with non-pathogenic E. coli. The EnPEC or their LPS also caused PID when infused into the uterus of mice with accumulation of neutrophils and macrophages in the endometrium. Infusion of EnPEC was only associated with bacterial invasion of the endometrium and myometrium. Despite their ability to invade cultured cells, elicit host cell responses and establish PID, EnPEC lacked sixteen genes commonly associated with adhesion and invasion by enteric or extraintestinal pathogenic E. coli, though the ferric yersiniabactin uptake gene (fyuA) was present in PID-associated EnPEC. Endometrial epithelial or stromal cells from wild type but not Toll-like receptor 4 (TLR4) null mice secreted prostaglandin E(2) and chemokine (C-X-C motif) ligand 1 (CXCL1) in response to LPS from EnPEC, highlighting the key role of LPS in PID.

Conclusions/significance: The implication arising from the discovery of EnPEC is that development of treatments or vaccines for PID should focus specifically on EnPEC and not other strains of E. coli.

Figure 1. Phylogeny of E. coli isolated from the postpartum uterus.

(A) Distribution of uterine E. coli isolates in phylogenetic Triplex-PCR group A (n = 37 isolates), B1 (n = 51), B2 (n = 3) and D (n = 23) between clinically unaffected animals (normal, □) and animals with pelvic inflammatory disease (PID, ▪). (B) Distribution of the uterine E. coli isolates in Triplex-PCR group A to D between weeks one to four post partum. (C) Distribution of bacteria in RAPD genotype A4 (n = 14), B1-1 (n = 20), B1-4 (n = 13) and D5 (n = 5) collected 1 to 4 weeks post partum between clinically unaffected (normal, □) or animals with pelvic inflammatory disease (PID, ▪). Bars represent the percent of isolates within a genotype.

Figure 2. Multilocus sequence typing of E. coli.

Dendrogram of E. coli isolates from the uterus of clinically unaffected or diseased (•) animals based on Multilocus sequence typing (MLST) using genes listed in Materials and Methods. The boostrap values are provide, the RAPD groups are in parenthesis and the MLST clusters of bacteria are indicated. Eleven reference strains of E. coli are also presented in the dendogram, including E. Coli K12 (K12), enteroinvasive E. coli (EIEC), enteropathogenic E. coli (EPEC), enterohaemorrhagic E. coli (EHEC; human or bovine origin), Shiga toxin-producing E. coli (STEC; bovine origin), uropathogenic E. coli (UPEC), avian pathogenic E. coli (APEC), enteroadherent E. coli (EAEC) and a bovine mastitis strain of E. coli.

Figure 3. Adhesion of E. coli to bovine endometrial cells.

Bacteria from MLST clusters 1 to 4 (n≥4 per cluster) were added to confluent (A) epithelial or (B) stromal endometrial cells, incubated for 1 h and the number of adherent CFU measured; experiments were repeated on 4 separate occasions with 4 to 6 isolates from each cluster. Values were compared to cluster 1. Adherence of bacteria to (C) epithelial or (D) stromal cells was inhibited by addition of 2.5% D-Mannose to the culture medium (□), which blocks the Type 1 fimbriae; values were compared to untreated control medium (▪), within each cluster. Adherence of the bacteria was also modulated by pre-treatment of (E) epithelial or (F) stromal cells for 48 h with medium containing 5 ng/ml progesterone (□) or 3 pg/ml estradiol (▪); values were compared to untreated control medium (▪), within each cluster. Bars represent the mean + SEM of four experiments, * P<0.05, ** P<0.01, *** P<0.001.

Figure 4. Invasion of host cells by uterine E. coli.

Bacteria from MLST clusters 1 to 4 (n = 4 per cluster) were added to confluent (A) epithelial or (B) stromal cells, incubated for 1 to 4 h before addition of gentamicin for 2 h, and the number of adherent CFU measured; experiments were repeated on 4 separate occasions with at least 4 isolates from each cluster. Values were compared to cluster 1 at the corresponding time point. The number of (C) epithelial or (D) stromal cells present at the end of each experiment was determined. Bacteria (arrow) were visualised using the Syto 9 DNA stain in (E) epithelial and (F) stromal cells after 4 h but not after 1 h of incubation. Bacterial invasion of (G) epithelial or (H) stromal cells was inhibited by the addition of cytochalasin D (□) or colchicine (▪) to the culture media; values were compared to untreated control medium (▪), within each cluster. Bars represent mean + SEM of four experiments, * P<0.05, ** P<0.01.

Figure 5. Host cell response to uterine E. coli.

Prostaglandin E₂ and interleukin-8 (IL-8) concentrations accumulated over 24 h in the medium of endometrial epithelial (A, C) and stromal cells (B, D) treated with LPS. Cells were treated with control media or media containing 1 µg/ml LPS purified from E. coli O111:B4 or from bacteria isolated from the uterus of animals and classified into MLST cluster 1 to 4 (n = 3 isolates per cluster). Prostaglandin E₂ and IL-8 concentrations were determined by RIA and ELISA, respectively. Data are presented as the mean + SEM of at least three experiments. Values with different superscripts differ significantly, P<0.05.

Figure 6. Uterine-derived E. coli cause PID in mice.

Mice were infused with vehicle, live bacteria or LPS for 24 h and the uterus collected and processed for histology. Inflammation in the endometrium of animals treated with bacteria or LPS is evident in H&E sections (A–C) and following immunohistochemical staining for neutrophils (D–F) or macrophages (G–I). Bacteria invasion of the uterus is evident using FISH for mice infused with E. coli into the uterine lumen (J–L).

Figure 7. Endometrial cells respond to LPS and the response is TLR4 dependent.

Prostaglandin E₂ and CXCL1 concentrations accumulated over 24 h in the medium of endometrial epithelial (A, C) and stromal cells (B, D) isolated from C57BL/6 wild type (WT) or TLR4^−/− mice. Cells were treated with control media or media containing 1 µg/ml LPS purified from E. coli O111:B4 (O111 LPS) or from bacteria isolated from the uterus classified into MLST cluster 4 (PID LPS; n = 2 isolates). Prostaglandin E₂ and CXCL1 concentrations were determined by RIA and ELISA, respectively. Data are presented as the mean + SEM for three experiments. Values differ significantly between WT and TLR4^−/−, * P<0.05, ** P<0.01, *** P<0.001.