Secretion of Alpha-Hemolysin by Escherichia coli Disrupts Tight Junctions in Ulcerative Colitis Patients

Crohn’s disease (CD) and ulcerative colitis (UC) are two different forms of chronic inflammatory bowel disease (IBD), the etiology of which is still unknown. CD and UC are distinguished by their clinical manifestations and inflammatory profiles.1 UC is a chronic inflammatory disorder of the colorectal mucosa, whereas CD is a chronic, segmentally localized granulomatous disease of the gastro-intestinal tract CD may even affect non-intestinal tissue such as lymph nodes and skin. Clinical practice has seen that in both diseases chronicity is interrupted by acute flares, bloody diarrhea, relapses, and remission. IBD can appear at any age, however, most often in the third decade of life.2 The highest reported prevalence values for IBD are in Europe (UC, 505 per 100,000 persons; CD, 322 per 100,000 persons).3 Genome-wide association studies in IBD have identified genetic polymorphisms contributing to susceptibility to IBD. Many of these gene polymorphisms are associated with pathways involved in intestinal homeostasis, linking host genetics to deregulated host responses to the microbiota.4 The concordance rate among monozygotic twins was 6.3% for UC and 58.3% for CD.5 This clearly indicates a role of genetic factors in CD, but also indicates an important role for environmental factors, particularly in UC. An abnormal microbiota composition and decreased complexity of the gut microbial ecosystem (commonly referred to as dysbiosis) are common features in patients with CD or UC.6 These observations have fueled efforts to identify opportunistic gut pathogens (or pathobionts) that may have a role in the pathogenesis of IBD. Escherichia coli pathobionts exhibiting pathogen-like behaviors are more frequently cultured from IBD patients with active disease due to their selective growth advantage in inflammatory conditions.7 Moreover, adherence of the B2 phylotype E. coli to human intestinal epithelial cells is mediated through the type 1 pili interaction with mannosylated carcinoembryonic antigen-related cell adhesion molecule 6 (CEACAM6). Interestingly, CEACAM6 expression by cultured intestinal cells was shown previously to be upregulated after treatment with interferon γ and tumor necrosis factor-α.8 These findings indicate that inflammatory conditions in the gut support E. coli colonization via increased CEACAM6 expression and offer an explanation for their more frequent isolation from patients with active disease. Among the Proteobacteria, adherent invasive variants of the B2 phylogroup E. coli (AIEC) have been proposed to have a role in the pathophysiology of IBD,9 owing to their capacity to adhere to intestinal epithelial cells, to invade intestinal epithelial cells via a macropinocytosis-like process, and to survive and replicate intracellularly in epithelial cells and macrophages.10 Others have, likewise, found increased numbers of B2 phylogroup E. coli isolated from IBD patients.11 Petersen et al.12 showed that E. coli isolates from fecal samples of primarily UC patients with active disease frequently belong to the B2 phylogenetic group and harbor genes commonly associated with extra-intestinal pathogenic E. coli (ExPEC) causing urinary tract infection and meningitis.13 Recently, a hemolysin (Hly) producing E. coli strain was shown to induce localized defects in epithelial integrity colonic cell monolayers and rat colon tissue ex vivo. Additionally, wild-type (WT) and colitis susceptible IL-10−/− mice colonized with an HlyA-expressing E. coli had elevated inflammation scores and an increased epithelial permeability compared with mice colonized with the HlyA-deficient mutant. Furthermore, qPCR analysis revealed that lesions (focal leaks) in mucosal samples from the human colon were associated with 10-fold higher levels of hlyA DNA, suggesting that Hly-expressing E. coli have a role in the pathology of intestinal inflammation in IBD.14 The aim of this study was to extend the above observations to isolates of B2 phylogroup E. coli from IBD patients by testing their effects on permeability, tight junction stability, and viability of human intestinal cell epithelial monolayers cultured in vitro. For comparison, we also tested the effect of prototype AIEC strain LF82 and the probiotic E. coli Nissle on permeability and viability of polarized human Caco-2 cells. As some strains of B2 phylogroup E. coli isolated from UC patients also possess a gene encoding cytotoxic necrotizing factor type 1 (cnf1), we investigated the role of cnf1 and hlyA in causing epithelial damage by the construction and testing of genetic mutants in cellular assays. Detailed information regarding extent of disease and current medication among the included patients has previously been described,12 neither controls nor patients had received antibiotics within the last 2 months before inclusion and all patients had an established diagnosis of IBD before inclusion in our study. Fecal samples were cultured at Statens Serum Institut (SSI): bacteriological analysis, E. coli phenotypic characterization, determination of phylogenetic group, and ExPEC virulence gene detection were performed as described previously in Petersen et al.12 E. coli clinical isolates p7, p10, p13, p19A, p22, p25, p26, p27, and p32; healthy control E. coli isolates C2, C4, and C6 were characterized by PCR for virulence genes (data not shown) in this study. The probiotic E. coli Nissle 1917 and the pathogenic E. coli LF82 (ref. 17) were used as a negative and positive control, respectively. Patient characteristics and diseases association and location are described in Table 1. Caco-2 cells (between passages 55 and 60) were seeded at a density of 2.6 × 105 cells/cm2 in a 24-transwell system containing Tissue Culture-treated filter (0.4 μm pore size, BD Biosciences Falcon type # 353494, Erembodegem, Belgium) and grown for 16 days until they differentiated into polarized monolayers. After 14 days, the transepithelial electric resistance (TER) reached 600–800 Ohms/cm2 (Volt/Ohm meter; World Precision Instruments, Sarasota, FL). For bacterial co-incubation experiments, the medium was removed by aspiration from the Transwell filters, and the filters were inserted in the cellZscope apparatus (Nanoanalytics, Münster, DE). Cell-culture medium without antibiotics was then added to the upper chambers (450 μl) and lower chambers (800 μl), and the apparatus was placed in a humidified incubator at 37°C containing 5% CO2/95% O2 atmosphere for 2 h before the addition of bacteria. Bacteria were grown overnight in LB (Luria broth) media at 37 °C, centrifuged, resuspended in DMEM, and added to the upper chambers (filter) in the cellZscope at a multiplicity of infection (MOI) of 10. TER measurements were recorded continuously for up to 24 h, and TER values were normalized to the initial TER value (100%) and absolute TER is mean of four independent measurements. As a control, TER was measured for uninfected Caco-2 cell monolayers (controls in figures). Three independent experiments were performed for p19A WT and its mutant strains. Hemolytic titration assays were performed with bacteria and bacterial culture supernatants in order to investigate the hemolytic activity of clinical isolate E. coli p19A WT and the hly and cnf1 mutants. The hemolytic activity was only completely abolished in the double mutant lacking hly clusters I and II (P<0.05) (Figure 2). Deletion of hly cluster II only partially decreased hemolytic activity, compared with the WT, suggesting that hly cluster I does contribute to the overall hemolytic activity of the WT strain, despite the fact that no reduction in hemolysis was observed in the hly cluster I mutant. RT-PCR was used to quantify the relative amounts of the hly transcript in E. coli p19A WT and the different hly deletion mutants, using rpoA transcripts as an internal control. The relative expression of hly was twofold higher in the WT p19A than in the ΔhlyI mutant and fourfold higher than in the ΔhlyII mutant. As expected, only the double ΔhlyI, II mutant of p19A lacked hly gene expression (Figure 3). The epithelial cell layer is an essential constituent of the gut and a highly specialized interface between the host and its environment. Desmosomes, adherence junctions, and tight junctions hold the cells of the intestinal epithelial layer together. Tight junctions are important in controlling paracellular permeability to ions and small molecules and preventing translocation of luminal antigens and bacteria into the lamina propria.24 In this paper, we demonstrated that IBD-associated E. coli strains from UC patients who produce α-hemolysin cause disruption of epithelial tight junctions of intestinal cell monolayers, leading to the loss of TER. Three of five UC-associated E. coli strains (p7, p19A, and p22) isolated from patients with active UC induced a rapid loss of TER at low MOI without any loss of cell viability. The IBD-associated strains causing loss of epithelial integrity were all of the phylotype B2 and consistent with previous reports showing an increased abundance of the phylotype B2 E. coli in UC and CD patients with active disease.11, 12, 25, 26 The role of E. coli pathobionts in the pathophysiology in IBD was attributed to their capability to adhere and invade epithelial cells and replicate in macrophages, and the most well-studied prototype strain is LF82. In contrast to p19A, strain LF82 does not cause rapid dissolution of epithelial tight junctions, clearly indicating that the phylotype B2 of UC-associated strains differs markedly in pathogenic mechanisms. The type 1 fimbriae of AIEC were shown to bind to CEACAM6, which is expressed at higher levels in inflamed intestinal epithelial cells of IBD patients.27 Our UC-associated E. coli p19A strain has the same capacity as LF82 to adhere to epithelial cells (data not shown). All the UC-associated E. coli strains that caused loss of tight junctions in epithelial cell monolayers were hemolytic. Four types of hemolysin have been demonstrated in E. coli: alpha-hemolysin (HlyA), plasmid- and phage-carried enterohemolysin (EhxA and HlyA) and silent hemolysin (SheA); EhxA and HlyA belong to the RTX (repeats in toxin) related family, which lyse erythrocytes from different mammalian species.28, 29, 30 It is known that a number of E. coli pathotypes, i.e., urinary tract pathogenic E. coli, enteropathogenic E. coli, and enterotoxigenic E. coli are all able to produce α-hemolysin.20 The E. coli α-hemolysin is known to be able to lyse erythrocytes through binding to the surface protein glycoporin,31, 32, 33 but also other cell types including leukocytic cells, bladder, and renal tubular cells in a dose-dependent manner.25, 26, 27, 28, 29 Lysis of immune cells is greatly influenced by the presence of cell receptors CD11a and CD18, which are expressed on B and T cells, as well as neutrophils monocytes and dendritic cells.30, 31 The role of HlyA in tight junction disruption was further investigated in E. coli strain p19A, which possessed two hlyA clusters as previously reported for some isolates of uropathogenic E. coli belonging to phylotype B2.34 We showed that both hlyA gene clusters in p19A contributed to the damaging effects on the epithelial integrity, suggesting that intestinal E. coli strains possessing more than one hlyA locus may have increased pathological consequences in intestinal inflammation. Although our UC-associated strains did not induce epithelial cell apoptosis, an hly-expressing uropathogenic E. coli was previously shown to cause localized regions of apoptosis in HT29/B6 cell monolayers. The difference between these findings and our results may be due to the use of a higher MOI than in our study, the use of different strains, or the amount of hlyA expressed.35 Our study showed that around 50% of phylotype B2 E. coli isolated from UC patients can adhere to epithelial cells and disrupt epithelial tight junctions via an HlyA-dependent mechanism, provides strong evidence that this is an important novel pathogenic mechanism in UC; and distinct of AIEC LF82 in CD. Lesions in tight junctions of intestinal epithelium from IBD patients with active disease have been associated with a reduction in several tight junction proteins including claudin 1 and 4, occludin and tricellulin,36 and the synthetic octapeptide (AT1001), which prevents the opening of tight junctions, improves colitis in susceptible IL-10−/− mice.37 Further evidence for the importance of HlyA in the epidemiology of IBD comes from a previous study showing that an HlyA-producing strain of E. coli but not an HlyA-deficient mutant was a potentiator of inflammatory activity in the colon of susceptible IL-10−/− mice and monocolonized germ-free mice due to its effects on the epithelial barrier function.14 During active UC and high inflammation and increased CEACAM6 expression, binding of specific E. coli is facilitated. This study is a mandate for further investigation of epithelial barrier disruption in other UC cohorts and geographic locations. Preliminary evidence from genomic sequencing suggests that some E. coli strains carry large conjugative plasmids, suggesting that lateral gene transfer of hly loci could contribute to the spread of pathogenic traits. A recent meta-study including 10 randomized trials from CD patients and 9 randomized trials from UC patients yielded an odds ratio of 2.17 (95% confidence interval, 1.54–3.05) in favor of antibiotic therapy.38 These results suggest that antibiotics improve clinical outcomes in patients with IBD. Another meta-study published in 2011 by Khan et al.39 concluded that antibiotic therapy may induce remission in active CD and UC, although the diverse number of antibiotics tested means the data are difficult to interpret. This systematic review proposed further trials of antibiotic therapy in IBD. Approaches for combating bacteria that adversely affect the barrier function (e.g., HlyA-expressing E. coli) might provide new treatment options for IBD. This might include antibiotic therapy, vaccination or competition by probiotic bacteria lacking HlyA and other virulence factors that can cause harm to the host. Guarantor of the article: Karen Angeliki Krogfelt, PhD. Specific author contributions: Participated in the design of the study: Karen Angeliki Krogfelt, Andreas Munk Petersen, Hengameh Chloé Mirsepasi-Lauridsen, Zhengyu Du, Carsten Struve, Jurgen Karczewski, and Jerry M. Wells; drafted the manuscript: Hengameh Chloé Mirsepasi-Lauridsen, Andreas Munk Petersen, Jerry M. Wells, Carsten Struve, and Karen Angeliki Krogfelt; responsible for the experimental setting: Hengameh Chloé Mirsepasi-Lauridsen, Zhengyu Du, Carsten Struve, Godefroid Charbon, and Jurgen Karczewski. All authors have read and approved the final manuscript. Financial support: None. Potential competing interests: None. We thank the Met-Vet-Net Association for a travel grant to H.C.M.-L. We also thank post doc Mette Elena Skindersø, SSI, Nico and Anja Taverne-Thiele, University of Wageningen, for their help with cell assays and for helpful discussions. G.C., Copenhagen University, was supported by the Lundbeck foundation. We also thank the laboratory staff at Zodiac (J. Wells’ laboratory) for their help and support during H.C.M.-L. stay in their laboratory. Marian Jørgensen is thanked for proof reading the manuscript.