Oxygen and contact with human intestinal epithelium independently stimulate virulence gene expression in enteroaggregative Escherichia coli

Abstract Enteroaggregative Escherichia coli (EAEC) are important intestinal pathogens causing acute and persistent diarrhoeal illness worldwide. Although many putative EAEC virulence factors have been identified, their association with pathogenesis remains unclear. As environmental cues can modulate bacterial virulence, we investigated the effect of oxygen and human intestinal epithelium on EAEC virulence gene expression to determine the involvement of respective gene products in intestinal colonisation and pathogenesis. Using in vitro organ culture of human intestinal biopsies, we established the colonic epithelium as the major colonisation site of EAEC strains 042 and 17‐2. We subsequently optimised a vertical diffusion chamber system with polarised T84 colon carcinoma cells for EAEC infection and showed that oxygen induced expression of the global regulator AggR, aggregative adherence fimbriae, E. coli common pilus, EAST‐1 toxin, and dispersin in EAEC strain 042 but not in 17‐2. Furthermore, the presence of T84 epithelia stimulated additional expression of the mucinase Pic and the toxins HlyE and Pet. This induction was dependent on physical host cell contact and did not require AggR. Overall, these findings suggest that EAEC virulence in the human gut is modulated by environmental signals including oxygen and the intestinal epithelium.

Shiga toxin-producing EAEC strains are emerging, causing potentially fatal systemic disease that cannot be treated with antibiotics. The severity of this has been underlined by a large EAEC outbreak in Germany in 2011 that resulted in 1,000 hospitalisations and 50 deaths (Bielaszewska et al., 2011).
Despite their considerable impact on human health, the mechanisms of how EAEC cause disease remain unknown. This is partly due to a lack of suitable animal models that reflects their specificity for the human host (Philipson, Bassaganya-Riera, & Hontecillas, 2013). In addition, EAEC are a heterogenous group, and not all strains cause human disease (Jenkins, Chart, Willshaw, Cheasty, & Tompkins, 2007;Nataro et al., 1995). The differences in pathogenicity among EAEC isolates can probably be attributed to their traditional classification based on "stacked brick"-like aggregative adherence to HEp-2 cells, which does not necessarily reflect their ability to cause human disease (Nataro et al., 1987). Nevertheless, research on EAEC so far indicates that bacterial adherence to intestinal epithelium, biofilm formation, release of toxins, and mucosal inflammation likely contribute to pathogenesis and diarrhoea (Estrada-Garcia & Navarro-Garcia, 2012).
Protein interaction studies have identified several AAF host receptors including cytokeratin 8, MUC1, and extracellular matrix proteins (Boll et al., 2017;Izquierdo et al., 2014). Similar to AAF, dispersin is plasmid-encoded and regulated by AggR and has been linked to adherence and biofilm formation (Sheikh et al., 2002). Structural studies suggest that dispersin binding to outer membrane lipopolysaccharide masks its negative charge and allows positively charged adhesins such as AAF to bind more distant sites, thereby promoting dispersal of adherent bacteria (Velarde et al., 2007). Although AAF are strongly linked to aggregative adherence, many EAEC isolates have no AAF allele, and the phenotype is believed to be multifactorial (Jønsson et al., 2015). For example, the E. coli common pilus (ECP) found in many E. coli pathotypes (Rendón et al., 2007) has been implicated in aggregative adherence, especially in AAF-negative strains (Avelino et al., 2010). In addition to adhesins, EAEC produce several toxins, which are important for the induction of diarrhoeal symptoms. The enterotoxin EAST-1 has similarities to the heat-stable enterotoxin STa of enterotoxigenic E. coli and is proposed to function in a comparable way via interference of cGMP signalling and dysregulation of anion secretion (Ménard, Lussier, Lepine, Paiva de Sousa, & Dubreuil, 2004). Haemolysin E (HlyE) is a pore-forming toxin mediating cytolytic and cytopathic effects in cultured human cells. As it is also found in nonpathogenic bacteria, the role of HlyE in EAEC pathogenesis remains unclear (Navarro-Garcia & Elias, 2011). Another enterotoxin associated with EAEC virulence is the plasmid-encoded SPATE Pet that degrades the structural protein spectrin, leading to cytoskeletal disruption in epithelial cells (Boisen, Ruiz-Perez, Scheutz, Krogfelt, & Nataro, 2009;Canizalez-Roman & Navarro-García, 2003). In addition, some EAEC strains express the SPATE Pic that cleaves mucins and complement proteins and stimulates mucus hypersecretion in the gut (Henderson, Czeczulin, Eslava, Noriega, & Nataro, 1999;Navarro-Garcia et al., 2010).
Despite the identification of these and other putative virulence factors, genotypic studies have failed to consistently associate a single gene or combination of genes with EAEC pathogenicity (Estrada-Garcia & Navarro-Garcia, 2012 For this study, we employed the well-characterized EAEC prototype strains 17-2 and 042, which have been used in human volunteer studies (Nataro et al., 1995). To select the most suitable intestinal epithelial cell line for the VDC infection model, we first evaluated EAEC adherence in in vitro organ culture (IVOC) of human mucosal biopsies from different parts of the intestine (proximal small intestine to distal colon).
Although both EAEC strains demonstrated aggregative adherence to tissue from the transverse and sigmoid colon after 7 hr of incubation, only few bacteria bound to biopsies from the terminal ileum (Figure 1), and no bacteria were detected on duodenal samples (Table 1). Aggregative adherence to colonic mucosa was specific for EAEC as no epithelium-bound bacteria were detected on biopsy samples incubated with E. coli K12 (Figure 1). Aggregative adherence and colonisation of colonic epithelium was confirmed by infecting human colon carcinoma-derived T84 cells with EAEC for up to 5 hr (Figure 2), and this cell line was subsequently used for further studies.

| Establishment of the MA VDC system
In our previous studies, we have shown that T84 cells grown on Snapwell supports form polarised and well-differentiated epithelial monolayers that can be maintained in the MA VDC for at least 6 hr without loss of barrier function or cell viability (Schüller & Phillips, 2010 MA levels (5%) after 4 hr, a period of 3 hr was chosen for subsequent bacterial gene expression analysis. Differences in bacterial respiration status between AE and MA conditions were confirmed by analysis of terminal oxidase expression. In E. coli, AE respiration in oxygen-rich environments is mediated by the low affinity cytochrome bo 3 oxidase complex (cyoABCDE), whereas the high affinity cytochrome bd oxidase (cydAB) is utilised at low-oxygen tensions (Cotter, Chepuri, Gennis, & Gunsalus, 1990;Morris & Schmidt, 2013). As shown in Figure 3c

| Oxygen induces virulence gene expression in EAEC 042
To determine the influence of oxygen concentrations on bacterial virulence gene expression, polarised T84 cells were infected with strains 17-2 or 042 for 3 hr under AE or MA conditions. After harvesting nonadherent bacteria from the apical media, RNA was extracted and transcription of selected virulence genes was determined by qPCR (Table 2)

| Host cell contact enhances EAEC virulence gene expression
We next characterized the influence of host cells on virulence gene expression. Polarised T84 intestinal epithelia were infected in the VDC as described above. After 3 hr, nonadherent (planktonic) and adherent bacteria were harvested for RNA extraction, and gene transcription was analysed by qPCR. In strain 042, all selected genes except aggR and ecpA were significantly induced in adherent versus To determine if physical contact between bacteria and host cells was required for virulence gene induction, T84 cells were seeded out in 12 well plates. EAEC were either added directly to the cells or prevented from direct cell contact by addition to a Transwell insert with a porous membrane enabling exchange of soluble mediators.
For comparison, bacteria were incubated in Transwell supports in 12 well plates without T84 cells. After bacterial RNA extraction, qPCR was performed for a subset of virulence genes, and gene expression was compared between (a) EAEC in Transwells with and without T84 cells (no bacteria-host cell contact) and (b) adherent and nonadherent EAEC in well plates without Transwells (bacteria-host cell contact). As shown in Figure 5d, expression of all virulence genes except pet was increased in adherent versus nonadherent bacteria, which paralleled our findings in the VDC system. In contrast, no induction in gene expression was observed when bacteria were separated from the T84 epithelium by a Transwell insert, in which case, even a significant reduction in expression of aafA (042) and aap (17-2) was detected ( Figure 5d).

| Dependence of virulence gene induction on AggR regulation
To determine the dependency of oxygen-and contact-induced viru-   reporter constructs for 5 hr. GFP expression in adherent and nonadherent bacteria was determined by fluorescence intensity and normalised to colony forming units (CFU). Fluorescence is displayed as relative fluorescence units (RFU) per 10 4 CFU (n = 3 in triplicate). (c) Polarised T84 cells were infected with strain 042 for 5 or 7 hr under AE conditions. Expression of dispersin (Aap) in adherent (A) and nonadherent bacteria (NA) was determined by Western blot analysis. Bacterial lysates of an isogenic aggR mutant (ΔaggR) were included as negative control. Expression of GroEL was used to normalise total protein amounts. Band intensities were quantified with ImageJ, and protein expression is indicated as fold change in adherent versus nonadherent bacteria (n = 3). (d) T84 cells were grown in 12 well plates, and EAEC were added directly to the cells or prevented from direct cell contact by insertion of a porous Transwell insert. After 3 hr, expression of selected virulence genes was quantified by qPCR and is expressed as fold change in EAEC in Transwell inserts with and without T84 cells (no contact) or in adherent versus nonadherent EAEC in plates without Transwells (contact; n = 3 in duplicate). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 Garcia, 2003;Dudley, Thomson, Parkhill, Morin, & Nataro, 2006).
T84 colon carcinoma cells were chosen for the establishment of the VDC system as they exhibit structural similarity to intestinal crypt cells and form highly polarised columnar epithelia (Madara, Stafford, Dharmsathaphorn, & Carlson, 1987). In addition, previous studies have shown pAA-dependent aggregative adherence of EAEC 042 to T84 but not to Caco-2 cells (Nataro et al., 1996). Oxygen availability has been recognised as an important environmental signal for the modulation of virulence in enteropathogenic bacteria (Marteyn, Scorza, Sansonetti, & Tang, 2011). Salmonella typhimurium demonstrates increased host cell adherence and invasion when grown under low-oxygen tension (Lee & Falkow, 1990). In addition, the FNR transcriptional regulator involved in sensing low-oxygen levels is required for full Salmonella virulence in mice and modifies expression of a type III secretion system (T3SS) required for pathogenesis (Fink et al., 2007). Similarly, oxygen modulates T3SS expression and subsequent adherence of EHEC to intestinal epithelium (Schüller & Phillips, 2010). In Shigella flexneri, lack of oxygen in the gut lumen enhances expression of the T3SS but suppresses secretion of virulence proteins resulting in their accumulation inside bacteria. As S. flexneri approaches the mucosal surface, oxygen released by the epithelium triggers targeted T3S and subsequent bacterial invasion (Marteyn et al., 2010).
Here, we demonstrate an oxygen-dependent induction of genes associated with adherence (aafA, aap, and ecpA), which might prime the bacteria for host cell binding as they approach the intestinal epithelium. The corresponding upregulation of aggR, aap, and aafA suggests that modulation of the aggR regulon is linked to oxygen sensing. This effect is independent on the presence of T84 cells and therefore not mediated by hypoxia-related changes in epithelial cell Transcription of selected virulence genes in adherent and nonadherent EAEC was determined by qPCR and is indicated as fold change in adherent versus nonadherent bacteria (n = 4 in triplicate). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 function, which have been shown to influence infection by enteropathogenic bacteria (Zeitouni et al., 2016). Notably, induction of adherence gene expression was only evident in strain 042 but not 17-2. This might be related to strain-specific differences in gene regulation as an earlier study showed different temperature-dependent AAF expression in strains 042 and 17-2 (Hinthong et al., 2015).
In contrast, adhesion to host cells enhanced expression of not only colonisation-associated virulence genes but also those encoding toxins and SPATEs (astA, hlyE, pet, and pic). Stimulation of gene expression in adherent EAEC was largely independent of oxygen levels and evident in both strains, although the extent of induction of specific genes differed between 042 and 17-2. Regulation of virulence by sensing chemical cues has been widely reported in EHEC (Barnett Foster, 2013), and previous studies on EAEC have shown enhanced pet expression in nutrient-rich versus minimal media (Betancourt-Sanchez & Navarro-Garcia, 2009). In our study, however, physical contact with the epithelium rather than host-secreted soluble compounds were responsible for gene induction in adherent EAEC.
Notably, this was more pronounced in Mutant strains 042 ΔaggR and 042 aggR::pBAD30 have been described previously (Sheikh et al., 2002). Strain DFB042TC was constructed by sequentially disrupting the cat and tetA genes in EAEC 042, which encode for chloramphenicol and tetracycline resistance, respectively.
The cat gene was disrupted by introducing an internal stop codon into cat, using the suicide plasmid pCVD442 (Donnenberg & Kaper, 1991).
The tetA gene was disrupted using gene doctoring methodology (Lee et al., 2009)

| Cell culture and infection
The

| Vertical diffusion chamber
Infections in the VDC system were performed as described previously (Tran et al., 2014). Briefly, 5 × 10 5 T84 cells were seeded on collagen-  based on evaluation of four E. coli reference genes ( Figure S1). Ct values for genes of interest were normalised using the geometric mean Ct of the two reference genes. Fold expression levels in treated samples were calculated relative to matched non-treated controls using the formula 2 −ΔΔCt .

| Construction of GFP reporter strains and analysis of promoter activity
The promoter fragments aafD100 and aap500 were amplified upstream of aafD and aap, respectively, from EAEC 042 as described previously (Yasir et al., 2018). PCR fragments were cloned into the low copy number GFP reporter plasmid pRW400 via EcoRI and HindIII restriction sites (Alsharif et al., 2015), and constructs were verified by and imaged with a FluorChem E Imager (ProteinSimple). Densitometric analysis of band intensities was performed using ImageJ software (https://imagej.nih.gov/ij/).

| Statistics
Statistical analysis was performed using GraphPad Prism software (version 5.04). Student's paired t-test was used to determine differences between two groups. One-way or two-way ANOVA with Tukey's multiple comparisons test was used for multiple groups. A P value of <0.05 was considered significant.