Ulcerative Colitis-associated E. coli pathobionts potentiate colitis in susceptible hosts

ABSTRACT Ulcerative colitis (UC) is a chronic inflammatory condition linked to intestinal microbial dysbiosis, including the expansion of E. coli strains related to extra-intestinal pathogenic E. coli. These “pathobionts” exhibit pathogenic properties, but their potential to promote UC is unclear due to the lack of relevant animal models. Here, we established a mouse model using a representative UC pathobiont strain (p19A), and mice lacking single immunoglobulin and toll-interleukin 1 receptor domain (SIGIRR), a deficiency increasing susceptibility to gut infections. Strain p19A was found to adhere to the cecal mucosa of Sigirr -/- mice, causing modest inflammation. Moreover, it dramatically worsened dextran sodium sulfate-induced colitis. This potentiation was attenuated using a p19A strain lacking α-hemolysin genes, or when we targeted pathobiont adherence using a p19A strain lacking the adhesin FimH, or following treatment with FimH antagonists. Thus, UC pathobionts adhere to the intestinal mucosa, and worsen the course of colitis in susceptible hosts.


Introduction
Inflammatory bowel diseases (IBD) are chronic inflammatory conditions traditionally divided into Crohn's disease (CD) and Ulcerative Colitis (UC). CD is a chronic, segmentally localized and penetrating granulomatous inflammatory disease that can affect any part of the gastrointestinal (GI) tract 1 . In contrast, UC is a relapsing, superficial mucosal inflammatory disease restricted to the colon that is characterized by bloody diarrhea during flares of the disease. The etiology of IBD is still unknown, but thought to reflect the convergence of environmental stimuli and genetic susceptibility factors that together promote chronic intestinal inflammation. A variety of IBD susceptibility genes have been identified, many of them encoding proteins that regulate innate immune control over enteric bacterial pathogens, such as NOD2, NALP3 and ATG16L1 2 . Correspondingly, numerous investigations suggest the characteristic intestinal inflammation seen in IBD patients occurs concurrently with, and is potentially exacerbated by dysbiotic changes in the intestinal microbiome 3 . In many studies, the loss of potentially beneficial microbes such as Faecalbacterium prausnitzii and Lactobacilli species, as well as a reduction in microbial diversity have been suggested as pathogenic steps toward the development of IBD.
Even so, the loss of beneficial microbes alone does not explain the complicated role played by the gut microbiota in IBD. For example, placebo-controlled studies have shown that antibiotic treatment can induce remission in some IBD patients 4,5 . Moreover, the overgrowth of specific Escherichia coli species has been suspected since the 1970's as the reason for relapses in some patients with IBD 6 . In 1988, Burke et al 7 showed a significantly higher frequency of adherent E. coli isolated from UC and CD patients undergoing disease relapse, as compared to healthy persons. Several additional studies have demonstrated an increased prevalence of E. coli with virulence properties in the GI tracts of IBD patients, especially within IBD patients undergoing disease relapse 8,9,10 , Recent studies indicate the involvement of adherent-invasive E. coli (AIEC) in the pathogenesis of CD 11,12,13,14, with demonstrations that AIEC can invade intestinal epithelial cells as well as persist within macrophages 15,16 . The development of several mouse models of AIEC infection has aided in defining the virulence properties of AIEC, as well as the host factors that control susceptibility to these microbes 17,18 .
Although AIEC is frequently seen in CD patients, it is uncommon in UC patients. Instead, diffusely adherent E. coli (DAEC) have been linked to UC 19 . Bacteriological analysis of biopsies and fecal samples from UC patients shows the increased prevalence of E. coli species belonging to the B2 phylogenetic group that harbor extra-intestinal pathogenic E. coli (ExPEC) genes 20, 21 , including αhemolysin genes, as well as the gene encoding the adhesin FimH. Notably, studies focusing on a representative UC-associated E. coli strain termed p19A found that upon phagocytosis, this hemolysin-expressing microbe triggered the death of dendritic cells as well as strongly stimulated the secretion of the proinflammatory cytokines TNFα, IL-6 and IL-23 22 . Studies of p19A interactions with the Caco-2 intestinal epithelial cell line have also shown that p19A induces α-hemolysin dependent damage to the tight junction protein occludin and thereby disrupts tight junctions in Caco-2 cells, increasing barrier permeability 20,23 . Despite these pathogenic properties, the potential for these E. coli pathobionts to promote the course of UC remains unclear due to the lack of relevant animal models.
Recently, we have made significant progress developing mouse models of enteric infection, using mice deficient in Single IgG IL-1 Related Receptor (SIGIRR). SIGIRR is highly expressed by intestinal epithelial cells, acting as a negative regulator of innate signaling through most Toll-like receptors (TLR) as well as IL-1 R 24 . The absence of SIGIRR results in an increased inflammatory tone within the GI tract under baseline conditions, as well as exaggerated inflammatory and antimicrobial responses to murine and human enteric bacterial pathogens 25,26 . Similarly, many IBD patients, even in remission, have a higher inflammatory tone in their GI tracts. Interestingly, upon Citrobacter rodentium infection, Sigirr -/-mice exhibited exaggerated antimicrobial responses that were largely ineffective against this pathogen, but instead depleted competing resident commensal microbes. We also found Sigirr -/-mice to be uniquely susceptible to infection by the clinically important foodborne pathogen Campylobacter jejuni, developing overt gastroenteritis upon infection 26 .
Based on these findings, we explored the potential to establish a mouse model of GI infection by the UC-associated E. coli strain p19A, as well as characterize the pathogenic features of p19A. Initial studies showed that mice orally gavaged with high and/or repeated doses of p19A suffered severe damage within the small and large intestine as well as systemic spread of the bacteria, leading to high mortality rates 27 . Interestingly, these responses were at least partially dependent on αhemolysin. In our current study, to develop a mouse model more representative of UC, mice were pretreated with vancomycin and then infected with a single moderate dose of p19A. Indeed, vancomycin pretreatment of wildtype C57BL/6 mice enabled p19A to persistently colonize the intestinal lumen, whereas the same treatment of Sigirr -/mice also led to p19A adherence to the cecal mucosal surface. While p19A infection caused modest cecal inflammation in non-DSS treated Sigirr -/-mice, it dramatically worsened the course of colitis in DSS treated Sigirr -/-mice, with p19A showing increased adherence to the inflamed intestinal mucosa. Notably, this adherence, and its ability to worsen colitis was lost with a p19A strain lacking the adhesin FimH, or when mice were given FimH antagonists to block p19A adhesion. Moreover, a p19A strain lacking α-hemolysin genes (∆hlyI∆hlyII) was attenuated in its ability to promote DSS colitis. Our findings thus provide evidence that UC associated E. coli strains can readily and persistently colonize the intestines of susceptible hosts, and significantly worsen the course of colitis in a manner dependent on specific virulence factors, including α-hemolysin and the type 1 fimbrial adhesin FimH.

Vancomycin pretreatment facilitates persistent p19A intestinal colonization of C57BL/6 mice
A variety of B2 phylogenetic group E. coli strains have been isolated from patients with either active or inactive UC 20 . Whether these E. coli strains play a pathogenic role in UC remains unclear. To address this, we examined the ability of p19A, one of the best-characterized human UC isolates 23 , to colonize the intestines of mice. Groups of C57BL/6 wildtype (WT) mice were left untreated, or pretreated with vancomycin for 6 hours, followed by infection with p19A. We then quantified the number of p19A in the stool by plating. As shown in Figure 1a), mice not receiving vancomycin showed only moderate shedding of p19A in the stool (10 5 colony forming units (CFU)/gram) at 1 day postinfection (dpi), but by 5 and 7 dpi, p19A was greatly reduced (~10 3 CFU/gram) or even cleared from the stool. By 14 dpi, p19A was cleared from the stool of all mice. In contrast, pretreatment with vancomycin dramatically increased p19A shedding to 10 9 CFU/gram of stool at 1 dpi, and similar levels (5 x 10 8 CFU/gram) were shed at 5 dpi. The shedding of p19A gradually decreased to an average of 10 6 CFU/gram by 7 dpi and to 10 5 CFU/gram of stool by 14 dpi. This level of shedding continued out to 60 dpi, and potentially beyond, showing that p19A can persistently colonize the intestines of mice following antibiotic treatment.
Sigirr -/-mice show increased susceptibility to p19A colonization of the GI tract.
Our group has previously shown that Sigirr-/mice display heightened susceptibility to several intestinal pathogens 25,26 . We tested whether these mice would also be susceptible to p19A infection. WT and Sigirr-/-mice were pretreated with vancomycin for 6 hours and then infected with p19A. At 5 dpi, the numbers of p19A recovered from the ceca and colonic lumen of Sigirr-/-mice were almost 10 fold higher than those of WT mice, although similar levels of p19A were seen in the colons of both groups of mice ( Figure 1b). We then examined the localization of p19A within the GI tract. Vancomycin-pretreated WT and Sigirr-/-mice were gavaged with a p19A derived luciferase-expressing strain (p19A-lux) for 5 days, followed by imaging of bioluminescence signals in excised intestinal tissues. Interestingly, strong luminescence signals were detected in the intact GI tracts (following euthanization) of both WT and Sigirr-/-mice (Figure 1c), but upon opening and vigorously washing away the luminal content from the intestinal tissues, the luminescence signals were largely lost from the ceca of WT mice, but remained strong in the ceca of Sigirr-/-mice. This suggests that the majority of p19A in WT mice are luminal, whereas p19A in Sigirr-/-mice may be adherent to the cecal mucosal surface. To confirm this, we infected vancomycin-pretreated WT, and Sigirr-/-mice with a p19A derived GFPexpressing strain (p19A-GFP), and collected cecal tissues for immunostaining. Antibodies against GFP, along with actin were used to visualize p19A-GFP bacteria and the intestinal epithelial surface, respectively. As shown in Figure 1d, GFP expressing bacteria were restricted to the cecal lumen of WT mice, whereas GFP positive bacteria were in close proximity to, or adherent to the cecal mucosal surface (indicated by arrows) of Sigirr −/mice.
We next examined if p19A colonization induced cecal inflammation in mice. While vancomycin-pretreated WT, and Sigirr-/-mice showed no inflammation in their ceca at baseline (Figure 1e), at day 5 post p19A infection, higher levels of inflammatory cell infiltration, increased submucosal edema and epithelial damage were seen in the ceca of Sigirr-/-mice as compared to WT mice (Figure 1e,f). Consistent with these microscopic changes, the mRNA levels of proinflammatory cytokines (Il-6, Tnf-α) in the cecal tissues of Sigirr-/-mice were double that of WT mice at 5 dpi ( Figure S1). These data indicate that Sigirr-/-mice are susceptible to p19A colonization of their intestinal mucosal surface, with p19A adhering to their intestinal epithelial cells and inducing modest inflammation in their ceca.

p19A infection worsens subsequent DSS-induced colitis in vancomycin treated Sigirr -/-mice
Since p19A colonized the intestinal mucosal surface of Sigirr-/-mice, causing modest pathology, we asked whether infection with p19A would have an effect on subsequent intestinal inflammation induced by other noxious stimuli, such as dextran sodium sulfate (DSS). Notably, the DSSinduced colitis model has been widely used as a mouse model of human UC 28 , mimicking many of the pathologic features seen in UC patient tissues. After testing a range of doses, we found that 2.5% DSS consistently caused a very mild colitis in Sigirr-/-mice in our animal facility. Vancomycin-pretreated Sigirr-/-mice were subsequently gavaged with the control E. coli DH10B (a derivative of commensal strain K-12) or p19A. One day after infection, mice were exposed to 2.5% DSS in their drinking water for another 4 days. Over the following days, body weights were recorded. As shown in Figure 2a, p19A +DSS treated mice lost significantly more body weight than DH10B+DSS treated mice, starting at 4 dpi. Consistent with this, the disease activity index (DAI) score (based on body weight loss, stool consistency and blood in the stool) of p19A +DSS treated mice was significantly higher than that of the DH10B+DSS treated groups at day 4 and 5 dpi (Figure 2b). When mice were euthanized at 5 dpi, a striking difference was noted in the large intestines of these mice. The cecal and colonic tissues of the DSS+ p19A treated mice were overtly shrunken and inflamed, and their colons were significantly shorter than those of mice given DH10B+DSS (Figure 2c,d).
Colonic shortening typically reflects severe mucosal inflammation. Thus, while H&E staining of cecal and colonic tissues of the DH10B+DSS treated mice revealed only mild inflammatory cell infiltration and epithelial cell damage, this was in stark contrast to the widespread inflammatory cell infiltration, severe edema, loss of crypt structures, and broad areas of deep ulceration seen in the p19A+DSS treated mice ( Figure  2e). Correspondingly, the cecal and colonic pathology in the p19A+ DSS mice was significantly greater than that seen in the DH10B + DSS mice (figure 2f). We also examined levels of fecal lipocalin 2 (Lcn-2), since it functions as a sensitive, noninvasive marker for intestinal inflammation 29 . Lcn-2 levels in the p19A+ DSS mice were almost two times those of the DH10B + DSS mice ( Figure  S2). We then examined if DSS-induced colitis affected the growth of p19A within the gut. Vancomycin-pretreated Sigirr-/-mice were infected with p19A-lux for 1 day, and exposed to DSS treatment for 4 days. In vivo imaging analysis revealed stronger bioluminescence signals emanating from DSS treated mice as compared to non-DSS treated mice (Figure 2g), and upon removal of the tissues, and repeated washing, stronger luminescence signals were detected in the ceca and colons of the DSS treated mice. Confirming this observation, when ceca and colons of these mice were collected for plating, the numbers of p19A bacteria recovered from the p19A+DSS treated mice was 2-4 log higher than that recovered from mice receiving p19A, but no DSS ( Figure  2h). Notably, despite the inflammation, p19A were still found to adhere to areas of intact intestinal mucosa, an observation confirmed through electron microscopy ( Figure S3). Collectively, these results demonstrate that prior infection with p19A aggravates DSS-induced colitis in vancomycin-pretreated Sigirr-/-mice, and this colitic response further promotes p19A growth in the gut.

UC-E. coli isolate p7 also aggravates subsequent DSS-induced colitis in Sigirr -/-mice.
To test whether p19A's ability to aggravate DSSinduced colitis was shared by other UC-associated pathobionts, we tested if prior infection with the UC isolate p7 intensified colitic responses induced by DSS. Notably, the pathobionts p7 and p19A both belong to the B2 phylogenetic group, but they display different sero-type profiles 20 . Vancomycin pretreated Sigirr-/-mice were infected with E. coli DH10B or p7 for 1 day, followed by exposure to DSS as above. As shown in Figure S4, p7+ DSS treated mice showed significantly worsened clinical symptoms as compared to DH10B+DSS treated mice, including greater body weight loss, higher DAI, and exaggerated tissue pathology. Thus, E. coli isolates from different UC patients are conserved in their ability to worsen DSS-induced colitis in Sigirr-/-mice.

p19A's ability to worsen DSS-induced colitis is largely dependent on α-hemolysin
Our previous studies showed that much of the damage caused by p19A and other UC pathobionts to cells in culture was attributable to the production of α-hemolysin 23 . To test whether the pro-colitic actions of p19A showed a similar basis, Sigirr-/mice were pretreated with vancomycin and then gavaged with either wildtype p19A, or the ∆hlyI∆hlyII mutant strain (a derivative of p19A that lacks hemolysin genes hlyI and hlyII). The ∆hlyI∆hlyII mutant has been previously shown to suffer a significant loss of hemolytic activity 23 . As outlined earlier, at 1 dpi, the infected mice were given 2.5% DSS in their drinking water. Mice infected with the ∆hlyI∆hlyII mutant showed significantly less body weight loss at 4 and 5 dpi, and a lower DAI score at 5 dpi (Figure 3a,b). Following euthanization, the mice infected with the ∆hlyI∆hlyII strain showed less macroscopic intestinal damage (larger ceca and longer colons), and suffered significantly less histological damage than mice receiving wildtype p19A (Figure 3c-e). While reduced pathology and inflammatory response was seen in ∆hlyI∆hlyII infected mice, the shedding of the ∆hlyI∆hlyII mutant and wildtype p19A in the stool was comparable ( Figure 3f). Thus, the E. coli pathobiont p19A worsens the course of DSS colitis through the production of α-hemolysin, a virulence factor previously linked to this microbe's ability to trigger inflammatory responses in vivo and damage intestinal epithelial barrier function in vitro 23, 30, 31 .

p19A's potentiating effects on colitis depend on the adhesin FimH
The ability of p19A to adhere to the intestinal mucosal surface of Sigirr-/-mice suggests that p19A may express certain adherence factor(s) mediating its interaction with host cells. Indeed, the genome of p19A (unpublished data) consists of many genes encoding adherence factors, including pili and fimbriae 32 . Of note, p19A contains fimH, a gene encoding a mannose-binding lectin named FimH. FimH is expressed by many other E. coli strains, including AIEC, and uropathogenic E. coli (UPEC), mediating their adherence to epithelial cells 33,34,35 . Comparing amino acid sequences of FimH from different E. coli strains, we found that p19A FimH was phylogenetically most similar to E. coli K-12 FimH, containing two mutations (V51A and R190H) ( Figure S5). To test whether FimH was a pathogenic factor of p19A, we generated an isogenic strain of p19A lacking fimH (∆fimH), and infected Sigirr-/-mice as outlined earlier. As shown in Figure 4a-e, compared to mice receiving wildtype p19A+DSS, ∆fimH+DSS treated mice showed significantly improved clinical symptoms, losing less body weight, displaying lower DAI scores, and suffering dramatically less macroscopic and histologic damage in their ceca and colons. Remarkably, the colitis developed in ∆fimH+DSS treated mice was similar to that seen in the control DH10B+DSS treated mice. To better understand how the mutation in fimH attenuates p19A pathogenicity, we compared p19A and ∆fimH burdens in infected mice. Although ∆fimH and wildtype p19A showed comparable levels of shedding in the stool at 1 dpi, by 5 dpi (4 days post-DSS), the levels of ∆fimH were dramatically reduced, to 4 logs lower (10 4 CFU/gram) than those of wildtype p19A (10 8 CFU/gram) ( Figure  4f). Immunostaining of infected cecal and colonic tissues showed that wildtype p19A was present in both the lumen and on the intestinal epithelial surface, whereas ∆fimH was rarely detected, and only in the intestinal lumen ( Figure 4g). These results suggest that FimH is not only critical for p19A to remain within the inflamed gut and be shed at high levels in the stool, but the mucosal adherence/lasting presence of p19A within the inflamed gut is also essential for p19A to aggravate DSS colitis.

FimH antagonists prevent p19A from aggravating DSS-induced colitis
FimH contains a carbohydrate recognition domain (CRD) that interacts with mannosylated proteins expressed on the surface of epithelial cells 36   while 7 CD-HM is an heptavalent glycoconjugate where the HM ligands are grafted onto a βcyclodextrin core. At 1 dpi, DSS was given to these mice as outlined earlier, and they were monitored for another 4 days. As shown in Figure  5a-e, mice receiving either antagonist showed significantly less weight loss, lower DAI scores, as well as less macroscopic and microscopic damage,  although 7 CD-HM appeared to be more effective than 1A-HM. To determine if these FimH antagonists affected the interaction of p19A with intestinal epithelial cells, cecal and colonic tissues were opened and washed extensively, followed by homogenization and plating. The numbers of p19A recovered from cecal and colonic tissues of antagonist-treated mice were significantly lower than those recovered from vehicle treated mice ( Figure  5f), suggesting that p19A were prevented from adhering to the intestinal mucosa in the presence of antagonists. Thus, FimH antagonists, by inhibiting the adherence of p19A to the intestinal mucosa, reduced p19A's ability to aggravate DSS-induced colitis.

Discussion
The etiology of IBD is complex, and appears to involve the interaction of genetic, environmental and immunologic factors. Also, luminal bacteria or their products play a critical role in the instigation and exacerbation of the chronic intestinal inflammation seen in IBD 10,11,41,42 . For instance, AIEC is associated with CD, whereas a variety of recently isolated B2 phylogenetic group E. coli strains are associated with UC 20,31 . Many in vitro and in vivo models have been developed to define how AIEC pathogenicity factors and the host response to this microbe act in concert to potentially contribute to the etiology of CD. In contrast, the potential contribution of B2 phylogroup E. coli strains to UC has received less attention, with their pathogenesis primarily studied in cell culture 23 . Robust animal models that utilize these UC associated E. coli strains in a relevant fashion are clearly needed. Here, we found that the clinically isolated UC pathobiont p19A was able to colonize the intestines of genetically susceptible mice (Sigirr-/-) following antibiotic treatment. While p19A caused only modest intestinal inflammation on its own, it was found to adhere to the intestinal mucosal surface and worsen subsequent DSS-induced colitis.
Using this robust animal model, we also characterized key pathogenicity factors of p19A involved in colitis development.
Our results indicate an increased susceptibility of Sigirr -/-mice (compared to C57BL/6 mice) to p19A infection following antibiotic treatment. This may not be surprising, considering that this mouse strain is also more susceptible to many other enteric pathogens, including C. rodentium, S. Typhimurium and C. jejuni 25,26 . Upon infection with these pathogens, Sigirr -/-mice display exaggerated inflammatory and/or antimicrobial responses that promote their expansion and colonization in the gut. Sigirr -/-mice infected with p19A also exhibited slightly higher inflammatory responses than WT mice, as well as modest gut inflammation. We reason that p19A causes only modest inflammation on its own because it does not express the typical pathogenicity factors used by intestinal bacterial pathogens to induce strong inflammatory responses, such as a type III secretion system and the capsular polysaccharides expressed by many Gram-negative bacterial pathogens 43,44 . From an evolutionary perspective, lacking the expression of these factors may benefit p19A, allowing this microbe to survive and thrive within the GI tract without excessively agitating its host.
Vancomycin pretreatment enabled p19A to heavily colonize the intestines of both C57BL/6 mice and Sigirr -/-mice, suggesting that the microbial dysbiosis caused by antibiotic treatment created an environment that favored the expansion of p19A in both mouse strains. Even so, immunostaining of intestinal tissues showed that p19A was able to adhere to the intestinal mucosa of Sigirr -/mice (but not C57BL/6 mice). The basis for this difference is unclear, but could reflect the greater inflammatory response in p19A+vancomycin treated Sigirr -/-mice (as compared to C57BL/6 mice). Similar to the report by Winter et al. 45 , the heightened inflammatory tone of Sigirr -/-mice may differentially induce p19A's virulence factors, allowing it to colonize the surface of the inflamed gut. It is also possible there are key factors expressed on the inflamed intestinal mucosa of Sigirr -/-mice that permit p19A binding. For example, AIEC infection has been shown to up-regulate the expression of the carcinoembryonic antigen related cell adhesion molecule (CEACAM6) by intestinal epithelial cells, thus facilitating its adherence to the ileal mucosa of CD patients 18 . The fact that antibiotic pretreatment alone was sufficient to increase p19A colonization levels in the gut suggests a potential for p19A to worsen already existing UC, or exacerbate disease relapses in the clinical setting. Consistent with our findings in animal studies, p19A and other similar E. coli pathobionts are often seen in UC patients with inactive, as well as active disease 20,31 .
While p19A infection alone caused only modest intestinal inflammation in Sigirr -/-mice, it dramatically potentiated subsequent DSS-induced inflammation. Moreover, p19A numbers bloomed during DSS-induced intestinal inflammation, as evidenced by the recovery of significantly higher numbers of p19A from the ceca and colons of p19A+DSS treated mice as compared to mice not receiving DSS. These data suggest that p19A, or other UC pathobionts could dwell at low numbers within the gut, but upon the onset of intestinal inflammation triggered by other noxious stimuli, take advantage of the inflammation and expand to aggravate disease. Interestingly, the pro-colitic effect of p19A was also seen with another B2 phylogenetic group E. coli strain termed p7. Whether other UC-associated E. coli strains show similar pro-colitic effects awaits further investigation.
The ability of p19A to potentiate colitis depends on key pathogenic factors. One of these is αhemolysin, a pore forming toxin that can cause cell lysis and induce tissue damage. Notably, αhemolysin-expressing E. coli strains are often found in UC biopsies 30,32 . We and others have shown that many E. coli strains (including p19A), through α-hemolysin production, disrupt tight junctions and cause focal lesions in cultured epithelial cells 23,46 . Correspondingly, we found that the ∆hlyI ∆hlyII mutant of p19A lost its ability to potentiate DSS-induced colitis. Using a mouse model that does not involve DSS treatment, Bucker et al. 30 have also demonstrated that E. coli α-hemolysin impairs intestinal barrier function, although this is through the induction of focal lesions in the colonic epithelium. Thus, both mouse models highlight the potential for p19A's αhemolysin to cause tissue damage and/or intensify ongoing inflammation and disease.
Previous studies have found that the FimH adhesin aids other E. coli strains in attaching to epithelial cells and adapting to different environmental conditions 47,48,49 . Building on these findings, our study identified FimH as an essential pathogenicity factor for p19A. The ∆fimH mutant (derived from wildtype p19A) failed to colonize the intestinal mucosal surface. Moreover, mice treated with ∆fimH+DSS showed almost no tissue pathology at 5 dpi. Similar to our findings, inactivation of fimH in various extraintestinal pathogenic E. coli strains (including UPEC and AIEC) decreases their colonization in the gut and/or reduces their ability to promote inflammation 34,50,51,52,53 .
We also saw a remarkable decrease in the fecal shedding of the ∆fimH mutant from infected Sigirr -/-mice over time. One possible explanation is that the ∆fimH mutant may be unable to attach to the intestinal mucosal surface, and is flushed out of the intestinal lumen by the flow of feces, water or mucus. Alternatively, since FimH is important for biofilm formation in many other bacteria 54,55 , the ∆fimH mutant may have been unable to form biofilms in the gut, and was thus unable to persistently colonize the inflamed intestine. Interestingly, FimHmediated adhesion of UPEC to the urothelial surface causes the activation of the nuclear factor-κB (NF-κB) signaling pathway, leading to the secretion of pro-inflammatory cytokines 56 . In future studies, it will be interesting to determine whether p19A also aggravates DSS-induced colitis through FimHdependent activation of NF-κB signaling.
While there are only two point mutations separating the FimH of p19A from that of the commensal E. coli K-12, we speculate that these mutations are sufficient to mediate p19A's ability to adhere to the intestinal mucosal surface of Sigirr -/-mice and worsen their course of colitis. Supporting this, point mutations in FimH also confer AIEC bacteria a higher ability to persist and induce intestinal inflammation in genetically susceptible transgenic mice 34 . Therefore, as proposed previously 34 , analyzing single nucleotide polymorphisms of fimH could potentially predict the pathogenicity of E. coli strains isolated from CD and UC patient biopsies.
To further characterize the role of FimH in p19A pathogenicity, we treated Sigirr -/-mice with two FimH antagonists, and found that they significantly improved the clinical symptoms induced by p19A +DSS treatment. The colonization by p19A of the intestinal mucosa was significantly reduced in mice treated with these antagonists. It is possible that these antagonists are able to inhibit the interaction of FimH with mannosylated proteins on the intestinal mucosa, but have no or minimal effect on the ability of p19A to form biofilms in vivo (since the fimbriae structure is still present). In addition to suppressing the pathogenicity of p19A in our colitis model, FimH antagonists have been shown to alleviate AIEC induced colitis 37 and deplete UPEC from infected mice 38 . Thus, FimH represents a very promising therapeutic target in the fight against intestinal infection/inflammation caused e1847976-12 by a wide range of E. coli strains, including UCassociated E. coli strains. The FimH antagonists targeting one type of E. coli could prove broadly effective, as seen for 1A-HM and 7 CD-HM, which also attenuated the pathogenicity of AIEC LF82 in a different mouse model of colitis 37 .
Taken together, we have developed a robust mouse model of colitis through the use of the UCassociated E. coli p19A. While it is conceivable that similar pathobionts may initiate UC in susceptible individuals, it appears more likely that they arise in the context of intestinal inflammation, exacerbating and promoting the chronicity of the disease, as well as potentially converging with other inflammatory stimuli to promote disease relapse. We further show the pro-colitic effects of p19A are mediated by at least two key factors -α-hemolysin and FimH. Selectively inhibiting FimH activity using antagonists prevented p19A from worsening colitic responses. These findings led us to propose a working model whereby p19A contributes to UC development in genetically susceptible hosts ( Figure S6). While p19A is isolated from UC patients, LF82 is isolated from CD patients. They both belong to the B2 phylogenetic group and use FimH to potentiate colitis. However, in contrast to p19A, LF82 does not exhibit hemolytic activity or cause tight junction disruption 23 . Future work will be needed to address other genetic similarities and differences between these two pathobionts. Nevertheless, our pre-clinical model should not only facilitate studying how bacterial-host interactions play a role in the pathogenesis of UC, but will also significantly advance the development of novel therapeutic strategies that target UC.

Bacterial strains
UC-associated E. coli B2 strains p19A and p7, and the ∆hlyI∆hlyII mutant (generated from p19A) lacking both hemolysin genes, have been described previously 20,23 . A p19A-derived ∆fimH mutant was generated by allelic exchange using the gene doctoring technique as previously described 57 . Briefly, primers GDupfimH and GDdwfimH (primer sequences are summarized in Table 1) were used to amplify a kanamycin cassette from the plasmid pDOC-K. The GDupfimH primer contains 50 bp of homology to the DNA sequence upstream of fimH and 20 bp of sequence (K-FWD) from pDOC-K. The GDdwfimH primer contains 50 bp of homology to the sequence of the terminal region of fimH and 18 bp (P-REV) of sequence on pDOC-K. The amplified kanamycin cassette was cloned into the plasmid pDOC-C, which was subsequently transformed into p19A carrying the helper plasmid pACBSCE encoding the λ-Red recombinase. To induce recombination, transformed bacteria were grown in Luria-Bertani (LB) broth containing 0.5% L-arabinose. Recombinant bacteria were further grown on LB agar containg 50 μg/ml kanamycin and 5% sucrose to select for the ∆fimH mutant. Correct allelic replacement of the wild type fimH leaving only a 39 bp scar of the fimH gene was verified by PCR. The functional loss of FimH by the ∆fimH mutant was also confirmed with a yeast agglutination assay as described by others 58 .
Since p19A is susceptible to many different antibiotics, it is difficult to selectively recover it from p19A-infected tissues. To circumvent this limitation, we inserted a chloramphenicol resistance marker on the chromosome of wildtype p19A. Briefly, a triple mating involving wildtype p19A, E. coli MFDλpir containing pMAC5 (a chloramphenicol-marked Tn7 delivery vector) 59,60 , and E. coli MFDλpir containing a helper plasmid pTNS2 61 was performed on LB agar containing diaminopimelic acid (DAP). Conjugants were selected by plating 24 h conjugation mixture onto LB agar containing chloramphenicol but lacking DAP, and screened for the proper Tn7 transposition as before 61 . The p19A strain carrying the chloramphenicol resistance marker in the right position of the chromosome was named p19A-Chl r . Using the same method, we also generated a p19A-lux strain expressing the the Photorhabdus luminescens lux operon 62 , and a p19A-GFP strain expressing GFP on the chromosome of p19A. The lux operon or gfp was expressed under the control of the promoter PLtetO 63,64 . Notably, p19A-Chl r , p19A-GFP, p19A-lux and the parental strain p19A showed similar potential to potentiate DSS colitis (data not shown). Note that all bacterial strains were grown from single colonies on LB plates, and cultured in 5 ml of LB broth without antibiotics, or with chloramphenicol (30 μg/ml), kanamycin (50 μg/ml), or DAP (0.3 mM) at 37°C overnight with shaking at 200 rpm. The next day, overnight cultures were subcultured at 1:250 dilution into 5 ml of LB at 37°C at 200 rpm until the OD 600 values reached 0.8 ~ 0.85, followed by centrifugation of 1 ml of the cultures at 4000 rpm for 5 min. Culture pellets were resuspended in 1 ml of sterile PBS, and 100 μl of the suspension was used for oral infection of mice as detailed below.

Mouse infection experiments
The C57BL/6 mice (originally from Charles River Laboratories) and the single immunoglobulin and toll-interleukin 1 receptor (TIR) domain (Sigirr) -/-mice used in these studies were bred and maintained under specific pathogen-free conditions at BC Children's Hospital Research Institute. All mice were housed in a temperature-controlled (22 ± 2°C) animal facility with a 12-h light-dark cycle. Male mice at 6-10 weeks of age were orally gavaged with 0.1 ml of a 50 mg/ml vancomycin solution suspended in PBS (5 mg per mouse). Six hours later, each mouse was infected by oral gavage with approximately 7 × 10 7 CFU of wildtype p19A, p19A-Chl r , p19A-GFP, p19A-lux, ∆hlyI∆hlyII, ∆fimH, or E. coli DH10B. One day after infection, mice were exposed to 2.5% dextran sulfate sodium (DSS) in their drinking water. For the FimH antagonist experiment, 1A-HM and 7 CD-HM were orally administered in a volume of 0.1 ml of PBS at a dose of 10 mg/kg, at 2 h, day 1, 2, and 3 post-infection. Disease activity index (DAI) scores were recorded according to the following criteria (body weight, stool consistency and blood in the stool): Score 0 -no weight loss, hard stool and no blood detection; Score 1 -less than 10% body weight loss, hard stool and no blood in the stool; Score 2-10 − 15% body weight loss, loose stool and fecal occult blood; Score 3 -15 -20% body weight loss, loose stool with gross blood; Score 4 -> 20% body weight loss, diarrhea and rectal bleeding. A maximum score was 12. Mice were euthanized at day 4 or 5 post-infection, and tissues collected for further analysis.

Ethics statement
All procedures involving the care and handling of the mice were performed according to protocol number A15-0206, approved by the University of British Columbia's Animal Care Committee and in direct accordance with the Canadian Council of Animal Care (CCAC) guidelines. Mice were monitored daily for mortality and morbidity throughout their infection and euthanized if they showed signs of extreme distress or more than 20% body weight loss.

Bacterial counts
To enumerate bacteria within large intestinal tissues and lumen, the ceca and colons were opened longitudinally, and the luminal contents were collected in a 2.0 ml microtube. Tissues were washed in PBS extensively, before being placed in the microtube. Tissue and luminal contents were homogenized in a MixerMill 301 bead miller (Retsch) for a total of 6 mins at 30 Hz at room temperature. Stool samples were similarly processed to count bacteria in the feces. Homogenized tissues, luminal contents or fecal samples were plated onto LB agar containing chloramphenicol (30 μg/ml) for p19A WT strains, or kanamycin (50 μg/ml) for ∆hlyI∆hlyII and ∆fimH mutant strains. After culturing at 37°C overnight, bacterial counts were recorded.

In vivo imaging
In vivo imaging was performed using the Ami-x platform (Spectral Instruments Imaging, AZ). Grayscale reference images taken under low illumination were collected and overlaid with images capturing the emission of photons from the luxexpressing p19A-lux using AMIView software. Live mice were anesthetized and images were taken. Thereafter, the mice were euthanized, followed by imaging of the intestinal tissues (before and after washing with PBS).

Electron microscopy (EM) fixation and imaging
Sigirr-/-mice were pre-infected with p19A for 1 day, and then exposed to 2.5% DSS in their drinking water. After 2 days of DSS exposure, cecal and colonic tissues of p19A infected mice were removed and processed. EM studies of tissue samples and overnight p19A cultures (37°C in LB) were performed as previously described 66 .

Real time qPCR analysis
Total RNA from cecal and colonic tissues was extracted using the Qiagen RNeasy kit following the manufacturer's instructions, and quantified using a NanoDrop Spectrophotometer (ND1000). 1 μg of RNA was reverse-transcribed using a Qiagen OmniscriptRT kit (Qiagen), followed by dilution of the cDNA at 1:5 in RNase/DNasefree H 2 O. The qPCR was carried out using a Bio-Rad CFX connect Real-time PCR detection system, with the specificity for each of the PCR reactions confirmed by melting point analysis. Quantitation was performed using CFX Maestro TM software (Bio-Rad). The expression of genes was normalized to that of Rplp0. Primer sequences and reaction conditions for all genes analyzed are summarized in Table 1.