Colonization of the Human Gut by E. coli and Colorectal
Cancer Risk
Mathilde Bonnet1,2, Emmanuel Buc1,2,3, Pierre Sauvanet1,2,3, Claude Darcha3, Damien Dubois1,2,3,
Bruno Pereira3, Pierre Dechelotte3, Richard Bonnet1,2,3, Denis Pezet1,2,3, and Arlette Darfeuille-Michaud1,2,3
Abstract
Purpose: The intestinal microbiota is potentially involved in the development of colorectal carcinoma via various mechanisms. Escherichia coli are commensal bacteria of the human gut microbiota, but some pathogenic strains have acquired the ability to induce chronic inflammation and/or produce toxins, such as cyclomodulin, which could participate in the carcinogenesis process. Here, we analyzed the E. coli population associated with mucosa of patients with colon cancer in relation to clinicopathologic char-acteristics. We assessed carcinogenic properties of a colon cancer–associated E. coli strain in multiple intestinal neoplasia (Min) mice.
Experimental design: Mucosa-associated or internalized E. coli were quantified and characterized from tumors and mucosa of patients with colon cancer and the healthy mucosa of diverticulosis controls. Min mice were inoculated with a colon cancer–associated E. coli strain (11G5). The number of colonic polyps was evaluated at 7 weeks after infection.
Results: An increased level of mucosa-associated and internalized E. coli was observed in the tumors compared with normal tissue. A relationship between poor prognostic factors for colon cancer (tumor– node–metastasis stage) and colonization of mucosa by E. coli was observed. Pathogenic cyclomodulin-positive E. coli strains were more prevalent on mucosa of patients with stages III/IV than those with stage I colon cancer. Proliferative index and E. coli colonization level of the mucosa distant from the tumor significantly correlated. Min mice infected with the E. coli strain 11G5 displayed a marked increase in the number of visible colonic polyps compared with controls.
Conclusion: These findings support that pathogenic E. coli could be a cofactor in pathogenesis of colorectal cancer. Clin Cancer Res; 20(4); 859–67.2013 AACR.
Introduction
Colorectal cancer is the third most commonly diag-nosed cancer worldwide and is responsible for more than 600,000 deaths every year. Because of its high incidence and mortality rate, colorectal cancer is a major public health problem. Survival and risk of recurrence are con-ditioned by the extent of the tumor, stratified according to the prognostic classification of the American Joint
Com-mittee on Cancer (AJCC; ref. 1). When the tumor is confined to the colon (stages I and II), surgery is curative with a 5-year survival rate up to 80%; however, in the case of dissemination (nodal involvement, metastatic disease) the prognosis is dramatically reduced (2). Thus, the development of new diagnostic tools for the early detec-tion of lesions is attractive.
Colorectal carcinogenesis was first defined as a classic adenoma–carcinoma sequence by Fearon and Vogelstein in 1990 and several risk factors were identified to favor this tumoral development (3). Among them, pathogens seem to play a major role. Indeed, approximately 20% of the global cancer burden can be linked to infectious agents (4, 5). Bacteria and their related products could participate in initiation or progression of sporadic colon cancer by driving a variety of mechanisms, including the induction of proinflammatory and procarcinogenic pathways in epithe-lial cells, the production of genotoxins and reactive oxygen species, and the conversion of procarcinogenic dietary factors into carcinogens (6–8). In colorectal cancer, various bacteria have been associated with carcinogenesis, includ-ing Streptococcus bovis, Enterococcus spp., Helicobacter pylori, the enterotoxigenic Bacteroides fragilis, and various
Authors' Affiliations: 1Clermont Universite, UMR 1071Inserm/Universite d'Auvergne;2
Institut National de la Recherche Agronomique (INRA), USC-2018; and3
Centre Hospitalier Universitaire, Clermont-Ferrand, France Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
M. Bonnet and E. Buc contributed equally to this work.
Corresponding Author: Mathilde Bonnet, Clermont Universite, UMR 1071Inserm/Universite d'Auvergne, Centre Biomedical de Recherche et Valorisation, 28 Place Henri Dunant, INRA, USC-2018, 63005 Clermont-Ferrand, France. Phone: 33-473178376; Fax: 33-473178371; E-mail: mathilde.bonnet@udamail.fr
doi: 10.1158/1078-0432.CCR-13-1343
pathogenic Escherichia coli (7–11). Despite the fact that E. coli is a commensal bacteria of the human microbiota and represents the most common cultivable, gram-negative, aero–anaerobic bacteria, various studies have demon-strated a clear link between mucosa-adherent E. coli and colorectal cancer (8, 10, 11). E. coli are divided into four phylogenic groups (A, B1, B2, and D) according to the acquisition of factors of virulence. E. coli from phy-logroups A and B1 are generally not pathogenic, whereas phylogroups B2 and D are involved in intestinal and extra-intestinal diseases. Interestingly, some strains of E. coli from phylogroup B2 are associated with Crohn disease, a chronic inflammatory bowel disease known to be a risk factor for colorectal cancer (12, 13). Moreover, two studies showed that associated and mucosa-internalized E. coli were observed more often in patients with colorectal cancer than in controls, supporting the central role of these bacteria in the development of colorectal cancer (10, 11).
Pathogenic E. coli strains synthesize various virulence factors, including several toxins called cyclomodulins such as cytolethal distending toxins (CDT), cytotoxic necrotizing factor (CNF), cycle inhibiting factor, and colibactin (6). Cyclomodulins are genotoxic and/or modulate cell-cycle progression, proliferation, cell differentiation, and apopto-sis (14–19). In particular, colibactin, a hybrid polyketide-nonribosomal peptide encoded by the pks genomic island, has genotoxic properties causing DNA double-strand breaks and chromosomal instability in human eukaryotic cells (18, 19). With regard to the cyclomodulin-producing E. coli, a recent study revealed an increased prevalence of cyclomo-dulin-producing B2 E. coli in colon tumor biopsies, suggest-ing a possible role of such pathogenic E. coli in colon carcinogenesis (20).
The aim of the present study is to compare the bacterial colonization of the colonic mucosa in patients screened for colorectal cancer with noncolon cancer controls. Given that E. coli represent the majority of the cultivable bacteria in the colon and are positively selected for during chronic inflam-mation (21), we restricted our analysis to these bacteria. We evaluated two parameters of mucosa-associated and inter-nalized E. coli in the normal mucosa and tumors of patients with colon cancer compared with the normal mucosa of patients receiving an operation due to diverticulosis, a nontumoral intestinal pathology. We analyzed the rela-tionship between the levels of mucosa-associated E. coli, including genotypic characterization of the isolated strains (phylogroup and search for cyclomodulin-encoding genes), and clinicopathologic characteristics. Finally, we assessed the carcinogenic properties of an E. coli strain colonizing the tumor and the mucosa in a patient with colon cancer, using the multiple intestinal neoplasia (Min) mice animal model.
Materials and Methods
Patients and resection specimen
Patients with resectable colon cancer (n ¼ 50) and uncomplicated diverticulosis (noncancer controls; n ¼ 33) were prospectively included in the Surgical Digestive Unit of Clermont-Ferrand (France) between March 2007 and July 2010. All patients were adult volunteers and signed informed consent. Patients with diverticulosis were operated on to prevent any inflammatory recurrence at least 10 weeks after an episode of diverticulitis. Exclu-sion criteria included neo-adjuvant chemotherapy, histo-ry of previous colonic resection, emergency surgehisto-ry, and use of antibiotics within 4 weeks before surgery. The clinical data were collected from medical files and are summarized in Tables 1 and 2. The sex ratio (M/F) was 1.22 and 0.74 for patients with colon cancer and diver-ticulosis, respectively. The age range was 35 to 95 years for cancer patients (median age, 70 years) and 34 to 81 years for controls (median age, 58 years). Of note, the diver-ticulosis group comprised younger subjects than the colorectal cancer group (Table 1).
Surgery was performed either by laparotomy or laparos-copy. None of the patients received mechanical bowel preparation before the surgery. All patients received cefox-itin (2 g i.v.) at the time of incision. A specimen of mac-roscopically normal mucosa (absence of diverticules) was collected from the patients with diverticulosis and a non-necrotic fragment from the peripheral areas of the tumor was collected from the patients with colon cancer. In par-allel, fresh normal mucosa samples adjacent to the tumor (10 cm from the tumor) were collected from the patients with colon cancer.
Pathologic analysis
After colonic resection, fresh specimens were transported to the pathology laboratory, fixed in buffered 4% formalin, embedded in paraffin, cut into 5-mm slices, and stained with hematoxylin/eosin/safranin. For the patients with colon cancer, tumors were staged according to the seventh Translational Relevance
The gut microbiota and dysbiosis have recently been associated with colorectal cancer. We observed here a significant relationship between the colonization of mucosa by pathogenic Escherichia coli and poor prognos-tic factors for colorectal cancer, such as tumor–node– metastasis stage, suggesting that this colonization by E. coli could be an additional prognostic factor of poor outcome in patients. In addition, a strong association was observed between the proliferative index of epithelial cells and the colonization of the mucosa by E. coli, suggesting a role for these bacteria in regenerative epithelial cell growth and tumor progression. We dem-onstrated that infection with a pathogenic colon cancer– associated E. coli strain accelerates tumor development in multiple intestinal neoplasia mice, thereby providing evidence of the strain’s procarcinogenic properties. Our findings show that mucosa colonization by pathogenic E. coli could be a crucial factor in colorectal carcinogenesis and serve as a new prognostic marker.
AJCC/Union for International Cancer Control (UICC) tumor–node–metastasis classification (TNM stage; ref. 1). The acronym is indicative of the following criteria: (T) tumor size and nearby tissue invasion, (N) regional lymph node involvement, and (M) distant metastasis. The degree of tumor differentiation (poor, moderate, or high), peri-neural invasion, vascular embols, the mucinous colloid component, and the presence of necrosis were also assessed. In patients with diverticulosis, the extent of mucosal inflam-mation (congestion of the mucosa or the submucosa, ulcerations, polynuclear infiltration, or abscesses) and pres-ence of polyps or tumors were assessed.
Immunohistochemical staining of the human Ki-67 anti-gen was performed on the colonic mucosa sections from patients with colon cancer. Five-micrometer sections of paraffin-embedded, healthy mucosa were labeled with the anti–Ki-67 MIB-1 antibody (Dako) and Ultra View Detec-tion Kit (Ventana) on Benchmark XT stainer (Ventana). Sections were counterstained with Harris’s hematoxylin.
For each patient, the proliferation index was calculated as the percentage of Ki-67–positive epithelial cells in propor-tion to the total number of cells in a minimum of 10 entire crypt columns as previously reported (22). Ki-67 expression was then classified as a low (Ki-67–positive cells10%) or high (Ki-67–positive cells>10%) proliferative index. Microbiologic analyses
Specimens (tumor, normal mucosa) were immediately processed for the analysis of mucosa-associated and inter-nalized E. coli colonization as previously described (20). The genotypic characteristics (phylogroup typing and pres-ence of cyclomodulin-encoded genes) of these E. coli strains were assessed using PCR, and cyclomodulin expression was determined by assessing a specific cytopathic effect on cultured epithelial HeLa cells (20).
Electron microscopy
Human epithelial cell (T84) monolayers were infected with the colon cancer–associated E. coli strain 11G5 in vitro for 3 hours at a multiplicity of infection (MOI) of 50 bacteria per cell. The cells were then fixed in 1.6% glutaraldehyde for 1 hour, washed in cacodylate buffer (0.2 mol/L, pH 7.2), and postfixed in 1% OsO4 for 1 hour. After washing, the samples were dehydrated in graded alcohol and embed-ded in epoxy resin. Thin sections (60 nm) cut with a diamond knife in an ultramicrotome were stained with uranyl acetate and lead citrate and observed using a Hitachi H7650 TEM (Hitachi High Technologies Corporation). Sample preparations and observations were performed in the Centre Imagerie Cellulaire Sante (CICS) platform. Animal model
Studies were performed using 5- to 6-week-old C57BL/ 6J-ApcMin/þ(Min mice; The Jackson Laboratory) or wild-type (WT) C57BL/6J female mice in accordance with the French and European Economic Community guidelines (86-60, EEC) for care of laboratory animals. Studies were approved by the French Regional Ethical Animal Use Committee (No. CE-2912). All mice were housed in specific pathogen-free conditions at the animal care facil-ity of the Universite d’Auvergne (Clermont-Ferrand, France). To enhance E. coli strain colonization, we admin-istered streptomycin (2.5 g/L) for 3 days before inocula-tions of peroral bacteria (1 108bacteria in PBS) or PBS alone (controls).
Two E. coli strains were tested: the ampicillin- and kana-mycin-resistant E. coli strain 11G5 isolated from a patient with colon cancer as the colon cancer–associated strain and the commensal rifampicin-resistant E. coli strain K-12 MG1655 (laboratory stock) as the nonpathogenic control. We periodically quantified fecal bacterial colonization as colony-forming units (CFU) per stool as previously described (23). Fifty days after bacterial inoculation, the mice were sacrificed. The colons were then removed from the caecum to the rectum, flushed in PBS, and splayed longitudinally. Then, the tumors were counted. The colons were swiss-rolled from the distal to proximal end and fixed
Table 1. Clinical data from patients with colon cancer and diverticulosis in the search for mucosa- or tumor-associated and internalized E. coli Patients with colon cancer Patients with diverticulosis Patients,n 50 33 Age 70.5 11.4 58.0 11.9b Sex, male/female 28/22 13/20 Presence of colon-associatedE. coli, n (%)
Mucosa Yes 45 (90) 29 (88) No 5 (10) 4 (12) Tumora Yes 41 (93)
No 3 (7)
Presence of colon-internalizedE. coli, n (%) Mucosa Yes 27 (54) 16 (48)
No 23 (46) 17 (52) Tumora Yes 38 (86)c
No 6 (14) Presence of B2 phylotypeE. coli strains
Yes 31 (65)d 13 (39)
No 17 (35) 20 (61) Presence of cyclomodulin-positiveE. coli strains
Yes 26 (60)e 6 (26) No 17 (40) 17 (74)
a
Six are undetermined.
b
P ¼ 0.03.
cP < 0.0002; tumor versus mucosa of patients with colon
cancer or diverticulosis.
d
P ¼ 0.04; in both the mucosa and tumors of patients with colon cancer versus diverticulosis.
e
P ¼ 0.01; in both the mucosa and tumors of patients with colon cancer versus diverticulosis.
overnight in 10% formalin. Paraffin-embedded sections were cut into 5mm sections, and the tissue sections were prepared for hematoxylin–phloxin staining and routine pathologic analysis.
Statistical analysis
The bacterial colony counts among multiple disease populations were compared using the Kruskal–Wallis non-parametric analysis followed by the Dunn posttest. The Fisher exact test orx2analysis was used for comparisons of the proportion of patient samples positive for E. coli across multiple disease populations. Unpaired Student t tests and Mann–Whitney tests were used for comparisons of 2 groups where appropriate (normality verified by the Shapiro–Wilk test and homoscedasticity by the Fisher– Snedecor test). All tests were performed using Graph Pad Prism 5 STATA (StataCorp) or R software (http://cran.r-project.org/). We considered P values of<0.05 as statisti-cally significant.
Results
Patients and pathologic data
Fifty patients with adenocarcinoma (22 proximal and 28 distal samples) and 33 with diverticulosis were prospec-tively included. Most patients with colon cancer had tumors invading the peritoneal serosa (T3 stage, 60%) and negative lymph nodes (N0, 62%; Supplementary Table SA). Tumors were most often classified as moderate or well-differentiated (42% and 54%, respectively) adenocarcinoma as well as lacking perinervous encasement (91%), vascular embols
(82%), mucinous component (84%), or intratumoral necrosis (64%; Supplementary Table SA). Acute (n¼ 5) or chronic (n¼ 6) inflammation was observed in 11 patients with diverticulosis. None of the patients with diverticulosis displayed signs of neoplastic transformation or adenoma-tous polyps.
Mucosa-associated and internalizedE. coli in colon cancer and noncancerous tissues
Mucosa-associated E. coli were detected in 88%, 90%, and 93% of diverticulosis mucosa, colon cancer mucosa, and colon cancer tumor specimens, respectively (Table 1). A significant increase in mucosa-associated E. coli was observed in colon cancer tumors compared with colon cancer and diverticulosis mucosa (P ¼ 0.01 and P < 0.001, respectively; Fig. 1A). Mucosa-internalized E. coli were detected in 48%, 54%, and 86% of diverticulosis, colon cancer mucosa, and colon cancer tumor specimens, respectively (P < 0.0002). In addition, the number of internalized E. coli was significantly increased in colon cancer tumors compared with colon cancer and diverticu-losis mucosa (P< 0.001; Fig. 1B).
Colonization and clinical data
Colonization by associated and internalized E. coli was assessed according to the patients’ clinical data (Supple-mentary Table SA). The sex and the age of the patients with diverticulosis and colon cancer did not influence the col-onization by associated and internalized E. coli (data not shown).
Table 2. Characteristics ofE. coli strains isolated from the colon cancer samples according to TNM status TNM status
I II III/IV
Number of patientsa,n 14 15 19
Presence ofE. coli, n (% of patients) No 3 (21%) 1 (7%) 1 (5%) Yes 11 (79%) 14 (93%) 18 (95%) Phylogroup,n (% of patients) A 6 (43%) 2 (13%) 6 (32%)
B1 5 (36%) 4 (27%) 4 (21%)
B2 6 (43%) 12 (80%)c 13 (68%)c
D 5 (36%) 4 (27%) 5 (26%)
Cyclomodulin-positive strainb,n (% of patients) þ 5 (45%) 9 (64%)c 12 (67%)c
6 (55%) 5 (36%) 6 (33%)
pks-positive strain, n (% of patients) þ 4 (36%) 8 (57%)d 9 (50%)d
7 (64%) 6 (43%) 9 (50%)
a
Undetermined for 2 patients (one patient with stage I tumor and other with stage III tumor).
bThe presence of genes coding for the cyclomodulins CNF, CDT, CIF, and colibactin (
pks island) was analyzed.
c
P < 0.05.
d
P < 0.05.
Colonization was associated with the TNM stage and each parameter separately (T, N). Given the low number of patients with metastasis (n¼ 3), statistical analysis was not performed on this category. Colonization by associated E.
coli was significantly increased in the normal mucosa distant from the tumor at advanced stages (stages III/IV) compared with stage I (P< 0.01; Fig. 2A). No significant variation in internalized bacteria was observed in relation to TNM stages
Figure 1. Mucosa-associated and internalizedE. coli are significantly increased in colon cancer tumors. Quantification of colon-associated (A) and internalized (B)E. coli in human colonic mucosa of patients with diverticulosis and colon cancer as well as the tumors of patients with colon cancer.
,P < 0.05. 108 106 104 CFU/g of tissue 102 100 108 106 104 CFU/g of tissue 102 100 I II III/IV T 1/2
A
B
C
D
T 3/4 T 1/2 N+ N– T 3/4 108 106 104 CFU/g of tissue 102 100 108 106 104 CFU/g of tissue 102 100 Figure 2. Significant relationshipbetween mucosal colonization by E. coli and TNM stage classification or its parameters T and N. A, mucosa-associatedE. coli according to the TNM stage (from the TNM classification of AJCC/ UICC). B and C, mucosa-associatedE. coli according to T (B) or N status (C). D, mucosa-internalizedE. coli according to T status.,P < 0.05.
(data not shown). When each parameter of the TNM clas-sification was considered, colonization by associated E. coli was significantly increased in nontumor mucosa of patients with locally advanced tumors (T3 and T4; P¼ 0.004) and lymphatic spreading (Nþ; P¼ 0.03; Fig. 2B and C). With regard to associated E. coli, the level of internalized bacteria was increased in nontumor mucosa of patients with advanced tumors (T3 and T4) compared with noninvasive tumor (T1 and T2; P¼ 0.045; Fig. 2D). In the tumor, no significant variation in the colonization by either associated or internalized E. coli was observed in relation to TNM classification or parameters. For other clinical data (differ-entiation status, perinervous encasement, vascular inva-sion, mucinous component, and intratumoral necrosis), no significant association was observed with associated or internalized bacteria colonization in either tumor or nor-mal mucosa. In addition, we measured the percentage of the Ki-67 positive cells in the normal mucosa of patients with colon cancer (Fig. 3). The proliferative index of the mucosa did not correlate with the TNM characteristics of the corre-sponding tumors (Supplementary Fig. SA). However, the
proliferative index significantly correlated with the levels of both mucosa-associated and internalized E. coli (P< 0.02 and P< 0.04, respectively; Fig. 3).
We then extended the study by searching for a correlation between the presence and characteristics of the E. coli strains with regards to TNM classification. Of note, the frozen subcultures of mucosa-associated bacteria from two patients, one with TNM stage I and the other with stage III, were not viable. No E. coli was isolated in 21%, 7%, and 5% of patients with stage I (3 of 14), II (1 of 15), and III/IV (1 of 18) tumors, respectively (Table 2). In most cases, similar clones of E. coli strains were found in the normal mucosa and tumor tissue of patients with colon cancer and most strains were from the B2 phylogroup. Interestingly, the proportion of mucosa and tumor colonization by B2 phylogroup E.coli was significantly increased in patients with stage II (80%) and III/IV disease (68%) compared with stage I (43%; P< 0.05; cf. Table 2). TNM stage was not associated with other phylogroups. As previously reported for tumor tissue (20), E. coli strains harboring cyclomodulin-encoding genes (cdt, cif, and/or cnf) and/or the pks island were significantly more fre-quently observed in normal mucosa of patients with colon cancer (26/43) than diverticulosis (6 of 23; 60% vs. 26%, respectively; P ¼ 0.01). Of great interest, the proportion of mucosa and tumor colonization by cyclo-modulin-positive E. coli strains was significantly higher in TNM stage II (64%) and III/IV (67%) patients than stage I patients (45%; P< 0.05; cf. Table 2). The presence of B2 phylogroup or cyclomodulin-positive E. coli did not influence the overall level of E. coli colonization in the mucosa or tumors (Supplementary Figs. SB and SC). Colon cancer–associatedE. coli 11G5 stimulates colon tumor development in Min mice
The E. coli strain 11G5, which exhibited high level col-onization in the healthy mucosa and tumor tissue of a patient with an advanced stage tumor (T4) and liver metas-tasis, was selected as a representative colon cancer–associ-ated E. coli strain. The strain belongs to phylogroup B2 and possesses the pks island encoding colibactin. Electron microscopy analysis of infected T84 intestinal cells indicat-ed that E. coli 11G5 induces the elongation of eukaryotic cell membranes at bacterial adhesion sites. In addition, bacteria were internalized within endocytic vacuoles (Fig. 4A). Infec-tion of Min mice with E. coli 11G5 resulted in high-level colonization of the gut as indicated by the number of bacteria in the feces (4.95 109CFUs per gram feces) 2 days after infection (Fig. 4B). Similar levels of colonization were measured in the stools of WT and Min mice (P¼ 0.79). In contrast, significantly reduced bacterial levels (1.69 107 CFUs per gram feces) were detected in stools of mice infected with E. coli K-12 (P< 0.001; Fig. 4C). At the time of sacrifice (50 days after infection), colonic polyps were detected in 92% of Min mice infected with E. coli 11G5, 57% of Min mice with K-12 E. coli, and 50% of uninfected Min mice (P< 0.05). The histologic analysis indicated that all polyps were adenocarcinomas (Fig. 4D) and no sign of obvious inflammation in the 11G5-colonized mice was
Figure 3. Significant relationship between colonization of the mucosa by E. coli and the mucosal proliferative index. A, representative Ki-67 immunostainings from the normal colonic mucosa of patients with colon cancer with a low (left) or high (right) proliferative index. B and C, mucosa-associated (B) and mucosa-internalized (C)E. coli according to the proliferative index.
detected. Infection experiments with strain 11G5 in WT mice did not lead to any neoplastic changes (data not shown). Markedly significant increases in the number and in the size of tumors were observed in the 11G5-colonized Min mice (Fig. 4E and F).
Discussion
Bacterial and viral infections have long been established as important factors in the etiology of several human cancers (4, 5). More recently, the gut microbiota and dysbiosis have been associated with colon cancers. This article presents evidence for the presence of pathogenic E. coli strains in the mucosa and tumors of patients with colon cancer that could play a role in colorectal carcinogenesis. We observed that
mucosa-adherent E. coli were not restricted to the tumor site of carcinoma as previously reported (8, 10, 24). We dem-onstrated that colon cancer tumors have increased numbers of associated E. coli than colon cancer and diverticulosis mucosa. It could be argued that local modifications of the mucosal properties induced by the tumor, such as loss of the mucosal barrier and epithelial cell polarity as well as expres-sion of abnormal surface antigens in patients with cancer, promote increased adhesion and internalization of bacteria in the tumor tissue. Although colon cancer–specific surface antigens remain uncharacterized, we can hypothesize that the carcino-embryonic antigen cell adhesion-related mole-cule 6 (CEACAM6), which is upregulated in patients with colorectal cancer (25), could be involved. Indeed, we have
Figure 4. The colon cancer– associatedE. coli strain 11G5 persists in intestinal epithelial cells and stimulates colon tumor development in Min mice. A, ultrastructural analysis of T84 intestinal epithelial cells infected with theE. coli strain 11G5 (MOI ¼ 50 per cell) for 3 hours. Analysis by transmission electron microscopy shows internalization of 11G5 bacteria located within endocytic vacuoles (a and b) and control, noninfected T84 cells (c). B, level of colonization measured in the stools of mice infected with the 11G5 colon cancer–strain or K12 commensal bacteria 2 days postinfection (p.i.). C, the level of colonization by the 11G5 strain measured in the stools of WT and Min mice. D, histologic analysis of colonic polyps observed in K12- or 11G5-infected mice. E, the number of visible colonic tumors in mice infected with the colon cancerE. coli strain 11G5, the K12-commensalE. coli strain, or control PBS. F, the size of tumors in infected mice.,P < 0.05.
previously reported, in Crohn disease that type I pili are among the bacterial virulence factors potentially implicated in increased E. coli adhesion based on binding to CEA-CAM6, which is abnormally expressed in mucosa of these patients (26). Despite noting no significant variation in the total number of E. coli between the normal mucosa of patients with colon cancer and diverticulosis, we observed changes in the E. coli populations. We previously showed that pathogenic cyclomodulin-positive E. coli strains belonging to B2 phylogroup were significantly more prev-alent in patients with colon cancer tumors than in the mucosal samples of diverticulosis (20). We demonstrated here that pathogenic E. coli strains were also detected on normal mucosa of patients with colon cancer. Our findings are consistent with results by Arthur and colleagues in a smaller cohort of patients with colon cancer (8). This finding prompted us to explore the effect of colon can-cer–associated E. coli infection in the Min mice model for which the microbiota has been reported to play a role in carcinogenesis (27). The representative pathogenic pks-positive E. coli strain (11G5) significantly induced tumorigenicity with an increase in the size and number of adenocarcinoma tumor. The genotoxic effects of colibactin requires bacteria–host cell contact (18, 19), and it has been reported that mRNA expression of the pks island occurs in the bacteria in closest contact with the epithelial cells (28). We observed similar levels of colonization with E. coli 11G5 in Min and WT mice, indicating that tumor development induced by infection with pks-positive bacteria requires a first hit, such as an apc mutation which is consistent with results previously observed in the interleukin (IL)-10/ murine model (8, 29, 30). In parallel to these results observed in mice, a strong association between the levels of E. coli colonization in human normal mucosa and the proliferative index (Ki-67 immunostaining) was noted, suggesting that these bacteria play a role in regenerative epithelial cell growth and tumor progression which is previously demonstrated in murine models (24, 27). We also observed a relationship between the levels of mucosa-associated E. coli colonization and the tumor cell dissem-ination process (axillary lymph node invasion and invasive tumor status) and a higher prevalence of the pathogenic strains in aggressive form of cancer. Given that abnormal bacterial colonization was previously reported on the mucosa of patients with adenoma (10, 24), all these data on human mucosa suggest that pathogenic E. coli could promote the development of an aggressive form of carci-noma at the early steps of enterocyte transformation by cyclomodulin genotoxic agent effect, such as colibactin, and/or the induction of chronic inflammatory reactions.
We report, for the first time, a statistically significant relationship between high levels of mucosa-associated E.
coli and poor colorectal carcinoma prognostic factors, such as TNM stages. The TNM staging system is still the most widely used in patient selection for adjuvant chemotherapy after curative resection. At present, it is not possible to accurately differentiate between patients with a favorable or poor prognosis in stages II and III colon cancer (31). Various clinical factors have been associated with a higher risk of recurrence in stage II and III colon cancer, including tumor obstruction and/or perforation, the number of lymph nodes identified in the surgical specimen, T4 stage, lymphovascular invasion, undifferentiated histology, and lymph node micrometastasis (32–36). None of these mar-kers are able to clearly separate patients into distinct prog-nostic groups. Although other microbes likely participate in the progression of colon cancer, our findings demonstrate that mucosal colonization by pathogenic E. coli could serve as a new and crucial prognostic factor of poor outcome. Large multicenter prospective clinical studies are needed to evaluate whether pathogenic E. coli colonization in the mucosa is a prognostic marker.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: M. Bonnet, E. Buc, R. Bonnet, A. Darfeuille-Michaud
Development of methodology: M. Bonnet, P. Sauvanet, P. Dechelotte, R. Bonnet
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Bonnet, E. Buc, P. Sauvanet, D. Dubois, P. Dechelotte, D. Pezet
Analysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): M. Bonnet, E. Buc, P. Sauvanet, B. Pereira, A. Darfeuille-Michaud
Writing, review, and/or revision of the manuscript: M. Bonnet, E. Buc, P. Sauvanet, B. Pereira, R. Bonnet, A. Darfeuille-Michaud
Administrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): M. Bonnet, E. Buc, P. Sauvanet, D. Pezet
Study supervision: D. Pezet, A. Darfeuille-Michaud Histologic analysis: C. Darcha
Acknowledgments
The authors thank the CICS platforms from Universite d’Auvergne for assistance with the transmission electron microscopy, as well as Alexandra Rezard and Abdelkrim Alloui (Animal Facilities) for animal care.
Grant Support
This study was supported by the Ministere de la Recherche et de la Technologie, Inserm and Universite d’Auvergne (UMR1071), INRA (USC-2018), and grants from the Conseil Regional d’Auvergne.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received May 15, 2013; revised November 22, 2013; accepted November 26, 2013; published OnlineFirst December 13, 2013.
References
1. Edge SB, Compton CC. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 2010;17:1471–4.
2. O'Connell JB, Maggard MA, Liu JH, Etzioni DA, Ko CY. Are survival rates different for young and older patients with rectal cancer? Dis Colon Rectum 2004;47:2064–9.
3. Fearon ER, Vogelstein B. A genetic model for colorectal tumorigenesis. Cell 1990;61:759–67.
4. Dalton-Griffin L, Kellam P. Infectious causes of cancer and their detection. J Biol 2009;8:67.
5. Zur Hausen H. The search for infectious causes of human cancers: where and why. Virology 2009;392:1–10.
6. Lax AJ. Opinion: bacterial toxins and cancer–a case to answer? Nat Rev Microbiol 2005;3:343–9.
7. Collins D, Hogan AM, Winter DC. Microbial and viral pathogens in colorectal cancer. Lancet Oncol 2010;12:504–12.
8. Arthur JC, Perez-Chanona E, Muhlbauer M, Tomkovich S, Uronis JM, Fan TJ, et al. Intestinal inflammation targets cancer-inducing activity of the microbiota. Science 2012;338:120–3.
9. Tjalsma H, Boleij A, Marchesi JR, Dutilh BE. A bacterial driver-pas-senger model for colorectal cancer: beyond the usual suspects. Nat Rev Microbiol 2012;10:575–82.
10. Swidsinski A, Khilkin M, Kerjaschki D, Schreiber S, Ortner M, Weber J, et al. Association between intraepithelialEscherichia coli and colo-rectal cancer. Gastroenterology 1998;115:281–6.
11. Martin HM, Campbell BJ, Hart CA, Mpofu C, Nayar M, Singh R, et al. EnhancedEscherichia coli adherence and invasion in Crohn's disease and colon cancer. Gastroenterology 2004;127:80–93.
12. Darfeuille-Michaud A, Neut C, Barnich N, Lederman E, Di Martino P. Presence of adherentEscherichia coli strains in ileal mucosa of patients with Crohn's disease. Gastroenterology 1998;115:1405–13. 13. Darfeuille-Michaud A, Boudeau J, Bulois P, Neut C, Glasser AL, Barnich N, et al. High prevalence of adherent-invasiveEscherichia coli associated with ileal mucosa in Crohn's disease. Gastroenterology 2004;127:412–21.
14. Thelestam M, Frisan T. Cytolethal distending toxins. Rev Physiol Biochem Pharmacol. 2004;152:111–33.
15. Falzano L, Filippini P, Travaglione S, Miraglia AG, Fabbri A, Fiorentini C. Escherichia coli cytotoxic necrotizing factor 1 blocks cell cycle G2/M transition in uroepithelial cells. Infect Immun. 2006;74:3765–72. 16. Malorni W, Fiorentini C. Is the Rac GTPase-activating toxin CNF1 a
smart hijacker of host cell fate? Faseb J 2006;20:606–9.
17. Jubelin G., Chavez CV, Taieb F, Banfield MJ, Samba-Louaka A, Nobe R, et al. Cycle inhibiting factors (CIFs) are a growing family of functional cyclomodulins present in invertebrate and mammal bacterial patho-gens. PLoS ONE 2009;4:e4855.
18. Nougayrede JP, Homburg S, Taieb F, Boury M, Brzuszkiewicz E, Gottschalk G, et al.Escherichia coli induces DNA double-strand breaks in eukaryotic cells. Science. 2006;313:848–51.
19. Cuevas-Ramos G, Petit CR, Marcq I, Boury M, Oswald E, Nougayrede JP.Escherichia coli induces DNA damage in vivo and triggers genomic instability in mammalian cells. Proc Natl Acad Sci U S A 2010;107: 11537–42.
20. Buc E, Dubois D, Sauvanet P, Raisch J, Delmas J, Darfeuille-Michaud A, et al. High prevalence of mucosa-associatedE. coli producing cyclo-modulin and genotoxin in colon cancer. PLoS One 2013;8:e56964. 21. Winter SE, Lopez CA, Baumler AJ. The dynamics of gut-associated
microbial communities during inflammation. EMBO Rep. 2013;14: 319–27.
22. Fujimori Y, Fujimori T, Imura J, Sugai T, Yao T, Wada R, et al. An assessment of the diagnostic criteria for sessile serrated adenoma/ polyps: SSA/Ps using image processing software analysis for Ki67 immunohistochemistry. Diagn Pathol 2012;7:59.
23. Carvalho FA, Barnich N, Sivignon A, Darcha C, Chan CH, Stanners CP, et al. Crohn's disease adherent-invasiveEscherichia coli colonize and induce strong gut inflammation in transgenic mice expressing human CEACAM. J Exp Med. 2009;206:2179–89.
24. Shen XJ, Rawls JF, Randall T, Burcal L, Mpande CN, Jenkins N, et al. Molecular characterization of mucosal adherent bacteria and associa-tions with colorectal adenomas. Gut Microbes 2010;1:138–47. 25. Jantscheff P, Terracciano L, Lowy A, Glatz-Krieger K, Grunert F,
Micheel B, et al. Expression of CEACAM6 in resectable colorectal cancer: a factor of independent prognostic significance. J Clin Oncol. 2003;21:3638–46.
26. Barnich N, Carvalho FA, Glasser AL, Darcha C, Jantscheff P, Allez M, et al. CEACAM6 acts as a receptor for adherent-invasiveE. coli, supporting ileal mucosa colonization in Crohn disease. J Clin Invest 2007;117:1566–74.
27. Li Y, Kundu P, Seow SW, Matos CT, Aronsson L, Chin KC, et al. Gut microbiota accelerate tumor growth via c-jun and STAT3 phosphor-ylation in APCMin/þ mice. Carcinogenesis 2012;33:1231–8. 28. Homburg S, Oswald E, Hacker J, Dobrindt U. Expression analysis of
the colibactin gene cluster coding for a novel polyketide inEscherichia coli. FEMS Microbiol Lett 2007;275:255–62.
29. Uronis JM, Jobin C. Microbes and colorectal cancer: is there a relationship? Curr Oncol 2009;16:22–4.
30. Wu S, Rhee KJ, Albesiano E, Rabizadeh S, Wu X, Yen HR, et al. A human colonic commensal promotes colon tumorigenesis via acti-vation of T helper type 17 T cell responses. Nat Med 2009;15: 1016–22.
31. Xu F, Wang F, Di M, Huang Q, Wang M, Hu H, et al. Classification based on the combination of molecular and pathologic predictors is superior to molecular classification on prognosis in colorectal carcinoma. Clin Cancer Res 2007;13:5082–8.
32. Mulcahy HE, Toner M, Patchett SE, Daly L, O'Donoghue DP. Identi-fying stage B colorectal cancer patients at high risk of tumor recurrence and death. Dis Colon Rectum 1997;40:326–31.
33. Noura S, Yamamoto H, Ohnishi T, Masuda N, Matsumoto T, Takayama O, et al. Comparative detection of lymph node micrometastases of stage II colorectal cancer by reverse transcriptase polymerase chain reaction and immunohistochemistry. J Clin Oncol 2002;20:4232–41. 34. Quah HM, Chou JF, Gonen M, Shia J, Schrag D, Landmann RG, et al. Identification of patients with high-risk stage II colon cancer for adjuvant therapy. Dis Colon Rectum 2008;51:503–7.
35. Chin CC, Wang JY, Changchien CR, Huang WS, Tang R. Carcinoma obstruction of the proximal colon cancer and long-term prognosis– obstruction is a predictor of worse outcome in TNM stage II tumor. Int J Colorectal Dis 2010;25:817–22.
36. Stocchi L, Fazio VW, Lavery I, Hammel J. Individual surgeon, pathol-ogist, and other factors affecting lymph node harvest in stage II colon carcinoma. Is a minimum of 12 examined lymph nodes sufficient? Ann Surg Oncol 2011;18:405–12.
