Inflammation and microbiota are critical components of intestinal tumorigenesis. To dissect how the microbiota contributes to tumor distribution, we generated germ-free (GF) ApcMin/+and ApcMin/+;Il10−/− mice and exposed them to specific-pathogen-free (SPF) or colorectal cancer-associated bacteria. We found that colon tumorigenesis significantly correlated with inflammation in SPF-housed ApcMin/+;Il10−/−, but not in ApcMin/+mice. In contrast, small intestinal neoplasia development significantly correlated with age in both ApcMin/+;Il10−/− and ApcMin/+ mice. GF ApcMin/+;Il10−/− mice conventionalized by an SPF microbiota had significantly more colon tumors compared with GF mice. Gnotobiotic studies revealed that while Fusobacterium nucleatum clinical isolates with FadA and Fap2 adhesins failed to induce inflammation and tumorigenesis, pks+Escherichia coli promoted tumorigenesis in the ApcMin/+;Il10−/− model in a colibactin-dependent manner, suggesting colibactin is a driver of carcinogenesis. Our results suggest a distinct etiology of cancers in different locations of the gut, where colon cancer is primarily driven by inflammation and the microbiome, while age is a driving force for small intestine cancer. Cancer Res; 77(10); 2620–32. ©2017 AACR.
Colorectal cancer, the third most common type of malignancy and the third leading cause of cancer-related deaths in the United States (1), involves both genetic and environmental factors. Among the genomic changes associated with colorectal cancers, loss-of-function mutations in the Apc (adenomatous polyposis coli) gene, a regulator of the WNT signaling pathway, are the most prevalent and are considered the initiating event in approximately 80% of colorectal cancers (2). Of the environmental factors, the gut microbiota is increasingly appreciated as a key player in colorectal cancer pathogenesis. Colorectal cancer patients often carry a distinct microbiota from the healthy population (3). Microbes can modulate colorectal cancer development directly by generating genotoxins, or indirectly and more commonly by mediating inflammatory and immune responses (4, 5). Inflammation is not only a hallmark of colorectal cancer (5), but also an established risk factor for colorectal cancer as supported by epidemiological data from individuals with inflammatory bowel diseases (6).
While human studies provide valuable correlation data on colorectal cancer, much of the mechanistic insight into the disease etiology is obtained from mouse models. Mouse colorectal cancer models can be categorized into two classes: spontaneous and chemical induced (7). Spontaneous colorectal cancer mice carry mutations in genes frequently mutated in human colorectal cancers. The multiple intestinal neoplasia (Min) mouse (referred to as ApcMin/+ hereafter), a commonly used animal model of intestinal carcinogenesis, carries a point mutation in one allele of the Apc gene and is susceptible to spontaneous intestinal adenoma formation, although predominantly in the small bowel, without exhibiting chronic intestinal inflammation (8). A general protumorigenic role for the microbiota was demonstrated in the ApcMin/+ model, as the mice display reduced tumor load in the small and/or large intestine when derived under germ-free (GF) conditions (9, 10). Noticeably, inflammation also enhances development of colon cancer in this model, as seen with the use of dextran sulfate sodium (DSS; ref. 11), by specifically deleting the Apc gene in epithelial cells (12), and by genetically introducing defective IL10 signaling (13–15). Inflammation and colonic polyposis in mice with Apc deficiency and T cell–specific deletion of Il10, ApcΔ468;CD4CreIl10f/f mice can be attenuated by antibiotic treatment (14), suggesting that microbiota-driven inflammation underlies colitis-associated colorectal cancer.
The mechanisms by which microbes have been shown to promote development of colorectal cancer are diverse and somewhat specific to each microorganism. For example, enterotoxigenic Bacteroides fragilis promotes colorectal cancer through induction of the Th17 response (16), polyketide synthase (pks)+ island carrying E. coli via production of the genotoxin colibactin (17), and Fusobacterium nucleatum adhesins that either bind E-cadherin to promote tumor growth (FadA) or promote immune invasion and localization to tumors (Fap2; refs. 18–20). Although genetics, inflammation, and microbes play a role in promoting intestinal carcinogenesis in preclinical models, it is unclear how these factors interact and how much each contributes to promotion of colorectal cancer.
To stringently evaluate the relationship between genetic susceptibility to inflammation, microbial status, and cancer, we utilized specific-pathogen-free (SPF) and gnotobiotic ApcMin/+and ApcMin/+;Il10−/− mice. In this study, we found that inflammation and the microbiota is essential for colorectal but not small intestinal neoplasia in SPF ApcMin/+;Il10−/− mice. We also demonstrated that colorectal cancer-associated bacteria have differential abilities to promote colorectal cancer with colibactin-producing E. coli, but not F. nucleatum, inducing colon tumors in ApcMin/+;Il10−/− mice.
Materials and Methods
The University of Florida Institutional Animal Care and Use Committee approved all animal experiments (Protocol #201308038). 129/SvEv ApcMin/+ mice were derived GF and crossed to GF 129/SVEv Il10−/− mice to generate GF ApcMin/+;Il10−/− mice. GF ApcMin/+;Il10−/− and ApcMin/+ mice were transferred to the SPF breeding suite and bred for 2 to 3 generations. SPF ApcMin/+;Il10−/− and ApcMin/+ mice were either transferred to an SPF housing suite after weaning or remained in the breeding suite, mice transferred to the SPF housing suite were sacrificed at 12, 16, and 20 weeks of age. SPF ApcMin/+;Il10−/− and ApcMin/+ mice older than 20 weeks were retired breeders from the SPF breeding suite.
Bacterial strains and culture conditions
F. nucleatum strains were provided by Dr. Emma Allen-Vercoe (University of Guelph), including the inflammatory bowel disease clinical isolate EAVG_016, and colorectal cancer isolates CC53, CC7/3JVN3C1, CC7/5JVN1A4, CC2/3Fmu1, CC2/3FmuA, and CC7/4Fmu3 (used for the 20-week colonization experiment in ApcMin/+ mice). E. coli NC101 or NC101 ΔclbP were cultured from glycerol stocks in LB broth, then diluted 1:10 in fresh LB medium and cultured at 37°C before harvesting for gavage. F. nucleatum strains were cultured in Brain Heart Infusion Broth (BHI; AS-872, Anaerobe Systems) statically at 37°C in an anaerobic chamber (type B Vinyl, Coy Laboratory). Enumeration of F. nucleatum was done by anaerobically plating serial dilutions of culture or fecal materials on fastidious anaerobic agar supplemented with 5% sheep blood.
PCR detection of F. nucleatum adhesins
All F. nucleatum strains were screened for the FadA and Fap2 adhesins by PCR using the following primers: Fn 16S_F GGATTTATTGGGCGTAAAGC; Fn 16S_R, GGCATTCCTACAAATATCTACGAA; fadA_F CAAATCAAGAAGAAGCAAGATTCAAT; fadA_R, GCTTGAAGTCTTTGAGCTCT (18); fap2_F, AGCCTCTGAGGGTACAAGGT; fap2_R, TGAGCCCCTCCTTCTTCTGA. The screening revealed 4 F. nucleatum colorectal cancer isolate strains were fadA+, fap2−, and 2 were fadA+, fap2+ (Supplementary Fig. S1).
E. coli NC101 clbP mutation
Inactivation of gene clbP was performed by using the lambda Red recombinase method (21) using primers clbP-P1 (TTCCGCTATGTGCGCTTTGGCGCAAGAACATGAGCCTATCGGGGCGCAAgtgtaggctggagctgcttc) and clbP-P2 (GTATACCCGGTGCGACATAGAGCATGGCGGCCACGAGCCCAGGAACCGCCcatatgaatatcctccttag). The allelic exchange was confirmed by PCR using primers IHAPJPN29 and IHAPJPN30 (22).
SPF microbiota preparation
Cecal and fecal contents were collected from wild-type 129/SvEv mice that were housed under SPF conditions in the animal facility at the University of Florida. One gram of the contents was suspended in 10 mL sterile PBS, broken down using pipette tips, and vortexed. After settling for 2 minutes, the supernatant was transferred to a new tube, mixed with equal volume of sterile 20% glycerol, and frozen at –80°C.
Seven- to 12-week GF ApcMin/+ and ApcMin/+;Il10−/− were transferred to SPF conditions or gnotobiotic isolators as described above. SPF stock microbiota was diluted 1:106 and 200 μL of this mixture was gavaged to each mouse. E. coli NC101 or NC101 ΔclbP was gavaged at 108 CFU/mouse. For AOM/Il10−/− experiments, mice received 6 weekly intraperitoneal AOM (A5486, Sigma) injections (10 mg/kg) starting 1 week post monoassociation with E. coli and mice were sacrificed 20 weeks post monoassociation. F. nucleatum was gavaged at 108 CFU/mouse when a single strain was used, or 108 CFU per strain per mouse when a mixture of strains were used. BHI medium weekly gavaged mice were used as control for F. nucleatum experiments.
Mice were euthanized at indicated time points. The small intestine, cecum, and colon were cut open longitudinally and macroscopic tumors were counted. About 1 × 0.5 cm snips were taken from the proximal and distal colon, flash frozen in liquid nitrogen, and stored at −80°C until analysis. The rest of the colon was Swiss rolled and fixed in 10% neutral buffered formalin solution. Swiss rolls were processed, paraffin-embedded, sectioned and hematoxylin and eosin stained by the Molecular Pathology Core at the University of Florida. Histological scoring of inflammation was performed blindly as described previously (17) and calculated as the average between the proximal and distal colon region scores.
IHC was performed as described previously (17). Briefly, Swiss roll sections were deparaffinized, rehydrated, and boiled in 10 mmol/L citrate buffer for antigen retrieval. For CTNBB1, the mouse anti-CTNNB1 antibody (1:300 overnight; 6101503, BD Transduction Laboratories) and mouse on mouse (M.O.M.) peroxidase kit (PK-2200, Vector Labs) were used. For PCNA, sections were blocked with 1% BSA, incubated with anti-PCNA clone PC10 (M087901-2, Dako) mouse monoclonal antibody (1:300, 30 minutes), followed by 1:1,000 goat anti-mouse biotin secondary antibody (31800 Fisher), and then incubated with streptavidin–horseradish peroxidase (18–152, Millipore). Liquid DAB+ (K3467, Dako) was used according to the manufacturer's instructions for development.
Fecal DNA extraction and 16S qPCR
DNA was extracted using phenol:chloroform separation followed by the DNeasy Blood and Tissue Kit (69506, Qiagen). qPCR was performed on the CFX384 Touch Real-Time PCR Detection System (1855485, Bio-Rad) using the SsoAdvanced Universal SYBR Green Supermix (1725274, Bio-Rad). The following primers were used: Fuso_F GGATTTATTGGGCGTAAAGC, Fuso_R GGCATTCCTACAAATATCTACGAA; and Eubacteria_F GGTGAATACGTTCCCGG, Eubacteria_R TACGGCTACCTTGTTACGACTT.
16S rRNA sequencing
The V1–V3 region hypervariable region of the 16S rRNA gene was amplified using primer pair 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 534R (5′-ATTACCGCGGCTGCTGG-3′). Both the forward and the reverse primers contained universal Illumina paired-end adapter sequences, as well as unique individual 4 to 6 nucleotide barcodes between PCR primer sequence and the Illumina adapter sequence to allow multiplex sequencing (Supplementary Table S1). PCR products were visualized on an agarose gel, before samples were purified using the Agencourt AMPure XP kit (A63881, Beckman Coulter) and quantified by qPCR with the KAPA Library Quantification Kit (KK4824, KAPA Biosystems). Equimolar amount of samples was then pooled and sequenced with an Illumina MiSeq. See Supplementary Data for description of 16S rRNA sequencing analysis.
qPCR examination of inflammatory cytokines
RNA was extracted from frozen tissue snips using TRIzol reagent followed by phenol:chloroform separation. After DNA removal using the Turbo DNA-free Kit (AM1907, Ambion), 10 to 1,000 ng of RNA was used for cDNA synthesis using the iScript cDNA Synthesis Kit (1708891, Bio-Rad).
qPCR was performed on the CFX384 Touch Real-Time PCR Detection System (1855485, Bio-Rad) using the SsoAdvanced Universal SYBR Green Supermix (1725274, Bio-Rad). The following primers were used: IL6_F CGGAGGCTTGGTTACACATGTT, IL6_R CTGGCTTTGTCTTTCTTGTTATC; TNFα_F ATGAGCACAGAAAGCATGATC, TNFα_R TACAGGCTTGTCACTCGAATT; IFNγ_F ACGCTTATGTTGTTGCTGATGG, IFNγ_R CTTCCTCATGGCTGTTTCTGG; IL1β_F GCCCATCCTCTGTGACTCAT, IL1β_R AGGCCACGGTATTTTGTCG; IL17A_F GCCCTCAGACTACCTCAACC, IL17A_R ACACCCACCAGCATCTTCTC; IL22_F CATGCAGGAGGTGGTGCCTT, IL22_R CAGACGCAAGCATTTCTCAG; 36B4_F TCCAGGCTTTGGGCATCA, 36B4_R CTTTATTCAGCTGCACATCACTCAGA; and GUSB_F CCGATTATCCAGAGCGAGTATG, GUSB_R CTCAGCGGTGACTGGTTCG. 36B4 and Gusb were used as references. Relative fold gene expression was calculated using the delta delta Ct method.
Data are expressed as mean ± standard deviation. Significance thresholds of ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05, or NS: not significant (P > 0.05) are shown. Statistics for all figures except for the 16S rRNA sequencing analysis were calculated with GraphPad Prism using the Mann–Whitney nonparametric unpaired two-tailed t test. For correlational analyses, the Spearman nonparametric correlation analysis with a confidence interval of 95% and a two-tailed significance test with alpha = 0.05 was used.
All sequences have been uploaded to the NCBI SRA (National Center for Biotechnology Information Sequence Read Archive) under BioProject PRJNA350319. See Supplementary Table S3 for individual accession numbers.
Inflammation promotes development of colorectal cancer in ApcMin/+;Il10−/− mice
To investigate the interaction between inflammation and colorectal cancer development, we interbred Il10−/− mice to ApcMin/+ mice (129SvEV background) to generate ApcMin/+;Il10−/− mice. Colon and cecal tumors increased dramatically in ApcMin/+;Il10−/− compared with ApcMin/+ mice (colon tumor mean= 5.03 vs. 0.73, respectively P < 0.0001; cecal tumor mean = 0.51 vs. 0, respectively, P < 0.0005; Fig. 1A and C), whereas small bowel lesions remained similar between the two genotypes (mean = 0.84 vs. 2.23 P = 0.2914; Fig. 1D). Histological assessment showed presence of colonic neoplastic lesions in ApcMin/+;Il10−/− mice (Fig. 1E) and, as expected, increased inflammation in ApcMin/+;Il10−/− mice compared to ApcMin/+ mice (combined score mean = 1.32 vs. 0.23, respectively, P < 0.0001; Fig. 1B). There was a significant correlation between development of colorectal cancer and extent of colon inflammation in ApcMin/+;Il10−/− mice (r = 0.7441; P < 0.0001), whereas no such correlation is observed in ApcMin/+ mice (r = −0.09608; P = 0.6135; Fig. 1F). Furthermore, development of small intestinal neoplasia did not correlate with the state of colitis in either model (ApcMin/+;Il10−/−, r = 0.2364; ApcMin/+, r = 0.2526; P > 0.05; Fig. 1F). Interestingly, endpoint age (see Materials and Methods for description of the chosen endpoint age range) was a significant contributor to tumorigenesis in the small bowel of both ApcMin/+;Il10−/− and ApcMin/+ mice (r = 0.6458, 0.8208, respectively, P < 0.0001) but only weakly contributed to neoplasia in the large bowel of ApcMin/+;Il10−/− mice (r = 0.2561; P = 0.0256; Fig. 1G). Thus, genetic susceptibility to inflammation promotes colon but not small intestinal tumorigenesis in SPF ApcMin/+;Il10−/− mice, while age appears to be the primary factor contributing to small intestinal tumorigenesis in SPF ApcMin/+;Il10−/− and ApcMin/+ mice.
Due to its role in promoting cellular proliferation, we evaluated the distribution of nuclear catenin beta 1 (CTNNB1) and proliferating cell nuclear antigen (PCNA) in actively inflamed and neoplastic regions of ApcMin/+;Il10−/− and ApcMin/+ colons. Nuclear CTNNB1 and PCNA staining was mostly restricted to the crypt bases in ApcMin/+ mice (Fig. 2B and D; Supplementary Fig. S2B and S2D). In contrast, the colonic mucosa from ApcMin/+;Il10−/− mice showed areas of CTNNB1 and PCNA staining extending the full crypt length in dysplastic regions (Fig. 2A and C; Supplementary Fig. S2A and S2C). In addition, expression of proliferative and inflammatory IL6, TNFα, IFNγ, IL1β, IL22, and IL17a mRNA increased in ApcMin/+;Il10−/− compared with ApcMin/+proximal colon tissues (Fig. 2E). Furthermore, ApcMin/+;Il10−/− mice with a high number of tumors (>2) had significantly increased levels of TNFα, IFNγ, and IL1β mRNA compared to low tumor number (≤2) ApcMin/+;Il10−/− mice (Fig. 2F). Taken together, these data suggest that the heightened inflammatory and proliferative state observed in ApcMin/+;Il10−/− compared with ApcMin/+ mice increased propensity for colorectal tumor formation and progression.
Differential microbial composition within ApcMin/+;Il10−/− mice correlates with tumor multiplicity and inflammation status
We previously showed that inflammation altered microbial composition (17), a condition important for colorectal cancer development. Using 16S rRNA sequencing, we found that the SPF ApcMin/+;Il10−/− stool microbiome was associated with both colon inflammation score and tumor number (Fig. 3A), especially along the first principal coordinates analysis (PCoA) axis (Fig. 3B). Using mixed linear models with either colon tumor number or colitis score as the independent variable, we identified 17 genera significantly affected by both colon tumor number and colitis score, including Allobaculum, Anaerotruncus, Butyrivibrio, Clostridium IV, Clostridium XI, Enterococcus, Oscillibacter, Pseudoflavonifractor, and Syntrophococcus (Fig. 3C; Supplementary Table S2). Only 5 genera were affected by colon tumor number but not combined inflammation score (Akkermansia, Coprobacillus, Escherichia/Shigella, Marvinbryantia and Robinsoniella) while 4 genera were only affected by combined inflammation score and not colon tumor number (Alistipes, Enterorhabdus, Flavonifractor, and Roseburia; Fig. 3C). Interestingly, approximately 1/3 of these genera (10/26) were significantly increased in mice with colon tumors and/or inflammation while approximately 2/3 were decreased (16/26). Therefore, the microbial community composition is correlated with higher tumor numbers and inflammation, suggesting specific bacteria play a role in the development of colon inflammation and tumorigenesis in ApcMin/+;Il10−/ mice.
Bacteria are essential for development of colon tumorigenesis in ApcMin/+;Il10−/−mice
To stringently evaluate the impact of bacteria on colorectal cancer development, we derived ApcMin/+;Il10−/− and ApcMin/+mice in GF conditions and then performed microbial manipulation by either gavaging the mice with SPF biota or allowing them to naturally acquire a microbiota in SPF conditions. Importantly, colon tumorigenesis was practically abolished in GF ApcMin/+;Il10−/− mice (mean = 0) compared with SPF conditions (Fig. 4A,D). Interestingly, SPF gavage enhanced colon tumor loads compared to passive SPF colonization of ApcMin/+;Il10−/− mice (mean = 3.86 vs. 1, respectively, P = 0.0126; Fig. 4A and D), although colitis scores and most proliferative/inflammatory cytokine expression (IL6, TNFα, IFNγ, IL22, and IL17a) were not significantly different (colitis score means = 1.86 vs. 2.36 respectively, P = 0.46; Figs. 4B and 5D). Colon inflammation and tumors were negligible or absent in GF and SPF gavaged ApcMin/+mice (Supplementary Fig. S3A and S3B), suggesting inflammation is a key component of bacteria-mediated colon tumorigenesis. Development of small bowel neoplasia in ApcMin/+;Il10−/− and ApcMin/+mice (Fig. 5C; Supplementary Fig. S3C) was not significantly impacted by microbial colonization, suggesting a less pronounced role for the microbiota in the small intestine compared to the colon in this model. GF ApcMin/+;Il10−/− colons had reduced nuclear CTNNB1 and PCNA (Fig. 5A–C; Supplementary Fig. S4A–S4C) and decreased inflammatory cytokine expression (Fig. 5D) compared to SPF mice, indicating bacteria play a significant role in the increased inflammatory and proliferative state in SPF ApcMin/+;Il10−/− colons.
Gnotobiotic experiments reveal specific microbial requirements for colorectal cancer development in ApcMin/+;Il10−/−mice
Fusobacterium spp. have been linked to the development of colorectal cancer (4) and recent studies showed increased carcinogenesis in F. nucleatum-colonized ApcMin/+ mice (23–25). To investigate the interplay between microbiota and F. nucleatum in colorectal cancer, we transferred GF ApcMin/+ to SPF conditions and gavaged them with SPF microbiota followed by weekly gavage with a fadA+, fap2−F. nucleatum colorectal cancer clinical isolate for 20 weeks. Interestingly, and in contrast to previous studies (23–25), the presence of F. nucleatum failed to enhance carcinogenesis in these mice (Fig. 6A and C). We next transferred GF ApcMin/+;Il10−/− mice to SPF conditions, gavaged them with SPF microbiota, and then introduced a mixture of 6 F. nucleatum strains (fadA+, fap2− or +) obtained from colorectal cancer patients by weekly gavage for 16 weeks. Although ApcMin/+;Il10−/− mice developed more inflammation and tumors than ApcMin/+ mice, presence of F. nucleatum species did not influence intestinal carcinogenesis or colitis (Fig. 6B and C). To rule out the possibility that the SPF biota downmodulates F. nucleatum carcinogenic properties, we transferred GF ApcMin/+ mice to a gnotobiotic isolator and associated these mice with a mixture of 6 F. nucleatum colorectal cancer clinical isolates (fadA+, fap2− or +; single gavage). Again, the presence of F. nucleatum isolates failed to enhance intestinal carcinogenesis in ApcMin/+ mice (Fig. 6D) despite the presence of high colony-forming unit (CFU) counts (mean = 107 CFU/g of stool). Although we did not see evidence of histological inflammation, to further confirm the effect of F. nucleatum on host inflammatory response, we examined inflammatory cytokine gene expression via qPCR. We found comparable levels of TNFα and IL1β in the mouse distal colon between germ-free and monoassociated ApcMin/+ mice, whereas IL6, IFNγ, IL22, and IL17a levels were undetectable (data not shown).
To further study the relationship between microbial status and carcinogenesis in gnotobiotic ApcMin/+;Il10−/− mice, we colonized these mice with E. coli NC101, a strain producing the colibactin genotoxin. We previously showed that removing pks from E. coli NC101 decreased development of colorectal cancer in the azoxymethane (AOM)/Il10−/− mouse model (17). We next monoassociated ApcMin/+;Il10−/− mice by oral gavage (108 CFU/mouse) with an E. coli NC101 mutant deficient for clbP, the pks gene necessary for colibactin activation (ΔclbP). We found that wild-type NC101-colonized mice developed significantly more colon tumors than E. coli NC101 ΔclbP-associated mice (mean = 1.71 vs. 0.17, respectively, P = 0.0023; Fig. 7A). The finding that NC101 ΔclbP has diminished carcinogenic capacity compared to NC101 was confirmed in the AOM/Il10−/− model (mean = 2 vs. 5 tumors, respectively, P = 0.039; Supplementary Fig. S5). Importantly, deletion of clbP did not compromise the ability of E. coli NC101 to induce inflammation (colitis score mean = 2.675, 2.5 respectively; P = 0.76; Fig. 7B and D). Presence of a functional pks did not influence development of small intestinal tumors in ApcMin/+;Il10−/− mice (mean = 0.29, 0.8, respectively; P = 0.22; Fig. 7C). Overall, these findings show that ApcMin/+;Il10−/− mice are sensitive to microbial status and develop site specific tumors, with colibactin-producing E. coli but not fadA+, fap2−/+ F. nucleatum enhancing colon tumor development.
Genetics and environmental factors play an important role in colorectal cancer development, with increasing attention directed toward the intestinal microbiota as a key environmental component (26). In general, the microbiota is thought to play a procarcinogenic role in colorectal cancer with numerous colorectal cancer mouse models demonstrating tumor reduction in antibiotic treated or GF mice (27). Here we utilized gnotobiotic ApcMin/+and ApcMin/+;Il10−/− mice to define the relationship between inflammation, microbial status, and tumorigenesis. We observed that despite genetic susceptibility in both ApcMin/+and ApcMin/+;Il10−/− mice, colonic inflammation in the latter mice fosters development of colon tumors, which was associated with an altered luminal microbiota composition. Gnotobiotic experiments revealed that E. coli colibactin but not F. nucleatum FadA and Fap2 adhesins promote colon tumorigenesis, suggesting an intricate interaction between host genetics and bacteria.
We observed an inflammation-dependent increase in colon tumorigenesis in 129 SvEv ApcMin/+;Il10−/− mice, which is in line with previous reports on Il10-deficient C57BL/6 ApcMin/+ and ApcΔ468 mice (13, 14). However, in the small intestine compartment, tumors developed at a comparable rate regardless of Il10 status in ApcMin/+ mice, which is in contrast to previous findings showing a delay in small intestinal polyp formation in Il10 deficient ApcΔ468 mice (15). Possible explanations for the differences in small intestine tumor formation may be due to genetic background differences that have been shown to strongly modulate tumor multiplicity, particularly in the small intestine of ApcMin/+mice (28). Nevertheless, our findings that ApcMin/+;Il10−/−and ApcMin/+ small intestine tumor development significantly correlated with age but not inflammation suggest that age-related factors are a primary driver of small intestinal tumorigenesis
We observed changes in the abundance of 26 genera that correlated with colon inflammation and/or tumor number. In contrast to previous sequencing results with ApcMin/+ and T cell–specific ApcMin/+;Il10−/−mice, no significant increases in the Bacteroides or Porphyromonas genera were observed (14). These differences are likely due to a combination of factors including sampling location (stool vs. tissue), mouse genetic background, husbandry and 16S sequencing methods (V1–V3 region vs. V3–V4). Importantly, Akkermansia, Blautia, Dorea, Enterococcus, and Escherichia/Shigella, which positively correlated with tumor number, have been previously associated as increased in mucosal tissue (17, 29, 30) or stools from either human colorectal cancer patients (31–33) or AOM/DSS mice (34, 35). Conversely, some of the genera that negatively correlated with tumors and inflammation or just inflammation included Clostridium XIVa, Lachnospiracea and Roseburia, which have been implicated as butyrate producers and were decreased in colorectal cancer stool (31, 33) or stools from AOM/DSS mice transplanted with stools from human patient samples (34). Thus, our sequencing results suggest there are changes in the microbiota that are associated with colon inflammation and tumor status.
The interaction between bacteria and the host in the context of intestinal carcinogenesis is complex. One study, using the chemical AOM/DSS regimen, reported that GF mice developed more colonic tumors than mice colonized with a complex biota, suggesting certain bacteria can have a beneficial role in colorectal cancer (36). Because microbial composition is a key determinant of colon tumor burden in AOM/DSS mice (35, 37), this chemical model may better capture the protective functions of bacteria than the genetic ApcMin/+mouse model.
Nevertheless, the role of bacteria in ApcMin/+intestinal tumorigenesis is complex, with one report showing fewer tumors in the middle region of the small intestine in GF ApcMin/+mice (9) while another report showed reduced tumors throughout the intestine in GF ApcMin/+mice (10). The difference in tumor distribution is not clear. Our finding that bacteria promote colon tumors in ApcMin/+;Il10−/−mice is in line with a study showing reduced colon polyp numbers in ApcΔ468;CD4CreIl10f/f mice following broad-spectrum antibiotic treatment (14).
Numerous studies have implicated Fusobacterium spp., in particular F. nucleatum, as carcinogenic, based on associative studies showing the presence of the bacterium in the luminal and mucosal compartment of human colorectal cancer patients using genomic analyses (3, 4, 24, 33). In addition, daily gavage of F. nucleatum (strain EAVG_002; 7/1 or ATCC 25586) for 8–24 weeks was shown to promote intestinal tumorigenesis in C57BL/6 ApcMin/+mice (23–25). Surprisingly, ApcMin/+ and ApcMin/+;Il10−/−mice colonized with fadA+, fap2+/−F. nucleatum isolates from colorectal cancer patients failed to promote intestinal tumorigenesis, in the presence (SPF) or absence of complex biota (gnotobiotic). The absence of tumorigenesis in monoassociated ApcMin/+mice was not due to poor colonization, because a high load of F. nucleatum (107 CFU/g) was recovered from these mice. The discrepancy between our study and previous F. nucleatum ApcMin/+studies (23–25) is unclear but could be the result of strain specific properties (EAVG_002; refs. 19, 38) and ATCC 25586 (39, 40) vs. strains tested here), mouse genetic background differences and different microbial environments, as microbial communities are notoriously different between institutions. Nevertheless, our gnotobiotic approach clearly showed that presence of FadA and Fap2 adhesins in F. nucleatum is not sufficient to induce either inflammation or cancer, as opposed to E. coli pks+ monoassociated ApcMin/+;Il10−/−mice. Thus, it is possible that only a select group of F. nucleatum strains possess carcinogenic abilities, which require interactions with other specific members of the microbial community. It will be important to define these interactions and test a larger set of F. nucleatum strains to determine the role of these bacteria in colorectal cancer pathogenesis.
Several studies have found an association between pks+E. coli and human colorectal cancer patients (17, 29). Furthermore, pks+E. coli isolates from mice or human colorectal cancer patients have a protumorigenic effect in GF AOM/Il10−/−, SPF ApcMin/+, and SPF AOM/DSS mice (17, 29, 41). However, because the pks-associated clbA gene is implicated in the production of siderophores located in the enterobactin (ent) and yersiniabactin (HPI) loci (42), and our previous observation was based on removal of the entire pks island, it was unclear whether the decreased tumorigenesis observed in AOM/Il10−/− mice was the consequence of dual siderophore/colibactin impairment, or solely due to abolished pks activity. Using a mutant with defective ClbP, the key enzyme implicated in pre-colibactin cleavage and generation of the active form (43), we demonstrated the colibactin-producing E. coli murine isolate NC101 is responsible for the protumorigenic effect of the bacterium in ApcMin/+;Il10−/− mice. Whether clbA contributes to colibactin-mediated tumorigenesis is still unclear and would need to be investigated, especially because a recent in vitro study showed that iron levels and E. coli iron sensors regulate clbA transcription and colibactin production (44). Because our studies were performed using a monoassociation approach, and therefore without competitive pressure from other microorganisms, the full extent of iron acquisition on E. coli pks+ induced carcinogenesis remains unclear.
Recent studies have attempted to dissect the contributions of intrinsic (organ specific stem cell division rates, aging) and extrinsic factors (hereditary mutations, lifestyle, environmental exposure, etc.), to overall cancer risk in humans (45–47). However, the interplay between all these factors makes it difficult to tease out the various contributions using epidemiological data. Nevertheless, these studies suggest that small intestine cancers with a relatively low lifetime risk are driven by intrinsic risk factors, while 82.9% of the mutation signatures in colorectal cancers are from extrinsic factors, correlating with a much higher lifetime risk (46). We postulate that one of the extrinsic factors contributing to colorectal cancer risk is the microbiota, which not coincidentally is also affected by lifestyle and environmental factors (48, 49). Interestingly, the concentration of bacteria increases along the gastrointestinal tract with 103–104 bacteria/mL in the small intestine to 1011 bacteria/mL in the colon, mirroring the distribution of cancer risk along the human intestinal tract (50). Similarly, in the ApcMin/+;Il10−/− model, age strongly correlates with small intestine tumor numbers while inflammation and bacteria composition play a strong role in colon tumorigenesis. Elucidating the mechanisms by which specific bacteria interact with other microbiota members to promote carcinogenesis will generate important insights into the pathophysiology of colorectal cancer.
Conception and design: S. Tomkovich, Y. Yang, M. Mohamadzadeh, E. Oswald, C. Jobin
Development of methodology: S. Tomkovich, Y. Yang, P. Martin, G.P. Wang, C. Jobin
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Tomkovich, Y. Yang, J. Gauthier, X. Sun, X. Liu, C. Jobin
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Tomkovich, Y. Yang, K. Winglee, A.A. Fodor, C. Jobin
Writing, review, and/or revision of the manuscript: S. Tomkovich, Y. Yang, K. Winglee, M. Mohamadzadeh, G.P. Wang, E. Oswald, A.A. Fodor, C. Jobin
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Gauthier
Study supervision: C. Jobin
Other (scored and evaluated histology): M. Mühlbauer
Other (financial contribution): C. Jobin
The authors would like to thank the University of Florida Animal Care Services, particularly the Germ-Free Services division for assistance with SPF and gnotobiotic mouse experiments and the staff at the Molecular Pathology Core for assistance with histology. We also thank Dr. Emma Allen-Vercoe for sharing the F. nucleatum strains. Finally, we thank Dr. Elena Verdu (McMaster University) for derivation of Apcmin/+ mice into germ-free conditions.
This research was supported by NIH grants R01DK047700, R01DK073338, and R21 CA195226 to C. Jobin. Y. Yang was supported by the Crohn's & Colitis Foundation of America (CCFA) research fellowship award (CCFA Ref. #409472). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
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.