Manipulation of the Gut Microbiota Reveals Role in Colon Tumorigenesis

Abstract

There is growing evidence that individuals with colonic adenomas and carcinomas harbor a distinct microbiota. Alterations to the gut microbiota may allow the outgrowth of bacterial populations that induce genomic mutations or exacerbate tumor-promoting inflammation. In addition, it is likely that the loss of key bacterial populations may result in the loss of protective functions that are normally provided by the microbiota. We explored the role of the gut microbiota in colon tumorigenesis by using an inflammation-based murine model. We observed that perturbing the microbiota with different combinations of antibiotics reduced the number of tumors at the end of the model. Using the random forest machine learning algorithm, we successfully modeled the number of tumors that developed over the course of the model on the basis of the initial composition of the microbiota. The timing of antibiotic treatment was an important determinant of tumor outcome, as colon tumorigenesis was arrested by the use of antibiotics during the early inflammation period of the murine model. Together, these results indicate that it is possible to predict colon tumorigenesis on the basis of the composition of the microbiota and that altering the gut microbiota can alter the course of tumorigenesis. IMPORTANCE Mounting evidence indicates that alterations to the gut microbiota, the complex community of bacteria that inhabits the gastrointestinal tract, are strongly associated with the development of colorectal cancer. We used antibiotic perturbations to a murine model of inflammation-driven colon cancer to generate eight starting communities that resulted in various severities of tumorigenesis. Furthermore, we were able to quantitatively predict the final number of tumors on the basis of the initial composition of the gut microbiota. These results further bolster the evidence that the gut microbiota is involved in mediating the development of colorectal cancer. As a final proof of principle, we showed that perturbing the gut microbiota in the midst of tumorigenesis could halt the formation of additional tumors. Together, alteration of the gut microbiota may be a useful therapeutic approach to preventing and altering the trajectory of colorectal cancer.

Keywords: 16S rRNA gene sequencing; azoxymethane; colorectal cancer; dextran sodium sulfate; microbial ecology; microbiome; murine models.

FIG 1

Antibiotic perturbation drives changes in microbial community structure and the final tumor burden. The AOM-DSS model was administered to C57BL/6 mice reared under SPF conditions with different antibiotic perturbations that were applied during the period covered by each of the rectangles; The arrowheads indicate the times when fecal samples that were used for our analysis were obtained (A). The mice were treated with all of the possible combinations of metronidazole, streptomycin, and vancomycin to create eight treatment groups, which resulted in a continuum of tumor burdens in the mice (B to D). The stars indicate which treatments yielded a significantly (P < 0.05) different number of tumors compared to the treatment with the vertical line. The antibiotic treatments resulted in variation in the taxonomic structure of the communities at the start of the model (day 0) (D). The two-dimensional nonmetric multidimensional scaling ordination had a stress of 0.20 and explained 84.0% of the variation in the distances (E).

FIG 2

A random forest model successfully predicted the number of tumors in the mice at the end of the model on the basis of their microbiota composition at the start of the model. The model included 12 OTUs and explained 67.7% of the variation in the data.

FIG 3

Relationship between the initial relative abundance of the most informative OTUs from the random forest model and the number of tumors found in the mice at the end of the model. The vertical gray line indicates the limit of detection. Panels are ordered in decreasing order of the percent increase in the mean square error (MSE) of the model when that OTU was removed. The color and shape of the plotting symbols correspond to those used in Fig. 2.

FIG 4

The murine microbiota is dynamic, but the amount of change is not associated with the final number of tumors. The structure of the gut microbiota of the untreated, Δmetronidazole-treated (open red circles), and Δvancomycin-treated (open blue circles) mice changed the most throughout the model as measured with the θ_YC distance metric (A). OTUs 1 and 2 were among the most dynamic OTUs across all of the treatment groups; here we depict the change in their relative abundance across the model in those treatment groups that experienced the greatest overall change in community structure (B). The plotting symbols and characters are the same as those used in Fig. 1. In panel B, the median relative abundance is indicated by the plotting symbol and the range of observed relative abundances is plotted by the vertical bar. The vertical blue regions indicate when the DSS treatments were applied.

FIG 5

Antibiotic intervention prior to a second administration of DSS alleviates the tumor burden. Interventions with an antibiotic cocktail of metronidazole, vancomycin, and streptomycin were performed as depicted in Fig. 1A, and enumeration of tumors was performed at the endpoint of the model (A). Representative images of tumors in the distal colons of mice from each treatment group (B).