Characterization of Mucosal Dysbiosis of Early Colonic Neoplasia

Abstract

Aberrant crypt foci (ACF) are the earliest morphologically identifiable lesions in the colon that can be detected by high-definition chromoendoscopy with contrast dye spray. Although frequently associated with synchronous adenomas, their role in colorectal tumor development, particularly in the proximal colon, is still not clear. The goal of this study was to evaluate the profile of colon-adherent bacteria associated with proximal ACF and to investigate their relationship to the presence and subtype of synchronous polyps present throughout the colon. Forty-five subjects undergoing a screening or surveillance colonoscopy were included in this retrospective study. Bacterial cells adherent to the epithelia of ACF and normal mucosal biopsies were visualized by in situ hybridization within confocal tissue sections. ACF showed significantly greater heterogeneity in their bacterial microbiome profiles compared with normal mucosa. One of the bacterial community structures we characterized was strongly correlated with the presence of synchronous polyps. Finally, using DNA mass spectrometry to evaluate a panel of colorectal cancer hotspot mutations present in the ACF, we found that three APC gene mutations were positively associated with the presence of Instestinibacter sp., whereas KRAS mutations were positively correlated with Ruminococcus gnavus. This result indicates a potential relationship between specific colon-associated bacterial species and somatically acquired CRC-related mutations. Overall, our findings suggest that perturbations to the normal adherent mucosal flora may constitute a risk factor for early neoplasia, demonstrating the potential impact of mucosal dysbiosis on the tissue microenvironment and behavior of ACF that may facilitate their progression towards more advanced forms of neoplasia.

Keywords: Cancer prevention; Predictive markers.

Fig. 1

Direct visualization of tightly colon-associated bacteria within the colonic mucosa. Proximal ACF and adjacent normal mucosa were directly examined for the presence of colon-associated bacteria using 16S universal FISH probes. Clusters of bacteria were observed within the mucous layer on epithelium. Red: bacteria (EUB338-cy3 probe), Magenta: E-cadherin, Green; Mucin-2, Blue: DAPI. Bacteria are marked with white arrows. Bacteria were observed within the mucous layer associated with the epithelium

Fig. 2

β-Diversity comparison of the microbiome between normal mucosa and ACF based on thetaYC distance. In polyp-free subjects, principle coordinate analysis (PCoA) plots showed no significant spatial separation as determined by AMOVA between proximal ACF lesions and control mucosa taken from both ACF-free control and paired normal samples from the same ACF subject a. The distance measured pairwise between each sample, however, showed significant differences in microbiome community structure b. In polyp subjects, PCoA plots showed no significant spatial separation tested using AMOVA between proximal ACF lesions and control mucosa from both ACF-free controls and paired normal samples taken from the same ACF subject c. However, ACF samples from patients with polyps were better rsolved from control samples (blue) compared with that in a. In addition, the distance measured pairwise between each sample showed significant differences in microbiome community structure d, with more significant p-values from b when comparing the ACF site from polyp subjects (purple) to control group (blue) using Mann–Whitney test

Fig. 3

Significantly abundant bacterial taxa by DNA mutation types. Intestinibacter sp. was significantly increased in ACF samples with an APC mutation, whereas R. gnavus was significantly increased in ACF samples with a KRAS mutation compared with samples with no mutation detected

Fig. 4

Microbiome profiles of the entire sample set. Heat maps were generated using Morpheus based on the top 100 OTUs found in the biopsy specimens. Samples are shown in columns, while each OTU is depicted in the rows. Only significantly different taxa in each cluster are shown. The color scale appears on the top right side of the figure. Unsupervised hierarchical clustering (complete linkage) shows two clusters, Microbiome Cluster A (cyan) and Microbiome Cluster B (magenta). Annotation for the presence of a polyp(s) in the subject, gender, and obesity, as well as the presence of a polyp(s) in the ascending colon are indicated by unique shapes when positive. Significance was determined after a Benjamini–Hochberg multiple comparison adjustment

Fig. 5

Microbiome signature shapes the development of polyps. Two distinct microbiome clusters (Microbiome Clusters A and B) defined by bacterial community signature were depicted in the Sankey diagram, demonstrating the “flow” of each biopsy sample. Microbiome Cluster B is associated with precancerous lesions. Over 90% of samples with Microbiome Cluster B yielded malignant polyps in the same subject. Less than 50% of samples with Microbiome Cluster A were associated with the presence of any polyp in the same subject (Top). Bacterial relative abundances in each category were depicted in the bar graph comparing groups from the Sankey diagram above. There were bacterial community differences by Microbiome Cluster type, ACF/polyp status (no lesion, ACF, and ACF with polyp), and the specific type of polyp present in the subjects. Microbiome Clusters A and B were significantly correlated with demographic or clinical characteristics, such as obesity status, gender, acetaminophen, and calcium intake. Antibiotic intake was not significantly correlated with Microbiome Clusters A and B (Table 1)

Fig. 6

Predictive functional profiling annotation between Microbiome Clusters A and B. 16S rRNA gene sequencing data were clustered as OTUs with 97% identity and predictive functional profiling was performed via PICRUSt using KEGG KO as a reference. Several virulence factors were predicted as functional pathways associated with Microbiome Cluster B (green)