The gut microbiota in conventional and serrated precursors of colorectal cancer

Abstract

Background: Colorectal cancer is a heterogeneous disease arising from at least two precursors-the conventional adenoma (CA) and the serrated polyp. We and others have previously shown a relationship between the human gut microbiota and colorectal cancer; however, its relationship to the different early precursors of colorectal cancer is understudied. We tested, for the first time, the relationship of the gut microbiota to specific colorectal polyp types.

Results: Gut microbiota were assessed in 540 colonoscopy-screened adults by 16S rRNA gene sequencing of stool samples. Participants were categorized as CA cases (n = 144), serrated polyp cases (n = 73), or polyp-free controls (n = 323). CA cases were further classified as proximal (n = 87) or distal (n = 55) and as non-advanced (n = 121) or advanced (n = 22). Serrated polyp cases were further classified as hyperplastic polyp (HP; n = 40) or sessile serrated adenoma (SSA; n = 33). We compared gut microbiota diversity, overall composition, and normalized taxon abundance among these groups. CA cases had lower species richness in stool than controls (p = 0.03); in particular, this association was strongest for advanced CA cases (p = 0.004). In relation to overall microbiota composition, only distal or advanced CA cases differed significantly from controls (p = 0.02 and p = 0.002). In taxon-based analysis, stool of CA cases was depleted in a network of Clostridia operational taxonomic units from families Ruminococcaceae, Clostridiaceae, and Lachnospiraceae, and enriched in the classes Bacilli and Gammaproteobacteria, order Enterobacteriales, and genera Actinomyces and Streptococcus (all q < 0.10). SSA and HP cases did not differ in diversity or composition from controls, though sample size for these groups was small. Few taxa were differentially abundant between HP cases or SSA cases and controls; among them, class Erysipelotrichi was depleted in SSA cases.

Conclusions: Our results indicate that gut microbes may play a role in the early stages of colorectal carcinogenesis through the development of CAs. Findings may have implications for developing colorectal cancer prevention therapies targeting early microbial drivers of colorectal carcinogenesis.

Keywords: Adenoma; Cancer; Colorectal; Microbiome; Microbiota; Polyp; Serrated.

Figures

Fig. 1

α-Diversity and community types of colonoscopy-screened participants. a Violin plots of species richness and b Shannon diversity index by polyp histology (controls n = 322, CA cases n = 144, HP cases n = 40, SSA cases n = 33), location (distal CA n = 55, proximal CA n = 87), and advancement level (non-advanced CA n = 121, advanced CA n = 22). These indices were calculated for 500 iterations of rarefied (4000 sequences per sample) OTU tables, and the average over the iterations was taken for each participant (1 control excluded due to sequencing depth = 2088). p values from multiple linear regression are shown. c Fitting to the DMM [24, 56] model indicates optimal classification into 5 community types. d Principal coordinate analysis of Jensen-Shannon divergence values between participants, colored by community type. Green community type 1, blue type 2, purple type 3, yellow type 4, red type 5. e Distribution of the community types in groups distinguished by histology, f location, or g advancement level. p value from Fisher’s exact test with Monte Carlo simulation is shown. CAs conventional adenomas, HPs hyperplastic polyps, SSAs sessile serrated adenomas

Fig. 2

Heatmaps of OTUs that were differentially abundant between colorectal polyp cases and controls. All OTUs with q < 0.10 for comparisons of any case group (all CA, non-advanced CA, advanced CA, distal CA, proximal CA, HP, SSA) vs. controls are included in the figure. a Heatmap shows fold change from controls in the DESeq2 models, with white star indicating q < 0.10 for the comparison. b Heatmap shows OTU counts in each participant. For display, counts were normalized for DESeq2 size factors and log2 transformed after adding a pseudocount of 1. n = 1 and n = 2 CA cases were missing advanced status or location information, respectively. CAs conventional adenomas, HPs hyperplastic polyps, SSAs sessile serrated adenomas

Fig. 3

Microbial community ecology in controls and conventional adenoma cases. Correlation network of OTUs differentially abundant between a controls and all CA cases, b controls and distal CA cases, and c controls and proximal CA cases. Spearman’s correlation coefficients were estimated using counts (normalized for DESeq2 size factors) and calculated among the samples under comparison. Lines shown between OTUs indicate Spearman’s correlation ≥0.3 (green) or ≤-0.3 (red). Direction of enrichment in relation to all CA/distal CA/proximal CA cases vs. controls was determined from DESeq2 models. OTUs are colored according to family membership. Line thickness represents strength of the correlation, in steps of 0.3–0.4 (thinnest), 0.4–0.5, 0.5–0.6, 0.6–0.7, and >0.7 (thickest). CA conventional adenoma