Functional Changes in the Gut Microbiome Contribute to Transforming Growth Factor β-Deficient Colon Cancer

Abstract

Colorectal cancer (CRC) is one of the most treatable cancers, with a 5-year survival rate of ~64%, yet over 50,000 deaths occur yearly in the United States. In 15% of cases, deficiency in mismatch repair leads to null mutations in transforming growth factor β (TGF-β) type II receptor, yet genotype alone is not responsible for tumorigenesis. Previous work in mice shows that disruptions in TGF-β signaling combined with Helicobacter hepaticus cause tumorigenesis, indicating a synergistic effect between genotype and microbial environment. Here, we examine functional shifts in the gut microbiome in CRC using integrated -omics approaches to untangle the role of host genotype, inflammation, and microbial ecology. We profile the gut microbiome of 40 mice with/without deficiency in TGF-β signaling from a Smad3 (mothers against decapentaplegic homolog-3) knockout and with/without inoculation with H. hepaticus. Clear functional differences in the microbiome tied to specific bacterial species emerge from four pathways related to human colon cancer: lipopolysaccharide (LPS) production, polyamine synthesis, butyrate metabolism, and oxidative phosphorylation (OXPHOS). Specifically, an increase in Mucispirillum schaedleri drives LPS production, which is associated with an inflammatory response. We observe a commensurate decrease in butyrate production from Lachnospiraceae bacterium A4, which could promote tumor formation. H. hepaticus causes an increase in OXPHOS that may increase DNA-damaging free radicals. Finally, multiple bacterial species increase polyamines that are associated with colon cancer, implicating not just diet but also the microbiome in polyamine levels. These insights into cross talk between the microbiome, host genotype, and inflammation could promote the development of diagnostics and therapies for CRC. IMPORTANCE Most research on the gut microbiome in colon cancer focuses on taxonomic changes at the genus level using 16S rRNA gene sequencing. Here, we develop a new methodology to integrate DNA and RNA data sets to examine functional shifts at the species level that are important to tumor development. We uncover several metabolic pathways in the microbiome that, when perturbed by host genetics and H. hepaticus inoculation, contribute to colon cancer. The work presented here lays a foundation for improved bioinformatics methodologies to closely examine the cross talk between specific organisms and the host, important for the development of diagnostics and pre/probiotic treatment.

Keywords: Helicobacter hepaticus; Smad3; bioinformatics; butyrate; colon cancer; gut inflammation; gut microbiome; host-pathogen interactions; metagenomics; metatranscriptomics; polyamines.

FIG 1

Materials and methods flow chart. (A1) Four groups of mice were used for DNA/RNA extraction: a mouse that was Smad3^+/+ and H. hepaticus negative (S+H−) (top left), a mouse that was Smad3^−/− and H. hepaticus negative (S−H−) (top right), a mouse that was Smad3^+/+ and H. hepaticus positive (S+H+) (bottom left), and a mouse that was Smad3^−/− and H. hepaticus positive (S−H+) (bottom right, three panels from one histological section). These three panels display the types of lesions found in the S−H+ mice: hyperplasia, adenoma, and mucinous carcinoma. Inflammatory infiltrates are indicated by arrowheads. Four out of 10 mice in the S−H+ group had tumors at the time of sacrifice. The three double-headed red arrows indicate the Smad3^−/−, H. hepaticus, and combined effects. (A2) Dissection of mouse large intestine from the S−H+ group showing tumor locations in dotted red lines and site of cecal matter removal. (B1) Flow chart showing number of reads at each step of DNA analysis. Reads were aligned to PATRIC (48) with Taxoner64 (95). The last step of the DNA analysis feeds into the second step of the RNA analysis. (B2) Flow chart showing number of reads at each step of RNA analysis. Reads were aligned with Bowtie2 (97) and quantified/normalized with cuffquant/cuffnorm, part of the Cufflinks suite of tools (99).

FIG 2

Sum of estimated expression by KEGG pathway. Bubble chart of RNA count differences between sample groups among various KEGG pathways (102). The scale at bottom shows estimated RNA counts as outputted by cuffquant and cuffnorm (part of the Bowtie2/Cufflinks suite of tools). Labels at top show the sample groups. The heat map behind the bubbles shows log₂ fold changes (log₂ FC) of pathways for each of the effects. Fold changes are mapped to a blue-yellow color spectrum with bright yellow having the greatest increase in RNA counts and bright blue having the greatest decrease in RNA counts under the given condition versus control. Color behind the control bubbles represents no change and is provided for reference.

FIG 3

Species contribution to genes in butanoate (butyrate) metabolism. RNA count changes for buk (butyrate kinase) (EC 2.7.2.7), categorized by species/strain contribution. The y axis shows RNA counts, and the x axis shows sample groups. Species with majority contributions to count bars are numbered for clarity. Bottom right shows partial pathway with a red box to highlight gene. Partial pathway adapted from KEGG pathway database (102).

FIG 4

Taxonomic changes at phylum, family, and species levels. Changes in fractions of total bacterial counts at different taxonomic levels for the S+H− (Control), S−H− (Smad3^−/−), S+H+ (H. hepaticus only), and S−H+ (Combined) groups. (A) Changes for the phyla of Bacteroidetes, Deferribacteres, Firmicutes, and Proteobacteria. (B) Changes for the families of Bacteroidaceae, Deferribacteraceae, Lachnospiraceae, Lactobacillaceae, and Helicobacteraceae. (C) Changes for the species of Mucispirillum schaedleri, Lachnospiraceae bacterium A4, Lactobacillus plantarum, Lactobacillus murinus, Parabacteroides distasonis, and Helicobacter hepaticus.

FIG 5

Species contribution to genes in arginine and proline metabolism. RNA count changes for aguB (N-carbamoylputrescine amidase) (EC 3.5.1.53) and nspC (carboxynorspermidine decarboxylase) (EC 4.1.1.96), categorized by species/strain contribution. The y axis shows RNA counts, and the x axis shows sample groups. Species with majority contributions to count bars are numbered for clarity. Bottom right shows partial pathway with a red box to highlight genes. Partial pathway adapted from KEGG pathway database (102).

FIG 6

Species contribution to lpxC and lpxD in LPS biosynthesis. RNA count changes for lpxC [UDP-3-O-(3-hydroxymyristoyl) N-acetylglucosamine deacetylase] (EC 3.5.1.108) and lpxD [UDP-3-O-(3-hydroxymyristoyl) glucosamine N-acyltransferase] (EC 2.3.1.191), categorized by species/strain contribution. The y axis shows RNA counts, and the x axis shows sample groups. Species with majority contributions to count bars are numbered for clarity. Bottom right shows partial pathway with a red box to highlight genes. Partial pathway adapted from KEGG pathway database (102).

FIG 7

Correlation of mouse gene expression with bacterial gene expression. RNA expression correlation between mouse TLR genes and bacterial genes in the LPS pathway. Expression in arbitrary units due to normalization. Mouse gene expression based on qRT-PCR data; bacterial gene expression based on transcriptome sequencing from this study. P values: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

FIG 8

Species contribution to genes in oxidative phosphorylation. RNA count changes for nuoA to nuoN (NADH ubiquinone oxidoreductase A to N) (EC 1.6.5.3) and ppk (polyphosphate kinase) (EC 2.7.4.1), categorized by species/strain contribution. The y axis shows RNA counts, and the x axis shows sample groups. Species with majority contributions to count bars are numbered for clarity. Dividing lines within species represent multiple subunits for that gene. Bottom right shows partial pathway with a red box to highlight genes. Partial pathway adapted from KEGG pathway database (102).