Paired microbiome and metabolome analyses associate bile acid changes with colorectal cancer progression

Abstract

Colorectal cancer (CRC) is driven by genomic alterations in concert with dietary influences, with the gut microbiome implicated as an effector in disease development and progression. While meta-analyses have provided mechanistic insight into patients with CRC, study heterogeneity has limited causal associations. Using multi-omics studies on genetically controlled cohorts of mice, we identify diet as the major driver of microbial and metabolomic differences, with reductions in α diversity and widespread changes in cecal metabolites seen in high-fat diet (HFD)-fed mice. In addition, non-classic amino acid conjugation of the bile acid cholic acid (AA-CA) increased with HFD. We show that AA-CAs impact intestinal stem cell growth and demonstrate that Ileibacterium valens and Ruminococcus gnavus are able to synthesize these AA-CAs. This multi-omics dataset implicates diet-induced shifts in the microbiome and the metabolome in disease progression and has potential utility in future diagnostic and therapeutic developments.

Keywords: CP: Cancer; CP: Microbiology; bile acids; colorectal cancer; conjugated bile acids; high-fat diet; metabolome; microbiome.

Figure 1.. Genetics and diet reshape the gut microbiome

(A) α diversity of wild-type (WT) and APC^min/+ mice maintained on normal chow diet (ND) and high-fat diet (HFD). Within-sample diversity is measured by Faith’s phylogenetic diversity. Metrics from shotgun metagenomics sequencing data of cecum samples are presented by genotype-diet combination. Whiskers represent 1.5x interquartile range of data. (B) Unweighted Unifrac measures of b diversity in mice from (A). Metrics from shotgun metagenomic sequencing data are stratified by genotype and diet factors and visualized using principal-coordinate analysis (PCoA). (C and D) Ultrametric phylogenetic tree generated from shotgun metagenomics data of cecum samples in mice from (A). Microbial features colored by Songbird genotype (C) and diet (D) differentials. Red indicates positive association, while blue indicates negative association (both relative to all other features). (E and F) The differential rank plot of selected microbial features separating samples by genotype (E) and diet (F). Features in red correspond to those in the numerator, while those in blue correspond to features in the denominator. Features that are colored gray are not factored into the log ratio calculations. (G) Log ratios of selected microbial families separating samples across genotype (left) and diet (right). Family selection was performed by using Qurro to inspect differentially abundant microbial groups according to Songbird differentials. p values calculated from two-sided t test.

Figure 2.. Genetics and diet affect serum and fecal metabolomes

(A) Principal-component analysis (PCA) of cecum metabolites from WT and APC^min/+ mice maintained on ND and HFD. (B) Partial least square-discriminant analysis (PLS-DA) score plots of cecum (left) and serum metabolites (right) from mice in (A), with a model p <0.01. (C and D) Songbird differential rank plots of the association of metabolites with genotype (left) and diet (right). Differentials were calculated with multinomial regression and validated by comparison to a null model. See also Figures S2–S4.

Figure 3.. Genetics and diet affect fecal bile acids

WT and APC^min/+ mice were maintained on ND or HFD from 4 weeks of age. (A and B) Total fecal bile acids of 12 weeks old mice after 8 weeks of ND or HFD (A) and proportions of primary and secondary bile acids in feces (B). Each data point represents samples pooled from 5 mice (one cage) (n = 15 per arm). (C and D) Progressive changes in bacterially mediated conversion of tauro-cholic acid (T-CA) to CA and deoxycholic acid (DCA) in serum (C) and feces (D) from APC^min/+ mice. Serum samples were collected from mice at the indicated time (n = 3 per time point). Fecal data points represent samples pooled from 5 mice (n = 5 per time point). (E) Temporal changes in intestinal tumor burden (n = 3) and fecal DCA levels in APC^min/+ mice on ND (n = 3–5). (F) Temporal changes in bacterial load and DCA in feces from APC^min/+ mice on ND (pooled sample, n = 3–5). (G) Temporal changes of fecal DCA levels in WT (left panel) and APC^min/+ mice (right panel) maintained on ND and HFD during tumor progression (16–24 weeks) (each data point represents samples pooled from 5 mice (one cage), n = 5). (H) Fecal bile acid levels in WT and APC^min/+ mice on ND and HFD treated with the FXR agonist FexD (50 mg/kg/day) or vehicle for 8 weeks (n = 6 per arm). Data represent the mean ± SEM. For two-group comparation, Student’s unpaired t test was used. For more than two-group comparation, one-way ANOVA was used. *, #p < 0.05; **, ##p < 0.01; ***, ###p < 0.005. See also Figure S5.

Figure 4.. Non-classic amino acid-conjugated bile acids in cecum sample

(A) MS/MS spectra network analysis of the detected 7 novel bile acids. Chemical structure and molecular weight are presented. (B) MMvec microbe-metabolite co-occurrences study of tumor progression in APC^min/+ mice on ND (adenoma) and HFD (adenocarcinoma). Conditional probabilities exhibit a biclustering pattern between bile acids and gOTUS corresponding to ND and HFD. Connections between microbes and metabolites correspond to increased or decreased co-occurrence probability relative to all other microbes. Association was assigned by comparing cluster features to both metagenomic and metabolomic Songbird differentials. (C) Biplot of MMvec results from APC^min/+ mice. Points represent metabolites, and arrows represent most informative microbial features. Color of points corresponds to the Songbird-calculated association of each metabolite with the HFD compared with the ND. Novel bile acids are highlighted with different colors. Spearman correlation between PC1 of the MMvec ordination and the HFD differential was 0.77. See also Figure S6.

Figure 5.. Biological activity of non-classic conjugated Bas

(A) Dose-dependent activation of exogenous mouse FXR by amino acid-conjugated CA species. Luciferase activity in HEK293 cells expressing a luciferase reporter gene functionally linked to an FXR-responsive element (FXRE-Luc). n = 8 per concentration. (B) Dose-dependent activation of exogenous mouse TGR5 by amino acid-conjugated CA species. Luciferase activity in HEK293 cells expressing a luciferase reporter gene functionally linked to a cAMP-responsive element that is downstream of TGR5. n = 8 per concentration. (C) Dose-dependent proliferation of intestinal organoids from WT mice treated with CA (left), Ala-CA (center), and Trp-CA (right), measured by luminescent cell viability assay. n = 6 per concentration. (D) Relative expression of FXR and TGR5 target genes and intestinal stem cell marker genes in intestinal organoids from WT mice treated with amino acid-conjugated cholic acid species (10 μM for 24 h, n = 3 per treatment). (E) Cellular transport of amino acid-conjugated cholic acid species, as determined by the efflux ratio in Caco2 cells. Atenolol and propranolol serve as negative and positive controls, respectively. Digoxin serves as a positive control for P-glycoprotein-mediated efflux (n = 2). Data represent the mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.005. Student’s unpaired t test. See also Figure S7.

Figure 6.. Bacterial synthesis of non-classic conjugated Bas

(A) Dose-dependent generation of conjugated CA species in anaerobic cultures of cecal bacteria from WT and APC^min/+ mice on ND and HFD. Cultures were supplemented with increasing concentrations of CA for 48 h prior to mass spectral analysis. (B) Dose-dependent generation of conjugated cholic acid species in anaerobic cultures of Ruminococcus gnavus (left panel), Clostridium scindens (middle panel), and Ileibacterium valens (right panel). Cultures were supplemented with increasing concentrations of CA for 48 h prior to mass spectral analysis. (C) Dose-dependent generation of conjugated CA species in anaerobic cultures of Ileibacterium valens. Cultures were supplemented with increasing concentrations of T-CA for 48 h prior to mass spectral analysis. (D) Changes in Ileibacterium valens, Ruminococcus gnavus, and Clostridium scindens levels in ND- and HFD-fed APC^min/+ mice treated with FexD or vehicle for 8 weeks, determined by qPCR. Data represent the mean ± SEM. For two-group comparation, Student’s unpaired t test was used. For more than two-group comparation, one-way ANOVA was used. *, #p < 0.05; **, ##p < 0.01; ***, ###p < 0.005. See also Figure S8.