Shifts in the Gut Metabolome and Clostridium difficile Transcriptome throughout Colonization and Infection in a Mouse Model

Abstract

Antibiotics alter the gut microbiota and decrease resistance to Clostridium difficile colonization; however, the mechanisms driving colonization resistance are not well understood. Loss of resistance to C. difficile colonization due to antibiotic treatment is associated with alterations in the gut metabolome, specifically, with increases in levels of nutrients that C. difficile can utilize for growth in vitro. To define the nutrients that C. difficile requires for colonization and pathogenesis in vivo, we used a combination of mass spectrometry and RNA sequencing (RNA Seq) to model the gut metabolome and C. difficile transcriptome throughout an acute infection in a mouse model at the following time points: 0, 12, 24, and 30 h. We also performed multivariate-based integration of the omics data to define the signatures that were most important throughout colonization and infection. Here we show that amino acids, in particular, proline and branched-chain amino acids, and carbohydrates decrease in abundance over time in the mouse cecum and that C. difficile gene expression is consistent with their utilization in vivo. This was also reinforced by the multivariate-based integration of the omics data where we were able to discriminate the metabolites and transcripts that support C. difficile physiology between the different time points throughout colonization and infection. This report illustrates how important the availability of amino acids and other nutrients is for the initial stages of C. difficile colonization and progression of disease. Future studies identifying the source of the nutrients and engineering bacteria capable of outcompeting C. difficile in the gut will be important for developing new targeted bacterial therapeutics. IMPORTANCE Clostridium difficile is a bacterial pathogen of global significance that is a major cause of antibiotic-associated diarrhea. Antibiotics deplete the indigenous gut microbiota and change the metabolic environment in the gut to one favoring C. difficile growth. Here we used metabolomics and transcriptomics to define the gut environment after antibiotics and during the initial stages of C. difficile colonization and infection. We show that amino acids, in particular, proline and branched-chain amino acids, and carbohydrates decrease in abundance over time and that C. difficile gene expression is consistent with their utilization by the bacterium in vivo. We employed an integrated approach to analyze the metabolome and transcriptome to identify associations between metabolites and transcripts. This highlighted the importance of key nutrients in the early stages of colonization, and the data provide a rationale for the development of therapies based on the use of bacteria that specifically compete for nutrients that are essential for C. difficile colonization and disease.

Keywords: Clostridium difficile; amino acids; intestinal colonization; metabolomics; peptides; transcriptomics.

FIG 1

Cecal metabolome during C. difficile colonization and infection. (A) Variable-importance plot of the top 50 metabolites identified by Random Forest analysis. The mean accuracy value decrease is a measure of how much predictive power is lost if a given metabolite is removed or permuted in the Random Forest algorithm; thus, the more important a metabolite is to classifying samples into time point categories, the further to the right its point is on the graph. Metabolite points are color-coded according to the KEGG superpathway in which they belong. Metabolite names are labeled red if their level increased throughout infection, black if they were variable, and green if the level decreased. (B) Heat map showing the relative abundances of the metabolites identified in panel A. Each column corresponds to the cecal metabolome from an individual mouse, and each row corresponds to a given metabolite. Unsupervised hierarchical clustering was used to cluster metabolites with similar abundance profiles over time. The heat map scale ranges from −3 to 3 on a log₂ scale.

FIG 2

C. difficile transcriptome during colonization and infection. (A) Venn diagram showing the differentially expressed genes that were shared or unique between the three time points. (B to D) Volcano plots highlighting genes whose transcript levels changed by greater than 2-fold and met the significance threshold P adj. = <0.05. Genes highlighted in red had increased transcript levels, while those highlighted in green had decreased levels. Points in black represent genes whose results failed to meet the significance threshold.

FIG 3

KEGG pathway analysis of the C. difficile DEGs throughout colonization and infection. Protein sequences of the DEGs at 24 h (A) or 30 h (B) relative to 12 h were imported into Blast2GO, and data corresponding to the predicted enzymes were loaded onto KEGG pathway maps. The numbers on the x axis correspond to the number of predicted enzymes used to map to a given pathway and whether the enzyme’s transcript was increased or decreased in expression.

FIG 4

Multivariate-based analysis of the gut metabolome and C. difficile transcriptome during colonization and infection. A loading plot of the features selected in each component is provided. The top row indicates the features in the first component for the metabolites (left) and transcripts (right). The bottom row indicates the features in the second component for the metabolites (left) and transcripts (right). The values corresponding to the specific bar magnitudes are indicated in Table S5. The color indicates the expression levels of each variable according to each class where blue represents 12 h, orange represents 24 h, and gray represents 30 h.

FIG 5

Plot of the correlations between the metabolome and C. difficile transcriptome. A Circos plot displays the positive and negative correlations (r = >0.7) between the selected features with blue and red lines, respectively. The values corresponding to the exact weight for each line are indicated in Table S5. The metabolites are indicated in purple (top right quadrant), and the transcripts are indicated in green. Each individual feature name is labeled in the block. The outer lines indicate the expression levels of each variable according to each class where blue represents 12 h, orange represents 24 h, and gray represents 30 h. CDS, coding sequence; PLP, proteolipid protein.