Metabolic landscape of the male mouse gut identifies different niches determined by microbial activities

Abstract

Distinct niches of the mammalian gut are populated by diverse microbiota, but the contribution of spatial variation to intestinal metabolism remains unclear. Here we present a map of the longitudinal metabolome along the gut of healthy colonized and germ-free male mice. With this map, we reveal a general shift from amino acids in the small intestine to organic acids, vitamins and nucleotides in the large intestine. We compare the metabolic landscapes in colonized versus germ-free mice to disentangle the origin of many metabolites in different niches, which in some cases allows us to infer the underlying processes or identify the producing species. Beyond the known impact of diet on the small intestinal metabolic niche, distinct spatial patterns suggest specific microbial influence on the metabolome in the small intestine. Thus, we present a map of intestinal metabolism and identify metabolite-microbe associations, which provide a basis to connect the spatial occurrence of bioactive compounds to host or microorganism metabolism.

Fig. 1. Biogeography of the male SPF mouse metabolome.

a, Hierarchical clustering analysis of metabolite abundances from luminal content samples of male SPF mice. Abundances for all 128 quantified metabolites are shown as z score normalized concentrations, averaged from five mice, across the 15 sampling sites. Clustering based on Euclidian distance identified three main clusters corresponding to the three main physiological regions of the digestive tract: stomach, small intestine and large intestine. Metabolites are colour-coded according to MetaCyc (bottom right). b, PCA of metabolite concentrations from individual male SPF mice based on all 150 luminal and mucus samples covering the entire intestine. PCA was performed on z score normalized metabolite concentrations, with five mice per sample. Colours indicate sites grouped by intestinal region, symbols indicate lumen or mucus and the large intestine content cluster is highlighted in grey. c, Differential analysis of metabolite concentrations between luminal content and mucus samples. Concentrations were averaged across all 15 sampling sites for five male mice within the respective habitat. Positive or negative fold changes indicate higher concentrations in lumen or mucus, respectively. P values were calculated using a two-sided paired-sample Student’s t-test with Benjamini–Hochberg correction for multiple testing and are displayed as −log₁₀ transformed. Metabolites with significantly differing concentrations (absolute log₂(fold change) ≥ 1.5, corrected P ≤ 0.05) are coloured according to the MetaCyc classification, as defined in the box below. Lumen and mucus sampling types are represented schematically on the left, aligned to the corresponding parts of the volcano plot. Abbreviations: FC, fold change. Source data

Fig. 2. Metabolome differences between small and large intestine of male SPF mice.

a, Differential analysis of metabolite concentrations between small and large intestinal samples in lumen (upper) and mucus (lower). Data points represent mean fold change values calculated between ten small intestinal and four large intestinal sites, for five male mice. Negative and positive values represent higher concentrations in the small or large intestine, respectively. The y axis displays −log₁₀-transformed P values, calculated using a two-sided paired-sample Student’s t-test with Benjamini–Hochberg correction for multiple testing. Metabolites with significantly different concentrations (absolute log₂(fold change) ≥ 2, corrected P ≤ 0.05) are colour-coded according to the MetaCyc classification, as defined in the box (bottom left). b, Significantly changing metabolites between the small and large intestine in the lumen and mucus from differential analysis in Fig. 1a (absolute log₂(fold change) ≥ 2, corrected P ≤ 0.05). P values were calculated using a two-sided paired-sample Student’s t-test with Benjamini–Hochberg correction for multiple testing. Dot colours denote fold change and dot size denotes significance. Metabolites are classified according to MetaCyc, as defined in the box (bottom left). c,d, Spatial profiles of histidine and tryptophan (c) and glycerate and 4-hydroxyphenylacetate (d) over 15 intestinal sites in SPF mucus or lumen, respectively. Lines with shaded areas indicate the moving average of the mean ± s.e.m. of concentration measurements from five male mice. Abbreviations: cec, caecum; col, colon; duo, duodenum; ile, ileum; jej, jejunum; sto, stomach; TMAO, trimethylamine N-oxide. Source data

Fig. 3. Longitudinal metabolite pattern along the intestine of male SPF mice.

a–c, Example profiles of metabolites with differential concentration in the small intestinal content (a) and mucus (b), or in the luminal content or mucus of the large intestine (c). Lines with shaded areas indicate the moving average of the mean ± s.e.m. of concentration measurements from five male mice. Significantly different metabolite concentrations from one region in the small intestine compared with the neighbouring region are marked by asterisks (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001). Fructose intestinal content concentration in the duodenum versus jejunum: P = 0.0245; riboflavin intestinal content concentrations in the jejunum versus ileum: P = 0.0432; allantoin intestinal content concentrations in the jejunum versus ileum: P = 0.0220; guanine mucus concentrations in the duodenum versus jejunum: P = 0.0005; guanosine mucus concentrations in the duodenum versus jejunum: P = 0.0129; creatine mucus concentrations in the jejunum versus ileum: P = 0.0278. P values were calculated using a two-sided paired-sample Student’s t-test. Abbreviations: sto, stomach; duo, duodenum; jej, jejunum; ile, ileum; cec, caecum; and col, colon. Source data

Fig. 4. Microbiota effect on large intestinal metabolism.

a, Abundance fold changes of 31 metabolites with higher large intestine than small intestine concentrations, thus hypothesized to be of microbial origin (significantly changing metabolites from differential analysis; Fig. 2a,b). The log₂-transformed fold changes for luminal content (left) and mucus (right) were calculated from averaged concentrations for the four large intestinal sampling sites of five male SPF mice versus five male germ-free mice. Boxplots are thus based on 20 data points, the median log₂(fold change) is indicated in red, boxes contain the 25th to 75th percentiles, and the whiskers extend to the most extreme data points not considered outliers. The grey area marks absolute log₂(fold change) ≤ 2. Symbols on the left indicate previous evidence: () common knowledge about certain metabolites, for example see Koh et al. and de Aguiar Vallim et al.^,; (Δ) evidence from Han et al. using the comparison of Swiss-Webster germ-free versus conventional mice to identify differential metabolite abundances; (□) evidence from Matsumoto et al. using germ-free and ‘ex-germ-free’ mice that were inoculated in the stomach with faecal suspension from SPF BALB/c mice to classify metabolites as mouse- or microorganism-derived; () evidence from Marcobal et al. who used germ-free and conventional Swiss-Webster mice to compare metabolite levels; () evidence from Sridharan et al. using reaction network models to predict microbiota-dependent metabolic products. All data used from published work including explanations can be found in Supplementary Table 7. In addition, (◯) denotes metabolites detected exclusively in male SPF mice. For the SPF exclusive metabolites, fold changes cannot be calculated, as indicated by ∞. b,c, Large intestinal concentration of 2-oxoglutarate (b) and hydrocinnamate (c) in SPF and germ-free luminal content and mucus. Solid bars show the mean concentration of measurements from five male mice, averaged over the four large intestinal sites. The 20 corresponding data points are displayed as circles. Abbreviations: FC, fold change; cont, luminal content; GF, germ-free; muc, mucus. Source data

Fig. 5. Microbiota effect on small intestinal metabolism.

a, Heatmap representation of metabolites with at least fourfold higher concentrations in male SPF mice than in male germ-free mice, specifically in one location of the small intestine versus the whole small intestine. The left-hand column shows averaged log₂-transformed fold changes of all ten small intestinal sites (SPF versus germ-free), and the next three columns depict averaged fold changes (SPF versus germ-free) in the duodenum, jejunum or ileum separately. From spatial profiles (as in Fig. 3, not all shown here), we identified the region (duodenum, jejunum or ileum) that varies from its neighbouring region. For that region, marked by an asterisk, we suspect microbial involvement causing the distinct difference, and consequently also a higher SPF over germ-free concentration. b, Spatial profiles of allantoin concentrations in SPF and germ-free luminal content. Lines with a shaded area indicate the moving average of the mean ± s.e.m. of concentration measurements from five male mice. c, Small intestine concentrations of guanine, guanosine and hypoxanthine in SPF and germ-free luminal content and mucus. Solid bars show the mean concentration of measurements from five male mice, over ten small intestinal sites. The 50 corresponding data points are displayed as circles. d, Spatial profiles of creatine concentrations in SPF and germ-free mucus. Lines with a shaded area indicate the moving average of the mean ± s.e.m. of concentration measurements with five male mice. Abbreviations: FC, fold change; GF, germ-free; vs., versus; sto, stomach; duo, duodenum; jej, jejunum; ile, ileum; cec, caecum; col, colon; cont – luminal content; muc – mucus; NaN, not a number (fold changes could not be calculated). Source data

Fig. 6. Metabolite–microbe associations.

a, Venn diagram showing the overlap of 128 measured metabolites classified as host- or microorganism-related, based on PICRUSt2 predictions of microbial metabolic functions and the mouse metabolic network. The majority of metabolites cannot be classified as either host- or microbe-associated, seven metabolites are host-related and thirteen metabolites are microorganism-related. Twenty-seven metabolites did not match any of the metabolic networks. b, Distribution of the correlation coefficients of 126,336 metabolite–microbe pairs. Zoomed-in bars show how applying thresholds on the P value, the SPF–germ-free fold change and the presence of metabolic enzymes reduces the number of predicted functional metabolite–microbe pairs to 148. c, Sankey diagram showing links between unique metabolites and the corresponding microbial orders in positively correlated metabolite–microbe pairs that meet the thresholds defined in b. Metabolites are colour-coded according to MetaCyc. The size of the linkage line denotes number of pairs. d, Spatial profiles of metabolite concentrations and associated microorganisms in SPF lumen samples. Lines with a shaded area indicate the mean ± s.e.m. of concentration measurements with five male mice, and the mean ± s.e.m. of relative microorganism abundance. Abbreviations: duo, duodenum; jej, jejunum; ile, ileum; cec, cecum and col, colon. Source data

Extended Data Fig. 1. Biogeography of male SPF mice.

a – Schematic outlining the 15 sampling sites used in this study: one site in the stomach, 10 sites in the small intestine (three in the duodenum, four in the jejunum and three in the ileum), and four sites in the large intestine (in the cecum and in the ascending, transverse and descending colon). b – Repartition of MetaCyc chemical classes among all 128 metabolites measured with our LC-TOF-MS method used in this study (Supplementary Table ST1). c – Hierarchical clustering analysis of metabolite abundances from intestinal mucus of male SPF mice. Abundances for all 128 quantified metabolites are shown as z-scores of normalized concentrations averaged from five mice, across the 15 sampling sites. The clustering was performed based on Euclidian distances on luminal samples (Fig. 1a) and resulted in the identification of three main clusters, corresponding to the three main known physiological regions of the digestive tract, namely stomach, small and large intestine. In this heatmap here, metabolites along the y-axis are sorted according to the order from Fig. 1a, for comparison. Metabolites in the three main clusters are color-coded according to MetaCyc (as in S1b; Supplementary Table ST1). d – Metabolite concentrations of arginine, ornithine and spermidine in SPF luminal content and mucus large intestinal samples. Plotted are 20 data points (four large intestine sites, five male mice), boxplots show median metabolite concentration with a thicker mark, 25^th and 75^th percentile are denoted by the box, and whiskers extend to the most extreme data points not considered outliers. Source data

Extended Data Fig. 2. Hierarchical clustering of male SPF small intestinal samples.

a, b – Hierarchical clustering analysis of metabolite abundances from 10 small intestinal luminal content (a) and mucus (b) samples of male SPF mice. Abundances of 128 metabolites are shown as z-scores of normalized concentrations averaged from five male mice and three duodenum, four jejunum or three ileum samples, respectively. Hierarchical clustering was performed based on these three averaged values (duodenum, jejunum, ileum). Significantly changing metabolites are marked with boxes and asterisks (p-value ≤ 0.05). P-values were calculated using a two-sided paired-sample Student’s t-test and can be found in Supplementary Table ST4. Color key of the heatmap indicates low to high relative abundance of metabolites throughout the small intestine. c – Distribution of peak concentrations in luminal content (top) and mucus (bottom) small intestine, with respect to the three regions. Three bars denote the three regions duodenum, jejunum and ileum, y-axis shows frequency of peak metabolite concentrations occurring in the respective region. Abbreviations: duo – duodenum, jej – jejunum, and ile – ileum. Source data

Extended Data Fig. 3. Hierarchical clustering of male SPF large intestinal samples.

a, b – Hierarchical clustering analysis of metabolite abundances from four large intestinal luminal content (a) and mucus (b) samples in male SPF mice. Abundances of 128 metabolites are shown as z-scores of normalized concentrations averaged from five male mice per sampling site in the colon (cecum, ascending, transverse and descending colon). Hierarchical clustering was performed only on these four values. Significantly changing metabolites are marked with boxes and asterisks (p-value ≤ 0.05). P-values were calculated using a two-sided paired-sample Student’s t-test. Color key of the heatmap indicates low to high relative abundance of metabolites throughout the large intestine. c – Distribution of peak concentrations in luminal content (top) and mucus (bottom) large intestine, with respect to the four regions is depicted. Four bars denote the regions, y-axis shows frequency of peak metabolite concentrations occurring in the respective region. Abbreviations: cec – cecum, acol – ascending colon, tcol – transverse colon, and dcol – descending colon. Source data

Extended Data Fig. 4. Microbiota effect on large intestinal metabolism.

a – Distribution of metabolite concentration fold changes in the SPF versus germ-free comparison in luminal content and mucus samples. Fold changes are calculated per metabolite in the respective habitat for five male mice and 15 sampling sites, log₂ transformed, and their distribution is displayed as contour plot along the x-axis. The frequency is denoted on the y-axis. A negative log₂ fold change indicates higher metabolite concentration in germ-free mice, a positive log₂ fold change indicates higher concentration in male SPF mice. In grey, all absolute fold changes ≤ 2 are marked. b – Large intestinal concentration of adenosine, 3-hydroxybutyrate, fucose/rhamnose, and caproate in SPF and germ-free luminal content and mucus. Solid bars show mean concentration of measurements from five male mice, averaged over the four large intestinal sites. The 20 corresponding data points are displayed as circles. Abbreviations: FC – fold change; GF – germ-free; duo – duodenum, jej – jejunum, and ile – ileum; cont – luminal content; muc – mucus. Source data

Extended Data Fig. 5. Small intestinal metabolism.

a – Log₂ transformed fold changes of the mean small intestinal concentration of 42 metabolites with longitudinal patterns in the small intestine (see Fig. 3 for example profiles, Fig. 5, Supplementary Table ST4). Fold changes were calculated between male SPF and male germ-free samples in luminal content (left panel) and mucus (right panel), averaged over the 10 sampling sites and five biological replicates. Boxplots are thus based on 50 data points, where the median log₂ fold change is indicated in blue, 25^th and 75^th percentile are denoted by the box, whiskers extend to the most extreme data points not considered outliers, and outliers are marked with the plus symbol. The grey area represents absolute log₂ fold changes ≤ 2. b – Spatial metabolite profiles of luminal content or mucus concentrations in SPF and germ-free mice. Lines with shaded area indicate the moving average of the mean ± s.e.m. of concentration measurements with five male mice. Abbreviations: FC – fold change; GF – germ-free; duo – duodenum, jej – jejunum, ile – ileum, cec – cecum, and col – colon. Source data

Extended Data Fig. 6. Microbiome composition of male SPF luminal content and mucus.

Community composition and differences throughout the gut of male SPF luminal content (upper panel) and mucus (lower panel) over five sites: duodenum, jejunum, ileum, cecum and colon. Displayed are relative microbe abundances in percent, on the family level. The top 95% of annotated community members per sample are displayed discretely, bacterial categories representing less than 5% of the community are grouped in “others”. Source data