A metabolomic view of how the human gut microbiota impacts the host metabolome using humanized and gnotobiotic mice

Abstract

Defining the functional status of host-associated microbial ecosystems has proven challenging owing to the vast number of predicted genes within the microbiome and relatively poor understanding of community dynamics and community-host interaction. Metabolomic approaches, in which a large number of small molecule metabolites can be defined in a biological sample, offer a promising avenue to 'fingerprint' microbiota functional status. Here, we examined the effects of the human gut microbiota on the fecal and urinary metabolome of a humanized (HUM) mouse using an optimized ultra performance liquid chromatography-mass spectrometry-based method. Differences between HUM and conventional mouse urine and fecal metabolomic profiles support host-specific aspects of the microbiota's metabolomic contribution, consistent with distinct microbial compositions. Comparison of microbiota composition and metabolome of mice humanized with different human donors revealed that the vast majority of metabolomic features observed in donor samples are produced in the corresponding HUM mice, and individual-specific features suggest 'personalized' aspects of functionality can be reconstituted in mice. Feeding the mice a defined, custom diet resulted in modification of the metabolite signatures, illustrating that host diet provides an avenue for altering gut microbiota functionality, which in turn can be monitored via metabolomics. Using a defined model microbiota consisting of one or two species, we show that simplified communities can drive major changes in the host metabolomic profile. Our results demonstrate that metabolomics constitutes a powerful avenue for functional characterization of the intestinal microbiota and its interaction with the host.

Figure 1

Ultra performance liquid chromatography (UPLC)-MS analysis reveals microbial-related metabolites in fecal and urine samples of HUM mice. (a) Experimental workflow for metabolomic data generation and analysis. (b) Number of features detected in feces and urine of at least one of three HUM mice by different UPLC-MS methods. (c) Number of significantly different features (P<0.01, >10-fold difference) between HUM (n=3) and GF (n=3) mice found in urine or fecal metabolome. (d) Percentage of predicted isotopes and adducts in the set of features characteristic of feces from HUM mice (RPLC-ESI+).

Figure 2

Humanized (HUM) mice fecal and urine metabolomes are distinct from those of conventional mice. (a, b) Principal component analysis (PCA) plot of fecal (a) and urine (b) metabolome from germ-free (GF, pre-humanization), conventional (CN) and HUM mice. (c, d) Hierarchical clustering of the fecal (c) and urine (d) metabolome from GF, CN and HUM mice samples.

Figure 3

Humanized (HUM) microbiota-dependent changes in fecal metabolites. Normalized mean values in GF and HUM mice±s.d. of three biological replicates are plotted for tryptamine, indoxyl glucuronide, a trisaccharide, creatine and creatinine.

Figure 4

Dietary change alters composition and function of HUM mouse microbiota. (a) Principal co-ordinate analysis (PCOA) plot of 16S rRNA-based gut microbial profiling from HUM mice fed a regular diet (RD) or polysaccharide-deficient diet (PDD; n=3 mice per group). (b) PCA plot of fecal metabolite profiling from HUM mice fed with two different diets. (c) Hierarchical clustering of fecal metabolomes from HUM mice fed with two different diets versus metabolomes of the respective mouse diets.

Figure 5

Simplified model microbiotas alter fecal metabolome. Heat-map of microbe-dependent metabolites from fecal samples of gnotobiotic mice colonized with B. thetaiotaomicron (BT), Bt and B. longum (BTBL) or humanized (HUM) mice. Features significantly increased or decreased in either BT or BTBL relative to GF are shown; fold change>10 and P-value<0.01 m/z values in red indicate GF-related features that disappear in all three colonization states.

Figure 6

Gut microbial diversity and human fecal metabolome are reconstituted in HUM mice. (a) Principal co-ordinate analysis (PCOA) plot of 16S rRNA-based gut microbial profiling from HUM mice from three different donors (n=4–8 mice per group) and the respective donor samples (open enlarged symbols). (b) Percent of features found in human feces, HUM mouse feces or both, for the three respective donor/recipient groups using RPLC-ESI+. (c) Venn diagram with metabolites common to the human and HUM mice samples for each donor/recipient group. (d) PCA plot of fecal metabolite profiling from HUM mice colonized with three different fecal samples.