Microbiota metabolism of intestinal amino acids impacts host nutrient homeostasis and physiology

Abstract

The intestine and liver are thought to metabolize dietary nutrients and regulate host nutrient homeostasis. Here, we find that the gut microbiota also reshapes the host amino acid (aa) landscape via efficiently metabolizing intestinal aa. To identify the responsible microbes/genes, we developed a metabolomics-based assay to screen 104 commensals and identified candidates that efficiently utilize aa. Using genetics, we identified multiple responsible metabolic genes in phylogenetically diverse microbes. By colonizing germ-free mice with the wild-type strain and their isogenic mutant deficient in individual aa-metabolizing genes, we found that these genes regulate the availability of gut and circulatory aa. Notably, microbiota genes for branched-chain amino acids (BCAAs) and tryptophan metabolism indirectly affect host glucose homeostasis via peripheral serotonin. Collectively, at single-gene level, this work characterizes a microbiota-encoded metabolic activity that affects host nutrient homeostasis and provides a roadmap to interrogate microbiota-dependent activity to improve human health.

Keywords: amino acid metabolism; glucose tolerance; gut microbiota and metabolic genes; human microbiota.

Figure 1.. Identify the candidate gut microbes involved in efficiently utilizing amino acids.

(A) Schematic design and optimization of the in vitro, live-cell based metabolomics pipeline to identify the candidate microbes for efficiently depleting amino acids. We first optimized the pipeline by screening a collection of 20 gut microbes (highlighted in B) using four different conditions (see STAR Methods for detailed information), and 104 gut commensals (spanning multiple phyla) isolated from healthy human stools were subject to large scale screening using Condition 4. (B) Phylogenetic tree (colored by Phyla) of the 16s rRNA sequences from the 104 microbes screened in this study. Numbers correspond to strain number in Table S1. Detailed information about their growth is available in Table S1. The 20 microbes used for pipeline optimization are highlighted with yellow dots. (C) Representative LC-MS traces of an efficient His and Pro metabolizer Clostridium senegalense and Clostridium sporogenes vs. a non-metabolizer Clostridium hathewayi. (D) Clostridium sporogenes, and Clostridium senegalense rapidly deplete His and Pro in vitro. We assessed their depletion efficiency by measuring the percent of corresponding amino acids consumed in the assay buffer at different time points (15 min, 30 min, and 60 min). (E) Heatmap of the LC-MS screening of 104 human gut commensals (see Table S3 for heatmap raw data). The heatmap is clustered by average-linkage hierarchical clustering, and the Euclidean distance is calculated. The depletion efficiency of each amino acid by individual gut commensal was color-coded in green if they could metabolize corresponding amino acids in vitro and pink if they synthesize. We assessed their depletion efficiency by quantifying the percent of corresponding amino acids consumed in the assay buffer at different time points (15 min, 30 min, and 60 min). A value of ‘1’ indicates that 100% supplemented amino acids were metabolized at this time point (or the left-over amino acid is below the detection limit of the LC-MS). A negative value suggests higher concentrations of the corresponding AAs are detected in the assay buffer compared to the buffer control (without microbial cells). The heatmap shows the average values calculated at 60 min from two separate experiments. Representative strains, including Blautia hydrogenotrophica, B. ovatus, C. hathewayi, C. senegalense, and C. sporogenes, were pointed out on the map. In the AA (amino acid) Group panel, the ‘Aliphatic chain’ represents aliphatic amino acids, including Ala, Val, Leu, and Ile. ‘Aromatic’ includes Phe, Tyr, and Trp. ‘Positively charged’ represents those positively charged at pH= 7, including Lys, His, and Arg. ‘Negatively charged’ represents amino acids negatively charged at pH= 7, including Glu and Asp. ‘Hydroxyl’ represents amino acids with a hydroxyl functional group, including Ser and Thr. ‘Sulfur’ represents sulfur-containing amino acids, including Hcy, Cys, and Met.

Figure 2.. Microbiota metabolism of intestinal amino acids affects host intestinal and circulatory amino acid homeostasis.

(A) Germ-free C57BL/6J mice (n = 4 or 5 per group, age and sex match) mono-associated with B. ovatus (Bo), C. senegalense (Ce), and C. sporogenes (Cs) wild-type strains have decreased intestinal Asn, Gln, His, Lys, Pro and Trp concentrations compared to the GF control. Fecal samples are normalized to the internal standard and total fecal weight. (B) Colonization of B. ovatus, C. senegalense, and C. sporogenes also reduces the serum concentrations of Asn, Gln, His, Lys, Pro, and Trp. Samples are normalized to internal standards. For (A), data are shown as mean ± SEM, The multiple unpaired t-test was performed, the FDR (False Discovery Rate)-adjusted p values were reported; for (B), Box and whisker plots show median values, 25th–75th percentiles, and range for n = 4–6 biological replicates, an unpaired two-tailed Student’s t-test was used. The asterisk indicates p-value <0.05 (*), < 0.01 (**), or < 0.001 (***).

Figure 3.. Identify the microbiota metabolic genes for efficiently utilizing amino acids.

Depletion and complementation of metabolic genes for amino acid metabolism in multiple nonmodel gut microbes. The same assay and calculation (Fig. 1) were applied to the bacterial wild-type, gene deletion mutants and complementation constructs to quantify their amino acid depletion efficiency. A total of 35 deletion mutants were generated (Table. 1). The mutants that affect microbial metabolism of the corresponding amino acids include: Arg- (a double deletion mutant of C. sporogenes adiA and adiB (ΔClospo_00894+ΔClospo_03346) that both encode an arginine deiminase), Asn- (a deletion mutant of B. ovatus L-aspariginase, ΔBACOVA_02827 ), BCAA- (two deletion mutants were generated and the BCAA aminotransferase in both C. senegalense and C. sporogenes were mutated, ΔCSGHG_RS19535 and ΔClospo_03201), Glu- (a deletion mutant of C. senegalense glutamate mutase, ΔCSGHG_RS11260), His- (a deletion mutant of C. senegalense histidine ammonia-lyase, ΔCSGHG_RS04365), Lys- (a deletion mutant of C. senegalense lysine 5,6-aminomutase, ΔCSGHG_RS06670), Met- (a deletion mutant of C. sporogenes methionine ammonia-lyase, ΔClospo_00030), Pro- (a deletion mutant of C. sporogenes proline reductase, ΔClospo_02527), Ser- (a deletion mutant of C. sporogenes serine dehydrase, : ΔClospo_03760), Thr- (a deletion mutant of C. sporogenes Threonine dehydratase, ΔClospo_00066), Trp- (a two-gene deletion mutant of aromatic aminotransferase and Trp decarboxylase ( ΔClospo_01732+ΔClospo_02083 ) for C. sporogenes to more efficiently block C. sporogenes Trp depletion, a single gene deletion mutant of aromatic aminotransferase (ΔRUMHYD_RS04985) for B. hydrogenotrophica ), Detailed information of the manipulated genes can be found in Table 1. ‘Comp.’ indicates complementation of the mutants by re-introducing the wild-type metabolic gene into the knocked-out background to restore the corresponding metabolic function in those specific mutants. Detailed information regarding genetic disruption and complementation can be found in SI. For each histogram, n=2 or 3, data is analyzed using a one-way ANOVA, Tukey’s multiple comparisons test. The asterisk indicates p-value < 0.05 (*) < 0.01 (**) 0.001 (***), or < 0.0001 (****), n.s.: not statistically significant. Error bar: standard deviation.

Figure 4.. In vivo modulation of microbiota amino acid metabolism at a single-gene level.

(A) GF C57BL/6J mice (n = 4 to 7 per group) were mono-associated with the wild type (WT) or the knock-out bacterial mutants. The colonization of the WT and KO strains was determined by CFU (B). Fecal (C) and serum amino acid (D) levels were quantified by LC-MS (after normalization). The knock-out mutants examined in the context of host colonization include Arg- (ΔClospo_00894+ΔClospo_03346), Asn- (ΔBACOVA_02827), BCAA- (ΔClospo_03201), Glu- (ΔCSGHG_RS11260), His- (ΔCSGHG_RS04365), Pro- (ΔClospo_02527), and Trp- (ΔClospo_01732+ΔClospo_02083). For (C) and (D), Box and whisker plots show median values, 25th–75th percentiles and range for n = 3–7 biological replicates. Data in (B), (C), and (D) were analyzed using an unpaired two-tailed Student’s t-test. The asterisk indicates p-value <0.05 (*), < 0.01 (**), or < 0.001 (***).

Figure 5.. Microbiota depletion of intestinal BCAAs and Trp modulates host glucose homeostasis via peripheral serotonin release.

(A) Experimental scheme of evaluating the effect of microbiota metabolism of BCAAs and Trp on host glucose homeostasis via oral glucose tolerance test OGTT. GF age-sex-matched C57BL/6J mice (n = 4 or 5 per group) were mono-associated with the C. sporogenes WT, BCAA-, or the Trp- mutant strains for two weeks before OGTT assays. (B) The mice colonized with the BCAA- mutant are more glucose tolerant compared to the C. sporogenes WT colonized mice. (C) The mice colonized with the Trp- mutant are less glucose tolerant compared to the WT colonized mice. (D) The BCAA- mutant colonized mice have decreased serum serotonin compared to the WT colonized mice. The Trp- mutant colonized mice have elevated serum serotonin compared to the WT colonized mice. (E) Experimental scheme examining whether the modulatory effect of glucose homeostasis by microbiota depletion of intestinal BCAA and Trp is via peripheral serotonin release. The WT and mutant groups were given a Thp1 inhibitor LP533401 for one week before the OGTT test. The differentiated glucose handling between groups, as observed in (B) and (C), are much less significant in (F) and (G), respectively. (I) A working model showing how microbiota depletion of intestinal BCAA and Trp could modulate host glucose homeostasis via peripheral serotonin release. (H) Glucose areas under the curve (AUC) in (B), (C), (F), and (G) are calculated from 0 to 120 min of the OGTT using the trapezoidal method and shown as mean ± SD. Glucose concentration curves in (B), (C), (F), and (G) are also analyzed using a two-way ANOVA, Bonferroni post hoc test. For (D), box and whisker plots show median values, 25th–75th percentiles, and range for n = 5 biological replicates, and data in (D) were analyzed using an unpaired two-tailed Student’s t-test. For(H), data is analyzed using a one-way ANOVA, Šídák’s multiple comparisons test. The asterisk indicates p-value < 0.05 (*), < 0.01 (**), < 0.001 (***), < 0.0001 (****).