Untargeted stable-isotope probing of the gut microbiota metabolome using ¹³C-labeled dietary fibers

Abstract

The gut microbiome generates numerous metabolites that exert local effects and enter the circulation to affect the functions of many organs. Despite extensive sequencing-based characterization of the gut microbiome, there remains a lack of understanding of microbial metabolism. Here, we developed an untargeted stable isotope-resolved metabolomics (SIRM) approach for the holistic study of gut microbial metabolites. Viable microbial cells were extracted from fresh mice feces and incubated anaerobically with ¹³C labeled dietary fibers including inulin or cellulose. High resolution mass spectrometry was used to monitor ¹³C enrichment in metabolites associated with glycolysis, the Krebs cycle, the pentose phosphate pathway, nucleotide synthesis, and pyruvate catabolism in both microbial cells and the culture medium. We observed differential use of inulin and cellulose as substrates for biosynthesis of essential and non-essential amino acids, neurotransmitters, vitamin B5, and other coenzymes. Specifically, use of inulin for these biosynthetic pathways was markedly more efficient than use of cellulose, reflecting distinct metabolic pathways of dietary fibers in the gut microbiome, which could be related with host effects. This technology facilitates deeper and holistic insights into the metabolic function of the gut microbiome (Metabolomic Workbench Study ID: ST001651).

Graphical Abstract

Introduction

The gut microbiota is a complex ecosystem of microorganisms that inhabits and critically maintains homeostasis of the gastrointestinal tract. Most of the contributions made by the gut microbiota to the physiology of the host superorganism are related to microbial metabolism^1,2. The critical contributions of the gut microbiota metabolites toward human health have just begun to be elucidated. The primary focus on gut microbiota metabolites to date has been on the short chain fatty acids (SCFAs), amines, sulfur compounds, indoles, and choline derivatives. For example, studies are revealing how the impacts of microbial metabolism extend beyond the GI tract, denoting the so-called gut-brain (neurotransmitters, such as GABA, serotonin, and histamine)³, gut-liver (bile acids)⁴, gut-kidney (sulfur compounds)⁵, and gut-heart (trimethylamine-N-oxide)^6,7 axes. More recent research highlights the involvement of specific gut microbiome derived metabolites in the regulation of host immunity homeostasis^8,9.

Although metabolism studies on the gut microbiome have increased, crucial knowledge gaps still remain in this area. Few studies have elucidated the metabolites derived specifically from the gut microbiome without the interference and contribution from host metabolism. Additionally, gut microbes act as a community and many pathways intersect to form a dense network. Specific metabolic interactions and pathways remain to be elucidated. Furthermore, metabolites that are excreted from the microbes to the extracellular pool play important roles in stimulating specific host effects. Therefore, a better understanding of the flux of microbial products from intracellular to extracellular pools will lead to insights about gut microbiome-host interactions.

Molecularly deciphering the gut microbial metabolome remains a daunting task given the large number of microbial species and the intertwined microbial pathways constituting vertebrate gut microbiomes. Metabolomics has emerged as a powerful system biology approach measuring small molecular weight metabolites, which provides a snapshot of microbial metabolite profiles^10,11. Such steady-state metabolite profiling is only part of the story in the quest for characterizing the gut microbiome metabolites. SIRM is a thriving approach that enables dynamic tracking of individual atoms through metabolic networks¹². Concurrently, advances in high-resolution mass spectrometry have provided an additional lens through which to identify stable isotope labeled metabolites with high accuracy and sensitivity, setting the stage for more rigorous studies¹³. Untargeted SIRM (i.e., comprehensive detection and characterization of SIRM) has been applied in investigating metabolism networks in various in vitro and in vivo models^14–16, yet there are thus far few applications in microbiome research¹⁷.

Diets are important to the microbiota and metabolome, as food is a major source of precursors for microbial metabolite production¹⁸. Dietary fibers are a diverse set of carbohydrate polymers. Higher intakes of dietary fiber are associated with healthy gut microbiome and reduced incidence and mortality from several non-communicable diseases¹⁹. Soluble fibers such as inulin are natural components of grains, fruits and vegetables. Insoluble fibers such as cellulose are a principal component of the cell walls of most plants²⁰. Using ¹³C-inulin and ¹³C-cellulose as representative fibers, we applied SIRM to investigate the gut microbial metabolome in an in vitro anaerobic incubation system. The findings from this investigation not only help to understand the nutrient cycling in gut microbial ecosystems, but also provide an important perspective in gut microbiome metabolism, with implications in health and disease of the host.

Methods

Animals and Fecal Collection

C57BL/6 breeders were given ad libitum access to food (2918 Teklad irradiated Global 18% Protein Rodent Diet) and water (Lexington city tap water treated by reverse osmosis). Normally reared litters remained undisturbed with the dam until weaning on postnatal day 21. After weaning female C57BL/6 mice (n = 3) were given ad libitum access to food (D12450J, 10% Kcal from fat, Research Diets, New Brunswick, NJ) and water for 20 weeks. Animal rooms were maintained at 21±2 °C and kept on a 14:10 light:dark cycle. All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Kentucky. After 20 weeks, fresh fecal pellets were collected from each mouse (23 weeks old) and immediately placed in a sterile microcentrifuge tube. Samples were quickly transferred to an anaerobic chamber for immediate experimental setup to ensure microbial viability.

Gut microbe separation and incubation with ¹³C-fibers

Culture medium was prepared as previously described²¹. Briefly, 240 mg KH₂PO₄, 240 mg K₂HPO₄, 480 mg NaCl, 480 mg (NH₄)₂SO₄, 100 mg MgSO₄·7H₂O, 64 mg CaCl₂ ·2H₂O, 600 mg cysteine hydrochloride, 500 mg yeast extract, and 1000 mg trypticas peptone were mixed in 1L water. Media was adjusted to a pH of 6.5 by adding NaOH, autoclaved to remove O₂, and cooled under CO₂, at which point 4.0 g Na₂CO₂ was added. Media was then anaerobically transferred to Hungate tubes, sealed and autoclaved for sterility. Fresh fecal samples (ca. 110 mg) from each mouse were weighed and dispensed in 3 mL of prepared medium separately in an anaerobic chamber. The samples were pestled to suspend the microorganisms and particles. The suspensions were subjected to low-speed centrifugation (500 g, 8 min) to remove larger particles of undigested material. The supernatants were then collected and centrifuged at 3,000 g for 10 min to pellet microbes. The microbial cells were then suspended in 4 mL of culture medium, and divided equally into two Hungate tubes (2 mL/tube, in a 100% CO₂ atmosphere). To each paired tube was amended with either ¹³C-inulin (from chicory, ≥97 atom % 13C, Sigma-Aldrich, St. Louis, MO) or ¹³C-cellulose (from maize, 97 atom % 13C, Sigma-Aldrich, St. Louis, MO) aseptically to achieve a final concentration of 4 g/L fibers. The sealed Hungate tubes were incubated in a water bath (37°C) for 24 h. After incubation, the samples were centrifuged (3,000 g, 5 min) to collect supernatant (culture medium). Each pellet was washed with 1 mL of fresh culture medium and centrifuged to collect the microbial cells. All procedures were performed under anaerobic conditions.

Preparation of culture medium for SIRM profiling

The collected medium was weighed, and equal amounts of medium (50% of total medium weight) were accurately aliquoted from each sample and lyophilized. The dried powder was stored at −80 °C and dissolved in methanol: water (8:2, v/v) before analysis.

Extraction of metabolites from microbial cells for SIRM profiling

The microbial cells were quenched using 450 μL cold methanol right after collection. After brief agitation by vortex, the samples were transferred into glass tubes. Then 5 mL of methyl tert-butyl ether was added to each tube and the samples were mixed on a multi-tube vortexer (600 rpm, 30 min). Phase separation was then induced by adding 1.25 mL of Millipore deionized water. Samples were then vortexed again for 1 min, incubated at 4 °C for 10 min to allow phase separation and centrifuged (3,000 g, 10 min, 4 °C). Polar fractions were collected into clean tubes and lyophilized. The dried powder was stored at −80 °C and dissolved in methanol: water (8:2, v/v) before analysis. The protein pellet was washed with acetone and extracted using 2% SDS as reported previously²². Protein contents were analyzed using Pierce BCA Protein Assay Kit (Thermo Scientific, Inc., Waltham, MA).

SIRM profiling using LC-high resolution MS

Liquid chromatography-mass spectrometry analysis was carried out using an Ultimate 3000 ultra high performance liquid chromatography and a Thermo Q-Exactive orbitrap mass spectrometer (MS) interfaced with a heated electrospray ionization source. The mass spectrometer was operated in positive and negative ionization modes. The full scan and data-dependent MSⁿ scan were collected at a resolution of 70,000 and 35,000, respectively. The heated capillary was held at 275°C, and the HESI probe held at 350°C. The sheath gas flow was set to 40 units, the auxiliary gas flow was set to 15 units, and the sweep gas flow was set to 1 unit. The MS data acquisition was performed in a range of 59–880 m/z. Chromatographic separation was achieved using Sequant ZIC-pHILIC column (2.1 × 150 mm, 5 μm, EMD Millipore) with a guard column (ZIC-pHILIC, 2.1 × 20 mm, 5 μm, EMD Millipore). Buffer A was 20 mM ammonium carbonate, 0.1% ammonium hydroxide; buffer B was acetonitrile. The chromatographic gradient was run at a flow rate of 0.150 ml/min as follows: 0–20 min, linear gradient from 80% to 20% B; 20–21 min, hold at 20% B min; 21–22 min, linear gradient to 80% B; 22–28 min, re-equilibrate at 80% B. The injection volume was 5 μL.

Metabolite identification and ¹³C stable isotope enrichment analyses

Initial metabolite identification was performed using Compound Discoverer 3.1 (Thermo Fisher Scientific). A custom designed workflow was established for spectra alignment, compound detection, grouping, metabolite identification, and pathway analysis. The intensity threshold for component extraction was 1e⁶, and the signal-to-noise ratio was 3. The maximum retention time shift for peak alignment was 0.2 min and the mass tolerance was 5 ppm. Metabolites were further confirmed by comparison of the ion features in the samples with an in-house reference library of authentic chemical standards that included retention time, precursor ion, and their associated product ion mass spectra (Table S1). This library allowed the rapid identification of metabolites with high confidence. Peak areas of metabolites and isotopologues were integrated and exported to Excel via the TraceFinder 5.0 software package (Thermo Fisher Scientific). The fractional ¹³C enrichment in metabolites was obtained after natural abundance stripping using the protocol described previously^15,23,24.

The LC-MS peaks areas of all isotopologues for metabolites that detected in both microbial cells and culture media were summarized and log transformed to correct for heteroscedasticity and balance distributions, then the data were normalized by protein weight. Partial least-squares discriminant analysis (PLS-DA) was used to identify initial trends and clusters in data sets, which was performed using the MetaboAnalyst 4.0 web portal (www.metaboanalyst.ca). The heatmap, implemented in MetaboAnalyst tool commonly used for unsupervised clustering, were constructed based on the potential candidates of importance, which were extracted with PLS-DA.

Statistical analyses

Statistical analyses of metabolite data were performed using the GraphPad Prism version 7.04 for Windows (GraphPad Software Inc., La Jolla, CA, USA). Two groups of data were compared using Multiple Student’s t test with Holm-Sidak correction²⁵, and P < 0.05 was considered statistically significant.

DNA Extraction and 16S rRNA Sequencing

DNA extraction and 16S rRNA gene sequencing of untreated mouse fecal samples (n=3) were conducted by the Environmental Sample Preparation and Sequencing Facility (ESPSF) at Argonne National Laboratory. Analysis was conducted using the program Quantitative Insights into Microbial Ecology (QIIME version 1.9). Abundance profiling of taxonomy was performed using MicrobiomeAnalyst²⁶. The sequencing data were deposited in the ArrayExpress database.

Results

Gut microbial populations and metabolites in culture medium and microbial cells

The microbial populations in mouse fecal samples were investigated using 16S rRNA sequencing. At the phylum level, Bacteroidetes and Firmicutes were the most abundant. At the genera level, Allobaculum, Bacteroides, and Lactobacillus are the top three most abundant (Figure S1). A total of 114 metabolites were detected in the culture medium and microbial cells, covering metabolites and pathways related to amino acids biosynthesis, organic acids, nucleotide biosynthesis, coenzyme biosynthesis, central carbon metabolism, one carbon metabolism, neurotransmitters, bile acids, indoles, and benzenoids (Figure S2). Among those metabolites/pathways, except for bile acids and indoles which were not labeled, the other 77 metabolites (67% of the total detected) were labeled with ¹³C to different extents. This indicates that carbon atoms derived from dietary fiber could be used by the microbial cells to generate a wide spectrum of metabolites. Detailed information of the metabolites is provided in Table S1.

¹³C-cellulose and ¹³C-inulin contribute differentially to the central carbon metabolism in microbial cells

To compare the utilization of inulin and cellulose by gut microbiota, fecal cell suspensions were incubated with ¹³C-inulin and ¹³C-cellulose, and the fate(s) of ¹³C were tracked into central carbon metabolism using LC-high resolution MS. Figure 1 shows the incorporation of the heavy atoms (¹³C, red circles) into the different pathways in the microbial cells, including the Embden-Meyerhof-Parnas (EMP) and Entner Doudoroff (ED) pathways (glycolysis), pyruvate catabolism, pentose phosphate pathway (PPP), Krebs cycle, and nucleotide synthesis (Figure 1A). Inulin derived ¹³C (ca. 36–82%) was incorporated to a much greater extent than that of derived from cellulose (ca. 8–37%) into related metabolites (Figure 1B).

Figure 1.

¹³C-inulin and ¹³C-cellulose differently participate in central carbon metabolism in the microbial cells. A: Labeling map depicting integration of ¹³C into glycolysis, PPP, nucleotide synthesis, the Krebs cycle, and pyruvate catabolism. Black circle: ¹²C; red circle, green circle: ¹³C from pyruvate dehydrogenase (PDH) and pyruvate carboxylase (PCB)-initiated Krebs cycle reactions, respectively. Grey circle: ¹³C derived from purine de novo synthesis. B: Fractional distribution of isotopologues for metabolites. Values shown are mean ± SEM (n = 3). * P <0.05; ** P < 0.01; *** P < 0.001, as indicated.

Fully ¹³C-labeled glucose-6-phosphate (G6P) was the major isotopologue in both ¹³C-cellulose and ¹³C-inulin incubated samples, and a significantly higher labeling fraction was observed in inulin incubated samples (Figure 1B) compared with that of cellulose. Similarly, sedoheptulose 7-phosphate (S7P), an intermediate in the pentose phosphate pathway (PPP), showed a predominant fully-labelled isotopologue at ¹³C₇ (Figure 1B).

Inosine monophosphate (IMP) is a common intermediate in purine nucleotide biosynthesis. The fraction of ¹³C enrichment in IMP was significantly higher for the inulin treatment compared to the cellulose treatment. The major isotopologues for IMP were ¹³C₄ and ¹³C₅ (Figure 1B). The dominant presence of ¹³C₅ isotopologues IMP is consistent with the ¹³C flux through PPP and labeling of ribose.

Tracking ¹³C metabolism via the Krebs cycle (Figure 1A), we observed substantially higher fractional ¹³C enrichment in the ¹³C₃ isotopologue of malate and fumarate in inulin treated samples (Figure 1B), which is generated from pathways catalyzed by pyruvate carboxylase (PCB) initiated Krebs cycle.

Once pyruvate is produced from cellulose and inulin, it can either be catabolized into succinate, lactate, or acetyl-CoA (Figure 1A). These intermediates can be further metabolized to produce the short-chain fatty acids (SCFAs) such as propionate and butyrate. Propionate could be produced by the succinate pathway and acrylate pathway²⁷. The ¹³C enrichment of lactate was comparable in inulin and cellulose incubated samples, with ¹³C₂ and ¹³C₃ fractions accounting for less than 40% of the total carbon pool (Figure 1B). In contrast, higher ¹³C enrichment in succinate was observed for inulin incubated samples than that of cellulose, with ¹³C₂ and ¹³C₃ being the dominant isotopologues contributing to 69% of the total carbon pool. Notably, higher fractional ¹³C enrichment of propionate was evident in inulin treatment versus cellulose treatment, with ¹³C₂ and ¹³C₃ accounting for 78% of the total carbon pool. Butyrate is produced from the combination of two molecules of acetyl-CoA, which explains the observed major isotopologues of ¹³C₂ and ¹³C₄ regardless of the fiber type. Higher ¹³C labeling of butyrate was observed in inulin treated samples compared to cellulose, with ¹³C₂ and ¹³C₄ fractions accounting for 58% of the total carbon pool (Figure 1B).

Dietary fibers fuel microbial biosynthesis of amino acids and Vitamin B5

A total of 36 amino acids and their metabolites were detected in gut microbial cells using the current method (Table S1), and among which 28 compounds were labeled in both ¹³C-cellulose and ¹³C-inulin treated samples (Figure S2), including γ-aminobutyric acid (GABA) and 6 essential amino acids, i.e., threonine, leucine, isoleucine (leucine and isoleucine cannot be well distinguished by the current LC-MS method), lysine, valine, and phenylalanine.

Leucine, isoleucine, and valine are branched-chain amino acids (BCAA). The biosynthesis of BCAA is linked to central carbon metabolism (i.e., pyruvate, acetyl-CoA, and oxaloacetate) (Figure 2A). Our data demonstrate obvious efflux of ¹³C toward BCAA biosynthesis in inulin treated samples, and ¹³C labeled BCAA contributes to 2–7% of total carbon pool, with ¹³C₄ being the major isotopologues (Figure 2B). In ¹³C-cellulose treated samples, less than 2% of BCAA were labeled.

Figure 2.

¹³C-inulin more actively fuels biosynthesis of amino acids and vitamin B5 in the microbial cells than ¹³C-cellulose. A: ¹³C is traced through BCAA biosynthesis, Vitamin B5 biosynthesis, GABA biosynthesis, and arginine and ornithine metabolism pathways. B: Fractional distribution of isotopologues for metabolites. Values shown are mean ± SEM (n = 3). * P <0.05; ** P < 0.01; *** P < 0.001, as indicated.

The biosynthetic pathway of BCAA also produces 2-ketoisovalerate, which is an intermediate for the synthesis of vitamin B5 (pantothenic acid, Figure 2A). Another precursor needed for the biosynthesis of vitamin B5 is alanine, which could be produced from aspartate via the decarboxylation pathway²⁸. In ¹³C-inulin treated samples, although the ¹³C enrichment in BCAA and alanine is less than 16%, we observed a high fractional ¹³C enrichment in vitamin B5, with labeled species accounting for 50% of the total carbon pool (Figure 2B). In contrast, the ¹³C enrichment in vitamin B5 was less than 1% in ¹³C-cellulose treated samples (Figure 2B). The results indicated carbon atoms from inulin rather than cellulose actively participate in the vitamin B5 biosynthesis.

Glutamate is a non-essential amino acid in humans that can be synthesized by many gut microorganisms from α-ketoglutarate generated by the Krebs cycle. Glutamate serves as a precursor for the biosynthesis of amino acids such as proline and arginine. In addition, the inhibitory neurotransmitter GABA is also synthesized from glutamate²⁹. Thus, glutamate is of great interest as it is involved in multiple metabolic pathways (Figure 2A). Total ¹³C fractional enrichment in glutamate and its following metabolites in ¹³C-inulin treated sample was significantly higher than that of ¹³C-cellulose (e.g., 26% vs 3% for glutamate). ¹³C₂ and ¹³C₃ were the major isotopologues for glutamate in ¹³C-inulin treated samples (Figure 2B). For GABA, the ¹³C labeled species contributed to 23% of the total carbon pool in inulin treated samples compared with 0.1% in cellulose treated samples. Although ¹³C₂ and ¹³C₃ isotopologues were observed in GABA, they accounted for only 8% of total carbon pool. Instead, ¹³C₄ was the major isotopologue for GABA (Figure 2B). Ornithine and arginine are precursors for putrescine, which can be metabolized to GABA by as-yet-unidentified enzyme(s)³⁰. In the current study, fully ¹³C labeled ornithine and arginine were detected in inulin-treated samples (Figure 2B). Those precursors could potentially contribute to the production of ¹³C₄-GABA.

Different ¹³C enrichment of microbial dinucleotide co-factors after ¹³C-cellulose and ¹³C-inulin treatment

Nicotinamide adenine dinucleotide (NAD⁺) and flavin adenine dinucleotide (FAD⁺) are two important cofactors that are involved in redox reactions. In salvage synthesis of NAD⁺, nicotinic acid, nicotinamide, and nicotinamide mononucleotide (NMN) are either imported from outside of cells or recycled from inside cells and converted into NAD⁺ (Figure 3A). The total ¹³C fraction of nicotinamide in inulin treated samples was 17% compared to 2% in cellulose treated samples, and the dominant fraction was the ¹³C₅ isotopologue (Figure 3B). ¹³C₅ and ¹³C₁₀ were the major isotopologues in NMN, with the former (¹³C₅-ribose) being observed in both cellulose and inulin treated samples, and the latter being detected only in the inulin treated samples. ¹³C₁₀ was derived from labeling of both ribose and nicotinamide (¹³C₅-ribose +¹³C₅-nicotinamide), which is consistent with the labeling pattern of nicotinamide after ¹³C-inulin treatment (¹³C₅ as the dominant isotopologue). Adenylation of NMN introduces a ribose ring and a nucleobase adenine to produces NAD⁺ (Figure 3A). Comparable ¹³C enrichment in NAD⁺ was observed for ¹³C₅ to ¹³C₁₀ isotopologues in cellulose and inulin treated samples, suggesting a similar extend of incorporation of two ¹³C₅-ribose into NAD⁺ regardless of fiber type. In contrast, the ¹³C₁₅ to ¹³C₁₇ isotopologues were significantly higher in inulin treated samples (Figure 3B). Detailed analysis of MS² fragment spectrum of ¹³C₁₅ NAD⁺ (Figure S3), the dominant isotopologue, showed that the labeling was predominantly on nicotinamide and the two ribose rings rather than the nucleobase (Figure 3C). For FAD⁺, high enrichment levels of ¹³C₁₄-¹³C₂₃ isotopologues were observed only in inulin treated samples (altogether, they contributed to 61% of the total carbon pool). Analysis of MS² fragment spectrum of ¹³C₁₇ FAD⁺ showed that the labeling was on riboflavin and ribose rather than the adenine (Figure S4 and S5). These results indicate that ¹³C₅-ribose derived from both ¹³C-inulin and ¹³C-cellulose could be used in the microbial biosynthesis of NAD⁺ and FAD⁺. However, only ¹³C from inulin could be used by the microbial cells in the syntheses of the nicotinamide substructure in NAD⁺ and the riboflavin substructure in FAD⁺.

Figure 3.

¹³C-inulin and ¹³C-cellulse contribute differently in the biosynthesis of NAD⁺ in the microbial cells. A: ¹³C is traced through de novo synthesis and salvage synthesis pathways. B: Fractional distribution of isotopologues for metabolites. Values shown are mean ± SEM (n = 3). * P <0.05; ** P < 0.01; *** P < 0.001, as indicated. C: Proposed fragment pathway and labeling pattern of the ¹³C₁₅ isotopologue. Red circle: possible ¹³C labeling site.

Modification of intracellular and extracellular metabolite pools after treating microbial cells with ¹³C-fibers

Metabolites excreted from the gut microbial cells can contact with host cells and therefore modulate host health and disease³¹. In order to understand the exchange of metabolites between intracellular and extracellular pools, metabolites in microbial culture media were analyzed using untargeted LC-MS method. The ¹³C labeled fraction of each metabolite was compared between the microbial cells and culture medium. In ¹³C-cellulose treated samples (Figure 4A), among the 30 representative metabolites, acyl-ornithine had the highest ¹³C fraction, followed by lactate and propionate. In ¹³C-inulin treated samples (Figure 4B), propionate and acyl-ornithine had the highest ¹³C fraction, followed by butyrate. Fourteen metabolites showed higher ¹³C fraction enrichment in culture medium compared to microbial cells (Figure 4B), including butyrate, nicotinamide, GABA, organic acids, amino acids, BCAA, vitamin B5, and purines. Those results suggest that active excretion of newly synthesized metabolites from intracellular to extracellular pools occurs. In ¹³C-cellulose treated culture medium samples, there were 5 metabolites with ¹³C fraction higher than 20% (Figure 4A). The number increased to 22 in ¹³C-inulin treated culture medium samples (Figure 4B), suggesting that a more diverse spectrum of metabolites could be synthesized by the gut microbial cells from treatment with inulin than cellulose. In addition, these metabolites could be actively exported to the extracellular pool, therefore impacting gut microbial ecology, host physiology, and health.

Figure 4.

¹³C labeled fraction of 30 representative metabolites in the microbial cells and culture medium. A: ¹³C-inulin treated samples. B: ¹³C-cellulose treated samples. The left side of each panel represents the metabolites detected in culture medium (orange bar), and the right side of each panel represents the metabolites detected in microbiome (blue bar). The metabolites were listed in descending order of ¹³C fraction in microbiome, with the highest labeled metabolite listing on the top of the graph. Values shown are mean ± SEM (n = 3). * P <0.05; ** P < 0.01; *** P < 0.001, as indicated.

In order to understand the differences in metabolite pool sizes after ¹³C-cellulose and ¹³C-inulin treatments, we summarized the isotopologues for each metabolite and compared their levels in microbial cells (Figure 5A and C) and culture media (Figure 5B and D). PLS-DA shows a clear separation of ¹³C-cellulose and ¹³C-inulin treated samples (Figure 5A and B). As shown in the heat map, compared to cellulose, inulin treatment elevated the pool size of coenzymes, vitamin B5, nicotinamide, GABA, and SCFAs in the microbial cells (Figure 5C) and/or culture medium (Figure 5D). In contrast, the pool size of BCAA (leucine/isoleucine, valine) was decreased in both the microbial cells (Figure 5C) and culture media (Figure 5D) with inulin treatment. These results suggest that inulin and cellulose differentially modulate the pool size of the microbiome-derived metabolites.

Figure 5.

Comparison of metabolite pool size in the microbial cells (A and C) and culture media (B and D) after ¹³C-inulin and ¹³C-cellulose treatments. Microbial cells separated from fresh mice feces were incubated with ¹³C-inulin and ¹³C-cellulose separately (n = 3). All isotopologues of each metabolite were summarized, and data were analyzed using MetaboAnalyst. A and B: PLS-DA indicated clear separation of metabolite pools after ¹³C-inulin and ¹³C-cellulose treatments. C and D: Heat map showed the metabolites that were differentially expressed in different groups.

Discussion

Despite mounting evidence linking the gut microbiome to physiological processes and diseases in animal models and humans, the mechanisms by which specific properties and functions of the microbiome can affect host physiology are incompletely understood. The systematic characterization of microbial metabolites and elucidation of metabolic pathways will undoubtedly be a key to deciphering mechanistic details. The current study demonstrates that SIRM analysis provides a direct read-out of gut microbial activity and metabolite variations due to different fiber sources. We found that the carbons from inulin were more actively assimilated into multiple metabolic pathways by the microbial cells compared to cellulose. These included central carbon metabolism, amino acids, vitamin, and coenzyme syntheses. Moreover, higher fractions of the de novo synthesized (¹³C labeled) metabolites were excreted into culture media in inulin treated microbial cells than in samples treated with cellulose (Figure 4), resulting in extracellular pools of abundant newly synthesized microbial metabolites (Figure 4 and 5). Those metabolites could interact with host cells either locally or systematically to play a role in microbiome-host communications. The different physicochemical characteristics of inulin and cellulose contribute to their differences in the observed metabolism by the gut microbiome. Bacteria possess carbohydrate-binding modules and an extensive set of enzymes that allow for the hydrolysis of a wide variety of fibers. Cellulose is an insoluble fiber that is poorly fermented by the gut bacteria of non-ruminant mammals³². Inulin is a soluble fiber that can be hydrolyzed by microbial inulinases into fructose, glucose and inulooligosaccharides³³.

It has been reported that the gut microbiota is an independent contributing source of elevated serum BCAA levels in common human states of insulin resistance³⁴. In the current study, we found that the pool size of BCAA was significantly decreased in both the microbial cells (Figure 5C) and medium (Figure 5D) after inulin treatment compared with cellulose. Consistent with this finding, the levels of circulating BCAAs were lower after inulin intervention in piglet³⁵, and supplementation of dietary inulin could ameliorate insulin resistance in patients with type 2 diabetes³⁶. We speculate that directly modulating BCAA production by the gut microbial cells may be one of the mechanisms through which inulin exerts beneficial effects on type 2 diabetes.

GABA and nicotinamide are metabolites related to host brain functions, and they were found to be synthesized in the gut microbiome from dietary fibers in our study. GABA serves as the main inhibitory neurotransmitter in the human cortex. Accumulating evidence demonstrate that microbiota-based interventions can alter GABA levels³⁷. A recent study reported that GABA producing pathways are actively expressed in Bacteroides, Parabacteroides, and Escherichia species, and GABA was mainly produced from glutamate by decarboxylation³⁸. Our study showed that inulin rather than cellulose fueled the de novo synthesis of GABA in the microbial cells, and pathways other than glutamate decarboxylation contributed to GABA biosynthesis. The fully labeled species (¹³C₄) was the major isotopologue for GABA. In contrast, ¹³C₂ and ¹³C₃ were the major isotopologues for glutamate, which could contribute to the formation of the minor isotopologues of GABA (¹³C₂ and ¹³C₃). Ornithine, another precursor for GABA, was also labeled with ¹³C, and the fully labeled species (¹³C₅) was the major isotopologue (Figure 2B). These results suggest that in addition to glutamate decarboxylation, other pathways, for example ornithine and arginine pathway could contribute to the production of GABA when feeding the gut microbial cells with inulin.

Nicotinamide, the amide form of vitamin B3, has been shown to have a role in neuronal development and function in the central nervous system³⁹. A recent study on neurodegenerative diseases showed that amyotrophic lateral sclerosis (ALS) mice given A. muciniphila led to attenuation of ALS by increasing in vivo nicotinamide production, and high level of nicotinamide was also found in the media of A. muciniphila cultures⁴⁰. In our ¹³C-cellulose treated culture medium, the ¹³C fraction of nicotinamide accounted for 13% of the total carbon pool. This value increased to 62% in ¹³C-inulin treated samples (Figure 3B). Additionally, the pool sizes of nicotinamide in the microbial cells (Figure 5C) and media (Figure 5D) were increased in inulin-treated samples. Nicotinamide is known to be synthesized from tryptophan³⁹. Since tryptophan was not labeled in either of the fiber treated samples (Table S1), other pathway(s) and mechanism(s) may contribute to the production of nicotinamide by the gut microbial cells using carbons from dietary fibers. Inulin as a dietary supplement was shown to modulate brain function and neuroinflammation in animal models^41,42. Our study demonstrates that inulin enhances microbial GABA and nicotinamide production by promoting the carbon flow into their biosynthetic metabolic pathways (Figure 2 and 3). Those microbial metabolites could potentially affect the gut-brain axis by contributing to the host metabolite pools.

Conclusion

Determining the microbiota metabolic characteristics is an essential step in understanding the microbiome-host communications and developing personalized plans for managing the gut microbiome for desirable functionality. Using untargeted-SIRM technology, we i) identified metabolites produced by the gut microbiome using dietary fibers as substrates, ii) revealed new insights into how dietary fibers can modulate microbial cell metabolism, and iii) discovered new potential biosynthetic mechanisms for microbiome-derived metabolites (e.g., GABA and nicotinamide). This approach provides a better understanding of the microbial ecology of metabolite formation. Our findings paved the road to the precision intervention to modulate metabolite formation catalyzed by the human gut microbiota.

Supplementary Material

Supporting Info

Table S1, LC-MS information of detected metabolites. Figure S1, gut microbial populations. Figure S2, the numbers and categories of metabolites. Figure S3, MS spectra of NAD⁺ and ¹³C-labled NAD⁺. Figure S4, ¹³C enrichment of FAD⁺. Figure S5, MS spectra of FAD⁺ and ¹³C-labled FAD⁺.