Tryptamine Metabolism and Functionalization in Gut Commensal Bacteria Expand Human Tryptamine Signaling Responses

Abstract

Gut microbes secrete specialized small molecules that broadly influence human physiology. Despite their potential significance, the variety of functional small molecules known in the gut is relatively limited. Here, we screened the supernatants from human fecal-derived bacterial cultures to explore their agonist effects on the human G protein-coupled receptors (GPCRs), melatonin receptor type 1A and 1B (MTNR1A, MTNR1B). Chemical analysis of the supernatant-soluble molecules of Clostridium sporogenes, a prominent gut commensal identified in the screen, led to the characterization of agonists for these two melatonin receptors. Specifically, through bioactivity-assisted isolation and characterization, we identified three small molecules 1-3, including two previously uncharacterized metabolites, which were synthesized to confirm their structures. While the structure of 1 features a urea core symmetrically disubstituted with tryptamine moieties, 2 and 3 harbor a monomeric tryptamine functionalized with methyl carbamate and N-acetyl groups, respectively. These structural characterization efforts illuminated downstream functional consequences of tryptamine metabolism in C. sporogenes. Additional GPCR screening analyses revealed that 2 activates melatonin receptors and the purinergic P2RY11 receptor, whereas 1 serves as an agonist for the semi-orphan receptor GPR55. Interestingly, 1 also exhibits significant inhibitory activity against inflammatory soluble epoxide hydrolase with a half-maximal inhibitory concentration of 420 nM. Single-cell RNA sequencing analysis of gut tissue from mice orally treated with 1 relative to solvent vehicle control revealed that 1 specifically decreased the frequency of GPR55- and granzyme K-expressing effector-like CD8 T cells in the intraepithelial lymphocyte population. Overall, this study broadens our understanding of tryptamine-derived signaling at the human-microbe interface.

Graphical abstract

INTRODUCTION

The human gut is a rich habitat harboring trillions of microbial cells and viral particles that closely engage with human physiology.^1–4 Recent investigations have greatly expanded our understanding of the gut microbiome’s taxonomy, structure, and metabolic functions.^1–4 The gut microbiome is continuously shaped by host factors such as age, diet, sex, and antibiotic use, and intestinal microbes participate in a variety of biochemical processes.^5–7 Deviations from gut microbiota homeostasis, known as dysbiosis, are closely linked to the onset, severity, and progression of myriad human diseases, including inflammatory diseases, metabolic disorders, cancers, and infectious diseases.⁸ This growing body of knowledge has spurred the development of microbiome-based treatments, such as fecal microbiota transplantation (FMT), and the establishment of novel diagnostic biomarkers.^9,10 However, despite these advances, the molecular mechanisms underlying the myriad interactions between gut microbes and their hosts remain poorly understood at the molecular level.

Gut microbes can produce a diverse range of small molecules known as specialized metabolites. These include polyketides, non-ribosomal peptides, ribosomally synthesized and post-translationally modified peptides (RiPPs), terpenoids, and saccharides, in addition to the processing of primary metabolites, such as amino acids.^6,11–14 These functional small molecules have significant implications for human health and disease, including regulating diverse human signal transduction systems.^15–17 Notable examples of gut microbe-derived molecules include short-chain fatty acids, trimethylamine N-oxide, bile acids, and indoles.^8,9,18–23 These metabolites can accumulate in the gut and often enter the bloodstream, positively or negatively influencing physiology at distal body sites.^18,24 Gut microbes are also capable of biotransforming diverse xenobiotic molecules, including antibiotics and dietary metabolites, into secondary small molecule signals.^18,25 Thus, gut microbes play a pivotal role in modulating human health and disease through the production and manipulation of small molecules.

GPCRs are the largest family of membrane receptors and are expressed in nearly all human tissues.²⁶ They detect a wide range of small molecule stimuli, such as neurotransmitters and hormones, and are frequently targeted by medical drugs; greater than 40% of all current small molecule drugs target GPCRs.²⁶ Among them, the melatonin receptors MTNR1A and MTNR1B recognize their namesake hormone to regulate sleep and circadian rhythm.^27,28 Dysregulation of melatonin signaling has been linked to a variety of disorders, including sleep disturbances, mood disorders, and metabolic dysfunction.²⁷ Small molecule metabolites secreted by gut microbes have significant untapped chemical potential, and given the diversity of GPCRs, there is an opportunity to discover novel endogenous small molecule ligands that specifically regulate these receptors. To identify bioactive small molecule metabolites secreted by gut microbes, we employed a function-guided discovery method on diverse supernatants from a bacterial culture collection. Specifically, we leveraged the high-throughput Parallel Receptor-ome Expression and Screening via Transcriptional Output-Tango (PRESTO-Tango) GPCR screening platform to uncover microbial metabolites that engage the GPCRs MTNR1A and MTNR1B.^15,29 In this study, we report the identification, isolation, structural characterization, and functional analysis of active tryptamine-derived metabolites from the anaerobic culture supernatant of C. sporogenes. We also describe their plausible routes of formation and potential biological activities.

RESULTS AND DISCUSSION

We isolated 132 bacterial strains (Table S1) from six human fecal samples under anaerobic conditions in a rich medium specialized for the growth of gut commensals (Gifu) to screen for regulatory properties on MTNR1A and MTNR1B using β-arrestin recruitment-based Tango assays.^29,30 We found that supernatants from multiple C. sporogenes isolates activated the melatonin receptors compared to medium controls, which was a focus of this study (Figure S1). We also found that a Ruminococcus gnavus isolate from the collection had comparable activity to C. sporogenes whereas a Bacteroides vulgatus isolate had reduced activity (Figure S2). LC-MS analysis showed similar chromatographic patterns across the C. sporogenes strains (Figure S3), and thus, we selected and grew two representative C. sporogene strains 661 and 110 in 1 L of Gifu medium for 2 days. To separate small molecules by polarity, the spent supernatants of the cultures were sequentially extracted with two organic solvents, methylene chloride (MC) and ethyl acetate (EA). The remaining aqueous layer was applied to Amberite XAD-7 resin followed by methanol extraction of the filtered resin. The three organic extracts were then tested for the activation of the melatonin receptors; unexpectedly, the processed organic extracts of the rich Gifu medium alone also exhibited background MTNR activity under these conditions (Figure S4). Therefore, we repeated the identical extraction and screening procedure using microbes cultivated in a defined minimal SACC medium (Table S2). Unlike Gifu medium, we clearly observed that the MC-soluble fractions from the bacterial cultures in SACC medium harbored strong activation activity on both MTNR1A and 1B with a single crude dose of 25 μg/mL while SACC medium extract controls were inactive, suggesting that active C. sporogenes metabolites are present in the MC-soluble fraction (Figure S5).

We next carried out metabolite profiling of the active MC-soluble fraction of representative isolate 661 compared to those of SACC medium alone. Crude MC-soluble materials were analyzed by C18 high-performance liquid chromatography (HPLC) connected to a low-resolution mass spectrometry (LR-MS) system. HPLC-MS data analysis displayed the presence of two distinct peaks 4 and 5, eluting at t_R 17.2 min ([M+H]⁺ at m/z 190) and t_R 16.2 min ([M+H]⁺ at m/z 188), respectively (Figure S6). High-resolution electrospray ionization-quadrupole time-of-flight-mass spectrometry (HR-ESI-QTOF-MS) analysis suggested that these molecules are indole-3-propionic acid (IPA, 4) and indole-3-acrylic acid (IAA, 5) with the molecular formulae C₁₁H₁₁NO₂ and C₁₁H₉NO₂, respectively (Figure S7), which were structurally validated through comparisons with their corresponding commercial standards (Figure S6). Further UV-MS analysis of the minor metabolites suggested that C. sporogenes produces other metabolites with the indole core. Since melatonin contains an indole moiety, we also performed an HPLC coelution experiment with a commercial melatonin standard; however, this metabolite was not detected in the active MC-soluble fraction (Figure S8).

To isolate the active metabolites, the MC-soluble extract was fractionated via C18 semi-preparative HPLC with a 1 min fraction collection interval, giving 60 subfractions. The fractions were dried, redissolved in dimethyl sulfoxide (DMSO), and examined for MTNR activity, leading to three major HPLC fractions 20, 26, and 30 with activity for both receptors (Figure 1A and Figure S9). LC-MS analysis combined with HR-MS data suggested that the major metabolite present in active fraction 20 was N-acetyl tryptamine (3) (Figure 1B), which was further supported by comparison with a commercial standard (Figure S10). We further purified the two indole-functionalized molecules from active fractions 26 and 30, yielding 2 and 1 with retention times of 27.2 and 31.5 min, respectively (Figure 1A).

Figure 1. Activity-guided discovery of metabolites 1–5 from culture extracts of human fecal-derived C. sporogenes that activates human melatonin receptors.

(A) Activity-guided metabolite fractionation and isolation were conducted on the MC-soluble fraction from C. sporogenes 661. Major metabolites 1-5 are marked on the chromatogram. Activity data for each fraction are presented as mean ± SEM from three technical replicates in this fractionation and screening workflow. (B) Chemical structures of metabolites 1-5. (C) Key COSY and HMBC NMR correlations of 1 and 2.

Structural characterization of 1 and 2 was achieved through interpretation of ¹H and 2D-NMR (gCOSY, gHSQC, and gHMBC) and HR-ESI-QTOF-MS spectral data (Figures S11–S19 and Table 1). Briefly, the molecular formula of 1 was determined to be C₂₁H₂₂N₄O ([M+H]⁺ at m/z 347.1866) based on an HR-MS measurement, which indicated that 1 contained 13 degrees of unsaturation. The ¹H NMR spectrum of 1 recorded in methanol-d₄ showed a partially pure spectrum; however, we were able to characterize distinct signals attributable to indole, including one olefinic methine proton and four aromatic proton signals, and an additional two methylene groups (Figure S11). The gHSQC spectrum of 1 enabled the assignment of all the protons to the directly bonded carbons (Figure S13). The gCOSY spectrum revealed the sequential COSY correlations from H-4 to H-7 and the HMBC correlations from H-5 to C-3a, H-6 to C-7a, and H-2 to C-3 established the presence of an indole motif (Figure S12 and S14). The key COSY correlations between two methylene groups H-8 and H-9 built the conjugation of these groups and further connectivity of the indole motif and these methylene groups by an HMBC correlation from H-2 and C-8 (Figure S12 and S14). Additionally, the HMBC spectrum showed a key three bond correlation from H-9 to a quaternary carbon C-10 (δ_C 162.0), suggesting the presence of two units of tryptamine with a urea core (Figures 1B and 1C), which is consistent with the HR-ESI-QTOF-MS data (Figure S19). The molecular formula of 2 was assigned as C₁₂H₁₄N₂O₂ ([M+H]⁺ at m/z 219.1128) based on HR-ESI-QTOF-MS spectroscopic data (Figure S19). The ¹H NMR spectral data were similar to that of 1. However, a distinct proton signal for the resonance of O-methyl protons suggested the existence of a methoxy group (Figure S15). Key HMBC correlations from H-9 to a quaternary carbon C-10 (δ_C 158.0) indicated the establishment of carbamate further functionalized with a methyl group, which is evident by the HMBC correlation from the methyl group (δ_H 3.55) to C-10 (Figure 1C and Figure S18).

Table 1.

¹H and ¹³C NMR data for compounds 1 and 2 in CD₃OD

	Compound 1			Compound 2
Position	δH mult (J, Hz)	δ C		δH mult (J, Hz)	δ C
2	7.02 s	121.9	2×CH	7.03 s	121.9	CH
3		111.8	2×C		111.8	C
3a		127.4	2×C		127.4	C
4	7.53 d (J = 8.0 Hz)	117.8	2×CH	7.54 d (J = 8.0 Hz)	117.8	CH
5	6.97 t (J = 7.5 Hz)	118.1	2×CH	6.97 t (J = 7.4 Hz)	118.1	CH
6	7.06 t (J = 7.5 Hz)	120.7	2×CH	7.06 t (J = 7.4 Hz)	120.8	CH
7	7.30 d (J = 8.0 Hz)	110.7	2×CH	7.30 d (J = 8.0 Hz)	110.7	CH
7a		136.7	2×C		136.7	C
8	2.88 t (J = 7.0 Hz)	25.7	2×CH2	2.90 t (J = 7.5 Hz)	25.4	CH2
9	3.39 t (J = 7.0 Hz)	40.4	2×CH2	3.36 t (J = 7.5 Hz)	41.4	CH2
10		160.0	C		158.2	C
O-methyl				3.61 s	50.9	CH3

The structures and activities of 1 and 2 were further confirmed by chemical synthesis. Briefly, tryptamine hydrochloride was dissolved in dimethylformamide (DMF) and stirred at room temperature followed by the addition of carbonyldiimidazole (CDI). Triethylamine was added and stirred for 1 h and reacted at 60 °C overnight, yielding 1. To synthesize 2, a solution of tryptamine hydrochloride in a mixture of EA and 1N NaOH was degassed, and methyl chloroformate was added dropwise under nitrogen gas purging. The reaction mixture was stirred for 1 h at room temperature to give 2. Comparative NMR data analyses of the natural and synthetic materials unambiguously confirmed the structures of the metabolites (Figures S20 and S21). To the best of our knowledge, although their chemical structures have been described in chemical synthesis^31,32, metabolites 1 and 2 are newly described from natural sources.

The enzymes responsible for the conversion of L-tryptophan to IPA (4) and IAA (5) in C. sporogenes were previously established by Fischbach and co-workers.²⁴ IPA derived from tryptophan by aromatic amino acid aminotransferase can be converted into indole lactic acid that subsequently undergoes dehydration and dehydrogenation reactions by bacterial phenyl lactate dehydratase, acyl-CoA ligase, and acyl coenzyme A dehydrogenase to yield 5 and 4, respectively (Figure 2A).²⁴ L-Tryptophan is also known to be converted into tryptamine by tryptophan decarboxylase encoded in C. sporogenes (Figure 2A).³³ Here, we propose downstream processing of tryptamine in C. sporogenes, in which N-acetylation yields the known metabolite 3. While this metabolite is known in both humans and human microbes, it was not previously reported in C. sporogenes (Figure 2A). The condensation of tryptamine with carbon dioxide/bicarbonate (or an activated carbonate source) could lead to carbamate-functionalized tryptamine as a potential intermediate, which was not detected under our experimental conditions (Figure 2A). Nevertheless, O-methylation or further tryptamine functionalization could lead to 2 and 1, respectively (Figure 2A).

Figure 2. Plausible routes of formation for the observed metabolites.

(A) Proposed biosynthesis and formation of identified metabolites. (B) Production of 1 with supplementation of tryptamine in cell-free SACC medium in both aerobic and anaerobic overnight incubations. m/z 347.1866 indicates the protonated mass-to-charge ratio value of 1. Intensity means peak integration values of the high-resolution extracted ion count (EIC) chromatogram around 347.1866 with a 10 ppm window. (C) Detection of 1 in Ruminococcus gnavus culture. Representative data for 1 and 1a is shown. 1a indicates phenethylamine dimer with urea core and the structure is shown in Figure 4B. Detailed analysis is presented in the supporting information (Figure S20). Data are mean ± SEM for three biological replicates. *P<0.05; **P<0.01; ***P<0.001 by two-tailed student’s t-test. nd, no detection.

Select molecular features in host-microbiota metabolomes derive from spontaneous reactions of common sets of core reactants. To test the potential non-enzymatic production of 1, we fed tryptamine into fresh SACC medium at varying concentrations in the absence of bacteria and incubated the mixture under aerobic and anaerobic conditions for 24 h. Notably, 1 could be detected under these conditions with high concentrations of tryptamine supplementation in both aerobic and anaerobic incubations (Figure 2B). We also analyzed the metabolome of an additional commensal species known to express tryptophan decarboxylase and convert tryptophan and phenylalanine into their corresponding amines, Ruminococcus gnavus.³³ Consistent with its capacity to produce tryptamine and phenethylamine (PEA) production, we also detected the respective dimer products 1 and 1a in R. gnavus cultures (Figure 2C and Figure S22). The PEA-derived dimer 1a was also chemically synthesized in a parallel route with that of 1. Although our data demonstrate that 1 can arise via spontaneous chemistry under the media conditions used here, we cannot rule out other cellular catalysts in culture.

All purified molecules 1-5 were tested for the activation of the two melatonin receptors, MTNR1A and MTNR1B (Figure 3A). We included L-tryptophan, tryptamine, and melatonin as controls. PRESTO-Tango screening results revealed that, as anticipated, the known ligand melatonin showed the strongest activation of MTNR1A and 1B with EC₅₀ values of 1.52 and 4.96 nM, respectively (Figure 3A). N-acetyl tryptamine 3 and compound 2 activated both receptors, although compound 2 exhibited significantly lower EC₅₀ values (3: EC₅₀ values; 1.29 μM for MTNR1A and 393 nM for MTNR1B; 2: EC₅₀ values; 16.9 μM for MTNR1A and 24.3 μM for MTNR1B). By contrast, compound 1 was much less active compared to 2 or 3 (Figure 3A). Importantly, the two major indole metabolites of C. sporogenes, IPA (4) and IAA (5), did not contribute to melatonin receptor activation in this assay (Figure 3A).

Figure 3. GPCR screening of metabolites.

(A) Dose response analysis of metabolites for the activation of melatonin receptors, MTNR1A (top) and MTNR1B (bottom), via PRESTO-Tango assays. (B) G protein recruitment assay revealed activity against GPR55 and P2RY11 by metabolites 1 and 2, respectively. Metabolites 1 and 2 were tested at 10 μM, while an equivalent amount of DMSO was used as a solvent vehicle control. (C) Dose response of metabolite 1 against GPR55 via G protein and β-arrestin (PRESTO-Tango) recruitment assays. LPI was used as a positive control. Data is expressed as the mean ± SEM based on three replicates. The activity results were reproduced in two independent experiments.

Because PRESTO-Tango screening only assesses one aspect of GPCR signaling (β-arrestin recruitment),²⁹ we next evaluated GPCRome-wide activation by 1 and 2 using three G protein-dependent reporter assays that assess activation of the Gα subunits, Gα₁₂, Gα_i, and Gα_s (Figure 3B). While none of these compounds engaged Gα_s signaling, we observed that 1 activated both Gα₁₂ and Gα_i reporters downstream of GPR55, whereas 2 activated the purinergic receptor p2ry11 receptor (initial test dose of 10 μM; Figure 3B). GPR55 is proposed to be a potential cannabinoid and lysophosphatidylinositol (LPI) receptor and is expressed in immune cells, including CD8+ intraepithelial lymphocytes (IELs)^34–36; however, GPR55 is still considered to be a semi-orphan receptor.³⁷ Thus, we further conducted a dose-response analysis of 1 for GPR55 activation, using the proposed endogenous GPR55 ligand lysophosphatidylinositol (LPI) as a positive control.^38,39 We verified the previously observed selectivity and established estimated EC₅₀ values of 41.2 and 14.8 μM for Gα₁₂ and Gα_i, respectively (Figure 3C). This estimated potency is comparable with that of LPI (EC₅₀ values; 41.2 μM for Gα₁₂ and 30.0 μM for Gα_i). These results suggest that urea-functionalized metabolite 1 is an endogenous gut microbiome-produced metabolite that selectively activates G protein signaling downstream of GPR55.

1,3-disubstituted urea core molecules, especially in the synthetic chemistry literature, are known to harbor high potency in inhibiting soluble epoxide hydrolase (sEH).^40,41 This enzyme is crucial in various diseases due to its role in metabolizing anti-inflammatory epoxy fatty acids (EFAs), which are produced by cytochrome P450s in the arachidonic acid signaling pathway (Figure 4A). sEH then converts EFAs into their corresponding vicinal diols that are known to be pro-inflammatory in both humans and animals (Figure 4A). Indeed, inhibition of sEH can sustain elevated levels of EFAs, ultimately offering anti-inflammatory effects.⁴⁰ Given the urea core of metabolite 1, we further evaluated 1 and 1a for their inhibitory effects on human sEH enzyme activity in vitro (Figure 4B). Compounds 1 and 1a harbored significant inhibitory activities with IC₅₀ value of 420.5 nM and 2.2 μM, respectively, relative to solvent vehicle controls set at 100% activity. These studies illuminate endogenous microbiota-produced molecules that effectively inhibit inflammatory sEH (Figure 4B).

Figure 4. Inhibition of soluble epoxide hydrolase (sEH) by 1 and 1a.

(A) Representative sEH reaction converting EETs to DHETs. EETs are derived from arachidonic acid via cytochrome P450 oxidation. (B) Potency of 1 and 1a on sEH inhibition. Abbreviations are as follows: EETs: Epoxyeicosatrienoic acids; DHETs: Dihydroxyeicosatrienoic acids. Data are mean ± SEM for three biological replicates.

To explore the potential physiological impacts of 1 on IELs in vivo, we treated two groups of specific pathogen-free (SPF) C57BL/6 mice with either vehicle or 1 by oral gavage. Intraepithelial immune cells were then isolated and profiled using single-cell RNA sequencing (scRNA-seq; Figure 5A). We identified 17 cell populations based on marker feature expression: Effector-like CD8 αα T cell, memory-like CD8 αα T cell, effector-like CD8 αβ T cell, memory-like CD8 αβ T cell, proliferating T cell, naïve CD4 T cell, activated CD4 T cell, resting CD4 T cell, B cell, plasma cell, natural killer (NK) cell, innate lymphoid cells 1 (ILC1), ILC2, ILC3, conventional dendritic cell (DC), plasmacytoid DC, and mast cell (Figure 5B). GPR55 is expressed by CD8 T cells and NK cells and is specifically enriched in effector-like CD8 αα T cells and memory-like CD8 αα T cells (Figure 5C). Thus, to explore the direct impacts of 1, we subset the effector-like CD8 αα T cells and memory-like CD8 αα T cells, and proliferating T cells followed by re-clustering. In this analysis, 1 specifically led to a decrease in a subpopulation of effector-like CD8 αα T cells (4: effector-like CD8 αα T cells E3; Figure 5D, E). Marker feature analysis revealed that effector-like CD8 αα T cells E3 express granzyme K in addition to granzyme A and granzyme B (Figure 5F, G). These data suggest that 1 may reshape the intraepithelial lymphocyte compartment by down-regulating the frequency of GPR55- and granzyme K-expressing T cells.

Figure 5. Metabolite 1 decreased Gzmk+ CD8 T cells in the gut intraepithelial lymphocytes.

(A) Schematic of the scRNA-seq pipeline for mouse gut IEL. Two groups of SPF C57BL/6 mice were gavaged with either vehicle (solvent for 1: 5% DMSO, 5% Tween 80 and 90% PBS) or 30 mg/kg 1 for ten days, ileal and colonic IELs were isolated and sorted followed by scRNA-seq. (B) UMAP plot of gut intestinal IELs combining mice from both groups. Cells were pooled from two independent scRNA-seq experiments. (C) GPR55 expression in the IEL population. (D) UMAP plot of CD8⍺+ T cell subpopulation, split by treatment. (E) Proportion of different T cell populations of total CD45+ cells. Individual values of each group represent data from each of the independent scRNA-seq experiments. (F) Expression of marker features in CD8+ T cell subset. (G) Violin plot of GPR55 expression in CD8⍺+ T cell subset. Data in panel B-F present pooled data from two independent scRNA-seq experiments. TD indicates metabolite 1.

In summary, we identified previously undescribed metabolites from the gut commensal C. sporogenes that selectively activate melatonin receptors and GPR55. Our structural and metabolic analyses revealed that tryptamine is further functionalized to form 1 and 2 with distinct urea and carbamate moieties, respectively, expanding our understanding of tryptamine-derived metabolism and functionalization. To the best of our knowledge, this type of downstream metabolic processing of tryptamine and its associated metabolites have not been described previously. Based on our non-enzymatic data analyses in complex medium conditions, these molecules could, in principle, be formed by producers that accumulate tryptamine, which was further supported by testing an additional microbial tryptamine producer R. gnavus. Functionally, these molecules serve as agonists of the human melatonin and semi-orphan GPR55 and may contribute to GPCR-associated host inflammatory signaling. Metabolite 1 also exhibits substantial sEH inhibitory activity, which is also associated with inflammatory signaling. Notably, single-cell RNA sequencing data suggested that 1 may participate in the regulation of IELs, particularly targeting GPR55- and granzyme K-expressing T cells. Overall, the structure-function-based study described here adds to the growing knowledge that gut bacteria are untapped producers of bioactive small molecule metabolites and expands our understanding of how tryptamine-derived microbiota metabolites may shape host biology.