Cellular stress-induced metabolites in Escherichia coli

Abstract

Escherichia coli isolates commonly inhabit the human microbiota, yet the majority of E. coli’s small molecule repertoire remains uncharacterized. We previously employed erythromycin-induced translational stress to facilitate the characterization of autoinducer-3 (AI-3) and structurally-related pyrazinones derived from ‘abortive’ tRNA synthetase reactions in pathogenic, commensal, and probiotic E. coli isolates. In this study, we explored the ‘missing’ tryptophan-derived pyrazinone reaction and characterized two other families of metabolites that were similarly upregulated under erythromycin stress. Strikingly, the abortive tryptophanyl-tRNA synthetase reaction leads to a tetracyclic indole alkaloid metabolite (1) rather than a pyrazinone. Furthermore, erythromycin induced two naphthoquinone-functionalized metabolites (MK-hCys, 2; and MK-Cys, 3) and four lumazines (7–10). Using genetic and metabolite analyses coupled with biomimetic synthesis, we provide support that the naphthoquinones are derived from 4-dihydroxy-2-naphthoic acid (DHNA), an intermediate in the menaquinone biosynthetic pathway, and the amino acids homocysteine and cysteine. In contrast, the lumazines are dependent on a flavin intermediate and a-ketoacids from the aminotransferases AspC and TyrB. We show that one of the lumazine members (9), an indole-functionalized analogue, possesses antioxidant properties, modulates the anti-inflammatory fate of isolated T_H17 cells, and serves as an aryl-hydrocarbon receptor (AhR) agonist. These three systems described here serve to illustrate that new metabolic branches could be more commonly derived from well-established primary metabolic pathways.

Graphical Abstract

Introduction

The human gut microbiota members express diverse metabolic pathways that result in the production of countless metabolites, most of which are structurally and functionally undefined.^1,2 Understanding mechanistically how microorganisms contribute to host phenotypes at the molecular level is limited by this gap in knowledge. One common member of the microbiota is Escherichia coli. While E. coli isolates only constitute about 0.1% of the human gut microbiota composition, it is found in over 90% of the human population.^3,4 While genetic variants of E. coli serve as paradigm organisms in molecular microbiology and biotechnology, natural isolates exhibit a diversity of lifestyles, engaging in commensal, probiotic, and pathogenic roles in their hosts.⁵ Indeed, E. coli isolates produce a wide range of small molecule metabolites to maintain colonization fitness and mediate interactions with other microbes and the host.⁶ Defining E. coli’s chemical repertoire remains integral to understanding the mechanistic connections between individual strains and host phenotypes. Many of the molecular mechanisms employed by E. coli to thrive in a human host are shared with other microbial species (e.g., the production and use of the siderophore enterobactin, the signaling molecule autoinducer-2, and the genotoxin colibactin, among many others).⁶ Thus, a more detailed characterization of E. coli’s small molecule output is valuable for more broadly understanding the metabolic regulatory contributions at the host-microbe interface.

Previously, we characterized the structure and biosynthesis of the small molecule termed autoinducer-3 (AI-3) in E. coli along with a family of structurally-related pyrazinone analogues, including an autoinducer reported from Vibrio cholerae.^7,8 The pyrazinone AI-3 at 5 nM enhanced the expression of the locus of enterocyte effacement (LEE) virulence genes in a qseC-receptor-dependent manner in the food-born pathogen enterohemorrhagic E. coli (EHEC). Due to the potency and low production levels of the signal, isolation of AI-3 had been a longstanding challenge in the field. Using the ribosome-inhibitory antibiotic erythromycin as a model translational stressor, we were able to enhance the production of AI-3 and its analogues to sufficient levels for isolation and characterization. Structurally, the AI-3 analogues detected were limited to the incorporation of non-polar amino acids, with all but the tryptophan analogue being detected. During the course of this work, we also noticed the stimulation of two other families of metabolites across several strains of E. coli. In this work, we investigate the ‘missing’ tryptophan AI-3 analogues and elucidate the structures, biosyntheses, and potential biological activities of the two uncharacterized erythromycin-stimulated families of metabolites.

Results and Discussion

TrpRS-mediated abortive tRNA synthetase reaction accounts for indole alkaloid biosynthesis.

The AI-3 pyrazinone analogues are formed via the spontaneous coupling of threonine dehydrogenase (Tdh)-derived aminoacetone with an aminoacyl-AMP derived from ‘abortive’ tRNA synthetase reactions.⁷ We analyzed the E. coli tryptophanyl-tRNA synthetase (TrpRS) for its potential ability to make the ‘missing’ AI-3 analogues in the presence of aminoacetone, L-Trp, and ATP. In doing so, we detected a protonated mass corresponding to the missing pyrazinone (m/z 240.1131); however, similar to previous studies, this metabolite was undetected in the E. coli extracts analyzed in this study (Figure S1). Intriguingly, when attempting to synthesize the tryptophan-derived pyrazinone standard from a linear Boc-protected tryptophan-aminoacetone dipeptide, we observed an alternative cyclization event involving a spontaneous Pictet-Spengler reaction (Figure 1A). This new tetracyclic indole alkaloid metabolite 1 (C₁₄H₁₆N₃O⁺, m/z 242.1288) was strikingly also detected in the TrpRS experiments, as well as, in representative E. coli Nissle 1917 cultures, though at minor levels in the latter under the conditions of our studies (Figure 1B–D). The retention times and tandem MS spectra were identical between the natural metabolite and synthetic standard, supporting the production of the indole alkaloid in vitro and in cell culture (Figures 1, S1, and S2). The identification of an indole alkaloid in E. coli and its derivation from an ‘abortive’ tRNA synthetase reaction was unexpected. Indeed, indole alkaloids are known to be widely produced by plants, fungi, animals, and cyanobacteria.^9–12 They have a long history of use in traditional medicine and are currently used to treat a range of medical conditions such as cancers, neurological disorders, and cardiovascular conditions.^9,13 The tetracyclic core of 1 is shared with several plant alkaloids including sarpagine, macroline, and ajmaline alkaloids.^14,15 Other than simple mono-alkaloids like tryptamine, the human microbiota is not generally thought of as an indole alkaloid source. These unexpected findings suggest that primary tryptophanyl-tRNA synthetases may moonlight in indole alkaloid biosynthesis.

Figure 1.

Biosynthesis and detection of the indole alkaloid 1. (A) Proposed biosynthesis of the Pictet-Spengler reaction-derived indole alkaloid product. (B) Indole alkaloid (C₁₄H₁₆N₃O⁺, m/z 242.1288) trace detected from E. coli Nissle 1917 culture with a matching retention time to that of the synthetic standard. (C) Results for the in vitro syntheses of linear l-Trp-aminoacetone (AA) and the cyclized indole alkaloid product. TrpRS, aminoacetone (AA), l-Trp, and ATP are required for in vitro synthesis of linear Trp-AA. Data are mean ± SD. nd, not detected. (D) Time-course analysis results of linear Trp-AA and indole alkaloid production from in vitro assays. Data are mean ± SD.

Identification of Menaquinone Pathway-Derived Metabolites in E. coli.

To identify the other erythromycin-induced metabolites, we treated the probiotic E. coli Nissle 1917, the pathogen adherent invasive E. coli (AIEC) LF82, and the commensal model strains E. coli MG1655 and BW25113 with a sublethal dose of erythromycin (25 μg/mL, Figure 2A) in cell culture and compared their mass spectral profiles to those from untreated cultures.⁷ We observed the differential stimulation of two metabolites with distinctive UV-visible chromophores (310 nm detection) that exhibited protonated masses of m/z 290.0481 (2, C₁₄H₁₂NO₄S⁺, calc. 290.0482) and 276.0323 (3, C₁₃H₁₀NO₄S⁺, calc. 276.0325) in all strains but AIEC LF82 (Figures 2B and S3). Neither of these small molecules were consistent with any previously described metabolite. Mouse fecal samples were examined to assess their potential in vivo relevance,¹⁶ which resulted in the detection of metabolite 2 but not 3 (Figure 2C). We also found that while major metabolite 2 was strongly detected in cultures grown for 24 hours, minor metabolite 3 was only detected after extended 48-hour growth periods in cell culture. Because of the strong detection of 2 both in cell cultures and in mouse stools, we elected to focus our studies primarily on metabolite 2.

Figure 2.

Detection and structural elucidation of MK-hCys (2) and MK-Cys (3). (A) Half-maximal inhibitory concentration (IC₅₀) analysis of E. coli Nissle 1917 in response to erythromycin. Blue area represents sublethal antibiotic range examined for metabolite production. (B) Detection of 2 and 3 in E. coli strains (Nissle 1917, MG1655, BW25113, and LF82) treated with and without erythromycin (25 μg/mL). Data are mean ± SD. *p < 0.05, ***p < 0.001, ****p < 0.0001. ns, not significant. nd, not detected. Two-tailed t-test. (C) Detection in C57BL/6 mouse fecal samples. n = 5 biological replicates. Data are ± SD. (D) Detection of ¹³C-isotopic labeling of 2 in cultures supplemented with universally-labelled [¹³C₅]-l-Met. (E) Reduced detection of 2 in luxS mutant strain (E. coli JW2662-1) relative to the parent strain (E. coli BW25113). Data are mean ± SD. *p < 0.05. Two-tailed t-test. (F) Structure of 2 (termed MK-hCys). (G) Key COSY and HMBC correlations of 2. (H) MK-hCys concentrations measured from E. coli cultures grown to stationary phase. Data are mean ± SD. (I) MK-hCys concentrations measured from E. coli Nissle 1917 cultures treated with erythromycin and grown to stationary phase. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant. Two-tailed t-test between no stress controls and erythromycin stress conditions. (J) Structure of metabolite 3. (K) Key COSY and HMBC correlations of 3. (L) Tandem MS of synthetic MK-Cys (top) and minor levels detected in E. coli BW25113 (bottom) using a collision energy of 30 eV. (M) Alternative cyclization products characterized while synthesizing 3 from l-Cys and 1,4-naphthoquinone.

To determine the chemical structure of metabolite 2, we first analyzed previously collected mass spectrometry data of E. coli Nissle 1917 extracts obtained from cultures supplemented with isotopically labeled amino acids, l-[U-¹³C]-Met, l-[U-¹³C]-Leu, l-[U-¹³C]-Thr, l-[U-¹³C]-Phe, and l-[U-¹³C]-Ala.⁷ We observed a 4 Da shift in the mass of 2 in cultures supplemented with universally-labelled [¹³C₅]-L-Met, indicating that four carbons in the structure are derived from methionine (Figure 2D). We predicted that the four-carbon shift in the product was due to a methionine-derived homocysteine precursor. To test this hypothesis, a luxS knockout strain was examined, which exhibited a 4.6-fold decrease in production of the compound compared to wildtype (Figure 2E). LuxS is a member of one of the two known pathways in which homocysteine is generated, where it catalyzes the cleavage of S-ribosylhomocysteine into homocysteine and 4,5-dihydroxy-2,3-pentanedione (DPD), the precursor to the quorum sensing signal, autoinducer-2 (AI-2).^17,18 The luxS quorum sensing system has been shown to regulate various functions across bacterial strains, including cell growth, biofilm formation, motility, and virulence.¹⁹ These studies provided genetic and metabolic support for homocysteine utilization in the biosynthesis of 2 and prompted structural characterization of this unknown metabolite.

To characterize the structure of metabolite 2, we cultivated E. coli Nissle 1917 (6 liters) in lysogeny broth (LB) for 48 hours at 37 °C under aerobic conditions and then extracted the molecule with ethyl acetate. The dried organic phase was subjected to repeated column chromatography purifications including (semi)preparative (C₁₈) reversed-phase high-pressure liquid chromatography (HPLC) leading to semi-pure 2 (0.2 mg). Based on the ¹H NMR spectrum and the predicted molecular formula determined from its high-resolution mass, a homocysteine-derivatized 1,4-naphthoquinone chemical structure 2 was proposed for the isolated purple compound (Figure 2F). To validate the structure, the proposed metabolite was chemically synthesized by reacting 1,4-naphthoquinone with l-homocysteine. Both the ¹H NMR spectrum and the LC-MS retention time of the synthesized compound matched that of the natural material (Table 1, Figure S3). Two-dimensional (2D) NMR analyses (gCOSY, gHSQC, and gHMBC) of the synthetic compound further supported the proposed structure (Table 1, Figure 2G). The concentration of compound 2, produced in the selected probiotic and commensal strains grown to stationary phase (24 hours), ranged from 0.7 – 1.8 μM (Figures 2H and 2I).

Table 1.

¹H [ppm, mult., (J in Hz)] and ¹³C NMR data of 2 isolated from E. coli Nissle 1917 and data of synthetic 2 and 3 in methanol-d₄.

pos.	2 (natural)	2 (synthetic)		3 (synthetic)
	δ H	δC (type)	δ H	δC (type)	δ H
1	-	182.7 (C)	-	180.7 (C)	-
2	-	116.1 (C)	-	111.2 (C)	-
3	-	147.5 (C)	-	n.d.	-
4	-	179.8 (C)	-	178.5 (C)	-
4a	-	131.5 (C)	-	131.6 (C)	-
5	8.03, brd (7.7)	127.2 (CH)	8.02, brd (7.7)	127.1 (CH)	8.03, dd (7.6, 1.4)
6	7.67, brt (7.5)	133.6 (CH)	7.67, td (7.5, 1.0)	133.6 (CH)	7.68, td (7.5, 1.4)
7	7.74, brt (7.5)	135.2 (CH)	7.73, td (7.5, 1.0)	135.1 (CH)	7.73, td (7.6, 1.4)
8	7.98, brd (7.5)	126.6 (CH)	7.98, brd (7.6)	126.5 (CH)	7.98, dd (7.8, 1.3)
8a	-	134.1 (C)	-	134.0 (C)	-
1′	-	174.8 (C)	-	172.3 (C)	-
2′	5.42, dd (11.4, 5.1)	55.9 (CH)	5.40, brd (6.8)	53.9 (CH)	4.65, brt (4.0)
3′a	2.50, m	32.2 (CH2)	2.50, tt (11.9, 5.8)	26.2 (CH2)	3.14, dd (12.9, 3.4)
3′b	2.22, m		2.23, td (12.4, 5.6)		3.27, dd (13.0, 4.6)
4′a	3.70, m	31.5 (CH2)	3.70, ddd (14.8, 11.4, 5.7)	-	-
4′b	3.12, m		3.11, dd (14.8, 6.6)	-	-

Based on the predicted molecular formula of minor metabolite 3, which only differs from the molecular formula of major metabolite 2 by a methylene group, we predicted that 3 could be a cysteine analogue of 2 (Figures 2J and 2K). While 3 was not isolated from culture, the tandem MS fragmentation pattern and LC-MS retention time matched that of the synthetic standard (Figures 2L). As expected, the NMR spectra of synthetic 3 was similar to those of metabolite 2, excluding the missing methylene (Table 1). In keeping with the low detection of 3 in vivo, compound 3 was only a minor cyclization product in the in vitro reaction with 1,4- naphthoquinone and l-Cys used to make the chemical standard. Instead, the formation of synthetic compound 4 was detected immediately under our reaction conditions, suggesting a preference for the 6-exo-trig ring closure over the 6-endo-trig ring closure (Figure S4). Compound 4 readily underwent decarboxylation to form reactive product 5 (Figures 2M and S5, Table S1). A more stable, methylated analogue 6 was observed when performing the reaction in heated methanol (Figure 2M, Table S1). Due to the instability/reactivity of these compounds, it is unclear whether these reactions occur in cells, as we were unable to detect them in cell cultures or in mouse stools. Given the spontaneous formation of 2–5 from 1,4-naphthoquinone with l-homocysteine or l-Cys, we assume that the absolute configuration of l-homocysteine or l-Cys would be retained in 2–4 and 6 as shown in Figure 2.

We propose that the biosynthesis of 2 is formed via the addition of l-homocysteine at the C2 and C3 positions of a naphthoquinone derivative. While plants synthesize hundreds of 1,4-naphthoquinone related metabolites through various pathways,²⁰ bacteria are known to harbor one naphthoquinone related biosynthetic pathway: the menaquinone biosynthetic pathway. Menaquinone (or vitamin K2) encompasses a family of naphthoquinones substituted at the C3 and C2 positions with a methyl group and a prenyl side chain of varying lengths, respectively (Figure 3A). In E. coli, menaquinone is an electron carrier in the respiratory chain during anaerobic growth.²¹ In humans, menaquinone is an essential nutrient involved in blood coagulation, bone metabolism, and cardiovascular health.^22,23 Menaquinones are abbreviated MK-n where n refers to the number of prenyl units in the side chain, MK-8 being the most common analogue in bacteria.²⁴ Here, the naphthoquinone core is instead substituted with homocysteine and cysteine (i.e., metabolites 2 and 3 are termed here MK-hCys and MK-Cys, respectively).

Figure 3.

Proposed biosynthetic origins of MK-hCys (2) and MK-Cys (3) and bioactivity. (A) Proposed biosynthetic pathway for 2 and 3. (B) MK-hCys and MK-Cys profiling in bacterial strains encoding the full menaquinone biosynthetic pathway (K. pneumoniae 700603, V. cholerae, and B. subtilis) and an incomplete pathway (P. aeruginosa PA01 and Lactobacillus rhamnosus LMS2-1). Data are mean ± SD. nd, not detected. (C) MK-hCys traces detected in wildtype (WT), menA mutant, and menB mutant E. coli BW25113 strains. The Extracted Ion Chromatograms (EICs) were generated with a 10-ppm error window. (D) MK-hCys is detected in the wildtype and menA mutant strains but not in the menB mutant strain. Data are mean ± SD. nd, not detected. (E) Supplementation of 4-dihydroxy-2-naphthoic acid (DHNA) into the menB mutant strain chemically complements MK-hCys production. Data are mean ± SD. (F) MK-Cys traces detected in wildtype, menA mutant, and menB mutant E. coli BW25113 strains. The EICs were generated with a 10-ppm error window. (G) MK-Cys is detected in the wildtype and menA mutant strains but not in the menB mutant strain. Data are mean ± SD. nd, not detected. (H) Cell viability of colon epithelial HCT116 cells after treatment with DHNA, 1,4-naphthoquinone (NQ), and MK-hCys determined by a resazurin fluorometric assay. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (I) Formation of MK-hCys from a reaction between l-homocysteine (1 mM) and DHNA (1 mM) (left) and from a reaction between l-homocysteine (1 mM) and 1,4-naphthoquinone (1 mM) (right). Data are mean ± SD.

Naphthoquinone itself is not a proposed intermediate of the menaquinone pathway. Instead, a reduced analogue, 1,4-dihydroxy-2-naphthoic acid (DHNA), is prenylated by the prenyltransferase MenA followed by methylation by UbiE to afford menaquinone.²⁵ To produce DHNA, 1,4-dihydroxy-2-naphthoyl-CoA synthase (MenB) converts o-succinylbenzoyl-CoA to DHNA-CoA via an intramolecular Claisen condensation.²⁶ A thioesterase, YdiI (MenI), has been proposed to then hydrolyze DHNA-CoA into DHNA.²⁷ Naphthoquinols, like DHNA, are reported to be prone to spontaneous oxidation due to their low midpoint redox potentials.²⁸ Due to this chemical characteristic, we hypothesized that DHNA spontaneously oxidizes, especially under aerobic conditions, allowing for the non-enzymatic addition of homocysteine and cysteine. Others have also reported similar additions of thiols and amines into 1,4-naphthoquinone-2-carboxylic acid precursors.^29–31

Because menaquinone biosynthesis is widely spread in diverse bacteria, we examined other bacterial species for production of MK-hCys and found it to be present in the extracts of both the Gram-negative pathogens carbapenem-resistant Klebsiella pneumoniae ATCC 700603 and Vibrio cholerae El Tor N16961 and the Gram-positive nonpathogenic bacterium Bacillus subtilis BR151 (Figure 3B). MK-hCys was not detected in the Gram-negative bacterium Pseudomonas aeruginosa PAO1 nor the Gram-positive bacterium Lactobacillus rhamnosus LMS2-1, which is in agreement with the absence of a complete MK pathway in these species. Of these organisms, MK-Cys was only detected in B. subtilis. Thus, these studies suggest that major MK-hCys is the common side product derived from the menaquinone biosynthetic pathway, which is consistent with its ready detection in fecal samples.

To validate that DHNA is the precursor to MK-hCys, we constructed menB and menA deletion strains in E. coli BW25113. As predicted, the ΔmenB strain lost production of MK-hCys, while the ΔmenA strain retained production (Figures 3C,D). Production of MK-hCys was about 3.6-fold higher in the ΔmenA strain relative to the wildtype strain, suggesting a buildup of its DHNA precursor (Figure 3D). Consistent with these hypotheses, chemical complementation with DHNA restored MK-hCys production in the ΔmenB strain (Figure 3E). We likewise analyzed these mutant strains for their effects on the production of minor MK-Cys and found that its production was also lost in the ΔmenB strain as anticipated (Figures 3F,G).

The wide range of menaquinone producing bacteria in intestinal flora along with detection of MK-hCys in mouse fecal samples support the basis for MK-hCys production in vivo. Because our data support that MK-hCys is produced intestinally, we examined whether it exhibits cytotoxicity toward mammalian cells. To test this, we treated the colon epithelial cell line HCT116 with varying concentrations of MK-hCys and then measured their viability using the resazurin fluorometric assay (Figure 3H).³² We found no cytotoxic effects displayed by MK-hCys. However, the precursor DHNA and its oxidized analogue naphthoquinone both exhibited cytotoxic effects with IC₅₀ values of 71.2 μM and 6.4 μM, respectively.

Others have also reported DHNA cytotoxicity at similar concentrations on different cell lines, including young adult mouse colonic cells and skin cells.^33,34 DHNA has been regarded as a beneficial metabolite, shown to exhibit anti-inflammatory activity in the gut.^35–37 Even its mild cytotoxicity has the proposed benefit of acting as a potential treatment for psoriasis through apoptosis of human keratinocytes.³⁴ However, as we have observed, DHNA readily oxidizes to a more reactive and cytotoxic naphthoquinone. Based on our findings, we propose that the spontaneous formation of MK-hCys may be a non-glutathione-based cellular detoxification strategy to mitigate cytotoxic levels of oxidized DHNA in menaquinone-producing organisms, although establishing the physiological relevance of this hypothesis would require further experimentation. A non-enzymatic strategy has likewise been proposed for the detoxification of 4-hydroxynonenal (HNE), an end-product derived from lipid peroxidation.³⁸ The primary method of HNE detoxification is through conjugation to glutathione by the enzyme glutathione transferase. However, it has also been proposed that non-enzymatic conjugation to cysteine can serve as a detoxification strategy.³⁸ Indeed, we detect MK-hCys production within 30 minutes of incubating l-homocysteine with DHNA and immediately when incubating l-homocysteine with 1,4-naphthoquinone (Figure 3I). While we observe the non-enzymatic formation of MK-hCys in vitro, we cannot rule out enzymatic mechanisms in cells. Nonetheless, this rapid reaction suggests that homocysteine can serve as a trap to scavenge oxidized DHNA.

Characterization of Lumazine Metabolites in E. coli.

The second family of erythromycin-stimulated metabolites in E. coli shared a 380 nm chromophore with protonated masses of m/z 407.1196 (7, C₁₇H₁₉N₄O₈⁺, calc 407.1197), 391.1252 (8, C₁₇H₁₉N₄O₇⁺, calc 391.1248), and 430.1362 (9, C₁₉H₂₀N₅O₇⁺, calc 430.1357) (Figures 4A and 4B). These metabolites were also detected in erythromycin-treated Pseudomonas aeruginosa PA01 and Xenorhabdus bovienii, along with a fourth metabolite with a protonated mass of m/z 444.1504 (10, C₂₀H₂₂N₅O₇⁺, calc 444.1514) observed only in X. bovienii (Figures 4A, S6, and S7). To structurally characterize these metabolites, we cultured 6 L of X. bovienii that upregulated all four metabolites under erythromycin translational stress for metabolite isolation. The molecules were extracted from the culture using Amberlite^® XAD-7 HP resins, purified by high-performance liquid chromatography, and characterized using 1D and 2D NMR spectroscopy (¹H, COSY, HSQC, HMBC, and LR-HSQMBC) (Table 2, Figure S8). NMR analysis indicated that metabolites 7, 9, and 10 were lumazines with variable aromatic side chains as depicted in Figure 4C. To confirm our structural assignments, we synthesized the compounds (7–9) that were conserved between E. coli and X. bovienii_. Acetonide-protected 5-amino-6-ribitylamino-2,4-(1H,3H)-pyrimidinedione (12) was synthesized by combining a procedure published by Bender and colleagues with a modified procedure by Winestock and Plaut (Figure S18).^39,40 Acetonide-protected 12 was then coupled to glyoxylate substrates 11a–c and deprotected to obtain lumazines 7–9 (Figure S18). The synthesized standards shared identical NMR spectral profiles, tandem MS spectra, and LC retention times with the isolated bacterial metabolites (Figures S9–S11). While 8 was not isolated from culture, the LC-MS retention time and tandem MS fragmentation pattern matched that of the synthetic standard (Figure S10). A literature search revealed that molecules 7 and 9 were initially discovered in Achromobacter petrophilum and Pseudomonas ovalis in 1969 and 1970, respectively.^41,42 More recently, Lewinsohn and colleagues identified 9 in E. coli and showed that it was antigenic, acting as a ligand for MR1-restricted T cells (MR1Ts).⁴³ Since there is no NMR data of 7 and 9 in these three research papers, we present full NMR assignments of Lum-406 (7) and Lum-429 (9) in this study (Table 2). This is also the first reporting to our knowledge of new bacterial metabolites Lum-390 (8) and Lum-443 (10).

Figure 4.

Detection, structural elucidation, and biosynthetic origins of lumazines 7–10. (A) Detection of 7–10 in E. coli strains (Nissle 1917, MG1655, BW25113, and LF82), X. bovienii, and P. aeruginosa treated with and without erythromycin (25 μg/mL). Data are mean ± SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns, not significant. nd, not detected. Two-tailed t test between no stress controls and erythromycin stress conditions. (B) Dose-dependent regulation of 7–9 in E. coli Nissle 1917 by erythromycin. n = 3 biological replicates. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns, not significant. Two-tailed t test between no stress controls and erythromycin stress conditions. (C) Proposed pathway for lumazines 7–9. 5-amino-6-ribitylamino-2,4-(1H,3H)-pyrimidinedione (5-A-RU), 4-hydroxyphenylpyruvic acid (4-HPPA), phenylpyruvic acid (PPA), indole-3pyruvic acid (I3P). (D) Spontaneous production of lumazine 9 from of I3P and 5-A-RU in vitro. n = 3 biological replicates. (E) Comparison of metabolite production levels between wildtype E. coli BW25113 (WT) and its tyrB, aspC, and tyrB-aspC mutants. n = 3 biological replicates. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns, not significant; nd, not detected. Two-tailed t test between wildtype and mutants. (F) Cellular stress induction of AspC/TyrB-dependent bacterial metabolites.

Table 2.

¹H [ppm, mult., (J in Hz)] and ¹³C NMR data of 7, 9, and 10 isolated from X. bovienii.

pos.	7 (in methanol-d4)		9 (in methanol-d4)		9 (in DMSO-d6)		10 (in DMSO-d6)
	δC (type)	δ H	δC (type)	δ H	δC (type)	δ H	δC (type)	δ H
2	n.d. (C)	-	n.d. (C)	-	159.8 (C)	-	159.8 (C)	-
4	n.d. (C)	-	n.d. (C)	-	148.7 (C)	-	n.d. (C)	-
6	146.4 (C)	-	n.d. (C)	-	142.5 (C)	-	142.1 (C)	-
7	n.d. (C)	-	n.d. (C)	-	153.7 (C)	-	n.d. (C)	-
9	n.d. (C)	-	n.d. (C)	-	140.7 (C)	-	n.d. (C)	-
10	n.d. (C)	-	n.d. (C)	-	108.2 (C)	-	108.3 (C)	-
1′a	46.7 (CH2)	4.62, dd (14.9, 2.4)	46.7 (CH2)	4.68, dd (14.4, 3.0)	45.3 (CH2)	4.52, brd (14.6)	45.3 (CH2)	4.52, brd (14.3)
1′b	-	4.31, dd (14.3, 9.1)	-	4.35, dd (14.6, 8.5)	-	4.23, m		4.22, m
2′	71.4 (CH)	4.20, ddd (8.7, 4.6, 3.0)	71.7 (CH)	4.22, ddd (8.0, 4.6, 2.9)	68.7 (CH)	4.02, m	68.8 (CH)	4.02, m
3′	74.3 (CH)	3.75, dd (7.1, 4.5)	74.3 (CH)	3.76, dd (7.2, 4.6)	72.4 (CH)	3.63, m	72.4 (CH)	3.63, m
4′	73.9 (CH)	3.84, ddd (7.1, 5.7, 3.6)	74.0 (CH)	3.86, ddd (7.2, 5.7, 3.6)	73.3 (CH)	3.63, m	73.4 (CH)	3.63, m
5′a	64.4 (CH2)	3.81, dd (11.2, 3.6)	64.5 (CH2)	3.82, dd (11.3, 3.6)	62.7 (CH2)	3.60, m	62.7 (CH2)	3.60, m
5′b	-	3.68, dd (11.2, 5.6)	-	3.69, dd (11.3, 5.8)	-	3.45, m	-	3.45, m
1″	128.1 (C)	-	-	-	-	-	-	-
2″	131.7 (CH)	8.22, d (8.8)	131.6 (CH)	8.65, s	129.8 (CH)	8.61, d (2.9)	133.8 (CH)	8.61, d (2.9)
3″	115.4(CH)	6.83, d (8.8)	112.6 (C)	-	111.0 (C)	-	111.0 (C)	-
4″	160.2 (C)	-	124.2 (CH)	8.90, m	122.7 (CH)	8.86, brd (7.7)	122.9 (CH)	8.80, brd (7.9)
5″	115.4 (CH)	6.83, d (8.8)	121.7 (CH)	7.19, overlap	120.2 (CH)	7.17, brt (7.4)	120.5 (CH)	7.22, ddd (7.9, 7.1, 0.8)
6″	131.7 (CH)	8.22, d (8.8)	123.4 (CH)	7.21, overlap	122.1 (CH)	7.20, brt (7.4)	122.2 (CH)	7.27, ddd (8.1, 7.1, 1.2)
7″	-	-	111.9 (CH)	7.42, m	111.3 (CH)	7.46, brd (8.0)	109.7 (CH)	7.51, brd (8.1)
8″	-	-	137.6 (CH)	-	135.8 (CH)	-	136.4 (CH)	-
9″	-	-	127.4 (C)	-	125.7 (C)	-	126.2 (C)	-
N-CH3	-	-	-	-	-	-	32.7 (CH3)	3.88, s
1-NH	-	-	-	-	-	11.47, s	-	11.48, s
1″-NH	-	-	-	-	-	11.56, brs	-	-

While studying cellular stress-induced small molecules in E. coli, our group has also assessed AspC-dependent metabolites. AspC is an amino acid transaminase that reversibly converts amino acids to their α-ketoacid forms. Through genetic studies, aspC has been linked to a core ‘hormetic’ antibiotic stress response in bacteria.⁴⁴ Because the lumazines are stimulated by antibiotic stress and display amino acid moieties from the AspC substrates Tyr, Phe, and Trp, we proposed that AspC contributed to the production of 7–9. To test this hypothesis, metabolite levels were measured in an aspC mutant strain (Figure 4E). Additionally, the aromatic amino acid transaminase TyrB is known to accept similar substrates and therefore a tyrB mutant and two separate aspC-tyrB double mutants were also analyzed. Indeed, the aspC single mutant and to a lesser extent tyrB single mutant resulted in a marked reduction in 7–9 levels, while the double mutant abolished production. From these results, we conclude that the conversion of l-Tyr, l-Phe, and l-Trp to 4-hydroxyphenylpyruvic acid (4-HPPA), phenylpyruvic acid (PPA), and indole-3-pyruvic acid (I3P), respectively, by AspC and TyrB is required for the formation of lumazines 7–9 (Figure 4C).

To synthesize 7–9, the flavin intermediate 5-amino-6-ribitylamino-2,4-(1H,3H)-pyrimidinedione (5-A-RU) was coupled to the α-ketoacid synthetic substrates 4-hydroxyphenylglyoxylate, phenylglyoxylate, or indole-3-glyoxylate (I3G) (Figure S18). 5-A-RU is the precursor to 6,7-dimethyl-8-ribityl-lumazine (RL-6,7-diMe) in the riboflavin biosynthetic pathway.⁴³ Others have likewise shown that 5-A-RU can couple to substrates with vicinal carbonyl groups to form different lumazine compounds.^45–47 In these studies, we found that direct incubation of 5-A-RU with the AspC/TyrB-dependent transamination products 4-HPPA, PPA, and I3P leads to the spontaneous formation of 7–9 and that no other enzymatic steps are required (Figure 4D, S12). Given the prevalence of aromatic amino acid transferases and the riboflavin pathway across bacterial species, we also examined several other microbes for lumazine production under non-stimulatory conditions. Lumazine 9 was detected in laboratory strains including the human pathogenic gammaproteobacteria Vibrio cholerae, Vibrio parahaemolyticus, and Klebsiella pneumoniae, as well as the Gram-positive riboflavin producer Bacillus subtilis (Figure S13). Lumazines 7 and 8 were likewise detected in the Vibrio spp.

Our work here builds on prior studies examining AspC-associated metabolites in E. coli, including the colipterins and the indolokines.^16,48 The colipterns are antioxidants that induce IL-10 production in macrophages and improve colitis symptoms in a colitis mouse model,¹⁶ and the indolokines are human aryl hydrocarbon receptor (AhR) ligands that also serve as defense signals in bacteria, plants, and mammals.⁴⁸ While the colipterins are stimulated by the folate stress-inducer sulfamethoxazole, their production here is reduced by erythromycin (Figure S13). The indolokines on the other hand are stimulated by erythromycin (Figure S13), but to a far lesser extent than when treated with paraquat,⁴⁸ an agent that causes oxidative stress. Collectively, E. coli produces several bioactive aspC/tyrB-dependent metabolites that respond to different forms of cellular stress (Figure 4F).

Many pteridine-containing compounds serve as redox cofactors and possess redox-based activities (e.g., antioxidant, pro-oxidant, and radical scavenging activities).^16,49,50 As such, we examined lumazines 7–9 for radical scavenging activity against the stable radical 1,1-diphenyl-2-picrylhydrazyl (DPPH) (Figure 5A). Ascorbic acid was used as a positive control [half-maximum concentration (EC₅₀) = 30 μM]. While 7 and 8 only had slight radical scavenging activity, 9 showed comparable activity to ascorbic acid (EC₅₀ = 32 μM). Thus, as an antioxidant, 9 could assist in protecting cells from redox stress or serve an additional redox role that has yet to be explored.

Figure 5.

Lumazine bioactivity analysis. (A) Stable DPPH radical scavenging activities of lumazines 7–9; L-ascorbic acid was used as a positive control. Data are presented as mean ± s.d. (n = 3). (B) Schematic of isolated T_H17 cell polarization experiments. (C) The fold change in the percentage of IL-17A⁺, IL-10⁺ IL-17A⁺, and IL-10⁺ T_H17 cell populations observed after treatment of T_H17 polarized cells with lumazine 9 relative to those treated with a DMSO vehicle. Indicated populations were identified on the basis of reporter molecules: IL-17A (Katushka) and IL-10 (eGFP). Cells are pre-gated on viable CD4⁺ T cells. Fold change = (% T cell population after lumazine 3 treatment)/ (% T cell population after DMSO treatment). Data are mean ± SD (n = 4); ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, by two-tailed t test. (D) Lum-429 (9) concentrations measured from E. coli Nissle 1917 cultures treated with erythromycin and grown to stationary phase. Data are mean ± SD (n = 3); ns, not significant; ***p < 0.001, ****p < 0.0001. Two-tailed t test between no stress controls and erythromycin stress conditions. (E) Percentage of cells that underwent 7 or more divisions determined through CellTrace Violet and analyzed by flow cytometry. Cells are pre-gated on viable CD4⁺ T cells. Data are mean ± SD (n = 4); ns, not significant; **p < 0.01 by two-tailed t test. (F) AhR activation (fold increase) in response to lumazines 7–9 and ITE positive control using a transcription reporter assay (HT29-Luca^™ AhR cells) after 24 h. Data are mean ± SD (n = 3); ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, by two-tailed t test.

We next hypothesized that the lumazines may play a role in regulating T helper 17 (T_H17) cell differentiation. T_H17 cells are prevalent in the intestinal lamina propria where they protect the host from pathogens through pro-inflammatory cytokines and effector functions.⁵¹ Overstimulation of these pro-inflammatory T_H17 cells is associated with a risk of developing immune mediated inflammatory diseases. The capacity of T_H17 cells to acquire an anti-inflammatory state – characterized by the secretion of the immunosuppressive factor, IL-10 – aids in sustaining immunological tolerance.⁵² Several studies have shown that the microbiota can stimulate intestinal T_H17 cell development and modulate the balance between pro-inflammatory and anti-inflammatory T_H17 cell populations.^52–55

To test whether bacterial metabolites 7–9 promoted either pro-inflammatory or anti-inflammatory T cell functionality, we used an IL-17A^Katushka IL-10^eGFP Foxp3^RFP triple reporter mouse model (Figure 5B).⁵⁶ Purified naïve CD4⁺ T cells labeled with CellTrace Violet were polarized to T_H17 cells using TGF-β1, IL-6 and IL-23 for 1–3 days. The differentiated cells were then treated with the selected compounds for an additional 3 days prior to flow cytometry analysis. Lumazine 9 treatment resulted in an increase in the number of IL-10⁺ and IL-10⁺ IL-17⁺ cell populations by about 2- to 3- fold without affecting the IL-17A⁺ population (Figure 5C). Neither 7 nor 8 altered the IL-10⁺ and IL-10⁺ IL-17A⁺ populations (Figure S15). E. coli Nissle 1917 produces 9 at a concentration as high as 3 μM when stimulated by erythromycin, and given that these effects were observed at the lowest concentration tested (5 μM), we propose that these T_H17 modulatory effects by 9 may be physiologically relevant (Figure 5D). While there was a slight increase in CD4⁺ T cell proliferation when cells were treated with 5 μM of 9, there was not a significant increase when cells were treated with 10 or 20 μM, suggesting that 9 modulates the anti-inflammatory fate of T_H17 cells without enhancing or impairing cellular growth (Figure 5E). The lack of difference in cell proliferation also supports a T-cell receptor (TCR)-independent mechanism for the stimulation of IL-10⁺ T_H17 cells by 9.

We hypothesized that the biological activity of 9 may or may not be linked to its redox activity. To test this, the antioxidants ascorbic acid (vitamin C) and α-tocopherol (vitamin E) were similarly examined (Figure S16). Consistent with previously published reports, ascorbic acid displayed the reverse trend, with an increase in the IL-17⁺ and IL-10⁺ IL-17⁺ populations and no effect on the IL-10⁺ population, while α-tocopherol did not alter any of the T_H17 populations.⁵⁷ These data suggest that the activity of 9 is not due to a general redox mechanism.

Activation of the aryl hydrogen receptor (AhR) using the AhR ligand FICZ has been reported to mediate the conversion of mature T_H17 cells into IL-10 producing anti-inflammatory cells.⁵⁶ AhR ligands are generally aromatic in structure and include indole-functionalized metabolites produced by the human microbiota. To explore whether 9 serves as an AhR agonist, an AhR activation assay was performed using HT29-Lucia^™ AhR Cells (InvivoGen). Lumazine 9 was observed to significantly induce AhR activity in a dose-dependent manner, supporting an AhR-mediated mechanism (Figure 5F). Metabolites 7 and 8 failed to initiate IL-10 expression in T_H17 cells, which was consistent with their only slight AhR reporter response that was not dose-dependent (Figure S17). Thus, we propose that lumazine 9 enhances the isolated population of anti-inflammatory T_H17 cells through an AhR-mediated pathway.

Conclusion

In this follow-up paper to our work on AI-3,⁷ we explore the fate of tryptophan when coupled to aminoacetone via a TrpRS-mediated abortive tRNA synthetase reaction. In doing so, we discover a new metabolite (1) featuring an indole alkaloid core scaffold previously unreported in bacteria formed via a Pictet-Spengler cyclization. While this metabolite was only detected at low levels in E. coli cultures, we hypothesize that it could be more abundantly produced by other microbes. We also revisit the metabolomic data collected from E. coli cultures subjected to erythromycin treatment.⁷ From this data set, we had originally observed the stimulation of AI-3 and structurally-related pyrazinone analogues. Here, we report the stimulation of two other families of metabolites. The first are previously unreported metabolites (MK-hCys and MK-Cys) derived from homocysteine and cysteine, respectively, and DHNA, an intermediate of the menaquinone biosynthetic pathway. We propose that the formation of MK-hCys (the major and widely detected metabolite of the two) may act as a 1,4-naphthoquinone-2-carboxylic acid detoxification strategy. The detection of MK-hCys in mouse fecal samples indicates a potential in vivo biological role that can now be explored. The second family of metabolites that were stimulated by erythromycin stress are a family of lumazines. Two of these lumazines (4 and 6) had been previously reported as bacterial metabolites, while the other two (5 and 7) are newly reported. We show here that the aminotransferases AspC and TyrB are essential for the production of these lumazines, which appear to be formed through the spontaneous coupling of the riboflavin intermediate 5-A-RU and pyruvate-functionalized metabolites derived from tyrosine, phenylalanine, and tryptophan. One of the lumazine members (9), an indole-functionalized analogue, had been previously reported as an E. coli metabolite with MR1 antigenic properties.⁴³ However, we demonstrate that this indole-functionalized analogue also possesses antioxidant properties on par with vitamin C and modulates the anti-inflammatory fate of isolated T_H17 cells, likely by activating AhR. Overall, the characterization of previously undescribed or understudied metabolites, such as those reported here, bolster our understanding of E. coli’s small molecule repertoire and serve as a structural foundation for investigating biological mechanisms at the host-microbe interface.

Experimental Section

General Experimental Procedures.

Instrumentation.

High-resolution electrospray ionization mass spectrometry (ESI-MS) data were obtained using an Agilent iFunnel 6550 quadrupole time-of-flight (QTOF) instrument fitted with an ESI source coupled to an Agilent 1290 Infinity HPLC system and a Kinetex 5μ C₁₈ 100 Å column (250 × 4.6 mm) with a water:acetonitrile gradient containing 0.1% formic acid at 0.7 mL/min: 0–30 min, 5 to 100% acetonitrile. The mass spectra were recorded in positive ionization mode with a mass range from m/z 100 to 1,700. Extracted ion chromatogram (EIC) spectra using exacts masses with 10 ppm mass windows were used to calculate metabolite production levels. Isolation of metabolites was performed using an Agilent Prepstar high-performance liquid chromatography (HPLC) system. 1D (¹H and ¹³C) and 2D (gCOSY, gHSQCAD, gHMBCAD, and LR-HSQMBC) NMR spectral data were measured on an Agilent 600 (or 400) MHz NMR spectrometer equipped with a cold probe and a 3-mm tube, and the chemical shifts were recorded as δ values (ppm) referenced to solvent residual signals.

Expression and Purification of TrpRS.

A TrpRS-His₆-tagged gene cassette was cloned using the below primers and gDNA template from E. coli Nissle 1917. The PCR product was digested and ligated into a pETDuet-1 plasmid using the NdeI and XhoI restriction sites. The sequence-validated plasmid was transformed into chemical competent E. coli BL21(DE3) cells (New England Biolabs) and cultivated on lysogeny broth (LB) agar plates supplemented with 100 μg/mL ampicillin at 37 °C in a static incubator. For protein expression, an overnight culture derived from a single colony was diluted 1:200 into 1 L LB supplemented with 100 μg/mL ampicillin, then incubated at 37 °C and 250 rpm until the OD₆₀₀ reached 0.5. The culture was then induced with 1 mM IPTG and the temperature was reduced to 25 °C to allow for TrpRS protein expression (250 rpm for 24 h). TrpRS was purified using Ni-NTA Agarose (Qiagen) according to the manufacturer’s protocol, and purity was assessed by SDS-PAGE analysis. Protein concentrations were determined using absorbance at 280 nm.

TrpRS_Forward	5’-GTAAAAAATCATATGACTAAGCCCATCGTTTTTAG-3’
TrpRS_Reverse	5’-GTAAAAAATCTCGAGTTAATGATGGTGATGATGATGGCCGCCCGGCTTCGCCACAAAACCAATCG-3’

In Vitro Protein Biochemical Synthesis of Linear Peptide and Indole Alkaloid with Isolated TrpRS.

To 500 μL of aminoacylation buffer (100 mM HEPES, pH 7.2, 30 mM KCl, and 10 mM MgCl₂), the following components were added: 1 μM TrpRS (final), 1 mM aminoacetone (AA), 1 mM L-Trp, 10 mM ATP (pH 7.5), and 1 mM DTT. All reactions were run in triplicate. The reaction mixtures were incubated for 4 h at 37 °C in a standing incubator, after which 500 μL of ethyl acetate was added. The mixtures were vortexed for 10 s and centrifuged for 5 min at 18000 × g. The ethyl acetate layers (400 μL) were transferred and dried in vacuo. The dried materials were dissolved in 200 μL of methanol for LC-MS analysis. For time-course analysis, the reaction mixtures were extracted with ethyl acetate after 4, 24, 48, and 72 h incubations.

Cultivation of E. coli Nissle 1917 and Detection of Indole Alkaloid 1.

Small scale: E. coli Nissle 1917(ArdeyPharm) was grown for 48 hr at 37 °C under aerobic conditions (250 rpm) in 5 mL of lysogeny broth (LB) alongside a 5 mL LB media control. Ethyl acetate (5 mL) was added to the culture and control, which was vortexed for 10 seconds and centrifuged for 10 mins at 2000 × g. The organic layers were transferred and dried in vacuo. The dried extracts were dissolved in 200 μL of methanol for Q-TOF analysis. Large scale: E. coli Nissle 1917 (ArdeyPharm) was grown overnight at 37 °C under aerobic conditions (250 rpm) in 5 mL of lysogeny broth (LB). The overnight culture was used to inoculate 1 L of M9 (glucose, MgSO₄, CaCl₂, casamino acids). The 1 L culture and control were incubated at 37 °C with shaking (215 rpm) for 48 hr, then extracted with 2.1 L (3 × 700 mL) of ethyl acetate and dried in vacuo. The crude extract was fractionated by an Agilent Prepstar HPLC system with an Agilent Polaris C₁₈-A 5 μm (250 × 21.2 mm) column with a gradient elution from 10 to 100% aqueous acetonitrile with 0.01% trifluoroacetic acid over 60 mins and flow rate of 8 mL/min using a 1 min fraction collection time window. Indole alkaloid 1 was detected in fraction 18 using this method.

Erythromycin Stimulation of Metabolites 2–7 in Bacterial Cell Cultures.

E. coli strains (Nissle 1917, MG1655, BW25113, LF82, JW2662-1, JW0911-1, JW4014-2, and BW25113 ΔtyrB/ΔaspC), Pseudomonas aeruginosa PAO1, and Xenorhabdus bovienii were grown from single colonies overnight at 37 °C under aerobic conditions (250 rpm) in 5 mL of LB. Overnight cultures were used to inoculate three replicates of 5 mL of LB or LB + erythromycin (25 μg/mL, Acros Organics) at a 1:200 dilution and incubated at 37 °C with 250 rpm shaking for 48 h. 6 mL of n-butanol was added to each culture, which were then vortexed for 10 seconds and centrifuged for 10 mins at 1500 × g. The top 5 mL of the organic layer was transferred and dried in vacuo. The dried extracts were dissolved in 200 μL of methanol for QTOF LC-MS analysis.

The same procedure was repeated without erythromycin stimulation using the following strains: Klebsiella pneumoniae ATCC 700603, Vibrio cholerae El Tor N16961 ΔctxAB, Bacillus subtilis BR151, Pseudomonas aeruginosa PAO1, and Lactobacillus rhamnosus LMS2-1.

E. coli Nissle 1917 was also treated with varying concentrations of erythromycin (0.4, 1.6, 6.3, 25, or 100 μg/mL, Acros Organics), and metabolites were extracted into ethyl acetate as previously described.⁷

E. coli JW2662-1 (ΔluxS768::kan), E. coli JW0911-1 (ΔaspC745::kan), and E. coli JW4014-2 (ΔtyrB747::kan) were obtained from the Coli Genetic Stock Center (Yale University). E. coli BW25113 ΔtyrB/ΔaspC was constructed in our lab as previously described.¹⁶

Isolation of MK-hCys from E. coli Nissle 1917.

Overnight culture of E. coli Nissle 1917 in six 5 mL aliquots of LB were used to inoculate six 1 L cultures of LB in 4 L Elenmeyer flasks. These cultures were incubated at 37 °C with shaking (250 rpm) for 48 h pooling and extracting with 12 L (2 × 6 L) of ethyl acetate. The dried crude material was subjected to a Biotage^® SNAP Cartridge (KP-C₁₈-HS, 30 g) and separated using a step gradient with the following solvent composition (each 100 mL): Fraction 1, 20% aqueous methanol; Fraction 2, 40% aqueous methanol; Fraction 3, 60% aqueous methanol; Fraction 4, 80% aqueous methanol; Fraction 5, 100% methanol. Fraction 2 was further fractionated by an Agilent Prepstar HPLC system with an Agilent Polaris C₁₈-A 5 μm (250 × 21.2 mm) column with a gradient elution from 10 to 100% aqueous acetonitrile with 0.01% trifluoroacetic acid over 60 min and flow rate of 8 mL/min using a 1 min fraction collection time window. Fractions 2–37 were further separated using a Phenomenex Luna C₁₈ (2) 100 Å column (250 × 10 mm, flow rate 4.0 mL/min, a gradient elution from 10 to 100% aqueous acetonitrile with 0.01% trifluoroacetic acid to give 2 (0.2 mg).

MK-hCys (2): Purple gum; UV (MeCN/H₂O) λ_max 231, 291, 542 nm; ¹H NMR (600 MHz) data in methanol-d₄, see Table 1; HRESIMS (positive-ion mode) m/z 290.0481 [M+H]⁺ (calcd for C₁₄H₁₂NO₄S⁺, calc. 290.0482).

Isolation of Lumazines 7, 9, and 10 from X. bovienii.

Overnight cultures of X. bovienii in six 5 mL aliquots of LB were used to inoculate six 1 L cultures of LB + erythromycin (25 μg/mL) in 4 L Elenmeyer flasks. These cultures were incubated at 30 °C with shaking (250 rpm) for 48 h. After removal of the cells by centrifugation for 20 mins at 13,000 × g, 20 g/L of Amberlite^® XAD-7 HP resins were added to the supernatant, which was shaken using 150 rpm for 30 mins. The resins were recovered by filtration using a glass frit under vacuum and extracted with 1 L of methanol to give a crude extract (5 g). The dried crude material was subjected to a Biotage^® SNAP Cartridge (KP-C₁₈-HS, 30 g) and separated using a gradient elution from 10 to 100% aqueous methanol over 30 mins with a flow rate of 25 mL/min using 3 min fraction collection time window. Fractions 2 and 3 were further fractionated by an Agilent Prepstar HPLC system with an Agilent Polaris C₁₈-A 5 μm (250 × 21.2 mm²) column with a gradient elution from 15 to 30% aqueous acetonitrile with 0.01% trifluoroacetic acid over 60 min and flow rate of 8 mL/min using a 1 min fraction collection time window. Fractions 2–7, 3–10, and 3–15 were further separated using a Phenomenex Luna C₁₈ (2) 100 Å column (250 × 10 mm, flow rate 4.0 mL/min, a gradient elution from 10 to 100% aqueous acetonitrile with 0.01% trifluoroacetic acid to give 7 (0.5 mg), 9 (1.4 mg), and 10 (1.0 mg), respectively.

Lum-406 (7):Yellow gum; UV (MeCN/H₂O) λ_max 290, 368 nm; ¹H (600 MHz) and ¹³C (150 MHz) NMR data in methanol-d₄, see Table 2; HRESIMS (positive-ion mode) m/z 407.1196 [M+H]⁺ (calcd for C₁₇H₁₉N₄O₈⁺, calc. 407.1197).

Lum-429 (9):Yellow gum; UV (MeCN/H₂O) λ_max 215, 275, 314, 391 nm; ¹H (600 MHz) and ¹³C (150 MHz) NMR data in methanol-d₄ and DMSO-d₆, see Table 2; HRESIMS (positive-ion mode) m/z 430.1362 [M+H]⁺ (calcd for C₁₉H₂₀N₅O₇⁺, calc. 430.1357).

Lum-443 (10):Yellow gum; UV (MeCN/H₂O) λ_max 214, 275, 315, 391 nm; ¹H (600 MHz) and ¹³C (150 MHz) NMR data in DMSO-d₆, see Table 2; HRESIMS (positive-ion mode) m/z 444.1504 [M+H]⁺ (calcd for C₂₀H₂₂N₅O₇⁺, calc. 444.1514).

Standard Calibration Curve and Quantification of MK-hCys and Lumazine 9.

Pure compound (MK-hCys or lumazine 9) was added to 5 mL of LB medium in culture tubes to achieve final concentrations of 0.05, 0.1, 0.4, 1.6, 3.13, and 6.25 μM. n-Butanol (5 mL) was added to the LB medium and vortexed. The culture tubes were centrifuged for 10 minutes at 1500 × g. The top 4 mL of the n-butanol layer was transferred to a vial and dried in vacuo. This was performed in triplicate. The dried extracts were redissolved in 200 μL of methanol and subjected to LC-MS analysis. A 10 ppm mass window around the calculated exact mass (290.0482 for MK-hCys and 430.1357 for lumazine 9) was used to generate extracted ion chromatogram (EIC) graphs. The peak integrations were used to create the standard calibration curves.

Single colonies of E. coli BW25113, E. coli MG1655, and E. coli Nissle 1917 were inoculated in 5 mL of LB medium and grown at 37 °C (250 rpm). Overnight cultures were used to inoculate three replicates of 5 mL of LB at a 1:200 dilution. Cultures were incubated at 37 °C (250 rpm) for 24 h. Cultures were extracted and analyzed using the same method described above for the standards. The same procedure was repeated for E. coli Nissle 1917 treated with erythromycin (0.4, 1.6, 6.3, 25, or 100 μg/mL, Acros Organics).

Detection of MK-hCys and MK-Cys in E. coli ΔmenB and ΔmenA Mutants.

For detection of MK-hCys, E. coli BW25113 containing the lambda Red recombinase plasmid pORTMAGE-3 and the constructed single gene knockout strains ΔmenB::amp and ΔmenA::amp, both containing pORTMAGE-3, were inoculated into 5 mL LB in triplicate and grown for 24 hr at 30 °C (250 rpm). 5 mL of ethyl acetate was used to extract each culture. The ethyl acetate layer was dried in vacuo, redissolved in 200 μL of methanol, and subjected to LC-MS analysis. For the detection of minor MK-Cys, the same procedure was used except cultures were grown for 48 h and extracted with n-butanol.

Construction of Deletion Mutants.

The ampicillin resistance cassette of plasmid pETDuet-1 was PCR amplified using primers with sequence extensions homologous (50 nt) to the 5¢ and 3¢ ends of menB and menA. The PCR products were DpnI-digested and gel-purified using the QIAquick PCR Purification Kit (Qiagen). The plasmid pORTMAGE-3 containing the lambda Red recombinase system and a kanamycin (Kan) selection maker was transformed into E. coli BW25113. 3 mL of LB with 50 μg/mL Kan was inoculated with overnight seed culture at a 1:200 ratio. The culture was grown at 30 °C (250 rpm) until the OD₆₀₀ reached 0.5. Cultures were induced by incubating at 42 °C for 15 min. The cultures were split into 1 mL portions and cooled on ice. Cells were pelleted by spinning for 2 minutes at 7,000 × g at 4 °C. The supernatant was removed and cells were resuspended in 1 mL of chilled ddH2O. This process was repeated for a total of two washes before resuspending cells in 50 μL of chilled ddH2O. 1 μL of 100 μM oligonucleotide was added and cells were transferred to an electroporation cuvette with a 0.1 cm gap width. Cells were electroporated at 1800V with 25 μF capacitance and 200Ω resistance. Transformed cells were recovered in 3 mL of LB + Kan at 30 °C overnight and then plated on LB agar plates with 50 μg/mL Kan and 100 μg/mL ampicillin. Mutants were validated by amplifying the gene of interest with over-spanning PCR and sequencing.

Cytotoxicity Assay.

HCT 116 cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin at 37 °C in a humidified atmosphere with 5% CO₂. Cells at a density of 1 × 10⁴ per well were seeded in a 96-well plate and incubated for 24 h. The seeding medium was removed by inverting and flicking the plate over a waste container, then blotting the plate on a paper towel. The wells were refilled with 99 μL of fresh RPMI₁₆₄₀ and 1 μL of compound in DMSO at final concentrations of 0, 1.6, 3.1, 6.3, 12.5, 25, 50, 100, and 200 μM, in triplicate, and incubated for 48 h. The plate was inverted to remove the supernatant, refilled with a 44 μM solution of resazurin in DMEM, and incubated for 3 h. The 44 μM solution of resazurin was made by adding 1 mL of resazurin (440 μM in PBS at pH of 7.4) to 9 mL of the DMEM medium. After incubation, the fluorescence was recorded using a 560 nm excitation/ 590 nm emission filter set. Cell viability was calculated using the formula Cell viability % = 100 × (S_X-S_pos)/(S_neg− S_pos) where S_X is the signal from the experimental sample, S_pos is the signal when there are zero viable cells, and S_neg is the signal from the sample treated with the vehicle control.

In vitro Production of MK-hCys.

1,4-Napthoquinone (1 mM) or 4-dihydroxy-2-naphthoic acid (1 mM) was incubated with homocysteine (1 mM) in 50% methanol in water at 37 °C. At each time point (0, 0.5, 1, 3, and 6 hours), 50 μL of the reaction was transferred to a microcentrifuge tube and stored at −80 °C. Each tube was thawed immediately prior to LC-MS analysis.

2,2-Diphenyl-1-picrylhydrazyl (DPPH) Radical Scavenging Assay.

Stock solutions (100 mM) of lumazines 1–3 and ascorbic acid were prepared in DMSO. Freshly prepared 0.2 mM DPPH in methanol was dispensed into a 96-well plate with 198 μL added to the first column of wells and 100 μL to the rest. The antioxidants (2 μL) were added to the first column and quickly two-fold serially diluted across the row. The plate was incubated in the dark for 35 min, and radical scavenging activities were subsequently quantified via absorbance at 517 nm. In parallel, the same procedure was performed with methanol only (no DPPH) to measure the background absorbance of the antioxidants. The absorbance values from the methanol only plate were subtracted from the values of the DPPH/methanol plates. DPPH radical-scavenging activity percentage (PI%) was calculated using the formula PI = 100 × (A_control−A_sample)/A_control.

T_H17 Cell Experiments.

Triple reporter mice (Foxp3^RFP, IL-10^eGFP, IL-17A^Katushka) generated by the Flavell group were used in these studies.⁵⁶ Naïve CD4⁺ T cells were isolated from the spleen and lymph nodes of 2 mice with the mouse naïve CD4⁺ T cell isolation Kit (Stemcell). Cells were labeled with CellTrace^™ Violet (CTV) to monitor cell proliferation. The cells were cultured under T_H17 polarizing conditions, in Click’s medium supplemented with 5% FBS, 55 μM β-mercaptoethanol, Glutamax^™ (Gibco), 100 U/mL penicillin/streptomycin, 10 μg/mL anti-CD3, 2 μg/mL anti-CD28, 10 μg/mL anti-IFNγ, 10 μg/mL anti-IL-4, 20 ng/mL murine IL-23, 20 ng/mL murine IL-6, and 0.5 ng/mL murine TGF-β1. A DMSO vehicle or the select compound was added to the cultures after 1–3 days. After the addition, the cultures were cultured for an additional 3 days, followed by flow cytometry analysis.

AhR Activation Assay.

The activity of the AhR signaling pathway was measured using HT29-Lucia^™ AhR cells, purchased from Invivogen. The cell lines are engineered to study AhR induction by monitoring the activity of Lucia luciferase reporter protein. QUANTI-Luc^™ Gold (InvivoGen) was used to detect the secreted luciferase according to the supplier’s instructions. Briefly, 199 μL cells were seeded at a density of about 50,000 cells/well in a 96 well plate and 1 μL of the desired ligand (or DMSO control) was added. 2-(1H-Indol-3-ylcarbonyl)-4-thiazolecarboxylic acid methyl ester (ITE) was used as a positive control and unstimulated cells were used as a negative control. After a 24 h incubation, 20 μL of cell supernatant was transferred to a black plate and 50 μL of QUANTI-Luc^™ Gold was added immediately prior to reading the luminescence. The promotor activity was repeated in triplicate and values were expressed as fold-change relative to DMSO vehicle treated cells.

Table 3.

Primers used to construct menB and menA knockout strains.

Primer	Sequence (5′ to 3′)	Description
menB-F	GTAATCGCAAGTCTGGCGCAGATCTTGCGTTTCTGA CTAAAGGACACAATTTCAAATATGTATCCGCTCA	Forward primer for menB mutation
menB-R	ACCCCCGCGTCCATGGGGATCTGCCAGCGGTATACC TGCGCGCTACGCATTTACCAATGCTTAATCAGTG	Reverse primer for menB mutation
menA-F	CATCATTGTTTGATGGGGCTGAAAGGCCCCATTTTT ATTGGCGCGTATTTTCAAATATGTATCCGCTCA	Forward primer for menA mutation
menA-R	GCTATGTGGGCTGTTGGCAAAATCATCAATTGTTAA TTGATATTTGTCAGTTACCAATGCTTAATCAGTG	Reverse primer for menA mutation
menB-OF	ATTTATGTGGTGAACGTGACAGC	Forward primer for menB over-spanning PCR
menB-OR	TCTTCCCAGGTTTCCTGACTGAA	Reverse primer for menB over-spanning PCR
menA-OF	CCAGCGATTTAAGCGGTCAA	Forward primer for menA over-spanning PCR
menA-OR	AAAGTTGGAGAACAGCGGTT	Reverse primer for menA over-spanning PCR