Abiotic Azoreduction Assay
The capacity for H2S to abiotically reduce azo compounds common in foods and drugs was investigated in a sterile, anoxic phosphate buffer medium. Serum bottles were prepared by sparging with ultrapure (99.9%) N2 flowed over a heated copper catalyst to remove impurities27. Bottles were stoppered and capped under positive N2 pressure and autoclaved 3 times on successive days. The medium contained final concentrations of 3 mM sulfide delivered as Na2S, 50 mM NaPO4 buffer (pH 6.3), 25 μM FMN, and 500 μM azo-compound. Bottles were incubated at 37°C with shaking and sampled using sterile anoxic technique. To halt the reaction, 5 mM ZnAc was pre-dispensed into tubes so the sample was immediately submerged in ZnAc solution. Samples were vortexed for 1min and stored at -80°C. For mass spectrometry analysis, samples were centrifuged at 14,000xg for 5 min, and the supernatants were analyzed. Redox potentials of the experimental azo compounds were calculating using a multipotentiostat (CHI 1040C A2728)28. Three cycles of cyclic voltammetry scans were collected using a rod-shaped glassy carbon working electrode, carbon counter electrodes and silver/silver-chloride (Ag/AgCl) paste reference electrode connected to the multipotentiostat at various scan rate settings to observe the reduction potentials of test molecules.
LC-MS of Azo-compounds and Azoreduction Products
Chromatography was performed on an Agilent 1200-series HPLC coupled to a 6130 Single-Quadrupole Mass Spectrometer (Agilent Technologies, Santa Clara, CA). Liquid chromatography was performed using an Agilent Poroshell 120 EC-C18, 21x50mm, 2.7nm. The mobile phases used were 2mM NH4Ac in H2O +0.1% formic acid (mobile phase A), and 2mM NH4Ac in MeOH +0.1% formic acid (mobile phase B). The column was maintained at 25°C, while the autosampler was maintained at 4°C. Elution schemes for each drug and metabolite are reported in Supplementary Table 3.
Ions were analyzed using single ion monitoring in either positive or negative mode, using m/z for each metabolite reported in Supplementary Table 3. Drying gas was set at 12 L/min, nebulizer pressure at 35 psig, drying gas at 350°C, fragmentor at 70V, and capillary voltage at 3000V. Identity of all compounds was verified by retention time and m/z of authentic, commercially available standards, except for the azoreduction products of Yellow 6 and Congo red, which were unavailable commercially. For these compounds, we identified species with predicted m/z values that increased in intensity at the same relative rate at which the parent compound decreased in our abiotic experiment.
Statistical analysis was performed in R (version 3.6.3,29) and plotted with the package ggplot2 (ggplot2 3.3.3,30) or in Graphpad Prism 8.
Hydrogen sulfide analysis
Hydrogen sulfide was quantified using a modified methylene blue detection assay31,32. For liquid microbial culture, 30 µL sample was collected using strict anoxic technique and immediately dispensed into 157 µL ZnAc (28 g/L). This was diluted with 1.26 mL H2O, followed by the addition of 113 µL Cline’s Reagent (N,N-dimethyl-p-phenylenediamine and FeCl3·6H2O in 50% HCl). The reaction was incubated at room temperature for 20 minutes to allow complete color development. Samples were centrifuged to remove particulates and 100 µL supernatant was transferred into Corning Costar half area 96-well flat bottom plates (Corning Inc., Corning, NY). Hydrogen sulfide concentrations were quantified using UV-vis detection at a wavelength of 670 nm on a BioTek Synergy H4 plate reader (BioTek Instruments, Winooski, VT) and quantified using Gen5 version 1.11 software.
E. coli pure cultures
E. coli K-12 W3100 (Carolina Biological, Burlington, NC) was grown in anoxic LB supplemented with 25 mM FMN and 30 mM succinate for all experiments. To minimize headspace and encourage hydrogen sulfide partitioning to the liquid fraction, 8 mL cultures were grown in 10 mL serum bottles under an N2 atmosphere, sealed with butyl rubber stoppers and aluminum crimps, and incubated at 37°C. All conditions were established in triplicate. Samples were extracted from sealed bottles by syringe using sterile anaerobic technique.
Live E. coli
Sulfide azoreduction and enzymatic azoreductase activity were compared in growth phase E. coli cultures. Cysteine served as a readily metabolized thiol source. The scheme of culture conditions can be seen in Supplemental Table 4. Experimental conditions received 2 mM L-cysteine and Red 40 amended at 3.0 mM, 0.5 mM, or 0.25 mM. To quantify the maximum potential sulfide generated from cysteine, cultures were spiked with 2 mM cysteine but no Red 40 (“No Red 40”). Three control conditions were established. The first, “Enzymatic Control”, monitored E. coli azoreduction and hydrogen sulfide production in bottles receiving 0.25 mM Red 40 but no supplemental amino acid. The second set of controls received 2 mM L-serine as a nonsulfur cysteine analog, to account for any growth advantage effect of excess amino acid (“AA Control”). These control cultures served to quantify baseline enzymatic azoreductase activity, while the cysteine amended cultures would in theory azoreduce Red 40 via both enzymatic azoreductase activity and sulfide azoreduction. A sterile, uninoculated media control containing media and 0.25 mM Red 40 was also monitored for abiotic activity in the media itself (“Background”). Each condition was monitored for Red 40 loss and hydrogen sulfide generation.
Heat-inactivated E. coli
The E. coli experiment was replicated using heat-killed E. coli to eliminate all enzymatic activity, including azoreductases. Cultures of E. coli were incubated with the addition of either 2 mM cysteine or serine for 18 hours and then heat-inactivated by autoclave. Autoclaving the sealed bottles allowed for cell death and enzyme denaturing without breaching the headspace, maintaining both dissolved and gas phase H2S. Following 1 hour of cooling and equilibration, H2S concentration was measured and bottles then immediately amended with 3.0, 0.5, or 0.25 mM Red 40. An additional 25 mM FMN was supplemented to replace any portion lost to heat-inactivation. H2S and Red 40 concentrations were monitored for 6 hours.
Biological Culture Red 40 Analysis
Red 40 concentrations were quantified using UV-vis detection at a wavelength of 489 nm in Corning Costar half area 96-well flat bottom plates (Corning Inc., Corning, NY) on a BioTek Synergy H4 plate reader (BioTek Instruments, Winooski, VT) and quantified using Gen5 version 1.11 software. Timepoints were diluted 1:1 with water and quantification was performed by linear regression of known standards.
Fecal Microcosm Azoreduction
The ability of a complex fecal microbiome community to azoreduce Red 40 both with and without dietary and endogenous sulfur sources was determined in fecal microcosms. Microcosms were established as 1% feces (w/v) in defined chemical medium as described in detail below. A freshly voided fecal sample was collected from a volunteer with no known health conditions who had not used antibiotics within 6 months. This study was approved by the Einstein Institutional Review Board (IRB#2013-2895). Defined chemical medium was prepared under anaerobic conditions by boiling and sparging with ultrapure (99.9%) N2 scrubbed as before, with the following final concentrations: 9 mM Na2HPO4, 31 mM NaH2PO4, 24 mM NaHCO3, 0.3 mM (NH4)2SO4, 14.4 mM NaCl, 25 µM riboflavin, 20.5 µM CaCl2 ꞏ 2H2O, 50.5 µM MnCl2 ꞏ 4H2O, 98.4 µM MgCl2 ꞏ 6H2O, 4.2 µM CoCl2 ꞏ 6H2O, Widdel trace elements andvitamins41, 0.77 µM (0.5 µg/mL) hemin, 100 µM DTT and 20 mM glucose. 8 mL cultures were grown in 10 mL serum bottles under an N2 atmosphere, sealed with butyl rubber stoppers and aluminum crimps, and incubated at 37°C. All conditions were established in triplicate. Base media before addition of feces was added to serum bottles as sterile controls. Fecal sample was homogenized in 10mL media with sterile glass beads by vigorous shaking and added anoxically to the base media to a final concentration of 1% feces (w/v). This unamended fecal slurry was placed in serum bottles to serve as heat-killed controls and immediately autoclaved before incubating at 37°C for 12 hours. This process was repeated twice before supplementation with Red 40. For the Mucin, BSA, and Formaldehyde controls, fecal slurry was supplemented with 0.4% (w/v) type III porcine gastric mucin, 1% Bovine Serum Albumin, and 4% paraformaldehyde, respectively, before aliquoting into serum bottles. Cysteine and Serine conditions were supplemented with 10 mM of the appropriate amino acid; Background was not supplemented to show baseline hydrogen sulfide production. Once all culture bottles had been assembled, they were supplemented with Red 40 to a final concentration of 500 µM (Supplementary Table 5).
Cysteine Degrading Gene Analysis
Nine genes encoding enzymes with characterized cysteine-degrading activity17–19,24 were evaluated in human gut metagenomic samples and in commensal gut microbes: CBS (K01697), CSE (K01758), CysK (K01738), CysM (K12339), CyuA (COG3681), MalY (K14155), MetC (K01760), SseA (K01011) and TnaA (K01667).
Metaquery Analysis for Mean Copy Number Per Cell in Human Gut Microbiomes
The KEGG or COG orthologue of each gene was searched within the 2,522 annotated metagenomic samples in the Metaquery20 database. For each gene, the copy number per cell in each metagenomic sample was averaged and reported in Figure 3c.
Gut Commensal Genomic Analysis of Cysteine Degrading Genes
To identify the presence or absence of the nine cysteine degrading genes, databases of homologous sequences for each gene were constructed and searched against 24,758 NCBI bacterial genomes (Genbank, April 202133).
The reference protein database for each gene was derived from the sequence from the organism in which it was originally characterized17. A KEGG web server34 BLASTP search of these genes, limited to bacteria, provided 500 closely related orthologues (≤ 1e-75) from other species.
CyuA (formerly YhaM19) is not catalogued in the KEGG database. Reference sequences were instead retrieved from COG3681 in the EggNOG database (EggNOG 4.5.135). To reduce the number of sequences to approximately 500 to facilitate alignment, sequences were clustered at 77% identity with usearch cluster_fast (Usearch 8,36) and centroids were used in the reference database.
Sequences in each gene’s reference protein database were aligned using MUSCLE with default parameters (MUSCLE 3.8.1551,37) and alignments were used to construct Hidden Markov Models using hmmbuild with default parameters (hmmer 3.3.2,38). The 9 hmms were searched against the 24,758 bacterial genomes using hmmsearch with a stringent cutoff of 1e-10. One hit in a genome indicated presence of the gene in a particular genome; multiple hits were ignored (Supplementary Table 6). We then narrowed the output to a subset of bacteria commonly found in the human gut microbiome. To identify relevant gut microbes we procured 8,548 metagenomic samples of patients from 51 studies using the R package curatedMetagenomicData39. For each of these samples, we downloaded a list of relative abundances of 6,832 taxa annotated using the MetaPhlAn240 software for each of these studies. We computed the mean relative abundance across studies for each taxa and selected organisms with ≥0.05% mean relative abundance and manual curation. Finally, different strains within the same species with the same pattern of cysteine degrading genes were clustered. By doing so we were able to observe intra-species diversity of cysteine degrading genes while eliminating duplicates.
A phylogenetic tree of the stool organisms of interest was constructed using 16S rRNA nucleotide sequences associated with each genome assembly in NCBI. 16S sequences were aligned using MUSCLE with default parameters via the EMBL-EBI search and analysis tool 41. A maximum likelihood tree was constructed using the IQ-TREE web server with 1000 ultrafast bootstrap analysis, an automatically selected substitution model and otherwise default parameters42 and visualized in iTOL43.
Mouse Housing and Diet
To assess the influence of varying dietary cysteine concentrations on fecal H2S, three separate cohorts of 10 3-week old male weanling C57BL/6J mice were randomly assigned to each of three diets varying in cysteine (Control, 0 Cys, and High Cys; Supplementary Table 2) (N = 30, 10/diet). 2-5 mice per cage were co-housed with ad-libitum access to their assigned diet and water. We then followed this experiment with a crossover experiment adding Red 40 to the control diet as a potential redox partner for H2S. For the crossover experiment, two cohorts of 6 3-week old male weanling C57BL/6J mice were assigned to either the Control diet or Red 40 diet (Supplementary Table 2) (N=12, 6/diet). Each cohort of 6 mice was split into two cages, 3 mice per cage.
Mouse Fecal Sulfide Sampling
Fecal sulfide was measured twice per week in all mice consuming Control, 0 Cys, or High Cys diets. Mice co-housed in each condition were placed in a clean collection tray and allowed to freely defecate for one hour, at which point 7-9 random fecal pellets were collected and separately placed in pre-weighed tubes containing 157 µL of 28 g/L Zinc Acetate. These samples were vortexed and stored at 4°C until analysis by Cline assay. During the crossover experiment fecal sulfide was measured more frequently, 2-5 times per week. Each mouse was placed in a separate collection tray to ensure one pellet per mouse, and, following defecation, pellets were collected and stored as above.