Microbially derived glutathione from Eubacterium rectale alleviates oxidative stress and promotes intestinal epithelial recovery

Abstract

Background

Certain microbes inhabiting the gut have been implicated in maintaining gut homeostasis and promoting gut damage repair. Lachnospiraceae members were highly detected in dysbiotic IL-10 KO mice that displayed similar physiological outcomes as control mice. Lachnospiraceae is a highly diverse family of microbes that have been shown to display both commensal and pathogenic characteristics in the colon environment.

Results

We investigated the impact of genetic variation in five Lachnospiraceae strains on lowering cellular inflammation and reactive oxygen species (ROS) levels. Cell-free spent media (CFSM) from Eubacterium rectale resulted in lowered ROS, and nitric oxide levels in stressed colon cells and colon organoids. CFSM-treated organoids showed reduced ROS accumulation, improved epithelial integrity, and partial recovery of barrier function compared to oxidatively stressed controls. We demonstrated through shotgun metagenomics, metabolomics, host RNA sequencing, and molecular techniques that glutathione (GSH) biosynthesized by E. rectale alleviated host ROS damage. We showed downregulation of cell stress and immune response genes, indicating recovery from ROS stress. Chemical depletion of GSH in CFSM confirmed the role of microbial derived GSH in alleviation of ROS in colon cells.

Conclusions

In this study, we identify E. rectale as a potential probiotic by lowering colon inflammation and ROS damage through production of reduced glutathione. Microbially derived GSH has not been well established in the Lachnospiraceae family which are a large member of the overall gut microbiota. Understanding more about the impacts of microbial functions including GSH on lowering inflammation is needed to develop potential probiotics or therapies for chronic inflammatory conditions.

Download video file ^{(41.3MB, mp4)}

Video Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-026-02457-y.

Background

Microbial diversity [1, 2] is often associated with a healthy gut environment and lower chronic inflammation. A diverse microbiota has the ability to out crowd potential gut pathogens while stimulating the host immune system and providing many facilitatory metabolites and nutrients [3]. During perturbations, such as antibiotic usage or other medicinal exposures [4], gut infections [5, 6], or autoimmune disorders [7], there is generally a loss in gut microbial diversity, notably from the Firmicutes (Bacillota) phylum [2, 8]. Firmicutes are a functionally diverse phylum of bacteria, and contain the Lachnospiraceae family, also a diverse and prevalent family at around 10% of the gut microbiota [9]. As a member of the healthy gut microbiota, Lachnospiraceae have been implicated in gut inflammation alleviation [10, 11] with recent research focusing on butyrate production [12, 13] that is common among members of this bacterial family. The Lachnospiraceae family is a large complex family that has undergone reclassification with several genera being added to this class from Clostridium cluster XIVa [11, 14, 15]. Members of this family of bacteria are well known for producing short-chain fatty acids (SCFAs) [16], in addition to the SCFA butyrate [17, 18], secondary bile acids [19], essential amino acids [18], and several products involved in reducing radicals and oxidative stress [20].

Many of these Lachnospiraceae-derived products are implicated in the alleviation of host gut inflammation by regulating pro-inflammatory cytokine signaling [21], promoting epithelial repair [22], mitigating reactive oxygen species (ROS) damage [23], and maintaining homeostasis [24]. However, not all Lachnospiraceae members are metabolically equal, due to considerable metabolic and functional diversity within the species [14] resulting in wide variance of metabolism in each species of the Lachnospiraceae family [25]. Notably, phenotypic heterogeneity extends to the strain level within species, with considerable variation in butyrate production, anti-inflammatory capacity, and metabolite output observed even among isolates of the same species. Some species do not produce butyrate [11], and others have a variety of virulence genes [26] including antimicrobial resistance [27] and defensive genes [28]. Moreover, the presence of some species may reduce gut inflammation in ulcerative colitis (UC) but not in Crohn’s disease (CD), both subtypes of inflammatory bowel disease (IBD) [29]. This suggests that a specific community of Lachnospiraceae producing a wide range of beneficial metabolites and products are critical in lowering inflammation and promoting gut repair [25, 30]. The lack of a clear understanding of how Lachnospiraceae-derived metabolites could alleviate gut inflammation hinders the ability to develop targeted therapeutic interventions and diagnostic tools. In addition, it raises further questions about the specific mechanisms by which these metabolites interact with the host immune system to reduce inflammation.

Maintenance of redox homeostasis is a central component of intestinal health. Controlled levels of ROS regulate epithelial signaling, barrier repair, and immune responses, whereas excessive ROS accumulation disrupts epithelial integrity and drives inflammation. Host-derived glutathione (GSH) is well established in its antioxidant capability within the healthy gut, and its function as a node in the circuitry of intestinal signaling with connections to host-cell proliferation and differentiation [31, 32]. During chronic gut inflammation, host GSH synthesis is often impaired, leading to worsened outcomes [33]. Thus, GSH serves as a critical biomarker for host health and is important for promoting homeostasis and gut repair. This study builds upon existing theories that certain bacterial populations within the Lachnospiraceae family are able to induce host GSH synthesis through the production of SCFAs, and further showed that Lachnospiraceae can utilize cysteine, glutamic acid, and glycine, which are needed to biosynthesize GSH [34] to aid in host recovery. During inflammatory states, host-derived GSH synthesis can be suppressed by oxidative burden and pro-inflammatory signaling, impairing the endogenous antioxidant defense. In this context, microbially derived GSH may help restore redox balance and support epithelial repair. Thus, the involvement of Lachnospiraceae-derived amino acids, used in biosynthesis of GSH, that promotes gut homeostasis through eliminating ROS are of essential importance for designing probiotics or other therapeutics [34, 35].

Here, we conducted an in-depth characterization of Lachnospiraceae-derived metabolites, and their ability to alleviate gut inflammation and ROS damage, which provides insights into the immune response of the inflamed host after exposure to microbial metabolites. This study combines a multi-omic in vivo approach with microbiological and molecular in vitro techniques to propose metabolic pathways of interest in Lachnospiraceae that could aid in alleviation of the inflamed gut. We observed that the Lachnospiraceae-derived metabolites have unique profiles that support alleviation of inflammation; however, metabolites derived from Eubacterium rectale lowered cellular stress more effectively than the other strains. We observed a significant reduction of GSH in the cell-free spent media of E. rectale compared with the other strains. Following that, we focused on the metabolic mechanisms of E. rectale-derived GSH and incorporated the host response to shed light on host-microbe interactions and mechanisms that may be involved with stress relief. Our findings mechanistically link associations with Lachnospiraceae-derived GSH, other crucial microbial metabolic pathways, and host’s immune markers of gut health to provide insights into the role of E. rectale in potential protection against gastrointestinal disorder.

Methods

Murine methodology

IL-10 KO mice

Mice in this study were all IL-10 KO Jax stock #002251/C57BL/6 J genetic background, consistent with other studies that emulate human gut inflammation in a mouse model [36, 37]. All mouse experiments were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at Kansas State University (Protocol Approval Number: APN #4391). Mice were bred and housed in the Kansas State University mouse facility under certified Helicobacter hepaticus-free conditions with quarterly sentinel testing for pathogen free status, a facility of the Johnson Cancer Research Center. Murine housing conditions included a 12-h light–dark cycle with each cage holding up to 4 littermates all the same biological sex. This experiment consisted of two different sets of mice, control mice (Control), mice with no dysbiosis, and mice with vertically transmitted dysbiosis (Dysbiotic), to investigate the impact of the resilient microbial populations on inherited dysbiosis (Fig. 1A). Dam’s autoclaved drinking water was supplemented with a broad-spectrum antibiotic, cefoperazone sodium salt (CPZ; 0.5 mg/mL). This exposure began when they were in the third week of gestation, and lasted until 21 days of pup age, to instill inherited dysbiosis [38]. Mice were monitored for signs of colitis daily. Fecal content from (n = 75) pups at weeks 4 (Dysbiotic n = 7, Control n = 6), 6 (Dysbiotic n = 8, Control n = 11), 9 (Dysbiotic n = 8, Control n = 19), and 24 (Dysbiotic n = 7, Control n = 9) of age were used for shotgun metagenomic sequencing, to capture development and effects of inherited dysbiosis as well as before and after treatment with dextran sulfate sodium (DSS). By week 23, the remaining mice (n = 16) received 2.5% DSS for 5 days, with water changed as needed. After a 16-day recovery period from the 2.5% DSS treatment, all mice were sacrificed and tissues, and serum was collected. The difference reflects animal survival and sample availability during longitudinal sampling, as dysbiotic groups had fewer animals due to smaller litter sizes and attrition during DSS treatment.

Fig. 1.

Dysbiotic mice display similar histology scores, metabolite profiles, and gene expressions as control mice. A Graphic of vertically transmitted dysbiosis IL-10 KO murine-induced colitis model. Vertically transmitted dysbiosis was instilled via cefoperazone antibiotic (CPZ) exposure in drinking water of dams during the third week of gestation through weaning of the pups, resulting in dysbiotic pups for the experiment. Control pups did not receive any treatments. Created in BioRender. Lee, S. (2026) https://BioRender.com/j18k679. B Histology images at low magnification and high magnification in both groups of mice. Scale bars are 300 uM in top left image, 600 uM in top right, 150 uM in bottom left, and 150 uM in bottom right image. C Boxplots of histology scores for summary (0–9) scores and broken into categories of inflammation (0–3), surface damage (0–3) and crypt hyperplasia (0–3) with no statistical differences between treatment groups. D Boxplots of colon content metabolites between treatment groups of mice with no significant differences. E Subset of murine serum cytokine violin plots of cytokines typically highly expressed during colon inflammation also result in no statistical difference between groups. Full set of murine serum cytokine are provided in Supplementary Table S1. F Gene Ontology (GO) analysis of gene family expressions clustered by phenotype and genetic diversity of gene classes present in the visualization, in both treatment groups. Statistical difference marked by bolding and text size with blue colored text corresponding to downregulation of a gene family in dysbiotic mice; red text is upregulation of a gene family in dysbiotic mice compared to control. Statistics were obtained through using a one-way ANOVA with Tukey’s HSD, p-value of < 0.05 considered significant

Cytokines from murine serum

Blood was collected via heart puncture and centrifuged to separate out the serum. Serum was stored at − 80℃ until murine cytokines were quantified using a 31-Plex assay (Eve Technologies, Canada, catalog: #33836).

IL-10 KO colon histopathology

A small section of the distal colon from each mouse in the study was fixed in 10% neutral formalin, thinly sliced, embedded into paraffin, cross-sectioned, and finally stained with hematoxylin and eosin. Colon cross-sections were scored for surface damage, type, and quantity of infiltrating inflammatory cells, in addition to gland damage like dilation, inflammation, and hyperplasia. Each category was blindly graded on a scale from 0 (normal), 1 (mild), 2 (moderate), or 3 (severe).

IL-10 KO colon RNA sequencing

Another small portion of the distal colon was stored in TRIzol reagent at − 80℃ until sequencing by CD Genomics (Shirley, NY). CD Genomics performed the extraction and QC analysis. Next, cDNA library preparation (poly-A selection) was completed for n = 12 samples, which were then sequenced via NovaSeq PE150 for a total of 20 million pair-end reads per sample.

Colorectal cells methodology

Bacterial cultivation

We obtained 5 strains of Lachnospiraceae members from ATCC: Anaerostripes hadrus strain B2-52, Blautia coccoides strain CLC-1, Blautia producta strain VPI 4299, Eubacterium rectale strain VPI 0990, and Lachnoclostridium phytofermentans strain ISDg. These strains were decided based on their genomic similarity (80–90% ANI) to Lachnospiraceae MAGs detected in dysbiotic mice, serving as cultured representatives of the dominant taxa identified in the metagenomic analysis. We cultivated these strains anaerobically in a Coy Laboratory Products anaerobic chamber (Coy Labs) in a gas mix (10% H₂, 10% CO₂/N₂) for 48 h in chopped meat media using manufacturers’ suggestions (Anaerobic Systems).

Caco-2 cell culture

We obtained colorectal adenocarcinoma cells (Caco-2) from ATCC and maintained them at 37℃ with an atmosphere of 5% CO₂, 95% air, in 100-cm culture dishes. We maintained the Caco-2 cells in routine cell media consisting of filter-sterilized Gibco Advanced DMEM supplemented with 10% heat inactivated fetal bovine serum, 1% 10⁵ U/L penicillin, and 100 mg/L streptomycin, 1% Amphotericin B (fungizone). Cell generations 8–70 were used for this study with routine image comparisons ensuring no change in visible morphology. For experiments, we plated Caco-2 cells at 1 × 10⁶ cells per well in 24-well tissue culture plates and grew to confluence over 7 days with media changes 3 times per week. Undifferentiated Caco-2 cells were used to model an inflammatory epithelial state, as undifferentiated cells exhibit heightened sensitivity to inflammatory stimuli and are routinely used in this context [39]. To induce inflammation, we introduced cells to 5% high molecular weight dextran sulfate sodium (DSS) (MP Biomedicals) at 3 h and 24 h [40, 41].

Lachnospiraceae-derived metabolite challenge to inflamed Caco-2 cells

We plated Caco-2 cells (generations 25–70) at 1 × 10⁶ cells per well in 24-well tissue culture plates and grew to confluence up to 7 days with media changes 3 times per week. We collected bacterial spent media cultures and centrifuged them for 10 min at 10,000 × g, with supernatant collected, filter sterilized, and diluted in DMEM with DSS for all experiments. Caco-2 cells were washed, with all media aspirated and 5% final DSS solution or 5 mM of hydrogen peroxide in DMEM and bacterial spent media were placed on the Caco-2 cells [42, 43]. We incubated the Caco-2 cells for 3 h, and performed the Promega Griess Assay, measuring nitric oxide (as nitrite) according to the manufacturer’s protocol. We also performed the Promega LDH non-toxic assay according to the manufacturer’s protocol at 6-h timepoints with both assays read on the Accuris smartreader 96. Finally, the DCFDA H2DCFDA Cellular ROS Assay Kit (ab113851) was used to measure intracellular ROS levels, with manufacturer’s protocols followed at 6 h post-introduction and measured using a Biotek Synergy H1 reader.

Metabolomics of cell supernatant

Caco-2 cells were seeded in 24-well plates grown to confluence for 7 days. On the day of the metabolite exposure, we incubated the Caco-2 cells in 1% DSS in cell culture media for 3 h. We incubated with spent media from each Lachnospiraceae strain that were grown for 72 h anaerobically in BHI broth, spun down, and separated. BHI broth was specifically chosen for bacterial cultivation in the metabolomics experiment rather than chopped meat medium, as BHI provides a more chemically defined and reproducible background that reduces the risk of medium-derived metabolite interference in the LC–MS analysis. We then added spent media, and Caco-2 cells were incubated for 3 h under cell culture conditions. A total of 0.5 mL of the cell culture supernatant was frozen at − 80℃ until analysis. Samples were extracted in 1:4 methanol for liquid chromatography-mass spectrometry (LC–MS) analysis (Metware Bio, Woburrn, MA). Ultra performance liquid chromatography (UPLC) (ExionLC 2.0) and quadrupole time-of-flight MS (UPLC QTOF-MS) (TripleTOF 6600 +, AB SCIEX) were performed for widely targeted detection. CFSM from E. rectale with and without chemical depletion of GSH was included in the widely targeted metabolomics following the same methods as above. CFSM was collected after 48 h of anaerobic growth, centrifuged at 5000 × g for 10 min, and filter-sterilized through 0.22 um. Diethyl maleate (DEM) was used for glutathione depletion and was added to CFSM for a concentration of 5 mM and incubated at room temperature under anaerobic conditions for 24 h where it was then flash frozen for analysis [44–46].

Depletion of CFSM reduced glutathione experiments

Spent media was collected after 48 h of bacterial growth, centrifuged, and filter-sterilized. A subset was also being treated with 5 mM of diethyl maleate (DEM) (Sigma-Aldrich Cat. D97703) and incubated for 24 h to deplete reduced GSH levels through a radical oxygen donor [47, 48]. The CFSM with and without reduced GSH depletion was placed on confluent undifferentiated Caco-2 cells that had been exposed to 5 mM of hydrogen peroxide. After 4 h of co-incubation of the CFSM on colon cells, we performed the ROS assay according to protocols mentioned previously.

RNAseq of Caco-2 with Lachnospiraceae-derived metabolites

We seeded the Caco-2 cells at 0.3 × 10⁶ cells per well in 6-well plates overnight using generations 29–50. We then washed the plates with buffer, and treatments were added to the wells and incubated for 6 h in cell culture conditions. All Caco-2 cells except the control received 5 mM of hydrogen peroxide, in addition to the treatment of E. rectale spent media, 15 mM GSH or chopped meat media that E. rectale was cultivated in. After the incubation period, we washed the Caco-2 cells with a cold buffer on ice and scraped in the buffer. Caco-2 cells were then spun down at 1600 rpm and the supernatant was removed. Pellets were then flash frozen and stored in − 80℃ with RNA extraction and sequencing performed by CD Genomics (Shirley, NY). Briefly, total RNA was extracted using TRIzol reagent, followed by DNase I treatment to remove genomic DNA contamination. RNA quality and concentration were assessed using a NanoDrop spectrophotometer and Bioanalyzer (Agilent), with samples requiring an RNA Integrity Number (RIN) ≥ 7 for library preparation. Sequencing libraries were constructed using poly-A selection to enrich for mRNA, followed by first- and second-strand cDNA synthesis, end repair, adapter ligation, and PCR amplification. Libraries were sequenced on the Illumina NovaSeq platform using paired-end 150 bp reads, generating a total of 20 million paired-end reads per sample.

Human colon organoids methodology

Colon organoids experiment

Human colon organoids (batch number: NIO005-258M-25BV, AcroBiosystems, Newark, DE) was maintained in human intestinal organoid medium supplemented with 10 µM Y-27632 and 1% Primocin, at 37 °C in a humidified 5% CO₂ incubator. Enhanced Wnt pathway support was used to support organoid outgrowth. Intestinal organoid maintenance medium (IOMM) was replaced every 2–3 days, and organoids were passaged every 6–10 days to maintain robust growth and crypt-like budding morphology. For the experiment, the organoids were cultured in low-attachment 48-well plates. At 0 h, the control group received 300 µL of IOMM, while H₂O₂, CFSM, and GSH treatment groups received 300 µL of IOMM supplemented with hydrogen peroxide to a final concentration of 100 µM H₂O₂. All treatment groups were incubated at 37 °C with 5% CO₂ under gentle shaker culture conditions for 12 h. Following the initial H₂O₂ exposure, a second 300 µL treatment was added to each well as follows: the control group received 300 µL of fresh IOMM; the H₂O₂ treatment group received 300 µL of IOMM containing 100 µM H₂O₂; the CFSM treatment group received 300 µL of thawed CFSM; and the GSH treatment group received 300 µL of IOMM containing 30 mM GSH, resulting in final concentration of 15 mM GSH. Organoids were incubated for additional 12 h under the same conditions prior to downstream organoid surface area, transepithelial electrical resistance (TEER), and ROS detection measurements and analyses.

Surface area measurements

To quantify changes in colon organoid surface area, individual organoids were transferred into 1.5-mL microcentrifuge tubes containing approximately 1.0 mL of calcium and magnesium-free PBS. The initial liquid level in each tube was marked prior to organoid transfer, and after the organoid was introduced, excess PBS above the marked level was carefully aspirated until the meniscus returned to the reference line, accounting for the displaced volume. The aspirated PBS was collected into a pre-weighted 200 µL microcentrifuge tube and its mass determined. Given the density of PBS is approximately 1 g/mL, the measured mass was numerically equivalent to the displaced volume of the organoid, approximating organoids as spheres; the surface area of the organoids was calculated as surface area = 4ℼr₂, where r = (3 V/4ℼ)^1/3 where V represented the calculated organoid volume. Surface area change was then determined by pre-treatment surface area from the surface area measured after 24 h of treatment.

Transepithelial electrical resistance measurements

To evaluate the effect of the treatments on transepithelial electrical resistance (TEER) in the colon organoids, measurements were obtained prior to treatment and 24 h after the treatment. For each measurement, the organoids were transferred into 6-cm culture dishes pre-filled with PBS, and TEER was recorded using a transepithelial electrical resistance meter. Prior to measuring organoids, the two needle electrodes were immersed in PBS alone to obtain a blank resistance. Subsequently, one electrode was inserted into the center of each organoid while the second electrode remained in the surrounding PBS, and the resistance reading was recorded. The corrected TEER value for each organoid was then obtained by subtracting the blank PBS resistance from the measured value. The same procedure was repeated 24 h following treatments, and changes in TEER were calculated by comparing post-treatment values to the corresponding baseline measurements. The final TEER value for each organoid was calculated by multiplying the measured resistance difference with the estimated surface area of the corresponding organoid.

Reactive oxygen species measurements

Intracellular reactive oxygen species (ROS) levels in intestinal organoids were quantified using a ROS detection assay kit according to the manufacturer’s instructions. Briefly, organoids were incubated with the ROS assay reagent, diluted in the recommended buffer for 30 min at 37 °C under standard culture conditions. Following incubation, fluorescence images were acquired using a fluorescence microscope (Agilent BioTek Lionheart Imager) to visualize ROS-associated signal intensity. For quantitative analysis, three organoids were randomly selected from each experimental group, and one representative image was captured per organoid. Images were processed using ImageJ software, and regions of interest (ROIs) of identical size were applied consistently across all samples to measure fluorescence intensity. To normalize ROS levels, the mean fluorescence intensity of the three control organoids was first calculated to establish a baseline reference. The fluorescence intensity of each treated organoid was then divided by the control group mean to generate a relative fluorescence intensity value. The relative intensities within each treatment group were subsequently averaged to obtain a group mean, and percentage changes in ROS levels were calculated.

Bioinformatic pipeline

Shotgun metagenomics

We used a pipeline that has been described in detail in Richie et al. [49]. Briefly, we automated the metagenomics workflows using the program “anvi-run-workflow” in anvi’o ver 7.1 [50], as well as other programs [51–63]. We assigned the metagenomes into 2 experimental groups (control, n = 45; dysbiotic, n = 30) for co-assembly. We moved forward with all bins that are > 70% complete and < 10% redundant and define these as metagenome-assembled genomes (MAGs). All downstream analyses included only non-redundant MAGs as defined above. We used a “detection” metric to assess the occurrence of MAGs in each murine sample. Detection of MAGs is the proportion of the nucleotides that were covered by at least one short read. We considered a MAG was detected in a metagenome if the detection value was > 0.25.

Human metagenomic analysis

We obtained the genomes from the discovery PRISM cohort (NCBI BioProject number PRJNA400072) with 161 adult patients enrolled with either ulcerative colitis (UC) or Crohn’s disease (CD) and non-IBD control adults [64]. We then mapped the five ATCC Lachnospiraceae genomes we obtained for molecular testing to the PRISM cohort genomes using Bowtie2 v2.3.5 [60] and samtools v1.9 [61] for converting to BAM files. Bowtie2 was used to map metagenomic short reads to each isolate genome, and the proportion of mapped reads was calculated using SAMtools, flagstat, and bedtools genomecov to determine coverage breadth and depth per sample. This allowed us to assess whether these species are differentially present in healthy controls vs UC/CD patients.

BV-BRC analysis

We uploaded and annotated all non-redundant MAGs in Bacterial and Viral Bioinformatics Resource Center (BV-BRC) [65, 66]. All 73 MAGs resolving to the Lachnospiraceae family were grouped, and for each MAG, we used the subsystem genes for comparison of potential functions and virulence factors between the Lachnospiraceae MAGs related to the ATCC cultures obtained. For each of the 73 Lachnospiraceae MAGs, we tallied the number of genes annotated to each BV-BRC subsystem functional category. These counts were used to generate a comparative heatmap and to identify enriched KEGG pathways using the BV-BRC Pathway tool.

IL-10 KO RNASeq analysis

This pipeline was previously described in Richie et al. [49]. We used FastQC [67], multiQC [68], and Trimmomatic [69]. Trinity ver 2.13.1 [70] was utilized for de novo assembly, with RSEM [71] and DESeq2 [72]. Annotation of sequences was carried out using blast + 2.11 and R version 4.0.0. The transcripts compared in DESeq2 were first blasted against mouse reference genome C57BL/6NJ (http://ftp.ensembl.org/pub/release-107/fasta/mus_musculus_c57bl6nj/cdna/), with the top hit kept. Finally, we used GO_MWU [73] to analyze the annotated transcripts by the treatment group of mice.

Caco-2 RNAseq analysis

A nf-core/RNAseq pipeline v3.12.0 based on nextflow (v 23.04.3.5875) with a singularity container was chosen for the RNA data analysis. Default parameters and tools are used in the pipeline except that the raw reads were mapped with STAR and quantified using RSEM [71] with the reference genome GRCh38 release-110. DESeq2 [72] package was used for the downstream statistical analysis. Gene ontology (GO_MWU) enrichment analysis [73] was used to analyze annotated transcripts.

Colon organoids RNAseq analysis

RNA sequencing of human colon organoids was performed on the Illumina NovaSeq platform (paired-end, 150 bp). Raw reads were assessed using FastQC [67], and processed with fastp [74] to remove adapter sequences, reads containing ambiguous bases, and low-quality bases (Q < 20), generating high-quality clean reads. Clean reads were then aligned to the human reference genome using HISAT2 [75], and mapping quality was evaluated using RseQC [76] and Qualimap [77]. Gene expression levels were quantified using featureCounts [78], and normalized expression values (TPM/FPKM) were calculated. Differential gene expression analysis between treatment groups was performed using DESeq2 [72], with significance determined using Benjamini–Hochberg adjusted p values (FDR < 0.05). Alternative splicing events were analyzed using rMATS [79], and transcript assembly with novel isoform prediction was conducted using StringTie [80] followed by GffCompare [81]. Functional enrichment analyses of differentially expressed genes were performed using topGO [82] for KEGG pathway analysis.

Manuscript statistical analyses

Rstudio was used for all analyses, the one-way ANOVA with Tukey HSD test, unless stated otherwise. Prism 11 was utilized for statistical analysis of Fig. 7; one-way ANOVA with Tukey HSD was conducted. Normality was assessed with the Shapiro–Wilk test and variance homogeneity with Levene’s test. Two-group comparisons used Student’s t-test or Wilcoxon rank-sum test as appropriate. Multi-group comparisons used one-way ANOVA with Tukey’s HSD. p-values of less than 0.05 were considered statistically significant. We used the program Data Integration Analysis for Biomarker discovery using Latent cOmponents (DIABLO) to build a model to identify key components in the metagenomic data and host RNASeq data [83, 84]. We (i) first identified key omics variables to further investigate; (ii) then maximized the common and correlated information between the datasets; and (iii) visualized the results to identify a network of relevant MAG detections compared with host genomic markers. Univariate and multivariate analyses were performed on metabolomics data to identify the metabolites from different treatments (Metware Bio, Woburrn, MA).

Fig. 7.

Colon organoids validate cell line study demonstrating CFSM leads to homeostatic physiology and lower ROS levels as well as higher expression of repair gene pathways reducing ROS. A Bar plot of organoid surface areas by treatment group (n = 3 per group for all tests). B Microscopy images of organoid edges demonstrating rounded edges of control organoids and serrated edges of the H₂O₂-treated organoids. Scale bars indicate 500 um. C Bar plot of Transepithelial/Transendothelial Electrical Resistance (TEER) of colon organoids in each treatment group. D Bar plot of fluorescent intensity of reactive oxygen species (ROS) of each treatment group. E Heatmap of differentially expressed genes using log fold change of RNA sequencing from colon organoids of each treatment. F Venn diagram of differentially expressed gene counts in each treatment category and comparison. G Venn diagram of transcript per million (TPM) expression analysis demonstrating a core and grouped set of genes by treatment comparison. H Bar plots displaying significantly different KEGG gene pathways between organoids treated with CFSM and H₂O₂ bar colors indicate pathway type. I Bar plot of highest enriched gene pathway categories shown by the enrichment factor and Q-value labeled as the color. J Bar plot of KEGG-enriched pathways within glutathione metabolism in the CFSM-treated organoids with Q-value displayed as the color of the bars. One-way ANOVA with Tukey HSD was utilized and key statistics are marked on the graph with * < 0.05, ** < 0.005, and *** < 0.0005

Results

Mice in the dysbiotic group displayed vertically transmitted dysbiosis from birth due to dam exposure to cefoperazone, and murine pups developed for 23 weeks until 2.5% DSS was administered at the endpoint of the study to induce colitis in all mice. Control mice originated from untreated IL-10 knockout mice and were monitored for 23 weeks until DSS exposure in drinking water (Fig. 1A; Supplementary Table S1). We used genome-resolved metagenomics for high-resolution characterization of the gut microbiota, combined with host gene expression and microbial metabolite analyses to explore the impact of Lachnospiraceae members on host recovery from dysbiosis.

Shotgun metagenomic sequencing of fecal samples resulted in over 1,000,437,831 sequences with an average of approximately 14,000,000 reads per sample. Assembly of sequences from the dysbiotic group (n = 30) and control group (n = 45) resulted in 439,024 contigs longer than 1000 nucleotides. We observed a clear pattern of vertically transmitted dysbiosis belonging to several microbial classes across each sample compared to the control mice immediately after cefoperazone exposure ended (Supplementary Figure S1). Our ultimate goal of this study was to monitor the outcomes of vertically transmitted dysbiosis throughout mouse development and investigate potential microbial members and their mechanisms influencing recovery.

Dysbiotic mice showed minimal differences from control mice in colon histology and cytokine markers

To gain insights into the host response after vertically transmitted dysbiosis, we performed histology and a broad cytokine panel on murine colons and serum respectively (Supplementary Table S1). As seen in previous studies, we expected dysbiotic mice with induced colitis to display more severe clinical signs of inflammation compared with control mice with induced colitis, which highlights the importance of a diverse gut community on immune regulation [85–87]. Surprisingly, we observed that offspring from dysbiotic dams and control mice displayed similar low-grade colon inflammation with signs of neutrophil infiltration and slight surface damage at week 24, due to the chemically induced colitis from DSS (Fig. 1B). Overall, dysbiotic mice were not significantly different from mice in the control group in all grading categories including inflammation, surface damage, and crypt hyperplasia scores (Fig. 1C). It was surprisingly evident that the dysbiotic mice might have recovered from dysbiosis during development as there are no significant differences in colon inflammation between groups.

We further evaluated serum systemic markers using a 31 multiplex pro-inflammatory cytokine panel for all mice that survived until week 24. Similar to our observations regarding colon inflammation, 30 of the 31 serum cytokines measured displayed no significant difference between dysbiotic and control mice (Fig. 1E). Monokine induced by gamma interferon (MIG) was significantly higher in dysbiotic mice (p = 0.0061) than in control mice. MIG (also referred to as CXCL9) is a chemokine that is implicated in gut inflammation, however is usually accompanied with other cytokines including IL-1ɑ, IL-10, IL-17, and TNF-ɑ during an immune response [88]. MIG is also activated in response to the introduction of a gut microbiota in both murine and cell models [89]. Because this was the only cytokine with elevated levels in the dysbiotic group, we suspect that this increased level of MIG in dysbiotic mice was due to the reappearance of gut microbes during development [90, 91]. TNF-ɑ was significantly higher in control mice (p-value = 0.047), further suggesting that the dysbiotic mice were not displaying an increased pro-inflammatory trend. In summary, we see little differences in cytokine profiles between treatment groups. Thus, we surmise if there was a microbial associated mechanism involved in the resulting similarity in cytokine profiles of the dysbiotic mice and the control mice in our study.

To ensure a persuasive hypothesis that the dysbiotic mice displayed a similar host status to control mice, we conducted a global metabolite panel and host RNAseq from each treatment group. There were no significant differences between the measured metabolites between treatment groups, including short chain fatty acids like acetate, propionate, and butyrate, or important amino acids involved in gut repair and homeostasis [92, 93], including cysteine and glutamate (Fig. 1D). To gain perspective on the host response, RNA sequencing on mouse colons from each treatment also resulted in minimal gene family regulation changes, with 63 significantly regulated gene families between treatments (Fig. 1F). These data suggest that there is not a clear difference of gene family regulation, and together with the lack of inflammatory gene upregulation, point to the recovery of the mice in the dysbiotic group.

Increase in Lachnospiraceae bacterial populations in dysbiotic mice associated with lowered host pro-inflammatory cytokine

The following work stems from the fact that there were no statistical differences in the histology, cytokine expressions, metabolite profiles, and host gene regulations, which suggests mice recovered from vertically transmitted dysbiosis. We were then curious if the gut microbiota played a part in the similar status of these dysbiotic mice to control mice. Therefore, to characterize the gut microbiota of dysbiotic mice, we performed shotgun metagenomic sequencing on the fecal samples at weeks 4, 6, 9, and 24 of age (Supplementary Table S2). We resolved 158 non-redundant metagenomes-assembled genomes (MAGs) (Supplementary Figure S1), with 73 MAGs annotated to the Lachnospiraceae family (Supplementary Table S2). We noticed Lachnospiraceae MAGs displayed an interesting longitudinal pattern in the dysbiotic mice (Fig. 2A). Although low in detection until week 6 (Supplementary Table S2; Fig. 2A and B), we saw a significant increase in Lachnospiraceae in the dysbiotic mice after week 9 that persisted throughout the endpoint of the study compared to control mice (Supplementary Table S2; Fig. 2A and B). Control mice, on the other hand, did not display this pattern (Supplementary Table S2; Fig. 2A and B). Interestingly, we noticed that control mice had the highest detection of Lachnospiraceae at week 6, which implicated their presence earlier in life as opposed to the dysbiotic group where this increased detection was seen later at week 9 of age.

Fig. 2.

Increased Lachnospiraceae detection over murine development correlated with downregulation of pro-inflammatory genes. A Clustered heatmap with rows corresponding to mouse samples sorted by treatment group and age in weeks. Constant detection levels are observed in control mice; however, over time, Lachnospiraceae increase over time in the dysbiotic mice. B Boxplot of Lachnospiraceae MAG detections by week comparing control mice (green) to dysbiotic mice (gold). Significant p-values from one-way ANOVA with Tukey HSD marked by asterisks * < 0.05, ** < 0.005, *** < 0.0005, and **** < 0.00005. C Network analysis of discriminant analysis of RNA sequencing gene counts (red) coupled with MAG detection levels (gray). A correlation cutoff of 0.95 was used to only visualize strong associations with (green) lines indicating a negative correlation, and (red) lines marking a positive correlation to the detection of a MAG with the upregulation of a gene

We wanted to determine if an increase in Lachnospiraceae populations was associated with beneficial host response patterns. Thus, to link this pattern with the recovery of the dysbiotic mice, we performed a discriminant analysis on MAG detections and host RNA sequencing counts. Gene counts for pro-inflammatory cytokine receptors, including the pro-inflammatory IL-17, IL-6, and IL-1 receptors, as well as the anti-inflammatory IL-10 receptor [94] depicted in red circles, were negatively correlated (displayed with a green or dark red interaction line) with the higher detection of Lachnospiraceae MAGs depicted in gray circles (Fig. 2C). Our results indicated that Lachnospiraceae MAGs might interact with the host immune system to lower pro-inflammatory cytokine expression, which effectively reduced the host’s immune responses [95, 96]. However, with the high diversity and heterogeneity within the Lachnospiraceae family, it is difficult to dissect which members of Lachnospiraceae are driving the alleviation of oxidative stress or inflammation. In the next sections, we examined all 73 murine MAGs and compared them to five Lachnospiraceae strains that are implicated with a healthy gut, to further drive at a mechanism Lachnospiraceae employ in promoting gut homeostasis.

Metabolic pathways of Lachnospiraceae MAGs included antioxidant and secondary metabolite productions that may benefit host

With the Lachnospiraceae bacterial populations displaying a unique detection pattern longitudinally in our study, we further investigated the potential functions of these 73 MAGs that could result in possible host recovery. We used Bacterial and Viral Bioinformatics Resource Center (BV-BRC) [66] genome annotation and subsystems to visualize and quantify the metabolic potential of all 73 MAGs. The highest number of genes by pathway for all MAGs was GSH metabolism with 225 genes found among the 73 MAGs and resulted in 100% coverage of the pathway (Fig. 3A). Six other pathways included in the analysis involved the metabolism of amino acids, including glycine, serine, threonine, arginine, proline, lysine, and cysteine. Amino acid metabolism within microbes is often the beginning step toward production of SCFAs with the amino acids mentioned above being involved in the production of butyrate [97, 98]. With GSH metabolism being the highest represented pathway, we next investigated how microbial GSH assimilation could impact host outcomes. The following section will discuss the rationale for choosing Lachnospiraceae isolates that are relevant to human health and have similar metabolic products to the murine-derived MAGs.

Fig. 3.

Genus level match of Lachnospiraceae MAGs to commercially available isolates using ANI, presence of beneficial microbial genes and pathways, and isolates detected lower in diseased patients. A Balloon plot of metabolic pathway summary of all 73 murine Lachnospiraceae MAGs. Size of the balloon indicates either the number of genes in the corresponding pathway or the percent coverage of the entire pathway shown. B Average nucleotide identity (ANI) resulted in around an 80–90% match of available Lachnospiraceae to murine-detected MAGs. Murine-detected isolates are indicated in red; the available isolates are marked in the legend. C Mapping the genomes of the available 5 ATCC isolates to human samples from a subset of the Discovery cohort indicated that patients with ulcerative colitis and Crohn’s disease were significantly lower in almost every strain. Significant p-values from one-way ANOVA with Tukey HSD were marked with asterisks with * < 0.05, ** < 0.005, and *** < 0.0005. D Circulized genome of E. rectale marking two genes in the pathway of GSH biosynthesis with their location on the genome and the length of the genes. Boxplots represent the mean coverages of the bcp and ArgC genes in an IBD vs healthy patient cohort. Two tailed T-tests indicate significance of less than 0.05

Lachnospiraceae members are detected higher in healthy human samples as an indicator of beneficial members

Most MAGs resolving to the Lachnospiraceae family were not annotated to species level, except for the genera Blautia, Acetatifactor, Anaerostipes, Clostridium, Eubacterium, Enterocloster, Robinsoniella, Kineothrix, Lachnoclostridium, Anaerosacchariphilus, and Schaedlerella (Supplementary Table S2). An average nucleotide identity (ANI) (Supplementary Table S3) resulted in Lachnospiraceae MAGs matching around 80–90% to the available ATCC isolates, indicative of matching at the genus level, but not reaching 95% threshold for likely species level identification [99] (Fig. 3B). With such wide variety among the isolates and the MAGs, we obtained all five strains from ATCC that were genomically similar to the MAGs that were highly detected in dysbiotic mice: Anaerostipes hadrus B2-52, Blautia coccoides CLC-1, Blautia producta VPI 4299, Eubacterium rectale VPU 0990, and Lachnoclostridium phytofermentans ISDg. With isolates spanning 3 genera and a set within the same genus, these isolates provide insights into functional differences and determine which strain has more potential to alleviate host inflammation.

In order to understand the prevalence of these strains in human populations, we mapped the isolates’ genomes to the PRISM cohort study of IBD patients diagnosed with ulcerative colitis (UC) and Crohn’s disease (CD) (Supplementary Table S3), compared to non-IBD patient samples [64] (Fig. 3C). Anaerostipes hadrus was detected significantly higher in the healthy control samples (p-value < 0.05), but not in the UC group and the CD group (p-value = 0.57). Lachnoclostridium phytofermentans was detected significantly higher in the healthy control samples as compared to the CD group (p-value = 0.00001), but not statistically significant when compared to the UC group (p-value = 0.089). Interestingly, the Blautia genus displayed a different detection pattern. Blautia coccoides was detected at lower ratios in the non-IBD control samples, however, only compared to the CD group (p-value = 0.032), not the UC group (p-value = 0.059). Blautia producta was also detected lower in the non-IBD control samples compared to the UC group (p-value = 0.043) and in the CD group (p-value = 0.027). Eubacterium rectale was detected much higher in the non-IBD control samples than the UC group (p-value < 0.05) and the CD group (p-value < 0.05). A higher prevalence in the healthy group versus both disease groups suggests that these microbes may be involved in maintaining homeostasis and may also play a role in recovery from inflammation. These results clearly indicate the presence of Lachnospiraceae populations in control samples, and are depleted in patients with inflammatory diseases, which suggests the importance for the presence of these strains of Lachnospiraceae, especially A. hadrus and E. rectale.

Microbial-derived GSH production in Eubacterium rectale matching with higher detection of GSH pathway genes in healthy patients

Further investigation into the genome and potential functions of E. rectale revealed two genes involved in GSH biosynthesis and recycling, bcp and ArgC (Fig. 3D). Genomic analysis of E. rectale did not identify canonical gshA/gshB homologs, which are well-known microbial GSH biosynthesis pathways. Instead, E. rectale appears to utilize an alternative pathway involving bcp (bacterioferritin comigratory protein/glutathione peroxidase) and ArgC (N-acetyl-gamma-glutamyl-phosphate reductase), which we identified and functionally contextualized, making E. rectale’s GSH mechanism novel. The bcp gene is a thioredoxin peroxidase, as well as a glutathione peroxidase which allows for the microbial use of GSH against ROS, and is involved in recycling of oxidized glutathione (GSSG) back to reduced glutathione (GSH) to be utilized again [100, 101]. Interestingly, overexpression of the bcp gene leads to increased levels of GSH in the intracellularly and is generally considered a bacterial stress response to ROS levels [101, 102], suggesting a role in GSH production as well. In addition to the bcp gene, the ArgC gene is involved in arginine and ornithine biosynthesis, through glutamate utilization which converts the glutamate to glutamic acid, used in biosynthesis of GSH [103]. In correlation with the low detection of E. rectale in inflamed patients (Fig. 3C), we also observed that the average gene coverages of both bcp (p-value = 0.002) and ArgC (p-value = 0.023) were significantly lower in inflamed patients compared to the healthy control (Fig. 3D), suggesting that this pathway of GSH biosynthesis and recycling was found in commensal species like E. rectale.

With the gut environment having a large abundance of GSH up to the millimolar level [44], understanding the impact GSH has on the host as well as the microbiota in inflamed conditions is important in host recovery from events like chronic inflammation and oxidative stress. In this study, we show that E. rectale is able to biosynthesize GSH (Fig. 3C and D), and future investigations will benefit from the ability of microbial-derived GSH in aiding host alleviation of inflammation and oxidative stress.

Species level differences in gut oxidative stress alleviationin vitrousing induced ROS damage

With the understanding that each Lachnospiraceae strain is functionally diverse across multiple metabolic pathways, we investigated how these differences could affect host oxidative stress or inflammatory outcomes. Lachnospiraceae have long been known as major producers for SCFA and butyrate [13, 104], which can help in gut repair and lowering oxygen levels in the colon and establishing local microenvironment for inflammation recovery. We surmised that each of the five Lachnospiraceae strains in this study has unique metabolites that may impact the ability to aid in host recovery from oxidative stress resulting from two separate exposure tests to DSS or hydrogen peroxide (H₂O₂). These two stressor exposures, either DSS or H₂O₂, were utilized because the immune mechanisms affected by DSS are strikingly similar to colitis [105, 106], whereas H₂O₂ causes ROS damage often inflicted by more aerobic pathogenic gut microbes present in the dysbiotic gut [107, 108]. Because GSH is an antioxidant [35], we also sought to determine if GSH alone is capable of reducing cellular peroxide stress to the physiological level we observed in our study.

Immune activation and cell death were quantified by measuring nitric oxide [109, 110], an immune system by-product indicative of an immune response. Additionally, we quantified ROS levels, which indicates host damage and oxidative stress, as well as cytotoxicity to capture cell death [111, 112] (Supplementary Table S4). We observed that E. rectale (p-value = 1.4 × 10⁻⁶) and B. coccoides (p-value = 4 × 10⁻⁷) cell-free spent media (CFSM) significantly lowered nitric oxide levels compared to DSS-inflamed Caco-2 cells, while none of the strains lowered nitric oxide levels in the H₂O₂ exposed Caco-2 cells (Fig. 4A). We further noticed that GSH did not significantly lower nitric oxide levels at any concentration in either stress exposure (Fig. 4B). Our result suggests that GSH alone may not contribute to lowering the immune response; however, it is likely involved with alleviating ROS damage. On the other hand, CFSM from E. rectale and B. coccoides may be involved in diminishing the immune response.

Fig. 4.

Molecular assays quantifying immune status, ROS damage, and cytotoxicity of each Lachnospiraceae strain’s spent media or reduced glutathione (GSH) on Caco-2 cells inflamed by 5% DSS or ROS stressed with 5 mM hydrogen peroxide (H₂O₂). CFSM GSH depletion results in increased ROS in stressed colon cells, highlighting the role of reduced glutathione in alleviation of ROS. Top left panel indicates the legend with the top graphs displaying DSS stress and the bottom graph in each panel displaying H₂O₂ stress. (A) Griess assay was used to quantify the immune response by-product nitric oxide. Boxplots show that B. coccoides and E. rectale are significantly lower than the inflamed cell control and chopped meat inflamed control for DSS, and no strains significantly lowered nitric oxide levels when cells are stressed in H₂O₂. (B) GSH was tested at several half log concentrations in cells stressed with DSS and H₂O₂ with no statistical significance in lowering nitric oxide in either stressor. (C) Boxplots showing ROS levels highlight no strain’s spent media lowered ROS levels in DSS inflammation; however, E. rectale, B. coccoides, and B. producta were significantly lower than the chopped meat inflamed control in H₂O₂ stress. (D) GSH at any concentration did not lower ROS levels in DSS stress, but at 15 mM GSH and higher concentrations was significantly lower than H₂O₂-treated cells. (E) Cytotoxicity boxplots highlight the spent media lowered toxicity in DSS-inflamed cells, with A. hadrus, and B. producta-treated cells having significantly lower cytotoxicity than inflamed control. However, with H₂O₂-treated cells, cytotoxicity B. coccoides, B. producta, and A. hadrus, and E. rectale displayed higher cytotoxicity than the H₂O₂-treated control. (F) Increased levels of GSH over 15 mM induced high levels of cytotoxicity in both treatments, with lower levels showing no difference from control cells. Significance (p-value < 0.05) symbolized by comparison (a = significantly different from Caco-2 cells only; b = significantly different from the stress control DSS or H₂O₂; and c = significantly different from Chopped meat media control with stressor DSS or H₂O₂). Figure legend on the left side of the graph indicates the color scale representing the concentration of GSH, and indicates that the top graph is testing DSS and the bottom graph in each subfigure tested H₂O₂. Differences between groups were analyzed using one-way ANOVA with Tukey HSD. (G) Two-dimensional principal component analysis of non-targeted metabolomics of the CFSM and CFSM treated with 5 mM diethyl maleate (DEM) for 24 h. Overlap of both groups indicate no significant difference in the overall makeup of CFSM when treated with DEM. (H) Violin plot of observed oxidized glutathione levels resulting from metabolite analysis. (I) Boxplots of GSH concentration in bacterial media controls with and without supplementation of GSH and depletion of GSH through DEM. (J) Boxplots of peroxide-stressed colon cells challenged with CFSM or CFSM with the GSH depletion using DEM treatment. To further ensure the impact of DEM treatment on GSH, we also supplemented cells with 15 mM GSH and GSH treated with 5 mM DEM. One-way ANOVA with Tukey HSD was utilized and key statistics are marked on the graph with * < 0.05, ** < 0.005, and *** < 0.0005

CFSM from E. rectale displayed the lowest ROS levels with DSS stress and was not significantly different from control Caco-2 cells while other strains (B. coccoides p-value = 0.0002; B. producta p-value = 0.0001; A. hadrus p-value = 0.038) displayed significantly higher ROS levels in the DSS exposed Caco-2 cells. In H₂O₂-exposed Caco-2 cells, E. rectale (p-value = 1.8 × 10⁻⁶), L. phytofermentans (p-value = 0.041), and B. coccoides (p-value = 0.0074) showed significantly lower ROS levels than the H₂O₂ control (Fig. 4C). While there are studies on microbial-produced GSH [113, 114], our results illuminate the genetic differences among the various members in the Lachnospiraceae family, in the biosynthesis of GSH, with E. rectale shown as the largest producer of GSH. There were no significant differences in ROS levels stressed with DSS for all concentrations of GSH; however, concentrations at and higher than 15 mM (p-value = 9 × 10⁻⁵) did lower ROS significantly with H₂O₂-stressed Caco-2 cells (Fig. 4D). This suggests that a high amount of GSH (≥ 3.75 mM) is involved in lowering ROS levels when a ROS stress is implemented, but adding GSH did not lower ROS levels when there was an inflammation stressor. Cytotoxicity remained low for all DSS exposed Caco-2 cells; however, B. producta (p-value = 3.8 × 10⁻⁷) and A. hadrus (p-value < 0.05) CFSM were significantly lower than the stressed control. H₂O₂-exposed cells revealed similarly that B. coccoides (p-value < 0.05), B. producta (p-value < 0.05), and A. hadrus (p-value < 0.05) were all significantly lower than the stressed control, with E. rectale (p-value < 0.05) CFSM significantly higher in cytotoxicity (Fig. 4E). We also noticed that GSH showed an interesting pattern of increasing cytotoxicity after 25 mM for DSS-stressed cells, and over 10 mM in the H₂O₂-exposed cells (Fig. 4F). This is likely due to the low pH shift in solution after 15 mM GSH, while most cellular concentrations of GSH range from 1 to 10 mM [32]. In summary, our results highlight that in both stressors and outputs of inflammation, E. rectale CFSM was more effective in lowering levels of both nitric oxide and ROS, whereas other strains, like B. producta and A. hadrus, might have more protection against cytotoxicity rather than lowering inflammation and ROS damage. Here, we also showed that GSH is able to lower ROS levels significantly, but does not significantly impact nitric oxide levels, which further stresses its importance as a ROS alleviator and antioxidant at cellular concentrations.

Oxidation of reduced glutathione in Eubacterium rectale CFSM highlights role of GSH in lowering oxidative stress in colon cells

In order to quantify the impact GSH has on colon cell ROS alleviation in the complexity of the microbial CFSM, we performed microbial GSH depletion in the E. rectale CFMS using diethyl maleate (DEM). DEM is well established in targeting specifically reduced glutathione and reducing the availability through a covalent bond [46, 115, 116]. This action renders glutathione unable to accept radicals, thus eliminating its antioxidant function [45, 115, 116]. In this study, chemical depletion of microbial GSH was preferred to functional deletion of GSH production through gene knockout, due to the targeted nature of DEM binding to GSH, and not altering microbial metabolism [117]. In doing this, we can investigate and quantify the impact of E. rectale derived GSH in reducing cellular ROS stress. However, to ensure that treating the CFSM with DEM did not significantly alter the makeup of CFSM, we used non-targeted metabolomics on samples of CFSM and CFSM treated with 5 mM DEM for 24 h. The GSH depletion treatment resulted in very similar profiles of metabolites and did not differ from the untreated CFSM (Fig. 4G). Additionally, oxidized glutathione levels were measured in the CFSM and CFSM with DEM depletion to investigate off target impacts of DEM on similar biochemical structures (Fig. 4H). There was no significant difference after DEM treatment demonstrating the targeted nature of DEM treatment and maintains the overall complexity of the CFSM. We further showed that there was a significant decrease in concentration of E. rectale-derived GSH in the CFSM when treated with DEM (Fig. 4I).

Next, we aimed to illustrate the impact of E. rectale derived GSH within the CFSM on stressed colon cells by comparing CFSM to the GSH-depleted CFSM. Similarly to the above protocols, we stressed colon cells with H₂O₂ and challenged them with CFSM or CFSM with 5 mM DEM. Corroborating with our previous results, we showed that CFSM significantly lowers ROS compared with the H₂O₂-stressed colon cells (p-value < 0.05) (Fig. 4J). Eubacterium rectale-derived CFSM with DEM resulted in significantly higher ROS intensity (p-value = 0.014), however was still lower than the stressed cells (p-value = 0.02). This difference in CFSM and CFSM with DEM depletion of GSH directly illustrated the impact of E. rectale-derived GSH in lowering ROS intensity in stressed colon cells. Thus, highlighting the role of E. rectale derived GSH in alleviation of ROS as a potent antioxidant in oxidative stress.

Functionally diverse Lachnospiraceae have inflammatory alleviating metabolites and antioxidant GSH

With the results showing the CFSM of Lachnospiraceae alleviated cell inflammation, next we investigated the microbial products of these five strains. We used a widely targeted metabolite analysis to investigate the metabolic profile of CFSM from Lachnospiraceae, and to identify potential metabolic pathways that contribute to the alleviation of host inflammation (Supplementary Table S5). We observed that amino acids had the highest represented class at 56% of all detections in every Lachnospiraceae metabolite profile, with other homeostasis promoting classes [118] including fatty acids (3.5%), bile acids (1.2%), and carbohydrates (4.4%) also represented. Next, a comparison of metabolites of each Lachnospiraceae strain CFSM revealed a core of 82 metabolites in common. However, each strain consisted of unique metabolic products Blautia coccoides (n = 6), Blautia producta (n = 7), Lachnoclostridium phytofermentans (n = 6), Anaerostipes hadrus (n = 15), and Eubacterium rectale had the highest number of 47 unique metabolites compared to the other strains (Fig. 5A). Eubacterium rectale-specific metabolites include L-tryptophan, L-glutamine, L-histidine, and isobutyric acid. Two of the 47 unique metabolites also include oxaloacetic acid and indole-3-acetic, which are responsible for fatty acid synthesis and oxidative stress reduction [119], and known to alleviate colitis-associated immune responses, respectively [120, 121]. The distinct metabolic profile of E. rectale CFSM is evident in Fig. 5A and B, where E. rectale harbored the highest number of unique metabolites (n = 47), and was the predominant producer of key anti-inflammatory metabolites including glutathione, L-histidine, L-taurine, and L-cysteine, suggesting that this strain contributed to the attenuation of host inflammation through multiple metabolic pathways..

Fig. 5.

Metabolomic analysis of five Lachnospiraceae strains spent media indicative of functional diversity in microbial family, large production of amino acids, SCFAs, and bile acids. A Venn diagram of all five strains and a Caco-2 cell control, showcasing the unique metabolites of each strain’s spent media, with E. rectale having 47 unique metabolites from the other four strains. B Clustered heatmaps displaying relevant metabolites in gut repair including amino acids, glutathione pathways, bile acids, and short chain fatty acids in all five strains. Normalized scale bars to the right of each graph indicate the observations of each sample. Notably, E. rectale spent media consists of a variety of metabolites as well as having some of the highest counts of metabolites including glutathione, indole, taurine, and several carnitine derivatives. C Balloon map of gene subsystems counts for E. rectale by metabolism pathway highlighting potential functional capacity of production of GSH as well as several amino acids with circle size and color indicating number of genes in the pathway. D Boxplot of GSH levels in each sample from metabolomic analysis with significance from a one-way ANOVA with Tukey HSD marked by a p-value of * < 0.05, ** < 0.005, and *** < 0.0005

We investigated the metabolic profiles of each Lachnospiraceae strain to determine which strains are major producers of metabolic classes related to host recovery. Eubacterium rectale was a major producer with the highest counts of carnitine isomers, histidine, L-lysine, glutathione, and L-taurine compared to the four strains (Fig. 5B). Fatty acids were produced highly by all five strains of Lachnospiraceae with E. rectale producing the highest amount of 2-thio-PAF and oxaloacetate acid. Bile acids were also produced by all five strains, with A. hadrus being the major producer, and E. rectale having high measurements for mucic acid and 7-ketodeoxycholic acid. Finally, we examined GSH pathway metabolites and showed that all five strains have partial production of some key metabolites, like N-acetyl-S-propyl-L-cysteine, D-ornithine, and nonenal glutathione. However, E. rectale produced the endpoint metabolites including N1-acetylspermidine, N-acetylcysteine, and both reduced and oxidized forms of glutathione (Fig. 5B).

We used metabolomics and cellular assays to demonstrate that E. rectale CFSM significantly reduced intracellular ROS levels and nitric oxide production in oxidatively stressed Caco-2 cells while maintaining low cytotoxicity, effects that were likely mediated in part by its unique repertoire of metabolites with well-established antioxidant and gut-reparative properties, including L-carnosine, L-cysteine, glutathione, L-histidine, and L-taurine. L-carnosine is involved with gut integrity and wound repair [122], while L-cysteine is implicated in gut lining repair [92, 123]. L-histidine is implicated in suppression of pro-inflammatory cytokines and plays a role in diminishing colitis [124] in addition to L-taurine, which promotes goblet cell formation and aids in tight junction regulation [125, 126]. L-cysteine, L-histidine, and L-taurine have also been considered for supplementation to patients with IBD as these patients often have very low serum levels of these amino acids [33, 127]. GSH is a well-documented antioxidant aiding in ROS damage and driver of an anaerobic gut environment [31, 32]. We quantified the GSH levels across strains and the colon cell controls, showing that there was significantly more GSH in E. rectale than the other four strains of Lachnospiraceae (Fig. 5C). Our results strongly suggest that the highly effective ROS alleviating characteristics of E. rectale was in part due to the high production level of GSH. We further confirmed the biosynthesis capacity of GSH by E. rectale, and found several biosynthesis pathways of GSH, as well as the precursor amino acids like L-cysteine synthases (Fig. 5D). We also noticed several other microbial metabolism mechanisms that can aid to alleviate host inflammation, such as fatty acid synthesis, tryptophan, and several other amino acid biosynthesis pathways. We next focused on E. rectale and its impact on host response, as well as investigation of E. rectale-derived GSH on alleviation of cellular inflammation. GSH was chosen as the focus of investigation due to its significantly elevated levels measured in E. rectale spent media relative to the other strains, and its well-established potential as an antioxidant, in order to provide mechanistic insights into the protective effect of E. rectale on host oxidative stress and inflammation.

Cell response of peroxide damaged Caco-2 cells exposed to Eubacterium rectale CFSM and reduced glutathione showed downregulation of defense and repair genes

Expanding on the cellular assay experiment, we investigated the global host response when exposed to H₂O₂, compared to the response of ROS stress with treatment of E. rectale CFSM. We evaluated the antioxidant GSH in lowering ROS cell stress because it is a microbial product from E. rectale and is protective against ROS damage. We observed patterns of highly differentially regulated genes in each treatment comparison (Supplementary Table S6). We performed a gene ontology analysis (GO) to investigate gene class differential regulation, and showed that only four gene families, glycosyltransferase, hydrolase, RNA binding, and carbohydrase kinase, were (n = 3) upregulated in the E. rectale CFSM treatment. This demonstrated a differential regulation between ROS-stressed cells treated with E. rectale CFSM compared to control, unstressed cells (Fig. 6A). Our results indicate that treating ROS-stressed cells with E. rectale CFSM significantly reduced host damage, and upregulated gene repair pathways in ROS-stressed cells. Compared with E. rectale CFSM treatment, ROS-stressed cells treated with 15 mM of reduced GSH revealed 20 gene classes differentially regulated, with classes including transferases, signal receptor binding, electron transfer, and extracellular matrix structure constituents. Comparing the H₂O₂ treated with GSH and the unstressed controls, 18 classes were differentially regulated (Fig. 6A). GSH regulation patterns were different compared with the E. rectale CFSM; however, they also displayed a differential gene class pattern from inflamed cells (Supplementary Figure S2). GSH-treated cells exposed to H₂O₂ indicated 31 gene classes differentially upregulated (n = 17) compared to inflamed controls. Our data suggest that GSH increased the ability of different gene family pathways to reduce ROS damage to host cells, compared to E. rectale CFSM.

Fig. 6.

RNA sequencing of peroxide stressed Caco-2 cells treated with CFSM of E. rectale resulted in little gene differentiation from unstressed Caco-2 cells, and treatment with reduced GSH reduced several damage repair genes and environmental sensing genes. A Volcano plots of DEseq2 DEG comparison analyses with a p-value < 0.05 and fold change of > 0.6 with the top 30 differentially expressed genes labeled on each plot. Color scale depicts non-significantly regulated genes as gray, upregulated genes as red, and downregulated genes as blue. B Gene ontology (GO) analysis between treatment group comparisons indicating gene families that are differentially regulated based on p-values less than 0.05 and indicated with text size and fonts. C Clustered and normalized heatmap by row of gene count data show the top 50 significantly different genes between stressed cells, E. rectale CFSM versus 15 mM GSH

We identified differential host gene isoforms analysis to provide a higher resolution of the host response. Volcano plots highlighted the pattern of fewer gene regulation differences between control cells and E. rectale CFSM, with H₂O₂-exposed cells consisted of nearly entirely significantly upregulated genes compared with E. rectale CFSM, with pro-inflammatory and host repair genes like FOS [128, 129], MUC15 [130], EGR1 [131], SLC38A3 [132, 133] (Fig. 6B). GSH-treated cells displayed a different pattern with upregulation of some pro-inflammatory and repair genes, including C1S [134], PLAU [135], PPP1R15A [136], and downregulation of genes, including DAB2 [137], CKAP [138], and HHMR [139], which are involved in downregulation of the immune response, environmental stress sensing, and cell repair or migration, respectively. When comparing GSH-treated cells with E. rectale CFSM-treated cells, we observed several stress response genes upregulated in GSH-treated cells similarly to E. rectale CFSM-treated cells compared to inflamed control cells (Fig. 6B). This result clearly indicates that E. rectale CFSM alleviates host damage from H₂O₂, promotes host homeostasis, and downregulates the immune response to cell damage.

To compare host-expressed functions between treatment groups, gene counts between all treatment groups revealed upregulation of host damage and stress genes in damaged cells, including NECTIN [140], NGFR [141], MUC15, and EGR1. This immune response activation increases mucin production and activation of goblet cells. These genes were all expressed less in damaged cells treated with E. rectale CFSM, except for NECTIN1, but not downregulated in the H₂O₂ cells treated with 15 mM GSH (Fig. 6C). However, both E. rectale CFSM- and GSH-treated cells displayed a differential pattern with upregulation of several stress and repair genes that were not upregulated in H₂O₂-exposed cells. Most genes upregulated in the E. rectale CFSM treatment are NECTIN1, NGFR, and ERC1, whereas GSH-treated cells were lower expressed in those genes, indicating possible differences in immune activation between these treatments. GSH-treated damaged cells were highly upregulated in genes involving environmental sensing and solute transfer, like ATF3, SPRR2G, PGF, and heat shock protein genes, while displaying downregulation of genes NECTIN1, ERC1, CDK6, and NGFR (Fig. 6C).

Our work clearly shows that E. rectale microbial products were effective in reducing cell damage and repair gene classes. Our results not only quantify the effectiveness of GSH as a well-documented antioxidant and promoter of cell homeostasis [142], but further demonstrate that E. rectale produces multiple bacterial metabolites that function synergistically as antioxidants to alleviate inflammation and, driving a balanced state, promoting gut repair. Although this pattern was unique from E. rectale CFSM, our work also showcases the genetic variations among individual microbial species, and its impact on their derived products, highlighting the implications of investigation of microbial populations and metabolites on a community level. We identify a product of E. rectale, reduced glutathione, as an antioxidant, as well as potential reigns on the immune response during ROS-induced stress.

Eubacterium rectale microbially derived factors reprogrammed colon organoid redox and inflammatory gene expression

In our colon organoid experiments, we observed as expected that treatments with H₂O₂-induced oxidative injury to the colon organoids, as evidenced by morphological deterioration, increased ROS accumulation, and compromised epithelial integrity. After 20 h of oxidative challenge, organoids that were treated with H₂O₂ showed significantly reduced surface area (Fig. 7A, area change = − 0.09 ± 0.03 cm², p = 0.023), and serrated epithelial edges (Fig. 7B), which were consistent with epithelial damage and loss of tissue organization. Furthermore, these structural changes were accompanied by a marked significant reduction in TEER values (Fig. 7C, TEER change = − 0.858 ± 0.213 kΩ, p = 0.046), indicating a disruption of epithelial barrier function. In contrast, we observed that E. rectale CFSM treatment promoted clear recovery of the organoid physiology following oxidative stress, with the organoids showing improved structural integrity (Fig. 7A), maintaining smoother epithelial surfaces (Fig. 7B) and epithelial barrier integrity (Fig. 7C), relative to the H₂O₂ treatment group, suggesting reduced cellular injury. Most notably, we observed that ROS fluorescence intensity was significantly lower in E. rectale CFSM-treated organoids (Fig. 7D, relative intensity 1.16 ± 0.33) as compared to the H₂O₂ group (relative intensity 3.38 ± 0.46). Furthermore, there was no statistical difference in the ROS fluorescence in the E. rectale CFSM as compared to the control group (p = 0.68), suggesting that the CFSM augmented the host intrinsic antioxidant defenses rather than merely buffering extracellular oxidants. Collectively, we showed that E. rectale CFSM acted as a protective factor that facilitated organoid recovery from oxidative stress, stabilized epithelial physiology, and maintained functional barrier properties.

Next, we performed RNASeq of the colon organoids and generated approximately 52–85 million cleaned reads per sample, with Q30 values > 97% (Supplementary Table S8). We used hierarchical clustering of differentially expressed genes and showed clear transcriptional separation between oxidative stress and recovery conditions (Fig. 7E, Supplementary Table S8). Genes in the H₂O₂ vs Ctrl comparison were largely downregulated, whereas CFSM vs H₂O₂ and GSH vs H₂O₂ showed predominantly positive changes, suggesting that oxidative stress suppressed organoids’ epithelial transcriptional programs that were restored following E. rectale CFSM and GSH treatment. Furthermore, the similarity between CFSM and GSH transcriptional profiles (Fig. 7E, Supplementary Table S8) suggested that E. rectale CFSM induced a recovery associated gene expression loop that was consistent with GSH-mediated redox regulation. We further observed that the differential expression overlap analysis (Fig. 7F, Supplementary Table S8) showed distinct transcriptional responses among the treatments, with 1850 genes uniquely altered in CFSM vs H₂O₂, as compared to 713 genes unique to GSH vs H₂O₂, and 2734 genes unique to GSH vs CFSM, with smaller shared subsets in other comparisons (60, 37, and 32 genes) and no genes common to all comparisons. Despite observed transcriptional shifts, our expression analysis showed that there was a stable core of 15,762 genes expressed across all treatments (Fig. 7G, Supplementary Table S8), with smaller treatment specific subsets (118 CFSM specific genes, 314 control specific genes, and 64 H₂O₂ specific genes). Overall, our data suggested that E. rectale CFSM may induce a distinct transcriptional recovery response rather than simply reversing oxidative stress.

Our KEGG pathway categorization analysis (Fig. 7H, Supplementary Table S8) showed that E. rectale CFSM promoted a coordinated restoration of redox balance and cellular homeostasis in the human colon organoids. We observed a substantial proportion of differentially regulated genes fell within metabolism, particularly in amino acid metabolism, metabolism of cofactors and vitamins, energy metabolism, and xenobiotics biodegradation and metabolism; all of which are pathways that are tightly linked to GSH turnover, NADPH production, and ROS detoxification. We also observed an enrichment in the category “environmental information processing” such as signal transduction and membrane transport, which further supported our previous results on adaptive redox-responsive signaling that enabled the host’s cells to sense and buffer oxidative stress. Our results on the enrichment in the categories cell growth and death, transport and catabolism, and development and regeneration further suggested that the E. rectale CFSM did not merely dampen oxidative injury but might also actively support epithelial recovery and barrier stabilization.

Based on the topmost significantly enriched pathways, we observed that E. rectale CFSM induced broad, systems level transcriptional remodeling that supported epithelial integrity, adaptive signaling, and redox stabilization in the colon organoids (Fig. 7I, Supplementary Table S8). The most highly enriched pathways included Rap1 signaling (Rich Factor = 0.53, q-value = 3.01e − 09), regulation of actin cytoskeleton (Rich Factor = 0.49, q-value = 4.89e − 07), calcium signaling (Rich Factor = 0.47, q-value = 6.09e − 08), focal adhesion (Rich Factor = 0.46, q-value = 3.68e − 06), ECM-receptor interaction (Rich Factor = 0.49, q-value = 0.004), and tight junction (Rich Factor = 0.42, q-value = 0.007), which collectively indicated strengthened cell–cell adhesion, cytoskeletal organization, and barrier architecture. We surmised that the enrichment of P13K-Akt, MAPK, Ras, Hippo, cAMP, and phospholipase D signaling pathways highlighted coordinated stress-adaptive signaling networks, which are known to integrate redox cues and limit apoptosis under oxidative stress. Furthermore, supporting the hypothesis of coordinated stress adaption in the organoids, we also observed significant representation of lipid-related pathways (ether lipid metabolism and glycerophospholipid metabolism), which are critical for membrane stability and protection against lipid peroxidation.

Next, we analyzed the KEGG-enriched pathways for evidence that E. rectale CFSM directly promotes GSH-associated repair mechanisms, glutathione recycling, and metabolic redox homeostasis (Fig. 7J, Supplementary Table S8). Glutathione metabolism (Rich Factor = 0.42, q-value = 0.02) emerged as a central hub within this network. GSH is the primary intracellular antioxidant responsible for detoxifying hydrogen peroxide and lipid peroxides, and thus, we put forward that the enrichment of this pathway in the colon organoids suggested enhanced transcriptional support for GSH utilization, recycling, and redox buffering in response to E. rectale-derived GSH availability. More critically, we also observed a concurrent enrichment of the pentose phosphate pathway (PPP) (Rich Factor = 0.32, q-value = 0.05), indicating that organoids upregulated NADPH-generating capacity to sustain GSH recycling. Therefore, the coupling of PPP activation with GSH metabolism strongly suggested that E. rectale-derived GSH supplementation triggered a host metabolic adaptation and maintained an expanded and functional GSH redox pool. Interestingly, we observed that rather than a transient antioxidant effect, E. rectale-derived GSH appeared to induce a reinforced antioxidant regeneration loop within the epithelium (Fig. 7J, Supplementary Table S8).

We also observed that the microbial GSH-driven transcriptional remodeling in the organoids extended to upstream substrate and enzymatic systems (Fig. 7J, Supplementary Table S8). We surmised that the enrichment of cysteine and methionine metabolism (Rich Factor = 0.32, q-value = 0.03) in the organoids indicated increased sulfur amino acid flux in order to sustain endogenous GSH biosynthesis, suggesting that E. rectale-derived GSH stimulated the host’s pathways maintaining redox capacity. Finally, the enrichment of the peroxisome pathway (Rich Factor = 0.47, q-value = 0.01) further supported enhanced compartmentalized host’s ROS clearance through catalase and lipid-oxidation control.

Collectively, these enriched pathways described a coherent redox homeostasis loop, integrating GSH metabolism, NADPH regeneration, enzymatic and auxiliary detoxification, and antioxidant buffering. This coordinated transcriptional activation aligned with sustained suppression of intracellular ROS, restoration of TEER, and preservation of organoids architecture, further demonstrating that E. rectale CFSM established durable redox homeostasis and barrier resilience.

Discussion

A deeper understanding of the gut microbiota and its impact on host inflammation and promotion of homeostasis is critical in developing treatments for colitis and associated inflammatory conditions. The Lachnospiraceae family has had a tumultuous history with nomenclature reclassifications; there has also been a fundamental lack of research seeking to understand pathways of individual microbial metabolites with the host and identification of microbial-derived products that protect from or alleviate excessive inflammation and oxidative stress. In this study, longitudinal metagenomic sequencing of vertically transmitted dysbiotic mice revealed a potential role of the presence of Lachnospiraceae with recovery of induced dysbiosis and its resulting inflammation.

Treatment of colitis is challenging; dysbiosis is difficult to diagnose and often requires patient-specific diagnostic and therapeutic regimens. With the safety concerns of fecal microbiota transplantation (FMT) [143, 144], an alternative approach could be using a community of commensal microbes to re-establish healthy host-microbe interactions and immune modulation. To assemble a community of microbes to facilitate gut health, a thorough understanding of individual microbes’ contributions or potential harm is necessary. Many Lachnospiraceae family members possess characteristics with potential as next-generation probiotics, including SCFA production, modulation of immune cell activation, and production of amino acids important for mucosal healing.

We utilized genome resolved shotgun metagenomics to identify Lachnospiraceae members potentially driving recovery in the murine model. Although our MAGs exceeded 70% completeness, ANI comparisons between the MAGs and commercially available ATCC isolates yielded 80–90% similarity, reflecting genus level rather than species or strain level genomic consensus. This falls below the 95% ANI threshold conventionally used for species level assignment, and therefore, strain level functional deviations between the ATCC isolates and the in situ MAGs cannot be excluded. Despite this limitation, we pursued a deliberate strategy of breadth and chose five ATCC strains that showed highest genomic similarity to the MAGs, spanning three genera, to capture functional diversity. Critically, we validated our strain selection at the functional level, with comparative genomic analysis confirming that the MAGs harbor the GSH biosynthesis pathway genes, consistent with those characterized in E. rectale isolate, and metabolomic analysis directly further confirmed GSH production in E. rectale CFSM. The concordance of pathway level annotations and measured metabolite output between the MAGs and ATCC E. rectale isolate provides molecular level support for our strain selection, independent of ANI thresholds. Future investigations employing strain-resolved metagenomics and targeted isolation Lachnospiraceae are essential to achieve true strain level experimental accuracy. In this study, we identified E. rectale as a promising probiotic candidate; however, clinical applications of probiotics have resulted in side effects, like increased gastrointestinal symptoms, with the author’s highlighting that probiotics in a clinical setting are under quantified and not well understood [145]. Although there is a potential clinical risk, benefits of Lachnospiraceae have been described in a recent in vivo study that showed fecal microbiota transplantation enriches for Lachnospiraceae and results in mitigating liver injury through increased levels of N-acetylcysteine, a precursor to GSH [146]. Some members of the Lachnospiraceae family, particularly certain species within genera such as Ruminococcus, have been enriched in disease states, underscoring the importance of species- and strain-level characterization within this diverse family [10]. For the identification of microbes or microbial-derived products for the treatment of chronic inflammatory conditions, a fundamental understanding of individual microbes and their products is essential.

This study focuses on the impact of increased Lachnospiraceae in an IBD murine model with induced colitis with the outcomes of similar host status to control mice. We then transition to an in vitro model of Caco-2 cells and colon organoids with induced colitis and stress to understand the effects of individual microbes’ spent media on inflammation and ROS damage. An essential future direction would be to provide more insights into how these microbial metabolites impact an in vitro model that has a more primary metabolism, like organoids to determine more clinical relevance. After the identification of E. rectale as a promising probiotic candidate, and its ability to produce reduced glutathione (GSH), a critical next step of this research would be investigating the potential of E. rectale as an alleviator of colitis in vivo. More well-known microbial GSH biosynthesis pathways include two genes, gshA and gshB, which are glutathione synthases [147, 148]. They have been well studied in the context of driving microbial pathogenesis through activation of microbial stress response [149, 150]. Through these genes, microbes including Pseudomonas aeruginosa can make GSH to maintain oxidative balance, in a protective manner; however, increased glutathione biosynthesis is attributed to the virulence of P. aeruginosa in vivo as well as impacting biofilm formation, and activating the type III secretion system [150–152]. Thus, the production of microbial GSH is important for host homeostasis, pathogenesis, and commensal protection.

While in our study, we did not identify canonical gshA/gshB homologs [147, 148] in E. rectale, and instead, E. rectale appeared to utilize an alternative pathway involving bcp [101, 102] and ArgC [153], making this bacterium’s GSH mechanism novel. While this study focused on GSH as a primary antioxidant product of E. rectale, we recognize that the full protective effect observed in our cellular and organoid models is likely the result of synergistic interactions among multiple microbially derived metabolites. Notably, E. rectale CFSM is uniquely enriched in L-histidine, L-taurine, L-cysteine, and L-carnosine, which are metabolites with well-documented anti-inflammatory and gut-reparative properties [124, 154–156]. The convergence of these metabolites in E. rectale CFSM suggested that GSH-mediated ROS alleviation operated within a broader network of anti-inflammatory and cytoprotective signals, rather than as an isolated mechanism. Disentangling the individual and cumulative contributions of these metabolites represents an important and tractable future direction. Additionally, the mammalian host response to microbial metabolites in a complex in vivo gut environment may differ from what we observed in our in vitro and organoid models, and confirmation of these findings in DSS-treated or IL-10 KO mice colonized with E. rectale will be essential for establishing clinical relevance.

Eubacterium rectale is among the more abundant Lachnospiraceae members in the healthy human gut, typically comprising 1–5% of the fecal microbial community [157, 158], and is consistently depleted in IBD patients as demonstrated in our PRISM cohort analysis. At these physiological abundances, E. rectale could plausibly generate GSH and other anti-inflammatory metabolites at concentrations sufficient to influence the local redox environment of the colonic epithelium [156]. However, its functional output in vivo will inevitably be shaped by interactions with surrounding microbial community members and the inflammatory state of the host, suggesting that E. rectale may be most effective as part of a defined health-associated microbial consortium rather than as a standalone probiotic [159, 160]. Together, these future investigations will build toward the development of E. rectale-derived metabolite as therapeutic strategies for reducing colon damage and oxidative stress in inflammatory bowel conditions.

Conclusions

Our results demonstrate that E. rectale microbial products have the potential to alleviate ROS stress equivalent to that of 15 mM of GSH. GSH is a known ROS alleviator in rats [161] and is hypothesized as a suppressor of harmful inflammation in humans; however, dosage of up to 1000 mg (equivalent to 3.25 mM) a day has been shown to be ineffective in lowering oxidative stress markers [162]. Highlighting the importance of microbial-derived metabolites, we showed that E. rectale was also able to lower ROS levels during DSS-induced colitis, while GSH alone was not sufficient in lowering ROS in inflamed cells.

Additionally, we investigated the impact of solely E. rectale-derived GSH by comparing microbial GSH containing CFSM with GSH depleted CFSM. These results directly indicated that E. rectale-derived GSH was responsible for a significant alleviation of ROS in colon cells. While the GSH depletion experiments highlight the central role of E. rectale-derived GSH in ROS alleviation, we acknowledge that other bioactive metabolites present in the CFSM, including L-histidine, L-taurine, L-cysteine, and L-carnosine, and are likely to contribute synergistically to the overall protective effect observed, and their individual and combined contributions warrant further investigation. We acknowledge that our current findings are based on in vitro Caco-2 and human colon organoid models, and that confirmation in a relevant murine model will be necessary to establish physiological and translational relevance. Our study addresses both at the level of potential gene function during glutathione biosynthesis, and at the metabolomic profile of E. rectale, revealing microbial products and pathways from E. rectale that may contribute to lowering inflammation and host damage. While our findings are supported by converging lines of genomic evidence, including ANI-matched MAG annotations, significantly lower bcp and ArgC gene coverage in IBD patients relative to healthy controls, and direct metabolomic quantification of GSH in E. rectale CFSM, we acknowledge that dedicated proteomics or enzyme activity assays directly confirming the functional activity of Bcp and ArgC in E. rectale were not performed in this study. Transcriptomic and proteomic validation of these enzymes, as well as their expression dynamics under oxidative stress conditions, represent important and high-priority next steps in establishing the complete mechanistic picture of microbial GSH biosynthesis in this commensal species.

Furthermore, while a genetic knockout of GSH biosynthesis genes in E. rectale would ideally provide the strongest causal evidence for the role of microbially derived GSH, the essential nature of GSH for bacterial survival under oxidative stress, as demonstrated in other bacterial species [149, 150], means that full gene deletion is likely to compromise bacterial viability rather than yield a clean metabolite-null strain. The development of conditional or inducible knockdown strategies for bcp or ArgC in E. rectale, contingent on advances in genetic tools for obligate anaerobes, therefore, represents an important long-term research priority. In addition, RNA sequencing of peroxide induced cells treated with E. rectale CFSM lowered environmental stress sensing genes and mucin production, indicating alleviation of ROS stress. While previous studies have explored the potential benefits of GSH [35, 115, 163], investigators focused on host-derived GSH or dietary supplementation, and have not adequately addressed microbial metabolites in lowering host-experienced ROS stress. This study addressed the critical knowledge gap that identified glutathione as a microbial product of E. rectale, and its potential as a probiotic candidate for therapeutic intervention during gut inflammation.

Authors’ contributions

T.G.R. and S.T.M.L. designed this study. Sample collection was performed by T.G.R., C.I., L.H., A.K., S.P., K.M., T.S., H.W., S.S. Authors T.S., T.G.R., L.H., and H.W. completed DNA extraction including Nanodrop and Qubit quality analysis. Q.R. performed all RNAseq bioinformatic analysis. T.G.R. and S.T.M.L. performed anvi’o and BV-BRC bioinformatic analyses. Bacterial and cell culturing and isolation was done by T.G.R., H.W., B.V., and C.I., A.S. and E.C. provided support for metabolite profiles. Authors T.G.R., H.W., B.V., C.I., and S.T.M.L. attributed biological relevance, wrote the manuscript, prepared figures, and supplementary files. T.G.R., S.T.M.L., and B.L.P. performed major manuscript and figure refinement while remaining authors contributed to lighter refinement. B.L.P. performed all histological analyses. S.T.M.L. acquired fundings for this study. All authors read, contributed to manuscript revision, and approved the submitted version of this manuscript.

Funding

This project was supported by an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health under grant number P20 GM103418. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health. Gratitude is also extended to the University of Kansas Medical Center Genome Sequencing Facility for their expertise and assistance in sequencing including Clark Bloomer, Dr. Veronica Cloud, Rosanne Skinner, and Yafen Niu. We greatly appreciate assistance from the following sources: Kansas IDeA Network of Biomedical Research Excellence (K-INBRE), Kansas State University Johnson Cancer Research Center, Kansas State University Biology Graduate Student Association, Kansas Intellectual and Developmental Disabilities Research Center (NIH U54 HD 090216), the Molecular Regulation of Cell Development and Differentiation – COBRE (P30 GM122731-03)—the NIH S10 High-End Instrumentation Grant (NIH S10OD021743), and the Frontiers CTSA grant (UL1TR002366) at the University of Kansas Medical Center, Kansas City, KS 66160. Dr. Tanner Richie was supported by an NIH T32 5T32AI007496-30 at the University of Virginia School of Medicine, Charlottesville, VA. We also thank Dr. Sherry Fleming for her continuous support and advice on this study.

Data availability

Accession Number and figshare. Raw sequencing data for the mice metagenomes were uploaded to SRA under NCBI BioProject PRJNA1074399. All other analyzed data in the form of databases and fasta files are accessible at figshare https://doi.org/10.6084/m9.figshare.25702764.v1.

Declarations

Ethics approval and consent to participate

All mouse experiments were reviewed and approved by the Institutional Animal Care and Use Committee at Kansas State University (APN #4391).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.