Multi-omics unveils strain-specific neuroactive metabolite production linked to inflammation modulation by Bacteroides and their extracellular vesicles

Abstract

Bacteroides species are key members of the human gut microbiome and play crucial roles in gut ecology, metabolism, and host-microbe interactions. This study investigated the strain-specific production of neuroactive metabolites by 18 Bacteroidetes (12 Bacteroides, 4 Phocaeicola, and 2 Parabacteroides) using multi-omics approaches. Genomic analysis revealed a significant potential for producing GABA, tryptophan, tyrosine, and histidine metabolism-linked neuroactive compounds. Using untargeted and targeted metabolomics, we identified key neurotransmitter-related or precursor metabolites, including GABA, l-tryptophan, 5-HTP, normelatonin, kynurenic acid, l-tyrosine, and norepinephrine, in a strain- and media-specific manner, with GABA (1-2 mM) being the most abundant. Additionally, extracellular vesicles (EVs) produced by Bacteroides harbor multiple neuroactive metabolites, mainly GABA, and related key enzymes. We used CRISPR/Cas12a-based gene engineering to create a knockout mutant lacking the glutamate decarboxylase gene (gadB) to demonstrate the specific contribution of Bacteroides finegoldii-derived GABA in modulating intestinal homeostasis. Cell-free supernatants from wild-type (WT, GABA+) and ΔgadB (GABA-) provided GABA-independent reinforcement of epithelial membrane integrity in LPS-treated Caco-2/HT29-MTX co-cultures. EVs from WT and ΔgadB attenuated inflammatory immune response of LPS-treated RAW264.7 macrophages, with reduced pro-inflammatory cytokines (IL-1β and IL-6), downregulation of TNF-α, and upregulation of IL-10 and TGF-β. GABA production by B. finegoldii had a limited impact on gut barrier integrity but a significant role in modulating inflammation. This study is the first to demonstrate the presence of a myriad of neuroactive metabolites produced by Bacteroides species in a strain- and media-specific manner in supernatant and EVs, with GABA being the most dominant metabolite and influencing immune responses.

Keywords: Bacteroides; Extracellular vesicles; GABA; Gut microbiome; Immunomodulation; Multi-omics; Neuroactive metabolites.

Graphical abstract

Fig. 1

Neurotransmitter pathways distribution across the 18 neuroactive gut bacterial strains based on WGS data. A phylogenomic tree was constructed with GToTree software and visualized with iTOL. Blue circles indicate the bootstrap percentage of 1000 replicates. Pathway numbers were identified relying on the de novo assembled genomes. The last panel is the neurotransmitters and their precursors, as presence/absence, were extracted from the pathways data generated using pathway tools software.

Fig. 2

Detection of neurotransmitter-related neuroactive metabolites. (A) The untargeted metabolomics approach shows the abundance of numerous potential neuroactive metabolites released by Bacteroidetes strains. (B) Illustrates the differences between the 18 strains in producing different neuroactive compounds and their precursors as quantified by targeted metabolomics analysis. (C) Bacteroidetes strains showed separate clusters based on genus indicating that producing neuroactive metabolites is a strain-specific trait as shown by PCA plot depending on the targeted metabolomics data. R² is the coefficient of determination and p refers to the p-value.

Fig. 3

Comparative metabolomic and proteomic analyses of CFS and EVs between selected Bacteroidetes strains grown on FAB medium. (A) A Venn diagram of untargeted metabolites detected in CFS and EVs of 4 strains. (B) An interactive cluster heatmap comparison between CFS and EVs shows the COG categories of 4 strains. (C) Bar plots illustrate GABA production comparison between the 4 strains in CFS and EVs, B. finegoldii UO.H1052 and P. massiliensis UO.H1001 grown on (FAB) exhibited the highest and lowest GABA levels respectively. (D) A heatmap shows the abundance of some neuroactive metabolites produced by strains UO.H1052 and UO.H1001. (E) A heatmap explained the production of neuroactive metabolites based on untargeted metabolomics analysis while (F) is a heatmap that identified the protein profile for UO.H1052 and UO.H1001 EVs.

Fig. 4

Comparative quantification of SCFAs production by Bacteroidetes strains in MacFarlane (MFM) and FAB media. Paired plots revealed the significant impact of fermentation media on the production of different SCFAs between MFM and FAB media. (B) Pie plots show the variation in concentrations of SCFAs released by each strain relying on the fermentation medium. (C&D) PCA plots showing the clustering of Bacteroidetes strains in groups based on the growing medium (C) or the taxonomy level (D). ***p< 0.001, ****p< 0.0001. R² is the coefficient of determination and p refers to the p-value. The number of independent and technical replicates was n = 3.

Fig. 5

Obstruction of GABA biosynthesis by knocking out the glutamate decarboxylase gene (ΔgadB) using the CRISPR/FnCas12a system. (A) The PB025 plasmid was constructed by introducing sgRNA and upstream and downstream flanking regions of the gadB gene. (B) The constructed plasmid was transformed into E. coli S17–1, and then conjugated into B. finegoldii. (C) The expression of the Cas protein was induced with anhydrotetracycline (aTC) to activate the deletion of the gadB gene and confirmed by PCR and sequencing. (D-1) shows the differences in the growth curve between B. finegoldii WT and mutant strain (ΔgadB) in a minimal medium with xylose. (D-2) A bar plot illustrates the variations in GABA and glutamic acid production levels in WT and ΔgadB strains at stationary phase CFS relying on targeted metabolomics analysis.

Fig. 6

B. finegoldii CFS attenuates LPS-induced gut membrane disruption in Caco-2/HT29 co-cultures, while GABA-enriched EVs modulate inflammatory responses in RAW 264.7 macrophages. (A) Exposure of Caco-2/HT29 co-cultures to LPS resulted in a significant decrease in TEER after 24 h. The subsequent treatment with bacterial supernatants of WT and ΔgadB B. finegoldii strains showed a significant increase in transepithelial resistance. (B) Lucifer Yellow assay showed reduced florescence in wells treated with WT and ΔgadB, indicating attenuated leaky gut conditions compared to control groups. (C) LPS-induced oxidative stress in RAW264.7 macrophages was significantly reduced by WT and ΔgadB EVs and GABA based on GSH/GSSG ratio measurement. Contrarily, the CFS increased oxidative stress, with ΔgadB showing higher oxidative stress than WT and GABA. (D) ΔgadB EVs exhibited the lowest cell viability compared to WT EVs and CFS from WT, ΔgadB, and GABA, as measured by ATP levels. (E) Gene expression by qRT-PCR analysis with LPS-treated RAW264.7 macrophages showed that WT EVs and GABA downregulated TNF-α and upregulated IL-10 and TGF-β, with IL-10 significantly less upregulated in the ΔgadB mutant. EVs with GABA and chemically synthesized GABA similarly affected IL-1β and IL-6 expression. *p< 0.05, **p< 0.01, ***p< 0.001, ****p< 0.0001. All these experiments were performed in three (n = 3) independent and three technical replicates.