The gut microbiome regulates many aspects of human biology, including immunity and inflammatory diseases, yet mechanistic discovery and translation remain constrained by fragmented workflows and manual hypothesis integration. Here, we present Eubiota, an open-source modular agentic framework that decomposes microbiome inquiry using specialized agents for planning, execution, verification, and grounded synthesis. Coordinated through shared memory and domain-specific tools, Eubiota employs reinforcement learning to optimize multi-turn reasoning, achieving 87.7% benchmark accuracy and outperforming GPT-5.1 by 10.4%. Across four case studies, Eubiota enabled end-to-end discovery with experimental validation. Screening nearly 2,000 bacterial genes, Eubiota identified the uvr-ruv DNA repair axis as a fitness determinant under inflammatory stress, validated using transposon mutants and IBD metagenomes. It further designed a four-strain consortium that attenuated colitis severity in mice and generated a commensal-sparing antibiotic cocktail, demonstrating its utility in addressing community-level design challenges at cellular and molecular levels. Finally, Eubiota discovered diet-associated metabolites that suppress NF-kB signaling. Together, these results establish Eubiota as a scalable, tool-grounded scientific copilot for mechanistically driven microbiome discovery.

The oxygen reduction reaction (ORR) in living cells efficiently generates energy but is limited for sustainable applications due to inefficient electron transfer across membranes. Here, we report an interspecies cooperative mechanism, termed electro-mutualism, that enables efficient extracellular ORR using an electrotrophic Acinetobacter venetianus RAG-1 (RAG-1) and a non-CO2-fixing electrotrophic Shewanella oneidensis MR-1 (MR-1), sustained by inorganic carbon supplied via a bicarbonate–CO2 equilibrium. Under cathodic polarization, RAG-1 assimilates inorganic carbon and secretes lactate using electrode-derived electrons, which fuels MR-1. In turn, MR-1 releases flavins that interact with the RnfB complex of RAG-1, thereby accelerating the transmembrane electron transfer. This electro-mutualistic cooperation achieves, to our knowledge, the highest reported whole-cell ORR current density (20.9 A/m2) under O2-aerated, neutral-pH conditions, outperforming benchmark Pt/C and laccase cathodes evaluated under identical conditions. The universality of electro-mutualism is further supported by replacing MR-1 with Bacillus subtilis and is predicted to extend across proteobacteria, firmicutes, and actinobacteria. Electro-mutualism thus provides a metal-free, genetically unmodified catalytic strategy for green energy conversion.

Sequencing of ribosomal marker genes remains a cornerstone for profiling complex microbial communities. In recent years, there has been a shift from Illumina to long-read technologies, including PacBio and Oxford Nanopore Technologies (ONT). ONT is attractive due to its low startup cost and portability; however, historically high error rates have prevented direct amplicon sequencing variant (ASV) generation from raw nanopore reads. This has forced most workflows to rely on mapping raw reads against reference databases constraining analyses to taxa covered by these. With recent improvements in ONT sequencing accuracy, we sought to challenge this view by sequencing samples of increasing complexity using primer sets targeting amplicons of different lengths, and by sequencing the exact same PCR libraries on both PacBio and ONT. We demonstrate that error-free ASVs can now be generated directly from raw nanopore reads using standard denoising algorithms originally developed for Illumina data. Current ONT read quality enables reliable reconstruction of amplicons spanning ~250 bp to ~4,200 bp and allows resolution of intragenomic rRNA gene variants. These results extend beyond simple mock communities to complex fecal, anaerobic digester, activated sludge, and soil samples. When sequencing depth is sufficient, ONT accurately recovers all or nearly all intra-genomic 16S rRNA gene copy variants, showing perfect sequence identity to curated reference sequences in mock communities and to ASVs inferred from PacBio data in complex communities. Across the primer sets, ONT required higher sequencing depth than PacBio to fully resolve the communities, with this requirement increasing with amplicon length. For complex samples, ONT required approximately 2-3x more reads for V4 (~250 bp) and V1-V3 (~500 bp), 4.1-5.6x more reads for V1-V8 (~1400 bp), and 25-42x more reads for rRNA operon (OPR) amplicons (~4200 bp). Consequently, sequencing complex communities with OPR primers on ONT is currently not feasible due to the unrealistically high read depth required. This study provides evidence that ONT amplicon sequencing has matured to the point where true ASV-resolved profiling is practically and economically feasible, moving ONT amplicon analysis beyond reliance on OTU clustering or reference alignment to enable application in environments lacking comprehensive reference databases.

Exopolysaccharides (EPS) produced by lactic acid bacteria (LAB) and other microorganisms have attracted considerable interest due to their structural diversity and physicochemical properties, which makes them valuable across various industrial applications. To achieve high cell densities and maximize EPS yields, microorganisms are typically cultivated in nutrient-rich media containing yeast extract. However, yeast extract may contain high molecular weight polysaccharides that are not metabolized by the bacteria. This can lead to an overestimation of EPS yields and contamination of the bacterial EPS, potentially resulting in misinterpretation of their structure and biological activity. In this study, we demonstrate the presence of high molecular weight α-mannan and β-glucan in yeast extract in EPS isolates using both ultrafiltration and the commonly used trichloroacetic acid/ethanol (TCA/EtOH) precipitation method. These polysaccharides were characterized by size-exclusion chromatography, high-performance anion-exchange chromatography, and nuclear magnetic resonance spectroscopy. Their abundances were estimated to range from 10 to 50 mg/L in MRS medium, depending on the supplier of the yeast extract. The main contaminant identified was yeast α-mannan. By cultivating L. rhamnosus GG (ATCC 53103) and L. pentosus KW1 and isolating their respective EPS, we illustrate how these yeast extract contaminants affect the structural interpretation of the EPS and that the contaminants can be completely removed by ultrafiltration of the growth medium prior to bacterial cultivation. In conclusion, we emphasize the necessity of stringent controls in the production and purification of microbial EPS, with particular attention to the chemical purity of medium constituents.

The persistent global challenge of foodborne pathogens necessitates the development of ultrasensitive point-of-care (POC) detection technologies. However, conventional methods for detecting foodborne pathogens, such as culture-based and nucleic-acid-based techniques, are often limited by complex procedures and reliance on sophisticated instrumentation, respectively. Although immunoassays are more suitable for POC detection, they are frequently constrained by the high cost and limited stability of the antibodies. Therefore, there is an urgent need to develop innovative, ultrasensitive POC biosensing platforms for foodborne pathogens detection. This comprehensive review discusses the development of phage-based nano-sensors for advancing ultrasensitive POC detection of foodborne pathogens. Initially, the inherent properties of phages and phage engineering strategies for creating ideal recognition elements are briefly introduced. Subsequently, phage-based nanosensors for foodborne pathogen detection and their signal amplification mechanisms are comprehensively reviewed, including colorimetric, fluorescent, electrochemical, magnetic relaxation switch, chemiluminescent, and other modalities. Then, the integration of artificial intelligence into intelligent analysis is highlighted. Furthermore, current challenges and future prospects are outlined to bridge the gap between laboratory research and practical application. The integration of phages into nanosensors offers opportunities to overcome the limitations of conventional methods for monitoring foodborne pathogens and is expected to significantly enhance food safety in control systems.

Even though complex microbial communities are ubiquitous and provide essential services for natural and human-associated ecosystems, our knowledge about their assembly and dynamics is incomplete. There is an ongoing debate whether the behavior of complex communities can be predicted from the outcome of pairwise competition of species, and whether communities reach alternative stable states depending on the level and complexity of resource provided for growth. To estimate the effect of two resource gradients, total carbon availability and resource complexity, on the compositional dynamics of a complex microbial community, we conducted a 16-day serial passage experiment, transferring a 16-species synthetic community in 96 different resource environments. We observed that although both resource dimensions influenced community composition, total carbon exerted a considerably larger effect. Additionally, we saw the emergence of a tristable pattern along the total carbon gradient, a feature not observed for the resource complexity gradient. Using monoculture assays, we identified lag phase duration as the dominant predictor of competitive success at carbon extremes, with maximum growth rate increasing in importance as lag times converged. Total carbon availability thus structured community state transitions and regulated which growth trait governed competitive sorting. These results suggest the importance of total carbon level over resource complexity and identifying dominant species for the quest to successfully manage, maintain and manipulate complex microbial communities.

Collectively known as one-carbon metabolism (OCM), both the folate and methionine cycles are highly regulated to meet cellular demands. These cycles are key in the production and recycling of methyl groups to be used in many essential cellular processes such as the production of nucleotides, as well as S-adenosyl-l-methionine (SAM) the global methyl donor for DNA, RNA, and post translational modifications. Within the folate cycle, 5,10-methylenetetrahydrofolate is the main species through which methyl groups enter OCM. Therefore, 5,10-methylenetetrahydrofolate reductase (MTHFR), which reduces 5,10-methylenetetrahydrofolate into 5-methyltetrahydrofolate, is the central enzyme that directs methyl groups for use within the methionine cycle. MTHFR is an enzyme found in all domains of life, but unlike in prokaryotes, eukaryotic MTHFR activity is highly regulated by the level of SAM, to balance the one-carbon needs of the cell. In this perspective, we review the catalytic mechanism of MTHFR, evolutionary differences, and the regulatory mechanisms that have evolved to alter its activity. We also discuss recent structural findings that reveal a unique mechanism for inactivation by SAM as a feedback loop and its consequences for understanding inherited MTHFR deficiency.