Background: Antibiotic treatment has a well-established detrimental effect on the gut bacterial composition, but effects on the fungal community are less clear. Bacteria in the lumen of the gastrointestinal tract may limit fungal colonization and invasion. Antibiotic drugs targeting bacteria are therefore seen as an important risk factor for fungal infections and induced allergies. However, antibiotic effects on gut bacterial-fungal interactions, including disruption and resilience of fungal community compositions, were not investigated in humans. We analysed stool samples collected from 14 healthy human participants over 3 months following a 6-day antibiotic administration. We integrated data from shotgun metagenomics, metatranscriptomics, metabolomics, and fungal ITS2 sequencing.Results: While the bacterial community recovered mostly over 3 months post treatment, the fungal community was shifted from mutualism at baseline to competition. Half of the bacterial-fungal interactions present before drug intervention had disappeared 3 months later. During treatment, fungal abundances were associated with the expression of bacterial genes with functions for cell growth and repair. By extending the metagenomic species approach, we revealed bacterial strains inhibiting the opportunistic fungal pathogenCandida albicans. We demonstrated in vitro howC.albicanspathogenicity and host cell damage might be controlled naturally in the human gut by bacterial metabolites such as propionate or 5-dodecenoate.Conclusions: We demonstrated that antibacterial drugs have long-term influence on the human gut mycobiome. While bacterial communities recovered mostly 30-days post antibacterial treatment, the fungal community was shifted from mutualism towards competition.

Candida species overgrowth in the human gut is considered a prerequisite for invasive candidiasis, but our understanding of gut bacteria promoting or restricting this overgrowth is still limited. By integrating cross-sectional mycobiome and shotgun metagenomics data from the stool of 75 male and female cancer patients at risk but without systemic candidiasis, bacterial communities in high Candida samples display higher metabolic flexibility yet lower contributional diversity than those in low Candida samples. We develop machine learning models that use only bacterial taxa or functional relative abundances to predict the levels of Candida genus and species in an external validation cohort with an AUC of 78.6–81.1%. We propose a mechanism for intestinal Candida overgrowth based on an increase in lactate-producing bacteria, which coincides with a decrease in bacteria that regulate short chain fatty acid and oxygen levels. Under these conditions, the ability of Candida to harness lactate as a nutrient source may enable Candida to outcompete other fungi in the gut.

High-throughput RNA sequencing (RNA-seq) is routinely applied to study diverse biological processes; however, when performed separately on interacting organisms, systemic noise intrinsic to RNA extraction, library preparation, and sequencing hampers the identification of cross-species interaction nodes. Here, we develop triple RNA-seq to simultaneously detect transcriptomes of monocyte-derived dendritic cells (moDCs) infected with the frequently co-occurring pulmonary pathogens Aspergillus fumigatus and human cytomegalovirus (CMV). Comparing expression patterns after co-infection with those after single infections, our data reveal synergistic effects and mutual interferences between host responses to the two pathogens. For example, CMV attenuates the fungus-mediated activation of pro-inflammatory cytokines through NF-κB (nuclear factor κB) and NFAT (nuclear factor of activated T cells) cascades, while A. fumigatus impairs viral clearance by counteracting viral nucleic acid-induced activation of type I interferon signaling. Together, the analytical power of triple RNA-seq proposes molecular hubs in the differential moDC response to fungal/viral single infection or co-infection that contribute to our understanding of the etiology and, potentially, clearance of post-transplant infections.

Food, especially plant-based diet, has complex chemical diversity. However, large-scale phytonutrient-metabolizing activities of gut bacteria are largely unknown. Here we integrated and systematically analysed multiple databases containing information on enzymatic reactions and food health benefits, and 3,068 global public human microbiomes. Transformation of 775 phytonutrients from edible plants was associated with enzymes encoded by diverse gut microbes. In vitro assays validated the biotransformation activity of gut species, for example, Eubacterium ramulus. The biotransformation of phytonutrients demonstrated high interpersonal and geographical variability. Machine learning models based on 2,486 public case–control microbiomes, using the abundances of enzymes associated with modification of phytonutrients present in health-associated foods, discriminated the health status of individuals in multiple disease contexts, suggesting altered biotransformation potential in disease. We validated the association of microbiome-encoded enzymes with the anti-inflammatory activity of common edible plants by combining metagenomics and metatranscriptomics analysis in specific-pathogen-free and germ-free mice. These findings have implications for designing precise, personalized diets to guide an individual towards a healthy state.