Integrative metagenomic and metabolomic profiling identifies faecal biomarkers of prolonged social stress in pigs

Abstract

Stressors significantly impact human and animal health, increasing the risk of physical and mental disorders, in part by affecting the gut-brain axis. Although a link between stress, alterations in gut microbial composition, and the serum metabolite profile has already been established in humans, multiomics studies integrating the faecal microbiome and untargeted metabolomics remain unavailable. The objectives of the present study were twofold: first, to identify microbial and metabolic signatures associated with prolonged stress, and second, to evaluate the potential of integrative multiomics approaches to predict key metabolites and discover non-invasive faecal biomarkers of stress in pigs (n = 60). Gut microbial profiles were obtained by shotgun metagenomic sequencing, while faecal metabolites were analysed by untargeted reverse-phase liquid chromatography quadrupole time of flight mass spectrometry, followed by partial least squares discriminant analysis. Metabolite prediction from microbial features was performed using the machine learning method based on neural ordinary differential equations. Eleven discriminant metabolites were identified. In the control group, neurotransmitters such as serotonin and metabolites such as 2-acetamidophenol and sinapine (which possess anti-inflammatory and antioxidant properties) were the most prominent. Conversely, the stressed group exhibited elevated levels of xanthosine, pyrimidine bases (thymine and uracil), n-octadecylamine, and N-α-acetyl-L-lysine. N-octadecylamine (r = 0.37) showed a positive, and serotonin (r = −0.32) a negative correlation with hair cortisol. The results revealed interspecific interactions that modulated microbial and metabolic shifts between the control and stressed pig groups. Feature selection further identified 64 microbial genes that improved classification accuracy between control and stressed pigs to 91.06% and enhanced the prediction of key metabolites, including serotonin and xanthosine. Overall, this integrative multiomics framework elucidates complex microbiome-metabolite interactions and identifies non-invasive biomarkers of prolonged stress-induced metabolic dysregulation, pro