Personalized whole-body modeling links gut microbiota to metabolic perturbations in Alzheimer's disease

The human gut microbiome has been linked to metabolic disturbances in Alzheimer's disease (AD). However, the mechanisms by which gut microbes might influence metabolic dysfunction in AD remain poorly understood. Previously, gut microbiome-personalized whole-body models of human metabolism have been applied to predict how altered gut microbiome compositions may influence metabolites in the blood of healthy aging individuals with increased risk of AD. However, these previous results have not been validated in AD. In this study, we aimed to test these prior predictions in a cohort of AD dementia patients and individuals with mild cognitive impairment (MCI) and a probable AD diagnosis. Therefore, we created gut microbiome-personalized whole-body metabolic models for 34 AD dementia patients, 51 MCI patients, and 298 healthy controls. These in silico models were profiled to predict the metabolic influences of gut microbiomes on blood metabolites with previously reported alterations in AD. We found increased capacities of the in silico host-microbiome co-metabolism to produce S-adenosyl-L-methionine, L-arginine, creatine, taurine, and formate in the blood of AD patients. The metabolic predictions were then linked to key microbial taxa using a novel method that combines modeling-informed prediction sensitivity to alternative microbial abundances with LASSO-based taxonomic stability selection and elastic net regressions. This method found that increased relative abundances of Bacteroides uniformis and Bacteroides thetaiotamicron in AD were major factors driving the predicted metabolic changes. Furthermore, the metabolic predictions were associated with allelic variations in the APOE risk gene in healthy individuals, confirming our previous findings. In conclusion, we identified blood metabolites with known links to AD that were differentially influenced by gut microbiota in AD, and identified possible microbial drivers of these predicted shifts in host-microbiome interactions. These findings may facilitate the development of microbiome-informed treatments of AD.