ABSTRACT
The human microbiota impacts a variety of diseases and responses to therapeutics. Due to a lack of robust in vitro models, detailed mechanistic explanations of host-microbiota interactions cannot often be recapitulated. We describe the design and development of a novel, versatile and modular in vitro system that enables indirect coculture of human epithelial cells with anaerobic bacteria for the characterization of host-microbe secreted metabolite interactions. This system was designed to compartmentalize anaerobes and human cells in separate chambers conducive to each organism's requisite cell growth conditions. Using perfusion, fluidic mixing, and automated sample collection, the cells continuously received fresh media, while in contact with their corresponding compartments conditioned supernatant. Supernatants from each chamber were collected in a cell-free time-resolved fashion. The system sustained low oxygen conditions in the anaerobic chamber, while also supporting the growth of a representative anaerobe (Bacteroides thetaiotaomicron) and a human colonic epithelial cell line (Caco-2) in the aerobic chamber. Caco-2 global gene expression changes in response to coculture with B. thetaiotaomicron was characterized using RNA sequencing. Extensive, targeted metabolomics analysis of over 150 central carbon metabolites was performed on the serially collected supernatants. We observed broad metabolite changes in host-microbe coculture, compared to respective mono-culture controls. These effects were dependent both on sampling time and the compartment probed (apical vs. basolateral). Coculturing resulted in the depletion of several important metabolites, including guanine, uridine 5'-monophosphate, asparagine, and thiamine. Additionally, while Caco-2 cells cultured alone predominantly affected the basolateral metabolite milieu, increased abundance of 2,3-dihydroxyisovalerate and thymine on the basolateral side, occurred when the cells were cocultured with B. thetaiotaomicron. Thus, our system can capture the dynamic, competitive and cooperative processes between host cells and gut microbes.
Subject(s)
Bioreactors , Coculture Techniques , Epithelial Cells , Humans , Bioreactors/microbiology , Epithelial Cells/microbiology , Epithelial Cells/metabolism , Caco-2 Cells , Gastrointestinal Microbiome/physiology , Bacteroides thetaiotaomicron/metabolismABSTRACT
Type 2 diabetes (T2D) is a global public health issue characterized by excess weight, abdominal obesity, dyslipidemia, hyperglycemia, and a progressive increase in insulin resistance. Human population studies of T2D development and its effects on systemic metabolism are confounded by many factors that cannot be controlled, complicating the interpretation of results and the identification of early biomarkers. Aged, sedentary, and overweight/obese non-human primates (NHPs) are one of the best animal models to mimic spontaneous T2D development in humans. We sought to identify and distinguish a set of plasma and/or fecal metabolite biomarkers, that have earlier disease onset predictability, and that could be evaluated for their predictability in subsequent T2D studies in human cohorts. In this study, a single plasma and fecal sample was collected from each animal in a colony of 57 healthy and dysmetabolic NHPs and analyzed for metabolomics and lipidomics. The samples were comprehensively analyzed using untargeted and targeted LC/MS/MS. The changes in each animal's disease phenotype were monitored using IVGTT, HbA1c, and other clinical metrics, and correlated with their metabolic profile. The plasma and fecal lipids, as well as bile acid profiles, from Healthy, Dysmetabolic (Dys), and Diabetic (Dia) animals were compared. Following univariate and multivariate analyses, including adjustments for weight, age, and sex, several plasma lipid species were identified to be significantly different between these animal groups. Medium and long-chain plasma phosphatidylcholines (PCs) ranked highest at distinguishing Healthy from Dys animals, whereas plasma triglycerides (TG) primarily distinguished Dia from Dys animals. Random Forest (RF) analysis of fecal bile acids showed a reduction in the secondary bile acid glycoconjugate, GCDCA, in diseased animals (AUC 0.76[0.64, 0.89]). Moreover, metagenomics results revealed several bacterial species, belonging to the genera Roseburia, Ruminococcus, Clostridium, and Streptococcus, to be both significantly enriched in non-healthy animals and associated with secondary bile acid levels. In summary, our results highlight the detection of several elevated circulating plasma PCs and microbial species associated with fecal secondary bile acids in NHP dysmetabolic states. The lipids and metabolites we have identified may help researchers to differentiate individual NHPs more precisely between dysmetabolic and overtly diabetic states. This could help assign animals to study groups that are more likely to respond to potential therapies where a difference in efficacy might be anticipated between early vs. advanced disease.