Selectivity of OATs and OATPs for Endogenous Metabolites and Signaling Molecules In Vivo

Organic anion transporters (OATs, SLC22) in the kidney and organic anion-transporting polypeptides (OATPs, SLCO) in the liver play crucial roles in the disposition of small molecule drugs that are organic anions. According to the Remote Sensing and Signaling Theory, these multispecific "drug" transporters are also central to crosstalk between the liver, kidney, and other organs via endogenous small molecules (e.g., metabolites, signaling molecules, gut microbiome products). These multispecific drug transporters govern access of small molecules with high informational content across multiple scales (organism to organelle). Previous chemoinformatic and machine learning methods have proven useful for identifying molecular properties of organic anion drugs that predispose them to handling by the OAT (renal) and the OATP (hepatic) transporters. This is important for understanding pharmacokinetics (ADME) in the context of chronic kidney disease (CKD) and liver disease. Given that OATs and OATPs are involved in many metabolic diseases, we sought to determine whether molecular properties could be identified for distinguishing OAT- versus OATP-interacting endogenous metabolites in vivo. This is essential for understanding endogenous small molecule communication between the kidney proximal tubule and hepatocytes in a larger Remote Sensing and Signaling System. We analyzed in vivo metabolomics data from OAT and OATP knockout mice, focusing on endogenous metabolites selective for OATs (e.g., OAT1 or SLC22A6; OAT3 or SLC22A8) vs OATPs (including the locus containing Oatp1b2, the closest homologue of human OATP1B1 or SLCO1B1 and OATP1B3 or SLCO1B3). Applying chemoinformatic methods to a data set of 210 metabolites based on knockout mouse metabolomics (92 OAT-selective, 118 OATP-selective), we identified a set of distinguishing molecular properties (e.g., MolLogP, RingCount, NumRotatableBonds). We then used machine learning approaches (e.g., Random Forest, Naive Bayes, Logistic Regression) to classify OAT vs OATP metabolites, achieving over 75% accuracy. These results support the view that transporter knockout mouse metabolomics can help define selectivity of SLC drug transporters for endogenous metabolites, signaling molecules, antioxidants, nutrients, and gut microbiome products. In the context of the Remote Sensing and Signaling Theory, we discuss the implications for understanding organ crosstalk and interorganismal communication as well as drug disposition, drug-metabolite interactions, and metabolite-based drug design.