The liver is one of the key organs for exogenous and endogenous metabolism and is often a target for drug- and chemical-driven toxicity. A wide range of experimental approaches has been established to model and characterize the mechanisms of drug- and chemical-induced hepatotoxicity. A number of microfluidics-enabled in vitro models of the liver have been developed, but the unclear translatability of these platforms has hindered their adoption by the pharmaceutical industry; to achieve wide use for drug and chemical safety evaluation, demonstration of reproducibility and robustness under various contexts of use is required. One of these commercially available platforms is the PhysioMimix LC12, a microfluidic device where cells are seeded into a 3D scaffold that is continuously perfused with recirculating cell culture media to mimic liver sinusoids. Previous studies demonstrated this model’s functionality and potential applicability to preclinical drug development. However, to gain confidence in PhysioMimix LC12’s robustness and reproducibility, supplementary characterization steps are needed, including the assessment of various human hepatocyte sources, contribution of non-parenchymal cells (NPCs), and comparison to other models. In this study, we performed replicate studies averaging 14 days with either primary human hepatocytes (PHHs) or induced pluripotent stem cell (iPSC)-derived hepatocytes, with and without NPCs. Albumin and urea secretion, lactate dehydrogenase, CYP3A4 activity, and metabolism were evaluated to assess basal function and metabolic capacity. Model performance was characterized by different cell combinations under intra- and inter-experimental replication and compared to multi-well plates and other liver platforms. PhysioMimix LC12 demonstrated the highest metabolic function with PHHs, with or without THP-1 or Kupffer cells, for up to 10–14 days. iPSC-derived hepatocytes and PHHs co-cultured with additional NPCs demonstrated sub-optimal performance. Power analyses based on replicate experiments and different contexts of use will inform future study designs due to the limited throughput and high cell demand. Overall, this study describes a workflow for independent testing of a complex microphysiological system for specific contexts of use, which may increase end-user adoption in drug development.

Background/Objectives: The kidneys are essential for eliminating drugs and chemicals from the human body and renal epithelial cells are particularly vulnerable to damage caused by xenobiotics and their metabolites. Drug-induced kidney toxicity is a major cause of drug attrition during preclinical and clinical development and the ability to predict renal toxicity remains a pressing challenge, necessitating more predictive in vitro models. However, the abundance of commercially available renal proximal tubule epithelial cell (RPTEC) sources complicates the selection of the most predictive cell types. Methods: This study compared a wide range of RPTEC sources, including primary cells (Lonza) and various RPTEC lines from different vendors, such as ciPTECs (Cell4Pharma), TERT1/RPTECs (ATCC), and HEK293 (GenoMembrane), including OAT1-overexpressing variants. HepG2 cells were included for a comparison of organ specificity. The different cells were cultured in 96- or 384-well plates and exposed to 12 drugs for 72 h at a concentration yielding a response (0.3–300 µM) to evaluate their ability to predict clinical outcomes. The CellTiterGlo^® assay was used to measure cell viability, and transcriptome data from unexposed cells was analyzed using the TempO-seq^® S1500+ platform. Results: Gene expression data showed that the primary kidney cells most closely matched the transcriptome of the human kidney medulla, followed by the TERT1 and ciPTEC lines, with the HEK lines showing the lowest similarity. The RPTEC sources showed clustering by cell type, with OAT1 overexpression driving changes in metabolic, detoxification, and immune pathways, especially in TERT1 cells. Cell viability data were used to determine points of departure (PODs) which were compared to human serum Cmax values to assess safety margins. The TERT1 and ciPTEC RPTEC lines demonstrated the highest predictive performance for nephrotoxicity, with OAT1 overexpression significantly enhancing sensitivity, accuracy, and overall predictive power (MCC scores: 0.764 and 0.667, respectively). In contrast, HepG2 cells showed the lowest performance across all metrics, highlighting the critical role of cell type and transporter expression in nephrotoxicity prediction. Conclusions: This study highlights important differences among RPTEC sources and their utility in drug safety studies of the renal proximal tubule. We show that while improved cell options for renal proximal tubule are needed, OAT1-overexpressing RPTECs are a superior model to the background cell type.