Predicting Early Diabetic Kidney Disease Using a 3D Mesangial Model and Bayesian Mechano-Chemical Stratification

Abstract

Diabetic kidney disease (DKD) is often diagnosed only after irreversible damage, limiting early treatment. The kidney mesangium, a structure highly sensitive to mechanical forces, plays a key role in early fibrosis development, yet its response to early disease cues like stiffness and biochemical changes is poorly understood. This is due to limitations in current in vitro models, poor in vivo accessibility, and static biopsy samples. To address this, we developed a 3D in vitro model of the mesangial microenvironment using stiffness-tunable gelatin methacrylate (GelMA) hydrogels that mimic healthy and fibrotic kidney tissue. Exposing mesangial cells to glucose and TGF-{beta}1 led to altered cell shape, increased dry mass, and elevated expression of fibrotic markers (-SMA and collagen IV), especially under stiffer conditions, indicating a synergistic effect of biochemical and mechanical stress. These responses were integrated using Gaussian process regression to create a 3D "severity cube" that maps DKD progression across mechanical and chemical inputs. This system quantifies early mesangial transitions and reveals key fibrosis-related mechanisms. By combining organotypic modeling with interpretable inference, our platform offers a predictive tool for early disease stratification and a basis for studying subclinical fibrosis progression in DKD.