Towards a 3D self-assembled hiPSC-derived inner blood-retinal barrier model on a microfluidic chip to study diabetic retinopathy

Diabetes is a highly prevalent metabolic disorder that significantly increases the risk of developing microvascular complications. Among them, diabetic retinopathy (DR) is one of the leading causes of vision loss. DR results from dysfunction of the inner blood-retinal barrier (iBRB), which is crucial in maintaining retinal homeostasis by regulating flux between the circulation and the retina. The pathology of early stages of the disease is not fully understood, and current therapies primarily target symptoms of the disease at later stages. Drug development and further research of the pathomechanism is often restricted to animal or 2D cell culture models. However, animals do not fully recapitulate the human ocular system and its pathologies in DR, while 2D cell culture models lack complex cellular interactions in a physiological 3D environment. Therefore, there is a need to develop human-based models, which provide interactions of multiple cell types, and can be used in modelling early stages of the disease. This would allow advances in studying the molecular mechanisms of DR and provide a platform for drug discovery, while reducing the use of animal models. Here, I established a 3D self-assembled microvascular network (MVN)-on-a-chip model to study diabetic vasculopathies. The use of human induced pluripotent stem cells (hiPSCs) provides an unlimited, stable and donor independent cell source. I successfully differentiated all four cell types involved in the formation of the iBRB: endothelial cells (ECs), neural pericytes (PCs), Müller glia (MüGl) and astrocytes (ACs). The first part of my work was the optimization and characterization of the MVN-on-a-chip model with the primary building blocks of microvasculature, namely ECs and PCs. This included adjustments of medium and gel composition, integration of interstitial flow and a thorough investigation of the role of PCs in the MVN. Moreover, I established an alternative 3D MVN­drop model which offers a cheaper and faster method in screening objectives. Next, I used the MVN-on-a-chip model in a diabetic setup to assess changes in expression pattern under hyperglycaemic conditions. High glucose levels clearly caused an upregulation of inflammatory pathways in the MVN. Therefore, I used hyperglycaemia, inflammatory cytokines and a combination to mimic a diabetic environment in the MVN-on-a-chip model and assessed their effect on vascular integrity. Both, hyperglycaemic and inflammatory treated MVN-on-a-chips showed characteristics of diabetic vasculopathies: vascular regression, acellular capillaries and reduced PC coverage. However, ECs and PCs were differently affected by high glucose levels and inflammation. PCs were more sensitive to hyperglycaemia while inflammatory cytokines mostly affected ECs. Some of these observations were confirmed in 2D cell cultures, but not those influenced by high glucose levels. Gene expression analysis of sorted ECs and PCs from hyperglycaemic treated MVN-on-a-chip systems suggests, that high glucose level induced upregulation of IL-1β in PCs which leads to an activation of NF-κB signalling pathway in ECs. Altogether, this provides evidence of the necessity of 3D co-culture models to provide physiological relevant insights in cellular mechanisms and pathologies and the meaningful application of MVN-on-a-chip model in diabetic disease modelling. In a next step, I integrated MüGl and ACs in the MVN-on-a-chip model to compare their effect on vascular formation and integrity. Both cell types seemed to result in comparable changes of the vascular architecture, such as decreased vessel diameter. Moreover, MüGl and ACs were observed in close contact with ECs and PCs, suggesting cellular crosstalk. Finally, I used the MVN-on-a-chip model with MüGl to specify it to an iBRB-on-a-chip model and assessed the previously established diabetic culture conditions. Similarly, the iBRB-on-a-chip model cultured in diabetic conditions exhibited changes associated with vascular regression. In conclusion, I established an MVN-on-a-chip model, an MVN-drop and an iBRB-on-a-chip model which can be used in a variety of studies. These models provide a platform to investigate vascular dysfunction in high glucose levels and inflammation, providing deeper insight into molecular changes on specific cell types.