A renewable supply of human retinal endothelial cells could help researchers probe the inner blood–retina barrier, model diabetic retinopathy, and develop new regenerative treatments for ischemic retinal disease. In a new Nature Biomedical Engineering study, the Duke University-based biomedical engineers describe a method for deriving functional retinal endothelial cells from human induced pluripotent stem cells (hiPSCs) – and show that these cells can model disease, form perfusable microvascular networks, and repair ischemic retina in mice.
A major barrier to studying these processes has been the limited availability of human retinal endothelial cells that retain their tissue-specific identity in culture. The team addressed this by harnessing Wnt–β-catenin signaling, particularly the Norrin–Frizzled4 axis, which is known to be crucial for retinal vascular development. Starting with hiPSCs, the researchers induced mesoderm, introduced ETV2 messenger RNA to drive endothelial fate, and then added Norrin and vitronectin to specify retinal identity. A second sorting step enriched for CD31-positive, Frizzled4-positive cells, producing induced retinal endothelial cells, or iRECs.
These iRECs expressed canonical endothelial markers, including CD31, VE-cadherin and von Willebrand factor, as well as retinal barrier-associated proteins such as ZO-1, Occludin, Claudin-5 and Claudin-2. Functionally, they showed stronger barrier properties than non-tissue-specific induced endothelial cells, with higher transendothelial electrical resistance, lower permeability, and evidence of retinal-relevant transporter activity, including GLUT1-mediated glucose uptake and p-glycoprotein efflux.
The cells also behaved differently under diabetic retinopathy-like stress. When exposed to high glucose and hypoxia, iRECs showed reduced junctional localization of key barrier proteins, decreased barrier function, and disrupted vascular network architecture. In three-dimensional collagen hydrogels, diabetic conditions produced shorter branches, smaller network volumes, narrower lumens, disconnected vessels, and signs consistent with glycocalyx disruption. Non-retinal induced endothelial cells showed a weaker disease response, underscoring the value of tissue-specific models.
In an oxygen-induced retinopathy mouse model, intravitreal injection of iRECs reduced vaso-obliteration and pathological neovascularization compared with PBS-treated contralateral eyes. The transplanted cells integrated with host vasculature, formed lumenized human and human–mouse hybrid vessels, and helped restore vascular morphology toward healthier capillary dimensions. Treated retinas also showed reduced permeability, while dextran perfusion confirmed that iREC-derived networks could connect functionally with host circulation.
Beyond transplantation, the investigators built an iBRB-on-a-chip. iRECs self-assembled into perfusable three-dimensional microvascular networks with tight junction proteins and lower permeability than control endothelial networks. The team then derived retinal pericyte-like cells from hiPSCs using Norrin supplementation and incorporated them into the chip. These pericytes surrounded and supported iREC vessels, reduced lumen size toward physiological capillary dimensions, and further strengthened barrier function.
The work remains preclinical, and the study authors note important limitations, including the need for better tissue-specific markers, incorporation of additional retinal cell types such as astrocytes, and further validation of disease and transplantation models. Even so, their research offers an unusually complete platform: a source of human retinal endothelial cells, a disease-relevant iBRB model, and a candidate cell therapy strategy. For retinal microvascular disease, it points toward a future in which patient-specific cells could be used not only to understand barrier breakdown, but also to repair it.
