Diabetic retinopathy (DR) is a leading cause of blindness in working-age populations. Proper in vitro DR models are crucial for exploring pathophysiology and identifying novel therapeutic targets. This study establishes a rational in vitro diabetic retinal neuronal-endothelial dysfunction model and a comprehensive downstream validation system.Human retinal vascular endothelial cells (HRMECs) and retinal ganglion cells (RGCs) were treated with different glucose concentrations with mannitol as matched osmotic controls. Cell proliferation and viability were evaluated by the Cell Counting Kit-8. Cell migration was measured using a transwell migration assay. Cell sprouting was assessed by a tube formation assay. The VEGF expression was assessed by ELISA. RGCs were labeled by neurons and RGC markers TUJ1 and BRN3A for quantitative and morphological analysis. Apoptosis was detected using PI/Hoechst staining and TUNEL assay and quantified by ImageJ.Cell proliferation and migration in HRMECs were significantly higher in the 25 mM glucose-treated group (p < 0.001) but lower in the 50 mM and 100 mM groups (p < 0.001). The permeability and the apoptotic index in HRMECs were statistically higher in the 25 mM, 50 mM, and 100 mM groups (p < 0.05). The tube formation assay found that all the parameters were significantly higher in the 25 mM and 50 mM groups (p < 0.001) concomitant with the elevated VEGFA expression in HRMECs (p = 0.016). Cell viability was significantly lower in the 50 mM, 100 mM, and 150 mM groups in RGCs (p 50mM = 0.013, p 100mM = 0.019, and p 150mM = 0.002). Apoptosis was significantly elevated, but the proportion of RGCs with neurite extension was significantly lower in the 50 mM, 100 mM, and 150 mM groups (p 50mM < 0.001, p 100mM < 0.001, and p 150mM < 0.001).We have optimized glucose concentrations to model diabetic retinal endothelial (25-50 mM) or neuronal (50-100 mM) dysfunction in vitro, which have a wide range of downstream applications.Copyright © 2021 Qiyun Wang et al.