Divergence Between Marker Expression and Phenotype During Human Corneal Endothelial Cell Expansion

Abstract

Purpose: The purpose of this study was to compare whether conventional marker-based assessments adequately reflect phenotypic stability during serial passaging of two established culture protocols for primary human corneal endothelial cells (CEnCs). Methods: Primary CEnCs were isolated from paired donor corneas from 25 individuals (18-44 years of age), and were cultured using 2 established methods. Cells were evaluated at passages 0 to 2 using quantitative morphology (cell area, form factor, nucleus-to-cytoplasm ratio, and cell density), RT-qPCR of endothelial-associated genes, flow cytometry, immunofluorescence, and ELISA quantification of secreted thrombospondin-2 (TSP2). Statistical comparisons were performed using 1- or 2-way ANOVA with multiple comparisons (P < 0.05). Results: CEnCs cultured with method 1 maintained higher cell density, smaller cell area, and greater hexagonality across passages compared with method 2 (P < 0.01 to P < 0.0001). In contrast, transcript levels of selected endothelial-associated markers (ATP1A1 and TJP1) and ZO-1 protein expression were higher in method 2 at later passages (P < 0.01), despite progressive morphological deterioration and loss of membrane localization. Proteins associated with loss of the CEnC phenotype (CD44 and CD10) were significantly elevated in method 2 (P < 0.05 to P < 0.0001), and TSP2 secretion was increased at passage 2 (P < 0.001). Across both methods, most canonical endothelial genes were ≥75% downregulated compared with native tissue. Conclusions: Culture-induced phenotypic drift can occur in CEnCs despite preserved canonical endothelial marker expression. Divergence among marker expression, morphology, localization, and endothelial-to-mesenchymal transition (EndMT) readouts highlights the limits of marker-based evaluation alone. Integrated, multidimensional phenotyping should guide assessment of cultured CEnCs and optimization of expansion protocols, although functional validation is needed to link these readouts to endothelial performance.