Quantifying progression of disease in aging mouse and human RPE

Abstract

Purpose: RPE sheet patterns change with age and progression in diseased RPE. We developed quantitative predictors of age and genotype from the RPE sheet, and built a model of how the RPE sheet changes in disease. Using functional principal component analysis (FPCA) and multidimensional scaling (MDS), we classified RPE sheets into major categories of genotypes and age groups.

Methods: Over 130 mice of many different ages and genotypes were used, including WT C57BL/6J and mutant rd10, rpe65 KO, rd12, IRBP KO, and tvrm148, which were congenic with C57BL/6J. Over 16 human cadaveric eyes were tested. ~10,000 cells per eye were analyzed. Borders of RPE cells were identified with anti-ZO-1 or phalloidin. Thresholding was by Ridler-Calvard adaptive method. RPE segmentation and analysis with 23 metrics were done with CellProfiler. Four classification methods were used after FPCA and MDS.

Results: We compared the morphology of RPE cells from eyes across various mouse strains and normal human and AMD eyes of different ages. Disrupted RPE had distinctive characteristics with great variability in size and shape of cells. In old RPE, there were gaps, atrophy, and interruptions in the sheet with significant show-through of the choroid. Joint aspect ratio and cell area provided principal components predictive of age (< or > 70 days) and genotype (WT or rd10) in mice with high accuracy (>98%). With further division of mice into four age groups and spatial locations, PCA and MDS showed that no single metric was significant as a signature. But spatial information revealed regional changes. At >P180 days, there was extensive deformation and a subpopulation of large cells. A mathematical model of RPE cell death recapitulated changes in vivo in AMD: Normal RPE progressed to abnormal only when multiple local clusters of RPE cells were killed. The death of single cells or sparse clusters of RPE cells were handled effectively by stretching the sheet. Preliminary analysis on human normal RPE showed no age-related trends or changes in morphology.

Conclusions: Morphometrics readily discriminates young vs old, normal vs mutant RPE sheets. Classification accuracy was high and depended on metric and location of the RPE cell. Tissue-stretching robustly maintains layer integrity. However, death of multiple local clusters of RPE cells is overwhelming and results in highly variable appearance and eventually frank holes.

Keywords: 701 retinal pigment epithelium • 696 retinal degenerations: hereditary • 599 microscopy: light/fluorescence/immunohistochemistry