Abstract
Purpose :
RPE becomes stressed and dysmorphic in retinal degeneration mice, evident in flat-mount preparations of the RPE. Over-expression of Nrf2 in rd1 mice RPE slows vision loss and preserves morphology of the RPE. Quantification of improvement of RPE morphology, preferably automated, would allow more rapid analysis of degeneration tissue. It would also allow determination of whether this may be a useful biomarker for rescue by other candidate therapeutic molecules.
Methods :
Eyes from mice with the rd1 mutation were collected across multiple timepoints (p40-p127). Some eyes were transfected with adeno-associated virus (AAV) Best1 Nrf2 driven by BEST1 promotor. Outlines of RPE were identified with phalloidin staining and imaged with spinning disk confocal microscopy. A customized eight-step Cell Profiler pipeline automated segmentation and measurement of RPE shape. RPE outlines were also manually traced using Fiji to extract similar shape properties. Three areas were sampled per eye: a location of greatest rescue of RPE morphology by Nrf2 and the corresponding location in the contralateral unrescued eye, and a conserved central and peripheral location across all eyes. Mean, variance, and skew of the area and aspect ratio distributions were calculated. Area under ROC curves (AUC) for these variables were used to quantify the differences between treated and untreated mice.
Results :
Rescue by Nrf2 results in a 27% decrease in mean cell area (p=0.0015), 89% decrease in cell area mean, sample variance (p=0.0069), and a 37% decrease in area sample skewness (p=0.0002), corresponding to AUCs of 0.68, 0.75, and 0.76 respectively. Additionally, Nrf2 rescue results in 22% increase in mean aspect ratio (p<0.0001) and 48% decrease in aspect ratio mean sample variance (p<0.0001), corresponding to 0.96 and 0.92 respectively. CellProfiler data provided less sensitivity, but still resulted in a 5% increase in mean aspect ratio corresponding to an AUC of 0.77.
Conclusions :
We are able to demonstrate a quantitative difference between rescued and non-rescued RPE utilizing parameters of cell area and cell aspect ratio. Although manual segmentation of RPE offers more sensitivity, CellProfiler is able to detect differences in rescued vs non-rescued eyes. Further improvements to the CellProfiler pipeline may bring its performance closer to that of human segmentation.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.