Abstract
Purpose: :
Quantification of cells and changes in cell number in retinal flatmounts is an important method in basic and translational research. Unblinded and subjective image analysis, however, may yield biased counting results. Using the example of astrocytic density in the murine retina we demonstrate a novel method for valid and objective cell quantification avoiding user-dependent selection bias.
Methods: :
Mice expressing histone-bound GFP under the Pdgfra promoter are used to identify astrocytic nuclei in the retina. Retinal vessels in flatmounts are stained with an antibody raised against collagen IV. GFP signal is photographed in the green channel while collagen IV is photographed in the red channel. Images are evaluated by an automated ImageJ macro (Astro Count Macro) which allows for blinded selection of regions of interest (ROIs) and a documented counting process. ROIs are chosen solely on the basis of the collagenIV-stained vasculature in the murine flatmount without the GFP signal of retinal astrocytes being visible to the user. The ROIs from the vascular image are then automatically projected onto the astrocyte image on which the counting process is performed. Finally, the red and green channels are merged and the counted cells labeled. The numeric results can easily be exported to spreadsheet calculation programs.
Results: :
The nuclear GFP signal expressed under the Pdgfra promoter allows quantitative analysis of retinal astrocytes. Using the Astro Count Macro, we show that astrocytes are not equally distributed during the physiological development of the murine retina. The density of astrocytes in the periphery at P7 exceeds their density in the central retina. The blinded selection of ROIs with the Astro Count Macro yields user-independent (R²= 0,96) and time-consistent results (R²= 0,98). Manual and automated counting processes yield highly concordant results (R²=0,95).
Conclusions: :
The Astro Count Macro provides an objective and documented cell counting method which is able to detect differences in retinal cell density. The automated counting requires reasonably good image quality but is not limited to astrocytes. The method might therefore be similarly useful for objective analysis of other labeled retinal cells or structures like ganglion cells, apoptotic cells, and vessels.
Keywords: astrocyte • retinal glia • image processing