Quantitative Approaches to Study Retinal Neurogenesis

Abstract

The study of the development of the vertebrate retina can be addressed from several perspectives: from a purely qualitative to a more quantitative approach that takes into account its spatio-temporal features, its three-dimensional structure and also the regulation and properties at the systems level. Here, we review the ongoing transition toward a full four-dimensional characterization of the developing vertebrate retina, focusing on the challenges at the experimental, image acquisition, image processing and quantification. Using the developing zebrafish retina, we illustrate how quantitative data extracted from these type of highly dense, three-dimensional tissues depend strongly on the image quality, image processing and algorithms used to segment and quantify. Therefore, we propose that the scientific community that focuses on developmental systems could strongly benefit from a more detailed disclosure of the tools and pipelines used to process and analyze images from biological samples.

Keywords: imaging; quantitative biology; retinogenesis.

Figure 1

(A). Three-dimensional reconstruction (using FIJI plug-in 3DViewer) from confocal sections of a zebrafish retina at 44 h post fertilization. The tissue was mounted and stained with Topro3 (DNA marker) (see methods section) in toto. Three-dimensional reconstruction is generated. The three axes that define the developing retina are represented in three colors. (B). Illustration representing the shift from qualitative to quantitative studies in retinal neurogenesis. As spatial and/or temporal dimensions are taken into account, the amount of data generated and processed to obtain reliable and robust quantification increases.

Figure 2

Automated segmentation of images from developing zebrafish retina. (A) Section of a developing zebrafish retina at 44 h post fertilization stained with nuclear marker Topro3 (see methods section). (B) Segmentation of the same section using Fiji base algorithms (auto-threshold + watershed). Individual objects not correctly identified are highlighted with different colors. (C) Image segmentation of a section of a developing zebrafish retina at 44 HPF (using FIJI base algorithms) following the pipeline: (1) automatic thresholding; (2) binary morphological opening (open function); (3) binary morphological closure (close function); (4) watershed algorithm (watershed function). (D) Segmentation using the exact same filters, with the only difference being that the order between closing and opening filters is reversed. Nuclei segmented differently by these two very similar pipelines are highlighted in colors.

Figure 3

(A) Bar plots comparing manual quantification of the same image from trained users, using FIJI Cell Counter to assist with the quantification. Difference between users can be up to 35% of the mean value. (B) Increase in the number of data points to be processed when taking into account different dimensional levels of the developing vertebrate retina (vertical axis in log scale to help visualize the changes in terms of orders of magnitude). (C) Automated quantification of the number of nuclei of three different slices of a 3D stack of a given retina, showing differences up to 44% of the mean value. (D) Quantification of the nuclei density in terms of the average distance to the five closest nuclei show variations of 25% across the three different sections analyzed. Images are processed in FIJI, using an auto-threshold followed by the watershed algorithm. The number of objects is obtained from FIJI’s Analyze Particles.

Figure 4

Schematic representation of the pipeline followed to reconstruct objects in three-dimensional images by OSCAR. (Left): 3D representation of the original image (FIJI-3D Viewer). (Center): representation of how neighboring sections from the same 3D object are connected. Each colored line follows the trajectory of a given nucleus along the z-axis. (Right): 3D representation of the reconstructed 3D image. Colored objects represent the nuclei that correspond to the colored lines in the central panel. A processing and segmentation pipeline is applied to the original image in order to obtain a more reliable reconstruction (Gaussian blur sigma = 20 pix; Minimum filter sigma = 20 pix; Maximum filter sigma = 20 pix; Image Calculator—Original image subtract filtered; Manual thresholding; Binary Open; Watershed; 3D Objects Counter; 3D Draw Shape—All 3D objects represented as 10 pix radius spheres; 3D Viewer).