Unbiased analysis of mouse brain endothelial networks from two- or three-dimensional fluorescence images

Abstract

Significance: A growing body of research supports the significant role of cerebrovascular abnormalities in neurological disorders. As these insights develop, standardized tools for unbiased and high-throughput quantification of cerebrovascular structure are needed. Aim: We provide a detailed protocol for performing immunofluorescent labeling of mouse brain vessels, using thin ( $25\mu m$ ) or thick (50 to $150\mu m$ ) tissue sections, followed respectively by two- or three-dimensional (2D or 3D) unbiased quantification of vessel density, branching, and tortuosity using digital image processing algorithms. Approach: Mouse brain sections were immunofluorescently labeled using a highly selective antibody raised against mouse Cluster of Differentiation-31 (CD31), and 2D or 3D microscopy images of the mouse brain vasculature were obtained using optical sectioning. An open-source toolbox, called Pyvane, was developed for analyzing the imaged vascular networks. The toolbox can be used to identify the vasculature, generate the medial axes of blood vessels, represent the vascular network as a graph, and calculate relevant measurements regarding vascular morphology. Results: Using Pyvane, vascular parameters such as endothelial network density, number of branching points, and tortuosity are quantified from 2D and 3D immunofluorescence micrographs. Conclusions: The steps described in this protocol are simple to follow and allow for reproducible and unbiased analysis of mouse brain vascular structure. Such a procedure can be applied to the broader field of vascular biology.

Keywords: angiogenesis; cerebrovascular; computation; connectivity; endothelium; image analysis; mouse brain; networks; unbiased.

Fig. 1

Primary antibody incubation on slides. Image depicts slide incubated with a small volume ($200\mu \mathrm{l}$) of primary antibody solution covered with a small piece of parafilm. Parafilm is large enough to cover all of the sections but slightly smaller than the width of the microscope slide. This method allows for the use of small volumes of antibody solution without the possibility of evaporation.

Fig. 2

Procedure for preparing mouse brain for vibratome sectioning. Brains are cut in half along a sagittal line. The olfactory bulbs and cerebellum are removed. The cortex is placed in PBS with the striatum facing up, and then the striatum along with the hippocampus is removed. The cortex is then placed on the non-coated side of the slide with the striatum side of the cortex facing down (step 1). Two glass separators, which were obtained by cutting a microscope slide into small pieces, are placed on either side of the slide and a second microscope slide covers the brains as shown in the figure above (step 1). Cortexes are placed in 4% PFA overnight in a plastic cell culture dish. The following day, the cortexes are embedded in agarose (step 2). The embedded cortex is then sectioned tangentially with a vibratome (step 2), stained, and mounted (see also Refs.  and 12).

Fig. 3

(a) Maximal intensity projection of CD31-stained vessels obtained by $10\text{-}\mu \mathrm{m}$-deep $z$-stacks. (b) Skeleton (see Sec. 3) representing the vasculature shown in (a) clearly captures all vessels within the $10\mu \mathrm{m}$ depth. (c) Illustration of cortex regions (green boxes) where images are captured for 2D analysis.

Fig. 4

(a) The tangential serial cortexes on the slide illustrate the anterior (blue box), parietal (green box), and occipital (gray box) cortical regions that could be imaged. (b) By pushing exposure and brightness, layer IV of the primary somatosensory barrel cortex is easily visible via background immunofluorescence. Tangential serial sections above and below layer IV are considered as layer II/III and V, as indicated in (a). (c) The primary somatosensory barrel cortex (S1) can be used as a landmark to identify other neighboring brain regions, namely, frontal/motor cortex (F/M), auditory cortex (A1), and visual cortex (V1).

Fig. 5

Illustration of the methodology used for characterizing vascular networks. The middle column shows the overall steps required for the analysis. The procedures used for implementing each step are shown on the right. Example images are shown on the left.

Fig. 6

Illustration of the pruning procedure. (a) The smallest branch (red) is detected. (b) The removal of the smallest branch generates a new branch, which is also removed because its arc-length is smaller than a threshold size. The branches are removed until there are no branches smaller than the threshold. The final result is shown in (c).

Fig. 7

Illustration of the methodology for calculating blood vessel tortuosity. (a) The skeleton of the vascular network is divided into segments. (b) The tortuosity at a reference pixel $p_c$ (in orange) is calculated as the average length of the red lines shown in the figure, which represent the smallest distances between the pixels and line $r$ (in purple).

Fig. 8

Example of application of the method to one of the samples. (a) Original sample. (b) Detected vascular network, represented as a binary image. (c) Skeleton of the vascular network. (d) Graph representing bifurcations and terminations (shown in blue), as well as their connections (shown in green).

Fig. 9

Measurements obtained for three selected 2D samples. Sample (a) has the largest density among the selected samples, whereas (b) has the lowest density and the largest average tortuosity. Sample (c) has low average tortuosity.

Fig. 10

Example of 3D vascular network detection and digital reconstruction. (a) Original sample. (b) Visualization of the detected blood vessels. The colors indicate the diameter of the blood vessels, with brighter colors representing thicker vessels.

Fig. 11

Measurements obtained for six selected 3D samples. Sample (a) has the largest density among the selected samples, whereas (b) has the lowest density. The vasculature in (c) has large average tortuosity, which is a consequence of its irregular blood vessel segments. Sample (d) has relatively low average tortuosity. Samples (e) and (f) have intermediate values of the considered measurements. The images shown are maximum $Z$-projections of the original samples.

Fig. 12

Distribution of the (a) vessel density, (b) density of branching points, and (c) average tortuosity for the considered set of 3D samples.

Fig. 13

Tortuosity values calculated for individual pixels of blood vessels. The tortuosity of the vasculature shown in (a) was calculated at a scale of (b) $d=10\mu \mathrm{m}$ and (c) $d=20\mu \mathrm{m}$.