VESGEN 2D: automated, user-interactive software for quantification and mapping of angiogenic and lymphangiogenic trees and networks

Abstract

Quantification of microvascular remodeling as a meaningful discovery tool requires mapping and measurement of site-specific changes within vascular trees and networks. Vessel density and other critical vascular parameters are often modulated by molecular regulators as determined by local vascular architecture. For example, enlargement of vessel diameter by vascular endothelial growth factor (VEGF) is restricted to specific generations of vessel branching (Parsons-Wingerter et al., Microvascular Research72: 91, 2006). The averaging of vessel diameter over many successively smaller generations is therefore not particularly useful. The newly automated, user-interactive software VESsel GENeration Analysis (VESGEN) quantifies major vessel parameters within two-dimensional (2D) vascular trees, networks, and tree-network composites. This report reviews application of VESGEN 2D to angiogenic and lymphangiogenic tissues that includes the human and murine retina, embryonic coronary vessels, and avian chorioallantoic membrane. Software output includes colorized image maps with quantification of local vessel diameter, fractal dimension, tortuosity, and avascular spacing. The density of parameters such as vessel area, length, number, and branch point are quantified according to site-specific generational branching within vascular trees. The sole user input requirement is a binary (black/white) vascular image. Future applications of VESGEN will include analysis of 3D vascular architecture and bioinformatic dimensions such as blood flow and receptor localization. Branching analysis by VESGEN has demonstrated that numerous regulators including VEGF(165), basic fibroblast growth factor, transforming growth factor beta-1, angiostatin and the clinical steroid triamcinolone acetonide induce 'fingerprint' or 'signature' changes in vascular patterning that provide unique readouts of dominant molecular signaling.

Fig. 1. VESGEN 2D Analysis of Branching Vascular Trees

(A) To obtain binary vascular images for VESGEN 2D analysis in the CAM, the arterial end-point region is generally imaged in grayscale by brightfield microscopy at low magnification (here, 12.5× total) to maximize the field of vascular branching. Magnification was also selected to exclude the CAM capillary bed (typically visualized at ≥ 20× total), because the capillaries are organized as a symmetric, homogeneous vascular network, and are thus morphologically distinct from the branching arterial and venous trees. The E8 CAM specimen illustrates how aldehyde fixation results in the retention of blood containing red blood cells (RBCs) within arteries (arrows), but not in veins (arrowheads). Arterial trees are thereby conveniently separated from overlapping venous trees, and (B) extracted by semi-automatic computer processing as a binary (black/white) vascular pattern. (C) VESGEN 2D generates the skeletonized (linearized) image of vascular connectivity from (B), the single input binary image. Vessel skeletons in this and all other figures have been widened for illustration (by definition, skeletons are one pixel in width). The software then maps and quantifies the vascular tree using various image processing algorithms to analyze these two geometric representations (B, C). (D) This VESGEN 2D output displays the major arterial tree with its automatically determined ROI (in black) and classification of vessels into ten successively smaller branching generations, G₁-G₁₀. For this E8 control specimen, for example, (B) vessel area density (A_v) = 0.136 (cm²/cm²) and fractal dimension (D_f ) = 1.66; (C) vessel length density (L_v) = 25.2 (cm/cm²), vessel branch-point density (Br_v) = 380 (cm⁻²) and D_f = 1.40. For the extracted major vascular tree (D), A_v1-3 = 0.057 (cm²/cm²), L_v1-3 = 2.8 (cm/cm²), and N_v1-3 = 11.4 (cm⁻²), compared to A_v7-10 = 0.040 (cm²/cm²), L_v7-10 = 15.1 (cm/cm²), and N_v7-10 = 397.3 (cm⁻²).

Fig. 2. VESGEN 2D Analysis of Vascular Networks

Obtaining a VESGEN 2D analysis of a vasculogenic or mature capillary network is similar to the previous example of a branching arterial tree (Fig. 1). (A) The vasculogenic capillary network from a P15 postnatal mouse retina visualized by Tie2/GFP expression at 10× total magnification was binarized and for this example, (B) a Euclidean distance map (EDM or DM) was generated from the binary pattern by VESGEN 2D. As indicated by the look-up table (LUT), colors within the vascular DM correspond to pixel distance to edge of vessel. For this vascular network, VESGEN 2D measured the vessel area density (A_v) as 0.14, vessel branch point density (Br_v) of 164 per retinal field, average vessel diameter (D_v) of 4.83 pixels, 50 enclosed avascular spaces, and avascular area fraction of 0.86. (C) By another vascular mapping option, the colorized skeleton was overlaid on the binary pattern where colors again correspond to thickness (or radius) of the vessel at the location of each skeletal point as given by the LUT (B). Averaging DM values at points along the skeleton can be used to estimate the average radius of local vessel segments as well as of the entire vessel network.

Fig. 3. Extraction of Vascular Trees from Clinical Images of the Human Retina

Vascular trees were extracted from (A, C) grayscale ophthalmic images obtained by clinical FA using semi-automatic computer processing (B, D). Arterial and venous trees were distinguished from each other according to FA guidelines for retinal arterio-venous anatomy, fluorescein filling, and vessel connectivity, as well as well-established morphological characteristics such as the smaller diameters and increased vessel tortuosity of arteries compared to veins. (B, D) The appearance of the overlapping arterial and venous trees suggests that significant vessel dropout occurred during the progression from mild/moderate (A) to severe (C) nonproliferative diabetic retinopathy (NPDR) as diagnosed clinically using other (secondary) vascular markers such as microaneurysm density.

Fig. 4. VESGEN 2D Analysis of NPDR Progression in the Human Retina

By VESGEN 2D measurements in generation-specific maps of (A, C) arterial trees and (B, D) venous trees, the largest difference between the mild/moderate NPDR and severe NPDR images was a large decrease in the number and density of smaller arteries (G5-G9), compared to decrease in the density of larger arteries (G1-G4). For example, vessel number density N_v5-9 decreased by 76% (from 629 arteries per normalized retinal field to 155), and N_v1-4, by 44% (from 36 arteries per retinal field to 20) in the severe NPDR arterial tree (C) relative to mild/moderate NPDR (A). Throughout the entire arterial trees of (C) relative to (A), vessel area density (A_v1-9) decreased by 30%, vessel length density (L_v1-9) by 51%, and vessel branch-point density (Br_v1-9) by 78%. Br_v typically correlates closely with N_v. Differences between the numbers and density of venous vessels were much smaller. Within the entire venous trees of (D) relative to (B), A_v1-9 increased by 8%; L_v1-9 and Br_v1-9 decreased by 15% and 30%. Decreased arterial density in the severe NPDR image could result from several causes, including arterial oblation, and/or a decrease in arterial diameter below the limit of imaging resolution, which is approximately 40 µm.

Fig. 5. In the CAM, Anti-Angiogenic Regulation by TA Decreases Selectively the Number of Smaller Vessels, but Decreases Vessel Diameter throughout the Vascular Tree

Binary and skeletonized images of a representative E8 control (A, B) and TA-treated CAM (D, E) suggest that application of TA decreased both vessel density and vessel diameter. (C, F) Quantification by VESGEN 2D demonstrated TA altered vascular architecture by two major morphological mechanisms: (1) vessel diameters decreased throughout the branching vascular tree, accompanied by (2) selective decrease in the density of smaller blood vessels as indicated by vascular maps of extracted major arterial trees within the ROIs (in black).

Fig. 6. The Anti-Angiogenic Steroid TA Decreases Vessel Density and Enlarges Avascular Spaces within Vasculogenic Capillary Networks of the Developing Mouse Retina

Compared to PBS-treated control (A–C), vessel density and diameter decreased considerably in the peripheral vasculogenic capillary networks of the developing mouse retina (D–F) following treatment with TA prior to P15. As a consequence, the number of avascular spaces also decreased. Network ROIs are labeled in black (B, E). However, only the numbers and areas of closed avascular spaces (illustrated in white, C, F) are quantified by VESGEN 2D (unclosed avascular spaces, appearing in gray, are excluded). Retinal vasculature was labeled by perfusion of lectin-conjugated fluorescein injected into the heart before sacrifice and after fixation, dissection and mounting, was imaged by fluorescence microscopy at 10× total magnification.

Fig. 7. VEGF₁₆₅ Disorganizes Lymphatic Architecture and Enlarges Vessel Diameter and Avascular Spaces of Lymphatic Networks in the CAM

(A, E) Compared to E8 control, vessel diameter increased–although with large variability–in lymphatic networks of the quail CAM following treatment with VEGF₁₆₅ for 24 hrs. CAM specimens were labeled with antibodies identifying VEGFR-2 (green) and a quail hematopoietic epitope (red). Arrows point to blood vessels, arrowheads to lymphatic vessels and networks, and hollow arrowheads (E) to a large, more homogeneous lymphatic network not associated with blood vessels that is typical of earlier stages of lymphatic development (B). As visualized by confocal microscopy at relatively low magnification, large numbers of isolated vascular progenitor cells appear as punctate staining and are being recruited to growing lymphatics (Parsons-Wingerter et al., 2006a). As shown by (C, F) vascular distance maps (DM) and (D, G) colorized skeleton overlays generated by VESGEN 2D network analysis, normal lymphatic architecture consists of highly organized lymphatic networks surrounding large blood vessels and smaller lymphatic vessels encircling smaller blood vessels. The total area of an image, as indicated by large non-lymphatic regions of the CAM (C and F, black) is used to quantify lymphatic vessel density and other density parameters according to standard practices for normalizing vascular quantities by tissue area. In contrast, white areas (C and F) indicate the avascular ‘island’ spaces which are totally enclosed within (compared to the areas of gray avascular spaces which extend beyond the field of the microscopic image); these avascular ‘island’ spaces are useful for characterizing vascular/avascular relationships within the network (LaRue et al., 2003). (A, B & E) reprinted with permission of Anatomical Record A.

Fig. 8. Development of Embryonic Coronary Vessels from an Immature Vasculogenic Network into a Mature Vascular Tree

(A–C) The coronary vasculature begins as an immature, undifferentiated vasculogenic network when the embryonic heart first grows large enough to require metabolic support by its own blood supply. This coronary network is from an E13.5 mouse heart in which the vessels were labeled by FITC-conjugated CD-31 and imaged by confocal microscopy at 400×. As in previous examples, the coronary network is represented by its DM and a colorized skeleton indicating vessel thickness overlaid upon the binarized network. (D–G) By E15.5, the murine coronary network has remodeled into a vascular network-tree composite (as visualized by brightfield stereo microscopy at low magnification, approximately 9×). According to VESGEN 2D tree analysis, branching generations are not yet well defined by successive, orderly, tapering decreases in vessel diameter. The vascular composite contains both tapering branches and anastomatic loops that are characteristic features of trees and networks, respectively. (H–I) By HH35 (approximately E9) in the chicken heart, which matures more quickly than the mouse heart, the coronary vasculature invested by smooth muscle has developed into a relatively mature tree composed of continuously tapering branches. Coronary vessels were labeled by Cy3-conjugated alpha smooth muscle actin (αSMA) and imaged by fluorescence stereo microscopy at 5×. The absence of smaller vessels within this tree results primarily from labeling by αSMA, which is not expressed abundantly on most smaller vessels, the low magnification, and the maturational investment of smaller blood vessels beneath the epicardium. The image in (A) is reproduced with permission from (Barbosky et al., 2006).