Coalescent angiogenesis—evidence for a novel concept of vascular network maturation

The generation of blood vessels is a vital process for embryonic development and organ growth, as well as in various pathophysiological conditions such as wound healing, ischemic heart disease and cancer. Two main processes for vascular development are recognized: vasculogenesis and angiogenesis. Vasculogenesis occurs during embryonic development, when the so-called primary vascular plexus is generated by differentiating mesodermally derived angioblasts. This de novo synthesis results in lumenized vascular structures or capillary meshes that generally lack hierarchical organization. In contrast, angiogenesis refers to the generation of new blood vessels from pre-existing ones [1] and supports the expansion of the primary vascular plexus into a more mature vascular network [2, 3]. Two distinct modes of angiogenesis have been described: In sprouting angiogenesis (SA), vessel outgrowth from existing vessels is initiated by specialized endothelial cells termed tip cells. Migration and proliferation lead to an elongation of the initial tip which may eventually connect, or anastomose, with another vessel to form a perfused vascular branch [4, 5]. SA is prominent in situations where perfusion is low and hypoxia is present such as during de novo generation of tissues in embryonic development or wound healing, and is primarily regulated by angiogenetic signals related to local hypoxia [6, 7].

In contrast to SA, intussusceptive angiogenesis (IA) or splitting angiogenesis is a non-sprouting mechanism, first described by Burri et al. for the development of the pulmonary vasculature in postnatal rats [8]. In IA, transluminal endothelial pillars form within a blood vessel and progressively split it into two daughter vessels. These transluminal pillars are formed by evaginations of endothelial cells, creating contact zones between opposite capillary walls. This process allows rapid expansion of a capillary network while maintaining perfusion and, thus, tissue function [9‐11] and has since been observed in many tissues including skeletal muscle and kidney [12, 13] as well as in pathological conditions such as COVID-19 [14]. Generation of new blood vessels by this mechanism has also been termed intussusceptive microvascular growth. In contrast, intussusceptive arborization and intussusceptive branching remodeling [15] involve the reorganization of an existing capillary bed via expansion or pruning of vessel branches to optimize branching geometry and hemodynamics in existent vascular trees [16‐18].

In addition, a process of vascular fusion was described in the context of blood vessel generation [19‐21]. Under physiological conditions all three processes can occur in parallel to meet the functional needs of a vascular bed [22‐25].

Sprouting angiogenesis has been intensively studied using intravital imaging approaches, such as the zebrafish model [26, 27]. This has led to substantial knowledge of the underlying basic physiological and molecular mechanisms as well as the cellular dynamics and time courses especially for embryonic development [28, 29]. Due to a paucity of appropriate model systems, such information is less readily available for intussusceptive angiogenesis and the generation of vascular beds in solid organs, including the lung. Consequently, our insight into the basic mechanisms by which such existing capillary beds expand and grow is still scarce. In-depth understanding of these processes requires studies delineating the morphological dynamics of IA vessel generation – similar to those previously reported for SA—to analyze the angiogenic processes and their interaction during organ growth (e.g., of pancreas, lung, or kidney) and in embryonic development.

The chick chorioallantoic membrane (CAM) presents a versatile in vivo model for the study of angiogenesis and vascular adaptation [30] as it facilitates optical access to a developing vascular system in a transparent, two-dimensional layout. Importantly, the CAM allows for extended observation periods for up to 1–2 days, permitting to follow the successive steps of angiogenic development over time [31]. In the fertilized chicken egg, the CAM is created through fusion of the allantois and the chorion from day 3 of embryonic development and its development continues until embryonic day 13 [32]. The CAM serves as the respiratory organ of the embryo [33], and as such is characterized by functional similarities to the human lung, including the transport of deoxygenated blood in arteries and oxygenated blood in venular vessels as well as similar regulatory mechanisms such as vasoconstriction in response to hypoxia [34]. The CAM thus potentially provides a unique model to study the IA in a longitudinal manner and was thus utilized here to delineate the spatio-temporal dynamics of IA as the basis for future mechanistic analyses. In the present study, intravital microscopy of the CAM over a period of 33 h was used to monitor rearrangement and growth of blood vessels and the capillary mesh. Unexpectedly, these observations revealed evidence that mature vessel trees are generated by a previously unrecognized mode of angiogenesis, which is distinct from intussusception and which we termed here “coalescent angiogenesis”.

Figure. 1A shows CAM vascular patterns as seen in bright-field intravital microscopy. Arterial and venous trees are arranged in an interdigitating pattern and exhibit a clear hierarchical organization and diameter distribution. Capillary structures are difficult to assess on still bright-field images (Fig. 1B, C) but become apparent in video sequences due to the movement of red blood cells (Supplementary Material 1). Image processing of such sequences allows for differentiation between vascular and non-vascular areas (Fig. 1D, E). The resulting images show that capillaries are organized in polygonal isotropic meshes, with capillaries encircling non-perfused tissue islands and link to feeding or draining vessels via areas with enlarged capillary patterns. These observations are supported by additional image processing which revels perfusion images as described and shows semi-quantitative flow patterns in the vascular network (Fig. 1F). The algorithm permits adequate visualization of the local perfusion in bigger blood vessels but not of the flow of the boundary regions of the microvascular blood vessels including the tissue islands.

To follow angiogenetic processes in the CAM, time series of selected CAM areas were repeatedly scanned over up to 33 h (Fig. 2). The progression from capillaries into a definitive venule can be differentiated into four phases of blood vessel development. The process starts from a predominantly isotropic capillary mesh in which capillary diameters, length and perfusion exhibit relatively narrow distributions (phase I, isotropic mesh). Within this mesh preferential pathways develop, which are characterized by increased vessel diameter and perfusion (phase II, preferential pathway) with a tendency for a reduction in capillary density in adjacent areas. In phase III (emergent vessel), parallel capillaries coalesce to form a single vessel structure constituting a structural pathway. In this process, tissue islands initially located within the preferential pathway progressively disappear (Fig. 2, middle panels). The last phase (phase IV, established vessel) is characterized by the presence of a definitive vessel with few side branches that now forms an integral part of a hierarchical vessel tree. Microscopic investigation at different focus levels shows that a capillary mesh is partially re-established above the new vessel, which sinks to a slightly lower level toward the yolk sac (see Supplementary Material 2). Essential morphological and hemodynamic properties of the observed phases are summarized in Table 1.

These sequential phases of coalescent angiogenesis can not only be observed over time at a given location but are also evident at one single time point along the length of a branching vessel tree, with the early phases corresponding to distal portions of the tree and later phases to its proximal portion (Fig. 3A). Following the vessel tree from distal to proximal, phases I to IV can be clearly distinguished. In the frame corresponding to phase III, a single remnant tissue island can be seen within the coalescent vessel. Such structures were regularly observed. From an individual picture, this structure could be interpreted as a developing tissue pillar, according to the concept of intussusceptive angiogenesis. As shown in the time series (Fig. 3B and Supplemental Material 2), however, these structures actually represent tissue islands that are eliminated in the process of coalescent angiogenesis, rather than newly generated intravascular pillars as would be postulated in intussusceptive angiogenesis. As such, coalescent angiogenesis presents a novel mechanism of network organization that may be adequately described as inverse intussusception.

As shown in the supplemental material (S-Figure S1 and S-Figure S2), there was no significant change of net area and capillary density in isotropic meshes in the central parts of the CAM microvascular network investigated in the present study. Thus, the observed processes of coalescent angiogenesis occur on the background of a constant tissue area and stable capillary topology.

Vascular networks in growing tissues must develop rapidly to meet increasing oxygen and nutrient demands. After the initial formation of vessels by vasculogenesis, the enlargement of a vascular system requires angiogenesis [5] with concomitant vascular remodeling and pruning to achieve hemodynamically efficient network structures [25, 39]. However, information regarding the dynamic steps of intussusceptive angiogenesis is lacking, especially in tissues undergoing fast growth or development. For such conditions, as in the growing lung, skeletal muscle or liver, intussusceptive angiogenesis (IA), i.e., division of pre-existing vessels by newly emerging tissue pillars, has been proposed as a central mechanism [13]. Previous studies have reported intussusception for both single capillaries of the CAM and in the development of larger blood vessels [10, 16, 40, 41]. Such studies were, however, usually performed by analysis of structurally fixed blood vessels or using microscopic observations at a single time point and hence, lack any temporal information regarding the dynamics or – importantly – the direction of this process.

Continuous or repeated observation over periods relevant for angiogenetic processes can resolve this knowledge gap and as such, provide new insights into angiogenic modes and mechanisms. In the present study, the repeated observation of the microcirculation in the chick chorioallantoic membrane (CAM) yields evidence for a novel mode of angiogenesis which we termed coalescent angiogenesis (CA). In CA, blood vessel tree extension occurs via parallel processes of vessel maturation and regression of capillary mesh segments. As such, CA in the CAM of embryonic chicken eggs adds a novel mode of microvascular growth to previously described patterns of angiogenesis (Fig. 4).

These different modes are adapted for specific functional requirements and conditions. In developmental sprouting angiogenesis, vascular structures are formed de novo during embryonic development. The corresponding stimuli are generated by a highly conserved sequence of events, leading to a largely genetically determined vasculature. A typical example of developmental sprouting is the generation of the main vessels in the embryonic zebra fish [42]. The angiogenetic stimuli leading to reactive sprouting angiogenesis are generated in the postembryonic phase in response to pathophysiological and physiological events, including wound healing, inflammation, cancer, and muscle training [43, 44]. Due to the stochastic processes involved, resulting vascular structures are quite heterogeneous and require intensive vascular adaptation and pruning to obtain functionally adequate vascular networks [25].

Both forms of sprouting angiogenesis initially generate vessels with non-connected free ends which do not contribute to perfusion until they connect to a respective draining or feeding vessel. In contrast, intussusceptive (splitting) and coalescent angiogenesis allow the fast expansion of vascular beds while maintaining perfusion, tissue function, and permanent functional remodeling. Due to the continuous perfusion of the vascular bed, angiogenetic signals derived from local hemodynamics are likely to play a more central role in intussusceptive (splitting) and coalescent angiogenesis, as compared to hypoxia-derived signals characteristic for sprouting angiogenesis [39, 45, 46]. This specific mode of predominant regulation by hemodynamic rather than hypoxic cues may be particularly prominent in organs where lack of perfusion does not cause hypoxia such as the CAM and the lung.

Intussusceptive angiogenesis was first described by Burri et al. in the pulmonary vasculature and describes the division of existing vessels, leading to two parallel segments and thus to relatively uniform vessel networks [8, 9]. The central element in intussusceptive (splitting) angiogenesis is the intravascular formation of tissue pillars which fuse to generate a permanent wall between the new parallel vessels. Referred to as an intussusceptive pillar, this endothelial-lined intravascular structure is commonly not visible by light microscopy but has been studied largely by corrosion casting and scanning electron microscopy [16, 41, 47].

Coalescent angiogenesis shares many properties with splitting angiogenesis, albeit with a very different morphological process for vessel generation. CA starts from an isotropic capillary network which is transformed to give rise to hierarchical and functionally adequate arterial and venous vessel trees. This process not only remodels existing vessels with respect to their diameter and wall composition but generates novel vascular structures required for effective perfusion. In this process, capillaries of the isotropic mesh fuse to establish feeding or draining vessels via the elimination of tissue islands separating them. In this process, the tissue island attains a morphological appearance that resembles the tissue pillars seen in splitting angiogenesis (Fig. 3B and Supplemental Material 2). However, in coalescent angiogenesis, these structures are progressively eliminated over time to give rise to definitive vessels, while in intussusceptive (splitting) angiogenesis, they grow and fuse to split an existing vessel. Accordingly, coalescent angiogenesis may be seen as the inverse process to splitting angiogenesis.

In the past, analysis of non-sprouting angiogenesis has mainly focused on splitting or intussusception. However, it is important to note that the respective studies are almost exclusively based on static images which cannot discriminate vessel division from vessel fusion. This dilemma was recently highlighted in a review by Gifre-Renom and Jones who proposed vessel fusion as a mode of non-sprouting vascular adaptation [48]. In this context, they point out the principal difficulty to distinguish fusion processes (combination of two or more small vessels to a larger one) from intussusceptive angiogenesis (splitting of a large vessel into two or more smaller ones). Individual micrographs representing a single point in time show anatomical features which could fit to both processes: The small roundish or ellipsoidal structures seen in microvessels, e.g., close to branching points (see Figures 1E and 3A), may represent tissue islands in the process of elimination during ‘coalescent angiogenesis’ or tissue pillars which grow to separate an existing vessel in ‘splitting angiogenesis’.

Only by following the anatomical development at a given location over time, as done in the present study, coalescent angiogenesis by vascular fusion can be clearly separated from splitting angiogenesis (intussusception). For the vascular bed studied, this approach provides direct evidence for coalescent angiogenesis of established vessels as the mechanism for vessel enlargement. Future time series studies are needed to (re-) define the respective roles of coalescent angiogenesis and splitting angiogenesis in other tissues and conditions.

Coalescent angiogenesis seems to have partial similarities with intussusceptive arborization. Intussusceptive arborization segregates the various vessel generations of capillary mesh networks by formation of vertical pillars in rows and narrow tissue septa formed by pillar reshaping and pillar fusions segregate the new vascular entities. Also the formation of horizontal pillars and folds which separate the new vessels from the capillary plexus is described as intussusceptive arborization to form hierarchical microvascular segments restructuring the primitive capillary meshwork to a vascular tree [47]. In Phase II of coalescent angiogenesis, a rearrangement of tissue islands along the preferred pathway is observed along with the enlargement of tissue islands in adjacent regions in phase IV. However, in intussusceptive arborization, pillars and similar structures are generated de novo, whereas in coalescent angiogenesis they represent remnants of tissue islands. Further studies are needed to clarify these aspects.

The CAM is functionally a two-dimensional analog to the lung, with both structures exchanging oxygen and carbon dioxide with the exterior via a dense mesh of capillaries that are fed and drained by dichotomous hierarchical arterial and venous vessel trees, respectively. Thus, it is tempting to speculate that the mode of vascular growth observed here—i.e., coalescent angiogenesis—may also be relevant for the development and/or adaptation of the pulmonary circulation. Yet although lung intravital microscopy has been successfully established and advanced in several labs including ours, such studies pose a considerable technical challenge due to the complex three-dimensional structure of the pulmonary vasculature, the inherent problems associated with prolonged imaging of the lung, and motion artifacts over the respiratory and cardiac cycle [49, 50]. Even for the CAM model, the concept of coalescent angiogenesis requires further studies and in-depth analysis, e.g., with respect to the relevance of hemodynamic and metabolic stimuli, to the cellular and molecular mechanisms involved in the reorganization of vascular structures, as well as to the generation of the initial capillary mesh. The present results (Supplementry Figures S1 and S2) for regions remote from the CAM’s growing border zone indicate that capillary meshes are stable with respect to vessel area and length. However, in the late stages of coalescent angiogenesis (see Supplementary Material 2), the capillary mesh is reconstituted on top of the main vessels which slightly sink into the yolk sac.

There is evidence that angiogenesis in adult lungs is generally limited, except in pathological conditions such as tumor growth, wound healing, or fibrosis. With respect to tumor growth, there is a non-angiogenic process described as vessel co-option, which is an alternative mechanism of generating the blood supply that differs fundamentally from the well-known sprouting angiogenesis and occurs in a significant fraction in highly vascularized organs like lung, brain, and colon [51‐55]. Here tumor cells grow into pre-existing mature blood vessels independently of stimulating new blood vessel formation. Vessel co-option, as well as vascular mimicry presents non-angiogenic processes that have been directly visualized in glioblastomas by intravital imaging [56]. Coalescent angiogenesis and vessel co-option are processes of neo-angiogenesis which leads to the fast generation of new blood vessels to increase the transport of oxygen and nutrients independently of angiogenic stimuli. In contrast to the pathological process of vessel co-option, coalescent angiogenesis is a physiological process which does not need invasive tumor cells. Several previous studies have highlighted the usefulness of the CAM model for the study of tumor invasion and metastasis [57, 58].