Manipulating the Microvasculature and Its Microenvironment

Abstract

The microvasculature is a dynamic cellular system necessary for tissue health and function. Therapeutic strategies that target the microvasculature are expanding and evolving, including those promoting angiogenesis and microvascular expansion. When considering how to manipulate angiogenesis, either as part of a tissue construction approach or a therapy to improve tissue blood flow, it is important to know the microenvironmental factors that regulate and direct neovessel sprouting and growth. Much is known concerning both diffusible and matrix-bound angiogenic factors, which stimulate and guide angiogenic activity. How the other aspects of the extravascular microenvironment, including tissue biomechanics and structure, influence new vessel formation is less well known. Recent research, however, is providing new insights into these mechanisms and demonstrating that the extent and character of angiogenesis (and the resulting new microcirculation) is significantly affected. These observations and the resulting implications with respect to tissue construction and microvascular therapy are addressed.

I. INTRODUCTION

Numerous efforts have been underway to assemble tissue-engineered solutions for clinical use, with several under development for wound repair or the repair and/or replacement of cardiac muscle, skeletal muscle, spinal cord, tendons, cartilage, skin, and bladder.^1–5 A majority of the successful efforts have involved thin, sheet-like tissue constructs, such as regenerative patches,⁶ heart valves,⁷ and replacement skin,⁸ likely because of the limits of diffusion in cell-seeded constructs.⁹ In most implants, nutritional maintenance relies on diffusion from the implant site until the implant is vascularized by the circulation of the host.^10,11 This is more difficult to realize in larger tissues in which tissue volumes, even as thin as 300 μm, can develop necrotic cores after 1 week of implantation.^12,13 Strategies for functionally supporting larger constructs have included the implantation of artificial constructs in areas of high vascularity,¹⁴ addition of proangiogenic factors, ¹⁵ or seeding with supportive cells that produced chemotactic signals for recruiting host cells and microvessels. ^16–18 Additional strategies have focused on prevascularizing the tissue space and/or tissue construct through the incorporation of either intact microvessels^19–21 or cultured vascular cells.^16,22–24 These approaches highlight the broad goal of establishing a functional associated vascular supply as a means to support the health and functionality of the implanted tissues.

The primary focus of inducing or constructing a vascular supply has been to form new microvessels; vascularization is increasingly being viewed as one of the primary challenges in tissue engineering of large 3-dimensional (3D) scaffolds.^25,26 However, an effective microcirculation depends on the density of the microvessels, the organization of these microvessels within a network, and the morphology and functional character of the microvessels—in other words, the formation of a microvascular network. A single capillary brings blood into close proximity of an area that is a few cell layers thick. A network of capillaries and other microvessels is necessary to provide sufficient exchange to support an entire tissue bed.^21,27 Therefore, unlike conduit vessels, where a single vessel meets the primary purpose (i.e., the bulk transport of blood), it is the microvessel network that fully meets the needs of the tissue. The successful manipulation of complex 3D vasculatures necessitates control over microvascular morphology and an understanding of the factors involved in this process. This review focuses on tissue vascularization and its regulation, with an emphasis on the development and control of microvessel morphology and organization.

A. Microvasculatures

The vascular system plays an important role in the maintenance of tissue homeostasis by providing oxygen and nutrients to different organs, promoting the removal of metabolic waste, integrating hormonal communication between distant tissues, and allowing for the rapid response of the immune system to distal sites of the body.²⁸ Its main component, the blood vessel, is complex and highly structured. The basic structure of the smallest microvessels (capillaries) is an endothelial cell (EC) monolayer organized as a tube. All microvessels have an integral basement membrane that, depending on the microvessel type,^29,30 consists of laminin, collagen IV, and other extracellular matrix (ECM) molecules.^31–33 In addition to forming the core structure of the microvessel, ECs also mediate blood cell activation, fibrinolysis, and coagulation; blood-tissue exchange; and vascular tone.³⁴ As vessel caliber and complexity increase, perivascular (mural) cells such as smooth muscle cells and pericytes envelop the ECs. Perivascular cells have an important role in vessel stabilization and integrity, and impaired perivascular cell coverage leads to hemorrhagic and hyperdilated vessels, resulting in edema, diabetic retinopathy, and even embryonic lethality.³⁵

Establishing an effective microvasculature or maintaining a pre-existing one involves a complex progression of events involving new vessel growth (angiogenesis) coordinated with vessel adaptation to meet the functional needs of the tissue³⁶ (Fig. 1). In this context, single vessels will grow from existing vessels, persist or regress, change diameter, and/ or modulate the mass of the vessel wall based on environmental cues.³⁷ These processes, often called neovascularization and angioadaptation, involve revising and pruning the network to achieve an efficient perfusion circuit and can depend on the interplay of vascular responses to angiogenic stimuli, the metabolic status of the tissue, and the hemodynamic forces exerted by blood flow.^36–39 Hallmark features of a stable network are morphological and functional heterogeneity among vessels in the network.⁴⁰ This heterogeneity reflects a vessel’s position within the network (e.g., an arteriole vs. a venule) as well as specific cues unique to the tissue environment. Any successful strategy for constructing a new microcirculation should provide a means by which proper network organization and functional specificity can occur.

FIGURE 1.

Neovascular progression in implanted microvascular constructs. A: Progression of implanted microvessels through the stages of angiogenesis and network organization, yielding a new heterogeneous microvascular bed composed of arterioles (a), venules (v), and capillaries (endothelium = green, smooth muscle = red) (Reprinted with permission from Nunes et al.³⁶). B: Schematic summarizing the integrated input stimuli determining neovascularization outcomes. While similar stimuli affect neovascularization at multiple places in the progression, the vascular responses to these stimuli are likely different

B. Microenvironment: Global versus Local

Most tissues contain vascular networks with specific architectures reflecting the varied function of those tissues. For instance, the microvasculature in skeletal muscle is highly ordered, with many of the distal vessels following a specific orientation generally aligned with myocytes.⁴¹ The mechanisms by which vasculatures are directed into a unique architecture are unclear. However, elements intrinsic to the tissue environment, such as growth factor gradients, oxygen gradients, ECM structure, stromal organization, and mechanical force gradients, are most certainly involved. In the broadest sense, there are 2 possible general levels of control of vascular organization: global and local. Global control systems represent a higher-order control that patterns the overall organization of the network within the tissue as well as the overall network boundaries. For example, during embryo development, genetic systems control gene expression gradients across entire tissue beds to organize the tissue and pattern the developing vasculature.^42–45 Studies of tracheal development in Drosophila indicate the existence of a genetic control for at least the primordial branching morphogenesis, which in subsequent stages could be modulated by local feedback signals.⁴⁶ Similar mechanisms have been envisaged in developing vasculature, where an additional differential expression of genes of the ephrin family has been observed in arterial and venous angioblasts even before the advent of blood flow (see Ref. 47 for a review). Whether analogous global systems operate in adult tissues is not known. But genes involved in higher-order regulation in the embryo do play a role in adult angiogenesis.⁴⁸ While signals generated at the global level ultimately act on vascular cells, local control reflects signals generated near or at the vessel wall that provide local instruction. The production of angiogenic factors by local stromal or tumor cells is perhaps the best-known example of local control. However, other important local signals include matrix or substrate composition and properties, microstructural changes, cell-cell interactions, and mechanical forces (including fluid flow).^30,49–54 Local control influences more the vascular organization within a network, such as structural adaptation in response to network hemodynamic forces, vessel densities within the network, and vessel specification.^55,56 In general, global regulation acts to organize and pattern local stimuli, which influence elements of individual vessels in the network.⁵⁷ Our recent studies have clearly demonstrated that a persistent patterning influence present during normal network maturation and remodeling events, probably induced by hemodynamic alterations, can alter the distribution of its morphological subtypes: arteries, veins, and capillaries.^58–60

C. Microenvironment: Complexity

In the native tissue, a variety of integrated stimuli act at the local level to give rise to a functional microvasculature. Autocrine and paracrine growth factors/cytokines, in gradients or otherwise, signal to vascular cells in the context of the structural constraints and mechanical influences of an ECM. In addition, vascular cells physically interact with each other and tissue stromal cells to form the vessel structure and navigate the tissue space. The spatial and temporal control of these varied stimuli in the microvascular microenvironment is critical to the dynamics of vascular form and function. For example, ECs exposed to collagen IV, collagen VIII, and laminin of the microvessel basement membrane contribute to the structure of the vessel wall, whereas exposure of the extravascular space to collagen I is thought to trigger angiogenesis.⁶¹ Also, without the recruitment of a perivascular cell layer to the EC tube via orchestrated platelet-derived growth factor–β signaling, microvessels are dilated, fragile, and prone to hemorrhage.^62,63 Conversely, direct heterotypic contact with mural cells and a concomitant dependence of vascular endothelial growth factor (VEGF) has been reported to be essential for formation of human umbilical vein EC networks.⁶⁴ Such cell-cell interactions may be necessary to determine the angiogenic phenotype, as highlighted in various sections of this review. Finally, the distribution of mechanical forces through a collagen I matrix and the corresponding organization of collagen I fibrils seem to contribute to the guidance of angiogenic neovessels and overall microvessel position.^65,66

The formation of a new, stable microvasculature requires the coordinated balance between those stimuli promoting expansion of a vascular bed (i.e., angiogenesis) and those stimuli promoting vascular stability and maturation. While endothelial cell angiogenic processes are stimulated by a variety of growth factors, VEGF has the broadest impact on the vascular endothelium.⁶⁷ While not fully demonstrated, the different cell responses to VEGF likely depend on the bioavailability, spatial distribution, and gradient structure of VEGF as well as its coincident signaling with other factors.^64,68–71 As a consequence, because of the potent provascular activity of VEGF, a common strategy to promote vascularization of a tissue construct has been to incorporate it into the 3D matrices of the construct. For instance, matrix-bound VEGF has been used to induce ECs into elongated, sprout-like morphologies in fibrin,⁷² and porous poly(lactic-co-glycolic) acid.⁷³ In addition, environmental modifications such as oxygen tension, which regulates VEGF levels, have been investigated with the fibrin-bead system.⁷⁴ Concomitant with responses to growth factors, an EC must respond to the mechanical loading caused by other cells either through direct cell-cell contact or indirect matrix remodeling.⁷⁵ The contact inhibition of EC-EC interaction is a dominant regulating influence, maintaining endothelial quiescence.⁷⁶ Also, EC shape may influence proliferation, and even a linear increase in the area over which cells spread can cause an exponential increase in the DNA synthesis, unlike simple receptor occupancy without any cell extension.⁷⁷ Since EC shape is maintained by a balance between the tension exerted on the focal adhesion sites and the resistance offered by the substrates (tensional homeostasis), it stands to reason that a mechanochemical coupling exists, resulting in additive effects or increased responsiveness.^78,79 Furthermore, cells impart strains onto matrices to facilitate cell growth,⁸⁰ and cell alignment is associated with static and dynamic growth within stressed and unstressed matrices.^32,81,82 In a similar way, the degradation and focal proteolysis of microvessel basement membranes is hypothesized to change the ECM compliance, thereby altering the tensional homeostasis of the ECs with their neighbors, initiating sprout formation inherent in angiogenesis.⁸³ From these limited examples, it is clear that the cellular microenvironment consisting of other cells, growth factors, matrix structure and composition, and the mechanical milieu are critical in determining vascular cell phenotype and function.

II. MICROVASCULAR ASSEMBLY

A common goal for the assembly of vasculatures involves in vitro blood vessel fabrication of both small-and large-caliber vessels. In addition, such constructed vessels and vasculatures can be used to replace both damaged vascular beds and dysfunctional arteries and veins and further our understanding of the mechanisms directing vascular dynamics and remodeling. There are 2 general approaches in microvessel formation: vasculogenesis (the formation of vessel structures by the coalescing of cells)¹³ and angiogenesis (the formation of new vessels from preexisting blood vessels).³¹ Two-dimensional, cultured, cell-based systems generate the cell sources needed and provide a means to investigate relevant responses of cultured microvascular cells (e.g., ECs, smooth muscle cells, and pericytes) to growth factors,^55,84–86 other signals produced by cells,^87,88 cell-cell interactions, ^64,89,90 and combined growth factor–matrix signals. ^91–93 For example, confluent layers of cultured ECs spontaneously form tube-like structures, which invade the underlying matrix and enclose a lumen (vacuolar or otherwise) to form structures resembling capillary sprouts.^30,33,94 While these and similar approaches provide many of the mainstays of the investigative approach into components of the angiogenic process (and microvascular assembly), many of the models suffer from the lack of a physiologically relevant physical environment. However, with the realization that cell phenotypes in isolated cell systems differ between 2-dimensional substrates and 3D scaffolds,⁶¹ many models have incorporated a 3D variation of this tubulogenesis or capillary-like structure– forming approach. In addition, other models involve the use of vascular structures, which comprise the necessary vascular cells already in a native vessel structure.^31,95–97 In all cases, maintaining cell viability and functionality can be difficult, in particular when using human cells. Modification of the vascularizing cells often is required, either through genetic modification to increase resistance to cell death⁹⁸ or in combination with circulating progenitor cells, to improve the viability and behavior of current and future engineered cells.⁹⁹

A. Isolated Cell Systems

The majority of engineered microvessel systems, both in vitro and in vivo, use isolated (and usually cultured) vascular or prevascular cells as the basis for forming new microvessels. Here, vascular precursors are induced to assemble vessel-like structures that can be capable of progressing into mature microvasculatures. The use of isolated cells in vascularizing tissue constructs makes it possible to populate a wide variety of scaffold or matrix systems as well as use flexibility in construct design. The cultured cells typically are placed in an environment that promotes assembly into “capillary-like” structures.¹⁰⁰ A common approach to forming neovessels is to culture vascular cells on polymer beads embedded within an facilitating ECM such as fibrin,^101,102 collagen,¹⁰³ or Matrigel.¹⁰⁴ In these systems, EC chains grow out from the bead-culture. In addition to providing a strategy for assembling new vessels in regenerative applications, such in vitro systems have proven useful as simple angiogenesis assays. A majority of the cells used in this generic approach are vascular cells such as ECs,¹⁰⁵ vascular smooth muscle cells,^106,107 and pericytes,^84,89,108 with some also using stem cells.^109,110 However, applications using only a single type of vascular cell may lack appropriate cellular feedback between the different cells thought necessary to assemble microvessels, ³⁰ prompting the incorporation of perivascular cell precursors with the ECs.¹¹¹ Such combinations of cells have been used, for example, in the engineering of vascularized skeletal muscle tissue to maintain viability, with about half of the total number of vessel-like structures perfused at 2 weeks.¹¹²

B. Isolated Vessel Systems

Different from isolated cell systems, isolated vessel systems use “organ culture” of elements native to the vessel. Suspension of the aortic ring or inferior vena cava segments in a 3D matrix such as collagen is a common strategy in angiogenesis models.^95–97 With this technique, ECs extend from the edges of the excised large vessels, forming capillary-like morphologies that project into the surrounding matrix. As the EC cords develop, the aortic ring assay provides insight into cell recruitment, matrix degradation, cell proliferation, and cell migration.¹¹³ Such assays have been used to evaluate tumor growth and the effects of angiogenic and antiangiogenic treatments.¹¹⁴ The culture or reimplantation of in vivo prevascularized constructs is a variation of such assays. When implanted, ECs extend from such constructs to form cord-like networks and have been shown to inosculate with each other and the host circulatory system.¹¹⁵ In another approach, isolated microvessels—instead of macrovessels—are collected and seeded into collagen I gels.³¹ The characteristic feature of this approach is the spontaneous, in vivo–like sprouting angiogenesis that occurs from the intact microvessels, which serve as parent vessels for the angiogenic sprouts. Once implanted, the nascent neovessels spontaneously connect to the host circulation and remodel to form a perfused microcirculation with proper in-flow and out-flow pathways.^19,36,116 The progression of these implanted neovessels, whether derived from parent microvessels or other precursor sources, into a perfused, hierarchical network highlights the intended outcome of any vascularizing strategy (Fig. 2). Regardless of the source (aorta or microvessel or isolated cells), successful strategies must consider dynamics after angiogenesis to derive a new, functional microvasculature.^117,118

FIGURE 2.

Progression of maturation after angiogenesis in an implanted neovasculature. Vascular casts of progressively longer neovessel implants made from host animals perfused with India ink: Soon after implantation (day 1) (a), limited neovascular network segments are perfused via the host circulation. Days later (day 2, b; day 3, c) nearly all of the immature network is perfused and eventually remodels into a more hierarchical microcirculation by day 28 (d). (Reprinted with permission from Shepherd et al.¹⁹).

C. Bioreactors for In Vitro Vascularization

The onset of flow is a key mediator of microvessel development in vivo.^19,119,120 Therefore, to replicate such development, a method to expose constructs to flow first and then maintain those conditions to propagate network development is required. Simpler experiments circulate media across cell monolayers^121–123 or thin films of cells suspended in a 3D matrix to create maturing shear forces.^124–129 Although capillary-like structures have been identified within these systems, functional networks of heterogeneous diameters are yet to be realized. In more complex biochambers, cells are suspended in media and circulated within a sterile container¹³⁰ or held within hollowed 3D scaffolds sodded with cells.^131,132 Hollow fiber bioreactors also have been used to assess cellular function in the presence of flow.^133,134 While these systems provide insight into cellular behavior in these more complex conditions, they are less conducive to vessel fabrication in vitro. Still more complex bioreactors incorporate perfused, harvested vessels or vessel constructs; engineered blood vessels mimics such as decellularized cadaver vessels,¹³⁵ porcine arteries,¹³⁶ and human saphenous veins¹³⁷; peripheral blood–derived blood vessel mimics^138,139; and synthetic cylindrical polymer scaffolds¹⁴⁰ to generate vessels as small as 4 mm in diameter.

Although promising for large-vessel applications, 4 mm is still orders of magnitude larger than typical arterioles and venules found in the microvasculature. To reduce the size of experimental in vitro vessels, smaller-caliber channels have been formed by drilling cylinders into collagen sponges¹¹⁷ or polymerizing collagen around mandrels of varied caliber.^118,141 After sodding the structures with microvascular cells, the constructs contained vessel-like structures with internal diameters a small as 150 μm. Similar results have been obtained by directly sodding cells onto nylon tubing.²³ In all of these models, sodded cells formed endothelial monolayers on the luminal surface of the perfusion conduits. In many cases, ECs and mural cells migrated out into the surrounding matrix, forming networks of cord-like structures adjacent to the internal lumen.^{23,105,118,141} While endothelial sprouts are capable of invading the surrounding matrix,^117,118,142 mimicking angiogenic behavior,¹⁴³ evidence of network patency and function has not been definitively provided. Therefore, true vascular network development in vitro has yet to be realized.

Toward this goal, bioreactor designs that incorporate a means to provide intravascular perfusion of in vitro neovessels are being explored. An interesting challenge is the drop in scale, often in diameter, in the flow paths needed to perfuse the narrow neovessel segments. Assuming the neovessels are interconnected, appropriately stepping down from the external pump tubing (millimeters in diameter) to neovessels that are less than 10–15 μm in diameter is difficult. One approach has been to use micro-flow systems, such as those in micro-electromechanical systems platforms, to scale down the entire perfusion circuit to the microvascular scale.^144,145 While very useful in a variety of applications, these “chip-based” microvascular systems are difficult to integrate into a vascularized implant. In a different approach designed to accommodate larger tissue volumes (e.g., >3 mL), a macroscale bioreactor designed to incorporate a contrived conduit or “feed vessel” into the 3D tissue construct can be used. Under the appropriate conditions, growing neovessels could inosculate with this feed conduit and, assuming lumen continuity, become perfused.¹⁴³ It is interesting that angiogenesis in this larger system occurred only under what would seem to be suboptimal conditions for oxygen and nutrient delivery,¹⁴³ highlighting the complexities of forming vascularized tissue systems or mimics in vitro.

D. Prevascularization of Implant Sites and Accelerated Biomaterial Vascularization

An alternate strategy to prevascularizing tissue constructs involves implanting avascular constructs into prevascularized tissue sites. This commonly involves the use of implanted growth factor depots to stimulate host angiogenesis at the implant site that subsequently progresses into the construct. For example, incorporation of fibroblast growth factor-2 within 3D matrices, including heparin and alginate,¹⁴⁶ silk fibroin,¹⁴⁷ collagen,¹⁴⁸ and fibrin,¹³¹ has been used to increase neovessel formation and host cell infiltration in a variety of applications such as bone implants,¹⁴⁹ cultured skin,¹⁵⁰ and hernia patches.¹⁵¹ VEGF has been incorporated into similar materials^119,152 and bioresorbable polymers to slowly release it into the surrounding tissues.⁷³ Delayed or controlled release of growth factors can increase the infiltration of host vessels into implanted tissue constructs.^153,154 An alternate approach relies on the targeted placement of vascularizing cells that form vessel structures and/or release angiogenic factors at the implant site preceding construct implantation.^73,155 For instance, a porous poly(ethylene glycol), poly(lysine), and poly(lacticco-glycolic) acid scaffold was sodded with neural progenitor cells and ECs and then implanted into severely compromised immunodeficient mice.¹⁵⁶ After 6 weeks, high densities of tube-like structures were found within the matrix. Similar results were attained by implanting a Matrigel-based EC and 10T1/2 cell construct into severely compromised immunodeficient mice.¹¹⁹ After 4 weeks, neovessels formed within the matrix and were perfused by the host circulatory system. In another example, an implant consisting of Vicryl mesh (90:10 poly[glycolide:lactide]) infiltrated with human fibroblasts secreted a number of growth factors into the surrounding tissue, thereby increasing microvessel density 2-fold.¹⁵⁷ An approach involving the preimplantation of cells can also take advantage of genetic modification strategies to customize factor production at the implant site.¹⁵⁸ Finally, implant sites are being prevascularized through the surgical manipulation of existing vascular beds. In one technique, a graft between the femoral artery and femoral vein of a rat created an arteriovenous loop within a semisealed chamber, in which a functional microvasculature expanded over time.¹⁵⁹ Tissue specific microvessel beds can similarly be excised as whole tissues and processed for subsequent reimplantation.^160–162

E. Microvessel Derivation from Tissue Explants

A variety of strategies for deriving new microvessels in culture from explanted tissues can be used. In these approaches, all necessary vascular precursors (e.g., ECs, murals) are included in the tissue harvest, facilitating the formation of new vessels. For example, in cultured murine myocardial tissue harvests, new interconnected vessel segments arose following 3 weeks of culture.¹⁶² In other approaches, the tissue explant is used to establish the stromal architecture supporting microvascular networks. After decellularizing harvested segments of porcine bowels, repopulating the tissue scaffold with endothelial progenitor cells “re-derived” the microvascular network. ¹⁶¹ In a similar way, placental tissue has been excised, decellularized, and resodded to form in vitro networks.¹⁶⁰ Others have harvested retinal microvasculature and directly perfused them to study microvessel behavior in vitro.¹⁶³ These procedures each provide ex vivo experimental microvasculatures and serve as useful platforms for network investigation and as biologically derived scaffolds for tissue generation, yet these techniques lack the ability to easily modify construct dimension and scale. The systems also require lengthy procedures to harvest tissue and weeks of sample conditioning before the networks are ready for experimental use. While these models retain patent microvasculatures for in vitro perfusion, the networks are organ and tissue specific, with vessel concentration and topology optimized for the original tissue, and offer limited opportunity to direct neovessel growth. Therefore, these systems may not be conducive to the development of varied tissue types. They may also limit potential construct designs by requiring future engineered tissues to be designed around the available vessel topology of the excised network. There may be no ideal in vitro angiogenesis system that is a perfect physiological mimic,¹⁶⁴ but the very recognition of this fact helps in the careful interpretation of results from the various prevalent models.

III. EXTRACELLULAR MATRIX

The ECM is a significant determinant of microvessel form and function. Every aspect of the ECM, including its composition, density, stiffness, fiber orientation and spacing, perceived porosity, and tensional state, all have a regulatory influence on vascular cells. Either as a simple biochemical signal or a spatially structured milieu of matrix-bound factors, the matrix regulates a variety of vascular cellular activities such as migration, proliferation, polarity, maturation, and phenotypic expression. However, the relationship between ECM ligand signaling and mechanical coupling in vascular biology remains incompletely described. It is clear that ECM ligand and cell-receptor binding alone are insufficient to determine microvascular pattern and growth. For a given substrate, cells may be directed to proliferate or differentiate depending on the mechanical attributes. ⁹⁴ Hence, both the ECM structure and its mechanical properties are relevant to development of a 3D vasculature. Regardless, given that initiation, elongation, and guidance of the angiogenic sprout or vasculogenic tube critically involve the spectrum of cellular activities,^165–167 any strategy to manipulate a functional microcirculation via microvessel assembly must consider the ECM.

A. ECM Composition

The composition of the ECM can modulate EC activities (proliferation, monolayer confluence, basement membrane formation, and tube-like structure formation).³³ ECs from adipose tissue tufts cultured on basement membrane collagens (IV, V) aggregated and progressed to form branched, tube-like structures, whereas those on interstitial collagens (I, III) simply proliferated and spread to confluence. ECs showed a similar response on the basement membrane and stromal side of amniotic membranes.³³ Montanez et al.¹⁶⁸ found differences in lumen morphogenesis between Matrigel- and collagen I–based angiogenesis models, which may be related to the difference in stiffness between these 2 substrates, among other factors.¹⁶⁹ Collagen as a substrate for growing vascular cells was pioneered by Ehrmann and Gey¹⁷⁰ in 1956 and, along with fibrin, has since become one of the ubiquitous substrates in the study of ECs and network formation. In addition to collagen I and basement membrane proteins (collagen IV and laminins), other relevant ECM molecules include collagen VIII, a promigratory molecule with low celladhesiveness¹⁷¹ that is associated with rapidly proliferating cells¹⁷²; the proteoglycan Decorin, which is causally involved in formation of EC cords with lumen (sprouts)^173–175; and Tenascin C, which is cytoprotective and promotes EC migration, cord lengthening and branching, and network formation. ^176,177

B. Vascular Cell–ECM Coupling

While not the only mechanism involved, the integrin-cytoskeleton axis is the primary molecular system coupling cells to the ECM. Integrins are multi-domain transmembrane proteins serving a variety of roles in the mediation of cell-ECM interactions.^83,178 Integrins serve as the interface between the intracellular structural framework of the cell, composed of actin filaments and microtubules (i.e., cytoskeleton), and the 3D structural framework of the tissue (i.e., the ECM). It is this interfacial location that implicates integrins in mechanotransduction by not only transmitting external forces to the cell’s structural elements but also via intracellular second messenger systems activated by these perturbations via their binding to specific ligand moieties on the ECM. The integrins are central regulators of vascular cells; deletions of the β₁ subunit lead to embryonic lethality due to vascular defects.¹⁷⁹ ECs treated with anti-integrin antibodies (α-chain specific) reduced proliferation and increased tube formation.¹⁶⁶ Integrin subunit involvement varies with the type of substrate and cellular activity. The α₂ and α₅ subunits are the predominant collagen- and fibronectin-specific receptor complex components.¹⁸⁰ During angiogenesis, the integrin expression profile of ECs changes from α₁β₁ to αvβ₃ (or αvβ₅), which is dramatically upregulated compared to their near absence on quiescent ECs.^{178,181–183} Furthermore, in response to VEGF, integrins α₁β₁ and α₂β₁ are upregulated, thus showing a modulation of cell surface adhesivity.^{178,181–183} Finally, Malan et al.¹⁷⁹ demonstrated that the lack of a β₁ integrin subunit can prevent the deposition of ECM elements such as collagen IV, laminin, and fibronectin in a pattern that either is supportive or indicative of vascular tube-like structures. This reveals a critical feedback between integrins and ECM ligands.

C. ECM Structure and Stiffness

Numerous examples demonstrate that ECM stiffness influences cell behaviors. In general, cells migrate more slowly and spread better on stiffer matrices,^184–186 whereas proliferation is reduced on softer matrices.¹⁸⁴ Fibroblasts cultured on flexible substrates of similar biochemical composition showed increased motility or lamellipodia ruffling but were less spread, as on stiffer substrates.¹⁸⁷ Furthermore, cellular phenotype can depend upon the type of ECM substrate and its stiffness.^188–190 However, phenotypic modulation can depend upon the cell type and other conditions such as cell density. For example, IMR90 fibroblasts (an immortalized fibroblast line) form cords and contract collagen gels, but only in stiffer matrices, whereas primary fibroblasts do so in a matrix of lesser stiffness.⁵⁴ Lo et al.¹⁸⁶ observed that such stiffness-dependent cell phenomena were observed only at lower cell-seeding densities, pointing to a more complex situation involving cell-cell interaction and subtle changes in local substrate mechanics. It is interesting that matrix microstructure is thought to determine, in part, the amount of strain transferred to the cells, thus relating matrix structure and stiffness to the regulation of cell function.¹⁹¹ Cell migration involves directed cytoskeletal contractions and tension that are a function of substrate stiffness.^192,193 Related to this, cell stiffness, presumably due to a concomitant increase in the organization of actin fibers, increases with increased substrate stiffness.¹⁸⁴ Also, integrin-cytoskeletal linkages are strengthened in response to perceived changes in substrate rigidity.^187,194 This linkage strength is dynamic and a function of integrin organization. Such dynamic alterations in focal adhesion complexes support the idea that these complexes are important sensors and regulators of the mechanotransduction process.^{187,189,194–196}

The ability of a substrate of defined stiffness to sustain multicellular retraction can also affect the formation of capillary-like structures (CLSs).94 While relatively low matrix stiffness resulted in rounded cells with poor viability, high matrix stiffness supported normal cell viability but prevented the formation of CLSs, suggesting that an ideal range of matrix stiffness is important for cell aggregation. Both the matrix density and cellular mechanical environment influences the formation and structure of CLSs, reinforcing both the role of the integrin-cytoskeleton axis as well as the possible existence of a balance in the cell forces and matrix stiffness.^{94,182,197,198} In a more complex model of neovessel sprouting, we demonstrated that the normalized dynamic stiffness of vascularized collagen I constructs decreases sharply during sprouting and early network formation but returns to higher values during neovessel elongation and further network growth.¹⁹⁹ It is interesting that the total normalized proteolytic activity in the construct correlated inversely with changes in construct stiffness but the expression of matrix metalloproteases remained elevated.¹⁹⁹ While the changes in matrix stiffness did not seem to affect the integrity of the neovessels formed (which does not involve a vasculogenic process), they did correlate with changes in neovessel activity (sprout formation vs. neovessel elongation and network formation). The potential involvement of matrilysis in changing matrix stiffness, however, agrees well with earlier studies of tubulogenesis or resorption.^200,201 It should be noted that the ECs of the microvascular construct model interact with the interstitial collagen of the surrounding gel as intact, growing neovessels and not as individual cells.³¹ Changes in matrix density and stiffness can also affect EC structure morphology and network and topology. For example, pulmonary microvascular ECs cultured on collagen matrices of different stiffness show remarkable differences in invasiveness and the topology of CLS networks due to pH-dependent polymerization changes.¹⁶⁹ More flexible (i.e., less stiff) matrices promoted the formation of thinner superficial networks with “vacuolar” lumens, whereas stiffer matrices support thicker and deeper networks with true cell-lined lumens.¹⁶⁹ In a similar way, stiffer matrices of fibrin with higher ECM protein concentrations produce less extensive capillary-like networks.^102,202 The dependence of ECs on the matrix structure for 3D assembly is further reflected in the markedly slower rates of tube formation in 2-dimensional (days) versus 3D (hours) matrices (reviewed in Ref. 166).

How matrix stiffness influences vascular cell behavior remains unclear. Evidence from the current literature suggests the existence of a dynamic mechanical balance between the cell and the surrounding ECM. However, Ghajar et al.²⁰² proposed that differences in hydrogel rigidity can influence diffusion kinetics, which in turn can affect network growth. It is also possible that changes in the stiffness or density of a matrix may alter proteolytic activity.²⁰³ Care should be taken when using proteolytic inhibitors to address this since inhibition of matrix metalloproteinases (MMPs) can themselves lead to direct inhibition of angiogenesis.^204,205 Porosity of the matrix environment, a function of matrix density, could also be involved. However, it is likely that porosity plays a permissive role such that the pores are either too small to permit cell invasion or large enough to accommodate a cell (or microvessel) presence. Vascular expanded polytetrafluoroethylene grafts with an intermodal distance of 60 μm readily permitted the microvasculature to grow into the graft wall.^206–208 A membrane with porosities ≤0.22 μm clearly does not permit cell invasion.²⁰⁹

D. ECM Remodeling

While the matrix environment influences vascular cells, the activities of vascular cells also influence the matrix, which can in turn reinfluence the cells. Matrilysis during angiogenesis, particularly by MMPs, is a well-described phenomenon²¹⁰ and likely represents the best known example of matrix remodeling. Recent evidence from a 3D model of angiogenesis correlated the changes in mechanical properties of the tissue environment with total proteolytic activity during angiogenesis, suggesting that matrilysis of the bulk matrix may be important as well.¹⁹⁹ In addition, new matrix production and the deposition of collagens, fibronectin, laminins, glycosaminoglycans, and others occur throughout angiogenesis and neovessel formation (see Ref. 178). A third means by which vascular cells remodel matrix is by restructuring matrix features and components. For example, the spontaneous formation of CLSs from cell monolayers involves the generation of condensed collagen fibrils called adhesive tendrils,⁹⁴ traction fibers,^211,212 or matrix migration pathways,^32,211,212 which are thought to direct cells to migrate and proliferate to form tube-like structures. ⁸² Stopak and Harris²¹³ reported the development of linear tracts of collagen fibers as long as 4 cm by contracting fibroblasts, which were more pronounced in stressed matrices. These changes in collagen structure were due to local contraction of the fibroblasts,²¹³ which can happen as soon as 30 minutes after seeding.²¹⁴ The changes in the lattice structure of collagen fibers in these examples is likely due to simple reorganization of existing collagen since there was minimal new collagen turnover associated with this large structural change.²¹⁴ During angiogenesis in a 3D collagen matrix, collagen fibers are bundled along the wall of the growing neovessel and oriented along the axis of the neovessel ahead of the advancing tip.⁶⁵ While not examined in this study, presumably the restructuring of the collagen lattice was due to vascular cell contractions related to the elongating neovessel, similar to that described for fibroblasts. ECs and fibroblasts actively contract collagen gels and artificial scaffolds.^81,215 Interestingly, contraction of collagen matrices by ECs is inversely proportional to the starting matrix stiffness.¹⁰²

An interesting point to consider is the integration of simultaneous matrilysis and the generation of tractional force that is depending on cell attachment. The matrix protease MT1-MMP and β₁ integrin colocalize on the forward leading edge of cell processes, followed by a trailing proteolytic zone responsible for the actual matrix remodeling.²¹⁶ Transient clusters of MT1-MMP at the leading edge of the cell do not cleave fibrillar collagen (thought to be due to stearic hindrance) and serve as adhesive anchors, whereas similar coclustering of integrin and MT1-MMP leads to active matrilysis.²¹⁶ A similar regulation of MMP-2 activity via a non-RGD (Arginine – Glycine – Aspartic Acid)-dependent binding to integrin αvβ₃ could also be involved in focal matrilysis.²¹⁷ The idea that adhesion and matrix remodeling are coupled is reinforced by the observation that subconfluent and migratory cells generated higher traction than confluent or quiescent cells and that the effect was diminished by MMP inhibition.²¹⁵ This might explain how ECs within a growing angiogenic neovessel can simultaneously exert traction and process ECM. Coordinately, the active process of pulling or compressing matrix molecules as a result of cell-dependent tensional forces may affect matrilysis. Stretching the ECM can alter protease binding and/or cleavage sites of individual matrix molecules, thereby changing the ability of proteases to degrade the molecules.²¹⁸

E. Contact Guidance

Since its first description, contact guidance has been recognized as a determinant of cell locomotion and orientation. In vivo, the oriented ECM fibers and 3D organization of the stromal tissue provide cells with a well-defined substrate for growth and form the basis for contact guidance. Spatially controlled cell adhesion on surfaces is one of the prevalent forms of cell patterning.²¹⁹ Printed guidance channels of ligands such as fibronectin,^220,221 surface etched features like ridges and grooves,^{195,222–224} or engineered fibers^225,226 have formed the mainstay of these orientation strategies. However, while cells in general prefer to grow along such adhesion or contact guidance paths, there seems to be a differential sensitivity to guidance features such as width, groove dimensions, pore morphology, and ligand density.²¹⁹ Central nervous system neuroblasts and peripheral nervous system neurons cultured on patterned surfaces mimicking neurite bundles showed orientation in both directions with a dependence on width and depth.²²⁷ Broadly speaking, substrates with ridges or grooves narrower than the diameter of the cell produce maximal alignment of cells along the substrate features.¹⁹⁵ Alignment of cells along grooved substrates increases with increasing feature depth (200 m to 5 μm)¹⁹⁵ but reduces with decreasing groove width.²²² However, the migration velocity remains unaffected, possibly as a function of the typical sizes of the focal adhesion complexes, and their distribution relative to the ridges and troughs.²²² It is interesting to note that the percentage of aligned cells on all grooved surfaces changes during longer culture periods.¹⁹⁵ Narrow adhesive patterns (about 20 μm wide) were able to constrain migration and orient growing ECs along their long axis better than relatively wider tracks, similar to that seen in ridges or groove-like patterns at confluence.²²⁸ Such cells were maximally elongated on 24-μm track widths with a range of adhesion widths.²²⁹ Although it is difficult to gauge the height of these patterns, it is clear from other studies that even a depth of a few hundred nanometers up to 1 μm is sufficient to generate a patterned migration.²³⁰ Still, it is interesting to note that even at the widest adhesive channel (130 μm), the cells show preferential orientation of their long axis along the pattern’s long axis at confluence.²²⁸

Guided contact of cells involves a balance between intracellular forces, cell adhesion, and the mechanical microenvironment. Harris et al.²³¹ demonstrated that tractional forces of cells vary with cell type; the more invasive and mobile types generate the least traction. In a similar way, EC motility has a biphasic dependence on cell adhesion such that there is an optimal, intermediate adhesion strength that gives rise to maximal migration.^232,233 Cell-cell interactions in higher-density cultures can modulate cell response to substrate stiffness via formation of adherens junctions²³⁴ and likely influence cell stiffness.²³⁵ These relationships reflect the dynamic equilibrium between adhesion density and cellular tension. Ingber⁷⁹ and Chicurel et al.²³⁶ proposed a tensegrity model to explain this equilibrium, whereby forces transduced across integrin adhesion sites are simultaneously distributed throughout the cell cytoskeleton, resulting in adjustments in molecular activity. Contractile cells such as fibroblasts typically generate forces larger than those required to maintain locomotion.²³¹ Perhaps as a result, matrix fibers around the cells become differentially oriented, with fibers at the ends of elongated cells being aligned with the long axis and fibers at the center of the cells being aligned perpendicular to the long axis. Whether the fiber alignment simply reflects a change in matrix structure as part of a force-equilibrating process or serves as contact guidance tracks (or both) is not clear. Aligned collagen fibers, either by fresh deposition by aligned cells or via force-dependent remodeling, facilitate cell-ECM interactions and contact guidance.²²⁴ Whether such contact guidance phenomena influence vessel formation and angiogenesis is unknown. In a monolayer invasion assay, endothelial sprouts grew along collagen fibrils,²³⁷ suggesting a role for contact guidance in directing cell invasion. However, the cells also moved into interstitial spaces beyond the apparent intersections of collagen fibrils,²³⁷ suggesting that contact guidance is not necessary. In strained, 3D angiogenic cultures, neovessel growth was parallel to aligned collagen fibrils,⁶⁶ but because this occurred in an oriented strain field,²³⁸ it is yet unclear if neovessel orientation was due to contact guidance (via the aligned collagen fibrils) or the strain field.

IV. MECHANICAL LOADING

During angiogenesis, EC sprouts and growing neovessels are potentially exposed to both luminal and abluminal mechanical loading. In patent neovessels, luminal forces include fluid shear and circumferential stretch secondary to hemodynamic forces. Abluminal forces arise from vascular cell coupling to the ECM and/or interstitial fluid shear. In the assembly of any microvasculature, these forces and their consequences on the vasculature should certainly be considered. However, the manner in which forces are applied matters to the outcome. For example, shear-induced angiogenesis primarily progresses by intussusception, whereas the stretch-induced or exercise-induced angiogenesis is of the sprouting type with its dependence on proteolysis by MMPs for basement membrane degradation.^61,239–241 We have evidence that blood flow into neovessels is required for the successful progression from an immature network to a mature, hierarchical network.³⁶ Also, hemodynamic forces are critical in determining the caliber and wall morphology of individual microvessel segments within an existing mature network.²⁴² Finally, concomitant application of both circumferential hoop strain and shear stress has a synergistic and potentiating effect on EC alignment.²⁴³ Considerably less is known concerning the effects of interstitial shear and environmental mechanical loading on establishing microvascular form and function, but constraining the deformability of the matrix environment around a forming microvasculature has a profound influence on the topology of the final microvascular network.⁶⁰

A. Tensional Forces

It is well known that static or cyclic stretch can induce cellular orientation either in the direction of or orthogonal to stretch direction.^244–249 Constraining the ability of cell-seeded constructs to contract in a specified direction gives rise to both cell and fiber orientation in a preferred direction. ^{213,214,250–252} Shirinsky et al.²⁴⁵ demonstrated a resistance to deformation by ECs (with elongated morphology oriented orthogonal to strain direction) pregrown to confluence under intermittent strain, whereas monolayers established under stationary conditions showed zones of denudation when subjected to stretch. Many cell types, such as fibroblasts smooth muscle cells, and ECs, orient transverse to the direction of cyclic stretch coincident with the formation of similarly oriented actin fiber bundles.^{247,253–258} Under stressed conditions, reorganization of the actin cytoskeleton precedes cell orientation^245,259 and depends on the reorganization of integrin-based focal adhesions along the axis of stretch.^180,260 The orientation angle relative to the stretch direction depends on the magnitude of stretch rather than strain rate, at least in the early stages of the perturbation.^{247,253,256–258,261} In more complex stress fields, cells and fibers orient circumferentially when allowed to contract axially against radial anchors.²⁵¹ In cell-containing matrices assembled around but not adhered to mandrels, cells and fibers also aligned circumferentially; when the matrix was adhered to the mandrel they aligned radially.^250,252 Cells do respond to compression forces²⁵⁸ but show no specific orientation under equibiaxial planar stretching,²⁴⁷ suggesting that the stimulus for orientation is cell strain (i.e., a change in cell length) and not tension.

It is not clear why cells orient with or against different strain regimens. One of the prevalent hypotheses is that the cells align in an attempt to minimize exposure to large deformations. Eastwood et al.²⁶² suggested that cells orient in the direction of maximum principal strain to minimize exposure to the strain field, especially when adopting a long and thin morphology. Others have proposed a mechanism of stretch avoidance reaction in an attempt to minimize the strain acting on the cells and limit peak-to-peak axial deformation.^253–255 However, such a mechanism would involve thresholds because small strains (e.g., 2% stretch) did not induce a specific orientation.^253,256 Neidlinger-Wilke et al.²⁶¹ noted that the cells typically oriented perpendicular to the stretch direction, such that the resulting axial strains were less than 5–6%, indicating both a level of threshold strain and a window of strain magnitude for an orientation response. Takemasa et al.^257,263 argue that the oblique alignment of the stress fibers in a cyclically strained substrate acts to minimize the amplitude and dampen the oscillations of axial strain felt by the cells. The direction of strain, the strength of the cell attachment, the nature of substrate, 2D versus 3D strain fields, and other boundary conditions clearly contribute the a cell’s response to tensional forces. Unlike individual cells, cellular structures such as microvessels will likely respond differently to applied stress and strain fields. Not all cells in the structure will experience the same degree and character of force, and the large, relatively bulky elements will be less free to reorient. Thus, neovessels growing within a 3D construct experiencing constant isometric stress or strain along a single axis surprisingly aligned parallel to this axis.⁶⁶ They also do this in the presence of cyclic stretch,⁶⁶ unlike single cells, which reorient perpendicular to the direction of cyclic stretch.

B. Mechanical Equilibria

In any tissue, cells are in mechanical equilibrium with their microenvironment.¹⁹⁶ Disturbances in this equilibrium can lead to changes in cell phenotype and tissue character. For example, chronic unloading of tendons leads to an increase in protease production, contributing to tendon fragility. Low-magnitude, low-frequency cyclic loading of tendons removed from the body prevents this protease production.^264–266 In blood vessels, vascular smooth muscle cells, which undergo regular, periodic stretch, will experience hypertrophy and produce matrix following chronic increases in stretch (as in hypertension) (reviewed in Ref. 267). In assays involving the assembly of tube- or cord-like structures, the cells do not go on to form structures without a minimum level of mechanical loading because of the matrix’s resistance to cell contraction.²⁶⁸ In one example, a minimum force, corresponding to about 20% of the endogenous forces generated by the cells themselves, was required to elicit a cellular mechanical response.²⁶⁹ Matrix fibril reorientation and restructuring may represent “mechanical slippage,” causing cells to migrate toward stiffer areas of the matrix.^{185,186,270,271} The establishment of mechanical equilibrium suggests that cells have an ability to sense balanced tensional forces. This is likely intrinsic to the complex and integrated system of adhesion complexes, cytoskeleton, and cellular contractile machinery,^{82,186,213,250,272} which will continually adapt and organize until a mechanical homeostasis is established.^273–275 This may explain why even though fibroblasts orient monotonically with increasing alignment of collagen fibrils, there seems to be a threshold level of mechanical loading that is necessary even in highly oriented fiber matrices.²⁷⁶ Conversely, the relative magnitude of cell traction and adhesion dictates more capillary-like morphology than absolute substrate stiffness.¹⁹⁸

C. Mechanical Forces versus Contact Guidance in Angiogenesis

In nearly all cases, the architecture of the microvascular microenvironment is coupled to mechanical loading. Anisotropic matrices will transduce forces differently than isotropic matrices. Mechanical loading of a tissue or the presence of cells within ECM scaffolds can lead to changes in matrix structure and fibril organization (Fig. 3). Therefore, which of these general stimuli, mechanical loading or contact guidance (via matrix organization), is the dominant influencing factor on tissue and vascular cells is not always clear. As we manipulate the tissue microenvironment to assemble a neovasculature or alter an existing microvascular bed, it is important to understand the contributions of each to the process. While it is difficult to uncouple loading effects from contact guidance, it seems that in general contact guidance strongly influences vascular cell organization. Adhesion-dependent cells certainly require a substrate upon which to attach. As a consequence, it is reasonable to assume that the organization of these substrates will affect the organization of the attached cells regardless of the loading conditions. For example, in a 3D collagen gel containing aligned fibronectin strands experiencing stretch perpendicular to strand alignment, cells not directly in contact with the strands aligned along the direction of stretch (up to 80%), whereas those in contact with strands remained aligned with the strands.²⁷⁷ Interestingly, the cells that remained aligned with the strands increased MMP activity, possibly reflecting an attempt to remodel the matrix to equilibrate the new mechanical loads.²⁷⁷ On grooved surfaces, stretch applied perpendicular to the direction of the groove similarly increases the degree of attached cell orientation along the grooves.²⁷⁸ However, contact guidance stimuli are not necessarily the dominant influence on cell organization. For example, when the direction of fluid flow is not the same as the pattern of features (e.g., flow is perpendicular to the direction of grooves), contact guidance predominates under moderate shear whereas flow direction is predominant under high shear,²²² even though the actin cytoskeleton of the cell aligns in the direction of the grooves regardless of flow direction and extent.²²² In a similar way, ECs cultured on a micro-patterned substrate made up of step changes in collagen density preferentially move to the areas of higher substrate density down the haptotaxis gradient.^185,186 However, application of a suprathreshold fluid shear in the opposite direction causes the cells to move into a region with lower substrate density.^185,279 A similar change in direction of cell migration could be generated by locally deforming the substrate at the leading or trailing edge of the migrating cell.¹⁸⁶ All of this suggests that while the cell prefers to follow the structure of underlying substrate, likely via focal adhesion complexes coupled to an organization cytoskeleton,²⁷⁹ mechanical forces can superimpose some level of control.

FIGURE 3.

Matrix organization and microvascular angiogenesis. A: Filopodial extensions from the tips of 2 sprouting neovessels (green) forming a connection (arrows) coincident with organized collagen fibrils (B), as observed on reflectance confocal imaging. The asterisk serves as a marker to relate neovessel sprouts and collagen fibrils; projected 60× confocal image stacks).

While these examples highlight cell alignment and morphology responses to applied forces and structured substrates, these stimuli influence higher-order cellular organization. For instance, tenocytes cultured in stressed collagen I gels form an external sheath around the periphery, similar to what is observed in a native tendon, despite collagen microstructure.² The development of both parallel and perpendicular cellular processes formed by neuroblasts on microgrooved substrates is reminiscent of the neurite branching and organization of the in vivo tissue.²²⁷ While mechanical forces influence the direction of growth of individual neovessels during angiogenesis,^66,238 they also seem to influence overall network topology.⁶⁰ Specific to the microvasculature, it may be that these different stimuli influence the cells to varying degrees during the angiogenesis and postangiogenesis programs (Fig. 4). For example, as the new sprout forms and advances, the tip cells orient ECM structure⁶⁵ via lamellipodia, and cell traction may lead to progressive alignment of the growing neovessels. The subsequent development of tension and hemodynamic forces related to blood flow would then influence vascular cell phenotype, leading to the formation and adaptation of a functional vascular network.²⁴¹ We are currently examining the temporal sequence of events in neovessel orientation, matrix fibril orientation, and tissue loading to test this hypothesis.

FIGURE 4.

Schematic of the various stimuli influencing sprouting angiogenesis. Maintenance of vessels in quiescence, initiation and control of angiogenesis, and patterning of newly formed microvasculature depend on the integration of several inputs to the endothelial cell machinery that act locally and globally. BM – Basement membrane; ECM, extracellular matrix.

D. Shear (Laminar and Interstitial Flow) Forces

The effects of fluid shear stress applied to the apical surface of ECs are extensive and varied. Exposure to laminar shear stress in culture promotes EC quiescence, alters gene expression, changes thrombotic potential, promotes an anti-inflammatory phenotype, and induces flow alignment.^280–282 Conditions of high shear rates, such as in turbulent or disrupted flow, tend to bring about the opposite effects.²⁸² In subconfluent cultures, migration is directed along the axis of shear stress.²⁷⁹ When cultured on collagen I gels, applied shear stress promotes the invasion of ECs induced by growth factors.¹²⁶ In vessels, shear stress contributes significantly to vascular tone and structural adaptation via EC-dependent mediators. ^56,283 Underlying substrates can alter EC responses to shear stress. For example, the presence of ECM dampened or differentially delayed responses to fluid shear stress.^284,285 Flow-dependent reorientation of ECs requires a concomitant change in the organization of the cytoskeleton.^286–288 In contrast, cells align perpendicular to the direction of large interstitial flows.²⁸⁹ Interestingly, interstitial fluid flow stimulates vascular and lymphatic ECs differently than simple laminar fluid shear alone.²⁹⁰

V. PATTERNING VASCULAR CELLS

Unlike large-caliber vessels, where it is relatively easier to assemble the different layers of the vessel wall,¹⁴¹ the formation of the microvasculature requires assembly of small-caliber vessels (often with walls of only 2 cell layers) into a complex, organized network. The formation of microvessels has generally involved the uncoordinated assembly of cultured vascular cells or vascular cell precursors to form capillary-like cell structures²⁹¹ or cell networks.^127,292 Few, but still successful, strategies have involved using native vessel elements such as microvessel fragments³¹ or aortic rings.²⁹³ While most past efforts have focused on manipulating the microenvironment constituents and cell compositions, ^{98,119,294–296} only recently has there been an interest in spatially directing neovessel assembly and formation in 2 and 3 dimensions. While many vascular cell systems have some ability to self-assemble, the neovessels and early networks that form may not reflect proper vessel wall structure or tissue architecture, delaying proper vascularization or leading to network instability. Prepatterning the assembly of neovessels and networks might then lead to quicker and more successful implant integration and tissue health. In addition to using mechanical and contact guidance stimuli to control vascular geometry, other approaches are being used to physically direct vessel organization.

A. Microelectromechanical Systems–based Blood Vessels

A common approach to creating specific microvascular network geometries involves a combination of classic biological techniques with MEMS. Derived from the microelectronics industry, silicon wafers and photosensitive polymers are chemically manipulated to create planar patterns of channels.²⁹⁷ Fabricated designs range from simple lines or geometric shapes²⁹¹ to complex interconnected networks.²⁹⁸ Features within these patterns can range from nanometers to millimeters but usually reflect typical microvessel widths (5–200 μm). The typical approach involves the creation of a master mold, which is used to cast an elastomer,²⁹⁹ composite polymer,³⁰⁰ or biodegradable³⁰¹ replica of the pattern. By far, the most commonly used polymer in the molds for patterning microvascular cells^291,302,303 is the inert elastomer poly(dimethyl siloxane). However, biological polymers such as collagen,^304–306 laminin,³⁰⁷ and fibronectin³⁰⁸ can also be patterned by a mold generated by a MEMS. Cells then are placed within the patterned channels to form the microvessel elements with defined network geometries. Other MEMS or MEMS-like approaches include spraying cells through precise masks to form cell patterns on collagen films³⁰⁹ and assembling “off-set printed” channel networks and cultured cell monolayers over the patterns.³¹⁰

To manipulate vascular cell phenotype, channel surfaces have been chemically modified with solvents,³¹¹ proteins,³¹² or radiofrequency-generated plasma³⁰² to change surface-cell interactions. Taking advantage of the cellular responses to structured substrates, cell adhesion and growth also have been modified by manipulating surface micro-³¹³ and nanotopography.³¹⁴ In addition, researchers have expanded cell culture techniques by directing individual³¹⁵ and multiple cell types³¹⁶ to grow simultaneously on microfabricated surfaces in an effort to preassemble the microvessel wall. Other groups have worked to generate 3D layers of cells and ECM proteins in an attempt to replicate more complicated tissues, such as arterial walls.³¹⁷ We have used a MEMS approach to pattern angiogenic neovessels within defined geometries.^60,318 Continuing efforts to fabricate more complicated in vitro vessels and vessel networks, researchers have fabricated microchannels that shape biological proteins coated and/or seeded with microvascular cells.³⁰⁵ In these systems, flowing media is used to introduce growth factor gradients, pressure gradients, and shear effects to in vitro cell cultures. EC monolayers as well as networks of capillary-like structures have been formed within collagen hydrogels. Evidence of microbeads lodged within lumen-like structures suggests these nascent structures may form capillary-like structures that may be perfused via nearby microchannels. Similar network structures have been formed in multicellular cultures.³¹⁹ These results suggest that more complicated tissue mimics containing vessels may be possible in the near future.

B. Bioprinting Microvasculatures

In addition to “top down” MEMS microfabrication, a second class of technologies focuses on “bottom up” fabrication and direct incorporation of cells within 3D structures via printing of cells and substrates (“bio-printing”).³²⁰ While MEMS approaches are very effective at forming complex planar systems, it is difficult to build complex patterns in 3 dimensions.³²¹ The first efforts at bio-printing involved inkjet printers to dispense cells onto substrates. ³²² Inkjet cartridges and nozzles, modified to accommodate cells, dispense droplets of polymer suspensions or cell-polymer suspensions onto a surface. In a computer-controlled, layer-by-layer fashion, cells and proteins can be deposited onto various substrates, including glass, collagen, agar, and Matrigel, forming simple geometric shapes.³²³ More complex geometries recently have been printed with multiple cells encapsulated within hydrogels.³²⁴ An alternate system for fabricating 3D geometries is direct-write bio-printing.³²⁵ In contrast to inkjet cartridges and tubing, direct-write bio-printing uses syringes to directly dispense (typically via servo or pneumatic pressure) biologicals into 3D structures. ³²⁶ Similar to solid freeform fabrication, complex shapes are constructed layer by layer. Both direct-write and inkjet bio-printing can organize cells and polymers onto substrates. However, a major advantage of direct-write printing is a greater range of printable viscosities. For instance, we routinely print 40 wt% Pluronic F127 and biological suspensions of 3 mg/mL collagen, which are about 600 × 10⁶ mPa · s. Direct-write printing recently has been used to form individual microvessels. By printing stacking rings of cell “spheroids,”³²⁶ small, cellularized tubes have been constructed. Likewise, linear and branched planar, vessel-like structures also have been fabricated.³²⁷

VI. COMPUTATIONAL MODELING OF THE MICROVASCULATURE

Progress in microvascular engineering has resulted in several strategies for creating microvascular networks, primarily in vitro, and some have focused on imposing topological features. Evidence of vessel maturation, pruning, phenotypic changes to match local vascular beds, and topological adaptations all point clearly to the existence of complex regulators that are highly integrated spatiotemporally and difficult to separate experimentally into cause and effect models.^{36,60,116,328} Thus, computationally modeling these networks can offer interesting insights into these processes and provide new experimental directions to understand the factors involved in both the formation and maintenance of the microvascular bed and the development of novel pharmaceutical targets that influence the process of angiogenesis. In general, models of angiogenesis range in complexities from a focus on cell-level interactions with respect to biochemical variables to macrolevel biophysical dynamics. However, the interdependence of several factors that are considered to be driving forces behind angiogenesis and vascular patterning make this a nontrivial problem both experimentally and computationally. This review addresses only the role of computational modeling in the context of its impact on our understanding of effectors of microvascular topology and morphology.

A. Determinants of Microvascular Branching and Morphology

The development of microvascular branching patterns and network topologies in various tissue beds is not well understood. Over the past decade, several “rules” guiding the growth and maturation of a microvascular network have been put forth. Hansen-Smith³²⁹ proposed that capillary network patterns can be determined by the same mechanisms in all tissues but have tissue- and stimulus-specific regulators. Using symmetric bifurcation of each generation of vessels, with the parent and daughter vessel diameters related by Murray’s law, as a rule has yielded a basic model of branched microvasculature. ²⁴² Other models of vascular network branching and diameters were proposed based on the idea of minimum work.^14,80 In addition, that the vasculatures and their variations can be described by their fractal properties,^8,10 with dimensional values ranging from 1.70 in muscle to 1.89 in the kidney, may be indicative of a conserved and iterative branch morphogenesis mechanism.^8,10,242,330 While there are likely many molecular determinants, which are possibly developmentally preprogrammed,³³⁰ final branching patterns likely arise from secondary stimuli such as matrix orientation, matrix mechanics, tissue oxygen consumption, and hemodynamic influences superimposed on the genetic programs.

B. Angiogenesis

Given the complexity of angiogenesis with neovascular cells continuously responding to and influencing pro- and antiangiogenesis signals, matrix structure and composition, both chemo- and haptotactical gradients, and extravascular cell dynamics, computational models have proven useful in describing the relevant integrated behaviors and making predictions as to underlying mechanisms. Often multiscaled,³³¹ such models (discretized and/ or continuous) integrate biochemical, cellular, and ECM dynamics to predict the extent, direction, and progression of neovessel growth. Such models are particularly effective at simultaneously describing emergent neovessel behaviors across temporal and spatial domains. Many efforts at describing angiogenesis via computational models have focused on tumor angiogenesis, often involving EC migration and secreted angiogenic factors acting as chemotactic agents.^332–337 As more experimental findings are described, the spectrum of computational models simulating neovessel sprouting and behavior have increased (see Refs. 338–340). For example, the behavior of tip and stalk cells within a sprouting neovessel have been developed by building off of the growing experimental evidence describing this interesting dynamic and the signaling axis of VEGF-Notch-DII4.³⁴¹ Also, VEGF dynamics involving both matrix-bound and more freely diffusible forms have been modeled to capture the impact of the varied VEGF regulatory mechanisms on an overall angiogenesis response.³⁴²

Computational models of angiogenesis indicate that the ECM plays a strong role in neovessel sprouting and elongation beyond simply providing a substrate for cell adhesion and migration. In a tumor angiogenesis model involving secreted angiogenic factors and based on a cell-centric approach, matrix dynamics was predicted to significantly affect neovessel branching and anastomosis.³⁴³ In a model of collagen matrix deformation, neovessel alignment and morphology during angiogenesis similarly depended on the extent and axis of strain, which correlated with collagen fibril orientation.²³⁸ An interesting collective prediction from a variety of models is that angiogenesis (and the subsequent postangiogenesis phase) depends on both deterministic and stochastic processes. For example, in a model of network expansion via angiogenesis, an overabundance of new vessels formed stochastically accompanied by network refinement via structural adaptation and pruning through deterministic “rules.”³⁴⁴ More and more, investigators are creating computational models in which angiogenesis-related phenomena are integrated with those related to vessel adaptation and network maturation over a larger temporal and spatial scale.

C. Microvascular Structural Adaptation

Several modeling approaches exist that utilize the relationship between flow shear, circumferential stretch, hoop stresses, vessel diameters, interstitial flow, metabolites and chemokines, and upstream and downstream conductive responses of these stimuli along the microvessels to predict the structural adaptation of networks (see Ref. 345). In an interesting report, Pries and Secomb³⁴⁶ showed that local heterogeneous blood flow and metabolic patterns have only limited dependence on the oxygen demand, suggesting that local metabolic needs adapt to the availability of oxygen. Reglin et al.³⁴⁷ used a novel advection and diffusion–coupled model to enhance the distribution of oxygen in the tissue space, rather than the more simplistic axially distributed models, and demonstrated adequate oxygenation of cardiac tissue for physiologic rates of flow and consumption in spite of assuming a microvasculature more characteristic of skeletal tissue. Godde et al.³⁴⁸ included vessel regression and for the first time simulated the interdigitations between terminal components of arterial and venous trees; they also showed a higher influence of shear-driven dynamics rather than pressure-driven dynamics on the occurrence of these interdigitations and that pressure-driven systems tend to progress toward “arterialization” of the network, with almost total absence of the venous parts. The role of interstitial flow as a significant regulator of the microvascular morphology is gradually being recognized. Even subtle changes in interstitial flow can alter the chemotactic gradient of extracellular molecules in the microenvironment, mainly that of the convection-driven macromolecules and proteins rather than solutes such as oxygen, which are dependent on diffusion.³⁴⁹ In addition to surface fluid shear, interstitial flow also is responsible for changes in the mechanical microenvironment or redistribution of chemokines. Using a fluid dynamics model, Pedersen et al.³⁵⁰ predicted that the perpendicular alignment of matrix fibers to interstitial flow shields cells from forces such as shear and pressure. Jain et al.³⁵¹ used a mathematical model to address the interstitial pressures and fluid velocity of tumors; they reconciled several related observations pertaining to anticancer therapy to explain the relationship between normalization of tumor vasculature and reduced tumor metastasis and other beneficial effects. ¹⁹ These findings are congruous with subsequent conclusions by Pries et al.,²⁴² who demonstrated that subjecting normal microvasculature to parameters of a tumor network led them to exhibit tumor vasculature– like properties.

D. Modeling the Matrix Environment and Cell-Matrix Interactions

Artificial matrices, mostly composed of a hydrogel substrate, have been assigned a biphasic nature comprising a continuous fibrillar architecture interspersed with interstitial fluid.²⁵⁰ The intraphase, viscoelastic nature of such substrates is attributable to a bulk phase fiber-fluid interaction component and an intrinsic viscoelastic component of the fiber-forming protein structure, such as collagen I.^250,352 Using finite element analysis, the cell traction force from substrate deformations can be estimated and used as an input to determine the work done by the cell or its strain energy function, assuming an isotropic linear elastic substrate.^184,186 Many cellular modeling and fiber reorientation studies are based on the results from matrices seeded with fibroblasts,^{77,214,251,252,276,353,354} but ECs may interact differently with the ECM, especially as a part of the tubular microvascular structure. Models of tubulogenesis driven under cell traction fields suggest that the amplitude of strain field is too weak to influence tubulogenesis for conditions leading to mechanical remodeling of the ECM.³⁵⁵ Thus it may be inferred that although directional cell migration occurs in a strain field, it may not be essential in the native tubulogenesis process.^32,82,355

Modeling the interactions of growing microvascular networks and the microenvironment presents significant challenges because of the heterogeneity of cell types, varying vessel morphology, and the need to integrate several competing stimuli into a set of concerted responses. Guilkey et al.³⁵⁶ used a meshless method, called the material point method, to predict the forces and deformations at the growing tip of capillary sprouts. Here, a 3D particle representation of the complex network was created by converting volume-rendered confocal image stacks into discrete particles, with the all relevant information carried on the particles themselves; calculations were performed on a background grid or scratchpad. Using global values for mechanical properties of the construct, the model revealed an inhomogeneous stress distribution around the cellular elements, even in uniaxial tensile loading, demonstrating the efficacy of this method in studying the interactions between microvessels and its matrix environment. More recent models using a “controller” to provide a platform for integration and exchange of parameters between existing modules have been proposed.³³¹ This approach successfully predicted the effects of migration separate from proliferation on the tip cell and stalk cell movement driven both by a VEGF chemotactic gradient and a push-pull relationship between the sprouting tip and the following stalk cells differentially in a temporal event sequence.³⁵⁷

VII. SUMMARY AND FUTURE CONSIDERATIONS

It is now well accepted that manipulation of tissues, either to repair native tissues or build engineered tissues, must include a strategy establishing a functional vasculature. In tissue construction, where there is an absence of existing vessels, this strategy is primarily based on providing the cellular building blocks necessary for assembling/forming vessel elements. Coupled with both diffusible and solid-state factors that stimulate these vessel precursors, many strategies effectively form new vessel elements in a variety of tissue spaces and applications. New investigations into the affect of the microenvironment on vascularization outcomes indicate that the extravascular matrix and cellular environment significantly affects vascular character through processes acting coordinately with the biochemical signals intrinsic to these elements. Matrix microarchitecture, deformation, and mutability all influence the extent of neovessel growth, guidance, and patterning. Because of this growing complexity, many approaches used to understand angiogenesis in constructed and native tissues are combining computational and experimental strategies to discern mechanistic underpinnings and system-wide dynamics.

While much of the focus to date has been on forming new vessels (both macro- and microvessels), it is becoming clear that organization and performance of the vascular network are equally important. In native tissues, vascular organization complements tissue organization, leading to healthy tissue function. Furthermore, proper microvascular function depends on the assembly of individual microvessels into an effective network and perfusion circuit. Much is known concerning the mechanisms of vascular growth. However, further work is needed to understand how new vessels are patterned into a vascular tree, the roles contact guidance and force guidance play in vascular morphology, and the integrative nature of global and local control in structuring vascular beds. As more is learned about the mechanisms underlying not only angiogenesis but also vascular patterning and network adaptation, it should be possible to manipulate the microvasculature to achieve the desired therapeutic outcome.