In Vitro Models of Angiogenesis and Vasculogenesis in Fibrin Gel

Introduction

The assembly of endothelial cells (ECs) into microvascular networks is of interest for myriad reasons. Understanding of the basic biology and signaling required for such assembly provides insights into development and both normal and pathologic angiogenesis. Although in vivo models are highly relevant, it can be difficult to isolate specific phenomena, and animal studies can be quite expensive. Therefore, the development of in vitro models of endothelial assembly is extremely worthwhile. In addition to studying the mechanisms of endothelial assembly, fully developed models would enable the testing of anti-angiogenic drugs for cancer therapy, the study of leukocyte-vascular interactions, and, potentially, the fabrication of vascularized engineered tissues for disease treatment.

It is important to note that another method for creating models of vasculature exists besides EC assembly, which involves forming channels within the scaffold and lining them with ECs.^1,2 Although this strategy is simple conceptually and results in a network of EC-lined tubes that can be easily connected to a flow circuit, the 3-dimensional microfabrication required makes it difficult to form small (capillary-sized) microvessels, and the microvascular patterning must be designed. In contrast, when ECs are allowed to self-assemble into vascular networks, which technically simple (in comparison to microfabrication), they form complex topolgies patterned by nature with lumen sizes of the same order of magnitude as native capillaries. In addition, the mechanisms of microvascular assembly and lumen formation can only be studied by the latter models. For these reasons, models of microvasculature involving the self-assembly of ECs are preferred.

Fibrin has been long used as a matrix for the development of in vitro models of microvasculature. As the provisional matrix in wound healing, fibrin is naturally angiogenic, and growth factors included in the gel formulation are released over a several day period,³ stimulating endothelial assembly longer than if the factors were included in the culture medium. In addition, also likely due to its role in wound healing, fibrin promotes the cellular production of extracellular matrix (ECM) proteins including basement membrane, which is crucial for achieving stable microvascular networks, and collagen I,⁴ which is necessary for tensile strength if one is to construct engineered tissues from biopolymers.

Within endothelial assembly models, several methods exist, which are typically divided into angiogenic or vasculogenic models. Angiogenic models induce sprouting from existing monolayers of ECs, similar to sprouting angiogenesis in vivo. Several techniques have been used, including seeding ECs in a monolayer on a surface of the gel, entrapping EC-coated microcarrier beads within the gel, or entrapping EC spheroids, in which the outer layer of ECs behaves similarly to a monolayer, within the gel.^5-7 Vasculogenic models begin with ECs dispersed throughout the scaffold, which then, under certain conditions, spread and associate into microvessels (linear EC structures with lumens of capillary diameter) and also interconnect to form a network; this is similar to the process of vascular formation during development.^8-10 In both models, “support cells” that stabilize the microvessels are often included.^11-15 Additional details on all of these techniques are discussed below.

Many factors affect the biochemical and ultrastructural properties of a fibrin gel, which in turn affect cell behavior. This likely reflects cellular response to the varied gel mechanical properties associated with these biochemical and ultrastructural properties.^16-18 Therefore, comparisons between experiments can be tenuous. For example, a search of Sigma-Aldrich products yields greater than 10 types of fibrinogen and greater than 10 types of thrombin, from a variety of human and animal sources. This does not include specialty fibrinogen products including plasminogen-free, ultra-pure, and others, yet alone products from other suppliers. In addition, the fibril diameter, porosity, and other ultrastructural gel properties are affected by such factors as the absolute and relative concentrations of fibrinogen and thrombin, calcium ion content, ionic strength (other than calcium), temperature, and pH during gelation. For example, Dubey and colleagues reported fibril diameters ranging from 150 to 510 nm by varying calcium ion concentration from 1.2 to 30 mM.¹⁹ Other studies have indicated that increasing the ionic strength, the thrombin concentration or the pH during gelation results in finer fibrils with a decrease in gel permeability.^20,21 Although a decrease in temperature during gelation below room temperature affected the final gel properties, no differences were observed between gelation at 25°C and 37°C.²⁰ Additionally, the ultrastructure is affected by gel compaction, the extent and rate of which are controlled by numerous factors.^13-15 Finally, any additional components of the gel including the solvents for fibrinogen and thrombin (e.g. saline solutions or basal medium) as well as cytokines, other ECM molecules, or other bioactive reagents can radically affect the behavior of the entrapped cells. Fibrin fibrils can also be aligned if fibrillogenesis occurs in the presence of a strong magnetic field,^13,19 and fibrin gels can be stiffened by dityrosine formation using a cell-compatiable photo-crosslinking method.^22,23 All of these properties can be manipulated to study EC assembly behavior, which makes fibrin an attractive scaffold. However, the range of chemical and structural properties of fibrin gels makes it difficult to compare results across publications, even before the myriad types of ECs and support cells utilized are considered. This review will include details of each model system developed, and discern potential effects of fibrin gel ultrastructure, which may explain some of these differences.

Fibrin gel ultrastructural properties can be examined in a number of ways. Gathering data on gel turbidity relatively simple and informative, as they can give estimates of fibril diameter.^19-21,24 Permeability can also yield information on ultrastructure and fibril diameter, and it has been noted that thinner fibrils have much more effect on permeability than turbidity, so both measures may be important to understand.^20,25 Confocal microscopy of gels made with fluorescently tagged fibrinogen and scanning electron microscopy can also produce visual information on fibril size, density, and topology (e.g. connectivity).^19,21,26 It may be wise for those working with in vitro models of EC assembly to routinely assess the fibrin gel ultrastructure using at least one of these methods.

Endothelial Cell Types

A wide variety of EC types have been used to develop models of vascular assembly. Mature ECs have been the most commonly used, from a range of species sources including bovine, canine, porcine and human. Isolation locations include both large vessels (e.g. umbilical vein, jugular vein, aorta) and microvessels (e.g. dermis). Clearly differences in endothelial biology exist across species, but even within an individual, differences in expression patterns have been observed between arterial and venous ECs,^27,28 and between large and small vessel ECs, even within the same organ.²⁹ Recently, human ECs have been used almost exclusively, because they are relatively easy to obtain and are the most relevant.

Human umbilical vein ECs (HUVECs) are the most widely studied human EC type, and are easily isolated and cultured, which explains their popularity. PubMed lists over 5000 publications on HUVEC isolation, phenotype, and behavior. Over half of the reported studies of endothelial assembly in fibrin gel have utilized HUVECs. Human dermal microvascular ECs (HDMECs) are the second most commonly used cell type in studies of EC assembly, also due to their ease of isolation and culture.

More recently, a variety of stem cell or blood-derived ECs have been studied. For example, several groups have used cells derived from the mononuclear fraction of cord or adult blood. These ECs are isolated via selection for CD31+ cells,³⁰ or by plating in EC-favoring medium and selecting the late outgrowth cells.^13,31 Others have reported the isolation of EC-like cells from human blood using CD34 and CD133 as selection factors,³² but these have not as of yet been used in models of EC assembly. Variation in isolation and culture methods makes it difficult to compare between cell types, but current research indicates that there are two distinct types of endothelial-like cells that can be isolated from blood, termed early and late outgrowth cells, due to the timing of their appearance in cultures of blood mononuclear cells. Early outgrowth cells appear within one week of culture, while late outgrowth cells appear after 2-3 weeks. The two cell types are distinct morphologically and proliferate at different rates (late outgrowth cells have a shorter doubling time), but both groups express a wide range of EC markers.^33,34 However, early outgrowth cells also express hematopoietic markers and do not undergo tubulogenesis in vitro or incorporate into existing networks in vivo; the opposite is true for late outgrowth cells.³⁴ These results suggest that only late outgrowth cells are truly ECs.

Mesenchymal stem cells (MSC) exhibit an endothelial phenotype with certain stimuli, both in vitro^35-39 and in vivo.^40,41 Using vascular endothelial growth factor (VEGF) supplementation and/or endothelial growth medium, several groups have demonstrated induction of endothelial surface marker expression.^35,37,40 Laminar shear stress has been shown to similarly induce endothelial markers,³⁸ with shear stress magnitude being critical for achieving enhanced expression.³⁶ Synergistic effects of shear stress and growth factor stimulation on the progression of endothelial differentiation have also been reported.³⁹

Recently, the differentiation of induced pluripotent stem (iPS) cells into ECs has also been reported, although these have not yet been used in EC assembly models. To induce endothelial phenotype, the iPS cells were cultured in either a cocktail of VEGF and fibroblast growth factor (FGF) or on a feeder layer of mouse fibroblasts, and then were sorted via flow cytometry to purify the population.^42,43

In developing models of microvasculature for the study of microvessel behavior, HUVECs and HDMECs are as useful as any other cell type. However, despite their popularity in research, HUVECs and HDMECs have limited potential to be autologous for disease treatment. Models of microvasculature using the other possible EC sources, especially circulating populations and ECs derived from MSCs, iPS cells, or other stem cells, are most promising for use in engineered tissues. However, preparing models of microvasculature using any of these cell types will require much additional research prior to clinical use, so it is as of yet unclear which cell type(s) will ultimately succeed. For example, endothelial-like cells derived from MSCs have not yet been shown to form microvascular networks; it is possible they are not truly ECs. Also, there is some recent evidence for developmental plasticity between venous and arterial ECs,⁴⁴ which provides additional impetus to continue research on HUVECs.

Support Cell Types

In vivo, capillaries are surrounded by a sparse layer of supporting cells termed pericytes (PCs). PCs are highly mobile and their processes extend to reach many ECs; primary processes extend along the capillary axis while secondary processes extend circumferentially.^45,46 They are often located within the vascular basement membrane, and are directly connected to ECs via tight junctions.⁴⁵ During development these PCs are recruited to the forming capillaries, and this recruitment signals the end of the microvascular “plasticity window,” in which the microvessels freely grow and regress.^46,47 ECs recruit PCs by releasing PDGF-B,⁴⁸ and PC recruitment can be blocked by knocking out PDGF or its receptor.^48,49 The knockout phenotype includes lethality at birth due to microvascular hemhorrage and increased vessel diameter due to unchecked EC proliferation. Interestingly, however, there is no difference in vessel length or branching between knockout and wild type animals.⁴⁹ This suggests that PCs are responsible for regulating microvascular stability, permeability and diameter, but not microvessel patterning. The mechanism by which PCs have these effects is not fully known, but is thought to involve paracrine signaling from both soluble and insoluble factors. PCs release Ang-1, a vessel stabilizing protein, and in the absence of PCs, Ang-1 can improve microvessel stability, suggesting that the PC-derived Ang-1 is partially responsible for the stabilizing effects.⁵⁰ Insoluble factors include type IV collagen and laminin, among other components of the vascular basement membrane. PCs both produce and trigger ECs to produce basement membrane proteins, which are thought to be critical in the maintenance of capillary beds.⁴⁶

In vitro, it has consistently been reported that without the use of support cells microvessels will eventually regress. The support cell types used in the literature vary, including fibroblasts, smooth muscle cells, MSCs, and PCs.^11,12,51,52 The use of these cell types has been met with varying degrees of success, but in general they tend to improve microvessel formation and/or stability. Fibroblasts and MSCs have been shown to behave similarly to PCs in that some of them become recruited to the microvascular niche and promote stability.^11,12 Recent results suggest that the type of support cell used can drastically affect the EC networks that are formed.⁵³

Perhaps not unexpectedly, PCs excel at the role of stabilizing engineered microvessels, as shown in numerous publications from the laboratory of George Davis at the University of Missouri as well as our laboratory. PCs co-entrapped with either HUVECs or blood-derived ECs in either type I collagen or fibrin gel recruited to the forming vessels, promoted the assembly of basement membrane, maintained microvessel diameter within the physiological range, and stabilized the networks for at least several weeks.^{14,15,52,54-56}

As with ECs, the clinical utility of a various support cell types is of concern. Lung fibroblasts and brain PCs have less potential for autologous use than dermal fibroblasts or MSCs. However, dermis may also be a source of PCs,⁵⁷ and the use of support cells derived from other types of stem cells have yet to be explored. Therefore only time will tell which support cell type(s) are the most clinically relevant.

Angiogenic Models

All angiogenic models involve the sprouting of ECs into the gel from a monolayer. However, the monolayer arrangement relative to the gel can vary: the monolayer can be formed on one of the gel surfaces, on the surface of microspheres embedded within the gel, or by the formation of EC spheroids (in which the surface of the spheroid behaves as a monolayer⁷) which are then entrapped in the gel.

The group of Victor van Hinsbergh of the Netherlands⁵ developed a method for inducing sprout formation from HDMECs seeded in a monolayer at confluence on the top surface of a fibrin gel via stimulation with a cocktail of growth factors. This system was used in several subsequent studies, one of which noted that decreasing the pH during gel polymerization created thicker fibrin fibrils and enabled greater total sprout length.^58,59

Volker Nehls of the University of Wurzburg in Germany⁶ developed another angiogenesis assay in which bovine pulmonary artery ECs (BPAECs) were grown to confluence on microcarrier beads and embedded in fibrin gel. Limited sprouting occurred unless fibronectin or pro-angiogenic cytokines were included in the gel formulation. Both the pH of gelation and the ionic strength affected the number of sprouts longer than 150 μm, with increases of either increasing sprout number.⁶⁰ This model was modified by Nehls and others to use other EC types, fibrin densities, and culture media (see Table 1).^61-67

The addition of support cell types to the microcarrier bead angiogenesis model improved sprout formation, although the level of improvement varied by support cell type. Support cell types that have been used with this model include porcine epicardial fibroblasts,⁶¹ NIH 3T3 fibroblasts,⁶¹ glioblastoma cells,⁶⁶ human dermal fibroblasts,⁶² human lung fibroblasts^63-65,67 and human MSCs.⁶³ Recent studies have cultured gels containing HUVEC-coated microcarrier beads in endothelial growth medium (EGM)-2 (Lonza), which is supplemented with a variety of cytokines, yielding vastly increased sprout numbers and lengths.^62-67 Despite the background cytokine levels from medium supplements and serum, support cells appear to produce chemical promoters of EC sprouting. Sprouting was observed to be reduced with increased fibrin density when fibroblasts were cultured in a monolayer on the gel surface, but this reduction was not present when the support cells were dispersed throughout the gel.⁶⁴ A reduction in sprouting was also observed as the separation distance was increased between the ECs and fibroblasts cultured on the gel surface.⁶²

Thomas Korff and Helmut Augustin of the University of Gottingen in Germany⁷ first described the culture of HUVECs as spheroids, in which the outer surface of the spheroid behaves similarly to a monolayer. They subsequently entrapped the HUVEC spheroids in fibrin gel. In the presence of VEGF or FGF-2, sprouts grew from the spheroids into the fibrin. Subsequent publications by others noted improvement in sprout lengths with the inclusion of support cells (dermal fibroblasts or preadipocytes),^13,68 and in some cases the support cells were necessary to achieve sprouting.¹³ Additionally, the pH of fibrin gelation drastically affected sprouting; gels polymerized at pH 6.8 induced single cell migration of ECs from spheroids, whereas gels polymerized at pH 7.0 induced sprout formation.⁶⁰

Vasculogenic Models

In contrast to angiogenic models, in which ECs sprout from existing monolayers, in vasculogenic models, ECs are entrapped in fibrin gel as single cells, which under certain conditions then associate with nearby ECs to form microvessels (linear EC structures containing lumens) that connect to form networks. In some cases, ECs will form networks in the absence of support cells, but typically support cells are required.

The groups of Ernst Reichmann of the University of Zurich in Switzerland⁸ and Dylan Edwards of the University of East Anglia in England⁹ were both able to create vasculogenic systems in which support cells were not needed. In the first case, a small number of HDMECs were entrapped in a relatively high density fibrin gel and were cultured for an extended period of time before network formation (with lumens) was observed. The second system used HUVECs entrapped in low density fibrin gels, and microvessel formation occurred within 3 days in the presence of exogenous growth factors. Although additional differences were present between the systems (see Table 2), it appears that in the absence of support cells, a longer culture time is required if a smaller initial cell number or a higher gel density is used.

Bernhard Frerich and colleagues of the University of Leipzig in Germany¹⁰ developed a vasculogenic model in which adipose stromal cells were included but were prevented from fully interacting with HUVECs because they were adherent to microcarrier beads. The role of the adipose stromal cells was unclear because the authors did not report on conditions in which they were absent; however this system represents an interesting case to explore the function of support cells.

Finally, other groups have entrapped dispersed support cells with ECs, including MSCs (Jan Stegemann/Andrew Putnam, University of Michigan¹¹), human lung fibroblasts (hLFs; Steve George, University of California at Irvine^12,30,69), and human brain PCs (Robert Tranquillo, University of Minnesota^14,15/George Davis, University of Missouri¹⁵). In all cases, a relatively short culture period was required for network formation, and the support cells co-localized with the microvessels. A variety of unique characteristics have been incorporated into these systems such as interstitial flow,^14,69 microvessel alignment,^14,15 and defined medium.¹⁵

Potential Effects of Gel Ultrastructure

Tables 1 and 2 highlight the wide variety of gel components and cell types that have created successful models of endothelial assembly. It is clear that the gel ultrastructure can have a substantial effect on the resulting endothelial structures from studies in which these effects were directly tested. However, it is unclear whether the differences in gel ultrastructure between models can explain the variation in cell densities or cytokine concentrations that yield successful models. Although differences in variables (e.g. EC type, fibrin gel formulation, culture medium) often preclude firm conclusions, the comparison of factors known to affect gel ultrastructure across models may nonetheless explain some of the differences observed.

Among angiogenic models, those used by Steven George/Andrew Putnam and William Broaddus (highlighted in Table 1) are similar enough to compare. Both models used 2.5 mg/ml fibrin gel in which the EGM-2 was included in the gel formulation. HUVECs were seeded on microcarrier beads, and support cells (fibroblasts or glioma cells) were seeded on the top surface of the gel. EGM-2 with aprotinin was used for culture. The only variable known to affect gel ultrastructure that was different between the models was the thrombin concentration (0.5 vs. 1.25 U/ml). All else equal, this difference in thrombin concentration would be expected to produce a 2 fold difference in permeability (lower thrombin concentration yields higher permeability).²¹ A comparison of the results demonstrates longer sprouts (~300 μm) were present in the gels polymerized with 0.5 U/ml thrombin relative to those (~150 μm) present in the gels polymerized with 1.25 U/ml thrombin. These results could be explained by the initial permeability differences of the gels. Of course, other differences in the matrix such as fibril diameter or stiffness could also explain the effects.¹⁶

Another example of potential gel ultrastructure effects can be seen in the model used by George and Putnam described above. Ghajar et al. noted that HUVEC sprout length decreased with increasing fibrin concentration when fibroblasts were seeded on the top surface of the gel.⁶⁴ However, when the fibroblasts were instead dispersed throughout the gel, the sprout length did not vary with fibrin concentration. The authors posited that the fibroblasts produced paracrine factors that diffused more slowly in the denser gels. While this is certainly a possible explanation, another explanation is that the HUVECs needed a relatively high permeability to grow long sprouts, and that the dispersed fibroblasts degraded the fibrin,⁷⁰ effectively increasing the permeability and enabling sprout growth. These explanations are, of course, not mutually exclusive.

A final example of potential effects of gel ultrastructure is apparent in vasculogenic models. The models used by Stegemann/Putnam, George, and Tranquillo are similar in several respects: the gel formulation included cytokines, the culture medium was EGM-2, and support cells were dispersed throughout the matrix. However, different fibrinogen, thrombin and cell concentrations were optimal in each model. Although several possible explanations exist, one possibility involves the differences in initial gel ultrastructure. The model used by George involved both the highest fibrinogen concentration and the highest thrombin concentration, indicating that initial gel permeability was the lowest in this model. The high concentration of fibroblasts found to be optimal in this model may be required to increase the gel permeability via fibrin degradation.⁷⁰ Another difference between models is that a relatively low concentration of HUVECs was used in the Stegemann/Putnam model, in which the lowest thrombin concentration was also used (more than 10-fold lower than the other models). The high gel permeability obtained via the low thrombin concentration may have enabled the HUVECs to assemble into a network more easily.

Despite the ultrastructural differences in fibrin gels that are likely present between model systems, a large number of similarities exist across these systems in the mechanisms of angiogenesis and vasculogenesis. For example, vasculogenesis occurs via the same steps observed in vivo, including cord formation, pinocytotic vesicle formation, and vesicle fusing.⁸ Many other studies have also reported that fibroblasts support angiogenesis and vasculogenesis in fibrin via cytokine release.^61,62,64 A final example is the importance of MT1-MMP in endothelial assembly in fibrin gels, which has been reported by several authors.^9,63

Conclusion

A wide variety of in vitro fibrin-based models of EC assembly exist, spanning both angiogenic and vasculogenic models. In addition to differences in cell types and cytokines used, the variation in fibrin gel ultrastructure that can directly result from changes in fibrinogen, thrombin or calcium concentration, ionic strength, or pH, or indirectly result from gel compaction and fibrinolysis due to the ECs and support cells, make it difficult to replicate results of existing models or compare results between models. Additional studies are needed to directly examine these effects. Nonetheless, fibrin provides a physiological substrate that is manipulatable and conducive to EC assembly, robustly leading to angiogenesis and vasculogenesis across a range of fibrin gel properties.