Vasculogenesis and angiogenesis are the fundamental processes by which new blood vessels are formed (Carmeliet, 2000; Risau, 1997; Risau and Flamme, 1995). Vasculogenesis is defined as the differentiation of precursor cells (angioblasts) into endothelial cells and the de novo formation of a primitive vascular network, whereas angiogenesis is defined as the growth of new capillaries from pre-existing blood vessels (Risau, 1997). In the embryo, blood vessels form through both vasculogenesis and angiogenesis. In the adult, the transient formation of new blood vessels is only observed under certain physiological situations (eg, in the female reproductive tract under control of the oestrous cycle, in the placenta during pregnancy, or during wound healing), and occurs mainly through angiogenesis. Dysregulated angiogenesis has been implicated in the pathogenesis of numerous diseases including vascular retinopathies, rheumatoid arthritis, and cancer (Folkman, 1995). The pioneering work of Folkman and his colleagues has convincingly established the concept that tumor development is dependent upon neoangiogenesis and has paved the way for the identification of several angiogenic molecules, including the fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF) families (Folkman and Shing, 1992). However, the recent characterization of circulating bone marrow-derived endothelial progenitor cells in the blood of adult animals and the demonstration of their incorporation into pathological neovascular foci indicate that vasculogenesis may also participate in pathological neovascularization (Isner and Asahara, 1999). Although major progress has been made during the last decade, our understanding of the molecular mechanisms of these processes is still incomplete.

Gene knock-out experiments have emphasized the pivotal roles played by VEGF, VEGF receptor 2 (VEGF-R2, also called flk-1 and KDR in mice and humans) and VEGF-R1 (also called flt-1) during embryonic vasculogenesis (Carmeliet et al, 1996; Ferrara et al, 1996; Fong et al, 1995; Shalaby, 1995). Heterozygous mice lacking one copy of the VEGF gene die in utero at E10.5, with aberrant blood vessel formation in the yolk sac and the embryo (Carmeliet et al, 1996; Ferrara et al, 1996). Mice lacking the VEGF-R2 gene present early defects (at E8.5) in angioblastic lineages (Shalaby, 1995). VEGF-R1 appears to play a role slightly later in embryonic vasculogenesis, because mice lacking the VEGF-R1 gene produce angioblasts but fail to correctly assemble endothelial cells into functional blood vessels and die at E8.5 (Fong et al, 1995). Thus, during embryonic angiogenesis, VEGF-R2 and VEGF-R1 are required sequentially in that order. In addition, several transcription factors, including ets-1 and Vezf1, have been proposed to play a role in the engagement of angioblasts toward the endothelial cell fate.

Two distinct mechanisms of angiogenesis have been described: sprouting and intussusception. Intussusceptive angiogenesis is caused by the insertion of interstitial cellular columns into the lumen of pre-existing vessels. The subsequent growth of these columns and their stabilization results in partitioning of the vessel and remodeling of the local vascular network (Risau, 1997). Sprouting angiogenesis entails two successive phases: neovessel growth and neovessel stabilization. During the initial phase, the following sequence of events usually occurs: dissolution of the basement membrane of the “mother” vessel and its surrounding interstitial matrix, migration of endothelial cells in this created space toward the angiogenic factor, proliferation of endothelial cells behind the front of migration, lumen formation within the endothelial sprout, and formation of loops by anastomoses of sprouts. The stabilization phase consists of arrest of endothelial cell proliferation, reconstruction of a basement membrane around the neocapillary, and investment and coverage of the immature capillary with pericytes. Both phases are equally important because, in the absence of vessel stabilization, the immature capillary will rapidly undergo apoptosis and regress (Benjamin et al, 1998). Whereas VEGF is important for the growth phase, transforming growth factor (TGF)-β, platelet-derived growth factor (PDGF)-BB, angiopoïetin-1, and their respective receptors are essential for the stabilization phase, as confirmed by both functional knockout of their genes in vivo and a number of in vitro observations (Beck and D’Amore, 1997). Invalidation of the genes encoding these factors and their receptors is lethal between E10.5 and birth as a result of edemas and hemorrhages, indicating that they are essential for vascular bed maturation. Several integrins are also involved in vascular morphogenesis (Bader et al, 1998). αvβ3, which is weakly expressed in quiescent blood vessels and highly increased in angiogenic vessels, is probably an important participant in tumor angiogenesis (Brooks et al, 1994a, 1994b, 1995). However, its role is less crucial during embryonic angiogenesis, as indicated by the viability of integrin β3 null mice (Hodivala-Dilke et al, 1999). Interendothelial adhesion molecules, such as vascular endothelial (VE)-cadherin (Carmeliet et al, 1999; Gory-Faure et al, 1999) and platelet-endothelial cell adhesion molecule (PECAM)-1 (DeLisser et al, 1997; Yang et al, 1999), also appear to be essential to the angiogenesis process.

As a common theme in biology, the action of the pro-angiogenic molecules listed above is counteracted by that of angiostatic factors. This has led to the concept of the angiogenic balance, according to which, overexpression of angiostatic or angiogenic factors controls endothelial quiescence or active angiogenesis, respectively (Iruela-Arispe and Dvorak, 1997). in vitro angiogenesis assays have been highly useful for the identification of a number of these factors and the characterization of their mechanism of action (Auerbach and Auerbach, 1994). Both natural proteins containing an angiostatic motif, such as the thrombospondin type I repeat (Adams and Tucker, 2000), and tumor-secreted proteolytic fragments, such as angiostatin and endostatin (Sage, 1997), have been identified as potent in vitro and in vivo angiogenesis inhibitors and have drawn attention as potential therapeutic agents (Klohs and Hamby, 1999). Several in vitro observations have shown that distinct angiostatic factors commonly act through induction of endothelial cell apoptosis, although via distinct molecular mechanisms (Claesson-Welsh et al, 1998; Jimenez et al, 2000).

Thus, if the activity of key molecules implicated in vascular morphogenesis needs to be evaluated through the gene invalidation approach, much can still be learned about their mechanism of action from in vitro assays (Ashton et al, 1999; Pelletier et al, 2000; Vittet et al, 1997). These usually encompass only one or a few specific steps of the whole morphogenetic processes. They nevertheless allow the distinction between direct (acting via endothelial cell receptors) and indirect (acting through a paracrine pathway) angiogenic factors. For example, cell proliferation can be assessed in vitro by the quantitation of tritiated thymidine incorporation into DNA of capillary endothelial cells or simply by cell counting. Migration assays consist primarily of the use of standard or modified Boyden chambers. Linear endothelial cell migration assays, such as those measuring the repair of a wound made across an endothelial cell monolayer (Ashton et al, 1999; Hoying and Williams, 1996), are also widely used. Proteolysis can be tested using zymographic assays (Pepper et al, 1987). Endothelial cell apoptosis can be measured either by the TUNEL (terminal desoxynucleotidyl transferase-mediated dUTP-biotin nick end-labeling) method or by the quantitation of specific markers such as caspases. Although these assays are very useful reporters for specific steps of the angiogenic process, they do not correctly model the complex interplay of multiple factors that is necessary for vessel development.

In contrast, several integrated assays have been developed that better recapitulate the multistep angiogenic and vasculogenic processes. Moreover, they offer the possibility of evaluating the effects of angiogenic and angiostatic factors at lower cost and experimental complexity than those of in vivo assays.

We will review here in detail the diverse in vitro models of vasculogenesis and angiogenesis that have been described to date and will critically address the advantages and limitations of each of them. Because of space constraints, we have chosen to focus our interest on morphogenesis models, ie, models in which endothelial cells undergo a morphological differentiation process leading to the formation of vascular structures. We have also attempted to highlight unresolved biological questions which may be amenable to such studies.

Vasculogenesis Models

As stated above, vasculogenesis occurs during both embryonic development and adult vascular growth by angioblast mobilization (reviewed in Carmeliet, 2000). Several in vitro systems have been developed for investigating the cellular events of vasculogenesis. Among them, embryo-derived mesodermal cell culture and embryonic stem (ES) cell differentiation assays enable researchers to investigate vasculogenesis virtually as it occurs in the embryo in vivo. Adherent cultures of dissociated cells from quail blastodiscs have been reported to generate both hematopoietic and endothelial cells that aggregate into characteristic blood islands and give rise to vascular structures in long-term culture (Flamme and Risau, 1992; Flamme et al, 1993). Vasculogenesis can also be observed when the same quail blastodisc cells are grown in suspension and form three-dimensional spherules (Krah et al, 1994). However, the formation of vascular structures in quail blastodisc cultures is strictly dependent on the presence of fibroblast growth factor-2 (FGF2) (Flamme and Risau, 1992; Flamme et al, 1993; Krah et al, 1994).

Perhaps the most promising system that has emerged in the past decade is the murine ES cell-derived embryoid body (EB) formation assay. This model system, whereby a primitive vascular plexus is formed (Fig. 1), provides an attractive tool for dissecting the mechanisms involved in the vasculogenesis process: angioblast differentiation, proliferation, migration, endothelial cell-cell adhesion, and vascular morphogenesis can all be evaluated.

Figure 1
figure 1

Schematic representation of in vitro embryonic stem cell technology as a model for vasculogenesis analysis. Embryonic stem (ES) cells, maintained undifferentiated in the presence of leukemia inhibitory factor (LIF) (see cell clumps as indicated by arrows), can undergo genetic manipulation leading to gene deletion or overexpression. Undifferentiated ES cells are induced to differentiate into complex embryoid bodies (EBs) after withdrawal of LIF from the culture medium. The morphology of 6-day- and 11-day-old EBs is shown by light microscopy (Panels A and B). During embryoid maturation, endothelial differentiation and vascular morphogenesis can be monitored. Panels C and D illustrate platelet-endothelial cell adhesion molecule (PECAM) whole mount immunostaining of EBs at Day 6 and Day 11. Arrows point to PECAM-positive cell clusters (Day 6) or vascular structures (Day 11), respectively. Scale bar = 200 μm.

Full size image

ES cells, which are derived from the inner cell mass of mouse blastocysts, are maintained in vitro as totipotent stem cells by culture in the presence of the cytokine leukemia inhibitory factor (LIF). When LIF is removed, the cells spontaneously undergo in vitro differentiation, resulting in the formation of embryo-like structures called embryoid bodies that have the potential to generate all embryonic cell lineages. Blood island formation and many aspects of normal endothelial differentiation and growth, leading to the formation of vascular channels, have been reported during ES-derived embryoid body development (Doetschman et al, 1985; Risau et al, 1988; Wang et al, 1992). Microscopic analysis has revealed that the vascular structures found within the walls of cystic embryoid bodies consist of endothelial cells that form tubular channels with typical endothelial junctions (Wang et al, 1992). In addition, these channels were found to connect cavernous areas that occasionally contain hematopoietic cells, evoking a primitive vasculature (Wang et al, 1992). More recent studies have indicated that endothelial development within ES-derived embryoid bodies follows an ordered sequence of genetic events that recapitulates murine vasculogenesis in vivo and leads to the formation of vascular structures evoking a primitive vascular network (Vittet et al, 1996). Up-regulation of the expression of vascular adhesion receptors such as intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule (VCAM) in response to inflammatory mediators was observed in embryoid bodies, indicating the presence of functional endothelial cells (Heyward et al, 1995). Endothelial differentiation and further vascular morphogenesis was observed regardless of the cell culture procedure used to obtain embryoid bodies: suspension culture (Risau et al, 1988; Wang et al, 1992), culture in semisolid medium (Vittet et al, 1996), ES cell aggregation in hanging drops (Goumans et al, 1999), or the spinner flask technique (Wartenberg et al, 1998). These observations indicate that this in vitro system contains most of the endothelial differentiation program and probably reflects the events taking place during in vivo endothelial differentiation in the embryo, which constitutes a significant technical advance for the study of vasculogenesis in a complex tissue environment.

Indeed, genetic modifications can be easily introduced into totipotent ES cells. The differentiation of genetically modified ES cells, in which gain-of-function or loss-of-function mutations have been introduced, offers excellent alternatives to in vivo studies on transgenic animals to analyze the consequences of specific mutations on the process of vascular development, especially when these mutations are lethal to embryos (Bautch et al, 2000; Schuh et al, 1999; Vittet et al, 1997). Analysis of flk-1−/− ES cell differentiation in vitro has been valuable for determining that flk-1 deficiency does not affect endothelial differentiation but rather impairs subsequent endothelial cell migration and organization into a vascular network (Schuh et al, 1999), issues that cannot be easily addressed in flk-1–deficient embryos. Similar in vitro differentiation experiments performed with heterozygous or homozygous VEGF-A mutant ES cells recently allowed the characterization of the stage-specific differentiation step at which vasculogenesis is blocked because of VEGF-A deficiency (Bautch et al, 2000). ES-derived embryoid bodies may also be useful for the development of genetically manipulated endothelial cell lines carrying gene mutations that are embryonically lethal (Balconi et al, 2000). Indeed, purified endothelial cell progenitors and endothelial cells can be easily separated from embryoid bodies at different maturation steps (Balconi et al, 2000; Hirashima et al, 1999).

In addition, endothelial differentiation in ES-derived embryoid bodies was found to be sensitive to both angiogenic and antiangiogenic agents (Sauer et al, 2000; Vittet et al, 1996; Wartenberg et al, 1998). Thus, the ES/EB model also appears particularly useful for the identification of factors potentially involved in the regulation of angioblast differentiation and further blood vessel formation in a three-dimensional tissue context.

Other experiments, performed by plating embryoid bodies primarily grown in suspension culture for 3 to 5 days onto gelatinized dishes or onto Matrigel, showed the formation of endothelial outgrowths characteristic of sprouting angiogenesis (Bielinska et al, 1996; Zhang et al, 1998). These observations indicate that the ES/EB system can also recapitulate some aspects of the angiogenic process and that this model appears to provide a unique in vitro system to gain further insight into the molecular and cellular mechanisms driving blood vessel formation.

Angiogenesis Models

Angiogenesis was first observed in vitro by Folkman and Haudenschild (1980) 20 years ago. After long-term culture of capillary endothelial cells, these authors observed the spontaneous organization of these cells into capillary-like structures (CLS). The presence of a lumen within these CLS was confirmed by phase contrast microscopy and transmission electron micrography. This report of angiogenesis in a culture dish provided the basis for the definition of in vitro endothelial angiogenesis: all the subsequently published assays referred to the presence of a lumen in the CLS as a criterion for the validation of an in vitro model. From a physiological point of view, an ideal in vitro model would take into account all the representative steps of in vivo angiogenesis, from detachment of endothelial cells from the vascular wall to final tubular morphogenesis, maturation, and connection to a functional vascular network. Furthermore, it should be rapid, easy to use, reproducible, and easily quantifiable (eg, CLS length, area covered by the capillary-like network, number of tubes, and complexity of the network). Depending on the way the cells reorganize, the assays described to date can be roughly classified into two categories represented in Figures 2 and 3: two-dimensional (when the cells develop tubular structures on the surface of the substrate) and three-dimensional (when the cells invade the surrounding matrix consisting of a biogel) assays. Most published in vitro angiogenesis assays are variations of the models listed in Table 1.

Figure 2
figure 2

Schematic representation of two-dimensional assays. A, Two-dimensional short term assay. Capillary-like structures (CLS) develop on the surface of a gel and adopt a planar configuration. B, Two-dimensional long term assay. CLS spontaneously develop on the bottom of the dish after several days or weeks of culture.

Full size image
Figure 3
figure 3

Schematic representation of three-dimensional assays. A, Three-dimensional assays: 1, activated endothelial cells seeded on a gel can invade the underlying matrix and form CLS; 2, vascular explants embedded in gel give rise to capillary sprouts; 3, cells sandwiched between two layers of gels migrate in the surrounding matrix and form CLS; 4, cells attached to microbeads or cell aggregates seeded in gels generate CLS. B, Molecular mechanisms that may be involved during in vitro CLS formation in three-dimensional assays. Note that in most in vitro assays, CLS are rapidly destroyed following step 6, because of the lack of maturation of the CLS.

Full size image
Table 1 In Vitro Models of Angiogenesis and Vasculogenesisa
Full size table

Two-Dimensional Models

Two-dimensional models refer to those in which the planar organization of the cells lies parallel to the surface of the culture plate (Fig. 2; Vernon and Sage, 1995). In such assays, endothelial cells are seeded onto plastic culture dishes that have eventually been coated with adhesive proteins (Feder et al, 1983; Ingber and Folkman, 1989b; Madri and Williams, 1983; Pelletier et al, 2000). Alternatively, they can be loaded on top of a gel made of either collagen, fibrin, or Matrigel (Kubota et al, 1988; Vernon et al, 1995; Vailhé et al, 1997). Many reports state that CLS formation could be observed spontaneously in long-term planar cultures (Feder et al, 1983; Folkman and Haudenschild, 1980). Later, it was observed and well documented in these two-dimensional models that distinct extracellular matrix components can promote CLS formation. Madri and Williams (1983) observed that endothelial cells either proliferate when seeded on Type I or III collagen or differentiate when seeded on Type IV/V collagen. Kubota et al (1988) showed that Matrigel, a laminin-rich matrix, promoted the rapid formation of CLS. Ingber and Folkman (1989b) demonstrated that CLS formation could be induced in planar cultures and modulated by variable substrates, such as fibronectin, collagen IV, or gelatin, depending on the density of the coating. They and others demonstrated that the biomechanical tension between endothelial cells and matrix may serve to regulate capillary development (Davis and Camarillo, 1995; Ingber and Folkman, 1989a; Vernon and Sage, 1995). Based on results obtained with Matrigel and collagen gels, Vernon and collaborators (Sage and Vernon, 1994; Vernon and Sage, 1995; Vernon et al, 1992, 1995) proposed that mechanical forces exerted by the cells onto the matrix drive its reorganization into cords and subsequent CLS formation. We and others observed that, when endothelial cells were seeded on fibrin, they reorganized into CLS (Olander et al, 1985; Vailhé et al, 1997). We showed that the mechanical properties of fibrin were pivotal parameters in inducing CLS formation in this system (Vailhé et al, 1997), because varying the concentration of fibrin within the gels (eg, the mechanical properties) modified the morphological behavior of the endothelial cells. However, fibrin degradation products are angiogenic (Thompson et al, 1985), and the structure of fibrin per se also plays a role (Chalupowicz et al, 1995; Sporn et al, 1995). Thus, it is now recognized that the mechanical signals sensed by cells from the substrate depend on the concentration and biochemical composition of the matrix and regulate the formation of CLS in two-dimensional models. These assays can be subdivided into two groups: short-term (Ingber and Folkman, 1989b; Kubota et al, 1988; Madri and Williams, 1983; Vailhé et al, 1997) and long term (Feder et al, 1983; Folkman and Haudenschild, 1980; Maciag et al, 1982; Pelletier et al, 2000; Vernon et al, 1995) two-dimensional cultures (Fig. 2).

In short-term models, CLS are observed within 1 to 3 days of culture, and cells require subconfluence to form networks (Vailhé et al, 1997). Thus, in such systems, the number of cells seeded (ie, cell density), cell proliferation, and the concentration and biochemical composition of the matrix are determining parameters. Endothelial cells must be seeded sparsely and proliferation should be limited in order to prevent cells from reaching confluence, a condition preventing rapid CLS formation (Vailhé et al, 1997). When necessary (on fibrin and collagen gels for example), cell proliferation can be limited by lowering serum concentration in the culture medium. However, it has been shown by Kubota et al (1988) that when cells are seeded on Matrigel they no longer proliferate. Furthermore, it should be noted that cell migration may not be necessary to form CLS when cells are cultured on gelified basement membrane matrices (Manoussaki et al, 1996). In this specific situation, cellular reorganization is dependent on proteolytic degradation of the matrix (Dubois-Stringfellow et al, 1994; Vailhé et al, 1998) and on cellular forces exerted on the substrate. If the proteolytic activity of the cells is blocked or, conversely, excessive, then CLS formation is inhibited. Thus, one should be aware that short-term cultures (24–48 hr) essentially recapitulate the morphogenesis step of angiogenesis but do not take into account the proliferation and migration steps. Furthermore, once CLS are formed on gels, they often quickly detach from the matrix because of gel disruption by cellular proteases, therefore rendering it difficult to maintain CLS over longer periods of time (Vailhé et al, 1998).

In long-term cultures, the process and the factors involved in spontaneous CLS formation are much less characterized (Vernon et al, 1995); the CLS develop on top of a confluent monolayer of cells (Vernon et al, 1995), and the morphogenic pathway may require cellular synthesis of extracellular matrix (Sage and Vernon, 1994). These models are less convenient for screening the angiogenic activity of molecules than the short-term assays because the CLS are not systematically observed and only a few cells undergo the process of morphological differentiation. Therefore, they lack a strong reproducibility. However, they are more convenient for the observation of stable tubular structures forming slowly over a long period of culture. Iruela-Arispe et al (1991) and Iruela-Arispe and Sage (1993) have developed a method to isolate the cell populations which exhibit this spontaneous tubular phenotype in long-term cultures. One strength of this approach is the development of subcultures which contain mostly angiogenic cells; it allows the specific analysis of cellular and molecular characteristics of activated endothelial cells (Iruela-Arispe and Sage, 1993; Thommen et al, 1997). It is noticeable that in both short-term and long-term two-dimensional models, few coculture systems have been developed. Recently, Pelletier et al (2000) described a long-term model of spontaneous human bone marrow angiogenesis, where pericytes are associated with the development of CLS, allowing the study of heterotypic cellular interactions during tubulogenesis.

Two-dimensional angiogenesis models have thus significantly increased our understanding of the role of extracellular matrix in vascular morphogenesis, but they obviously do not reflect all steps of physiological angiogenesis. Two-dimensional models lack the third dimension; yet, as discussed by Vernon and Sage (1995), the pattern of cellular organization observed in these assays could be representative of intussusceptive rather than sprouting angiogenesis (for a review on the mechanisms of angiogenesis and vasculogenesis, see Carmeliet, 2000; Risau, 1997). Despite this undetermined aspect, two-dimensional models (Table 1) are in general simple, designed from isolated cells, and particularly convenient for screening in vitro activity of angiostatic molecules. They can also be used to assess the ultrastructure of the CLS (Banerjee, 1998; Banerjee and Martinez, 1998), the synthesis of matrix by endothelial cells, or the role of cell adhesion molecules in the tubular morphogenesis process (Gamble et al, 1993). The two-dimensional fibrin model allowed the investigation of the putative role of the hemostatic system on angiogenesis (for review, see Browder et al, 2000). Among all the two-dimensional models, the only one standardized enough to permit satisfactory interlab experimental comparisons is still, in our opinion, the short-term assay on Matrigel. The Matrigel composition is not completely characterized (Zimrin et al, 1995), but it appears highly consistent because the results obtained are highly reproducible. It has allowed the screening of angiostatic molecules (Beckner and Liotta, 1996; Kuzuya and Kinsella, 1994; Morales et al, 1995; Stoltz et al, 1996a, 1996b; Wiederman et al, 1996) and the functional characterization of endothelial cell lines (Bauer et al, 1992; Hughes, 1996). The investigation of the putative role of cell adhesion molecules such as E-selectin (Gerritsen et al, 1996), PECAM-1 (Sheibani et al, 1997), cadherin-5 (Matsumura et al, 1997), and that of proteases (Schnaper et al, 1995) in tube formation has also benefited from this assay. Extracellular protein synthesis, vessel maturation (Haralabopoulos et al, 1994), the role of glycation products in diabetes (Yamagishi et al, 1997 and 1999) and that of matricellular proteins in angiogenesis (Dawson et al, 1997; DiPietro et al, 1994) have been studied as well using the Matrigel assay.

Three-Dimensional Models

Three-dimensional angiogenesis assays are based on the capacity of activated endothelial cells to invade three-dimensional substrates (Fig. 3). The matrix may consist of collagen gels, plasma clot, purified fibrin, Matrigel, or a mixture of these proteins with others. The culture medium may be added within the gel before polymerization or on top of the gel.

It is possible to embed and culture vascular explants such as aortic rings in gelified matrices and then to observe endothelial cells sprouting from the intima and forming CLS (Nicosia et al, 1982; Nicosia and Ottinetti, 1990a, 1990b). The model developed by Nicosia and its variations are still extensively used (Arthur et al, 1998; Brown et al, 1996; Hoying et al, 1996; Zhu et al, 2000). They closely fulfill the optimal conditions for an in vitro model because they allow the preservation of the vessel architecture during the in vitro assay and thus are close to an “ex vivo” model. They can be adapted to quantify angiogenic or angiostatic activities (Nissanov et al, 1995) or to study extracellular matrix reorganization during morphogenesis (Nicosia and Madri, 1987). However, the dissection of the precise role of each cell type (fibroblasts, pericytes, smooth muscle cells, endothelial cells) in the sequence of events leading to tubular morphogenesis is not easy.

Alternatively, one can observe the morphogenic response of isolated endothelial cells that have been seeded on or in gels. When confluent cells cultured on gels are stimulated by cytokines such as basic fibroblast growth factor (bFGF) or by phorbol esters, they invade the underlying gel and form CLS, switching from a planar confluent configuration to a differentiated three-dimensional one (Montesano and Orci, 1985; Montesano et al, 1986). Cells can also be directly overlaid by gels (Schor et al, 1983) or sandwiched between two gel layers before the polymerization (Chalupowicz et al, 1995; Montesano et al, 1983). In other assays, cells are seeded in gels before polymerization, either dispersed (Bayless et al, 2000; Madri et al, 1988), clustered as spheroids (Korff and Augustin, 1999), aggregated (Nicosia et al, 1986; Vernon and Sage, 1999), or attached onto microcarrier beads (Nehls and Drenckhan, 1995a, 1995b). The three-dimensional models are closer to the in vivo environment than the two-dimensional ones, because they take into account more steps of angiogenesis. In fact, depending on the culture media composition (percentage of serum, addition of cytokines), cells can be induced to sprout, proliferate, migrate, or differentiate in the 3-D configuration. Because the biogels are polymers, the concentration and the biochemical conditions of the matrix polymerization must be carefully defined because they may affect the density and the mechanical properties of the substrate (Ferrenq et al, 1997), leading to either proliferative, migratory, or tubular endothelial cell phenotypes (Nehls and Herrmann, 1996). Furthermore, as discussed above with the two-dimensional models developed on gels, proteolysis of the matrix must be controlled (Montesano et al, 1987) and the use of exogenous antiproteases may be required to limit gel degradation (Zhu et al, 2000). From this point of view, one clear advantage of the collagen matrix, compared with fibrin, is that it offers a better resistance to endothelial cell proteases, providing a better substrate than fibrin for long-term observations. Three-dimensional models have provided great advances in the understanding of angiogenesis. In particular, Montesano et al (1986) observed that bFGF induced a confluent monolayer of endothelial cells to form CLS in collagen gels, demonstrating the in vitro angiogenic properties of this factor. In addition, they showed that phorbol ester induced cellular invasion of collagen (Montesano and Orci, 1985) or fibrin matrices (Montesano et al, 1987) and characterized the synergistic properties of bFGF and VEGF (Pepper et al, 1992) as well as the biphasic effect of TGF-β1 on angiogenesis (Pepper et al, 1993). Three-dimensional models thus appear particularly appropriate for studying the effects of cytokines (Montesano et al, 1994; Montesano and Pepper, 1998; Pepper et al, 1994, 1996), the role of metalloproteases (Trochon et al, 1998), and that of the fibrinolytic pathway (Dubois-Stringfellow et al, 1994; Iwasaka et al, 1996; Koolwijk et al, 1996; Kroon et al, 1999; Lu et al, 1996; Van Hinsbergh et al, 1997) during tubulogenesis. Furthermore, they allowed the study of apoptosis (Korff and Augustin, 1998; Kuzuya et al, 1999; Schönherr et al, 1999) and showed the importance of the configuration and composition of the substrate (Nehls and Herrmann, 1996), the role of cell adhesion molecules (Bach et al, 1998; Bayless et al, 2000; Davis and Camarillo, 1996; Gamble et al, 1993; Trochon et al, 1996; Yang et al, 1999), and the effect of hypoxia (Phillips et al, 1995). They were also instrumental in the screening of angiogenic and angiostatic molecules (Bouloumié et al, 1998; Clapp et al, 1993; Koblizek et al, 1998; Papapetropoulos et al, 1997, Vasse et al, 1999).

Another important parameter can be investigated under the three-dimensional configuration: the bioavailability of angiogenic factors. The distance between the cells and the culture medium generates a gradient of diffusion of nutrients, oxygen, and stimulating factors. This is the case when the cells are seeded inside a gel (or sandwiched between two layers of gels) that is subsequently covered by culture medium. This gradient may represent what occurs during angiogenesis in vivo. Helmlinger et al (2000) recently demonstrated, on the basis of a sandwich system of collagen gels, that paracrine VEGF-induced morphogenesis of HUVECs depends on the distance of HUVECs from the edge of the sandwich culture. The cells retain their monolayer configuration at a distance of 0 to 2 mm from the edge of the sandwich, whereas a cell network is fully formed at the most hypoxic inner side of the sandwich (10–12 mm from the edge). This example shows that the three-dimensional configuration most completely models events occurring during angiogenesis in vivo: cell proliferation, migration, and tubulogenesis upon a gradient of nutrients and cytokines.

Future Perspectives

The first well-defined in vitro models of vascular morphogenesis were described in the eighties (Table 1). However, as previously discussed by Jain et al (1997), the origin and passage number of endothelial cells, the nature of the substrates (extracellular matrices), the angiogenic agents, and the levels of endotoxin have not been standardized enough to permit quantitative interlab comparison of these in vitro assays. Thus, one important goal would certainly be to standardize these assays in a more accurate way. To develop models which reflect more closely the in vivo situation is another important goal for the future. Recent data (Helmlinger et al, 2000; Korff and Augustin, 1999; Pelletier et al, 2000; Vernon and Sage, 1999) show that development of in vitro models remains an active and promising approach for identifying factors involved in vascular morphogenesis (Kahn et al, 2000). The availability of genetically modified cells from transgenic or knockout mice, and the preparation of cells transfected with inducible vectors should bring a new potential to these models; it is now possible to use these genetically modified cells to investigate in vitro the precise contribution of a gene to the vasculogenesis/angiogenesis process (Balconi et al, 2000). Conversely, gene expression profiling can be studied during in vitro formation of tubelike structures, providing some new insight into the potential role of new molecules and mechanisms in angiogenesis (Kahn et al, 2000; Pröls et al, 1998). It is remarkable to consider that most of the research to date has been focused on the elucidation of the mechanisms involved in vascular morphogenesis per se, but very little on the mechanisms implicated in the involution of the capillaries. Further investigation in this area may also give rise to useful new information. The model developed by Nicosia (Nicosia et al, 1982; Nicosia and Ottinetti, 1990a, 1990b) is, to our knowledge, the only available morphogenesis model adapted to the study of regression, remodeling, and involution of capillaries similarly to the context of wound repair (Zhu et al, 2000). Recently, Schechner et al (2000) described a very interesting method for grafting in vivo CLS which were preformed in vitro in collagen/fibronectin gels. Such “ex vitro” techniques are very promising for the study of vascular remodeling and the grafting of genetically modified capillaries.

In conclusion, in vitro vasculogenesis and angiogenesis models will undoubtedly benefit from the recent availability of cells derived from genetically modified animals and up-to-date techniques for studying the mechanism of action of factors regulating vascular morphogenesis.