The vascularization of tissues is accomplished by two distinct processes: vasculogenesis and angiogenesis. In vasculogenesis, vascular tubes develop from endothelial progenitor cells that coalesce and differentiate to form vessels. During angiogenesis, preexisting blood vessels are stimulated to form collateral sprouts, which then invade the surrounding tissue. Angiogenesis accounts for the vascularization patterns characteristic of embryonic brain and kidney, whereas vasculogenesis is more prominent in the formation of large vessels (Folkman, 1995; Risau and Flamme, 1995; Folkman and D'Amore, 1996; Risau, 1997). Vascular development in some tissues, such as the retina, occurs by a combination of the two processes (McLeod et al., 1987, 1996).
Although vasculogenesis and angiogenesis are fundamentally different processes, the end result in both cases is similar: the formation of a vascular wall composed of two principal cell types. Endothelial cells form the inner lining of a vascular tube, which is closely invested with an outer sheath of mural cells. Mural cells can be classified as either smooth muscle cells or pericytes, depending partly on their morphology and density, but also on the type of vessel in which they are found. Vascular smooth muscle cells invest larger vessels, whereas pericytes are associated with arterioles, venules, and capillaries.
The recruitment and differentiation of mural cells seem to depend on signals derived from endothelial cells and may occur by similar mechanisms in vasculogenesis and angiogenesis (Lindahl et al., 1997; Hirschi et al., 1998, 1999; Lindahl and Betsholtz, 1998). In turn, mural cells are important for the regulation and stabilization of developing endothelial tubes (Orlidge and D'Amore, 1987; Antonelli-Orlidge et al., 1989; Sato and Rifkin, 1989; Sato et al., 1990). Cross-talk between endothelial and mural cells is, therefore, critical in both vasculogenesis and angiogenesis (Nehls et al., 1992; Hungerford et al., 1996; Beck and D'Amore, 1997). This cross-talk seems to be mediated by a number of growth factors that are thought to act at various times during vascularization. Angiopoietin (Ang) 1 and 2, basic fibroblast growth factor (bFGF), platelet-derived growth factor (PDGF) AA and BB, and transforming growth factor (TGF-β) are growth factors that are reported to have effects on both the endothelial and mural cell types in developing vasculature (Risau and Flamme, 1995; Lindahl et al., 1997). Cell surface and extracellular matrix components that mediate or modulate cellular responses to these growth factors are, therefore, likely to be important players in vascular development.
The NG2 proteoglycan is a cell surface component that seems to be widely expressed in both vasculogenic and angiogenic neovasculature. During normal development, NG2 is found in large vasculogenic vessels such as the aorta (Grako and Stallcup, 1995), as well as in angiogenic microvessels of the central nervous system (CNS) (Miller et al., 1995; Nishiyama et al., 1996). NG2 is also found in the neovasculature associated with tumors and healing wounds (Schlingemann et al., 1990; Schrappe et al., 1991; Burg et al., 1999). A common theme of these studies is the up-regulation of NG2 in neovasculature and down-regulation of the proteoglycan in quiescent vasculature.
For understanding the role of NG2 in vascular development, it clearly is essential to know which vascular cell types express the proteoglycan. There has been some uncertainty in the literature concerning this issue. For example, NG2 expression in microvessels has been attributed to both pericytes (Schlingemann et al., 1990; Burg et al., 1999) and endothelial cells (Schrappe et al., 1991; Grako and Stallcup, 1995). In the current study, we resolve these discrepancies by describing in detail the spatial and temporal expression of the NG2 proteoglycan during several types of vascular morphogenesis. Our results show that, regardless of the mechanism of vascularization (vasculogenesis or angiogenesis) and regardless of the nature of the vasculature (macro or micro), NG2 is consistently expressed by the mural cell component of developing vascular structures. Therefore, NG2 is quite useful as a marker for identifying and studying the role of mural cells during vascular morphogenesis. In macrovasculature, NG2 is an important supplement to the list of existing markers for smooth muscle cells. In microvessels, NG2 seems to be one of the best available markers for the study of developing pericytes.
NG2 Expression by Cardiomyocytes in the Embryonic Heart
One of the earliest and most prominent locations for NG2 staining in the developing mouse embryo is the heart. Double staining with a number of markers at embryonic day (E) 10 suggests that cardiomyocytes are responsible for this NG2 expression. NG2 is not found in the endocardium, which is identified by the endothelial cell markers CD31 and flk1 (Drake and Fleming, 2000). Instead, NG2 is expressed by cardiomyocytes that are positive for α-smooth muscle actin (SMA) (Fig. 1A,B). Additional details of the NG2 expression pattern in cardiac myocytes were established by comparison with the localization of the atrial and ventricular isoforms of myosin light chain 2 (MLC2a and MLC2v). At this stage of development, MLC2a (Fig. 1F) labels the atrium and inflow tract more effectively than either SMA (Fig. 1B) or MLC2v (Fig. 1D) (Kelly et al., 1999). This method allows us to determine that NG2 expression is minimal in the atrium and the inflow tract but quite pronounced in the ventricle and outflow tract. This pattern of expression is maintained over the next few days of embryonic development, although the intensity of NG2 staining is somewhat diminished at E14 compared with E10.
NG2 Expression by Mural Cells of the Embryonic Macrovasculature
In the dorsal aorta, smooth muscle cells of the tunica media coexpress NG2 and SMA from E10 through E14 (Fig. 2A,B). NG2 is also colocalized with SMA in the tunica media of medium arteries such as the omphalomesenteric artery (E11) and intersomitic artery (E12) (data not shown). There seem to be some instances in which NG2 staining extends outside the borders of SMA labeling (see Fig. 2C). Although this observation still requires further experimental verification, such NG2 labeling may identify immature smooth muscle cells that have not yet begun to express SMA.
The vascular endothelium in arteries is positive for both CD31 and flk1 but negative for NG2, as shown for the aorta in Figure 2D and E. Superimposition of double-stained images consistently shows NG2 staining on SMA-positive smooth muscle cells (Fig. 2C) that lie external to the CD31-positive endothelial cells (Fig. 2F). Examination at higher magnification (60× oil immersion objective, N.A. 1.4) confirms the localization of several layers of NG2-positive smooth muscle cells outside the single layer of CD31-positive endothelial cells (Fig. 2G–I).
NG2 Expression by Mural Cells of the Embryonic Microvasculature
Capillaries expressing the endothelial cell markers can be seen in the hindbrain at E10 (Fig. 3B), followed by the appearance of capillary networks in the midbrain, forebrain, and vitreous cavity (hyaloid vessels) by E12. Confocal microscopy allows visualization of the intimate relationship between the CD31-positive endothelial cells and the NG2-positive pericytes in these CNS capillaries (Fig. 3A–C). When seen in cross-section at higher magnification (60× objective), the abluminal relationship of the NG2-positive pericytes to CD31-positive endothelial cells is even more apparent (Fig. 3G–I). Although the two cellular entities are clearly distinct, they are very closely apposed with an almost continuous investment of the endothelium by mural cells. These pericytes invariably coexpress PDGF β-receptor and NG2 (Fig. 3D–F), but are negative for SMA (data not shown). In contrast to the pericyte/endothelial cell relationship in CNS capillaries, pericyte ensheathment of endothelial tubes in embryonic limb bud (E10) is very discontinuous. Although NG2 and PDGF β-receptor still are coexpressed by pericytes located abluminally to CD31-positive endothelial cells in these capillaries, distribution of the pericyte markers is extremely scattered and patchy, suggesting incomplete pericyte investment of the endothelium (data not shown).
NG2 Expression by Mural Cells in the Vasculature of Newborn Mouse Eye
Microvessels in the postnatal mouse eye exhibit the same intimate relationship of pericytes and endothelial cells seen in brain microvasculature. At postnatal day (P) 7, NG2-positive, PDGF β-receptor–positive pericytes form a continuous investment of CD31-positive endothelial tubes in hyaloid vessels as well as in capillaries of the primary vascular plexus of the retina. This same spatial relationship between pericyte and endothelial cell markers is still seen in the primary (pp), secondary (sp), and tertiary (tp) capillary plexi present in the P11 retina. NG2 is colocalized with the PDGF β-receptor (Fig. 4A–C) but not with CD31 (Fig. 4D–F). The tight abluminal investment of CD31-positive endothelial cells by NG2-positive pericytes is more evident at higher magnification (Fig. 4G–I). Unlike embryonic brain microvessels, the larger retinal vessels contain some NG2-positive mural cells that express low levels of SMA (data not shown). It is not clear whether these cells represent smooth muscle cells or maturing pericytes, but in either case, these SMA-positive cells represent a small percentage of the the overall NG2-positive population at P11.
In contrast, there is extensive colocalization of SMA and NG2 in the tunica media of the P11 ophthalmic artery (Fig. 4J–L), similar to the pattern seen in embryonic macrovasculature. Some NG2 staining is also seen on SMA-negative cells in the tunica intima. These cells may be intimal pericytes that exhibit the same NG2-positive, SMA-negative phenotype observed for microvascular pericytes. Examination at higher magnification confirms that, as in the embryonic aorta, NG2-positive smooth muscle cells and intimal pericytes lie external to CD31-positive endothelial cells (Fig. 4M–O).
Our current survey resolves previous uncertainties about NG2 expression in vasculature, demonstrating that mural cells invariably are responsible for expression of the proteoglycan. This is true for vascular structures that are formed by the process of vasculogenesis, as well as for those that arise through the process of angiogenesis. Thus, NG2 is expressed in the developing heart by cardiomyocytes, in large vessels by vascular smooth muscle cells, and in microvessels by vascular pericytes.
The expression of NG2 by vascular mural cells may prove to be an especially useful tool in the case of pericytes, a cell population for which few reliable markers are available (Nayak et al., 1988; Schlingemann et al., 1996). The identification of NG2 as an effective pericyte marker will be very helpful in studying the development and function of this cell type. Although pericytes are widely regarded to be the microvascular equivalent of smooth muscle cells, the origin, development, and function of these cells seem to be variable and complex (Le Lievre and Le Douarin, 1975; Sims, 1986; Schor and Canfield, 1998; Allt and Lawrenson, 2001) and attempts to use smooth muscle markers for pericyte identification can be an oversimplification that leads to problems. For example, the absence of α-smooth muscle actin (SMA) on NG2-positive cells in CNS capillaries has led some workers to the erroneous conclusion that NG2 is expressed by endothelial cells, rather than mural cells, in this microvasculature (Schrappe et al., 1991; Grako and Stallcup, 1995). Our current results demonstrate that, although SMA is an excellent marker for smooth muscle cells, it is not always an appropriate marker for pericytes in vivo. Although a large percentage of pericytes in culture may develop a smooth muscle phenotype and express SMA, only a small percentage of pericytes in vivo (possibly mature, quiescent cells) appear to do so (DeNofrio et al., 1989; Nehls and Drenckhahn, 1991; Nehls et al., 1992; Boado and Pardridge, 1994). In our study, NG2 is expressed by SMA-negative pericytes both in microvessels and in the intimal layer of larger vessels.
Our studies also show that NG2-positive pericytes express the PDGF β-receptor. This finding is consistent with the identification of the PDGF β-receptor and its ligand PDGF-BB as elements that are critical for pericyte survival and development (Lindahl et al., 1997; Lindahl and Betsholtz, 1998). It should be noted that, during development, both NG2 and the PDGF β-receptor are found outside the vasculature on cell types other than pericytes (Nishiyama et al., 1991, 1996). Nevertheless the combined use of NG2 and PDGF β-receptor along with a bona fide endothelial cell marker such as CD31 allows identification of vascular pericytes with a high degree of confidence.
The use of NG2 as a marker for developing pericytes should lead to an improved understanding of the role of this cell type in vascular development. Two of the observations presented in this study seem noteworthy in this regard. First, pericytes appear to be present during the early phases of capillary development. This finding is especially evident in the CNS, where NG2-positive, PDGF β-receptor–positive pericytes are already evident at E10 in newly formed capillaries of the hindbrain. These capillaries can be no more than a day old and are already fully invested by pericytes. A similar situation exists in other areas of the developing CNS, including the postnatal retina. Between P8 and P11, secondary and tertiary capillary plexi develop as a result of collateral sprouting from capillaries in a primary plexus located at the level of the retinal ganglion cell layer. At P11, these new capillaries are fully invested with NG2-positive pericytes, suggesting that pericyte association with these angiogenic collaterals occurs very soon after their formation. These observations are likely to be relevant to the functional role played by pericytes in the development and maturation of vascular tubes. It is often observed in tissue culture that endothelial cells are capable of forming tube-like structures in the absence of mural cells, and a number of studies have presented evidence suggesting that mural cells play a relatively late role in stabilizing endothelial tubes, partly by inducing quiescence in the endothelial cell population (Orlidge and D'Amore, 1987; Sato and Rifkin, 1989; Sato et al., 1990; Beck and D'Amore, 1997; Hirschi et al., 1999). On the other hand, several studies contend that mural cells play a more active role in the very early formation of the vascular tube, perhaps even participating in the guidance of endothelial cell migration (Schlingemann et al., 1990, 1996; Nehls et al., 1992; Wesseling et al., 1995). Although our observations do not definitively resolve this issue, they do show that pericytes are associated with endothelial tubes at an early point in their development. By using NG2 as a marker, we may be able to perform more detailed developmental studies that will allow us to determine whether NG2-positive pericytes are present in association with nascent capillary sprouts.
Second, the degree of pericyte investment of endothelial tubes is not the same in all tissues. In developing CNS capillaries, we have documented virtually complete coverage of endothelial tubes by NG2-positive, PDGF β-receptor–positive pericytes. In contrast, developing capillaries in the limb bud appear to have only a sparse investment of pericytes. These findings are consistent with previous reports on differences in pericyte/endothelial cell ratios in different types of microvessels, ranging from a high ratio of investment in the CNS to low ratios in skeletal muscle and coronary capillaries (Speiser et al., 1968; Tilton et al., 1979; Sims et al., 1994; Balabanov and Dore-Duffy, 1998). It seems likely that the degree of pericyte investment of the endothelium will play a role in determining the structural and permeability properties of a given microvessel. Thus, extensive pericyte investment of CNS capillaries may contribute to the make-up of the relatively impermeable blood-brain barrier (Balabanov and Dore-Duffy, 1998), whereas sparse investment may be more appropriate in vessels where permeability is an asset. The use of NG2 as a pericyte marker should facilitate correlations between the functional properties of microvessels and the extent of their mural cell investment.
The expression of NG2 by developing cardiomyocytes also presents some opportunities for improved understanding of the development of this cell type. Cardiomyocytes may be likened to vascular smooth muscle cells in terms of both their expression of SMA and their overall pattern of histogenesis (Takahashi et al., 1996). Use of NG2 as a marker may help shed light on the question of how differentiating endocardial and myocardial cells become organized into their respective layers, followed by further morphogenesis of these layers into the discrete chambers of the mature heart. Because NG2 is not uniformly distributed in the myocardium but is more abundant in the ventricles and outflow tract, it provides an additional marker for studying the emergence of diversity among ventricular and atrial cardiomyocytes.
The high level of NG2 expression by vascular mural cells raises the question of the possible functional role of the proteoglycan in vascular development. Although this is still an open question, several properties of NG2 are suggestive of the types of roles it might play in mural cell biology. For example, NG2 functions as a cell surface receptor for some extracellular matrix components, including type VI collagen (Stallcup et al., 1990; Nishiyama and Stallcup, 1993; Burg et al., 1996, 1997; Tillet et al., 1997). Because type VI collagen is a component of the basement membrane in some types of vasculature (Rand et al., 1993; Kuo et al., 1997), NG2 may participate in organization of the basal lamina and/or in mural cell interaction with the basal lamina, both of which are required for vascular development. The interaction of NG2 with specific substrata induces cytoskeletal reorganization that leads to enhanced cell motility (Burg et al., 1997; Fang et al., 1999). NG2 engagement by type VI collagen-containing matrices may, therefore, be an important factor in mural cell migration during vascular morphogenesis.
Mural cell migration and proliferation may also be enhanced by the interaction of NG2 with angiogenic growth factors. NG2 binds tightly to PDGF-AA and bFGF, (Goretzki et al., 1999) and may function as a coreceptor for these growth factors, potentiating their effect on the respective cell surface signaling receptors (Grako and Stallcup, 1995; Grako et al., 1999). In this context, NG2 could be involved not only in the proliferation and migration of mural progenitor cells, but also in their recruitment to the specific mesenchymal phenotype that is required for interaction with endothelial cells during the process of neovascularization (Hirschi et al., 1998, 1999).
NG2 also binds with high affinity to the kringle domains of plasminogen (Goretzki et al., 2000). The interaction of NG2 with plasminogen promotes its activation to plasmin by means of the action of plasminogen activator. Plasmin is thought to be important not only for cell motility, but also for the activation of latent TGF-β, a key mediator of mural cell/endothelial cell cross-talk (Orlidge and D'Amore, 1987; Antonelli-Orlidge et al., 1989; Sato and Rifkin, 1989; Sato et al., 1990). Therefore, mural cell NG2 could be an important regulator of this cross-talk, stimulating the production of active TGF-β as a means of stabilizing developing vascular tubes.
Altogether, our studies suggest that NG2 is well suited temporally, spatially, and functionally to participate in the development of vascular mural cells and in cross-talk between the two major cellular components of the blood vessel wall. Further studies are warranted, not only to study the functional role of NG2 in neovascularization, but also to use the proteoglycan as a marker for studying the development and behavior of mural cells.
Embryos and Postnatal Retina Specimens
C57B1/6 mouse embryos were obtained from timed pregnancies. Noon of the day of discovery of a vaginal plug was designated as day 0 of gestation. At several stages of gestation ranging between embryonic days 10 and 14 (E10 and E14), pregnant females were sacrificed, the uterus was excised, and embryos were taken for processing. Retina and ophthalmic artery specimens ranging from postnatal day 7 to 17 (P7 to P17) were obtained by enucleation of timed postnatal C57B1/6 mice.
The embryo and eye specimens were fixed in 4% paraformaldehyde for 6 hr and then cryoprotected by incubation in 20% sucrose overnight. After snap freezing in OCT embedding compound (Miles, Inc., Elkhart, IN), the specimens were sectioned (15 μm thickness) by using a Reichert cryostat. Sections were air-dried on Superfrost slides (Fisher Scientific, Pittsburgh, PA).
Immunostaining and Confocal Microscopy
Immunostaining of tissue sections was performed as previously described (Nishiyama et al., 1996). Briefly, specimens were incubated overnight at 4°C with primary antibodies diluted in KPBS containing 2% goat serum and 0.1% Triton X-100. After thorough washing, specimens were incubated for 2 hr at room temperature with appropriate second antibodies. After additional washing, specimens were cover-slipped in Vectashield H-1000 (Vector Laboratories, Burlingame, CA) and examined by using a Bio-Rad MRC-1024MP laser scanning confocal microscope.
For the immunofluorescence survey of NG2 expression, affinity-purified rabbit and guinea pig antibodies against NG2 (Burg et al., 1999; Grako et al., 1999) were used. Expression of NG2 in vascular structures was evaluated in double-staining experiments with rat monoclonal antibodies against the mouse endothelial markers Flk1 (VEGF-R2) and CD31 (PECAM-1) (PharMingen, San Diego, CA). Rabbit antibodies against the atrial and ventricular isoforms of myosin light chain 2 (MLC2a and MLC2v), gifts from Dr. Ken Chien (University of California, San Diego, CA) and Dr. Steve Kubalak (Medical University of South Carolina), were used as markers for developing cardiac myocytes (Iwaki et al., 1990; O'Brien et al., 1993; Kubalak et al., 1994). Vascular smooth muscle cells were identified by staining with a fluorescein-labeled monoclonal antibody against α-smooth muscle actin (Sigma, St. Louis, MO). Fluorescein- and rhodamine-labeled secondary antibodies against rabbit, rat, and guinea pig immunoglobulins were obtained from Biosource International (Camarillo, CA) and Chemicon (Temecula, CA). For double-staining experiments, it was necessary in some cases to absorb these antibodies with appropriate immunoglobulins to eliminate unwanted species cross-reactivity.
For identification of pericytes, a rabbit antibody was prepared against the extracellular domain of the PDGF β-receptor, a component known to be critical for pericyte development (Lindahl et al., 1997; Lindahl and Betsholtz, 1998). For generating the necessary receptor fragment, we added a C-terminal his6 sequence to a cDNA clone coding for the mouse PDGF β-receptor extracellular domain (Yarden et al., 1986). This cDNA was ligated into the PCEP/4 vector and transfected into 293 EBNA cells (Tillet et al., 1997), followed by hygromycin B selection to obtain positive colonies. After establishment of confluent monolayers of the transfected cells, the his-tagged receptor fragment was purified from serum-free culture supernatant by chromatography on Ni++-agarose (Qiagen). Authenticity of the purified material was confirmed by amino acid sequencing. Rabbit antisera produced against this receptor fragment were affinity purified on a column constructed by coupling the purified receptor fragment to cyanogen bromide-activated Sepharose CL-4B (Pharmacia). Immunoblotting and immunofluorescence showed that the affinity-purified antibody was reactive against PDGF β-receptor expressed by B28 rat glioma cells, but not against PDGF α-receptor expressed by human MG63 osteosarcoma cells.
We thank Drs. Ken Chien and Steve Kubalak for the gift of antibodies against MLC2a and MLC2v. W.B.S. received support from the NIH. U.O. received support from the AHA.