In embryos, blood vessels form through a process that combines vasculogenic growth of early vessels, connection to circulation, and angiogenic growth (Wilting et al., 1995). Vasculogenesis is a multistep process in which endothelial precursor cells or angioblasts coalesce into clusters of cells that fuse to form primitive vascular endothelial networks (Drake et al., 1997). Angiogenesis is the process by which new endothelial tubes grow or sprout from existing mature blood vessels (Coffin and Poole, 1988). Whichever process gives rise to endothelial tubes, they, in turn, recruit smooth muscle and pericyte precursors to form vascular media and adventitia in larger vessels (Lindahl and Johansson, 1997; Dettman et al., 1998). In the developing heart, the coronary vessels have been observed to form within a transitory mesenchyme formed between the superficial epicardium and myocardium (reviewed in Morabito et al., 2002; Reece et al., 2002). Coronary vascular precursors migrate to the heart from extracardiac regions. Angioblasts are believed to migrate to the subepicardial mesenchyme from the liver primordium through the proepicardial organ. Smooth muscle and pericyte precursor cells migrate from the proepicardial organ to the epicardium and then undergo epicardial–mesenchymal transformation to form the subepicardial mesenchyme (Pérez-Pomares et al., 1997, 1998a; Dettman, et al., 1998; Gittenberger-de Groot et al., 1998).
The linear heart is avascular (Rychter and Ostádal, 1971), consisting of two epithelial cell layers, myocardium and endocardium separated by cardiac jelly (de Jong et al., 1987). During looping, the atrial and ventricular myocardium form and begin to proliferate at the outer curvature of the heart (Van den Hoff et al., 2001). The increased size of the beating myocardium requires enhanced oxygen delivery. To allow nourishment of the postlooped heart, the endocardial endothelium becomes progressively trabeculated, establishing a sinusoidal system that minimizes diffusion distances. With development of the compact myocardium, this sinusoidal system becomes insufficient and the coronary vasculature develops to maintain sufficient delivery of oxygen and nutrients to the myocardium.
Evidence from morphological analysis and cell lineage tracing experiments supports the hypothesis that the coronary vasculature arises by a vasculogenic process (Mikawa and Fischman, 1992; Mikawa and Gourdie, 1996). Embryonic formation of the coronary vascular bed is dependent on epicardial development (reviewed in Männer et al., 2001; Morabito et al., 2002). The epicardium is derived from the proepicardial organ, an outpouching of the septum transversum. The proepicardial organ attaches the dorsal aspect of the atrioventricular sulcus at chick stage HH16 (embryonic day [E] 2 or mouse E9.5; Virágh et al., 1993). After attachment, epicardial cells proliferate and migrate from the proepicardial organ cephalad to envelop the heart from the sinus venosus to the outflow tract. It has been well demonstrated in mice, by inactivation of genes encoding integrin alpha 4, VCAM-1, Wilms tumor-1, and RXRα that failure of epicardial development produces abnormal coronary vessels (Sucov et al., 1994; Kwee et al., 1995; Yang et al., 1995; Moore et al., 1999; Sengbusch et al., 2002). In these models, abnormal coronary artery development can be explained because the epicardium is an essential source for coronary vascular smooth muscle cells (CVSMC), pericytes, and intermyocardial fibroblasts, all derived through epithelial–mesenchymal transformation of the epicardium (Pérez-Pomares et al., 1997, 1998a; Dettman et al., 1998; Gittenberger-de Groot et al., 1998), a process that is both positively and negatively regulated by growth factors secreted from the myocardium (Morabito et al., 2001).
Coincident with the migration of the epicardium, endothelial cell precursors begin to colonize the subepicardial space (Mikawa and Fischman, 1992). Endothelial cells of the coronaries originate from an extracardiac source, probably from endothelial precursors residing in the liver sinusoids and the sinus venosus (Poelmann et al., 1993). Another possible source of endothelial cells in the subepicardial space is from epicardially derived mesenchyme, which has been proposed to differentiate into endothelial, smooth muscle, and perivascular fibroblast cells based on local inductive cues (Pérez-Pomares et al., 2002b). Endothelial cell precursors coalesce into a vascular plexus in regionally restricted patterns (Vrancken Peeters et al., 1997a). Once formed, the nascent coronary plexus invade the aorta to establish forward coronary perfusion (Bogers et al., 1988a, b, 1989; Waldo et al., 1990; Bouchey et al., 1996). Only then does blood flow in antegrade direction through the coronary arteries.
Although many studies have touched on aspects of this developmental process (Vrancken Peeters et al., 1997a, b; Ratajska et al., 1999, 2000), no comprehensive chronological description of the development of the subepicardial coronary vasculature has been reported. For this study, we used the embryonic quail heart to describe how the endothelial, hematopoietic, and smooth muscle cell precursors assemble to form the coronary vascular plexus and how this plexus is remodeled to give rise to the mature coronary arteries. Our specific goals were to (1) characterize the chronology and morphology of coronary vascular development as a prelude to understanding abnormal development; (2) understand the relationship of hematopoiesis to coronary vascular development; and (3) determine the timing and site of coronary smooth muscle differentiation.
We set out to define the spatiotemporal formation of the epicardial coronary vascular system in the developing avian heart. To delineate the formation of endothelial tubes and maturing blood vessels, hearts from staged quail embryos were fixed and stained in whole-mount for the endothelial antigen of the monoclonal antibody QH-1 (which recognizes a carbohydrate epitope on the surface of quail angioblasts and endothelial and hematopoietic cells; Pardanaud et al., 1987). Here, we define the morphological description of coronary vascular development into four distinct stages: (1) appearance of QH-1–positive cells in the proepicardial organ (Fig. 1); (2) appearance of QH-1–positive cells in the subepicardial mesenchyme (Fig. 2) as well as CD45-positive cells (Fig. 3); (3) formation of the subepicardial endothelial plexus (Fig. 4); (4) growth and remodeling of endothelial plexus (Fig. 5) and maturation of the vessel wall (Fig. 6). The patterns that we describe support a vasculogenic mechanism for coronary vessel development and a dorsal to ventral, cranial to caudal progression of coronary vascular development.
Appearance of QH-1–Positive Cells in the Proepicardial Organ and Superficial Epicardium
There are two proposed sources of endothelial cells that populate the developing subepicardial vascular plexus. Endothelial cells may be derived from angioblasts that migrate individually to the external surface of the heart from the septum transversum through the proepicardial organ (Poelmann et al., 1993). It is also thought that some endothelial cells may be derived from epithelial cells in the superficial epicardium, sharing a common lineage with vascular smooth muscle and perivascular fibroblasts (Pérez-Pomares et al., 2002a). To identify the temporal appearance and accumulation of coronary vascular endothelial cells in the proepicardial organ, we stained quail embryos from stages HH15–17 with QH-1 (Fig. 1). In embryos in which the proepicardial organ had not yet attached to the heart, we did not observe QH-1–positive cells in the proepicardial organ (Fig. 1A). However, we observed QH-1–positive cells in HH17 embryos, in which the proepicardial organ has attached to the heart (Fig. 1B). In whole-mount stained embryos, this finding appeared to identify individual cells; therefore, to confirm this observation, we sectioned HH17 embryos and stained them with QH-1 (Fig. 1C). We observed that, in cross-section, the QH-1 reaction was in individual cells not clusters of individual cells. Similar cells in sections through HH15 and HH16 proepicardial organs did not react with QH-1 (not shown). While consistent with the hypothesis that endothelial cells migrate through the proepicardial organ to the heart surface, we did not directly examine cell migration. As an indirect measure, we inspected sections of HH17 embryos to see if there are greater numbers of QH-1+ cells at the base of the proepicardial protrusion and fewer QH-1+ cells at the villous tips. As seen in Figure 1D, that appears to be the case. We conclude QH-1–positive cells appear to enter the basal aspect of the pyopericardium and either migrate or proliferate to populate the proepicardial villi near the time the proepicardial organ attaches to the sinoatrial pole of the heart.
After the proepicardium attaches to the heart, epicardial cells migrate cranially as an epithelial sheet over the surface of the heart (Virágh and Challice, 1981). We next inspected QH-1 staining in HH18–26 hearts to follow endothelial cells during the formation of the superficial epicardium (Fig. 2). In HH23 hearts, we observed endothelial cell precursors on the posterior surface of the heart, including the initial formation of tubules (Fig. 2A). On the anterior aspect, we observed a scattered distribution of endothelial cells over the ventricles and atrium in addition to an accumulation of cells at the border between the conotruncus and ventricles (Fig. 2B). Within 1 embryonic day, endothelial cells appear to have populated the entire surface of the heart, except the conotruncal surface (Fig. 2C,D). This distribution was heterogeneous with endothelial cells concentrated in the atrioventricular and atrioventricular (AV) groove regions and endothelial tubules in the posterior AV groove getting larger (Fig. 2C). Thus, we observed a slight delay in the formation of the plexus with vessels forming first on the posterior aspect and then on the anterior aspect.
It is thought that, in early vasculogenic beds of the avian embryo, angioblasts are committed to the endothelial lineage before they accumulate the QH-1 antigen (Drake et al., 1997). Therefore, we tested if the distribution of QH-1+ cells at the heart surface was similar to that of cells reactive for the common leukocyte antigen CD45 and the hemangioblast and endothelial marker VEGFR2 (Jaffredo et al., 1998; Eichmann et al., 1999). Even at the earliest stages of subepicardial coronary vascular development, blood cells, identified by CD45 expression, are found on the surface of the HH23 heart (Fig. 3). The pattern of expression of CD45 was closely reminiscent to that of QH-1 (Fig. 3A,B,A′,B′). CD45+ cells were scattered uniformly over the atrial and ventricular surfaces and did not appear exclusively in areas where early endothelial tubes were forming (Fig. 3B). The widespread appearance of cells in the hematopoietic lineage on the heart surface several days before coronary vessels are connected to the aorta was quite unexpected and suggests that hematopoietic and endothelial precursors arrive contemporaneously. In cross-sections, CD45+ cells were closely aligned with mature blood cells and blood islands (Fig. 3D). When we stained hearts with anti–VEGF-R2, the earliest marker of angioblasts and endothelial cells (Perez- Pomares et al., 1998b) we observed staining primarily in endothelial tubes (Fig. 3C) similar to those seen on the dorsal aspect of HH23 hearts stained with QH-1 (Fig. 3A). We also observed individual VEGF-R2+ cells in close association with endothelial tubes (C′) and blood islands (E). Thus, the pattern of accumulation of QH-1 overlaps both the expression of CD45, marking the blood cell lineage, and VEGF-R2, marking the endothelial lineage, consistent with, but not proof of, their origin from a common precursor cell.
Formation of the Subepicardial Endothelial Plexus
As individual or clusters of individual QH–1+ endothelial cells gather to form endothelial tubes, the vessels begin to coalesce into a vascular plexus in the subepicardial matrix (Fig. 4). From stages HH27 to HH35, the primitive endothelial vessels on the dorsal aspect of the heart organize into a nascent plexus starts that grows toward the atrioventricular groove, the outflow tract, and eventually into the interventricular groove. On the dorsal aspect, the plexus is disorganized and discontinuous, but by stage HH31, most of the plexus is mature and well connected (Fig. 4B). By HH33, the plexus begins to organize around the region of the future posterior descending artery and is not connected to the plexus around the future vessels forming over the interventricular groove (Fig. 4C). During these stages, we observed numerous individual endothelial cells or clusters of cells near each plexus (Fig. 4C,D). As we observed in the younger hearts, the plexus on the dorsal aspect matured slightly earlier than that on the ventral aspect. From stages HH29–HH35, the anterior plexus forms over the outflow tract eventually invading the proximal aorta (Fig. 4D,E) and growing toward the apex (Fig. 4F). The initial plexus grows from the population of pioneering QH-1+ cells found in the groove formed between the conotruncus and the ventricles (Fig. 2D). This plexus is characterized by endothelial segments (Fig. 4D) and clusters of individual cells (Fig. 4D,E). While scattered over the diaphragmatic surface of the ventricles, these clusters are concentrated near the outflow tract (Fig. 4D).
Remodeling of the Coronary Vascular Plexus
During stages HH34–HH38, we observed the formation of large diameter QH-1+ sinusoids in the posterior and anterior subepicardial plexuses (Fig. 5). These sinusoids usually were not connected to each other, thus confirming the vasculogenic mechanism though which they were assembled. Sinusoids were organized around sites of future, central vessels and over the ventricular surface (Fig. 5A,B,D,E). As hearts matured, we observed that the large sinusoids remodeled into smaller but mature vessels, creating several principal anterior branches (Fig. 5C,F). By stage HH38, remodeling had formed the principal channels where the definitive coronary vessels will persist later in adult life. At this stage, the vessels will have connected to the sinuses of Valsalva at the root of the aorta. Of particular note was the growth of vessels in and around the pulmonary artery and aorta (Fig. 5D–F). QH-1 staining was intense at the base of the great vessels and within the grooves defining the major trunks. Here, we observed individual QH-1+ cells and discontinuous segments that coalesced into vessels.
Growth and Maturation of the Coronary Vessels
As remodeling of the plexuses continues, maturation of the individual vessels begins. Larger vessels begin to express smooth muscle cell markers, starting with smooth muscle α-actin at HH27 (Fig. 6C). As the plexuses remodel into centrally dominated vessels, smooth muscle marker expression increases, to include smooth muscle myosin, calponin, caldesmon, and SM22-α. In our experience, cells expressing smooth muscle markers are found only associated with these remodeling (and maturing) vessels, never in situ in the proepicardial organ (Fig. 6A,B), the epicardium, or alone in the supepicardial matrix (data not shown).
From stages HH31 to HH39, we observed an increasing reaction with anti–smooth muscle α-actin as the coronary vessels grow in diameter (Fig. 6D–F). By this time, CVSMC marker expression was rich and concentrated in the largest vessels (Fig. 6C,D). By hatching (E17 days in the quail), vessels penetrate the myocardium at the anterior aspect of the heart. At the same time on the posterior surface of the heart, the principal posterior descending coronary and its ramifications extend through the interventricular groove. Surrounding these principal coronary vessels, a thinner network of vessels expressed QH-1.
Here, we report the morphological patterns of coronary vascular development in the quail embryo. We define four stages of this developmental process; appearance of endothelial cell precursors, formation of an endothelial plexus, remodeling of the plexus, and growth and maturation. The general patterns of coronary development reflect that vasculogenic mechanisms control coronary vascular formation. Throughout embryonic development of the coronary vasculature, we found significant overlap of these several stages, so that individual hemangioblast precursors were present even as the vascular plexus was being remodeled to the mature vessel stage. While this mimics the dorsal-to-ventral progression of vascular assembly in the early embryo (Drake et al., 1997), the simultaneous presence of individual or small clusters of hemangioblasts and relatively mature vessels is both striking and, to our knowledge, unique to coronary vascular development. We speculate that this finding is a reflection of the rather large distance that endothelial precursors must travel as well as the relative lateness of coronary development compared with other vascular beds.
A second important finding of our study is the demonstration of many CD45+ cells at the same time and in a similar distribution to QH-1+ cells on the atrial and ventricular surfaces of the developing heart. This finding supports the previous concept that blood islands codevelop with the coronary vasculature (Hiruma and Hirakow, 1989; Virágh et al, 1993; Tomanek, 1996). What is new is that CD45 expression appears to precede the association of these cells with the vasculature (Fig. 3D). CD45 is acknowledged to be a marker of the hematopoietic lineage after its divergence from the angioblast and it is usually first expressed at the center of clusters of hemangioblasts rather than on single cells as we find in the HH23 quail heart (Eichmann and Corbel, 1999). Thus, either CD45 expression precedes myelocyte commitment in the heart or there is novel migration of committed myelocytes to blood islands during heart development.
The vasculogenic origin of the coronary vasculature is suggested by anatomic and immunohistochemical analyses (Vrancken Peeters et al., 1997a, b; Ratajska and Fiejka, 1999; 2000) and was definitively shown by lineage traces performed by labeling proepicardial precursors with replication-defective retrovirus expressing lacZ (Mikawa and Fischman, 1992). Infection of the proepicardial organ results in subsequent labeling of the discrete segments of coronary endothelium or smooth muscle. That labeling of endothelium and smooth muscle does not occur within a given segment supports separate endothelial and smooth muscle origins (Mikawa and Gourdie, 1996). Further support of separate smooth muscle and endothelial origins was provided by lineage studies of labeled epicardium (Dettman et al., 1998). These studies were important, because the epicardium is a morphologic and antigenically uniform tissue while the proepicardium contains several cell types. The epicardial lineage trace demonstrated an epicardial to mesenchymal transformation in which the epicardium gives rise to smooth muscle, fibroblast, and pericyte lineages but not the endothelial lineage. What remained unclear is when smooth muscle or pericyte precursors become committed to those lineages.
In the present study, we show that smooth muscle markers appear only in relationship with coronary vessels. We were not able to identify smooth muscle-specific protein expression in the proepicardial organ or surrounding early endothelial tubes. While SMαactin does not uniquely label smooth muscle in the heart, because it is also expressed by mesenchymal cells of the AV cushions (Nakajima et al., 1997), it is the earliest described marker of smooth muscle differentiation (Owens, 1995; Landerholm et al., 1999). SMαactin expression appeared only in the media of the nascent blood vessels not in the proepicardium, subepicardial space, or blood vessel's surrounding mesenchyme.
The factors responsible for driving epicardially derived mesenchyme to the smooth muscle lineage are not yet known. Serum or growth factor treated proepicardial cells can be induced to express smooth muscle genes after they undergo epicardial–mesenchymal transformation in vitro (Landerholm et al., 1999; Lu et al., 2001; Dettman et al., 2003). Despite the eagerness of all proepicardial cells to become smooth muscle in vitro, we have observed that, in vivo, they do not do so until they reach the vascular wall of a blood vessel some days later. Moreover, other studies have shown that proepicardial and epicardial cells can differentiate into other cell types in the heart (Männer, 1999; Pérez-Pomares et al., 2002b). This finding indicates that there must be either negative influence on these cells to restrain differentiation (Dettman et al., 2003) or that the appropriate influences, such as platelet-derived growth factor-beta (Lu et al., 2001), do not effect epicardial cell fate until they reach their proper position within the heart.
A final issue is the role of mechanical forces in development of the vascular media and adventitia. Mechanical forces due to blood flow have been implicated in vascular growth and remodeling in many diseases (Xu, 2000). In the developing heart, the primary endothelial plexus forms and then invades the aorta. When this happens, anterograde flow is established within the developing coronary vessels. Subsequent accumulation of vascular smooth muscle during development may resemble the progressive extension of the muscle coat of the pulmonary arterial walls that occurs with increased pulmonary blood flow that follows birth (Hislop and Reid, 1981). The idea that physical forces imposed on newly formed blood vessels triggers the expression of molecules that influence smooth muscle differentiation and vessel remodeling is an attractive one. However, our observations do not support the hypothesis that coronary flow triggers smooth muscle differentiation, since we observed SMαactin staining in HH27 coronary vessels. This expression occurs at an earlier embryonic stage than described by Hood and Rosenquist (1992) and days before the primary plexus is connected to the aorta (Bogers et al., 1989; Waldo et al., 1990). While SMαactin may not accumulate exclusively in smooth muscle cells (Nakajima et al., 1997), we observed staining by anti-SMαactin in blood vessels unambiguously in HH31 (E7 hearts). We observed that SMαactin did not show a proximal to distal gradient of expression as has been reported previously (Vrancken Peeters et al., 1997a). By contrast, it appeared in a symmetric fashion and at an earlier developmental stage. Our results clearly indicate that smooth muscle accumulates around coronary vessels at developmental time before the establishment of coronary circulation.
However, this does not exclude the potential effect of mechanical forces, because it is likely that, as soon as the coronary vessels form fluid filled lumens, contraction of the heart creates a pulsatile pressure within them. This mechanical force, while distinct from shear forces caused by blood flow, does trigger molecular signals that are known to recruit and stimulate proliferation of smooth muscle (Ma et al., 1999; Sudir et al., 2001). Thus, the appearance of smooth muscle markers near early coronary vessels could be dependent on mechanical forces, even though anterograde flow has not yet begun. Our experiments suggest that this appearance occurs at or near HH27.
In summary, we have described in detail four stages of subepicardial coronary development in the avian model and how it compares with other vascular beds. Because coronary development is morphologically different, we speculate that coronary development might be regulated by different signals than other vascular beds. We also describe the association of hematopoiesis with coronary development. Curiously, commitment to the hematopoietic lineage appears to precede the formation of blood islands. Finally, we investigated the timing and location of coronary smooth muscle differentiation. Smooth muscle markers are not present in the proepicardial organ and only appeared in the coronary vessel's wall. Hence, although smooth muscle differentiation appears to occur in situ, it remains uncertain whether smooth muscle precursors are committed to this cell fate before their arrival at the vascular wall.
We used chick (Gallus gallus domesticus) and Japanese quail (Cotournix cotournix japonica) embryos. Eggs were incubated in humidified egg incubators at 37°C. Embryos were staged according to the criteria of Hamburger and Hamilton (HH) or in terms of embryonic day (Hamburger and Hamilton, 1951). At least 10 hearts or embryos from each developmental stage between HH15 and HH43 (E2 and E17) were removed then washed three times in phosphate-buffered saline (PBS). Explants were fixed in methanol and dimethyl sulfoxide (DMSO) 4:1 (Dent et al., 1989) overnight at 4°C. Pretreatment was achieved by using methanol-DMSO-30%H2O2 4:1:1 for 4 hr at room temperature to block endogenous peroxidase. Background binding was blocked incubating twice in PBS plus 2% bovine serum albumin (BSA) and 0.1% v/v Triton X-100 (PBT) for 1 hr each at room temperature. Diluted primary antibodies were incubated with tissue overnight at 4°C in PBT followed by five washes in PBT at 4°C for 1 hr each. Goat secondary antibodies coupled to horseradish peroxidase were incubated with tissue overnight at 4°C in PBT at a final concentration of 2 μg/ml. After five washes with PBT-BSA at 4°C for 1 hr each, embryos and hearts were incubated in diaminobenzidine tetrachloride/nickel-cobalt (Zymed Laboratories, Inc.) as the substrate for horseradish peroxidase. Color developed at room temperature between 4 and 8 min. Samples were visualized and photographed by using a Zeiss 2000-C stereo dissecting microscope. Embryos and hearts were dehydrated to 100% methanol and stored at −20°C. We used monoclonal antibody (mAb) QH-1 (Developmental Studies Hybridoma Bank) using a dilution of 1:100; anti-SMαactin 1:100 (clone 1A4, Sigma), anti-SM myosin mAb 1:100 (clone HSM-V, Sigma), anti-calponin mAb 1:100 (clone CP-93, Sigma), anti-caldesmon mAb 1:100 (clone CALD-5, Sigma). The anti-chick hematopoietic cells CD45 mAb 1:100 (Jaffredo et al., 1998) was obtained from ID-DLO, The Netherlands. The anti-quail VEGF-R2 (Quek-1 mAb) was a generous gift from Anne Eichmann; Institut d'Embryologie Cellulaire et Moléculaire, Collège de France. VEGF-R2 in whole-mount hearts was revealed by using the Tyramide Signal Amplification System (TSA, NEN Life Science) according to the manufacturer's protocol. Other procedures for whole-mount immunocytochemistry are as described above.
Immunocytochemistry in Tissue Sections
Hearts were obtained from quail embryos between HH15 and HH40. They were fixed in formaldehyde 4% (v/v) from paraformaldehyde (EM Sciences) at 4°C overnight. Tissue was dehydrated in ethanol series, cleared with Citrisolve (Fisher) and embedded in paraffin. Heart sections (7 μm) were prepared for IH by using Citrisolve and graded ethanol series, rinsed in tap water, and washed in PBS for 5 min. Endogenous alkaline phosphatase activity was inhibited by using 20% acetic acid at 4°C for 3 min. Then sections were incubated in PBT plus blocking serum for 1 hr, incubated with the primary antibody for 1 hr, washed in PBS, and incubated with the specific biotinylated secondary antibody for 1 hr. Vectastain avidin-biotinylated complex reagent (Vector Laboratories) was used for amplification of the signal and VectorRed alkaline phosphatase substrate kit was used for detection showing a red signal. Hematoxylin was used as counterstain. The samples were mounted with Gel/Mount (Biomeda). Samples were visualized and photographed by using a Leitz Orthoplan 2 microscope. Antibodies used for IH in tissue sections were QH-1 (Developmental Studies Hybridoma Bank), anti-SMαactin mAb 1:100 (clone 1A4, Sigma), anti-SM myosin mAb 1:100 (clone HSM-V, Sigma), anti-calponin mAb 1:100 (clone CP-93, Sigma), anti-caldesmon mAb 1:100 (clone CALD-5, Sigma), SM22α (a generous gift from Mario Gimona, Austrian Academy of Sciences, Institute of Molecular Biology, Salzburg), anti-cytokeratin pAb 1:400 (Dako), and CD45 mAb 1:100 (Jaffredo et al., 1998).
We thank Dr. Anne Eichman for her gift of the anti-VEGFR2 antibody, Dr. Chris Morabito for help with early phases of this work, and Dr. Nicolas Porta for insightful comments on the manuscript. R.W.D. received an American Heart Association Scientist Development Award, and J.B. received an American Heart Association Established Investigator Award and Grant-in-Aid.