Transforming growth factor-beta 1 (TGF-beta 1) has been implicated in the positive regulation of angiogenesis in vivo, whereas it inhibits the proliferation of endothelial cells in vitro. To reconcile these apparently contradictory effects, we have investigated the effect of TGF-beta 1 on bovine aortic endothelial cells that exhibit spontaneous angiogenesis in vitro. We show that concentrations of TGF-beta 1 which stimulate proliferation of cells that form endothelial cords and/or tubes inhibit proliferation of the same cells grown at subconfluent densities. An increase in cell number of 35% over control cultures was achieved with 0.5 ng TGF-beta 1/ml. The proliferative effect was blocked by antibodies against TGF-beta. Immunological detection of BrdU-labeled nuclei revealed an increase greater than 220% in cells treated with TGF-beta 1. Moreover, a population of cells within the cords appeared to be a selective target for this cytokine. The stimulatory effect was not restricted to bovine aortic endothelial cells, as similar results were obtained with endothelial cells derived from rat microvessels. Significant levels of active TGF-beta 1 were detected in cultures containing cords/tubes, whereas only latent TGF-beta 1 was detected in subconfluent cultures. We show further that endothelial cells exhibiting angiogenesis in vitro secrete plasminogen activator, an enzyme that regulates activation of TGF-beta. The major increases in mRNA transcripts for extracellular matrix proteins that are typically associated with TGF-beta 1 were not seen in cells exhibiting angiogenesis in vitro. Since the formation of tubular networks requires both invasion and proliferation, we propose that TGF-beta 1 is a major morphoregulatory factor in angiogenesis that specifically controls endothelial cell proliferation and extracellular matrix turnover.

Type VIII collagen was initially characterized as a biosynthetic product of bovine aortic endothelial cells in vitro and thus named Endothelial Collagen(EC). The protein has now been identified in avian and mammalian species as a product secreted by cultured corneal endothelial cells, keratinocytes, fibroblasts, astrocytoma and other tumor cells, and endothelial cells derived from several classes of blood vessels. In vivo, the distribution of type VIII collagen is restricted to specialized extracellular matrices such as meninges, periosteum, perichondrium, and Descemet's membrane of the cornea. It is also present in the subendothelium of large and small blood vessels and has been localized in capillary-like structures formed by sprouting cultures of endothelial cells. Data from protein and nucleotide sequencing have identified two distinct chains, termed α1 (VIII) and α2 (VIII); however, the assembly of these chains in the extracellular space is still poorly understood. Protein extracts from Descemet's membrane indicate that, in this tissue, type VIII collagen is a heterotrimer. Our understanding of the biological role of type VIII collagen is rather limited. Although type VIII collagen seems to perform a structural role in the Descemet's membrane, its temporally and spatially restricted expression during development suggests an involvement in tissue remodeling and differentiation.

The Mov-13 strain of mice was created by the insertion of the murine Moloney leukemia virus into the first intron of the alpha 1 (I) collagen gene. Consequently, Mov-13 embryos do not transcribe alpha 1 (I) collagen mRNA and lack type I collagen protein in the extracellular matrix (ECM). Homozygotes die within 12-14 days of embryonic development, in part from the rupture of large blood vessels, and also exhibit deficiencies in hematopoesis and assembly of the ECM (Lohler et al. [1984] Cell 38:597-607). Several matricellular proteins, proteoglycans, and growth factors bind to type I collagen, e.g., fibronectin, secreted protein acidic and rich in cysteine (SPARC), decorin, and transforming growth factor-beta. Here we investigate the expression and function of SPARC in the absence of type I collagen. We show that fibroblasts isolated from Mov-13 homozygous, heterozygous, and wild-type embryos transcribed and translated SPARC mRNA in vitro. However, accumulation of extracellular SPARC was severely affected in the tissues of Mov-13 homozygotes, whereas extracellular deposition of the secreted glycoproteins fibronectin and type III collagen was not altered. Since SPARC has been shown to be a regulator of cell shape, the functional consequences of the absence of extracellular SPARC were evaluated in collagen gel contraction assays. Fibroblasts isolated from homozygous Mov-13 mice did not contract native type I collagen gels as efficiently as fibroblasts from heterozygous littermates; however, addition of exogenous SPARC enhanced the contraction of collagen by homozygous Mov-13 fibroblasts. The stimulatory effect of SPARC was blocked by antibodies specific for the amino terminus of the protein. These results provide evidence that type I collagen is one of the major extracellular proteins that binds SPARC in vivo. Furthermore, the capacity of fibroblasts to contract ECM in vitro is enhanced by extracellular SPARC. We therefore propose that the remodeling of ECM by cells in vivo is regulated in part by a specific interaction between SPARC and type I collagen.

Vitronectin (Vn) is not only a major adhesive glycoprotein in plasma but also regulates cell-mediated proteolytic enzyme cascades, including the complement, coagulation, and fibrinolytic systems. This broad functional activity suggests that Vn may also play a critical role in development. To begin to investigate this possibility, we studied Vn gene expression during murine embryogenesis. In situ hybridization analysis of embryonic tissues revealed Vn mRNA primarily in the liver and the central nervous system (CNS). In the liver, Vn mRNA was detected by day 10, the level increasing at later developmental stages. In the CNS, Vn mRNA was also detected as early as day 10 and was confined to the floor plate. However, as development proceeded, high levels of Vn transcripts became prominent in the meninges of the cortex and spinal cord, and in close proximity to brain capillaries. The perikarya of most neurons lacked Vn mRNA. Unexpectedly, high levels of Vn mRNA were associated with capillaries of the CNS, but not with blood vessels of peripheral organs. These results indicate that Vn is expressed in a spatially and temporally distinct pattern during murine embryogenesis, and suggest that the Vn transcript may be a CNS-specific vascular marker.

Thrombospondin 1 is a secreted, trimeric glycoprotein that mediates interactions between cells and extracellular matrix and exhibits cell-specific effects on migration and proliferation. Recently, two additional thrombospondin genes (thrombospondin 2 and 3) have been identified. To study the functions of these proteins, we have used in situ hybridization and RNAse protection assays to compare the expression of the genes encoding thrombospondin 1, 2, and 3 during murine embryogenesis. Thrombospondin mRNAs were associated with ossification, neuronal organogenesis, and lung development, although transcripts were differentially expressed. Thrombospondin 1 was predominant from days 10 to 13. During this period, high but transient levels of expression were observed in the neural tube, head mesenchyme, and cardiac cushions. In contrast, a more constant level of thrombospondin 1 mRNA was apparent in resident megakaryocytes of the liver, as well as in circulating megakaryocytes; neither thrombospondin 2 nor 3 was detected in these cells. Thrombospondin 1 was also produced by cells of the developing kidney and gut. The expression of thrombospondin 2 was confined principally to organized connective tissue that included pericardium, pleura, perichondrium, periosteum, meninges, ligaments, and reticular dermis. Thrombospondin 2 was also produced by differentiating skeletal myoblasts and by cells of the kidney and gut. Moreover, high levels of expression were detected in blood vessels. Thrombospondin 3 mRNA was restricted to brain, cartilage, and lung. Although thrombospondin 1, 2, and 3 belong to a family of structurally related genes, the differences observed in the spatiotemporal distribution of the corresponding mRNAs indicate unique functions for these secreted proteins.

SPARC is a secreted glycoprotein that modulates cell shape and cell-matrix interactions. Levels of SPARC are increased at sites of somitogenesis, osteogenesis, and angiogenesis in the embryo and during wound repair in the adult. We have cloned and characterized SPARC from chicken embryo. A 2.2-kbp cDNA, obtained by a novel use of the polymerase chain reaction, was determined to encode a 298-residue protein that is 85% identical to human SPARC. Antigenic sites in particular appear to be highly conserved, as antibodies against C-terminal sequences of murine and bovine SPARC reacted with a 41-43 kDa protein in chicken embryo extracts. Chicken SPARC can be defined by four sequence signatures: (a) a conserved spacing of 11 cysteine residues in domain II, (b) the pentapeptide KKGHK in domain II, which is contained within a larger region of 31 identical residues, (c) a 100% conserved region of 10 residues in domain III, and (d) a C-terminal, calcium-binding EF-hand motif. SPARC mRNAs in the 10-day-old chicken embryo are represented by three sizes of 1.8, 2.2 and 3.0 kb. The relative steady-state levels for the 2.2-kb mRNA were determined as aorta > or = skeletal muscle > calvarium > vertebra > anterior limb > kidney > heart > brain > skin and lung >> liver. The relative abundance of the 1.8-kb and 2.2-kb mRNAs varied among tissues and indicated that differential processing of SPARC mRNAs might occur. All three RNA species were detected by a cDNA probe for the N-terminal part of the coding region. Thus, the three mRNA species appear to arise from differential 3' splicing and/or polyadenylation. Collective evidence demonstrates that SPARC has been well-conserved during vertebrate evolution, a finding that indicates a fundamental role for this protein in development.

In our recent studies, we have demonstrated that monolayer cultures of bovine aortic endothelial (BAE) cells that do not express type I collagen also fail to express and synthesize decorin, a small chondroitin/dermatan sulfate proteoglycan that interacts with type I collagen and regulates collagen fibrillogenesis in vitro. However, BAE cells exhibiting a spontaneous sprouting phenotype and a predisposition toward the formation of cords and tube-like structures (an in vitro model for angiogenesis) initiate the synthesis of type I collagen during their morphological transition from a polygonal monolayer to an angiogenic phenotype. In the present study, we examined whether BAE cells also initiate the synthesis of the proteoglycan decorin during this morphological transition. We show by Northern blot analysis and by immunochemical methods that BAE cell cultures containing sprouting cells and cords, but not monolayer cultures of these cells, express and synthesize decorin (M(r) approximately 100,000). We also show that type I collagen expression by BAE cell cultures is initiated concomitantly. However, the localization of decorin and type I collagen in cord and tube-forming BAE cell cultures is not completely identical. Type I collagen is detected only in sprouting BAE cells and in endothelial cords, whereas decorin is also apparent in BAE cells surrounding the cords and tubes. Our results indicate that the synthesis of decorin as well as type I collagen is associated with endothelial cord and tube formation in vitro.

1. Elastin was isolated from the bulbus arteriosus of a salmonid fish. Monoclonal and polyclonal antibodies, elicited against a CNBr digest of this protein, immunoprecipitated a polypeptide of Mr 43,000 from fish cell culture medium. 2. Cell-free translation of salmon poly A+ RNA produced a protein of approximately 43 kD that was immunoprecipitated with anti-elastin antibodies. The corresponding mRNA had an approximate Mr of 2 kb. 3. Despite similarities in amino acid composition, the differences in Mr between mammalian and salmon mRNA and protein suggest a divergence of fish and higher vertebrate elastins from an earlier ancestral gene.

Selected strains of vascular endothelial cells, grown as confluent monolayers on tissue culture plastic, generate flat networks of cellular cords that resemble beds of capillaries--a phenomenon referred to as "spontaneous angiogenesis in vitro". We have studied spontaneous angiogenic activity by a clonal population (clone A) of bovine aortic endothelial cells to identify processes that mediate the development of cellular networks. Confluent cultures of clone A endothelial cells synthesized type I collagen, a portion of which was incorporated into narrow, extracellular cables that formed a planar network beneath the cellular monolayer. The collagenous cables acted as a template for the development of cellular networks: flattened, polygonal cells of the monolayer that were in direct contact with the cables acquired spindle shapes, associated to form cellular cords, and became elevated above the monolayer. Networks of cables and cellular cords did not form in a strain of bovine aortic endothelial cells that did not synthesize type I collagen, or when traction forces generated by clone A endothelial cells were inhibited with cytochalasin D. In a model of cable development, tension applied by a confluent monolayer of endothelial cells reorganized a sheetlike substrate of malleable type I collagen into a network of cables via the formation and radial enlargement of perforations through the collagen sheet. Our results point to a general involvement of extracellular matrix templates in two-dimensional (planar) models of vascular development in vitro. For several reasons, planar models simulate invasive angiogenesis poorly. In contrast, planar models might offer insights into the growth and development of planar vascular systems in vivo.