Regulation of Gene Expression by SPARC during Angiogenesis in Vitro CHANGES IN FIBRONECTIN, THROMBOSPONDIN-1, AND PLASMINOGEN ACTIVATOR INHIBITOR-1*

Angiogenesis in vitro, the formation of capillary- like structures by cultured endothelial cells, is associated with changes in the expression of several extra- cellular matrix proteins. The expression of SPARC, a secreted collagen-binding glycoprotein, has been shown to increase significantly during this process. We now show that addition of purified SPARC protein, or an N-terminal synthetic peptide (SPARC4-23), to strains of bovine aortic endothelial cells undergoing angiogenesis in vitro resulted in a dose-dependent de- crease in the synthesis of fibronectin and thrombospon-din-1 and an increase in the synthesis of type l-plas- minogen activator inhibitor. SPARC decreased fibronectin mRNA by 76% over 48 h, an effect that was inhibited by anti-SPARC immunoglobulins. Levels of thrombospondin-1 mRNA were diminished by 80%. Over a similar time course, both mRNA and protein levels of type 1-plasminogen activator inhibitor (PAI-1) were enhanced by SPARC and the SPARC4-23 pep- tide. The effects were dose-dependent with concentrations of SPARC between 1 and 30 rg/ml. In contrast, no changes were observed in the levels of either type I collagen mRNA or secreted gelatinases. Half-maximal induction of PAI-1 mRNA or inhibition of fibronectin and thrombospondin mRNAs occurred with 2-5 pg/ml SPARC

Angiogenesis in vitro, the formation of capillarylike structures by cultured endothelial cells, is associated with changes in the expression of several extracellular matrix proteins. The expression of SPARC, a secreted collagen-binding glycoprotein, has been shown to increase significantly during this process. We now show that addition of purified SPARC protein, or a n N-terminal synthetic peptide (SPARC4-23), to strains of bovine aortic endothelial cells undergoing angiogenesis in vitro resulted in a dose-dependent decrease in the synthesis of fibronectin and thrombospondin-1 and an increase in the synthesis of type l-plasminogen activator inhibitor. SPARC decreased fibronectin mRNA by 76% over 48 h, an effect that was inhibited by anti-SPARC immunoglobulins. Levels of thrombospondin-1 mRNA were diminished by 80%. Over a similar time course, both mRNA and protein levels of type 1-plasminogen activator inhibitor (PAI-1) were enhanced by SPARC and the SPARC4-23 peptide. The effects were dose-dependent with concentrations of SPARC between 1 and 30 rg/ml. In contrast, no changes were observed in the levels of either type I collagen mRNA or secreted gelatinases. Half-maximal induction of PAI-1 mRNA or inhibition of fibronectin and thrombospondin mRNAs occurred with 2-5 pg/ml SPARC and approximately 0.05 mM SPARC4-23. Strains of endothelial cells that did not form cords and tubes in vitro had reduced or undetectable responses to SPARC under identical conditions. These results demonstrate that SPARC modulates the synthesis of a subset of secreted proteins and identify an N-terminal acidic sequence as a region of the protein that provides a n active site. SPARC might therefore function, in part, to achieve an optimal ratio among different components of the extracellular matrix. This activity would be consistent with known effects of SPARC on cellular morphology and proliferation that might contribute to the regulation of angiogenesis in vivo.

Vascular endothelial cells play a direct role in regulating
* This work was funded in part by National Institutes of Health Grants GM 40711 and HL03174. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Grant HL07312. 7 To whom correspondence should be addressed: Dept. of Biological Structure, SM-20, University of Washington, Seattle, WA 98195. Tel.: 206-685-1192;Fax: 206-543-1524. the growth of new vessels (angiogenesis) by their contribution to the synthesis and degradation of extracellular matrix (ECM)' macromolecules (Folkman, 1984;Ingber and Folkman, 1989;Liotta et al., 1991). The ECM, in turn, provides a variety of positive and negative signals to the resident endothelial cells. Confluent endothelial cells normally display low rates of replication and are noninvasive, properties that have generally been attributed to the presence of laminin and other components in the vessel wall (Furcht, 1986). Upon activation by phorbol esters or growth factors, endothelial cells secrete matrix-degrading enzymes and a variety of new extracellular gene products (Moscatelli et al., 1985;Ingber and Folkman, 1988). Among ECM proteins, fibronectin (FN) and several types of collagen seem to play definitive roles (Risau and Lemmon, 1988;Ingber and Folkman, 1988). Although changes in both the endothelial cell and its ECM are required for angiogenesis, many questions remain about how the extracellular environment functions to regulate specific features of cell behavior and the extent to which ECM signals are involved in the formation of new vessels.
The role of specific ECM proteins in angiogenesis has been particularly difficult to study in vivo due, in part, to our current lack of understanding of the molecular complexes that are formed in vivo and to problems in presenting these proteins in their native conformation. For example, the requirement of collagen synthesis for angiogenesis in vivo has relied on the use of inhibitors of protein processing which block collagen secretion but which can affect other cellular pathways as well (Maragoudakis et al., 1988;Ingber and Folkman, 1988). Similarly, evidence that ECM molecules such as thrombospondin (TSP) can act as endogenous regulators of vessel growth has been derived from the identification of a truncated form of the molecule (Good et al., 1990). In view of these limitations, a number of systems have been developed in vitro that allow the dissection of singular events involved in vessel formation (Jaye et al., 1985;Montesano et al., 1986;Kubota et al., 1988). We have characterized specific clones of endothelial cells in which cords and patent tubes develop spontaneously under standard culture conditions, after the cells have synthesized their own ECM (hela-Arispe et al., 1991a, 1991b). This system differs from previous models of angiogenesis in vitro in that it does not rely on the use of exogenous growth supplements or matrix substrata. It also duplicates many of the activities of endothelial cell behavior observed ' The abbreviations used are: ECM, extracellular matrix; bp, base pair; BAE, bovine aortic endothelial; FCS, fetal calf serum; FN, fibronectin; kb, kilobase pair; LPS, lipopolysaccharide (endotoxin); PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline; PAI-1, type 1-plasminogen activator inhibitor; PDGF, plateletderived growth factor; SDS, sodium dodecyl sulfate; TSP, thrombospondin-1; TGF-P, transforming growth factor$; DMEM, Dulbecco's modified Eagle's medium. during early angiogenesis, e.g. endothelial cell migration from a confluent monolayer, proliferation, organization of cords, and formation of lumina. Activities associated with late events in the stabilization of capillaries, such as the synthesis of laminin and the formation of a basement membrane, are not generally observed.
In previous studies we demonstrated that the morphogenesis of endothelial tubes was associated with changes in the expression of genes for specific extracellular proteins: an induction of type I collagen (hela-Arispe et al., 1991a), a decreased production of TSP, and an increased secretion of FN andSPARC (hela-Arispe et al., 1991a, 1991b). We have recently demonstrated that the production of type I collagen by these cultures is mediated by a de novo transcription of the d(1) collagen gene in cells that formed cords; unactivated cells in the same culture failed to transcribe detectable levels of this gene (Fouser et al., 1991). In a separate study, analysis of TSP protein demonstrated a specific interaction with cells forming capillary-like structures. The addition of anti-TSP antibodies to the cultures resulted in a specific increase in the number of cords. On the basis of these studies it was proposed that TSP might function to inhibit endothelial remodeling or to stabilize a noninvasive state (hela-Arispe et al., 1991~). Thus, models of angiogenesis in vitro can provide systems whereby the roles of individual molecules can be studied.
In this study we address the role of SPARC (also known as osteonectin, BM-40, and 43K), a protein that is induced during the formation of endothelial cords in vitro. SPARC is a secreted Ca2+-binding glycoprotein associated with tissue growth and remodeling (Termine et al., 1981;Holland et al., 1987;Sage et al., 1989a;Kelm and Mann, 1990). In cultures of subconfluent proliferating endothelial cells, it has been shown to induce changes in cell shape (Sage et al., 198913;Lane and Sage, 1990), delay entry into the S-phase of the cell cycle (Funk and Sage, 1991), and increase the production of endothelial type (type 1) plasminogen activator inhibitor (PAI-1) (Hasselaar et al., 1991). In addition, SPARC has recently been shown to modulate the activity of plateletderived growth factor (PDGF) through a direct binding interaction (Raines et al., 1992).
We were interested in cultures of endothelial cells that formed cords and tubes (i.e. angiogenic cultures) to elucidate potential activities of SPARC in vivo. We show that addition of exogenous SPARC decreased the synthesis of FN and TSP but had no apparent effect on the expression of type I collagen.
SPARC also stimulated the expression of PAI-1, a protein that can regulate the activation of latent TGF-/3 and the remodeling of matrix via its central role in the plasminogen activation pathway (Sat0 and Rifkin, 1989;Loskutoff et al., 1988). The response to SPARC required the activated morphology typical of angiogenic cultures of endothelial cells and appeared to be targeted specifically to cells involved in the formation of endothelial cords. Cultures of nonangiogenic BAE cells (i.e. cultures that did not form cords and tubes) showed markedly reduced responses. By using synthetic peptides corresponding to different regions of the molecule, we identified a peptide containing amino acids 4-23 of SPARC that had activities corresponding to those of the intact protein. A scrambled version of this sequence had no activity, and peptides from other regions of the protein were considerably less active, or inert, in these assays. We propose that the release of SPARC by endothelial cells could modulate the expression of ECM molecules in vivo in a paracrine and/or autocrine feedback manner. In view of the existing immunohistochemical evidence for the selective expression of FN, TSP, and plasminogen activators during angiogenesis in vivo, this form of regulation might be important for progression of the invasive angiogenic phenotype.
Purification of SPARC, Peptide Synthesis, and Antibody Production-SPARC was purified from conditioned PYS-2 cell media as described previously (Sage et al., 1989a), and SPARC peptides were prepared as described (Lane and Sage, 1990). The sequences used in this study are derived from the published sequence of murine SPARC (Mason et al., 1986) and are named according to their location in the amino acid sequence; the N-terminal alanine of the mature secreted protein is designated as 1. Sequences of peptides discussed in the text are described in Fig. 1. Two scrambled peptides, containing amino acid residues between 4 and 23 of the secreted protein, were made by inversion of the native sequence and a redistribution of the glutamic acid residues. As noted in the text, we used both the scrambled 19mer (SPARC6-23scr) and the 20-mer (SPARC4.23~~r) as control peptides in the experiments. Analysis of these sequences by the Mac-ProMass" program, developed by S. Vemuri and T. D. Lee (Beckman Research Institute, Duarte, CA), showed that the PI predicted for each of the SPARC N-terminal peptide isomers (SPARC,.zs, SPARC4.23scr, and SPARCs-z3scr) was 3.5.
Anti-SPARC antisera were produced against SPARC,.,, (Lane and In A are listed the amino acid sequences and names of SPARC peptides used in the present study. Peptides were synthesized and purity monitored by reverse phase high performance liquid chromatography as described (Lane and Sage, 1990). In B is shown a schematic representation of the amino acid sequence of SPARC in conjunction with functional domains identified with synthetic peptides. In some experiments, SPARCs-23scr, which lacked the glutamine residue ( Q ) , was used in place of SPARC4.23~~r; no difference in activity was detected. a, numbers indicate amino acids; the N-terminal amino acid of the mature protein has been designated as 1. b, total number of amino acids. c, Lane and Sage, 1990. d, Funk and Sage, 1991. e, present study. Sage, 1990). An immunoglobulin fraction from this serum specifically immunoprecipitated murine SPARC and has been shown to neutralize the anti-spreading activity of SPARC on endothelial cells. The antiserum has low cross-reactivity with bovine SPARC. Rabbit anti-I'AI-1 antiserum was a gift from Dr. David Loskutoff (Scripps Clinic and Research Foundation, La Jolla, CA). Rabbit anti-factor VI11 antigen (von Willebrand factor) was obtained from the Dako Corporation (Santa Barbara, CA). For immunohistochemical studies, the antisera were preincubated with FCS (1:lO dilution) for 24 h. This procedure effectively blocked the nonspecific reactivity of the anti-I'AI-1 antiserum toward high molecular weight components in FCS and had no effect on the immunoreactivity toward PAI-1; this specificity was assessed by immunoblotting. The content of LPS in preparations used in this study was determined by a Limulus amebocyte lysate assay (Hasselaar et al., 1991).
Riosynthetic Labe/ing of Secreted Proteins and Analysis by SDS-I'AGE-BAE cells were incubated in serum-free media for 12-24 h. Reagents were provided in fresh media every 24 h, and 50 pCi/ml L-[ 2,3,4,5-:'H]proline was added to each culture 12 h prior to collection o f the media. For analysis of collagen, medium contained 50 pg/ml sodium ascorbate and 64 pg/ml P-aminopropionitrile fumarate. Conditioned media were clarified of cell debris by centrifugation, and protease inhibitors were added to a final concentration of 0.2 mM phenylmethylsulfonyl fluoride and 5 mM N-ethylmaleimide. For analysis of radiolabeled proteins, media were dialyzed against 0.1 N acetic acid and lyophilized.
SDS-PAGE and Immunoblotting-Proteins from conditioned media were lyophilized and resuspended in SDS-PAGE sample buffer (Lane and Sage, 1990). An aliquot was removed for scintillation counting, and samples were reduced in 50 mM dithiothreitol for 3 min at 95 "C. Samples representing equivalent radioactivity were resolved on 4-15% SDS-polyacrylamide minigels. Proteins were stained with Coomassie Brilliant Blue R-250 (0.25% in 45% methanol and 9% acetic acid). Destained gels were impregnated with EN-HANCE'" and dried before exposure to x-ray film a t -70 "C. For immunoblotting, conditioned media were diluted with SDS-PAGE sample buffer containing dithiothreitol and were heated a t 95 "C for 3 min. Equivalent volumes were applied to 4-15% gradient minigels. Proteins were transferred to nitrocellulose and blocked with M T buffer (PBS, pH 7.7, containing 1% nonfat dried milk and 0.05% Tween 20). Blots were incubated with respective antibodies for 2 h a t room temperature. Antigen-antibody complexes were probed with 0.5 pCi/ml "'I-protein A. Radiolabeled proteins were detected as autoradiographic images on x-ray film exposed a t -70 "C with two intensifying screens.
RNA Isolation and cDNA Probes-Total RNA was isolated from cultures of BAE cells as described (Iruela-Arispe et al., 1991a). 5 or 10 pg of total RNA was denatured and resolved on 1.2% agarose gels containing 3% formaldehyde. T o verify the integrity of the samples, gels were stained for 5 min with 0.5 pg/ml ethidium bromide in diethylpyrocarbonate-treated water, destained, and photographed. RNA was transferred to nylon membranes under vacuum and crosslinked to the support using ultraviolet irradiation (Stratalinker", Stratagene, La Jolla, CA). RNAs were prehybridized for 16 h a t 42 "C in PH buffer (50% deionized formamide, 30% 20 X standard saline citrate (SSC) (1 X SSC is 0.15 M NaCI, 0.015 M sodium citrate), 50 mM sodium phosphate, 4% 50 X Denhardt's solution (1 X Denhardt's is 0.2 pg/ml Ficoll, 0.2 pg/ml polyvinylpyrrolidone, and 0.2 pg/ml hovine serum albumin), and 10 pg/ml yeast total RNA). Hybridization occurred under the same conditions in the presence of lo6 cpm/ml heat-denatured ['"PIcDNA probe. Final washing of hybridized RNA was performed in 0.1 X SSC, 0.1% SDS at 65 'C. In the case of the PAI-1 probe, transfers were washed a t 0.6 X SSC, 0.1% SDS at 65 "C.
Immunohistochemistry-BAE cells were grown on plastic coverslips until cords were clearly visible. Cultures were transferred to DMEM lacking FCS for 24 h and were subsequently treated with SPARC or SPARC peptides under serum-free conditions. After 20 h, cultures were washed 2 X with DMEM and fixed for 5 min with methanolic Carnoy's solution (30% methanol, 60% chloroform, 10% acetic acid). Fixed cell preparations were blocked in PBS, pH 7.65, containing 1% normal goat serum and were incubated with adsorbed anti-PAI-1 antiserum (1:250 final concentration, by volume) or antifactor VI11 antigen antiserum (1:lOO final concentration). Specific labeling was detected with biotinylated goat-anti rabbit IgG secondary antibody, avidin-biotin-peroxidase, and diaminobenzidine, as specified by the manufacturer (Vector Laboratories, Inc., Burlingame, CA). Preparations were photographed under bright field optics with a Zeiss photomicroscope equipped with a X 25 planapochromatic objective.
SDS Substrate Gels for Analysis of Proteases-Conditioned media were subjected to substrate gel electrophoresis on 10% SDS-polyacrylamide gels containing 0.1% gelatin or casein (Herron et al., 1986;Unemori and Werb, 1986). Media were mixed with SDS-PAGE sample buffer lacking dithiothreitol and urea, and samples were resolved at 15 mA/gel. SDS was removed by washing the gels in 2% Triton X-100 (twice for 15 rnin). Gelatinase activity was initiated and developed by incubating the gels in Tris-Ca" buffer (50 mM Tris-HCI, pH 8.0, 5 mM CaC12, 0.02% NaNd for 36 h at 37 'C. Gels were fixed and stained in 0.25% Coomassie Blue for 1 h and were subsequently destained. Gelatin-degrading enzymes were identified as clear zones on a blue background.

Three independent strains
of BAE cells that exhibited angiogenesis in vitro (termed angiogenic cultures) were used in these experiments; each culture displayed morphological and biosynthetic phenotypes similar to those described in previous studies (Iruela-Arispe et al., 1991a, 1991b. After passage, angiogenic BAE cells formed confluent monolayers within 2 days, and cord-like structures were apparent by 7 days (Fig. 2.4 (C, control)). In the presence of 1% FCS, the addition of up to 40 pg/ml SPARC ( Fig. 2A ( obvious change in morphology that was typical of the rounded phenotype described elsewhere with respect to nonangiogenic BAE cells and fibroblasts (Sage et ul., 1989b;Lane and Sage, 1990). Since our principal aim was to analyze the effect of SPARC on gene expression, we used quantities of protein that produced minimal alterations in the morphology of the angiogenic BAE cells. In addition, SPARC peptides a t concentrations of 0.1-0.4 mM, or anti-SPARC4.2R IgG at approximately 150 pg/ml, caused no apparent morphological alterations in these cells (data not shown). Some experiments were carried out in the absence of serum. Under these conditions, individual peptide preparations would occasionally result in altered cellular morphologies; however, the results were inconsistent and did not correlate with the changes in gene expression described below.
To analyze the biosynthetic profile of the cells in the presence or absence of exogenous SPARC, ["Hlproline-labeled culture media proteins were collected and analyzed by SDS-PAGE. As shown in Fig. 2B (lane C), cultures which did not receive SPARC exhibited a synthetic profile characteristic of angiogenic endothelial cells (hela-Arispe et al., 1991a). This secretory phenotype included high levels of FN (240 kDa), type I collagen (differential processing results in several bands between 100-200 kDa), and SPARC (40-43 kDa). After the addition of SPARC, there was a marked reduction in radiolabeled FN and an induction of a protein migrating a t 46 kDa (Fig. 2B (lane designated S P ) ) . The identity of the 240-kDa protein as FN was verified by immunoprecipitation and immunoblotting (not shown). The 46-kDa protein was identified as PAI-1, a protein previously shown to be induced by SPARC in subconfluent, but not in confluent cultures of nonangiogenic endothelial cells (Hasselaar et al., 1991). There was no apparent change in the production of SPARC, or of collagen, observed in these gels.
RNA was isolated from both angiogenic and nonangiogenic (control) cultures after treatment with native SPARC or with synthetic peptides derived from the amino acid sequence of murine SPARC. Initially, preparations of total RNA isolated from angiogenic cultures were probed with cDNAs corresponding to FN and PAI-1. SPARC and SPARC4-2R caused a significant inhibition of FN mRNA by 24 h. Consistent with the results from protein gels, PAI-1 mRNA was induced and F N mRNA was diminished within 24 h. When these blots were probed with cDNAs corresponding to other ECM gene products, we saw no effect on the levels of crl(1) collagen mRNA. However, TSP mRNA decreased with kinetics similar t o those of FN. In Fig. 3B, representative RNA transfers show the effects of SPARC, SPARC4-23, and SPARC2s4-273 on the expression of FN and PAI-1 mRNA in angiogenic BAE cell cultures. Addition of 20 pg/ml SPARC (Fig. 3B (lane 2 ) ) caused a decrease in the level of FN mRNA a t 24 h, which was more pronounced after 48 h (lane 6). PAI-1 mRNA was induced within 24 h (Fig. 3B (lane 2 ) ) and was sustained over a period of 48 h (lane 6). The peptide SPARC4-23 had activity similar to that of native SPARC (Fig. 3B (lanes 3 and 7)).
Treatment with this peptide resulted in a decrease in mRNA for FN and, concomitantly, an increase in that for PAI-1. By comparison, peptide SPARC2s1-273 from the C terminus of SPARC had no activity in these assays (Fig. 3B (lunes 4 and  8)).
Nonangiogenic BAE clones, those that did not form cords, had a markedly reduced response to exogenous SPARC or SPARC peptides. Under similar conditions, nonangiogenic cells became contact-inhibited and formed a typical cobblestone-like monolayer within days of plating. The confluent nonangiogenic BAE clones expressed low to undetectable  1 and 5 ) . After treatment with 20 pg/ml SPARC (lanes 2 and 6 ) , FN mRNA levels declined, whereas those for PAI-1 increased. Treatment with 0.4 mM SPARC4.2s (lanes 3 and 7) was also associated with a decrease in FN mRNA and an increase in PAI-1 mRNA. Treatment with 0.4 mM SPARC2s,.273 (lanes 4 and 8 ) resulted in mRNA levels similar to those of the PBS control. Angiogenic cultures of BAE cells were grown for 24 h in serum-free medium and subsequently exposed to various concentrations of SPARC (1 to 40 pg/ml) for 48 h. A shows a transfer of total RNA, hybridized sequentially with "'P-labeled probes as described in the legend to Fig. 3. B shows the results of densitometric scans of the autoradiogram presented in A. Absorbance values were corrected for total RNA on the blot by normalization to the 28 S rRNA signal. cDNA probes included TSP, PAI-1, FN, and 28 S rRNA (28 S ) . levels of TSP, FN, and PAI-1 mRNA and responded to exogenous SPARC with a markedly attenuated induction of PAI-1 as described previously (Hasselaar et ul., 1991). In Fig.  3A is shown RNA isolated from confluent nonangiogenic cultures. Control cultures expressed low levels of PAI-1 mRNA (Fig. 3A (lune 1 )) and undetectable levels of FN (not shown). Treatment with 30 pg/ml SPARC elicited a marginal increase in PAI-1 mRNA after 24 h (Fig. 3A (lane 2 ) ) .
Addition of varying amounts of SPARC protein to angiogenic cultures showed that the response seen in Fig. 3B was dose-dependent (Fig. 4). Cultures were grown for 24 h in serum-free media before treatment with SPARC. Levels of mRNA for TSP and FN decreased concurrently after addition of SPARC; significant diminution occurred in the presence of as little as 1 pglml(30 nM) SPARC. Quantitation of the RNA hybridization signals from these transfers revealed that TSP mRNA decreased approximately &fold, and that for FN, 4fold, after the addition of 10 pg/ml SPARC. In contrast, levels of PAI-1 mRNA increased 4-fold over the same concentration range. Preincubation of the SPARC preparation with a 10fold excess (by weight) of neutralizing anti-SPARC4_23 IgG resulted in a 50-60% recovery of FN and TSP mRNA levels and inhibited the induction of PAI-1 mRNA by a similar amount, compared with controls (not shown).
Experiments were conducted in the presence of peptides representing several domains of the SPARC protein; these results have been presented in Table I. Under serum-free conditions, SPARC and SPARC4-23 inhibited the production of FN and TSP mRNA by 70 and 65%, respectively, within 24 h. After 48 h of continuous treatment, cultures maintained a reduced production of FN and TSP mRNA; however, we were concerned about the effects of extended culture in the absence of serum. To confirm that the results obtained under serum-free conditions were valid, we conducted parallel experiments in 1% FCS to approximate conditions that are more compatible with long-term viability of these cultures. Culture in 1% FCS produced results similar to those conducted in the absence of FCS. After 48 h of continuous treatment with SPARC in the presence of FCS, levels of TSP mRNA declined to 12% of control levels. It was interesting that levels of TSP mRNA in cultures treated with SPARC4-23 appeared to recover by 48 h. Treatment of cultures with other peptides, including a scrambled isomer of SPARC5-23 (SPARC6-23scr), had no significant effect on the production of any mRNA for which we probed (Table I,  Results presented in Fig. 5 ( A and B ) show a dose dependence in the responses of FN and PAI-1 mRNA after treatment of angiogenic BAE cells with peptide SPARC4-23. PAI-1 mRNA increased &fold after 24 h with 0.1-0.2 mM peptide. The response of PAI-1 mRNA was saturable, and half-maximal induction occurred at approximately 0.05 mM. Halfmaximal inhibition of FN mRNA was also seen a t approximately 0.05 mM SPARC4-23, and maximal inhibition occurred a t approximately 0.2 mM. As shown with endothelial cells exposed to native SPARC (Fig. 4), FN and PAI-1 mRNA responded in a reciprocal fashion to SPARC4-2s. In addition, the half-maximal response of both FN and PAI-1 mRNA occurred a t a similar dose of SPARC (2-5 pg/ml; 0.15 pM) and of SPARC peptide (0.05 mM). The scrambled isomer (SPARC4-23~~r) had no apparent effect on the steady-state levels of FN or PAI-1 mRNA (Fig. 5C) or TSP mRNA (not shown).
Because of the relevance of plasminogen activators to various aspects of angiogenesis and tissue homeostasis, we continued to investigate the induction of secreted PAI-1 protein by SPARC in angiogenic cultures of BAE cells. An antiserum which recognizes bovine PAI-1 was used to probe blots of protein from media conditioned by cultures of angiogenic and nonangiogenic BAE cells. In accord with our analysis of PAI-1 mRNA, confluent nonangiogenic BAE cells expressed very low to undetectable levels of PAI-1 protein which increased only slightly in response to treatment with SPARC ( Fig. 3A and Hasselaar et al., 1991). In contrast, cultures of angiogenic BAE cells secreted low but detectable levels of PAI-1 protein which were increased considerably by treatment with SPARC. Fig. 6A is an immunoblot of PAI-1 secreted by angiogenic BAE cells cultured in the presence of increasing amounts of SPARC protein. Addition of exogenous SPARC correlated with the release of an immunoreactive protein that migrated with an apparent molecular weight of 46,000. Treatment of cultures with SPARC4-23 also stimulated the secretion of PAI-1 protein, whereas peptides derived from other domains of SPARC had significantly lower, or undetectable, activity (Fig.  6B). Different clones of angiogenic BAE cells displayed variability in the levels of PAI-1 released in response to SPARC (compare Fig. 6A with 6B), although all clones were stimulated; this effect was not investigated further.
To assess which cells in the culture were responding to exogenous SPARC, we localized PAI-1 protein by immunocytochemistry. Low levels of immunoreactive PAI-1 were associated with cords in cultures that were exposed to PBS (Fig. 7A (solid arrow)) or to a scrambled peptide (SPARC,-ascr) (Fig. 7B); the protein was not detected in cells associated with the monolayer. Both SPARC and SPARC,-z, induced a substantial increase in PAI-1 that was preferentially located in the endothelial cords (Fig. 7, C and D ) . By light microscopic examination the immunoreactivity appeared to coincide with the cytoplasm and the ECM of cells in cords. Cells associated with the monolayer of cultures treated with SPARC displayed a background level of reaction product and did not exhibit discernable levels of cytoplasmic immunoreactivity. The comparison of control (PBS or scrambled peptide) and SPARCtreated cultures provides an internal control for differences in path length, since the thickness of cords is unaltered in the presence of SPARC. Antisera against factor VI11 antigen (Fig.  7, E and F ) and against type I11 collagen (not shown) reacted with cells in the monolayer as well as with those in cords. The distribution or levels of these endothelial antigens did not appear to change following treatment with SPARC or SPARC peptides, and cells in cords were not stained selectively over those in the monolayer. By immunohistochemistry, these proteins do appear to concentrate slightly in cordforming regions, a result of multiple cell layers.
Due to concern over contamination by endotoxin in certain reagents, and the known sensitivity of bovine endothelial cells to this compound, we investigated the effects of bacterial lipopolysaccharide (LPS) on the gene products analyzed in this study. LPS, the pharmacologically active endotoxin from the surface of Gram-negative bacteria, is a broad spectrum activator of BAE cells and a potent inducer of PAI-1 (Sawdey et d, 1989). First, we measured the levels of LPS in various preparations of SPARC and found a range between 0.005 and 0.02 ng of LPS/pg of protein. We then determined the levels of LPS required for activation of PAL1 in BAE cells. As demonstrated by Hasselaar et ul. (1991), removal of FCS from the incubations inhibited the response to LPS and did not affect the response to SPARC. In angiogenic cultures of BAE cells, induction of mRNA for PAI-1 required 0.5 ng/ml LPS in the presence of 1% FCS (not shown). Fig. 8 shows the dosedependent activation of PAI-1 mRNA by LPS in the absence of FCS. The RNA was also hybridized with FN and TSP cDNA probes. Significant induction of PAL1 required between 10 and 50 ng/ml LPS (10 ng of LPS represented 50-200 times the LPS content measured in 10 pg of SPARC protein). TSP levels were unaffected by LPS below 50 ng/ml, and higher amounts of LPS were slightly stimulatory (2-fold in the presence of 100 ng/ml LPS). At the same time, LPS had a minimally inhibitory effect on mRNA for FN (Fig. 8B). These results indicated that contamination of the SPARC preparations by LPS was unlikely to account for the observed changes in the synthesis of FN or TSP, since LPS had either little effect or an effect opposite to that seen with SPARC. In addition, contaminating levels of LPS were too low to account for the observed induction of PAI-1 by SPARC (Fig. 8 and for 24 h in serum-free medium; cultures were then exposed to LPS (0.25-100 ng/ml ) for 48 h. A is a blot of total RNA from duplicate cultures, hybridized simultaneously with ["*P]cDNA probes for FN, TSP-1, and PAI-1. In B, hybridization signals from the blot in A were quantitated by scanning densitometry. Values are the average of duplicate samples normalized to the signal derived from the 28 S rRNA probe. Hasselaar et al., 1991). Peptides used in these assays contained no detectable LPS activity. Experiments were also conducted to determine whether SPARC could induce the synthesis of gelatin-degrading metalloproteases by angiogenic cultures of BAE cells. We tested the activities of several metalloproteases following SPARC or LPS treatment of BAE cells. Fig. 9 shows an example of gelatin zymography conducted with conditioned medium from these cultures. Cultures in the presence of PBS synthesized a 64-and a 92-kDa gelatinase in the presence (Fig. 9 ( l a n e 1 )) and absence of 1% FCS (Fig. 9 ( l a n e 4 ) ) . SPARC at 20 pg/ml had no apparent effect on the levels of either of these proteases (lanes 2 and 5). However, addition of LPS was associated with a significant increase in the 92-kDa gelatinase, although there was minimal change in the 64-kDa protease ( l a n e 3 ) . That serum factors were required to mediate this effect was demonstrated by the absence of induction of the 94-kDa gelatinase in serum-free cultures (Fig. 9 (compare lane 3 with  lane 6 ) ) . The gelatin-degrading activity of the conditioned medium could be attributed to metalloproteases, since enzymatic activity was inhibited by reducing agents such as dithiothreitol and by chelators of divalent cations such as EDTA and 1,lO-phenanthroline. The activity was insensitive to phenylmethylsulfonyl fluoride or N-ethylmaleimide (not shown). These results show that angiogenic BAE cells produce high levels of 64-and 94-kDa metalloproteases and that the 94-kDa metalloprotease was sensitive to LPS in a serumdependent manner. SPARC failed to activate the 94-kDa metalloprotease in either the presence or absence of serum. Thus, collagenase activity in BAE cells is not responsive to SPARC under the conditions employed in these experiments.

DISCUSSION
It is widely believed that the formation of new vessels by endothelial cells involves a series of events regulated by 1 2 3 4 5 6 -+ --+spmc + --+LPs + + + ---m s " FIG. 9. Effect of SPARC and endotoxin (LPS) on the secretion of gelatin-degrading metalloproteases by BAE cells. Experiments were conducted in the presence or absence of 1% FCS. Conditioned media were collected after 48 h of treatment and mixed directly with SDS-PAGE sample buffer. Samples were subjected to substrate gel electrophoresis under nonreducing conditions as described under "Experimental Procedures." Gels were stained with Coomassie Blue, and gelatinase activity (arrows marked I and I I ) was identified as clear bands on a background of undigested gelatin. and 4 ) , SPARC, 20 pg/ml (lanes 2 and 5), or LPS, 50 ng/ml (lanes 3 and 6). The gelatinase activities were identified as metalloproteases by sensitivity to reduction, ethylenediaminetetraacetic acid, and 1,lO-phenanthroline (not shown).

Treatments included PBS (lanes I
growth factors and ECM proteins. In a previous study, we proposed that SPARC might play an important role in this process, since it is synthesized a t high rates when endothelial cells organize into cords and tubes in uitro (Iruela-Arispe et al., 1991a, 1991b. We were interested in extending this observation to delineate a mechanism by which SPARC could act during the formation of endothelial cords and to identify domains within the SPARC molecule that could be involved in this function. Because the acquisition of an angiogenic phenotype is associated with increased secretion of SPARC, we were interested in studying possible feedback pathways that could be potentiated by the enhanced availability of this protein during angiogenesis in uitro. SPARC is a Ca2+-binding protein secreted by cultured endothelial cells isolated from normal blood vessels (Sage et al., 1981(Sage et al., , 1984 and by a variety of cells in the vessel wall following vascular injury (Raines et al., 1992). When added to cultured cells, SPARC inhibited the spreading of endothelial cells on plastic and collagenous substrates (Sage et al., 1989b). Subsequently, active sites in the molecule were identified by the use of synthetic peptides. Anti-spreading activity was located at the N terminus (SPARC4-2s) and was blocked by N-terminal-specific antibodies (Lane and Sage, 1990). A portion of the C terminus, SPARC2s4-273, also displayed antispreading activity and, in addition, competed for the binding of SPARC to collagens. These studies indicated that SPARC might function by blocking the interaction of cells with ECM, but the possibility of a direct action on cells mediated by surface receptors could not be ruled out. A recent study has demonstrated that SPARC increased the production of PAI-1 by subconfluent proliferating endothelial cells but had little effect on contact-inhibited nonangiogenic endothelial monolayers (Hasselaar et al., 1991). Since the induction of PAI-1 was apparently independent of overt changes in cell shape or adhesion, other mechanisms of signal transduction were implied.
In the present study we provide evidence that the synthesis of a small subset of ECM proteins, including FN, TSP, and PAI-1, can be modulated by SPARC and that the mechanism of regulation includes alterations in the steady-state levels of the respective mRNAs, as well as in amounts of secreted protein. The effects were most pronounced in BAE cells in an activated state of cellular migration and remodeling, as typically seen in angiogenic cultures. The response was muted or absent in nonangiogenic (contact-inhibited) endothelial cell monolayers. In addition, neutralizing anti-SPARC peptide antibodies directed against amino acids 4-23 significantly diminished the effects of exogenous SPARC. However, our experiments with anti-SPARC antibodies have generally proven more difficult to control than those in which SPARC peptides were utilized. It is possible that the high levels of endogenous SPARC made by these cultures, or complexities in the experimental model not currently understood, contributed to these problems.
Both SPARC and the N-terminal peptide SPARC4-23 had similar activities with respect to the regulation of FN, TSP, and PAI-1, an observation supportive of previous data which indicated that the N terminus of the protein might contain a bioactive domain. In the present experiments, SPARC4-23 was approximately 300-fold less active as compared with the native protein. Loss of activity has been a consistent finding in a variety of studies on peptide homologs of active sites in large proteins (Lane and Sage, 1990 and references therein). Although it is not clear what variables are most important for the retention of optimal peptide activity, it is likely that peptide fragments lose conformational constraints contributed by neighboring and distant structures within multidomain proteins. To control for the effects of peptide sequence and concentration, we show that peptides from other regions of SPARC, and a scrambled version of the N-terminal sequence, had little or no activity.
To determine the location of cells responding to SPARC in angiogenic cultures, we investigated the production of PAI-1 protein by immunocytochemical methods. These experiments revealed that cells associated with cords selectively expressed PAI-1, a result consistent with the active remodeling that is characteristic of these cells. PAI-1 immunoreactivity associated with endothelial cords was enhanced after treatment of the cultures with exogenous SPARC. We found no evidence to suggest that cells in the monolayer contributed significantly t o the production of PAI-1, in the presence or absence of exogenous SPARC. This observation is significant, because it is the cells that form cords and tubes, rather than quiescent cells in the monolayer, that produce SPARC in these cultures (hela-Arispe et al., 1991a). Endothelial cells involved in active remodeling might therefore produce proteins that selectively alter their extracellular environment and might in turn respond to this new environment in an autocrine manner. Other endothelial cells not directly involved in remodeling do not appear to respond. Although it is not clear why the cells in the monolayer were unaffected, it may be important that the responding cells were actively migrating and were not spread, a phenotype that is associated with a wide range of endothelial responses (Ingber et al., 1987;Ingber and Folkman, 1989;. Such differences might in fact be highly significant in regulating the extent of angiogenic phenomena in uiuo. FN and TSP are major secretory products of endothelial cells in culture (Sage et al., 1981;Raugi et al., 1982) and are important mediators of endothelial cell attachment (Murphy-Ullrich and Hook, 1989;Taraboletti et al., 1990). FN is present early in tissues undergoing angiogenesis and is thought to provide a provisional matrix for the migration of endothelial cells (Risau and Lemmon, 1988 and references therein). Synthesis of FN is increased in the presence of angiogenic factors (Jaye et al., 1985;Madri et al., 1988) and decreased by angiostatic steroids (Folkman and Ingber, 1987). Cell spreading on FN-coated plates resulted in elevation of cytoplasmic pH and increased rates of cell proliferation (Ingber et al., 1990). In addition, there is evidence that binding of FN to cells can activate cell surface Na+/H+ antiporters, a phenomenon that might explain some cellular responses to this protein (Ingber, 1990;Ingber et al., 1990). In the present study we demonstrated that exogenous SPARC inhibited the synthesis of FN mRNA and protein. Although it is not clear whether the activity of SPARC was direct or indirect, the functional significance of a decrease in extracellular FN could include the modulation of endothelial attachment, migration, and/or proliferation.
In contrast to FN, TSP does not appear to be associated with the earliest stages of endothelial budding but with more advanced capillaries, after basement membrane components have been deposited (O'Shea, 1987;O'Shea and Dixit, 1988).
TSP has also been shown to inhibit endothelial proliferation (Bagavandoss and Wilks, 1990) and to block the mitogenic response of capillary endothelial cells to serum (Taraboletti et al., 1990). In culture, the biosynthesis of TSP is highest in endothelial cells that are prevented from spreading  as well as in rapidly proliferating endothelial cells, but it is diminished as the cells reach confluence (Raugi et al., 1982;Mumby et al., 1984). Although the regulation of TSP has been studied in smooth muscle cells (Majack et al., 1985) and fibroblasts (Donoviel et al., 1990), information concerning its control in endothelial cells is minimal. Recently, we demonstrated that TSP has anti-angiogenic effects in our BAE cultures and might contribute to the stabilization of endothelial tubes (hela-Arispe et al., 1991~). In addition, a truncated form of TSP has been shown to suppress angiogenesis induced by basic fibroblast growth factor (Good et al., 1990). The present study shows that SPARC can functionally diminish mRNA for TSP in angiogenic BAE cells. As previously noted, SPARC and TSP inhibit the proliferation of endothelial cells in vitro (Funk and Sage, 1991;Taraboletti et al., 1990;Bagavandoss and Wilks, 1990); SPARC and TSP might therefore act cooperatively to regulate proliferative and adhesive properties of endothelial cells (reviewed in Sage and Bornstein, 1991). Additional members of the TSP family have recently been identified (e.g. ThbB;, but the expression of these gene products was not addressed in the present study. The induction of PAI-1 by SPARC has several implications for hemostasis and tissue remodeling. PAI-1 is a specific and potent inhibitor of both urokinase and tissue-type plasminogen activators (Loskutoff et al., 1988) and therefore plays an important role in maintaining the nonthrombogenic properties of the vessel wall. Plasmin has been implicated as an activator of latent TGF-fi (Sato and Rifkin, 1989;Lyons et al., 1990), and PAI-1 can inhibit the production of active TGF-fi in cell culture (Sato and Rifkin, 1989). Because TGFfi is known to increase endothelial PAI-1 and decrease production of plasminogen activators (Roberts et al., 1988), an effective feedback loop might exist among these factors. Activated plasmin has a demonstrated proteolytic activity against SPARC (Sage et al., 1984) and several ECM components. By inhibiting plasminogen activators, PAI-1 could function to protect ECM proteins from these proteases (Saksela and Rifkin, 1988) and to stabilize cellular attachments to matrix (Ciambrone and McKeown-Longo, 1990). In uiuo, increases in protease inhibitors induced by SPARC could function to protect specific matrix components. PAI-1 itself, by inhibiting local matrix turnover, might also cause a graded reduction in the angiogenic response.
ECM components act at many different levels to regulate angiogenesis. The binding of ECM to cell surface receptors can lead to the activation of a number of intracellular signal transduction pathways (Ingber and Folkman, 1989) and result in both direct and indirect modulation of cell shape, migration, and proliferative response. In addition, the stabilization or sequestering of specific growth factors by matrices of various compositions can strongly affect the distribution and activity of these factors (Rifkin and Moscatelli, 1989). Several growth factors that can potentially regulate angiogenesis include basic fibroblast growth factor (Folkman and Klagsbrun, 1987), TGF-/3 (Roberts et al., 1986), and platelet-derived growth factor-BB (PDGF-BB) (Fiegel et al., 1991;Madri et al., 1991). Although these morphoregulatory factors are not structurally related, they appear able to promote endothelial cell migration. They are also capable of promoting synthesis or turnover of ECM and are differentially sequestered by specific components of the ECM.
SPARC has demonstrated activities that are consistent with its classification as a morphoregulatory protein . Since it inhibits the spreading of endothelial cells and fibroblasts on collagen or plastic substrates (Sage et al., 1989b;Lane and Sage, 1990) and reduces the number of focal contacts made by BAE cells in culture (Murphy-Ullrich et al., 1991), SPARC could affect migratory or adhesive activities of the endothelium in viuo. It is not clear whether SPARC is a normal constituent of the ECM; however, the protein does interact with a variety of ECM components, a property that could regulate its distribution or stability. SPARC binds to several collagens, including types I, 111, IV, and V (Termine et al., 1981;Sage et al., 1989b, Nischt et al., 1991 and, to a lesser extent, TSP and FN (Clezardin et al., 1988;Sage et al., 1989b). Recently, SPARC was also shown to bind the B chain of PDGF (Raines et al., 1992). Specific binding of PDGF-B isoforms would suggest a novel function for SPARC in the regulation of this potent angiogenic and growth-regulatory factor.
Whereas aortic endothelial cells, such as those used in the present study, are not responsive to PDGF, capillary endothelial cells are stimulated to proliferate and migrate by PDGF-BB homodimers (Smits et al., 1989;Bar et al., 1989). It is therefore likely that endothelial cells derived from capillary beds would provide an optimal system to study the activities of both SPARC and We suggest that endothelial cells, activated by specific stimuli, have the ability to regulate angiogenesis through their responses to both positive and negative feedback pathways. During angiogenesis, the invasive endothelial phenotype is initiated by the production of matrix-degrading proteases and diminution of specific protease inhibitors such as PAI-1 (Montesano et al., 1990). However, the production of FN, type I collagen, and other substrate macromolecules that are permissive for endothelial migration and proliferation is also required. The angiogenic cycle is completed when basement membrane components are assembled that stabilize capillary structure and inhibit further migration and proliferation. Although many of the functions proposed for SPARC remain unproven in viuo, activities attributed to the protein based on in vitro experiments are consistent with actions a t sites of tissue remodeling. During angiogenesis, SPARC could function to regulate endothelial attachment and proliferation, possibly through the binding of specific growth factors. After the initiation of angiogenesis, SPARC could regulate the production or stability of ECM by modulating the synthesis of FN, TSP, and PAI-1.

PDGF-BB.
It is generally accepted that angiogenic factors such as basic fibroblast growth factor and TGF-/I act in part by fostering endothelial cell migration. Less understood are local control mechanisms that affect the extent of the signal and the magnitude of the response. Previous studies on the activity of SPARC have demonstrated diverse effects on attachment, spreading, proliferation, and the expression of PAI-1 by monolayers of nonangiogenic BAE cells. The results presented in this study indicate that release of SPARC by angiogenic endothelial cells might function to control a small subset of genes during the process of vascular remodeling. The fact that several extracellular proteins (including type I collagen and several metalloproteases) are not affected by SPARC indicates that the effects are specific and possibly confined to a limited set of genes. We suggest that SPARC derived from endothelial cells, platelets, or other cells in the vicinity of an angiogenic stimulus acts in part to control the synthesis of proteins involved in the modulation of endothelial cell morphology, matrix turnover, and proliferation.