Angiostatic peptides use plasma fibronectin to home to angiogenic vasculature

Abstract

A group of angiogenesis inhibitors are derived from fragments of extracellular matrix or blood proteins. Endostatin, antithrombin, and anastellin are members of this group of substances. The plasma adhesion proteins fibronectin and vitronectin serve as cofactors for these three antiangiogenic proteins. Anginex is a synthetic 33-amino acid peptide that was originally modeled to reproduce the β-sheet structure of antiangiogenic proteins. Here, we show that anginex initiates fibronectin polymerization and is inactive in mice that lack plasma fibronectin. Anginex shares these characteristics with anastellin. Fluorescein-labeled anginex and anastellin specifically localized in angiogenic vessels in vivo. This localization was dependent on plasma fibronectin and inhibited by an Arg-Gly-Asp peptide. Thus, anginex shares with several physiological angiogenesis inhibitors a dependence on plasma adhesion proteins. The role of the adhesion protein interaction apparently is to form integrin-binding complexes that deliver the antiangiogenic proteins to sites of angiogenesis. This functional convergence of several antiangiogenic factors has important implications for antiangiogenic therapies.

A ^balance between pro- and antiangiogenic factors tightly regulates the growth and regression of blood vessels. Under certain pathological conditions, as in diabetic retinopathy, rheumatoid arthritis, or during tumor growth, the balance is disturbed to induce excessive formation of new blood vessels (1). An increasing number of proteolytic fragments of extracellular matrix and blood proteins are emerging as important negative regulators of angiogenesis. Some of these physiological angiogenesis inhibitors, such as angiostatin (2) and endostatin (3), are in clinical trials as inhibitors of pathological blood vessel formation.

Anginex is a synthetic 33-mer peptide designed to incorporate structural features shared among several physiological angiogenesis inhibitors (4). Anginex is composed of numerous hydrophobic and basic amino acids, and it was recently reported that the bioactive form of anginex has a β-sheet structure (5). Anginex inhibits angiogenesis and tumor growth (4, 6, 7) and causes apoptosis in cultured endothelial cells (4). How anginex, or the physiological angiogenesis inhibitors, exert their effects in vivo is poorly understood.

Previous work from our laboratory identified a common feature of three physiological angiogenesis inhibitors: they depend on adhesion molecules for their in vivo activity (8). Anastellin, a 10-kDa fragment of the first type III repeat of fibronectin (9, 10) that is both antiangiogenic and antimetastatic (11, 12), interacts with fibronectin to produce a fibrillar, polymeric fibronectin matrix (10). The antiangiogenic activity of anastellin is manifested in normal mice but is not observed in mice that lack plasma fibronectin (8). Antithrombin is active in the plasma fibronectin-deficient mice but inactive in mice lacking vitronectin, whereas endostatin requires both fibronectin and vitronectin for its antiangiogenic activity (8). The fibronectin dependence of anastellin and the vitronectin dependence of antithrombin correlate with the binding of these proteins to fibronectin and vitronectin, respectively (9, 10, 13). Given the similarities of the structure and activity of anginex and the antiangiogenic proteins, we explored the possible role of adhesion proteins in the antiangiogenic activity of anginex.

We show here that the antiangiogenic activity of anginex is suppressed in mice that lack plasma fibronectin. We also show that intravenously injected anginex and anastellin specifically home to sites of angiogenesis in mice with normal plasma fibronectin but not in mice deficient in plasma fibronectin. Coinjecting an integrin-binding peptide inhibits the homing in normal mice. Thus, these antiangiogenic proteins form complexes with fibronectin, and the complexes bind to integrins that are overexpressed in angiogenic vessels. This knowledge of the shared adhesion protein dependence of antiangiogenic factors and their convergence on integrins will facilitate the design of improved antiangiogenic therapies.

Materials and Methods

Materials. Anginex was synthesized by using N-(9-fluorenylmethoxycarbonyl)-l-amino acid chemistry with a solid-phase synthesizer and purified by using HPLC. A control peptide comprising the central heparin-binding region of vitronectin (amino acid residues 348–361), KKQRFRHRNRKGYR (14), was also synthesized. The composition of the peptides was confirmed by mass spectrometry. We also used some anginex provided by K. H. Mayo (University of Minnesota Health Sciences Center, Minneapolis). Fluorescein-conjugated anginex was synthesized as described for other peptides (15). Anastellin was prepared as a recombinant His-tagged protein in bacteria and purified as described (10). Human plasma fibronectin was purchased from Chemicon and Gly-Arg-Gly-Asp-Ser-Pro (GRGDSP) from Calbiochem.

To prepare fluorescein-conjugated anastellin, a cysteine residue was added to the N terminus of anastellin by using the QuikChange mutagenesis kit (Stratagene) and following the manufacturer's protocol. The purified protein was labeled with fluorescein maleimide (Molecular Probes) according to the manufacturer's protocol at a molar fluorescein/protein ratio of 10:1 and overnight incubation at 4°C. The polymerizing activity of anastellin, anginex, and their conjugates was confirmed with a fibronectin polymerization assay (12, 16). The polymerization was quantified by measuring turbidity, and polymerized material was analyzed by SDS/PAGE. Fluorescein-conjugated fibronectin was prepared by adding 700-fold molar excess of fluorescein isothiocyanate to a 5 mg/ml fibronectin solution in 0.5 M sodium carbonate buffer (pH 9.5). The components were allowed to react for 30 min to 1 h while rotating end-over-end, followed by dialysis against a large excess of PBS to remove unconjugated fluorescein. Dialysis cassettes were from Pierce.

Mice and Matrigel Plug Angiogenesis Assay. Two transgenic plasma fibronectin-deficient mouse lines [C57BL/6-Fn(fl/fl); Alb-Cre⁺ and C57BL/6-Fn(fl/fl); Mx-Cre⁺] (17) were used. Mice carrying Alb-Cre have undetectable levels of plasma fibronectin 2–3 months after birth. In the Fn(fl/fl); Mx-Cre⁺ mice, the deletion of the fibronectin gene was induced in the liver by using poly(I):poly(C) (17). The vitronectin knockout mouse line originates from David Loskutoff's laboratory at The Scripps Research Institute (La Jolla, CA) (18). Wild-type C57BL/6 mice were used as controls for the vitronectin knockout mice (Harlan–Sprague–Dawley). Plasma levels of fibronectin and vitronectin were measured by immunoblotting (8).

Angiogenesis assays used bFGF-impregnated Matrigel plugs as described (8). In treatment experiments, 200 μg of anginex or 800 μg of anastellin in 200 μl of PBS was administered as daily i.p. injections for 7 days, starting on the day after the plug implantation (day 1). The mice were killed on day 8, and the plugs were processed for immunohistochemical analysis.

Immunohistochemistry. Blood vessels were visualized in histological sections with anti-CD31 antibodies (BD Biosciences). Fluorescent secondary antibodies (Alexa Fluor 488 and Alexa Fluor 594) were from Molecular Probes. Nuclei were visualized with DAPI-containing mounting medium (VECTASHIELD, Vector Laboratories).

Anginex and Anastellin Tissue Localization. Fluorescein-conjugated anginex (250 μg) or anastellin (750 μg) in 250–500 μl of PBS was i.v. injected into wild-type C57BL/6 mice, plasma fibronectin-deficient Mx-Cre mice, or their untreated littermates. The mice were killed 2 h later, and their Matrigel plugs and tissues were processed for histology.

Results

Anginex Polymerizes Fibronectin in Vitro. Anginex is similar to anastellin in that they both are relatively small β-sheet structured molecules with numerous exposed hydrophobic residues (5, 16). When mixed with fibronectin, anastellin induces the formation of fibronectin fibrils (Fig. 1A; 10). Because of the similarities between anastellin and anginex, we tested the effect of anginex on fibronectin. We found that mixing anginex with fibronectin caused increased turbidity (Fig. 1B). SDS/PAGE analysis of the insoluble polymer showed that anginex coaggregated with fibronectin (Fig. 2). Densitometric quantification of the gel bands indicated that the precipitate contained ≈50–100 moles of anginex per mole of fibronectin. For anastellin, this ratio was ≈5–10.

Fig. 1.

Polymerization of fibronectin by anastellin (A) and anginex (B). Various concentrations of anastellin or anginex were mixed with equal volumes of 0.5 mg/ml fibronectin in PBS. The polymerization of the protein mix was monitored by measuring the optical density at 590 nm. Because anginex has inherent turbidity at high concentrations, the optical density of anginex alone was monitored, and the values were subtracted from the anginex-fibronectin measurements. Anastellin was treated in the same manner and is shown here as a positive control.

Fig. 2.

Anginex and anastellin coaggregate with fibronectin. SDS/PAGE analysis of complexes of fibronectin with anginex (A) and anastellin (B). Fibronectin (1 mg/ml) was mixed with a 150-fold (anastellin) or 300-fold (anginex) molar excess of anginex or anastellin in PBS (lane 1). The aggregates were spun down, and the pellets were washed with PBS (lanes: 2, supernatant; 3–6, washes; 7, washed pellets). Control samples with anginex or anastellin alone (lane 8) or fibronectin alone (lane 9) were treated in the same manner as the complexes to monitor adsorption of the proteins to the tubes. Serial dilutions of known concentrations of the three proteins were run separately (not shown) to allow quantification of the proteins in the samples. Background binding to the tube (lane 8) was subtracted in the quantification of anginex and anastellin.

Anginex Requires Plasma Fibronectin for in Vivo Activity. Previous results demonstrate that anastellin is inactive in mice that lack plasma fibronectin as a result of a liver-specific conditional deletion of the fibronectin gene (8). The antiangiogenic activity of anginex also was severely impaired in plasma fibronectin-deficient mice (Fig. 3A). When the mice were treated with PBS as a control, the number of new blood vessels in Matrigel plugs was similar in fibronectin-deficient and control mice. In contrast, daily injections of 200 μg of anginex caused a significant reduction in angiogenesis in normal mice but were ineffective in mice lacking plasma fibronectin (P < 0.01). In contrast with the results obtained with the plasma fibronectin-deficient mice, we found that anginex is fully active in vitronectin-null mice (Fig. 3B). Previous work has shown that anastellin is also active in these mice (8). Thus, anginex displays a fibronectin dependence similar to that of anastellin and, like anastellin, does not require vitronectin for its activity.

Fig. 3.

Anginex requires plasma fibronectin, but not vitronectin, for in vivo activity. Mice lacking plasma fibronectin (pFN–) and their normal littermates (pFN+) (A), or vitronectin-null mice (VN–) and wild-type control mice (VN+)(B), were injected with Matrigel supplemented with bFGF. The mice were treated with daily i.p. injections of 200 μg of anginex in 200 μl PBS or PBS alone for 7 days. The number of blood vessels per microscopic field (×400) is shown. The P values represent the significance level of the differences between the treatment groups. NS, not significant.

Anastellin and Anginex Home to Angiogenic Sites Guided by Fibronectin. We hypothesized that anastellin and anginex may recruit fibronectin from the plasma and that the resulting fibronectin complex delivers them to angiogenic vasculature through integrin binding. Because the α5β1 and αvβ3 integrins, both of which bind fibronectin, are selectively expressed in angiogenic endothelial cells (19–22), the complexes could home to angiogenic vessels by binding to these integrins.

To begin to test the above predictions, we labeled anastellin and anginex with fluorescein and tested the ability of the labeled proteins to accumulate in angiogenic vessels after an i.v. injection. We found that both fluorescent anastellin and anginex accumulated in Matrigel plugs, colocalizing with the endothelial cell marker, CD31 (Fig. 4 A and E). Matrigel plugs implanted in the plasma fibronectin-deficient mice did not accumulate significant levels of anastellin or anginex fluorescence (Fig. 4 B and F), but the homing of these molecules was robust in their fibronectin-positive littermates (Fig. 4 C and G) and was restored in fibronectin-deficient mice that had been i.v. injected with purified fibronectin (Fig. 4 D and H). No anastellin or anginex fluorescence was present in the vessels of other major tissues (shown for skin, heart, and brain in Fig. 4 I–K and M–O), with the possible exception of the liver (Fig. 4 L and P). The reticuloendothelial system in the liver (and spleen) is known to take up fibronectin complexes (23). We also tested the homing to Matrigel plugs of anastellin–fibronectin and anginex–fibronectin complexes in which fibronectin was fluorescently labeled, and we found fluorescent complexes colocalizing with CD31 in the Matrigel plugs. Only occasional cells showed fluorescence when fluorescent fibronectin alone was injected as a control (data not shown). These results show that anastellin and anginex home to angiogenic vasculature in complexes with fibronectin.

Fig. 4.

Anastellin and anginex home to angiogenic vasculature in vivo. Normal or plasma fibronectin-deficient mice implanted with Matrigel plugs 8 days earlier were i.v. injected with fluorescein-labeled anastellin (750 μg; A–D and I–L) or anginex (250 μg; E–H and M–P), and the mice were killed 2 h later. Endothelial cells in the plugs and in various tissues were visualized with anti-CD31 staining. Yellow reveals colocalization of the injected fluorescent protein (green) and endothelium (red). Nuclei were counterstained with DAPI (blue). Anastellin and anginex fluorescence is observed in the plugs from wild-type mice (A and E). The vessels in the plugs of pFN–mice showed no significant anastellin (B) or anginex (F) fluorescence, whereas the plugs of the pFN+ littermates of the pFN–mice were positive (C and G). i.v. injection of 0.5 mg of fibronectin into the pFN–mice restored the homing of anastellin (D) and anginex (H) to the plugs. No anastellin or anginex fluorescence was detected in the s.c. tissue (the site of the Matrigel plug implantation; I and M), the heart (J and N), or the brain (K and O). Faint green fluorescence was present in the liver (L and P). (Original magnification: ×400.)

To explore the possible role of integrins in the homing of anastellin and anginex to angiogenic vasculature, we tested the effect of an Arg-Gly-Asp (RGD) peptide on the homing. α5β1 and αvβ3 are RGD-directed integrins, and their binding to fibronectin can be inhibited with RGD-containing peptides (24). If the binding of anastellin and anginex to angiogenic vessels is mediated by fibronectin, it should be possible to inhibit it with RGD peptides. This is indeed what we found. Neither fluorescent anastellin nor anginex was observed in Matrigel pellets when these proteins were coinjected with the pentapeptide GRGDSP (Fig. 5). An unrelated control peptide had no effect. Other controls demonstrated that the GRGDSP peptide had no effect on the ability of anastellin or anginex to polymerize fibronectin at the concentration used in the injections (not shown).

Fig. 5.

The homing of anastellin and anginex to angiogenic vasculature is inhibited by an RGD-containing peptide. Mice bearing Matrigel plugs were i.v. injected with fluorescein-labeled anastellin (A–C) or anginex (D–F) as in Fig. 4 together with PBS (A and D), 10 mg of GRGDSP (B and E), or an equal molar amount of an unrelated peptide (C and F). Little or no fluorescence is observed in the Matrigel plugs of mice coinjected with the RGD-containing peptide.

Discussion

The results described here show that the synthetic peptide anginex shares with several physiological angiogenesis inhibitors a dependence on a plasma adhesion protein cofactor. Anginex shares a particular similarity with anastellin in that both factors coaggregate with fibronectin and depend on plasma fibronectin for their antiangiogenic activity. We also provide evidence that the fibronectin complexes serve as a delivery vehicle for anastellin and anginex; the complexes bind to RGD-directed integrins in angiogenic vessels, providing a specific targeting function allowing anastellin and anginex to home in on angiogenic endothelium.

Fibronectin is secreted as a soluble protein but has the propensity to form insoluble fibrillar aggregates that incorporate into extracellular matrix. The fibril formation is the result of fibronectin–fibronectin interactions at cryptic binding sites, exposed by the unfolding of the molecule under stretch forces exerted by cells (25, 26). Anastellin is a fibronectin fragment that encompasses a site that is important in fibronectin fibril formation (10). Anastellin seems to be capable of initiating the fibronectin polymerization process, but a denaturing treatment (10) or stretching (R. Martin, M.E.A., T. Volberg, B. Geiger, E.R., and D. Hanein, unpublished data) is needed to complete the process. According to the present results, anginex shares the fibronectin polymerizing activity with anastellin. There is no sequence homology between anastellin and anginex that could explain their ability to specifically interact with fibronectin, but their β-sheet structures containing exposed hydrophobic residues are similar (5, 16). It seems that peptides with such a structure can interact with fibronectin, bringing about its aggregation.

The ability to interact with fibronectin is important to the antiangiogenic activity of anastellin and anginex. Anastellin requires plasma fibronectin to be active in vivo (8), and the present work shows that the same is true of anginex. We have shown earlier that antithrombin requires vitronectin, also a plasma adhesion protein, and that endostatin activity is impaired if either fibronectin or vitronectin is absent (8). Thus, we have tested four antiangiogenic peptides or proteins for the dependence of their in vivo activity on a plasma adhesion protein. The antiangiogenic activity has turned out to be adhesion protein-dependent in each case. It is remarkable that the adhesion protein binding and requirement for in vivo activity previously shown for anastellin, antithrombin, and endostatin extends to a completely de novo designed molecule, anginex. This finding underscores the central role of adhesion proteins in the functions of antiangiogenic proteins and suggests common features in the mechanism of action of various antiangiogenic peptides and proteins.

The aggregation of anastellin and anginex with fibronectin integrates these peptides in complexes that can bind to integrins. Anastellin and anginex do not directly bind to integrins. Neither peptide contains the integrin-binding RGD tripeptide. Moreover, although various types of cells, including endothelial cells, attach to anastellin and anginex, the attachment is not integrin-mediated (J.P., M.E.A., and E.R., unpublished data). The RGD sequence of fibronectin endows the complexes with the ability to interact with integrins, and such interactions are reinforced by the multivalency of the RGD presentation in the complexes. RGD-containing peptides and RGD peptidomimetics are commonly used to provide drug complexes, viruses, and nanoparticles with the ability to selectively home to angiogenic vasculature (22). As demonstrated here, their ability to complex with fibronectin in plasma provides anastellin and anginex, and likely other antiangiogenic proteins capable of interacting with adhesion proteins as well, a built-in RGD-dependent homing mechanism.

Although the integrin binding of the fibronectin complexes provides a means for the antiangiogenic peptides to home to angiogenic vessels, their proximate mechanism of action is not fully understood. The antiangiogenic peptides and proteins may or may not have any further function beyond bringing an aggregate of adhesion protein to angiogenic vessels. The adhesion protein aggregate could be the proximal antiangiogenic effector, or it could only serve a carrier function with the antiangiogenic factor as the effector. It has been shown that interference with the function of α5β1 and αvβ3 integrin results in apoptosis of endothelial cells and inhibition of angiogenesis (19, 20, 27). Furthermore, synthetic polymers containing RGD sequences can inhibit angiogenesis in vivo (28), and RGD peptides have been reported to have a proapoptotic activity in vitro (29–31). Therefore, exposed RGD sequences of fibronectin–anastellin and fibronectin–anginex complexes could be responsible for both the homing and apoptotic activities, leaving formation of the complexes as the only anastellin/anginex function. However, anastellin and anginex also have proapoptotic and cytotoxic effects that may be direct rather than mediated by adhesion proteins (7, 32). The exact mechanism of cell killing by anastellin and anginex, as well as by other antiangiogenic proteins, requires further study.

The present results provide a common mechanism whereby these proteins can specifically zero in on angiogenic vasculature and identify integrins overexpressed in angiogenic vasculature as the common target. Given the importance of angiogenesis in development, regeneration, and cancer, our results represent a significant advance in clarifying the poorly understood mechanisms of action of the various antiangiogenic proteins.