Distinct antitumor properties of a type IV collagen domain derived from basement membrane.

Vascular basement membrane is an important structural component of blood vessels. During angiogenesis this membrane undergoes many alterations and these changes are speculated to influence the formation of new capillaries. Type IV collagen is a major component of vascular basement membrane, and recently we identified a fragment of type IV collagen alpha2 chain with specific anti-angiogenic properties (Kamphaus, G. D., Colorado, P. C., Panka, D. J., Hopfer, H., Ramchandran, R., Torre, A., Maeshima, Y., Mier, J. W., Sukhatme, V. P., and Kalluri, R. (2000) J. Biol. Chem. 275, 1209-1215). In the present study we characterize two different antitumor activities associated with the noncollagenous 1 (NC1) domain of the alpha3 chain of type IV collagen. This domain was previously discovered to possess a C-terminal peptide sequence (amino acids 185-203) that inhibits melanoma cell proliferation (Han, J., Ohno, N., Pasco, S., Monboisse, J. C., Borel, J. P., and Kefalides, N. A. (1997) J. Biol. Chem. 272, 20395-20401). In the present study, we identify the anti-angiogenic capacity of this domain using several in vitro and in vivo assays. The alpha3(IV)NC1 inhibited in vivo neovascularization in matrigel plug assays and suppressed tumor growth of human renal cell carcinoma (786-O) and prostate carcinoma (PC-3) in mouse xenograft models associated with in vivo endothelial cell-specific apoptosis. The anti-angiogenic activity was localized to amino acids 54-132 using deletion mutagenesis. This anti-angiogenic region is separate from the 185-203 amino acid region responsible for the antitumor cell activity. Additionally, our experiments indicate that the antitumor cell activity is not realized until the peptide region is exposed by truncation of the alpha3(IV)NC1 domain, a requirement not essential for the anti-angiogenic activity of this domain. Collectively, these results effectively highlight the distinct and unique antitumor properties of the alpha3(IV)NC1 domain and the potential use of this molecule for inhibition of tumor growth.

until the peptide region is exposed by truncation of the ␣3(IV)NC1 domain, a requirement not essential for the anti-angiogenic activity of this domain. Collectively, these results effectively highlight the distinct and unique antitumor properties of the ␣3(IV)NC1 domain and the potential use of this molecule for inhibition of tumor growth.
The development of new blood vessels from pre-existing ones is generally referred to as angiogenesis (1). In the adult, new blood vessels arise via angiogenesis, a process critical for normal physiological events such as wound repair, the ovarian cycle, and endometrium remodeling (2). However, uncontrolled neovascularization is associated with a number of pathological disorders including diabetic retinopathy, rheumatoid arthritis, as well as tumor growth, and metastasis (3,4). Tumor growth and metastasis require angiogenesis (1), and this process is pivotal to the survival and subsequent growth of solid tumors beyond a few mm 3 in size (3). Expansion of tumor mass occurs not only by perfusion of blood through the tumor but also by paracrine stimulation of tumor cells by several growth factors and matrix proteins produced by the new capillary endothelium (3). Recently, a number of angiogenesis inhibitors have been identified, namely angiostatin, endostatin, restin, and pigment epithelium-derived factor (5)(6)(7)(8).
Basement membranes are thin layers of a specialized extracellular matrix that provide the supporting structure on which epithelial and endothelial cells grow and that surround muscle, fat, etc. (9). They are always associated with cells, and it has been well demonstrated that basement membranes do not only provide a mechanical support but also influence cellular behavior such as differentiation and proliferation. The major macromolecular constituents of basement membranes are type IV collagen, laminin, heparan sulfate proteoglycans, fibronectin, and entactin (10). Vascular basement membrane constitutes an insoluble structural wall of newly formed capillaries and is speculated to play an important role in regulating pro-and anti-angiogenic events (11). In general, type IV collagen promotes cell adhesion, migration, differentiation, and growth (11). Type IV collagen is expressed as six distinct ␣-chains, namely, ␣1-␣6 (12), assembles into triple helices, and further forms a network to provide a scaffold for other macromolecules in basement membranes. These ␣-chains are composed of three domains, the N-terminal 7 S domain, the middle triple helical domain, and the C-terminal globular noncollagenous domain (NC1) 1 (13). The ␣1 and ␣2 isoforms are ubiquitously present in * This work was supported in part by Grants DK-51711 and DK-55001 from the National Institutes of Health (to R. K.), the 1998 Hershey Prostate Cancer Research Award (to R. K.), the 1998 American Society of Nephrology Carl Gottschalk Research Award (to R. K.), the 1998 National Kidney Foundation Murray award (to R. K.), the 1998 Beth Israel Deaconess Medical Center Enterprise Award (to R. K.), and research funds from the Beth Israel Deaconess Medical Center. 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.
‡ Recipient of the 1999 Research Award for Young Scientists from the Inoue Foundation for Science of Japan.
ʈ Equity holder and consultant for Ilex Oncology, Inc., which is involved in the clinical development of tumstatin. To whom correspondence should be addressed: Nephrology Div., Dept. of Medicine, RW 563a, Beth Israel Deaconess Medical Center, 330 Brookline Ave., Boston, MA 02215. Tel.: 617-667-0445; Fax: 617-975-5663; E-mail: rkalluri@caregroup.harvard.edu. human basement membranes (9). The other four isoforms exhibit more tissue and organ-specific distributions (14,15). The distribution of the ␣3 (IV) chain is limited to certain basement membranes, such as glomerular basement membrane, several basement membranes of the cochlea, ocular basement membrane of the anterior lens capsule, Descemet's membrane, ovarian and testicular basement membrane (16), and alveolar capillary basement membrane (15,17,18). This chain is absent from epidermal basement membranes of the skin and the vascular basement membrane of liver (15).
The ␣3(IV) NC1 domain has been shown to bind and inhibit the proliferation of melanoma and other epithelial tumor cell lines in vitro (19). Han et al. (19) localized the binding site for melanoma cells to amino acids 185-203 of the ␣3(IV) NC1 domain. Monoclonal and polyclonal antibodies raised against this site were able to block melanoma cell adhesion and inhibition of proliferation (19). They also found that the specific sequence, -SNS-, located within amino acids 189 -191, was required for both the melanoma cell adhesion and inhibition of proliferation (19). Additionally, these investigators did not use the isolated human ␣3(IV) NC1 domain in these studies (19). In these studies, the 185-203 ␣3(IV) NC1 synthetic peptide was not tested on other cell types, including endothelial cells.
Recent studies have illustrated the anti-angiogenic properties associated with inhibitors of collagen synthesis, supporting the notion that basement membrane collagen assembly and organization is important for blood vessel formation (20,21). Furthermore, the NC1 domain of type IV collagen is speculated to play a crucial role in the assembly of type IV collagen to form trimers and thus influence basement membrane organization and modulation of cell behavior (10,11,(22)(23)(24). These prior observations coupled with the identification of endostatin as the type XVIII collagen NC1 domain (6) prompted us to examine the anti-angiogenic property of the NC1 domain of type IV collagen. In this regard, recently our laboratory identified a novel type IV collagen-associated inhibitor of angiogenesis and tumor growth, termed canstatin. Canstatin (NC1 domain of ␣2 chain) was identified as an endothelial cell-specific apoptotic agent (25).
In the present study, we demonstrate the pivotal role of the NC1 domain of the ␣3 chain of human type IV collagen (13,26) produced as a recombinant protein, in inhibiting the proliferation of capillary endothelial cells and blood vessel formation using in vitro and in vivo models of angiogenesis and tumor growth and also in inducing endothelial cell-specific apoptosis. We named this domain "tumstatin" (for its unique property of causing "tumor stasis") to add another member to the newly discovered family of endogenous inhibitors of angiogenesis derived from larger proteins, such as angiostatin, endostatin, restin, and canstatin (5,6,25). Among the six NC1 domains of type IV collagen, three exhibited promising anti-angiogenic activity with distinct mechanisms of action (25,27). The NC1 domain of the ␣3 chain (tumstatin) was most potent in inhibiting the proliferation of endothelial cells and causing apoptosis when compared with the other ␣(IV) chain NC1 domains. This newly discovered anti-angiogenic property of tumstatin, coupled with the previously reported antitumor cell activity, makes tumstatin a potentially useful therapeutic molecule in inhibiting tumor growth.

Recombinant Production of Tumstatin, Deletion Mutants, or Endostatin in Escherichia coli-
The sequence encoding tumstatin was amplified using polymerase chain reaction from the ␣3(IV)NC1/pDS vector (28). The resulting cDNA fragment was ligated into pET22b(ϩ) (Novagen, Madison, WI). This placed tumstatin downstream of the pelB leader sequence, allowing for periplasmic localization and expression of soluble protein. The 3Ј-end of the sequence was ligated in-frame with the polyhistidine tag sequence. Plasmid constructs encoding tumstatin were first transformed into E. coli HMS174 (Novagen, Madison, WI) and then transformed into BL21 for expression (Novagen). Overnight bacterial culture was used to inoculate a 500-ml culture in LB medium (Fisher). This culture was grown for ϳ4 h until the cells reached an A 600 of 0.6. Protein expression was induced by the addition of isopropyl-1thio-␤-D-galactopyranoside to a final concentration of 1 mM. After a 2-h induction period, cells were harvested by centrifugation at 5000 ϫ g and lysed by resuspension in 6 M guanidine, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0. Resuspended cells were sonicated briefly, and centrifuged at 12,000 ϫ g for 30 min. The supernatant fraction was passed over the nickel nitrilotriacetic acid-agarose column (Qiagen). Nonspecifically bound protein was removed by washing with both 10 and 25 mM imidazole in 8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0. Tumstatin protein was eluted from the column with increasing concentrations of imidazole (50,125, and 250 mM) in 8 M urea buffer. The eluted protein was dialyzed twice against PBS at 4°C and separated into insoluble and soluble fractions. Protein concentration was determined by the BCA assay (Pierce). Recombinant deletion mutants of tumstatin, tum-1-4, were produced in E. coli using pET28a(ϩ) (Novagen, Madison, WI) and purified as described above. Mouse endostatin was produced in E. coli using pET vector system and purified as described previously (29).
Immunoblotting-Soluble tumstatin was analyzed by SDS-PAGE and immunoblotting as described previously (30). Rabbit antibody raised against human ␣3(IV)NC1 was prepared as described previously (14). Goat anti-rabbit IgG antibody conjugated with horseradish peroxidase was purchased from Sigma.
Expression of Tumstatin in 293 Embryonic Kidney Cells-We used the pDS plasmid containing ␣3(IV)NC1 (28) to polymerase chain reaction amplify tumstatin in a way that it would add a leader signal sequence in-frame into the pcDNA 3.1 (Invitrogen, Carlsbad, CA). The leader sequence from the 5Ј-end of the full-length ␣3(IV) chain was cloned 5Ј to the NC1 domain to enable protein secretion into the culture medium. The tumstatin containing recombinant vectors were sequenced, and error-free cDNA clones were further purified and used for in vitro translation studies to confirm protein expression (data not shown). The tumstatin-containing plasmid and control plasmid were used to transfect 293 cells using the calcium chloride method (25). The cells were passed for three weeks in the presence of the geneticin (Life Technologies, Inc.) until no cell death was evident. Clones were expanded into T-225 flasks and grown until confluent. Then, the supernatant was collected and concentrated using an amicon concentrator (Amicon, Inc., Beverly, MA) and analyzed by SDS-PAGE, immunoblotting, and enzyme-linked immunosorbent assay for the tumstatin expression. Strong binding in the supernatant was detected by enzymelinked immunosorbent assay and Western blot using a tumstatin (␣3(IV)NC1) -derived synthetic peptide antibody raised in rabbit (14,28). Tumstatin containing supernatant was subjected to affinity chromatography using the same antibodies described above (14,28). described previously (25). Polymyxin B (Sigma) at a final concentration of 5 g/ml was used to inactivate endotoxin (32). Briefly, C-PAE cells (passages 2-4) were grown to confluence and kept contact inhibited for 48 h. 786-O, PC-3, HPE, and WM-164 cells were used as nonendothelial controls. Cells were trypsinized and a suspension of 12,500 cells in DMEM with 0.1% FCS was added to each well of a 24-well plate precoated with fibronectin. The cells were incubated for 24 h at 37°C, and medium was replaced with DMEM containing 20% FCS. Unstimulated control cells were incubated with medium containing 0.1% FCS. Cells were treated with various concentrations of tumstatin or deletion mutants. All wells received 1 Ci of [ 3 H]thymidine 12 h after the beginning of treatment. After 24 h, thymidine incorporation was measured using a scintillation counter. The methylene blue staining method was performed as described previously (25). All groups represent triplicate samples.
Endothelial Tube Assay-Endothelial tube assay was performed as described previously (25). Matrigel (Collaborative Biomolecules) was added (320 l) to each well of a 24-well plate and allowed to polymerize. A suspension of 25,000 MAE cells in EGM-2 without antibiotics was seeded into each well. The cells were treated with either tumstatin, BSA, or the 7 S domain of a ␣3 chain of human type IV collagen. The 7 S domain was extracted from human placenta basement membrane. Control cells were incubated with sterile PBS. Cells were incubated for 24 -48 h at 37°C and viewed using a CK2 Olympus microscope (magnification of ϫ 3.3 ocular, ϫ 10 objective). The cells were then photographed using 400 DK-coated TMAX film (Kodak). Cells were stained with diff-quik fixative (Sigma) and photographed again (33). Ten fields were viewed, and the number of tubes was counted by two investigators blinded for the experimental protocols and averaged.
Annexin V-FITC Assay-Annexin V, a calcium-dependent phospholipid-binding protein with a high affinity for phosphatidylserine was used to detect apoptosis (34). This assay was performed as described previously (25). CPAE cells (0.5 ϫ 10 6 /well) were seeded onto a 6-well plate in 10% FCS-supplemented DMEM. On the next day the fresh medium containing 10% FCS was added together with tumstatin ranging from 0.02 to 20 g/ml or 80 ng/ml TNF-␣. Control cells received an equal volume of PBS. After 18 h of treatment, medium containing floating cells was collected, and attached cells were trypsinized and centrifuged together with floating cells at 3000 ϫ g. The cells were then washed in PBS and resuspended in binding buffer (10 mM HEPES/ NaOH, pH 7.4, 140 mM NaCl, 2.5 mM CaCl 2 ). Annexin V-FITC (CLON-TECH) was added to a final concentration of 150 ng/ml, and the cells were incubated in the darkness for 10 min. The cells were washed again in PBS and resuspended in binding buffer. Annexin V-FITC-labeled cells were counted using a Becton-Dickinson FACStar plus flow cytometer. For each treatment 15,000 cells were counted and stored in listmode. These data were then analyzed using Cell Quest software (Becton-Dickinson).
Caspase-3 Assay-CPAE cells (0.5 ϫ 10 6 /well) were plated in 10-cm Petri dishes precoated with fibronectin (10 g/ml) onto a 6-well plate in 10% FCS-supplemented DMEM overnight. On the next day, medium was replaced with DMEM containing 2% FCS and incubated overnight at 37°C. Then cells were stimulated with basic fibroblast growth factor (3 ng/ml) in DMEM (2% FCS) also containing tumstatin (10 g/ml) or deletion mutants (5 g/ml) and incubated for 24 h. Controls received PBS buffer. TNF-␣ (80 ng/ml) was used as a positive control. After 24 h, the supernatant cells were collected, and attached cells were trypsinized and combined with the supernatant cells. Cells were counted and resuspended in cell lysis buffer (CLONTECH) at a concentration of 4 ϫ 10 7 cells/ml. The rest of the protocol followed the manufacturer's instruction (CLONTECH). A specific inhibitor of caspase-3, DEVD-fmk, was used to confirm the specificity of the assay. The absorbance was measured in a microplate reader (Bio-Rad) at 405 nm. Similarly nonendothelial cells (PC-3) were used and analyzed. This assay was repeated three times.
Matrigel Plug Assay-Five-to six-week-old male C57/BL6 mice (The Jackson Laboratories, Bar Harbor, ME) were obtained. All animal studies were reviewed and approved by the animal care and use committee of Beth Israel Deaconess Medical Center and are in accordance with the guidelines of the Department of Health and Human Services. Before injection into mice, matrigel (Collaborative Biomolecules) was mixed with 20 units/ml heparin (Pierce), 150 ng/ml basic fibroblast growth factor (R&D), and 1 g/ml tumstatin (E. coli-produced, soluble). Control groups received no angiogenic inhibitor. The matrigel mixture was injected subcutaneously. After 14 days, the mice were sacrificed, and the matrigel plugs were removed and fixed in 4% paraformaldehyde. The plugs were embedded in paraffin, sectioned, and hematoxylin and eosin stained. Sections were examined by light microscopy, and the number of blood vessels from 10 high power fields were counted and averaged. All sections were coded and observed by an investigator who was blinded for study protocols.
In Vivo Tumor Studies-PC-3 cells were harvested and injected subcutaneously on the back (5 ϫ 10 6 cells) of 7-9-week-old male athymic nude mice. The tumors were measured using Vernier calipers, and the volume was calculated using the standard formula (width 2 ϫ length ϫ 0.52) (5, 6). The tumors were allowed to grow to ϳ100 mm 3 , and the animals were then divided into groups of 5 or 6 mice. Tumstatin or mouse endostatin was intraperitoneally injected daily (20 mg/kg) for 10 days in sterile PBS. Tumor volume was calculated every 2 or 3 days. On day 10, mice were sacrificed, and tumor sections were obtained from control and tumstatin-treated groups and examined by light microscopy and TdT-mediated dUTP nick end labeling (TUNEL) assay.
786-O renal cell carcinoma cells (2 million cells) were injected subcutaneously on the back of 7-9-week-old male athymic nude mice. The tumors were allowed to grow to 600ϳ700 mm 3 , and animals were then divided into groups of six. Tumstatin was intraperitoneally injected daily (6 mg/kg) for 10 days in sterile PBS. In separate experiments, mice were intraperitoneally injected daily with 6 mg/kg truncated tumstatin lacking the N-terminal 53 amino acids (tum-1) expressed in E. coli (30) for 10 days. Each group consisted of five mice in this study. The initial tumor volume was 100 -150 mm 3 . In all experiments, the control group received vehicle injection (either BSA or PBS).
Light Microscopic Studies-Tumor tissue was fixed in 10% buffered formalin and embedded in paraffin. Sections (3 m) were stained with periodic acid-Sciff. In each sample, 30 blood vessels were examined, and the number of apoptotic cells was counted. An endothelial cell was considered morphologically apoptotic when it displayed loss of cell volume and chromatin condensation along the nuclear membrane with intensely basophilic staining (36).
TUNEL Method to Detect DNA Fragmentation in Tissue Sections-DNA fragmentation associated with apoptosis was detected in situ by the addition of digoxigenin-labeled nucleotides to free 3Ј-hydroxyl groups in DNA using the ApopTag In Situ apoptosis detection kit following the manufacturer's instruction (INTERGEN). The number of TUNEL-positive cells/blood vessel was counted, and the mean number of positive nuclei/30 cross-sections of blood vessels was determined.
Statistical Analysis-All values are expressed as mean Ϯ S.E. Analysis of variance (ANOVA) with a one-tailed Student's t test was used to identify significant differences in multiple comparisons. A level of p Ͻ 0.05 was considered statistically significant.

Expression and Purification of Human Tumstatin-Human
tumstatin was produced in E. coli using an expression plasmid, pET22b or pET 28a, as a fusion protein with a C-terminal six histidine tag. The E. coli-expressed protein was isolated predominantly as a soluble protein, and SDS-PAGE analysis revealed a monomeric band at 30 kDa (Fig. 1A). The additional 3 kDa arises from polylinker and histidine tag sequences. The eluted fractions represented by lanes 6, 7, and 8 were used in the following experiments. In nonreduced condition, a band was observed around 60 kDa representing a dimer of tumstatin. A single band around 30 kDa was observed in reduced condition (Fig. 1B). Tumstatin was immunodetectable by both antitumstatin (Fig. 1C) and anti-6-Histidine tag antibodies (data not shown). Human tumstatin was also produced as a secreted soluble protein in 293 embryonic kidney cells. This recombinant protein (without any purification or detection tags) was isolated using affinity chromatography (Fig. 1D), and a pure monomeric form was detected in the major peak by SDS-PAGE and immunoblot analyses (Fig. 1E). In our angiogenesis assays, we did not find significant differences in biological activities of E. coli-expressed and 293 cell-expressed recombinant protein. Therefore, because of the ease in production, we used E. coli-expressed protein for all of the studies. In some experiments, we also used 293 cell-expressed protein to confirm our results obtained by using E. coli-expressed protein.
Antiproliferative Effect of Tumstatin on Endothelial Cells-The antiproliferative effect of tumstatin on C-PAE cells was examined by [ 3 H]thymidine incorporation assay using E. coliproduced soluble protein. Tumstatin significantly inhibited 20% FCS-stimulated [ 3 H]thymidine incorporation in a dose-dependent manner with an ED 50 of ϳ0.01 g/ml ( Fig. 2A). No significant antiproliferative effect was observed with PC-3, 786-O, or primary human prostate epithelial cells (Fig. 2, B-D). When polymyxin B was used to inactivate endotoxins, the antiproliferative effect of tumstatin on C-PAE cells was not affected (Fig. 2E). Tumstatin had no effect on the proliferation of WM-164 melanoma cells even at a concentration of 20 g/ml (Fig. 2F). In previous studies, the proliferation of WM-164 melanoma cells was inhibited by synthetic peptide 185-203 derived from the tumstatin sequence (19). In these studies, full-length human tumstatin (␣3(IV) NC1) was not used (19).
Effect of Tumstatin on Endothelial Cell Tube Formation-When mouse aortic endothelial cells are cultured on matrigel, they rapidly align and form hollow tube-like structures (37). Tumstatin, produced in E. coli, significantly inhibited endothelial tube formation in a dose-dependent manner as compared with BSA controls (Fig. 3, A and C). The percentage of tube formation after treatment with 1 g/ml protein was 98.0 Ϯ 4.0 for BSA and 14.0 Ϯ 4.0 for tumstatin. Similar results were also obtained using tumstatin produced in 293 cells (data not shown). The 7 S domain of the ␣3 chain of type IV collagen had no effect on endothelial tube formation (Fig. 3B) and proliferation of mouse aortic endothelial cells (data not shown). However, we must point out that some recent studies have suggested that planar models of spontaneous angiogenesis (endothelial tube assays) stimulate invasive angiogenesis poorly (38).
Effect of Tumstatin on Inducing Endothelial Cell Apopto-sis-In the early stage of apoptosis, translocation of the membrane phospholipid phosphatidylserine from the inner surface of plasma membrane to outside is observed (34,39,40). Externalized phosphatidylserine can be detected by staining with a FITC conjugate of annexin V that binds naturally to phosphatidylserine (34). Tumstatin at 20 g/ml showed a distinct shift of annexin fluorescence peak after 18 h (Fig. 3D). The shift in fluorescence intensity was similar for tumstatin at 20 g/ml and the positive control TNF-␣ (80 ng/ml) (data not shown). Tumstatin at 2 g/ml also showed a mild shift in annexin fluorescence intensity, but concentrations below 0.2 g/ml did not demonstrate any annexin V positivity (data not shown). This shift of peak intensity was not observed when nonendothelial cells (PC-3) were used (data not shown).
Tumstatin Increases the Activity of Caspase-3-Caspase-3 (CPP32) is an intracellular protease activated at the early stage of apoptosis and initiates cellular breakdown by degrading structural and DNA repair proteins (41,42). The protease activity of caspase-3 was measured spectrophotometrically by detection of the chromophore (p-nitroanilide) cleaved from the labeled substrate (DEVD-p-nitroanilide). Tumstatin (20 g/ml) -treated cells exhibited a 1.6-fold increase in caspase-3 activity, whereas TNF-␣ gave a comparable (1.7-fold) increase compared with control (Fig. 3E). A specific inhibitor of caspase-3, DEVDfmk, decreased the protease activity to baseline indicating that the increase in the measured activity was specific for caspase-3. In nonendothelial cells (PC-3), there was no difference in caspase-3 activity between control and tumstatin-treated cells (Fig. 3F).
Effect of Tumstatin on Angiogenesis in Matrigel Plug Assay-To evaluate the in vivo effect of E. coli-produced soluble tumstatin on the formation of new capillaries, we performed a matrigel plug assay in mice (43). A 67% reduction in the number of blood vessels was observed with a dose of 1 g/ml tumstatin (Fig. 4C)  4B). The number of vessels/high power field was 2.25 Ϯ 1.32 for tumstatin and 7.50 Ϯ 2.17 for control (Fig. 4A).
Effect of Tumstatin on the Growth of Tumors in Mouse Xenograft Model-We examined the effect of tumstatin on established primary human tumor models in nude mice. Human tumstatin, produced in E. coli, significantly inhibited the growth of PC-3 human prostate carcinoma xenografts (Fig. 4D). Human tumstatin at 20 mg/kg inhibited tumor growth similar to mouse endostatin (20 mg/kg) (Fig. 4D). Significant inhibitory effect on tumor growth was observed on day 10 (control 202.8 Ϯ 50.0 mm 3 , tumstatin 82.9 Ϯ 25.2 mm 3 , endostatin 68.9 Ϯ 16.7 mm 3 ). Additionally, tumstatin at as little as 6 mg/kg inhibited the growth of 786-O human renal cell carcinoma xenografts as compared with the BSA control (Fig. 4E). A significant inhibitory effect on tumor growth was observed on day 10 (control 1096 Ϯ 179.7 mm 3 , tumstatin 619 Ϯ 120.7 mm 3 ).
Tumstatin Induces Endothelial Cell Apoptosis in Vivo-We evaluated the effect of tumstatin in inducing endothelial cell apoptosis in vivo using tumor (PC-3) tissue sections. The number of TUNEL-positive apoptotic cells in vessel walls was significantly increased in the tumstatin-treated group (Fig. 4H) at day 10 as compared with controls ( Fig. 4G) (control 2.0 Ϯ 1.0, tumstatin 13.0 Ϯ 1.5; Fig. 4F). Apoptosis in situ was also evaluated by the conventional cell morphology using light microscopy. The number of apoptotic nuclei in the blood vessels of the tumstatin-treated group (Fig. 4J) was also significantly higher than the control group (Fig. 4I). Tumstatin at 20 g/ml induced a distinct shift of fluorescence intensity peak compared with the control. This experiment was repeated three times, and the representative data are shown. Control and tumstatintreated cells were lysed, and caspase-3 activity was detected. Tumstatin increased the activity of caspase-3 as compared with control in C-PAE cells (E). When the specific inhibitor of caspase-3, DEVD-fmk was used, the activity went down to basal level in all groups. TNF-␣ was used as a positive control. Tumstatin did not increase caspase-3 activity in control, nonendothelial PC-3 cells (F). Each column represents the mean Ϯ S.E. of triplicate wells. This experiment was repeated three times.
Expression of Tumstatin Deletion Mutants-To further characterize the antitumor activity of tumstatin, three different FIG. 4. Matrigel plug assay and in vivo tumor studies. Sections of each matrigel plug stained by H&E were examined by light microscopy, and the number of blood vessels from 10 high power fields was counted and averaged. Tumstatin (1 g/ml) significantly inhibited in vivo neovascularization as compared with controls (treated with PBS). The difference between the mean percentage value of tumstatin-treated animals and control animals was significant (A). Each column represents the mean Ϯ S.E. of five to six mice/group. *, p Ͻ 0.05 by one-tailed Student's t test. Representative light microscopic appearance of matrigel plug (H&E staining, ϫ200 magnification) in control group are shown in B. Marked neovascularization (arrowheads) can be observed in the amorphous matrigel plug. Inset, high magnification view (ϫ800 magnification) of blood vessels. There was a lesser extent of neovascularization observed in the matrigel plug of the tumstatin-treated group (C). Inset, high magnification view (ϫ800 magnification) of nonvascular cells. Arrows indicate the position magnified in the insets. Daily intraperitoneal injection of tumstatin (20 mg/kg) or endostatin (20 mg/kg) inhibited the growth of PC-3 xenografts as compared with the PBS control (D). This experiment was started when the tumor volumes were less than 100 mm 3 . Daily intraperitoneal injection of tumstatin (6 mg/kg) inhibited the tumor growth of 786-O xenografts as compared with the control (E). This experiment was started when the tumor volumes were 600 -700 mm 3 . Each point represents the mean Ϯ S.E. of five to six mice. The mean number of TUNEL-positive cells in 30 blood vessels in a PC-3 section from day 10 are shown in F. *, p Ͻ 0.05 by one-tailed Student's t test. A higher number of TUNEL-positive endothelial cells were observed in the tumstatin-treated group (arrowheads in H) as compared with the control group (G) (ϫ400 magnification). Endothelial cell apoptosis was also evaluated by conventional cell morphological appearance (I, control; J, tumstatin-treated; PAS staining, ϫ400 magnification). Apoptosis was identified by the presence of typical condensation of nuclear chromatin and cytoplasm (arrowhead in J).
deletion mutants of tumstatin and the one described above were expressed in E. coli using the pET28a system, and soluble protein was isolated (Table I and Fig. 6). Tumstatin consists of 244 amino acids including 12 amino acids from the triple helical portion located in the N-terminal portion, and 232 amino acids derived from the NC1 domain. Tum-1 consists of 191 amino acids and is lacking the N-terminal 53 amino acids.
Tum-2 consists of 132 amino acids in the N-terminal half portion of tumstatin, and tum-3 is the C-terminal half portion (112 amino acids). Tum-4 consists of 64 amino acids in the C terminus of tumstatin, which includes the 185-203 peptide region (19).
Effect of Tumstatin Deletion Mutants on the Caspase-3 Activity of C-PAE Cells-tum-1 and tum-2 treatment of C-PAE cells increased the activity of caspase-3 (Fig. 8C). Tum-3 and tum-4 did not induce caspase-3 in these assays. Together, these results suggest that the antitumor activities associated with tumstatin are localized to distinct regions on this molecule (19). DISCUSSION The formation of new capillaries from pre-existing vessels, angiogenesis, is essential for the process of tumor growth and metastasis (2,3,48). The switch to an angiogenic phenotype requires both up-regulation of angiogenic stimulators and down-regulation of angiogenesis inhibitors (3). Vascular endothelial growth factor and basic fibroblast growth factor are abundantly expressed angiogenic factors in tumors. Vascularized tumors may overexpress one or more of these angiogenic factors, which can synergistically promote tumor growth. Inhibition of a single angiogenic factor such as the vascular endothelial growth factor with a receptor antagonist may not be enough to arrest tumor growth, because tumor subpopulations that produce angiogenic factors other than vascular endothelial  growth factor can still influence tumor growth (49). A number of angiogenesis inhibitors have been recently identified, and certain factors such as interferon-␣, platelet factor-4 (50), and PEX (51) are not endogenously associated with tumor cells. On the other hand, angiostatin (5) and endostatin (6) are tumorassociated angiogenesis inhibitors generated by the tumor tissue itself. In this report, we demonstrate the capacity of the NC1 domain of the ␣3 chain of human type IV collagen (tumstatin) (13,26) to inhibit the proliferation of vascular endothelial cells and the formation of new blood vessels using in vitro and in vivo models of angiogenesis and tumor growth. Our results indicate that tumstatin may be exerting its effect at different stages in the process of tumor angiogenesis. The specific inhibition of proliferation of endothelial cells by tumstatin strongly suggests that it may function via a cell surface protein/ receptor. Whether tumstatin functions by suppressing the activity of vascular endothelial growth factor and/or basic fibroblast growth factor remains to be elucidated. Induction of apoptosis in growth-stimulated endothelial cells by tumstatin was observed using annexin V-FITC. The induction of endothelial cell apoptosis by tumstatin was most pronounced when tumstatin was added to subconfluent monolayers, when cells were growing exponentially (data not shown). Conceivably, tumstatin is selective for tumor vasculature in which endothelial cells are activated. The pro-apoptotic effect of tumstatin was mediated by increased caspase-3 activity in endothelial cells. Additionally, tumstatin-treated tumors showed significantly increased apoptosis when compared with saline-injected control tumors. Matrix metalloproteinases have been implicated as key enzymes that regulate the formation of new blood vessels in tumors (52). Recently, it was demonstrated that PEX, a domain of matrix metalloproteinase-2, which can inhibit the interac-tion of matrix metalloproteinase-2 and ␣v␤ 3 integrin, suppressed tumor growth (51). Similarly, tumstatin may also function by inhibiting the activity of matrix metalloproteinases.
Recently, a synthetic peptide (19 amino acids) corresponding to the C-terminal portion of tumstatin was reported to bind to the ␣v␤ 3 integrin (53). Because angiogenesis depends on specific endothelial cell adhesion events mediated by the ␣v␤ 3 integrin (4, 54), it is possible that the anti-angiogenic effect of tumstatin is mediated by disrupting the interaction of proliferating endothelial cells to the matrix component, such as vitronectin and fibronectin, an event that is considered as an important anti-apoptotic signal (55). Interestingly, in these studies by Shahan et al. (53), only the synthetic peptide 185-203 was used and the full-length ␣3(IV) NC1 (tumstatin) was not used. In the present study, we speculate if the anti-angiogenic activity of tumstatin is mediated by the peptide 185-203 and its ␣v␤3 binding property. To address this issue, we employed deletion mutagenesis to generate truncated fragments of tumstatin. Our results indicate that although the isolated peptide 185-203 and tum-4 mutant containing this sequence inhibit melanoma cell proliferation and bind the ␣v␤3 receptor, it is not responsible for the anti-angiogenic activity of tumstatin.
In contrast, the mutant tum-2, which contains the N-terminal half of the sequence of tumstatin, but not the peptide sequence 185-203, exhibited anti-angiogenic properties with no antitumor cell activity. Collectively, our studies with deletion mutants of tumstatin indicate that the exclusive antiangiogenic activity is contained within amino acids 54 -132. Additionally, our experiments indicate the anti-angiogenic activity residing within these amino acids is effective even when it is part of a full-length folded tumstatin. Interestingly, the antitumor cell activity residing within the peptide sequence 185-203 is not available when present as part of the full-length tumstatin. The activity imparted on the melanoma cells by the amino acid sequence 185-203 is only realized when this peptide is exposed either by truncation of the molecule (as in this study) or by synthesis of a representative peptide (19). To our knowledge, this is the first report of a molecule with such distinct antitumor activities. Of course, many molecules have been reported to carry peptide activities not associated with the full-length protein consisting of the peptides, and one such example is the heparin binding peptides of the ␣1 NC1 domain of type IV collagen (23), although the ␣1(IV)NC1 domain itself does not exhibit a similar degree of binding capacity to heparin.
Because tumstatin possesses the pathogenic epitope for Goodpasture syndrome, it is possible that acute or chronic administration of tumstatin may induce this disease. We synthesized truncated tumstatin lacking the N-terminal 53 amino acids (tum-1) to remove this epitope, and this molecule continues to exhibit an inhibitory effect on the growth of 786-O xenografts. Additionally, this molecule did not bind autoantibodies from several patients with Goodpasture syndrome (data not shown). Tum-1 also potently decreased cell viability of endothelial cells. These results suggest that the anti-angiogenic region of tumstatin is conserved even when the N-terminal 53 amino acids are removed. Taken together with the results using tum-2 and tum-3 mutants, our studies indicate that the anti-angiogenic activity resides within the amino acids 54 -124, which differs from the Goodpasture epitope.
In conclusion, tumstatin inhibits angiogenesis in the in vitro and in vivo model, resulting in the suppression of tumor growth. Ultimately, we would like to develop alternative strategies to express the tumstatin gene in vivo directed to the tumor vasculature employing gene transfer approaches (56 -58). The antitumor activities associated with human tumstatin FIG. 8. Endothelial cells become less viable and undergo apoptosis when treated with Tum-1, Tum-2, and tumstatin. The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazoliumbromide assay was used to evaluate cell viability in C-PAE cells after treatment with tumstatin and the deletion mutants (A). Tum-1 decreased the cell viability in a dose-dependent manner. At dosages of 1 and 5 g/ml, tum-1 was significantly more effective than tumstatin at decreasing cell survival. Tum-4 was the only deletion mutant that decreased the viability of the WM-164 melanoma cell line (B). Each point represents the mean Ϯ S.E. of triplicate wells. Tum-1 and Tum-2 treatment increased the activity of caspase-3 in C-PAE cells as shown in C. TNF-␣ was used as a positive control. make it a promising therapeutic candidate for inhibition of tumor growth in cancer patients.