Cleavage of the Matricellular Protein SPARC by Matrix Metalloproteinase 3 Produces Polypeptides That Influence Angiogenesis*

SPARC, a matricellular protein that affects cellular adhesion and proliferation, is produced in remodeling tissue and in pathologies involving fibrosis and angiogenesis. In this study we have asked whether peptides generated from cleavage of SPARC in the extracellular milieu can regulate angiogenesis. Matrix metalloproteinase (MMP)-3, but not MMP-1 or 9, showed significant activity toward SPARC. Limited digestion of recombinant human (rhu)SPARC with purified catalytic domain of rhuMMP-3 produced three major fragments, which were sequenced after purification by HPLC. Three synthetic peptides (Z-1, Z-2, and Z-3) representing motifs from each fragment were tested in distinct assays of angiogenesis. Peptide Z-1 (3.9 kDa, containing a Cu2+-binding sequence KHGK) exhibited a biphasic effect on [3H]thymidine incorporation by cultured endothelial cells and stimulated vascular growth in the chick chorioallantoic membrane (CAM). In contrast, peptides Z-2 (6.1 kDa, containing Ca2+-binding EF hand-1) and Z-3 (2.2 kDa, containing neither Cu2+-binding motifs nor EF hands), inhibited cell proliferation in a concentration-dependent manner and exhibited no effects on vessel growth in the CAM. Reciprocal results were obtained in a migration assay in native collagen gels: peptide Z-1 was ineffective over a range of concentrations, whereas Z-2 or Z-3 stimulated cell migration. Therefore, proteolysis of SPARC by MMP-3 produced peptides that regulate endothelial cell proliferation and/or migration in vitro in a mutually exclusive manner. One of these peptides containing KHGK also demonstrated a concentration-dependent effect on angiogenesis.

cell proliferation, and one of them, KGHK, stimulated angiogenesis in vivo (18). Moreover, KGHK/Cu 2ϩ was shown to be released from native SPARC by plasmin (17), and peptide(s) containing this sequence were identified as natural cleavage products in the chick chorioallantoic membrane (CAM) model of angiogenesis (18).
In the present study we have asked whether certain members of the MMP family with established activities in angiogenesis could cleave SPARC into fragments with sequence motifs that could regulate one or more aspects of endothelial cell function, e.g. in a CAM assay to quantify vessel growth, the radial invasion of matrix by aggregated cells (RIMAC) assay, which measures migration of cells in native collagen gels, and an assay of DNA synthesis in vitro. We have shown that MMP-3 (stromelysin 1) cleaves SPARC into three major fragments, the sequencing of which allowed us to synthesize peptides with motifs previously identified as modulators of angiogenesis, as well as some additional ones, in an extended sequence context. The penultimate Cu 2ϩ -binding sequence was found to be an effective stimulator of the endothelial cell cycle and of angiogenesis in vivo. In contrast, a peptide comprising the Ca 2ϩ -binding EF hand-1 inhibited endothelial cell proliferation but stimulated migration in collagen gels. A third peptide, lacking both Cu 2ϩ -and Ca 2ϩ -binding motifs, also stimulated migration but was substantially less effective as a cellcycle inhibitor. Thus, we have shown that MMP-3 releases peptides from SPARC that contain sequence motifs with mutually exclusive effects, within a given range of concentration, on endothelial cell migration and proliferation. Additionally, a peptide containing the Cu 2ϩ -binding sequence KHGK exhibited a concentration-dependent effect on angiogenesis in vivo.

EXPERIMENTAL PROCEDURES
Recombinant Proteins-Recombinant human (rhu) SPARC with a C-terminal hexahistidine tag (SPARC-His) was prepared as described (19,20), with the following modifications. An 8 M urea extract of bacterial paste derived from 25 liters of a fed-batch fermentation (21) was applied to a 40-cm 3 column of nickel-nitriloacetic acid resin (Ni-NTA, Qiagen, Valencia, CA). Endogenous bacterial proteins were removed by washing with pH 6.0 buffer. rhuSPARC-His that was immobilized onto the Ni-NTA resin was renatured at pH 8.5 by a reverse gradient from 8 to 0 M urea in 0.1 M NaH 2 PO 4 , 0.05 M Tris-HCl, 0.5 M NaCl, 5 mM oxidized glutathione, 1 mM reduced glutathione, 0.5% octylthioglucoside. rhuSPARC-His was eluted at pH 4.5, dialyzed against 0.1 N acetic acid, and lyophilized.
The catalytic domain of human proMMP-3, lacking the N-terminal 34 amino acids and the C-terminal hemopexin-like domain (⌬ 248 -460 ), was expressed in Escherichia coli, refolded from inclusion bodies, purified as described previously (22), and provided as a gift from Dr. Hideaki Nagase (Imperial College, London, UK).
Cleavage of rhuSPARC-His with Active Form MMP-3 (⌬ 248 -460 )-1-5 mg rhuSPARC-His was digested with active form MMP-3 in 0.05 M Tris-HCl (pH 7.5), 0.15 M NaCl, 0.01 M CaCl 2 for 16 h at 37°C. Optimal digestion conditions were obtained at a 1:100 (enzyme:substrate) ratio by weight. Digestion was terminated by the addition of EDTA to 0.01 M. Following the addition of 3 volumes of 0.1% trifluoroacetic acid, the digestion mixture was injected onto a Vydac C4 reverse-phase HPLC column that had been repeatedly conditioned by the process described below. Development of the column was achieved at 1 ml/min by linear gradients from 0.1% trifluoroacetic acid (A) to n-propanol/0.1% trifluoroacetic acid (B), according to the following schedule: 0% B, 10 min; 0 -15% B, 15 min; 15-60% B, 100 min; 60 -100% B, 10 min; 100 -0% B, 20 min. Fractions were dried by vacuum centrifugation and were stored at Ϫ80°C.
Amino Acid Sequencing of rhuSPARC-His Fragments-Cleavage products of rhuSPARC-His were fractionated on 16.5% polyacrylamide electrophoretic gels that contained 0.1% SDS and were transferred to a polyvinyldenefluoride membrane in 10 mM morpholinosulfonic acid (pH 11.0) that contained 20% methanol by volume, for 30 min at 0.3 A. Transferred peptides were visualized by staining with Coomassie Brilliant Blue R-250, excised from the membrane, and subjected to Nterminal Edman degradation and sequencing on an Applied Biosystems 477A protein sequencer.
Synthetic Peptides-Peptides were synthesized by Zymogenetics, Inc. (Seattle, WA) and were purified by reverse-phase HPLC. Concentration and purity were determined by amino acid analysis. In all cases, peptides exhibited Ͼ90% purity, with no contaminating (unexpected) residues present. The sequences, molecular weights, and pIs of these peptides, designated Z-1, Z-2, and Z-3, are shown in Table I.
Proliferation and Migration Assays-BAE cells were isolated and subcultured as previously described (7). Cell proliferation was measured by the incorporation of [ 3 H]thymidine (10 mCi/ml, PerkinElmer Life Sciences, Boston, MA) according to (7). Confluent cells were released by incubation in trypsin, and were resuspended in Dulbecco's modified Eagle's medium (DMEM) containing 2% fetal calf serum at 5000 cells/ml. The cells were plated in triplicate (0.5 ml/well, 2 cm 2 ), and peptides were added immediately at concentrations ranging from 0.05-1.0 mM. After 18 h, 2 Ci/ml [ 3 H]thymidine was added to each well for an additional 2 h. After removal of the media, the cell layers were rinsed twice with ice-cold 5% trichloroacetic acid and were solubilized by shaking in 0.5 ml of 0.4 M NaOH. Incorporated cpm were measured by scintillation counting. Each experiment was performed a minimum of three times.
Endothelial migration and cord morphogenesis were assessed by the RIMAC assay (23). Aggregates of BAE or porcine aortic endothelial (PAE) cells (a gift of J. Waltenberger, Uppsala, Sweden) were placed in native type I collagen gels (0.5 mg/ml; BD Biosciences, Bedford, MA), in the presence or absence of fibroblast growth factor (FGF)-2 (20 ng/ml, Scios, Mountain View, CA). DMEM was used as control buffer and as diluent for the peptides. Assays were conducted in quadruplicate, and experiments were performed a minimum of three times.
Angiogenesis Assay-Quail CAM assays were performed and quantified as described in Parsons-Wingerter et al. (24). Peptides Z-1, Z-2, and Z-3, each at concentrations of 0.1 and 1.0 mM in a total volume of 500 l of PBS, were applied to CAMs cultured in 6-well tissue culture plates. Control CAMs received an equal volume of PBS alone. After 24 h, images were photographed digitally at 10ϫ, and arterial vessel growth was quantified by calculation of the fractal dimension (Df) of skeletonized binary images (24).

RESULTS
Published studies from several laboratories have identified a number of different functions of SPARC, many of which are related to de-adhesion and anti-proliferation of cultured cells (4 -6). The use of synthetic peptides representing domain-specific sequences of SPARC has confirmed some of the functional properties attributed to SPARC, and has identified smaller regions or subdomains and their sequences to which certain activities of SPARC could be assigned. However, most of the synthetic peptides were chosen arbitrarily, e.g. to isolate a Ca 2ϩ binding region or a sequence with a purportedly stabilized secondary structure, and prior to the crystallographic studies performed on domains II and III (16,27). Having identified peptides of SPARC with specific activities on several types of cells in vitro, we questioned whether one or more of these sequences were present in vivo, and whether natural proteolysis of SPARC could occur as a result of tissue injury. In this study we have therefore asked the following questions: (a) Are there peptides generated from the cleavage of SPARC by certain MMPs that influence the process of angiogenesis, and (b) are the previously identified short sequence motifs of SPARC active in the context of larger peptides? Fig. 1A is shown an incubation of 125 I-SPARC individually with three different MMPs (1, 3, and 9) that are relevant to angiogenesis. MMP-3, but not MMP-1 or 9, exhibited reproducible catalytic activity toward SPARC. The cleavage of SPARC by MMP-3 was concentration-dependent (Fig. 1B, lanes 2-5) and was inhibited by TIMP-1 (Fig. 1B, lane 6).

Cleavage of SPARC by MMP-3-In
Electroblotting of an MMP-3 digest of SPARC onto a nitrocellulose filter, followed by exposure to an antibody recognizing hSPARC peptide 2.3 (Thr 114 -Gly 130 ), revealed a band of approximate M r 14,000 (not shown). Interestingly, peptide 2.3 contains the sequence KKGHK, which has a high affinity for Cu 2ϩ and is angiogenic in vivo and in vitro (17,18,28). Incubation of peptide 2.3 with MMP-3 resulted in multiple and extensive cleavages, predominantly at Lys 123 , Leu 124 , and His 125 -Leu 126 , as detected by tandem mass spectrometry (not shown). These data indicate that this sequence is protected within the native structure of SPARC, a conclusion borne out by crystallographic data (27) and cell proliferation assays (29). In summary, MMP-3, but not MMP-1 or 9, displayed consistent proteolytic activity toward SPARC. One of the fragments released from SPARC was of approximate M r 14,000 and contained a Cu 2ϩ -binding sequence with reported angiogenic activity.
For the purpose of sequencing major cleavage fragments of SPARC produced by MMP-3, rhuSPARC was subjected to di-gestion under conditions that we had previously determined as optimal for this study: an enzyme-to-substrate ratio of 1:100 (by weight) for 16 h at 37°C, in a buffer containing 10 mM Ca 2ϩ for maintenance of native structure. Peptides were fractionated by HPLC on a C4 column, as shown in Fig. 2A, and peaks J1 and J2ϩJ3 were further resolved by SDS-PAGE (Fig. 2, inset), prior to electroblotting and Edman degradation. The peptide sequences obtained from this procedure are shown in Fig. 2B, and their size distributions are summarized in Table II. Peptide J1 (M r 9,300) covered the N-terminal-third of the protein (Gln 2 -Pro 87 ), whereas peptide J2 (M r 17,700) extended from Ile 88 to Pro 237 , with the remainder of the protein (49 amino acids) cleaved into smaller fragments. Importantly, EF hand-2 (Asp 257 -Glu 268 ) is absent from peptide J2, although J2 comprises much of the C-terminal EC domain of SPARC. In addition, it contains the Cu 2ϩ -binding sequence KKGHK (Lys 119 -Lys 123 ) and EF hand-1 (Asp 165 -Lys 176 ). Sequencing of peptide J3 (M r 10,600, Met 150 -Pro 237 ) revealed it to be derived from J2 (Fig. 2, A and B), which also contains EF hand-1 but lacks KKGHK. The net result of the limited digestion of rhuSPARC by MMP-3 is: (a) liberation of fragment J1 (Table II and Fig.  2A), (b) production of fragment J2-J3 (Table II and Fig. 2B), and (c) subsequent release of fragment J3 from J2-J3 (Table II and Fig. 2B).  Fig. 3C) is limited to domain III. Both, however, contain EF hand-1 (identified by Ca 2ϩ as orange spheres) (Fig. 3, B and C).
SPARC Peptide Z-1 Exhibits a Biphasic Effect on Angiogenesis-From the data shown in Figs. 2 and 3, we identified novel sequence motifs, or previously described motifs within extended sequences, as candidates for the production of synthetic peptides, termed Z-1, Z-2, and Z-3, as described in Table I. These peptides provided several advantages for assays of angiogenesis in vivo and in vitro: (a) a high degree of purity with no detectable endotoxin, (b) a molecular size and solubility properties advantageous for concentration-dependent assay in the CAM and RIMAC settings, and (c) copious amounts of standardized preparation that allowed for multiple experiments.
HPLC-purified peptides Z-1-Z-3 correspond to subfragments of 3 major MMP-3 cleavage products of rhuSPARC (Tables I  and II). Z-1 exhibited a stimulatory effect in the quail CAM assay (Fig. 4), with an increase in vessel density at a concentration of 0.1 mM, and an inhibitory effect (return to basal state) at 1.0 mM (Fig. 4, B and C, respectively). Peptide Z-1 contains the Cu 2ϩ binding sequence KHGK (Table I and Fig.  2B), for which a nearly identical effect on BAE cell proliferation was previously reported (7,29). In contrast, peptides Z-2 and Z-3 displayed minimal effects on vascular density in the CAM that were found not to be statistically significant (data not shown).
Biphasic and Inhibitory Effects of SPARC Peptides on Endothelial Cell Proliferation-Standard [ 3 H]thymidine incorporation assays were performed with BAE cells in the presence of different concentrations of peptides Z-1, Z-2, and Z-3, as shown in Fig. 5. Peptide Z-1 exhibited a biphasic effect on thymidine incorporation, with a 4-fold stimulation at a concentration of 0.2 mM (Fig. 5A). This result parallels the effect seen with peptide Z-1 in the CAM assay (Fig. 4). In contrast, both peptides Z-2 and Z-3 demonstrated only inhibitory activities in the Cleavage sites are indicated in bold type. EF hands-1 and -2 are underlined. Note that bacterial rhuSPARC lacks one amino acid (the N-terminal Ala) and is therefore shown as a sequence of 285 residues, rather than the 286 residues that comprise human SPARC. BAE cell proliferation assay (Fig. 5B). Peptide Z-2 exhibited an ED 50 at slightly less than 0.2 mM, whereas peptide Z-3 was considerably less effective in this assay, with an ED 50 Ϸ 0.5 mM (Fig. 5B). Since a 20-mer peptide containing a partial EF hand-1 sequence has previously shown minimal activity in proliferation assays, it is likely that the extended sequence of peptide Z-2, a 50-mer, has conferred additional stability on this nondisulfide-bonded Ca 2ϩ -binding sequence. Peptide Z-3, which lacks both EF hands, showed diminished activity in the thymidine incorporation assay, relative to Z-1 and Z-2 (Fig. 5B).

SPARC Peptides Z-2 and Z-3 Enhance Endothelial Cell Migration through Native Collagen Gels-Aggregates of PAE cells
were embedded in collagen gels, with different concentrations of peptides Z-1, Z-2, and Z-3 (singly or in combination), FGF-2, or buffer alone, and analyzed for distance of radial migration, as described for the RIMAC assay (23). In contrast to their null and inhibitory effects in the CAM and thymidine incorporation assays (Figs. 4 and 5, respectively), peptides Z-2 and Z-3 stimulated PAE cell migration in the RIMAC assay, whereas peptide Z-1, which exhibited biphasic effects in the CAM and proliferation assays, was ineffective (Fig. 6). At concentrations of 0.2 mM, peptides Z-2 and Z-3 stimulated PAE cell migration to essentially the same extent as seen with FGF-2 at 20 ng/ml (Fig. 6). None of the peptides was able to block the stimulatory effect of FGF-2 over a range of concentrations, and combinations of Z-2 or Z-3 (0.05-0.4 mM) with FGF-2 (20 ng/ml) were not synergistic (data not shown). Similar results were obtained with BAE cells (not shown).
A summary of the activities of peptides Z-1, Z-2, and Z-3, and their location within intact SPARC and within the peptides (J1, J2, and J3) derived from the proteolysis of SPARC by MMP-3, are shown in Fig. 7.

DISCUSSION
In this study we have shown that the matricellular protein SPARC can be cleaved in a reproducible, concentration-dependent manner by MMP-3, a reactivity that was inhibited by TIMP-1. Limited digestion of rhuSPARC with MMP-3 produced three major peptides (J1-J3) containing sequence motifs that influenced endothelial cell behavior and angiogenesis: Cu 2ϩ -binding KHGK, and Ca 2ϩ -binding EF hand-1. To facilitate study of these motifs on cultured endothelial cells and on CAMs, we synthesized shorter versions of J2 and J3 (termed Z-2 and Z-3) and tested them over a range of concentrations in assays of endothelial cell proliferation and migration, as well as angiogenesis. Peptide Z-1, containing the motif KHGK/Cu 2ϩ , exhibited a biphasic effect on [ 3 H]thymidine incorporation by endothelial cells in vitro and on angiogenesis, but did not affect the migration of endothelial cells through native collagen gels. In contrast, peptide Z-2, containing EF hand-1, and Z-3, lacking both Cu 2ϩ -and Ca 2ϩbinding motifs, exhibited inhibitory effects on proliferation but stimulated migration to levels seen with 20 ng/ml FGF-2. Peptides Z-2 and Z-3 were inactive in the CAM assay of developmental angiogenesis. Therefore, discrete motifs within major MMP-3-derived peptides of SPARC influenced a Length in amino acid residues; numbering based on the mature form of rhuSPARC produced in E. coli (19), which is missing N-terminal Ala; Pro 2 thus becoming Pro 1 .

FIG. 3. Major fragments produced by proteolytic cleavage of SPARC by MMP-3.
Shown are molecular models derived from A chain coordinates of Protein Data Bank Accession 1BMO (27). A, domain I (the structure of which is unsolved) is linked to the follistatin domain II (FS, blue), which is linked to the extracellular calcium-binding domain III (EC, green). His 6 -COOH designates the covalent linkage of a histidine hexamer to the C terminus of the recombinant protein (19). endothelial cell proliferation, migration, and angiogenesis in a mutually exclusive manner.
Proteolytic fragments of several components of the ECM, including collagen types XVIII (endostatin), IV (tumstatin, canstatin, arresten), XV (restin), and VIII (vastatin), as well as the laminin ␥1 chain, have been shown to regulate angiogenesis (15,30,31). In addition, fragments from molecules as diverse as calreticulin (vasostatin) (32), high molecular weight kininogen (kininostatin) (33), platelet factor 4, plasminogen (angiostatin), and prolactin have demonstrated activity on endothelial cells in vitro and on angiogenesis in vivo (reviewed in Ref. 15). Several of the matricellular proteins, which are characteristically expressed in areas of vascular remodeling and tissue repair, give rise to peptides that exhibit angiostimulatory or angioinhibitory activities in vitro and in vivo, e.g. os-teopontin, thrombospondin 1, and SPARC (1,15). One of the more remarkable examples is the plasmin-mediated release of the proangiogenic Cu 2ϩ -binding peptide KGHK from SPARC, which stimulated endothelial cell proliferation and cord formation in vitro and vessel growth in the CAM (17,18). It is important to note that some of the fragment(s) exhibit activities that differ from those of the parent protein, e.g. KGHK versus SPARC.
Limited digestion of SPARC with MMP-3 produced initially a cleavage at Pro 87 -Ile 88 , giving rise to peptides J1 and J2, as well as proteolysis of several peptide bonds at the C terminus of SPARC (beginning at Pro 237 -Leu 238 ) (Fig. 2). These latter small peptides were not recovered and comprise ϳ17% of the total sequence. Thus EF hand-2 (Asp 257 -Glu 268 ), previously characterized as peptide 4.2 (mThr 254 -Gly 273 ) that mimics most of the functions of SPARC (e.g. de-adhesion, cell cycle inhibition, dissolution of focal contacts) (34,35), was not a part of this study. Further limited proteolysis of SPARC by MMP-3 resulted in the identification of a scissile bond in J2 at Arg149-Met150, and gave rise to peptide J3 containing EF hand-1 and the same C-terminal amino acid (Pro 237) as J2 (Fig. 2B). EF hand-1 (Asp 165 -Lys 176 ) binds Ca 2ϩ but differs from the canonical EF hand-2 (see below). Previously studied as synthetic peptide 3.2 (mLys 154 -Lys 173 ), this sequence exhibited essentially no effect on focal adhesion dissolution (35). In the context of a larger synthetic peptide (termed Z-2 in this study, M r 6,100), the EF hand-1 motif was also weakly inhibitory with respect to thymidine incorporation but stimulated endothelial cell migration in native collagen gels (Figs. 5 and 6). Although the molecular basis for this effect is not known, enhancement of cell migration within collagen is a newly identified activity for this sequence motif in SPARC.
Peptide J1 (Glu 2 -Pro 87 ) (Fig. 7) comprises all of SPARC domain I, an acidic, low affinity Ca 2ϩ -binding region thought to interact with mineralized components of bone. Domain I (amino acids 1-52) is poorly conserved within the SPARC family.
Because several studies describing functions of this region have been published (reviewed in Ref. 6), we focused on the Cterminal sequence of J1 extending into the FS domain (II) and reexamined the Cu 2ϩ -binding motif KHGK, in the context of the synthetic peptide Z-1 (M r 3,900) (Fig. 7). The sequence KHGK was previously studied as a 20-mer peptide termed 2.1, and as a 10-mer, 2.1a (29). Both exhibited a biphasic effect on the incorporation of [ 3 H]thymidine by BAE cells, with stimulation between 0.025 mM and 0.2 mM and inhibition at concentrations in excess of 0.4 mM (29). In the present study, using the extended sequence Z-1, we also found a biphasic effect on thymidine incorporation by BAE cells (Fig. 5), which was ϳ2fold at 0.05 mM and maximal at 0.2 mM, as well as a stimulation of vascular growth in the CAM (at 0.1 mM), followed by a return to baseline at 1.0 mM (Fig. 4). Of interest was the lack of effect of a range of concentrations of peptide Z-1 on endothelial cell migration in collagen gels (Fig. 6). Thus, the effects of the high-affinity Cu 2ϩ -and Ca 2ϩ -binding motifs of SPARC on endothelial cell adhesion and proliferation appear to be mutually exclusive.
Previous experience with the use of peptides in vitro has demonstrated the necessity of testing a broad range of concentrations under a variety of culture conditions, for confirmation of biological activity (29,36). In earlier studies we had shown a stimulatory effect of SPARC peptide 2.1 on DNA synthesis on three different types of endothelial cells, over a concentration range of 0.025-0.2 mM, as well as an inhibitory effect at concentrations in excess of 0.4 mM (29). Moreover, a maximal effect on the dissolution of focal adhesions in BAE cells was observed at 0.2 mM peptide 2.1 (35). Our present study included an extended sequence of the 10-mer 2.1a (containing the active residues KHGK) that was denoted peptide Z-1 (Table I). Interestingly, Z-1 exhibited similar concentration-dependent effects on BAE cell proliferation, data indicating that KHGK-Cu 2ϩ is active within extended flanking regions of amino acids within the FS domain of SPARC (Fig. 7). KHGK (as peptide 2.1) has been shown either to mimic or to synergize with the EF hand-2 sequence of SPARC in many of the activities observed in vitro: binding to collagen and endothelial cells, focal adhesion dissolution, antiproliferation, and regulation of several proteins associated with G 1 3 S progression (10,34,35,37). These similar or synergistic effects between the two sequences are consistent with their location in the three-dimensional structure of SPARC (27,38), a factor that could influence the effective concentrations of peptide required for bioactivity in different assays. The concentrations of peptides Z-1 to Z-3 used in this study are similar to those used recently on vascular cells for fragments of two proteins: the laminin ␥1 chain (a 12-mer at 0.1 mM) (39) and the type III Ca 2ϩ -binding repeat of thrombospondin 1 containing RGD (0.1 mM) (40), but are considerably less effective than a 17-mer derived from kininostatin, which inhibited endothelial cell migration to vitronectin at an IC 50 of 0.2 M (33). The study by Colman et al. (33) also identified a second sequence within kininostatin (a 16-mer) that potently inhibited proliferation but not migration, and was also inhibitory in the CAM assay, results consistent with those reported here for peptide Z-1 (Fig. 7).
The MMPs are a large family (ϳ24 members) of structurally and functionally related zinc endopeptidases that exert decisive roles in the turnover of ECM in the context of tissue morphogenesis and remodeling (41). Several novel functions have recently been described for MMP-3 that underscore the importance of the identification of its substrates in various pathologies. For example, transgenic mice overexpressing MMP-3 in mammary tissue exhibited spontaneous development of malignant tumors that were reversed by coexpression of TIMP-1 (42). In apolipoprotein E-deficient mice (a model for atherosclerosis) that were also null for MMP-3, there was an increase in plaque size but a decrease in aneurysm formation, in comparison with MMP-3 ϩ/ϩ mice (43). Further evidence in vivo for the consequences of impaired degradation of ECM/ fibrillar collagen by MMPs can be found in a murine model of adipose tissue development, in which inhibition of MMP activity was associated with impaired adipose accumulation that was attributed to the formation of an excessive collagen-enriched ECM (44). Since SPARC-null mice exhibit both attenuated levels of collagen and increased amounts of adipose tissue (45, 46), a condition collectively mimicking the results of Lijnen et al. (44), it becomes interesting to speculate that the absence of one or more MMPs (e.g. MMP-3) might produce a phenotype in mice similar to the lack of an MMP-3 substrate, or a regulator of an ECM substrate, e.g. SPARC. Another matricellular protein, osteopontin, has recently been identified as a substrate of MMP-3 and MMP-7 (47). In analogy to SPARC, fragments of osteopontin were identified in tumor cell lines and in remodeling tissues, and cleavage of osteopontin by MMP-3 and -7 enhanced its stimulatory effects on cell adhesion and migration (47).
Under the reaction conditions that we used, we did not detect cleavage of native mSPARC or refolded, rhuSPARC by MMP-1 or -9, consistent with our previous work (16,48) but different from reports claiming proteolysis of mSPARC (termed BM-40 in these studies) by MMP-13 (murine homolog of human MMP-1) (49) and of rhuSPARC by MMP-2, 3, 7, and 9 (50). These discrepancies could be accounted for essentially by two factors: differences in protein conformation resulting from its source and purification, and differences in the conditions of proteolysis. As both our group (19,20) and Sasaki et al. (50) have used rhuSPARC that, despite differences in source and purification procedures, showed activity and/or physical chemical characteristics consonant with a correctly folded structure, we favor the latter possibility, i.e. the enzyme to substrate ratios and diminished incubation times. Whereas Sasaki et al. (50) report minimal cleavage of SPARC by MMPs (including MMP-3) at a 1:100 enzyme to substrate ratio (by weight) after 24 h, peptides J1-3 in our study were generated reproducibly at this ratio, within 16 h. Importantly, cleavage of SPARC by one or more of the MMPs used in studies from the Timpl laboratory has revealed an enhanced (ϳ10-fold) Ca 2ϩ -dependent binding to both fibrillar and basement membrane collagens (49,50). Moreover, antibodies specific for the cleaved neoepitope at Leu 198 revealed the presence of this fragment of SPARC in tissues (51). Further work will be necessary to clarify the ultimate specificity of MMP-3 for native SPARC, and the biological consequences thereof.
We have identified novel cleavage products of SPARC produced by MMP-3, including a sequence containing EF hand-1, a variant Ca 2ϩ -binding loop thought to function as a structural element in the native protein (38). This region (contained within peptides J2 and J3) was stimulatory for cell migration but inactive in both proliferation and CAM assays. Effects of SPARC on various stages of angiogenesis have been documented (reviewed in Refs. 1 and 6). Typically associated with growing/regressing capillaries rather than larger, established vessels (18), SPARC was identified as a highly and differentially expressed pan endothelial marker in both normal and tumor-associated endothelium (52). It is now generally accepted that angiogenesis on the part of the host is required for the growth and metastasis of most solid tumors, and many of the matricellular proteins have demonstrated roles in the regulation of neoplastic progression via their effects on neovessel formation (reviewed in Ref. 1). It is still not clear whether SPARC or specific fragments of SPARC are normally involved in the regulation of angiogenesis. Moreover, the presence of inflammatory cells, cytokines, specific proteinases, and their inhibitors, as well as a requirement for ECM, serves to complicate the interpretation of different injury models. It is undisputed, however, that MMPs are important contributors to disease progression and/or resolution vis-à -vis their effects on the activation of other proteases, ECM degradation, and growth factor release or activation (30,41,53). Conversely, SPARC can regulate the activation as well as the production and release of certain of the MMPs, including MMP-3 (12,13). Solving the regulatory loops in a tissue or model system, and the contribution of each variable to resolution of the biological process, remains a challenging but realistic endeavor.