The degradation of human endothelial cell-derived perlecan and release of bound basic fibroblast growth factor by stromelysin, collagenase, plasmin, and heparanases.

Perlecan is a modular heparan sulfate proteoglycan that is localized to cell surfaces and within basement membranes. Its ability to interact with basic fibroblast growth factor (bFGF) suggests a central role in angiogenesis during development, wound healing, and tumor invasion. In the present study we investigated, using domain specific anti-perlecan monoclonal antibodies, the binding site of bFGF on human endothelial perlecan and its cleavage by proteolytic and glycolytic enzymes. The heparan sulfate was removed from perlecan by heparitinase treatment, and the approximately 450-kDa protein core was digested with various proteases. Plasmin digestion resulted in a large fragment of approximately 300 kDa, whereas stromelysin and rat collagenase cleaved the protein core into smaller fragments. All three proteases removed immunoreactivity toward the anti-domain I antibody. We showed also that perlecan bound bFGF specifically by the heparan sulfate chains located on the amino-terminal domain I. Once bound, the growth factor was released very efficiently by stromelysin, rat collagenase, plasmin, heparitinase I, platelet extract, and heparin. Interestingly, heparinase I, an enzyme with a substrate specificity for regions of heparan sulfate similar to those that bind bFGF, released only small amounts of bFGF. Our findings provide direct evidence that bFGF binds to heparan sulfate sequences attached to domain I and support the hypothesis that perlecan represents a major storage site for this growth factor in the blood vessel wall. Moreover, the concerted action of proteases that degrade the protein core and heparanases that remove the heparan sulfate may modulate the bioavailability of the growth factor.

the pericellular environment and a ubiquitous component of basement membranes. These membranes are highly specialized structures that control tissue architecture, movement of cells/compounds between the bloodstream and tissues (1), and the binding of growth factors (2). The presence of basic fibroblast growth factor (bFGF) in the extracellular matrix of cultured endothelial cells was first demonstrated in 1987 (3). Subsequently, its release was shown to be modulated by the action of heparitinase or by competition with heparin (4), thus suggesting that the growth factor was bound to a HS component present in the matrix. Radiolabeled bFGF was also shown to bind ECM and be released by the enzyme collagenase (4), thrombin (5), or plasmin (6) and by incubation with unlabeled growth factor (3). When bound to HS, bFGF was found to be resistant to the action of proteases (7). Recently, perlecan was shown to bind bFGF and to present the growth factor to its high affinity receptors on the cell surface, which suggested that this HSPG has an important role in the regulation of cell growth (8). When perlecan was treated with heparinase I it could no longer act as a co-receptor, suggesting that it was the HS on perlecan that was responsible for this activity (8). This is not surprising in light of the fact that both acidic and basic FGF bind avidly to the closely related glycosaminoglycan, heparin. bFGF bound to heparin-Sepharose requires a NaCl concentration greater than 1 M for elution (3,9). Highly sulfated regions of heparin contain a bFGF-binding hexasaccharide sequence, which is comprised of two trisulfated disaccharides with the structure IdceA (2-OSO 3 )␣1-4GlcNSO 3 (6-OSO 3 ) followed by an IdceA (2-OSO 3 )␣1-4 anhydro-D-mannitol (10). A similar sequence has been identified in HS isolated from fibroblasts (11) and from smooth muscle cells (12), except that in both cases the proportion of the disulfated GlcNSO 3 (6-OSO 3 ) was lower, suggesting that the presence of the sulfate groups at position 6 of the N-sulfated glucosamine were not involved in mediating the high affinity binding. HS isolated from whole arterial tissue, which inhibited the growth of smooth muscle cells (13), probably by interfering with the presentation of bFGF to its high affinity receptors (14), had a high proportion of the sequence isolated from fibroblasts and smooth muscle cells. When the proportion of the IdceA (2-OSO 3 )␣1-4GlcNSO 3 disaccharide was decreased with respect to the amount of the IdceA (2-OSO 3 )␣1-4GlcNSO 3 (6-OSO 3 ), the fraction did not inhibit smooth muscle cell growth (13).
The protein core of perlecan has been divided into five domains on the basis of sequence homology to other proteins and the presence of repeating motifs (15,16). Domain I is unique to perlecan and contains three serine-glycine-aspartic acid sequences that may act as glycosaminoglycan attachment sites (16). Since these HS chains are needed for perlecan to bind FGF, the integrity of domain I is important for anchoring the growth factor activity to the ECM. We reasoned that if this domain were degraded or released by proteases, it could provide a mechanism whereby anchored growth factor molecules could be liberated. It has also been suggested that domain I may be oriented toward the cell surface, which would facilitate its co-receptor activity (17). Domain II has sequence homology to the low density lipoprotein receptor, domain III has homology to the A chain of laminin, domain IV is composed of immunoglobulin-like repeats that demonstrate homology to the neural cell adhesion molecule, N-CAM, and domain V has regions that have similarity to epidermal growth factor and the globular domains of the laminin A chain (15).
The fact that bFGF binds a HSPG in the matrix and is released by enzymes such as plasmin, thrombin, heparanase, and collagenase has been known for a number of years. The identity of the HSPG responsible for this binding, however, was unknown until recently when it was shown that perlecan bound bFGF (8). The major goal of this paper was to investigate which enzymes would degrade the protein component of perlecan and whether these same enzymes could facilitate the release of the bound bFGF. This was facilitated by the use of anti-perlecan monoclonal antibodies, which were characterized with respect to their domain specificity. Plasmin, stromelysin, and rat collagenase significantly degraded the protein core, reduced the immunoreactivity toward domain I, and released significant amounts of the growth factor from the HSPG. Plasmin cleaved the protein core to leave a product that probably contained domain III, whereas stromelysin and rat collagenase degraded perlecan core protein into many fragments. Heparitinase I released bound bFGF, whereas heparinase I did not, a finding consistent with the substrate specificity of heparinase I for highly sulfated regions of heparan sulfate. Platelet extract was the most efficient agent at releasing bound growth factor, and this may be due to the presence of many HS-degrading enzymes. This supports the hypothesis that degranulation of platelets at sites of injury may be very effective at mobilizing growth factor from the matrix, thereby aiding in the wound healing process.

EXPERIMENTAL PROCEDURES
Materials-Heparin (H3149), Tris, Triton X-100, ABTS, human plasmin, and human thrombin were purchased from Sigma. Heparinase I, heparitinase I (heparinase III), and chondroitinase ABC were from Seikagaku Corporation Co. All other chemicals were of analytical grade. Tissue culture plastic ware was from either Nunc or Corning. Medium 199 containing Earle's salts was from Life Technologies, Inc. All solutions for endothelial cell culture were prepared using pyrogen-free water (Baxters), and small aliquots were filtered through sterile Zetapore 0.2-m pore nylon membranes (Cuno Pacific). All reusable glassware was treated with E-Toxa clean (Sigma) to remove pyrogens, and following several washes in tap and distilled water, it was rinsed in pyrogen-free water before sterilization by autoclaving at 120°C for 1 h followed by dry heat at 170°C for 3 h. All handling of materials was done with surgical gloves. Precast 4 -15% polyacrylamide gels, high molecular weight prestained protein standards, the Protean II electrophoresis and blotting system, and Alcian blue were from Bio-Rad. "Rainbow" prestained molecular weight markers, recombinant 125 I-labeled bFGF, peroxidase conjugated to streptavidin, and biotin-conjugated rabbit anti-mouse IgG were purchased from Amersham Corp. Protein A-Sepharose was from Pharmacia Biotech Inc. Trans [ 35 S]methionine was purchased from ICN-Flow, and [ 35 S]sodium sulfate was purchased from DuPont. Peroxidase conjugated to rabbit anti-mouse IgG, unconjugated rabbit anti-mouse IgG, and alkaline phosphataseconjugated rabbit anti-mouse IgG were purchased from Dako. 5-Bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium substrate for alkaline phosphatase was purchased from Promega. Gelatin-Sepharose was from AMRAD-Pharmacia Biotech, Inc. Foetal calf serum was from P.A. Biologicals.
Human stromelysin-1 (MMP 3) purified from fibroblasts was supplied by Dr. Jack Windsor from the University of Alabama in Birmingham, Alabama. It degraded gelatin and casein by SDS-zymography. Human gelatinases (MMP 2 and MMP 9) from 12-O-tetradecanoylphorbol-13-acetate-stimulated HL-60 cells were supplied by Dr. Guy Lyons at the Kanematsu Institute, Royal Prince Alfred hospital, Sydney. The presence of active gelatinases was assayed by zymographic analyses. Rat collagenase was purified from rat mammary carcinoma cells as described previously (18) and supplied by Ms. A. Martorana from the University of Technology, Sydney. It was shown to degrade collagen using a dried collagen film assay (19) and gelatin by zymographic analysis. The human homologue of rat collagenase has been named collagenase-3 (MMP 13) (20). The platelet extract was a gift from Dr. Lloyd Graham of the CSIRO, Division of Biomolecular Engineering, and had a protein concentration of 9 mg/ml. One ml was prepared from 1.2 ϫ 10 11 platelets by freeze-thawing. This extract was assayed and shown to contain significant heparin/heparan sulfate-degrading activity using both a labeled and nonlabeled substrate assay (21). Human TIMP-1 was a gift from Dr. Kirby Bodden from the University of Alabama in Birmingham, Alabama.
Anti-perlecan Antibodies-The anti-perlecan monoclonal antibodies were the result of a fusion using Sp2/0 myeloma cells and spleen cells from mice immunized with whole bovine corneal endothelial cell ECM, described previously (22). Monoclonal antibodies A71, A74, A76, and A81 and irrelevant mouse IgG control were purified from ascites fluid by affinity chromatography on protein A-Sepharose, as described (23). The anti-human domain III monoclonal antibody, 7B5, was produced and characterized as described previously (24). It was supplied as lyophilized ascites fluid, which was reconstituted with water before use.
Perlecan Domain-specific Fusion Proteins-Fusion proteins specific for perlecan domains II, III, and V were produced as described previously (24). An additional fusion protein consisting of bacterial maltose binding protein fused to domain I of perlecan (bases 144 -659 of the cDNA (15)) was also used in this study. For ELISA, fusion proteins were purified over amylose resin columns as before (24). For immunoblotting, crude bacterial lysates were obtained by boiling pelleted bacteria in SDS-PAGE sample buffer prior to electrophoresis.
Cell Culture-Primary cultures of human umbilical arterial endothelial cells (HUAECs) were prepared from fresh umbilical cords delivered by caesarean section at Royal North Shore Hospital, Sydney, as described (25) using 0.1% bacterial collagenase. Cells were grown on tissue culture plastic precoated for 2 h at 37°C with chicken fibronectin (10 g/ml), purified from fresh plasma by affinity chromatography on gelatin-Sepharose as described (26). Culture medium was Medium 199 with Earle's Salts containing 20% pyrogen-free fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin sulfate, 100 g/ml heparin, and 2% bovine brain extract prepared as described (27). Cells were passaged at a 1:3 split ratio after disaggregation with 0.125% trypsin, 0.02% EDTA and used between passages 6 and 10. Radiolabeled conditioned medium was prepared by culturing endothelial cells with either [ 35 S]sodium sulfate (25 Ci/ml) or [ 35 S]methionine (25 Ci/ml) in complete medium for up to 72 h. The labeled medium was collected, centrifuged at 2000 ϫ g to remove cellular debris and stored at Ϫ20°C.
Preparation of Native ECM-Human endothelial cells were plated in 96-well microtiter plates at 0.5-1 ϫ 10 4 cells/well and cultured for 7 days in complete medium to allow deposition of ECM on the surface of the wells. Native ECM was prepared using the method of Gospodarowicz and Lui (28). The ECM coated plates were stored at Ϫ20°C under phosphate-buffered saline until used.
Immunochemical Procedures-Native ECM ELISAs were performed as described previously (29) using biotin-conjugated rabbit anti-mouse IgG and peroxidase conjugated to streptavidin to enhance the signal. ABTS substrate reaction absorbance was read at 405 nm in a Bio-Rad plate reader. The reference wavelength was 490 nm. Plates coated with the various fusion proteins at 2 g/ml were washed and incubated with freshly made 0.05% Tween 20, 1% bovine serum albumin, 5% lactose for 1 h at room temperature. This solution was removed, and the plates were drained and allowed to dry at room temperature for 4 h. The plates were stored at room temperature both while in transit and until used. ELISAs performed on the fusion proteins used peroxidase conjugated directly to rabbit anti-mouse IgG instead of using biotin and streptavidin enhancement.
For immunoprecipitation, protein A-Sepharose was saturated with rabbit anti-mouse IgG (RAM) by incubating 1 ml of a 50% suspension with 200 l of the antibody solution for 2 h at room temperature with rocking. The RAM-saturated protein A-Sepharose was washed with Tris-buffered saline (TBS) to remove unbound RAM, and TBS was added to give a final volume of 1 ml (50% suspension). Samples (1 ml) of radiolabeled HUAEC conditioned medium were precleared by a 2-h incubation at room temperature with 100 l of RAM-saturated protein A-Sepharose suspension. Simultaneously, 100-l samples of RAM-saturated protein A-Sepharose suspension were incubated with 200 l of purified monoclonal antibody (mAb) in TBS at 100 g/ml. The mixtures were microcentrifuged, and the mAb-loaded protein A-Sepharose was washed twice with TBS and resuspended in the precleared medium. After a 2-h incubation at room temperature with shaking, the protein A-Sepharose was washed four times with TBS and resuspended in 50 l of SDS sample buffer (10% (v/v) glycerol, 0.4% (w/v) SDS, 0.001% (w/v) bromphenol blue, 10 mM Tris, pH 6.8), boiled for 10 min, and microcentrifuged to pellet the protein A-Sepharose. The supernatant was electrophoresed on SDS-PAGE gels, dried, exposed to a phosphor screen (Molecular Dynamics) and imaged on a PhosphorImager (Molecular Dynamics) using ImageQuant software.
Immunoprecipitation of labeled bFGF-perlecan complexes was achieved using the same protocol as described above except that the HUAEC conditioned medium was not labeled and recombinant 125 Ilabeled bFGF was added to give a final activity of 0.2 Ci/ml of medium before precipitation with mAb-loaded protein A-Sepharose. Treatment of the labeled bFGF-perlecan-mAb-protein A-Sepharose complexes with NaCl or heparin was performed at room temperature for 16 h. Treatment with degradative enzymes was for 16 h at 37°C (as described below). After all treatments, the samples were microcentrifuged, and the supernatants were removed, counted in a ␥ counter (LKB 1275 Minigamma), and expressed as counts/l/min. The remaining pellet was washed twice with phosphate-buffered saline, 50 l of SDS-PAGE sample buffer was added, and the samples were prepared for SDS-PAGE analysis. The amount of labeled bFGF released by the various treatments was expressed as a percentage of the amount bound in the initial immunoprecipitate, which was estimated by extraction of the immunoprecipitate with 6 M urea, 0.2% SDS in phosphate-buffered saline. The values were corrected for the amount of 125 I-labeled bFGF released by buffer alone by applying the following formula: corrected % ϭ (uncorrected % Ϫ buffer only %) ϫ 100/(100 Ϫ buffer only %).
Protease and Endoglycosidase Treatment-Digestions with heparitinase I, heparinase I, or chondroitinase ABC were performed at 37°C for 16 h in 10 mM Hepes, 3 mM Ca(CH 3 COO) 2 , 0.1% Triton X-100, pH 7.0, 1 mM benzamidine, 1 mM ⑀-amino caproic acid, 1 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, as described by Lindblom et al. (30). Digestion with human plasmin, human thrombin, rat collagenase, or human stromelysin was performed using the same buffer without the addition of the protease inhibitors for 16 h at 37°C. The activity of these proteases was titrated against native matrix using the ELISA approach. The concentration of enzymes chosen was the minimum concentration that gave 100% of the degradative activity as demonstrated using the anti-domain V antibody, A74. In the cases of heparitinase I and platelet extract, reasonable maximum concentrations were selected. The final concentrations of the various enzymes were as follows: heparitinase I and heparinase I, 0.1 unit/ml; chondroitinase ABC, 0.5 units/ml; plasmin, 0.6 units/ml; thrombin, 1.2 units/ml; rat collagenase, 5 g/ml; gelatinases, 1 g/ml; and stromelysin, 2 g/ml. Platelet extract was used at a final concentration of 90 g/ml in enzyme buffer either with or without added protease inhibitors. For all enzyme digestions the aim was to perform exhaustive cleavage on the substrate in an attempt to demonstrate an absolute presence or absence of cleavage. Enzyme digestions of immunoprecipitated molecules were performed after the four washes with TBS in a volume of 50 l and were terminated by adding SDS-PAGE sample buffer and boiling the samples for 10 min. Enzyme digestions of native ECM were also performed in a volume of 50 l in the wells of the microtiter plates using the same conditions and final concentrations of enzymes. The plates were washed twice with phosphate-buffered saline prior to analysis by ELISA.
SDS-Gel Electrophoresis and Immunoblotting-SDS-PAGE was performed using 4 -15% polyacrylamide gradient gels as described by Laemmli (31). After electrophoresis the gels were washed with 25% (v/v) ethanol, 10% (v/v) acetic acid in H 2 O to remove any SDS and stained with 0.125% (w/v) Alcian blue in 25% (v/v) ethanol, 10% (v/v) acetic acid in H 2 O for 30 min at room temperature. After staining, the gels were destained using 25% (v/v) ethanol, 10% (v/v) acetic acid in H 2 O. The gels were dried, and radioactive bands were visualized using a PhosphorImager system. Crude bacterial lysates for Western blotting analysis were size separated on 10% polyacrylamide gels under reducing conditions before transfer to nitrocellulose membrane (32). Immunoreactive bands were detected using alkaline phosphatase-conjugated rabbit anti-mouse IgG as secondary antibody and 5-bromo-4-chloro-3indolyl phosphate with nitro blue tetrazolium as substrate.  Fig. 1 (lanes 1 and 4, respectively). The immunoprecipitated molecule only just entered the 4 -15% polyacrylamide gel, consistent with the expected size of perlecan. When [ 35 S]sulfate-labeled perlecan was digested with heparitinase I, the high molecular weight band disappeared (Fig. 1, lane 5), indicating that all the incorporated label was present as HS. When the [ 35 S]methionine-labeled perlecan was digested with heparitinase I, the labeled band moved further into the gel (Fig. 1, lane 2), yielding an estimate of the molecular mass of the protein core of human perlecan as 450 kDa. Chondroitinase ABC digestion had no effect on either the intensity or the mobility of either [ 35 S]sulfate-labeled (Fig. 1, lane 6) or [ 35 S]methionine-labeled endothelial cell-derived perlecan (Fig.  1, lane 3). Incubation with both heparitinase I and chondroitinase ABC gave the same results as those for heparitinase I alone (data not shown), supporting the suggestion that human endothelial perlecan contained only HS and no CS. The mobility of the protein core was unaffected by reduction with 10 mM dithiothreitol (data not shown). Results from immunoprecipitation experiments using any of four mAbs were identical (data not shown). The four antibodies reacted, by ELISA, with perlecan in native HUAEC matrix (data not shown). To determine further the specificity of the antibodies, they were screened by Western blotting and ELISA against recombinant fusion proteins encoding domains I, II, III, and V of the perlecan protein core. The fusion proteins were generated as bacterial maltose- binding protein fusions as described previously (24). Western blotting demonstrated that A71 reacted specifically with domain I (Fig. 2A, lane 1), whereas A74 (lane 2), A76 (lane 3), and A81 (lane 4) did not react with the domain I fusion protein ( Fig.  2A). There was some evidence of nonspecific reactivity with A71 and A76 ( Fig. 2A). All four antibodies were negative on Western blots to domain II, III, or V (data not shown). By ELISA, A71 reacted with domain I (Fig. 2B), which supported the results obtained by Western blotting, and A74 reacted with domain V (Fig. 2B). Both A76 and A81 showed no significant immunoreactivity (data not shown). In summary, these data indicate that we have specific monoclonal antibodies directed against different protein domains of perlecan.

Characterization of Monoclonal Antibodies
Enzymatic Cleavage of Perlecan-To establish sites of proteolytic cleavage in the perlecan protein core we used a variety of known proteases in combination with heparitinase digestion on immunoprecipitates. Heparitinase I treatment alone revealed the protein core of 450 kDa (Fig. 3, lanes 2 and 7). In the presence of both heparitinase I and plasmin the protein core was digested to a major product with an estimated molecular mass of ϳ300 kDa (Fig. 3, lane 3). In the absence of heparitinase, plasmin digestion produced a lower molecular weight smear with no major 300-kDa cleavage product (data not shown). Thrombin in combination with heparitinase I had no effect on the molecular weight of the protein core (Fig. 3, lane  4). In contrast, stromelysin (MMP 3) in the presence of heparitinase I degraded the perlecan core into several fragments of various molecular weights with no evidence of a major cleavage product (Fig. 3, lane 6). This enzyme also cleaved the whole proteoglycan (in the absence of heparitinase), however the large molecular weight smear seen with plasmin digestion was not evident, suggesting that the digestion with stromelysin was more complete (data not shown). Incubation of the perlecan protein core with 0.5 g/ml stromelysin showed incomplete degradation (data not shown). Incubation with rat collagenase also demonstrated cleavage of the protein core, although the digestion was not as complete, with a number of the smaller fragments being absent (Fig. 3, lane 8). A mixture of the two gelatinases (MMP 2 and MMP 9) had no effect on the size of the protein core (Fig. 3, lane 6) or the intact proteoglycan (data not shown). These findings indicate that perlecan has specific protease-sensitive sites in its protein core and that of the enzymes tested, plasmin, rat collagenase, and stromelysin (also known as a "proteoglycanase") were the most efficient at fragmenting the protein core. The activities of both rat collagenase and stromelysin toward the protein core of perlecan were completely inhibited when digested in the presence of a one molar excess (w/w) of the specific metalloproteinase inhibitor, TIMP-1 (data not shown). When the intact proteoglycan was incubated with platelet extract, in either the presence or absence of protease inhibitors, a protein core band identical in relative molecular mass to that obtained with heparitinase I was seen, suggesting that the major perlecan-degrading activity in platelet extract was that of a heparanase (data not shown).
Removal of Specific Perlecan Domains-The same major plasmin cleavage product was demonstrated when A71, A74, or A76 was used to immunoprecipitate the perlecan, suggesting that the generation of the 300-kDa band was not due to protection of the protein core by the antibody bound to a particular coli. were coated onto wells of an ELISA plate at 2 g/ml and probed with 2 g/ml of mAbs. Absorbances for A71 and A74 are shown and have been corrected for the absorbance reading obtained using the irrelevant mouse IgG control. mAbs A76 and A81 did not recognize any of the domain proteins by ELISA.

FIG. 3. Enzyme digestion of perlecan immunoprecipitates.
[ 35 S]methionine-labeled perlecan was immunoprecipitated from HUAEC conditioned medium as described under "Experimental Procedures." Immunoprecipitated material was incubated with buffer alone (lanes 1 and 7) or with the enzymes indicated by ϩ (lanes 2-6 and 8). Digestion products were resolved on 4 -15% polyacrylamide gels and visualized by phosphor imaging. This figure is a composite of three gels, two of which were run in the same experiment (lanes 1-6). The relative positions of molecular mass standards (in kDa) are shown on the left. Top, the origin of the running gel. mated size of the major plasmin cleavage product. Therefore, it was of interest to determine whether the plasmin cleavage product contained either domain III or V and whether the enzyme had removed domain I. The removal of domain I would remove the region that has the HS attached, thereby facilitating the liberation of growth factor activity. This was tested by digesting native HUAEC matrix with the various enzymes and probing with the domain-specific antibodies. Plasmin removed the immunoreactivity toward domain I and V, whereas it had less effect upon the immunoreactivity of domain III (Fig. 4). This suggested that domains I and V had been removed and supported the hypothesis that the 300-kDa plasmin cleavage product contained domain III. Stromelysin removed immunoreactivity toward domains I, III, and V (Fig. 4), supporting the SDS-PAGE cleavage data indicating that stromelysin cleaved perlecan at multiple sites thereby making it an efficient way of removing the HSPG from the matrix. Rat collagenase removed immunoreactivity toward domain V and reduced by approximately 80 and 50% the immunoreactivity toward domains I and III, respectively (Fig. 4). This data also supported the SDS-PAGE cleavage data obtained for rat collagenase. Digestion with thrombin had no effect on domain I reactivity, removed immunoreactivity toward domain V, and reduced by ϳ50% the immunoreactivity of domain III (Fig. 4). Heparitinase I had no effect on the immunoreactivity of domains I and III but did reduce by ϳ50% the immunoreactivity of domain V (Fig. 4). Chondroitinase ABC digestion had no effect on the immunoreactivity of domain I, III, or V (data not shown). Incubation with platelet extract containing protease inhibitors increased immunoreactivity toward domain I, suggesting that the platelet extract removed the HS that may have been masking this epitope (Fig. 4). Domain V immunoreactivity was reduced with platelet extract, consistent with those results obtained using heparitinase I. In fact, the A74 epitope was sensitive to exposure to any of the enzymatic treatments used. In summary, we demonstrated that immunoreactivity to domain I was removed by either plasmin, stromelysin or rat collagenase, whereas thrombin and heparitinase I had no effect, and platelet extract increased it by removing masking HS. Domain III immunoreactivity was less susceptible to protease digestion than domain I, except for treatment with stromelysin, which all but removed the immunoreactivity toward this domain. Domain V immunoreactivity was very sensitive to all of the enzymes tested, suggesting that the A74 epitope may be more reliant on the tertiary structure of perlecan.
Binding of Basic FGF to Perlecan and Competition by Heparin-To establish whether bFGF bound perlecan under physiologic conditions, we added 125 I-labeled bFGF to HUAEC conditioned medium and then immunoprecipitated the medium with mAb A76. Recombinant 125 I-labeled bFGF (18 kDa) was added both in the presence and absence of heparin (100 g/ml). In the absence of heparin, significant amounts of [ 125 I]bFGF were co-immunoprecipitated (Fig. 5, compare lane 3 with nonspecific control in lane 1). There was some evidence of the formation of dimers and trimers of bFGF, as bands at 36 and 54 kDa were also evident. When the conditioned medium also contained 100 g/ml heparin, the intensity of the co-immunoprecipitated 18-kDa band was reduced almost to background levels, i.e. the levels produced by an irrelevant mouse IgG (Fig.  5, compare lane 4 with lane 2). The reduction of the binding of bFGF to perlecan in the presence of heparin suggested that the latter was competing with the HS chains of perlecan for bFGF binding. This was confirmed in experiments where heparin was capable of eluting bound growth factor from the proteoglycan (Fig. 6, lane 2). Results obtained from immunoprecipitation experiments using the anti-domain I mAb, A71, were the same as those results obtained using A76 (data not shown). These results indicate a specific interaction between the soluble perlecan and bFGF.
Enzymatic Release of bFGF Bound to Perlecan-Because plasmin, stromelysin, and rat collagenase digested the protein component of perlecan and were capable of reducing the immunoreactivity of domain I in the matrix, it was of interest to determine whether these proteases, as well as those enzymes that removed the HS from domain I, were capable of liberating bound bFGF. To test this hypothesis, the immunoprecipitated perlecan-bFGF complexes were used as the starting material. After incubation with the various enzymes the amount of complexed bFGF remaining was visualized by exposing the polyacrylamide gels to phosphor screens and analyzing them on a PhosphorImager. The amount of bFGF complexed to perlecan was reduced from control levels (Fig. 6, lanes 1, 7, and 10), to those obtained by incubation with heparin (Fig. 6, lane 2), with heparitinase I (Fig. 6, lane 4), platelet extract (Fig. 6, lane 6), stromelysin (Fig. 6, lane 8), and rat collagenase (Fig. 6, lane  11). The intensity of the band after incubation with heparinase I (Fig. 6, lane 3) or plasmin (Fig. 6, lane 9) was reduced but was still greater than background levels. Thrombin (Fig. 6, lane 5) did not reduce the intensity of the bFGF band. Of note was the finding that both rat collagenase and stromelysin reduced the molecular mass of the released bFGF by approximately 3 kDa HUAEC ECM was isolated, and ELISAs were performed as described under "Experimental Procedures." The plates were treated with the various enzymes for 16 h at 37°C using the same concentrations as those used in Fig. 3. Domain I immunoreactivity was tested with mAb A71, that of domain III with mAb 7B5, and that of domain V with mAb A74.
FIG . 5. Perlecan binds bFGF, and the binding is competed by  heparin. [ 125 I]bFGF was added to HUAEC medium, which was conditioned in either the presence (lanes 2 and 4) or absence (lanes 1 and 3) of heparin (100 g/ml). The presence of heparin is indicated by a ϩ above the lane. Immunoprecipitated complexes were collected by centrifugation, resolved on SDS-PAGE, and visualized by phosphor imaging. Lanes 1 and 2 are immunoprecipitates using an irrelevant mouse IgG, and lanes 3 and 4 were obtained using mAb A76. The relative position of molecular mass standards (in kDa) are shown on the left. Top, the origin of the running gel. (Fig. 6, lanes 8 and 11, respectively) suggesting that these enzymes were cleaving the growth factor. Incubation with a one molar excess of the metalloproteinase inhibitor, TIMP-1, inhibited the release and cleavage of bFGF by rat collagenase and stromelysin (data not shown), demonstrating that these effects were due to the specific action of the metalloproteinases. Quantitative estimates of the amount of growth factor released by each treatment were obtained by expressing the radiolabel released as a percentage of total extracted from the immunoprecipitate with 6 M urea (Table I). Significant amounts of bFGF were released by incubation with platelet extract, heparitinase I, stromelysin, rat collagenase, and plasmin (p Ͻ 0.001; analysis of variance and Student Neuman Keuls tests). Platelet extract was the most effective treatment overall, and stromelysin was the most effective protease. Heparinase I and thrombin released small but significant amounts of the growth factor, whereas 5 M NaCl did not.
We tested the hypothesis that bound bFGF may protect the HS from cleavage by heparinase I by immunoprecipitating 35 SO 4 -labeled perlecan (from medium containing heparin that removed any endogenously bound bFGF) and digesting the immunoprecipitate with either heparitinase I or heparinase I in the presence or absence of an excess of bFGF. Both enzymes were effective at removing the heparan sulfate from immunoprecipitated perlecan (Fig. 7, lanes 2 and 5). The intensities of the various bands were quantitated, using the ImageQuant software, and compared with the intensity of the control band, which was taken as 0% digestion. Heparitinase I digested 96 and 98% of the 35 SO 4 -labeled HS in either the absence or presence of 0.5 g of bFGF, respectively (Fig. 7; compare lanes  2 and 3). Heparinase I was less effective at digesting the HS, as it degraded 73 and 75% in either the presence or absence of growth factor, respectively (Fig. 7; compare lanes 5 and 6). These data suggested that the growth factor was not protecting the HS from cleavage by either heparitinase I or heparinase I. The reason why bFGF did not protect the HS from heparinase I cleavage may be due to the fact that although heparinase I has a preference for highly sulfated regions, it may be capable of cleaving other regions of the HS chain.

DISCUSSION
Perlecan derived from human endothelial cells contained HS and no chondroitin sulfate, as demonstrated by its sensitivity to heparitinase I and its full resistance to chondroitinase ABC digestion. This is the most common form of the molecule (1), although some forms of the basement membrane proteoglycan have been described that contain chondroitin sulfate and heparan sulfate on the same core protein (33). We estimated the mass of the protein core of perlecan to be ϳ450 kDa, which was very similar to the published mass of 467 kDa deduced from cDNA cloning from human tumor and non-tumor cell lines (15,16). The anti-perlecan monoclonal antibodies were characterized with respect to their domain specificity and used to study the effects of protease treatment on the various domains. This study was performed in view of the fact that perlecan was recently shown to bind bFGF (8) and to have the necessary HS sequences to cross-link with the high affinity receptors on the   7. The presence of bFGF does not affect digestion of perlecan by heparinases. 35 SO 4 -labeled perlecan was immunoprecipitated as described for Fig. 1 using A76 antibodies. The immunoprecipitates were incubated with the treatments indicated by a ϩ above the lane, incubated in SDS-PAGE buffer, electrophoresed through a 4 -15% polyacrylamide SDS gel, stained with Alcian blue, dried, exposed to a phosphor screen, and analyzed with a PhosphorImager. The intensity of the various bands was quantitated using ImageQuant software. The concentration of both heparitinase I and heparinase I used was 0.1 unit/ml. 0.5 g of bFGF was added to the immunoprecipitate/enzyme mix in lanes 3 and 6. This figure is a composite of two minigels run on different days with their appropriate controls. The relative positions of molecular mass standards (in kDa) are shown on the left. Top, the origin of the running gel. cell surface. Since the HS is attached to domain I of perlecan (1), and since the removal of this domain by proteases would release a HS-growth factor complex, it was of interest to determine which proteases were effective at degrading the protein core and in particular which proteases removed domain I. It was of further interest to determine whether these same proteases were able to release bFGF that was bound to perlecan.
In the present study we demonstrated that stromelysin, rat collagenase, and plasmin reduced the immunoreactivity toward the domain I antibody and were also effective at releasing bound growth factor. Plasmin was previously shown to release bound bFGF from whole matrix preparations (6). It was not known, however, which component of the matrix functioned as the plasmin substrate or whether the serine protease released the growth factor indirectly by activating other latent proteases (34). We demonstrated here that plasmin cleaves immunopurified perlecan, leaving a major product with a molecular mass of ϳ300 kDa, which still contained domain III, thereby suggesting that the serine protease cleaved perlecan outside of this domain. In addition to the demonstration that plasmin cleaved perlecan, we demonstrated the involvement of metalloproteinases in perlecan degradation, as stromelysin (MMP 3) and rat collagenase (MMP 13) both degraded the basement membrane proteoglycan. MMP 3 was first described as a "proteoglycanase" because it degraded proteoglycans isolated from cartilage (35). More recently, it has been shown to cleave the aggregating cartilage proteoglycan, aggrecan, at a single site close to the N-terminal region of the G1 domain (36). In contrast, stromelysin cleaved perlecan at many sites, giving rise to fragments of various molecular weights. This enzyme was also very efficient at removing the immunoreactivity toward domains I, III, and V in HUAEC ECM and at liberating bound bFGF from perlecan. Collectively, the data suggest that the enzyme degraded the protein core at multiple locations. Rat collagenase also cleaved the protein core of perlecan, giving rise to a different digestion pattern, which showed no evidence of smaller fragments. The human homologue of rat collagenase has been cloned recently and termed collagenase-3. The human recombinant enzyme was found to degrade fibrillar type I collagen but not gelatin or casein (20). The work described in this paper demonstrates that the native rat enzyme has significant "perlecanase" activity. The cleavage of perlecan by stromelysin and rat collagenase was an attribute not shared by all metalloproteinases, as a mixture of the two gelatinases (MMP 2 and MMP 9) had no effect on core protein size or the release of bound growth factor.
The serine protease thrombin has been shown to release bFGF from native matrix (5). However, in our system, it had little effect on either the immunopurified protein core or the immunoreactivity of domains I and III of perlecan in matrix. Also, we showed that thrombin released very little of the growth factor from perlecan. Therefore, thrombin may have been removing bFGF from native matrix indirectly by degradation of the surrounding ECM, by activating ECM-bound proteases, or by stimulating heparanase activity (37). In contrast, heparitinase I (heparinase III) and platelet extract released significant amounts of bFGF from perlecan, whereas heparinase I released a smaller amount. This is consistent with results obtained elsewhere using native ECM as the source of binding HSPG (4,38). Platelet extract may contain more than one heparan/heparin degrading activity and has been used as a starting material to purify platelet heparitinase (39). The substrate specificity of the heparan sulfate lyases may explain why heparitinase I released bFGF from perlecan at a higher rate than heparinase I. Heparinase I has a specificity for degrading highly sulfated, heparin-like regions that have a high propor-tion of di-or trisulfated disaccharides (40). These same regions have been shown to bind bFGF (10,41) and have also been shown to inhibit bFGF-induced mitogenesis (14). When bFGF was present, however, our results showed that the growth factor did not interfere with the activity of either heparitinase I or heparinase I. The reason why a small amount of the growth factor was released from perlecan by digestion with heparinase I yet the enzyme was capable, although less so than heparitinase I, of degrading perlecan-HS in the presence of bFGF is unknown but may be due to differences between the two experimental systems (i.e. obtaining 125 I-labeled bFGF-perlecan complexes and adding enzyme versus obtaining 35 SO 4 -labeled perlecan and adding both enzyme and unlabeled bFGF). Heparitinase I (heparinase III), on the other hand, preferentially degrades the less sulfated regions of HS, which are more common in "non-heparin" HS (42) and often separate bFGF binding sequences. Therefore, the action of heparitinase I on perlecan-bFGF complexes was to liberate more efficiently bFGF bound to its HS binding sequence.
Bound bFGF can be released from heparin immobilized to Sepharose beads by treatment with 1.4 -1.6 M NaCl (3,9), and can be extracted from native matrix with 3 M NaCl (3). Our data demonstrate that 5 M NaCl does not remove significant quantities of the growth factor from perlecan, thereby suggesting that binding between perlecan and bFGF is very specific and avid. This apparent increase in avidity may be due to the fact that in our assay incubation mixture we have the whole proteoglycan bound to immunoglobulin, and incubation with 5 M NaCl may increase any hydrophobic attraction that may exist between bFGF and either the perlecan protein core or immunoglobulin. The previous use of 3 M NaCl to remove bound bFGF from matrix (3) may have been possible due to the presence of other HSPGs besides perlecan in the matrix that bind bFGF with a lower affinity. The finding that heparin removes bFGF from perlecan is likely due to the fact that in the competitive reaction mixture, perlecan was present in much smaller amounts when compared with the 100 g/ml of competing glycosaminoglycan. Heparin has also been shown to be a competitive inhibitor of bFGF binding to matrix at concentrations 10 times lower than those used in these studies (3). We could not liberate significant amounts of bound bFGF with up to 15 g/ml of "cold" growth factor. Due to economic constraints and the logistics of the assay (i.e. 50-l volume of perlecangrowth factor complex attached to protein A-Sepharose beads) we were unable to go above this concentration. Taking the concentration of bFGF required for 50% displacement of labeled growth factor as in excess of 15 g/ml, we calculated that the K d had to be less than 800 nM, which is consistent with previous estimates for the affinity of bFGF for extracellular matrix and HSPGs (4,43). We are planning to isolate perlecan in the absence of immunoglobulin and protein A-Sepharose and repeat these experiments.
The binding of bFGF to HSPGs in matrix and its release by incubation with heparin or digestion of the matrix with enzymes are not novel findings. However, in all of the previous experiments used to study bFGF-HSPG interactions, whole native matrix was used as the source of HSPG. Since the matrix may contain more than one HSPG as well as other proteoglycans, it was unknown which HSPG was responsible for the binding of the growth factor. Furthermore, these experiments were complicated by the presence in the matrix of endogenous proteases, heparanases, and other growth factors. Our data demonstrate directly that perlecan binds bFGF very tightly and that significant amounts are released by proteases that degrade the protein component of the proteoglycan, by heparan sulfate-degrading enzymes that remove the HS, or by incubation of the complex with a competitive ligand such as heparin. Because perlecan has been shown to possess the oligosaccharide sequences necessary to activate cell surface receptors (8), it has been assigned a co-receptor role. Therefore, growth factor bound to perlecan may be very relevant to processes involving cell growth and differentiation. The regulatory mechanisms involved in the synthesis/degradation of this proteoglycan may provide ways of controlling the bioactivity of the growth factor.