Active and Tissue Inhibitor of Matrix Metalloproteinase-free Gelatinase B Accumulates within Human Microvascular Endothelial Vesicles*

,

Human gelatinase B is involved in tissue remodeling and angiogenesis. It is thought to be synthesized and rapidly secreted as an inactive precursor. In this report, we have shown that human endothelial cells accumulate active forms of gelatinase B in the cytosol. Microvascular but not macrovascular endothelial cells dramatically increased the expression of cytosolic gelatinase B in response to phorbol myristate acetate. Western blotting showed that tissue inhibitor of metalloproteinase-1 (TIMP1) was also present in the cytosol. Whereas gelatinase B was complexed with TIMP1 in the conditioned medium, it existed as a free enzyme in the cytosol, suggesting that the formation of gelatinase B and TIMP1 complex occurs after their secretion. Immunogold electron microscopy revealed that gelatinase B was localized in secretory vesicles which were especially prominent in invading pseudopodia. In contrast, TIMP1 was found throughout the cytoplasm but was not present in the gelatinase vesicles. The accumulation of intracellular activated gelatinase B, ready for rapid release, may facilitate the migration of microvascular endothelial cells during angiogenesis.
The formation of capillaries from pre-existing microvessels (angiogenesis) occurs in a variety of normal and pathological conditions, including wound healing, tumor growth and metastases, and arthritis (1). During angiogenesis, microvascular endothelial cells secrete proteases, migrate through the underlying extracellular matrix, and proliferate to organize into new blood vessels. The initial step of angiogenesis requires focalized degradation of the basement membrane (2). The process is carried out by at least two matrix metalloproteinases (MMPs), 1 gelatinase A and gelatinase B, both of which degrade basement membrane collagens (3) and are produced by many cell types, including endothelial cells (4,5). They are unique among the MMPs in that their latent forms are thought to be secreted in physical association with their natural inhibitors, the tissue inhibitors of MMPs (TIMPs) (6,7). Pro-gelatinase B binds to TIMP1, whereas pro-gelatinase A complexes with TIMP2, both via the COOH-terminal domain of the enzymes (6,8).
With the exception of neutrophil gelatinase B, which is stored in cytoplasmic granules (9), MMPs are thought to be synthesized and rapidly secreted as latent precursors that must be processed extracellularly to their active forms to express enzymatic activity (10). Gelatinase A binds to and can be activated by the membrane-type MMP (MT1-MMP) at the cell surface of several tumors, transformed and normal cell types (11,12). The activation of gelatinase B is thought to occur in the extracellular milieu, although the mechanism is not clearly understood. Recent reports have suggested that stromelysin-1 may be responsible (6,14,15). Ginestra et al. (13) have recently shown that membrane vesicles shed from human HT1080 fibrosarcoma cells contained active gelatinase B. In this report, we have demonstrated that microvascular endothelial cells are capable of accumulating active gelatinase B in the cytoplasm. Furthermore, this active enzyme is separately compartmentalized from TIMP1 by being located in secretory vesicles.

EXPERIMENTAL PROCEDURES
Cells-Human microvascular endothelial cells (FSE) derived from neonatal foreskin obtained after circumcision were isolated using Ulex europaeus 1-coated Dynabeads as described previously (16). Macrovascular endothelial cells were obtained from human umbilical veins (HUVE) isolated as described by Jaffe (17). FSE were grown and maintained in Biorich medium (ICN Biomedicals, Aurora, OH) containing 30% normal pooled human serum (derived from healthy volunteers) plus 100 g/ml endothelial cell growth supplement prepared as described by Maciag et al. (18) and 50 g/ml heparin (Sigma). HUVE were grown in Biorich containing 20% fetal calf serum plus 50 g/ml endothelial cell growth supplement and 50 g/ml heparin. Cells were used at passage 4.
Experimental Protocol-HUVE and FSE (3 cell lines each) were cultured on 60-mm dishes (Becton Dickinson) at a density of 1.5 ϫ 10 6 cells/dish in growth medium. Endothelial cells were washed twice with Hanks' balanced salt solution and preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h. The cells were then replaced with fresh basal medium and incubated in the absence or presence of 100 ng/ml phorbol myristate acetate (PMA) (Sigma) for 24 h. Human serum was necessary to maintain an intact confluent monolayer of microvascular cells throughout the course of the experiment (5). Gelatinases present in human serum were removed by running the serum through a gelatin-Sepharose affinity column (3) (Pharmacia Biotech Inc.).
Conditioned Media and Triton X-114 Cell Extracts-After incubation, the conditioned medium was collected and the cytosolic fraction was separated from the membrane fraction by Triton X-114 extraction as described by Lewalle et al. (19). The extract was partitioned into the detergent (membrane fraction) and aqueous phases (cytosolic fraction) at 37°C for 5 min and centrifuged at 5000 ϫ g for 2 min. The aqueous phase was then collected. To eliminate any carry-over effect from the aqueous fraction, the detergent phase was repartitioned three times in 1.5% Triton X-114. It was then concentrated by mixing with 30 l of packed gelatin-Sepharose (Pharmacia) with end-over-end rotation for 30 min at 4°C. The bead suspension was centrifuged at 8000 ϫ g for 2 min and the pellet was then analyzed by gelatin zymography.
Flow Cytometric Analysis-Flow cytometry was performed as de-* This work was supported by the Northern Sydney Area Health Service, the Henry Langley Fellowship, the Wenkart Foundation, and the Rebecca L. Cooper Medical Research Foundation. 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.
‡ To whom correspondence should be sent. Tel.: 612-99266043; Fax: 612-99266269; E-mail: cjackson@med.usyd.edu.au. 1 The abbreviations used are: MMP, matrix metalloproteinase; TIMP, tissue inhibitor of MMP; MT1-MMP, membrane-type 1 MMP; FSE, neonatal foreskin microvascular endothelial cells; HUVE, human umbilical vein endothelial cells; PMA, phorbol myristate acetate; APMA, aminophenylmercuric acetate. scribed previously (20) with modifications. Cells were grown to confluence in 35-cm 2 dish and treated with either basal medium alone or basal medium containing 100 ng/ml PMA for 24 h. The cells were then detached from the monolayers using 20 mM EDTA in washing buffer (2% fetal calf serum in phosphate-buffered saline) for 10 min at 37°C and scraped using a rubber policeman. After centrifugation at 1100 rpm for 10 min, the pellet was resuspended and fixed in freshly prepared 2% paraformaldehyde in phosphate-buffered saline for 30 min at room temperature. The cells were then washed and permeabilized for 10 min with 10% Triton X-100, washed, centrifuged, and the primary monoclonal mouse antibody against gelatinase B (1:50 dilution, Oncogene Science, Uniondale, NY) was added for 60 min at room temperature. Fluorescein isothiocyanate-conjugated secondary antibody was then applied for 30 min and the cells were then analyzed by flow cytometry (Coulter Elite, Hialeah, FL).
Immunoblotting-MMPs and TIMP1 were detected by immunoblot analysis after SDS-polyacrylamide gel electrophoresis. Antibodies to stromelysin-1, gelatinase B, and TIMP1 were used at 1 g/ml (Oncogene Science).
Immunogold Electron Microscopy-FSE (7 ϫ 10 4 ) were seeded onto type I collagen gel and prepared as described previously (22) in an culture well insert for 24 h before PMA (100 ng/ml) was added for a further 24 h. The gel was removed, fixed with 2% paraformaldehyde and 0.05% glutaraldehyde in phosphate-buffered saline, pH 7.4, for 1 h at 4°C, and then dehydrated in ethanol and embedded in LR White acrylic resin (Emgrid, Australia). Single and double immunogold electron microscopy was carried out as described previously (23) using gelatinase B and TIMP1 antibodies (1:50 dilution) or nonimmune IgG antibody. Thin sections were examined with the Joel 100-S electron microscope.

Cellular Localization of Gelatinase B by Flow Cytometry-
Previous workers have shown that cultured human endothelial cells secrete gelatinase B into the conditioned medium in response to PMA (4,5). To determine whether gelatinase B was also localized to the endothelial cell, we performed flow cytometry using an antibody directed against the latent and active forms of gelatinase B. Results are shown in Fig. 1. Under basal conditions, FSE showed a 2.1-fold increase in gelatinase B expression compared with the nonimmune control. Stimulation of FSE with PMA resulted in a 9.6-fold increase in the expression of gelatinase B compared with unstimulated cells. In contrast, HUVE expressed markedly lower levels of gelatinase B.
Gelatinase B Present in Endothelial Cell Conditioned Medium and Cell Fractions-To determine whether gelatinase B was located on the cell membrane or intracellularly, the membrane and cytosolic fractions from endothelial cells were separated using Triton extraction as described under "Experimental Procedures." Gelatinase B levels were measured in these fractions as well as in the conditioned medium using zymography. Results are shown in Fig. 2. Under basal conditions, gelatinase B was not detected in the conditioned medium or membrane fraction of endothelial cells. However, endothelial cells expressed gelatinase B in the cytosolic fraction as three distinct bands at 88, 82, and 74 kDa, the latter two bands representing the active forms of gelatinase B. PMA markedly increased gelatinase B expression in FSE (7.1-fold compared with basal using scanning densitometry; mean, 3 cell lines) but not in HUVE. PMA also stimulated both the latent and active forms of gelatinase B in the membrane fraction of FSE. However, only the latent form was present in the conditioned medium of FSE and to a lesser extent in HUVE.
To assess whether the activation of gelatinase B in the cytosolic and membrane fractions was caused by the Triton extraction procedure, the conditioned medium from PMA-stimulated FSE, which contained only the latent enzyme, was incubated with the detergent for 16 h at 37°C. This treatment did not activate the latent gelatinase B present in the conditioned medium, thus indicating that gelatinase B was not activated by the extraction technique (Fig. 3).
In addition to gelatinase B there was a prominent band of gelatinolytic activity at 66 kDa in the conditioned medium and the cytosolic fraction of both FSE and HUVE, representing the latent form of gelatinase A (Fig. 2). Two additional bands at 62 and 59 kDa, being the active species of gelatinase A, were also present. The treatment of both FSE and HUVE with PMA for 24 h resulted in the further activation of gelatinase A. The total amount of latent and active forms of gelatinase A were increased in the membrane fraction by PMA (Fig. 2).
We next examined the time-course expression of gelatinases by FSE after 2, 8, and 16 h of PMA treatment (Fig. 4). In the conditioned medium, latent gelatinase B only became detectable after 16 h of treatment with PMA. In the cytosol, latent Confluent endothelial monolayers were preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h followed by incubation for 24 h in fresh basal medium in the presence of 100 ng/ml PMA or no test agent (Basal). The conditioned medium (Medium) was collected and the cytosolic fraction (Cytosol) and membrane fraction (Membrane) were extracted using Triton X-114 as described under "Experimental Procedures." Zymography was used to assess gelatinase activity in the conditioned medium and the cytosolic and membrane fractions of FSE and HUVE. These experiments were performed on three different FSE and HUVE cell lines, each with similar results. gelatinase B was first detected after 2 h, whereas the active forms of gelatinase B were detected after 8 h of PMA treatment, the levels of which were substantially higher at 16 h. Gelatinase A first appeared in the cytosol after 2 h but, in contrast to gelatinase B, the levels remained constant over the 16-h incubation period. This suggested that gelatinase A was secreted from the cell, whereas gelatinase B was stored in the cell.
Western blotting disclosed the presence of TIMP1 in the cytosol, which was markedly elevated in FSE but not HUVE after treatment with PMA (Fig. 5). Stromelysin-1 was also detected in the FSE cytosol under basal conditions, the levels of which were increased after treatment with PMA (data not shown).
Cytosolic Gelatinase B Is TIMP-free, whereas Secreted Gelatinase B Is Bound to TIMP1-Previous workers have reported that the aminophenylmercuric acetate (APMA) activation of the purified gelatinase B-TIMP complex results in the loss of an 8 -10-kDa NH 2 -terminal domain of the enzyme, generating a final product of 82 kDa (24,25). In contrast, activation of TIMP-free gelatinase B with APMA results in the formation of not only the intermediate 82-kDa species, but also the active species of 67 kDa and 40 -50 kDa (24 -26). To determine whether gelatinase B in the cytosolic fraction was complexed with TIMP1, the conditioned medium and the cytosolic fraction of PMA-stimulated FSE were incubated with 2 mm APMA at 37°C. Results are shown in Fig. 6. Incubation of the conditioned medium with APMA for 1 h resulted in the partial conversion of the latent gelatinase B to an 82-kDa species. No further activation occurred after 18 h. Treatment of the cytosolic fraction with APMA produced 82-, 74-, and 67-kDa species and diffusely resolved bands at 40 -50 kDa similar to those described by Senior et al. (26). Western analysis confirmed these results (Fig. 6b). These observations indicate that gelatinase B present in the cytosolic fraction is free of TIMP1, whereas the latent species present in the conditioned medium is complexed with TIMP1.

Gelatinase B Is Localized to Intracellular Vesicles-Because
both gelatinase B and TIMP1 were present in the cytosol yet were not complexed, we used immunogold electron microscopy to investigate whether TIMP1 and gelatinase B were separately compartmentalized within the cell. FSE were grown on type I collagen gel to provide support for maintaining cell integrity during processing. Immunogold electron microscopy confirmed the presence of gelatinase B and TIMP1 in the cytosol of FSE which had been treated with PMA for 24 h. Gelatinase B gold colloid particles were mainly localized in membrane-bound secretory vesicles (Fig. 7a). These vesicles were slightly more electron dense than the surrounding cytoplasm and were encapsulated by a clearly defined plasma membrane. In quiescent cells, the vesicles were usually found in close proximity to the cell membrane facing the collagen gel. Interestingly, the vesicles were more abundant in pseudopod extensions from the cells and, in some instances, the immunogold labeling showed gelatinase B being secreted from the tips of pseudopodia which were invading the collagen gel (Fig. 7c). Using double immunogold labeling, we found that TIMP1 was spread more diffusely throughout the cytoplasm (Fig. 7d). Although TIMP was occasionally present in small vesicles (Fig.  7d), it was never found in the gelatinase B-containing vesicles. DISCUSSION We have demonstrated that gelatinase B accumulates in human microvascular endothelial cells and that the enzyme is present in both its latent and active forms. Previous to this report, only stromelysin-3 and MT1-MMP were known to be FIG. 3. Effect of Triton X-114 extraction on the activation of gelatinase B. Confluent FSE were preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h followed by incubation for 24 h in fresh basal medium in the presence of 100 ng/ml PMA. Conditioned medium was then incubated with or without 1.5% Triton X-114 in Tris-buffered saline for 16 h at 37°C. Control represents conditioned media treated with Triton but not incubated. The samples were then assessed for gelatinase activity by zymography.

FIG. 4. Time course of gelatinase B expression by FSE.
Confluent FSE were preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h followed by incubation for 24 h in fresh basal medium in the presence of 100 ng/ml PMA for 2, 8, or 16 h. The conditioned medium (Medium) and cytosolic fraction (Cytosol) were assessed for gelatinase activity by zymography.
FIG. 5. Western blot analysis of TIMP1. Confluent endothelial monolayers were preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h followed by incubation for 24 h in fresh basal medium in the presence of 100 ng/ml PMA or no test agent (Basal). TIMP1 antigen was measured in the cytosolic fraction of FSE and HUVE using an antibody to TIMP1.

FIG. 6. Activation and processing of gelatinase B by APMA.
Confluent FSE were preincubated in basal medium (Biorich plus 1% normal pooled serum) for 6 h followed by incubation for 24 h in fresh basal medium in the presence of 100 ng/ml PMA or no test agent (Basal). The conditioned medium (Medium) or cytosolic fraction (Cytosol) of PMA-stimulated FSE were incubated with 2 mM APMA for 0, 1, or 18 h at 37°C and then analyzed for gelatinase B by zymography (a) and Western blotting (b) using an antibody to gelatinase B. processed to their enzymatically active forms intracellularly (27,28). The activation of stromelysin-3 and MT1-MMP is regulated by the presence of an unusual 10-amino acid insert sandwiched between the propeptide and the catalytic domains of the enzyme, containing an amino acid sequence Arg-Arg-Lys-Arg which is recognized and activated by the Golgi-associated proteinase furin. However, it is not clear how the activation of gelatinase B, which does not have this amino acid insert, occurs intracellularly. Recently, Baramova et al. (29) have shown that the plasminogen activator-plasmin system may be involved in gelatinase B activation. Free gelatinase B, but not gelatinase B bound to TIMP, can also be activated by active stromelysin-1 (6,14,15) and by the 62-and 45-kDa active forms of gelatinase A (30). Stromelysin-activated TIMP-free gelatinase B generates the intermediate 82-kDa form, the active 67-kDa species, and the inactive 50-kDa species (6,24), whereas gelatinase A generates an 82-kDa active species as the final product. The active species of gelatinase B present in the cytosolic fraction of our study are similar to those generated by stromelysin, suggesting that this may be the activating agent. Further evidence for this was provided by our finding that stromelysin-1 was also present in the cytosol of FSE. Whether stromelysin-1 is responsible for the activation of intracellular gelatinase B needs further investigation.
Gelatinase B is unique among the MMPs in that it strongly interacts with TIMP1 in its latent form via the C-terminal domain (6). Murphy et al. (3) have proposed that gelatinase B-TIMP1 complexes are formed during intracellular protein folding. It is generally believed that the gelatinase B-TIMP1 complexes are then secreted from cells (24). Our study has verified that gelatinase B secreted into the conditioned medium of endothelial cells is bound to TIMP1; however, the gelatinase B expressed in the cytosolic fraction of FSE can exist uncomplexed. Thus, the formation of the gelatinase B-TIMP1 complex occurs after (or while) these molecules are secreted from microvascular endothelial cells. Further evidence for this was obtained using immunogold electron microscopy, which revealed that gelatinase B was separately compartmentalized from TIMP1 within the endothelial cell by being localized in secretory vesicles (Fig. 7).
The presence of gelatinase A on the cell membrane of human umbilical vein endothelial cells has been previously reported (19). Our experiments confirmed these findings and also found that both gelatinase A and gelatinase B were expressed in the membrane fraction of FSE in response to PMA. The association of gelatinase B with the plasma membrane has been previously shown in bone metastatic tissue (31) and in basal cell carcinomas (32), although the binding mechanism is not clear. In contrast, the binding and activation of gelatinase A to the cell membrane via TIMP2 and MT1-MMP are well documented (12,33). Brooks et al. (34) have also found that the C terminus of active gelatinase A is associated with the ␣ v ␤ 3 integrin receptor on the cell surface of angiogenic blood vessels and melanoma tumors. Similar interactions may exist for the binding of gelatinase B to the membrane.
In this report, we have demonstrated that FSE have the ability to accumulate and activate gelatinase B intracellularly, especially in the presence of the tumor-promoting chemical PMA. We and others (4,5) have previously shown that PMA induces the synthesis and secretion of gelatinase B by FSE. To date, there have been few reports on the effects of cytokines or angiogenic/growth factors on gelatinase B induction by human microvascular endothelial cells. Hanemaaijer et al. (4) have shown that tumor necrosis factor-␣ can enhance the effect of PMA, but it does not stimulate gelatinase B synthesis by FSE when used alone (4,5). Future studies need to investigate the interplay of physiological agents that can induce gelatinase B accumulation in microvascular endothelial cells.
We have demonstrated that gelatinase B can exist as a free active enzyme in the cytoplasm and to a lesser extent on the cell membrane of FSE. These findings are likely to be relevant in angiogenesis, a phenomenon which only occurs in microvascular endothelial cells. It is feasible that as endothelial cells migrate during angiogenesis, the active forms of gelatinase B are secreted from the cell in short bursts to locally degrade the basement membrane. They are then rapidly inhibited by TIMP and/or degraded in the extracellular milieu (3), rendering them inactive. This inhibitory mechanism is important, as it prevents uncontrolled proteolysis. Pepper et al. (35) have shown that if proteolysis goes uninterrupted the dissolution of the FIG. 7. Immunogold labeling of gelatinase B. Electron micrographs of immunogold labeling for intracellular gelatinase B and TIMP1 in FSE grown on a type I collagen gel (Coll) in the presence of PMA (100 ng/ml) for 24 h. Protein A20-nm colloidal gold particles were used to localize gelatinase B. a, gelatinase B (G) localized in membrane-bound vesicles (ϫ 60,000). b, negative control. When the antibody to gelatinase B was replaced by a nonimmune IgG there was minimal nonspecific binding of gold particles (ϫ 25,000). c, a pseudopod which was invading the collagen gel contained many gelatinase vesicles. One vesicle (S) at the tip of the pseudopod is releasing gelatinase B into the matrix (ϫ 96,000). d, double labeling. The protein A20-nm and protein A10-nm colloidal gold particles were used to localize gelatinase B (G) and TIMP1 (T), respectively. TIMP1 was located throughout the intracellular matrix, occasionally in vesicles (Tv), but was not present within the gelatinase B-containing vesicles (ϫ 58,000). matrix prevents endothelial cells from migrating and forming tube-like structures due to the absence of a scaffold. The ability of microvascular endothelial cells to accumulate active gelatinase B in secretory vesicles, ready for release, in addition to the presence of active species of gelatinase A and gelatinase B on the cell membrane, would enable microvascular endothelial cells to focalize proteolytic activity to the pericellular environment and thus be very effective in the process of cell migration.