Identification of the 170-kDa melanoma membrane-bound gelatinase (seprase) as a serine integral membrane protease.

The 170-kDa membrane-bound gelatinase, seprase, is a cell surface protease, the expression of which correlates with the invasive phenotype of human melanoma and carcinoma cells. We have isolated seprase from cell membranes and shed vesicles of LOX human melanoma cells. The active enzyme is a dimer of N-glycosylated 97-kDa subunits. Sequence analysis of three internal proteolytic fragments of the 97-kDa polypeptide revealed up to 87.5% identity to the 95-kDa fibroblast activation protein α (FAPα), the function of which is unknown. Thus, we used reverse transcription-polymerase chain reaction to generate a 2.4-kilobase cDNA from LOX mRNA with FAPα primers. COS-7 cells transfected with this cDNA expressed a 170-kDa gelatinase that is recognized by monoclonal antibodies directed against seprase. Sequence analysis also showed similarities to the 110-kDa subunit of dipeptidyl peptidase IV (DPPIV). Like DPPIV, the gelatinase activity of seprase was completely blocked by serine-protease inhibitors, including diisopropyl fluorophosphate. Seprase could be affinity-labeled by [3H]diisopropyl fluorophosphate, but the proteolytically inactive 97-kDa subunit could not, confirming the existence of a serine protease active site on the dimeric form. Proteolytic activity is lost upon dissociation into its 97-kDa subunit following treatment with acid, heat, or cysteine and histidine-modifying agents. We conclude that seprase, FAPα, and DPPIV are related serine integral membrane proteases and that seprase is similar to DPPIV, the proteolytic activities of which are dependent upon subunit association.

Seprase was originally identified from a human malignant melanoma cell line LOX, which exhibited aggressive behavior in experimental metastasis (1,2). It is a membrane glycoprotein with gelatinase activity that is expressed and localized at the invasion front during invasion into the ECM 1 by human melanoma, breast carcinoma cells, and chicken embryo fibroblasts transformed by Rous sarcoma virus (3)(4)(5)(6). The protease is a hydrophobic glycoprotein soluble in Triton X-100 and in SDS (3). Seprase shows gelatinolytic activity as demonstrated by gelatin zymography (3,4). The enzyme maintains maximal activity at neutral pH, which can be further enhanced by SDS buffer, EDTA, the cysteine protease inhibitor trans-epoxysuccinyl-L-leucylamido(4-guanidino)-butane and dithiothreitol. However, seprase has a peculiar protease inhibitor profile; it is inhibited by the protease inhibitors PMSF and NEM (3).
A major problem in purifying seprase was the apparent low level of this protein in cultured cell lines. Because LOX cells produced a large quantity of membrane vesicles that were enriched in seprase, we isolated seprase from the cell membranes and shed vesicles in LOX conditioned medium and characterized its subunit composition and its enzymatic activity. We found that the protease had an apparent molecular mass of 170 kDa composed of proteolytically inactive 97-kDa subunits. Amino acid analysis of internal proteolytic fragments of the 97-kDa polypeptide revealed similarity to the 95-kDa FAP␣ (7) and the 110-kDa subunit of human DPPIV (8 -10). RT-PCR analysis suggests an identical size for the mRNA encoding the 97-kDa seprase subunit from LOX cells as well as the 95-kDa FAP␣ from WI-38 fibroblasts. When a 2.4-kilobase cDNA amplicon generated by RT-PCR from LOX mRNA using FAP␣ primers was expressed in COS-7 cells, it produced a 170-kDa gelatinase that was recognized by mAbs directed against seprase. In addition, this report demonstrates that like DPPIV (11,12), seprase requires dimerization for its gelatinase activity. This gelatinase activity was completely blocked by the serine-protease inhibitors, DFP, PMSF, AEBSF, and APSF. Dimeric seprase could be affinity-labeled by [ 3 H]DFP, but the proteolytically inactive 97-kDa subunit could not. These data demonstrate structural regulation of protease activity by the formation of a serine protease active site upon association of its subunits. Based on its protease characteristics and on the deduced amino acid sequence (GenBank accession number U76833) from its cDNA, we suggest that seprase and DPPIV represent novel serine integral membrane proteases, the proteolytic activity of which is regulated by subunit association.

EXPERIMENTAL PROCEDURES
Materials-The human amelanotic melanoma cell line LOX was obtained from Professor Oystein Fodstad, Institute for Cancer Research, The Norwegian Radium Hospital, Oslo, Norway. The human embryonic lung fibroblast line WI-38 and African Green monkey kidney fibroblast line COS-7 were obtained from the American Type Culture Collection (Rockville, MD). Cell culture materials and most protease inhibitors were obtained as described (3). [ 3 H]DFP was purchased from DuPont NEN (Boston, MA). 35 S-labeled Tran label, DFP, and AEBSF were obtained from ICN Biomedical, and APSF was obtained from Dr. Jörg Sturzebecher (Erfurt, Germany). Three hybridoma cell lines, D8, D28, and D43, which secreted monoclonal IgG (all class 2a) antibodies, were derived from the fusion of rat myoloma Y3 cells and the spleen cells of Sprague-Dawley rats that were immunized for four 2-week intervals with 50 g of partially purified seprase derived from human placenta.
RT-PCR, DNA Cloning, and Expression in COS-7 Cells-Total RNA was isolated from LOX and WI-38 cells using the RNA Stat-60 kit (Tel-Test "B", Inc.). Reverse transcription was carried out with 6 g of total RNA using oligo(dT) [12][13][14][15][16][17][18] as the primer. The reaction was catalyzed by Superscript II RNase H Ϫ reverse transcriptase, as directed by the manufacturer (Life Technologies, Inc.). Two oligonucleotide primers were synthesized which correspond to a sense sequence within the 5Ј untranslated region (FAP 1) and an antisense sequence within the 3Ј untranslated region (FAP 6) of the published FAP␣ cDNA sequence (7). A ϳ2.4-kilobase amplicon generated with either LOX RNA or WI-38 RNA and the FAP 1 (5Ј-CCACGCTCTGAAGACAGAATT-3Ј (no. 161-181)) and FAP 6 (5Ј-TCAGATTCTGATACAGGCT-3Ј (no. 2523-2505)) primers using the Expand Long Template PCR System (Boehringer Mannheim) and a Perkin Elmer GeneAmp 9600 cycler was isolated from a 1% agarose gel using a Qiagen gel extraction kit. Purified cDNA (40 ng) was ligated to 30 ng of the pCR3.1 mammalian expression vector. Ligation, transformation, and selection of recombinant clones were carried out using the Eukaryotic TA cloning kit (Invitrogen). A recombinant plasmid, clone pA15, was purified using the Qiagen Plasmid Maxi kit. Transfection of COS-7 cells was carried out by electroporation using a Bio-Rad Gene Pulser II system (Conditions: 0.3 kV and 950 F). Each electro-transformation was carried out on ϳ5 ϫ 10 6 cells in a volume of 0.5 ml using 20 g of plasmid. Transfection efficiency was ϳ20%, as determined by immunofluorescence. To enrich for cells that express seprase, transfected cells were immunoselected using a panning procedure (13) with the mAbs D8 and D28.
Immunofluorescence Staining and Confocal Microscopy-LOX and transfected COS-7 cells were fixed with 3% paraformaldehyde in PBS for 15 min and stained with mAbs D8 or D28 for confocal microscopic analysis as described previously (4).
Immunoprecipitation of 170-kDa Gelatinase and Its 97-kDa Subunit-Protein A-Sepharose beads (15 l; Pierce) were coated with rabbit anti-rat IgG (24 g; Amersham Corp.), and rat mAbs D43 or D28 (25 g each) prior to incubation with cell detergent extract. Alternatively, mAbs D8, D28, and D43 were coupled to CNBr-Sepharose 4B (Pharmacia Biotech Inc.) according to the manufacturer's instructions. Protease preparations were incubated with the antibody-coated beads for 2 h at 4°C. The antigen-antibody complex-coated beads were collected by centrifugation at low speed (1300 ϫ g), supernatants were removed, and the beads were washed four times with 1% Triton X-100 and 1 mM EDTA in TBS, pH 8.0, at 4°C (buffer A). The beads were then washed with 1.0% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS in TBS, pH 8.0 (RIPA buffer). Bound proteins were eluted with 2 ϫ SDS sample buffer containing 0.1 M dithiothreitol at 37°C or 80°C for 10 min, and the SDS-solubilized samples were analyzed by gelatin zymography, SDS-PAGE autoradiography, or immunoblotting. For immunoprecipitations requiring metabolic labeling, cells were incubated overnight with 0.025 mCi/ml 35 S-labeled Tran label (ICN Biomedical) in methionine-free Dulbecco's modified Eagle's medium, 10% fetal bovine serum, 1% complete Dulbecco's modified Eagle's medium, 2 mM glutamine, and 1% penicillin/streptomycin. For assaying the gelatin-degrading activity of seprase, heat-denatured rat tail type I collagen (gelatin) was incubated at 37°C for 45 h in TBS, in the presence of seprase immobilized on mAbs D8, D28, and D43 coupled to Sepharose beads. Following digestion, the supernatants were collected, and the digested gelatin was resolved by SDS-PAGE (7.5% acrylamide gel). The gels were stained by Coomassie Brilliant Blue.
Immunoaffinity Purification-We isolated seprase from 100 liters of LOX cell conditioned media and 15 ml of LOX cell pellet using WGA and organomercurial chromatography as described (3). Alternatively, seprase was enriched by a 40% ammonium sulfate cut of LOX cell RIPA lysates. Protein suspensions were spun at 10,000 ϫ g for 30 min, and the pellet was resuspended in 1/5 volumes with TBS containing 1% octylglucoside. Resuspended proteins were dialyzed against 2 ϫ 500 ml of 1% octylglucoside/TBS. The ammonium sulfate concentrated cell lysate was cleared by centrifugation at 10,000 ϫ g for 30 min at 4°C. The cleared lysate was loaded onto a BSA precolumn and then onto a mAb D28 affinity column. Column chromatography was carried out at 4°C. The D28 mAb column was washed with five bed volumes of buffer A and five bed volumes of RIPA buffer. Seprase was eluted with 50 mM glycine buffer, pH 2.4, containing 1% octylglucoside, and fractions (1.5 ml) were immediately neutralized with 1.5 M Tris buffer, pH 8.8. Fractions were analyzed by SDS-PAGE, and protein bands were visualized using silver staining.
Partial Amino Acid Sequence Determination-Amino acid sequence analysis was performed as described by Matsudaira (14). The immunoaffinity purified 97-kDa subunit was reduced by 20 mM dithiothreitol and then alkylated by 200 mM NEM. About 10 g of the alkylated subunit was resolved on SDS-PAGE and then transferred to a polyvinylidene difluoride membrane (Boehringer Mannheim; sequencing grade) in the presence of 1% hydrogenated Triton X-100 (Sigma)/10% acetonitrile/100 mM Tris-HCl, pH 8.0, for 24 h at 37°C. Protein bands migrating at 97 kDa were stained with Ponceau S and excised from the membrane, and Lys-C digestion was performed as described (15). Following digestion, samples were centrifuged, and the supernatant was transferred to an HPLC injection vial (Applied Bio Systems). Peptide isolation (200 l) was performed by HPLC (Applied Bio Systems model 130) using a Vydac C 18 column (2.1 ϫ 250 mm) with a flow rate of 150 l/min at room temperature. HPLC buffers were: buffer A, 0.1% trifluoroacetic acid/Milli-Q water; and buffer B, 0.08% trifluoroacetic acid/ acetonitrile. The gradient was 1.6 -29.5% B (0 -63 min), 29.6 -60% B (63-95 min), and 60 -80% B (95-105 min). The column was then washed with 80% buffer B for 12 min at 150 l/min and re-equilibrated with 1.6% buffer B for 50 min at 300 l/min. Peptide elution was monitored at 220 nm. Individual polypeptide peaks on HPLC were collected and subjected to Edman degradation using the Hewlett Packard Sequencer (model G1000A) with an on-line PTH analyzer (Hewlett-Packard model 1090).
Isoelectric Focusing and Superose 12 Gel Filtration Liquid Chromatography-Seprase, isolated through phase partitioning and WGA-affinity chromatography from LOX cell Triton X-114 extracts, was further purified by isoelectric focusing (Rotofor cell; Bio-Rad). Ampholite solution (2%) in PBS containing 10% glycerol and 1% Triton X-100 was prefocused for 1 h. WGA-purified proteins were applied at the neutral pH range, and the Rotofor was run for another 4 h. Samples were harvested and analyzed by gelatin zymography and immunoblotting. Fractions, at pI 5, containing seprase were pooled and stored at Ϫ80°C. An aliquot was applied to a gel filtration column (Superose 12; Pharmacia Biotech Inc.) previously equilibrated with 10% glycerol and 1% Triton X-100 in PBS. The sample was loaded and eluted at a flow rate of 0.05 ml/min and 0.1 ml/min, respectively. The void volume was collected in 0.5-ml fractions, and the remaining samples were collected in 0.2-ml fractions.
Serine Protease Inhibitor-Affinity Labeling-The 170-kDa form of seprase isolated by isoelectric focusing, the 97-kDa subunit obtained from mAb D28-affinity chromatography, or seprase from LOX cell lysate immobilized on mAbs D8, D28, and D43 beads was incubated in the presence of 33 M [ 3 H]DFP (10 Ci, 6 Ci/mmol; DuPont NEN) in 0.1% Tween 20/TBS (TBS-T), pH 7.5, for 1 h at room temperature. The free [ 3 H]DFP was washed out with TBS by consecutive rounds of dilution/ concentration using a Centricon concentrator (Amicon Inc.), and the [ 3 H]DFP-labeled seprase was resolved by SDS-PAGE under nonboiling/ nonreducing conditions. [ 3 H]Seprase immobilized on antibody beads was washed with TBS-T and eluted with SDS by reducing and boiling. The eluate was then subjected to SDS-PAGE. Proteins were fixed with 40% methanol/10% acetic acid for 30 min. The gel was incubated with Amplify solution (Amersham Corp.) for 20 min at room temperature and dried. Autoradiograms were exposed for 3, 10, or 20 days at Ϫ80°C using Hyperfilm (Amersham Corp.).
DPPIV Substrate Membrane Overlay Assay-Proteases were electrophoresed on SDS gels and analyzed on a substrate overlay membrane (Enzyme Systems Products) coupled with the fluorescent substrate Ala-Pro-7-amino-4-trifluoromethyl coumarin (16). The membrane was moistened in 0.5 M Tris-HCl, pH 7.8, placed against the gel, and incubated at 37°C in a humidified chamber. The membrane was then removed from the gel and air dried. The DPPIV activity of individual proteases was monitored by detecting 7-amino-4-trifluoromethyl coumarin released from the substrate using a long wave length ultraviolet lamp.
Gelatin Zymography and Immunoblotting of Seprase in the Presence of Protease Inhibitors, Buffers at Different pH, and Different Temperatures-Seprase isolated from LOX lysates was treated with inhibitors for various classes of proteases, exposed to different pH buffers, or incubated at temperatures of 40°C, 50°C, or 60°C. To prepare the protease for exposure to different buffers, seprase in 1% Triton X-114 was partitioned three times using saline to replace the Tris buffer. The resulting detergent phase was incubated with buffers at pH 4 (0.06 M citric acid/0.08 M sodium phosphate), pH 5 (0.05 M citric acid/0.1 M sodium phosphate), and pH 6 (0.04 M citric acid/0.12 M sodium phosphate) on ice for 10 min. Treated protease preparations were subjected to gelatin zymography and immunoblotting using mAbs D8 and D43.
Other Methods-Protein determinations were made using the bicinchoninic acid assay (Pierce) with bovine serum albumin as the standard protein. The absorbance was measured at 562 nm using an LKB Ultrospec 4050 (Pharmacia Biotech Inc.) or at dual absorbance of 540 and 490 nm using a Bio-Rad microtiter plate reader. For immunoblotting, proteins were separated by SDS-PAGE with 7.5 or 10% polyacrylamide gels, transferred to nitrocellulose sheets (Schleicher & Schuell) by electroblotting, and subsequently probed with mAbs. Blots were incubated with hybridoma supernatants diluted 1:30 in TBS-T for 1 h at room temperature. Immunoreactive polypeptides were visualized using the enhanced chemiluminescence detection system (Amersham Corp.) as described previously (17).

RESULTS
Localization and Isolation of Seprase and Its 97-kDa Subunit-Seprase was detectable by immunofluorescence with mAbs D8, D28, and D43 on cell surface extensions and membrane vesicles of LOX cells (Fig. 1, A-E). Confocal microscopic analysis of cells cultured on glass coverslips and stained with mAb D28 shows intense seprase localization on membrane extensions at the leading edge, lamellipodia, of the cell (Fig. 1, A-C). When cells were cultured on fibronectin-coated crosslinked gelatin films and stained with mAb D28, seprase became concentrated at invadopodia, specialized protrusions of the ventral membrane, that contacted the film (Fig. 1D). In addition, seprase could be observed on shed vesicles and on surface extensions in the cell-gelatin film interface (Fig. 1E). These results support the observation that seprase can be localized to the cell surface lamellipodia, invadopodia, and on shed vesicles.
Immunoblotting analysis showed that mAbs D8 and D28 recognized seprase (170 kDa) and its 97-kDa subunit, whereas mAb D43 labeled only dimeric seprase (Fig. 1F). Immunoprecipitation of metabolically labeled cells resulted in elution of a major 97-kDa protein band following washing with both 1% Triton X-100/TBS and RIPA buffer (Fig. 2A, lane 1), but multiple components were eluted when only 1% Triton X-100/TBS was used for washing ( Fig. 2A, lane 2). Monoclonal antibody D28 was used for affinity purification of seprase from LOX cell membranes and shed vesicles. Fractions (1-3) were analyzed by silver staining (Fig. 2B, lanes 1-3) and immunoblotting with mAb D8 (Fig. 2B, lanes 4 -6). The conditions used to elute seprase from the mAb column (glycine buffer, pH 2.4) resulted in elution of the proteolytically inactive 97-kDa subunit (Fig.  2B). Analysis of seprase to determine the extent of its glycosylation indicated that the 97-kDa subunit undergoes a reduction of apparent molecular mass of approximately 20 kDa after exposure to N-glycosidase F (Fig. 2C, lanes 1 and 2). D8 (data not shown) and D28 mAbs recognized the deglycosylated 97-kDa subunit (Fig. 2C, lanes 3 and 4), indicating that their epitopes did not include the N-linked side chains. These results show that seprase is composed of monomeric, N-glycosylated 97-kDa subunits.
Partial Amino Acid Sequence Analysis of the 97-kDa Protease Subunit-Initial attempts to obtain protein sequence data from the 97-kDa subunit were not successful, suggesting that its N terminus may be blocked. Three internal proteolytic fragments were generated from the 97-kDa subunit by Lys-C digestion, purified by HPLC, and subjected to microsequencing. Comparison of the HPLC patterns from the proteolytic digest of the 110-kDa subunit of human placental DPPIV and the 97-kDa subunit of seprase revealed differences (data not shown), suggesting that seprase was distinct from DPPIV. Sequence analysis of the three peptides, which were 10, 12, and 8 amino acids in length, exhibited 80, 66.7, and 87.5% identities, respectively, with the corresponding deduced sequences of FAP␣. In addition, the 10-and 12-mers exhibited 70 and 33.3% identities with that of DPPIV, respectively (Fig. 3). These results suggest that the 97-kDa subunit of seprase may be highly homologous to the 95-kDa FAP␣ and related to the 110-kDa subunit of DPPIV.
Cloning and Expression of Seprase-We carried out RT-PCR

FIG. 1. Localization of seprase on membrane protrusions of LOX human melanoma cells and their shed vesicles.
A-C, confocal microscopic analysis of seprase distribution in a cell cultured on a glass coverslip that was stained with mAb D28. Images were taken at 4.5, 6.0, and 7.5 m, respectively. Intense staining can be observed at the leading edge, lamellipodia, of the cells. D, seprase distribution in a cell cultured on a fibronectin-coated cross-linked gelatin film that was stained with mAb D28. This image was taken at the level of gelatin film to show the seprase concentration at sites of invasion, invadopodia. E, seprase distribution in a cell cultured on a fibronectin-coated crosslinked gelatin film that was stained with mAb D8. This image was taken at the cell-gelatin film interface to show the seprase concentration on shed vesicles and surface extensions. Bar, 25 m. F, immunoblotting analysis of dimeric seprase (170 kDa) and the 97-kDa subunit derived from LOX cell membranes that were treated with SDS buffer at 37°C for 10 min (lanes 1, 3, and 6) or SDS buffer containing 10 mM dithiothreitol at 80°C for 10 min (lanes 2, 4, and 6).  1 and 2) antigens detected by immunoprecipitation of 35 S-labeled LOX cells. Lane 1, a major 97-kDa protein was isolated when bound antigen was washed with 1% Triton X-100/ TBS followed by a wash with RIPA buffer, then eluted by boiling. Lane 2, multiple components were isolated when only 1% Triton X-100/TBS was used for washing. B, analysis of fractions 1-3 of mAb D28 affinitypurified 97-kDa subunit from LOX cells. Protein bands were visualized by silver staining (lanes 1-3) and immunoblotting with mAb D8 (lanes 4 -6). C, N-glycosidase F digestion of the 97-kDa subunit of seprase. Lanes 1 and 2, silver-stained gels of the protein treated without (Ϫ) or with (ϩ) N-glycosidase F. Lanes 3 and 4, immunoblotting with mAb D28 of the protein treated without (Ϫ) or with (ϩ) N-glycosidase F. of LOX total RNA using oligonucleotide primers that corresponded to the 5Ј untranslated region (FAP 1) and the 3Ј untranslated region (FAP 6) of the FAP␣ cDNA (7). The resultant ϳ2.4-kilobase amplicon was subcloned into the pCR3.1 mammalian expression vector (clone pA15). To confirm that this amplicon encoded the 97-kDa subunit of seprase, we transfected COS-7 cells with pA15 or the vector alone (pA11). Two mAbs, D8 and D28, which had been shown previously (4) to recognize epitopes on both seprase and its monomeric 97-kDa subunit, were used for the detection of surface expression, Western blot analysis, and gelatinolytic activity upshift. As can be seen in Fig. 4, mAb D8 specifically stained cells that had been transfected with pA15 (Fig. 4A), whereas only background staining was observed for those cells transfected with vector alone (Fig. 4B). In a functional assay, detergent extracts of immunoselected pA15-and pA11 (vector alone)-transfected COS-7 cells and LOX cells were assayed for proteolytic activity by gelatin zymography. The pA15-transfected cells gave rise to a gelatinolytic band at ϳ170 kDa (Fig. 4C, lane 3) that corresponds with the region of lysis produced by the LOX cell detergent extract (Fig. 4C, lane 4) in contrast to the vector-transfected COS-7 cell extracts, which show no specific band of lysis (Fig. 4C, lanes 1 and 2). In addition, the identity of the gelatinolytic activity was confirmed by the formation of seprase-D8 (Fig. 4C, lanes 6 and 8) and seprase-D28 (Fig. 4C, lanes 7 and  9) complexes that upshifted the proteolytic activity. The immunoblot shows that pA15-transfected cells (Fig. 4D, lane 3) express the 97-kDa monomeric subunit of seprase that is also present in the control LOX cell extract (Fig. 4D, lane 2) but not in the mock-transfected cells (Fig. 4D, lane 1). Furthermore, RT-PCR with the FAP␣ primers, FAP 1 and FAP 6, generated identical amplicons from LOX and WI-38 fibroblast mRNA (data not shown). These results support our initial observation that the 97-kDa subunit of seprase is related to the 95-kDa FAP␣.
Substrate specificity studies using zymography have shown that seprase degrades gelatin but not laminin, fibronectin, fibrin, or casein (3). Thus, we examined possible ECM substrates, including type I collagen, type IV collagen, laminin, and fibronectin, using a soluble proteolytic assay. Seprase was immobilized on mAbs D8-, D28-, and D43-Sepharose beads. Native type I and type IV collagens, when incubated with the immobilized protease, were apparently not digested (data not shown), but thermally denatured type I (Fig. 5B, lanes 2-4) and type IV collagens (data not shown) incubated at 37°C were digested into smaller polypeptides. Seprase did not degrade fibronectin and laminin under these conditions (data not shown). Also, based on its protein sequence similarity with DPPIV, we investigated whether seprase possessed DPPIV proteolytic activity using DPPIV derived from placental tissue as a positive control (Fig. 5C, lane 2). We did not detect any DPPIV activity for seprase (Fig. 5C, lane 1).
Dimerization Is an Essential Structural Requirement for Generation of the Seprase Serine Protease Active Site(s)-The effects of acid, heat, and amino acid-modifying agents, such as NEM and DEPC, in inhibiting seprase gelatinase activity have been investigated. The sensitivity of the seprase dimeric structure and its gelatinase activity to buffers at different pHs (Fig.  7A), incubation at different temperatures (Fig. 7B), and treatment with the histidine-modifying agent DEPC (Fig. 7C) was investigated using gelatin zymography and immunoblotting. Gelatinase activity was measured by zymography, and changes in subunit association were detected by immunoblotting with mAb D8 that recognizes both seprase and its 97-kDa subunit and with mAb D43 that recognizes only seprase. At pH lower than 5 (Fig. 7A), incubation at temperatures above 50°C (Fig.  7B), or DEPC treatment at concentrations over 5 mM (Fig. 7C),   1 and 3-7), but the 97-kDa monomer was not (lane 2). The autoradiograms were revealed after exposure of films for 20 days (lanes 1 and 2), 3 days (lanes 4 -6), and 10 days (lanes 3, 7, and 8).
loss of gelatinase activity was always accompanied by the disappearance of the 170-kDa form and the appearance of the 97-kDa subunit. A similar effect to DEPC was observed for the cysteine-modifying agent NEM but not for the serine protease inhibitors PMSF, AEBSF, APSF, and DFP at the concentrations that inhibited the gelatinase activity (data not shown). Thus, acidic pH, heat, NEM, and DEPC treatment inhibit the seprase gelatinase by altering its subunit association, which is necessary for its catalytic activity. DISCUSSION This study focused on the characterization of melanoma seprase in terms of its amino acid sequence, expression of its cDNA, classification of its proteolytic activity, and structure/ activity relationships. Seprase is localized at cell surface invadopodia, lamellipodia, and vesicles. These results, together with the previous observation that LOX cells with higher levels of seprase display a more invasive phenotype than those with lower levels (4), make it important to identify and characterize this protease.
Previously, seprase was defined by its gelatinolytic activity rather than by its isolation (3,4). Here we have used nondenaturing chromatographic techniques followed by D28 immunoaffinity chromatography to purify a Triton X-100-soluble gelatinase from LOX cell membranes. We have also isolated seprase using isoelectric focusing and have determined that it has an isoelectric point of 5. We show here that heat-denatured type I collagen is a substrate for seprase (Fig. 5B) but not native type I collagen, supporting the notion that it is a gelati-nolytic endopeptidase. Also, we showed that seprase is a serine protease by virtue of its inhibition by the serine protease inhibitors DFP, PMSF, APSF, and AEBSF and by affinity-labeling with [ 3 H]DFP. Consistent with these results, the deduced amino acid sequence of a cDNA clone that encodes the 97-kDa subunit of seprase (GenBank accession number U76833) is highly homologous to those of the nonclassical serine protease DPPIV (8 -10) and the putative nonclassical serine protease FAP␣ (7) in their catalytic regions.
The affinity-purified material gave a single band at 97-kDa by silver staining and was recognized by mAbs D8 and D28. This 97-kDa subunit was devoid of proteolytic activity, as determined by gelatin zymography, a soluble gelatin degradation assay, and DFP affinity labeling. Lack of [ 3 H]DFP labeling of the 97-kDa form could be due to the fact that serine proteases need to be in an active conformation to accept phosphorylation of the OH group of their catalytic site serine (18). However, we do know that each 97-kDa subunit of seprase contains the consensus motif of serine proteases, GXSXG, as well as the other residues that make up the catalytic triad (GenBank accession number U76833).
Most secreted ECM-degrading enzymes are proteolytically activated from their zymogen precursors; however, formation of a dimeric structure appears to be necessary for expression of the high molecular mass dipeptidase activity of DPPIV (11,12,19). We have shown that like DPPIV (20,21), seprase required dimerization to exhibit its gelatinase activity, which was completely blocked by the serine-protease inhibitors DFP, PMSF, AEBSF, and APSF. In addition, seprase could be affinity-labeled by [ 3 H]DFP, but the proteolytically inactive 97-kDa subunit could not. These results are further confirmed by our COS-7 cell expression experiments in which the expressed 97-kDa subunit dimerized to produce the active form of seprase. These data demonstrate that there is structural regulation of the proteolytic activity of seprase; the formation of the serine protease active site(s) is dependent on the association of two subunits. However, the mechanisms by which seprase subunit association is regulated are not yet understood.
We speculate that this structural switch for proteolytic activity may be a common mechanism for the regulation of serine integral membrane proteases, including seprase and DPPIV. The requirement for a dimeric structure for seprase gelatinase activity is supported by our studies showing a correlation between structure stability and gelatinase activity at various pH, temperatures, and in the presence of the amino acid-modifying agents NEM and DEPC. We have shown that seprase was not active when samples were exposed to acidic conditions (pH Ͻ5), to high temperatures (Ͼ50°C), or to 5 mM NEM or DEPC. In all cases, protease inactivation was accompanied by protease dissociation into the 97-kDa subunits. Similarly, the dimeric 150 -220-kDa DPPIV has been reported to be active and accessible to DFP labeling, but the 110-kDa monomeric DPPIV was not (11,22). These properties of membrane-bound enzymes point to the possibility that subunits may assemble to form active enzymes at sites of protease action, i.e. the cell surface. Furthermore, the fact that seprase is sensitive to an acidic environment suggests that seprase may be inactivated in an endocytic compartment. Thus, increasing or decreasing the stability of an oligomeric complex may be a means for cells to regulate the proteolytic activity of membrane proteases.