Protein Interaction Analysis of ST14 Domains and Their Point and Deletion Mutants*

ST14 (suppression of tumorigenicity 14) is a transmembrane serine protease that contains a serine protease catalytic (SP) domain, an SEA domain, two complement subcomponent C1r/s (CUB) domains, and four low density lipoprotein receptor class A domains. Glutathione S-transferase fusion proteins with SP, CUB, and low density lipoprotein receptor domains and their corresponding mutants were generated to analyze protein interactions with these domains. Modified glutathione S-transferase pull-down assays demonstrated the interaction between the SP domain and hepatocyte growth factor activator inhibitor-1. With the same method, a CUB domain-interacting protein was isolated and turned out to be the transmembrane protein with epidermal growth factor-like and two follistatin-like domains 1 (TMEFF1). Quantitative real time PCR revealed that the expression of the TMEFF1 gene was dependent on the transfection of the ST14 gene in the RKO cell line. Our results also suggested that ST14 and TMEFF1 were co-expressed in the human breast cancer cell line MCF7, human placenta, kidney, and liver tissues. Interestingly, these two genes were co-up-regulated in kidney tumors versus normal tissues, consistent with our results that showed the dependence of TMEFF1 expression on ST14 in RKO cells. Finally, homology modeling studies suggested that TMEFF1 might form a complex with ST14 by an interaction between epidermal growth factor and CUB domains.

ST14 (suppression of tumorigenicity 14) is a multidomain transmembrane serine protease of the S1 trypsin-like protein family (1). It was first isolated as a novel matrix-degrading protease from human breast cancer cells by Shi et al. (2). Independently by using subtractive hybridization to isolate genes that are highly expressed in normal intestinal mucosa but not expressed or expressed at a lower level in colon cancers (3), we obtained a clone that is identical to ST14. We further assigned the ST14 gene to human chromosome 11q24 (4). Takeuchi et al. also cloned ST14 from a human prostate cancer cell line and designated it the membrane type serine proteinase 1 (5). ST14 was detected in colon carcinomas, immortalized breast epithelial cell lines, breast cancer cell lines, and prostate cancer cell lines but not in cultured fibroblasts or fibrosarcoma cells (3,6). The localization of ST14 to the cell membrane was demonstrated by surface biotinylation techniques (7).
ST14 is composed of an N-terminal transmembrane signal, a trypsinlike serine protease catalytic (SP) 2 domain, a SEA domain, two tandem repeats of the complement subcomponent C1r/s (CUB) domains, and four tandem repeats of the low density lipoprotein receptor (LDLR) class A domains (5,8,9). Friedrich et al. illustrated the crystal structures of the catalytic domain of ST14 and its complex with a bovine pancreatic trypsin inhibitor (8). Kunitz type serine protease inhibitors, such as hepatocyte growth factor activator inhibitor-1 (HAI-1) (5,9) and aprotinin (10), strongly inhibit ST14 through binding to its catalytic domain. Besides the catalytic domain, other regions are also required for ST14 activation (11). By mutation analysis, the CUB and LDLR domains were shown to be indispensable for the proteolytic activity of ST14 (11). In the present study, we generated GST fusion proteins that express three domains of ST14 and demonstrated the interaction between ST14 and HAI-1 using modified GST pull-down assays. Furthermore, we identified a novel interacting protein, TMEFF1, which binds the CUB domain of ST14. Quantitative real time PCR demonstrated that TMEFF1 was expressed only upon ST14 transfection into the RKO cell line. These two genes were found to be co-expressed in the human breast cancer cell line MCF7, human normal placenta, kidney tumor, normal kidney, liver tumor, normal liver, lung tumor, and breast tumor. Co-up-regulation of these two genes was observed in kidney cancers and remains an interesting phenomenon to be further explored. Finally, we demonstrated a potential complex formation between TMEFF1 and ST14 by using homology modeling approaches.

MATERIALS AND METHODS
Cell Lines and Monoclonal Antibody-Human colon carcinoma cell line RKO was transfected with ST14 in a pSecTag2A vector to establish the RKO-ST14 cell line and was transfected with the pSecTag2A backbone vector to establish the RKO-pSecTag2A cell line, respectively. The SW620-HAI-1 cell line, another human colon carcinoma cell line SW620 transfected with HAI-1, and the human monoclonal anti-HAI-1 antibody were generous gifts from Jingjia Ye (12). RKO-pSecTag2A and RKO-ST14 cells were cultured in Dulbecco's modified Eagle's medium high glucose medium supplemented with 10% fetal calf serum and 600 g/ml zeocin at 37°C in 5% CO 2 . SW620-HAI-1, the prostate cancer cell line PC3, the human breast cancer cell line MCF7, and Bcap37 were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum in the same condition.
Point Mutation and Deletion of GST-ST14 Domain Expression Vectors-A deletion mutant of pGEX-4T2-CUB that only contains the first CUB domain (residues 196 -362) was created by PCR amplification using the following primers: forward primer, 5Ј-GGATCCGTGGT-CACCTCAGTGGTGGCTTTC-3Ј; reverse primer, 5Ј-TCAGGT-GGGTAGTGGCCTGGGTAGTAG-3Ј. To make site-directed mutations in the SP and LDLR domains, overlap extension PCR was used with primers containing appropriate nucleotide changes. Table 1 lists the primers used for making the three site-specific mutations (H656A, D711A, and S805A) in pGEX-4T2-SP and the four mutations (D482Y, D519Y, D555Y, and D598Y) in pGEX-4T2-LDLR, respectively. A high fidelity Pyrobest DNA polymerase (Takara) with a proofreading capacity was used to minimize the introduction of erroneous bases in overlap extension PCR. All mutations were confirmed by DNA sequencing.
Expression and Purification of GST Fusion Proteins-All constructed vectors were transformed into Escherichia coli BL21 (DE3) cells and tested for protein expression. Glutathione-Sepharose 4B (Amersham Biosciences) was used in the purification of GST fusion proteins according to the manufacturer's protocol. Purified GST fusion proteins were verified by 10% SDS-PAGE with Coomassie Blue staining.
Determination of Enzyme Activity-GST-SP and GST-mutSP fusion proteins were tested for serine protease activity using tissue plasminogen activator chromogenic substrate (Sigma) according to the manufacturer's protocol.
GST Pull-down Assays with Beads-The bacterial supernatant containing GST fusion proteins was prepared as described (13). The final concentration of phenylmethylsulfonyl fluoride and aprotinin were 1 mM and 5 g/ml, respectively (aprotinin was not added to the lysate of GST-SP). GST fusion proteins were immobilized onto glutathione-Sepharose 4B beads and incubated with the whole cell extract of SW620-HAI-1 cells overnight at 4°C (14,15). Beads were washed three times, boiled for 5 min in 2ϫ SDS-PAGE loading buffer containing dithiothreitol, and separated by SDS-PAGE followed by Western blot analysis with monoclonal anti-HAI-1 antibody.
GST Pull-down Assays with Columns-To compare the possibility of protein interaction between the wild-type and mutant forms of GST-ST14 domains, modified GST pull-down assays were used with glutathione-Sepharose 4B columns as described (16). GST fusion proteins and their corresponding mutants were immobilized onto glutathione-Sepharose 4B columns, respectively. The whole cell lysates were loaded onto columns and incubated overnight at 4°C. The columns were then washed thoroughly with PBS to minimize the contamination. Finally, saline gradient elutions with a 5-column volume of PBS with NaCl concentrations of 200, 400, 600, 800, 1000, and 2000 mM were performed sequentially. Amicon Ultra-4 centrifugal filter units (Millipore Corp.) with a 5000-Da molecular mass cut-off were used to concentrate each elution product to 50 l. The proteins bound to GST fusion proteins or their mutants were analyzed by 10% SDS-PAGE with Coomassie Blue staining.
Peptide Mass Fingerprint Analysis-Protein bands of interest were excised from Coomassie Blue-stained gel and cut into small pieces. After being removed of SDS, gel pieces were digested by sequencing grade modified trypsin (Promega) at 37°C overnight. The peptides extracted from the gel were analyzed by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry) ProteinChip System (Ciphergen) according to the manufacturer's protocol.
Western Blot Analysis-Protein concentration was measured by the DC Protein Assay Kit (Bio-Rad). Equal amounts of proteins were resolved by 10% SDS-PAGE and transferred to polyvinylidene difluoride membrane (Bio-Rad). The membrane was first blocked in TBST (20 mM Tris-HCl, pH 7.4, 137 mM NaCl, and 0.1% Tween 20) with 5% nonfat dry milk and then probed with primary antibody (monoclonal anti-HAI-1 antibody, 1:1000 in TBST with 5% nonfat dry milk), followed by incubation with secondary antibody (horseradish peroxidase-conjugated antibodies; 1:5000 in TBST with 5% nonfat dry milk). The results were visualized by using the Lumi-Light Western blotting substrate (Roche Applied Science).

Expression and Purification of GST-ST14 Domain Fusion Proteins and Their
Mutants-GST-ST14 domain fusion proteins were designed essentially as described by Lin and co-workers (11) and are illustrated in Fig. 1. Mutations that substitute His-656, Asp-711, or Ser-805 with Ala in the SP domain would inactivate the serine protease activity (8). Deletion of the second CUB domain would result in a poor activation of ST14 (11). The point mutations that substitute an Asp with Tyr in each of the four LDLR domains would inactivate the Ca 2ϩ -binding cage in each domain (11). In the present work, SP, CUB, LDLR domains, and the deletion mutant of the CUB domain (only containing the first CUB) were cloned into the pGEX-4T2 vector. The whole cell lysate with or without expression of GST fusion proteins was analyzed by 10% SDS-PAGE and Coomassie Blue staining (Fig. 2, lane 3 and 4). We also made three point mutations (S805A, D711A, and H656A) in the GST-mutSP fusion protein and four point mutations (D598Y, D555Y, D519Y, and D482Y) in GST-mutLDLR by using overlap extension PCR (26) as illustrated in Fig. 3. GST fusion proteins were purified by using glutathione-Sepharose 4B and was again revealed by 10% SDS-PAGE (Fig. 2, lane 2). ST14 autoactivation was observed with GST-SP fusion proteins in both crude cell lysate (lane 3) and its purified form (lane 2). The upper band corresponds to the inactive GST-SP fusion proteins, and the lower band corresponds to the active GST-SP fusion proteins. The addition of protease inhibitors inhibited autoactivation through purification (Fig. 2, lane 2), whereas autoactivation was more obvious in the whole cell lysate (Fig. 2, lane 3). Similar autoactivation was reported when a Histagged fusion protein of ST14 was expressed in E. coli (5).
Enzyme Activity Assay of GST-SP Fusion Proteins and Mutants-We diluted purified GST-SP and GST-mutSP fusion proteins to the same concentration and tested their enzyme activities in parallel. GST-SP fusion proteins following incubation with aprotinin in a final concentration of 5 g/ml were tested as control. Aprotinin shows strong inhibitory activity on serine protease activity of ST14 (10). As illustrated in Fig. 4, aprotinin inhibited the activity of purified GST-SP fusion proteins completely (Fig. 4, curve 3). Similarly, no serine protease activity was detected in the GST-mutSP fusion proteins that were mutated in the catalytic triad of the SP domain (Fig. 4, curve 2). Purified GST-SP fusion proteins were autoactivated slowly, and this activity lasted for several days (Fig. 4, curve 1).
Detection of the Interaction between GST-SP and HAI-1 by GST Pulldown Assays-To confirm the interaction between the SP domain and HAI-1, GST-SP fusion proteins were immobilized onto the glutathione-Sepharose 4B beads and tested for their ability to pull down the HAI-1 proteins that were expressed in the SW620-HAI-1 cells. A parallel experiment was performed using GST-mutSP under identical condi-    Table 1, respectively; lane 8, products of the second round PCR to amplify the full-length of LDLR domain. Five PCRs with five pairs of primers (Table 1), respectively, were used to create four point mutations in the first round. The first-round PCR products were diluted to the same concentration (Mol) and used as the template for the second round PCR. Mutagenesis of the SP domain was performed by the same method (data not shown).

FIGURE 4. Serine protease activity assays of GST-SP and GST-mutSP fusion proteins.
Three samples (GST-SP, GST-mutSP, and GST-SP) incubated with aprotinin (5 g/ml), respectively, were tested in the experiment. A time-dependent increase of serine protease activity was observed in GST-SP (curve 1). By contrast, GST-mutSP and GST-SP incubated with aprotinin showed no serine enzyme activity under identical conditions (curves 2 and 3). Fig. 5A, HAI-1 proteins were pulled down by both GST-SP and GST-mutSP; however, GST-SP pulled down more HAI-1 proteins than GST-mutSP. After an additional wash and increased ionic strength of washing buffer were used, more HAI-1 was washed off from the GST-mutSP beads, whereas a substantial amount of HAI-1 was still bound to GST-SP (data not shown). To further confirm the interaction between the SP domain and HAI-I, we performed modified GST pulldown assays to test the interaction between the SP domain and HAI-1. As illustrated in Fig. 5B, HAI-1 proteins bound to GST-mutSP were eluted completely by 400 mM PBS-NaCl. By contrast, HAI-1 proteins bound to GST-SP were eluted by 600 and 800 mM PBS-NaCl. Therefore, modified GST pull-down assays can distinguish small differences in binding properties between proteins and their corresponding mutants.

tions. As illustrated in
GST-ST14 Domains as Bait to Capture Prey Proteins-Three cell lines, including Bcap37, RKO-pSecTag2A, and RKO-ST14, were used as sources of potential ST14-interacting proteins. ST14 is detectable in Bcap37 but not in RKO. Therefore, a full-length cDNA of ST14 was cloned into pSecTag2A and transfected into RKO cells. RKO-pSecTag2A was the RKO cell line transfected with the empty pSecTag2A vector and used as a control. The reverse transcription-PCR confirmed the successful transfection. Western blot analysis indicated that HAI-1 was not detectable in these cell lines (Fig. 6A).
Three pairs of GST fusion proteins (GST-SP, GST-CUB, and GST-LDLR along with their mutants) were used to capture their interacting proteins in modified GST pull-down assays. Either the proteins bound to both wild-type and mutant ST domains, or bound proteins not sufficient for identification were excluded from the candidate pool of ST14-interacting proteins. The protein bands that were present in the fraction bound to the wild-type domains of ST14 but not in that bound to their mutants were excised from the SDS-polyacrylamide gel for peptide mass fingerprint analysis. No such specific binding proteins were identified in the assays of SP and LDLR domains. However, we obtained a protein of ϳ37.6 kDa that was still bound specifically to the wild-type CUB domain in the buffer with 800 mM NaCl (Fig. 6B). This 37.6-kDa protein was detected only in the RKO-ST14 cells, and it was not the GST or GST fusion protein according to its molecular weight. Interestingly, this protein was not detected in Bcap37 or RKO-pSecTag2A cells.
Peptide mass fingerprint analysis of this protein was shown in Fig. 7. By searching a peptide mass fingerprint data base (Aldente; available on the World Wide Web at au.expasy.org/tools/aldente/), this protein was found to be most identical to TMEFF1 (rank 1). The other candidate proteins were the precursor of TMEFF1 (rank 2) and DKFZp761G1118 (rank 3). DKFZp761G1118 was a hypothetical protein that was 89% identical to TMEFF1. Pairwise alignments of their corresponding mRNA sequences also demonstrated high similarity between these three genes. We then examined the expression of TMEFF1 in RKO-ST14 and RKO-pSecTag2A cells. TMEFF1 was also detected in RKO-ST14 but not in RKO-pSecTag2A by quantitative real time PCR (Fig.  6C). This result further indicated that this protein is TMEFF1, although it remains unknown whether this protein is the processed form of TMEFF1 or its precursor.
Co-expression of ST14 and TMEFF1 in Cell Lines and Tissues-Human placental tissue and a breast cell line MCF7 expressed ST14 and TMEFF1 at different levels. Northern blot analysis showed TMEFF1 was expressed at a moderately low level in the prostate cancer cell line PC3 (27). The ST14 mRNA is undetectable in PC3 according to quantitative real time PCR results of Bhatt et al. (28), although Takeuchi et al. (5) cloned ST14 cDNA from PC3 cells. Thus, we chose placental tissue, MCF7, and PC3 cell lines in our study. As shown in Fig. 8, our results confirmed the co-expression of ST14 and TMEFF1 in human placental tissue. However, ST14 was expressed at a low level in MCF cells and was FIGURE 5. GST pull-down assays to test the interaction between GST-SP and HAI-1. Shown is a Western blot analysis using monoclonal anti-HAI-1 antibody. A, interaction between SP/GST-mutSP and HAI-1 by using GST pull-down assays with beads. Lane 1, negative control; lane 2, whole cell lysates of SW620-HAI-1 as the positive control; lane 3, HAI-1 pulled down by GST-SP; lane 4, HAI-1 pulled down by GST-mutSP. B, interaction between SP/GST-mutSP and HAI-1 by using GST pull-down assays with columns. GST-SP and GST-mutSP were immobilized onto two columns, respectively. PBS-saline gradient elution was performed by using PBS with NaCl at a concentration of 200, 400, 600, 800, 1000, and 2000 mM, sequentially. HAI-1 bound to GST-mutSP was eluted completely by 400 mM PBS-NaCl. By contrast, HAI-1 bound to GST-SP was eluted by 600 and 800 mM PBS-NaCl.  undetectable in PC3 cells. We further tested the expression of ST14 and TMEFF1 in the tumor and matched adjacent normal tissue of kidney, liver, lung, breast, and colon. Co-expression of ST14 and TMEFF1 was identified in normal kidney, kidney tumor, normal liver, liver tumor, lung tumor, and breast tumor. Interestingly, ST14 and TMEFF1 were co-up-regulated in kidney tumors versus normal kidney tissues. Consistently, as mentioned above, we observed that the expression of TMEFF1 depends on the transfection of ST14 to RKO cells (Fig. 6C). ST14 was up-regulated in liver tumor as compared with normal liver tissue, whereas TMEFF1 was expressed at the same level in both tissues. TMEFF1 was expressed neither in colon carcinoma nor in normal colon tissue (Fig. 7).
Modeling the Interaction of TMEFF1 with ST14-After evaluating the three-dimensional compatibility between the CUB domain and Protein Data Bank templates, the CUB1-EGF-CUB2 regions of Masp2 (Protein Data Bank accession number 1NT0) were chosen as the template of the CUB domain. Pairwise alignment, however, revealed a very low identity between Masp2 and the EGF domain of TMEFF1. 1XDT, which is 43% identical to TMEFF1, was selected as the template of the EGF domain. A complex of the CUB-EGF domains was created by superimposing two predicted models to the coordinates in the template derived from 1NT0 using Swiss-PdbViewer (25). As illustrated in Fig. 9, the CUB domain of ST14 and EGF domain of TMEFF1 could form a complex similar to the CUB1-EGF-CUB2 regions of Masp2. This analysis suggested that ST14 forms a complex with TMEFF1 through an interaction between the CUB domain of ST14 and the EGF domain of TMEFF1.

DISCUSSION
ST14 is an interesting multidomain protein, which has been implicated in cancer metastasis (29,30). Different research groups have reported different splicing variant forms of ST14 since 1993. As shown in Table 2, five forms of ST14 have been detected with a molecular mass of 95, 80, 70, 45, and 25 kDa, respectively. It is postulated that these five different forms of ST14 are composed of the full-length protein (with a calculated molecular mass of 95 kDa), residues 149 -855 (78 kDa), residues 190 -855 (73 kDa) or 205-855 (74 kDa), residues 190 -614 (45.7 kDa), and residues 615-855 (26 kDa), respectively. These different forms of ST14, which contain various domains, are conceived to play different roles in different subcellular locations. Some forms of ST14 localize to the cell membrane, and others are secreted (7,9,11,31).
The SP domain is important for the activation of hepatocyte growth factor and urokinase-type plasminogen activator. HAI-1, aprotinin, and ecotin inhibit the serine protease activity of ST14 by binding to this domain (5,10). The CUB and LDLR domains have been thought to be involved in protein-protein interactions or protein-ligand interactions (32,33), but no interactions were reported for these domains on ST14. The CUB domain was originally identified in the complement subcomponents C1s and C1r (32). Most of the CUB domain-containing proteins are involved in development processes (32). The combinations of CUB with a variety of other protein domains have been implicated in two biochemical functions: (a) protein or ligand binding when the CUB domain is combined with either EGF, LDLR, Sushi, FA58C, Speract, or MAM domains and (b) proteolytic activity when the CUB domain is present in proteins such as astacin and trypsin-serine protease (34).
In the present work, three pairs of wild-type and mutant forms of GST ST14 domain fusion proteins were used to investigate the interacting proteins of each domain of ST14. We identified a CUB interacting protein, TMEFF1. Our results suggest that a CUB domain might interact with the EGF domain of TMEFF1 and form a complex with the EGF domain similar to the CUB1-EGF-CUB2 domain of Masp2.
TMEFF1 is a transmembrane protein, which contains a unique EGF domain and two Kazal_1 domains (35). These domains implicate a role for TMEFF1 in growth factor signaling (27). Several proteins that participate in the regulation of growth factor signaling, CSF1, IGF2, and TGFA have also been detected at high levels in the cell line RKO-ST14 (36). It has been suggested that TMEFF1 inhibits nodal signaling through binding to the nodal co-receptor Cripto (37). Although the Kazal_1 domain is known to inhibit a number of serine peptidases of the S1 family (35) that includes ST14 (38), Yamasaki et al. (10) showed that pancreatic secretory trypsin inhibitor containing a Kazal_1 domain did not inhibit ST14. Similarly, the GST pull-down assays in our work also did not provide evidence to support the interaction between the SP domain and TMEFF1.
The interaction of ST14 and TMEFF1 was not yet verified directly in our study because of the difficulty in expressing GST-ST14 (about 117 kDa) in E. coli. Nonetheless, our current results have strongly suggested the possibility that TMEFF1 participates in the ST14 activation through binding to the CUB domain. Protein processing and activation of ST14 are important for its biological role. As shown in Table 2, ST14 can be processed at residues 615, 205, 190, and 149 (11). We conceive that a group of regulators (activators and inhibitors) is necessary to determine the ST14 activation in a timely and spatial fashion by interacting with ST14. It is likely that HAI-1 and TMEFF1 govern different steps of ST14 activation, respectively. Additionally, the fact that ST14 and TMEFF1  co-localize on the cell membrane surface (27,39) also supports the possible interaction, since mature ST14 is secreted into human milk in a complexed form (9).
Northern blot analysis has revealed that TMEFF1 is expressed predominantly in the brain. Interestingly, matriptase-3, which is 31% identical to ST14, is also expressed at a high level in the brain (40). The expression of TMEFF1 is down-regulated in brain tumors (27), but the relationship between matriptase-3, TMEFF1, and cancer diseases has not been established. ST14 and TMEFF1 are hypothesized as tumor suppressor genes in colon cancers and in brain cancers, respectively (3,27,29,36,41). ST14 was detected, however, at very low levels in normal tissues but at high levels in breast carcinomas (6,42). In the present work, co-expression of ST14 and TMEFF1 was identified in human placenta, kidney, liver tissues, and MCF7 cells. More interestingly, the association between the expressions of these two genes was suggested both in vitro in cultured RKO-ST14 cell lines and in vivo in kidney tumors and normal kidney tissues. Co-up-regulation of ST14 and TMEFF1 in kidney tumors is not necessarily in conflict with the hypothesis that TMEFF1 behaves as a tumor suppressor gene in brain cancers. Take together, various expression levels in different tissue sites suggest tissue-specific roles of ST14 and TMEFF1 and their interaction in tumorigenesis, although more clinical samples need to be examined for this hypothesis.
The GST pull-down assay is a classical method to identify proteinprotein interactions in vitro. In a typical pull-down assay, a GST fusion bait protein is bound to an immobilized glutathione support and used to detect prey proteins from cell lysates. The interacting proteins are then detected by SDS-PAGE or Western blot. Various washing steps have been described (14,15,43), and a universal protocol does not exist. Our modified GST pull-down assay has improved the classical assay in the situation that the bait protein differs not very much in binding to its partners from the control protein. It enabled us to search for novel partners of ST14 domains.
Serine protease activity of GST-SP was verified using tissue plasminogen activator chromogenic substrate. Since the prokaryotic expressing system lacks SP activation machinery, this process may be an autoactivation. It is consistent with the report of Takeuchi et al. (5), in which a His-tagged fusion protein of ST14 was expressed in E. coli. The biological function of SP autoactivation needs further attention.
As shown in Table 2, different forms of ST14 are detected in different cell lines and tissues. They have different molecular weights, cell localization, and protein partners. Further exploration of the wild-type ST14 in its different cellular forms and its corresponding mutants is currently being carried out to uncover the roles of these different forms of ST14. Such a study will be of great help for our understanding of the membrane-associated proteolytic system and the mechanism of cancer metastasis.