Title: The Activation of Matriptase Requires Its Noncatalytic Domains, Serine Protease Domain, and Its Cognate Inhibitor
Abstract: The activation of matriptase requires proteolytic cleavage at a canonical activation motif that converts the enzyme from a one-chain zymogen to an active, two-chain protease. In this study, matriptase bearing a mutation in its catalytic triad was unable to undergo this activational cleavage, suggesting that the activating cleavage occurs via a transactivation mechanism where interaction between matriptase zymogen molecules leads to activation of the protease. Using additional point and deletion mutants, we showed that activation of matriptase requires proteolytic processing at Gly-149 in the SEA domain of the protease, glycosylation of the first CUB domain and the serine protease domain, and intact low density lipoprotein receptor class A domains. Its cognate inhibitor, hepatocyte growth factor activator inhibitor-1, may also participate in the activation of matriptase, based on the observation that matriptase activation did not occur when the protease was co-expressed with hepatocyte growth factor activator inhibitor-1 mutated in its low density lipoprotein receptor class A domain. These results suggest that besides matriptase catalytic activity, matriptase activation requires post-translational modification of the protease, intact noncatalytic domains, and its cognate inhibitor. The activation of matriptase requires proteolytic cleavage at a canonical activation motif that converts the enzyme from a one-chain zymogen to an active, two-chain protease. In this study, matriptase bearing a mutation in its catalytic triad was unable to undergo this activational cleavage, suggesting that the activating cleavage occurs via a transactivation mechanism where interaction between matriptase zymogen molecules leads to activation of the protease. Using additional point and deletion mutants, we showed that activation of matriptase requires proteolytic processing at Gly-149 in the SEA domain of the protease, glycosylation of the first CUB domain and the serine protease domain, and intact low density lipoprotein receptor class A domains. Its cognate inhibitor, hepatocyte growth factor activator inhibitor-1, may also participate in the activation of matriptase, based on the observation that matriptase activation did not occur when the protease was co-expressed with hepatocyte growth factor activator inhibitor-1 mutated in its low density lipoprotein receptor class A domain. These results suggest that besides matriptase catalytic activity, matriptase activation requires post-translational modification of the protease, intact noncatalytic domains, and its cognate inhibitor. Matriptase is a multi-domain, transmembrane serine protease of the S1 trypsin-like family (1Lin C.Y. Anders J. Johnson M. Sang Q.A. Dickson R.B. J. Biol. Chem. 1999; 274: 18231-18236Abstract Full Text Full Text PDF PubMed Scopus (190) Google Scholar). It belongs to the type II transmembrane serine protease group, the members of which are characterized by an N-terminal transmembrane domain and multiple extracellular domains, in addition to the conserved extracellular serine protease catalytic domain (2Hooper J.D. Clements J.A. Quigley J.P. Antalis T.M. J. Biol. Chem. 2001; 276: 857-860Abstract Full Text Full Text PDF PubMed Scopus (317) Google Scholar). Matriptase is also called the membrane-type serine protease-1, the tumor-associated differentially expressed gene-15, and the suppressor of tumorigenicity-14 (3Takeuchi T. Shuman M.A. Craik C.S. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 11054-11061Crossref PubMed Scopus (231) Google Scholar, 4Tanimoto H. Underwood L.J. Wang Y. Shigemasa K. Parmley T.H. O'Brien T.J. Tumour Biol. 2001; 22: 104-114Crossref PubMed Scopus (75) Google Scholar, 5Zhang Y. Cai X. Schlegelberger B. Zheng S. Cytogenet. Cell Genet. 1998; 83: 56-57Crossref PubMed Google Scholar). The mouse, rat, and Xenopus laevis orthologues of matriptase have been cloned (6Kim M.G. Chen C. Lyu M.S. Cho E.G. Park D. Kozak C. Schwartz R.H. Immunogenetics. 1999; 49: 420-428Crossref PubMed Scopus (123) Google Scholar, 7Satomi S. Yamasaki Y. Tsuzuki S. Hitomi Y. Iwanaga T. Fushiki T. Biochem. Biophys. Res. Commun. 2001; 287: 995-1002Crossref PubMed Scopus (75) Google Scholar, 8Yamada K. Takabatake T. Takeshima K. Gene (Amst.). 2000; 252: 209-216Crossref PubMed Scopus (24) Google Scholar) and share significant sequence homology with the human protein. Expression of the protease has been detected in the epithelial components of most epithelium-containing tissues (3Takeuchi T. Shuman M.A. Craik C.S. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 11054-11061Crossref PubMed Scopus (231) Google Scholar, 4Tanimoto H. Underwood L.J. Wang Y. Shigemasa K. Parmley T.H. O'Brien T.J. Tumour Biol. 2001; 22: 104-114Crossref PubMed Scopus (75) Google Scholar, 6Kim M.G. Chen C. Lyu M.S. Cho E.G. Park D. Kozak C. Schwartz R.H. Immunogenetics. 1999; 49: 420-428Crossref PubMed Scopus (123) Google Scholar, 9Lin C.Y. Anders J. Johnson M. Dickson R.B. J. Biol. Chem. 1999; 274: 18237-18242Abstract Full Text Full Text PDF PubMed Scopus (230) Google Scholar, 10Oberst M. Anders J. Xie B. Singh B. Ossandon M. Johnson M. Dickson R.B. Lin C.Y. Am. J. Pathol. 2001; 158: 1301-1311Abstract Full Text Full Text PDF PubMed Scopus (168) Google Scholar, 11Oberst, M. D., Singh, B., Ossandon, M., Dickson, R. B., Johnson, M. D., and Lin, C. Y. (2003) J. Histochem. Cytochem., in pressGoogle Scholar). The expression pattern of matriptase suggests that the protease may play broad roles in the biology of surface lining epithelial cells. Recently, matriptase was shown to be essential for postnatal survival, because of a loss of epidermal barrier function, in newborn, matriptase–/– (knockout) mice (12List K. Haudenschild C.C. Szabo R. Chen W. Wahl S.M. Swaim W. Engelholm L.H. Behrendt N. Bugge T.H. Oncogene. 2002; 21: 3765-3779Crossref PubMed Scopus (281) Google Scholar). These matriptase knockout mice also had abnormal hair follicle development and disturbed thymic homeostasis, as indicated by increased lymphocyte apoptosis in the thymuses of newborn mice. Matriptase has been shown to proteolytically activate the hepatocyte growth factor, the urokinase plasminogen activator, and protease-activated receptor-2 in vitro (13Lee S.L. Dickson R.B. Lin C.Y. J. Biol. Chem. 2000; 275: 36720-36725Abstract Full Text Full Text PDF PubMed Scopus (350) Google Scholar, 14Takeuchi T. Harris J.L. Huang W. Yan K.W. Coughlin S.R. Craik C.S. J. Biol. Chem. 2000; 275: 26333-26342Abstract Full Text Full Text PDF PubMed Scopus (386) Google Scholar). Hepatocyte growth factor and urokinase plasminogen activator have been implicated in cancer invasion and metastasis for their roles in cellular motility, extracellular matrix degradation, and tumor vascularization (15Comoglio P.M. Boccaccio C. Semin. Cancer Biol. 2001; 11: 153-165Crossref PubMed Scopus (113) Google Scholar, 16Andreasen P.A. Egelund R. Petersen H.H. Cell Mol. Life Sci. 2000; 57: 25-40Crossref PubMed Scopus (830) Google Scholar). Matriptase is also thought to play a role in cancer invasion and metastasis through its extracellular matrix-degrading activity (7Satomi S. Yamasaki Y. Tsuzuki S. Hitomi Y. Iwanaga T. Fushiki T. Biochem. Biophys. Res. Commun. 2001; 287: 995-1002Crossref PubMed Scopus (75) Google Scholar, 17Lin C.Y. Wang J.K. Torri J. Dou L. Sang Q.A. Dickson R.B. J. Biol. Chem. 1997; 272: 9147-9152Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar) and its potential role in activating urokinase plasminogen activator and hepatocyte growth factor on the surfaces of cancer cells (13Lee S.L. Dickson R.B. Lin C.Y. J. Biol. Chem. 2000; 275: 36720-36725Abstract Full Text Full Text PDF PubMed Scopus (350) Google Scholar, 14Takeuchi T. Harris J.L. Huang W. Yan K.W. Coughlin S.R. Craik C.S. J. Biol. Chem. 2000; 275: 26333-26342Abstract Full Text Full Text PDF PubMed Scopus (386) Google Scholar). Additionally, overexpression of matriptase in gastric cancer cells significantly enhanced their lymph node metastasis in nude mice (18Ihara S. Miyoshi E. Ko J.H. Murata K. Nakahara S. Honke K. Dickson R.B. Lin C.Y. Taniguchi N. J. Biol. Chem. 2002; 277: 16960-16967Abstract Full Text Full Text PDF PubMed Scopus (171) Google Scholar), and matriptase antisense has been shown to decrease the invasion of an ovarian tumor cell line in an in vitro invasion assay (19Suzuki M. Kobayashi H. Tanaka Y. Hirashima Y. Kanayama N. Takei Y. Saga Y. Suzuki M. Itoh H. Terao T. J. Biol. Chem. 2003; 278: 14640-14646Abstract Full Text Full Text PDF PubMed Scopus (47) Google Scholar). The expression of matriptase in human primary tumors has been examined by our group and others (4Tanimoto H. Underwood L.J. Wang Y. Shigemasa K. Parmley T.H. O'Brien T.J. Tumour Biol. 2001; 22: 104-114Crossref PubMed Scopus (75) Google Scholar, 10Oberst M. Anders J. Xie B. Singh B. Ossandon M. Johnson M. Dickson R.B. Lin C.Y. Am. J. Pathol. 2001; 158: 1301-1311Abstract Full Text Full Text PDF PubMed Scopus (168) Google Scholar, 20Oberst M.D. Johnson M.D. Dickson R.B. Lin C.Y. Singh B. Stewart M. Williams A. al Nafussi A. Smyth J.F. Gabra H. Sellar G.C. Clin. Cancer Res. 2002; 8: 1101-1107PubMed Google Scholar, 21Kang J.Y. Dolled-Filhart M. Ocal I.T. Singh B. Lin C.Y. Dickson R.B. Rimm D.L. Camp R.L. Cancer Res. 2003; 63: 1101-1105PubMed Google Scholar). Tanimoto et al. (4Tanimoto H. Underwood L.J. Wang Y. Shigemasa K. Parmley T.H. O'Brien T.J. Tumour Biol. 2001; 22: 104-114Crossref PubMed Scopus (75) Google Scholar) found that tumor-associated differentially expressed gene-15 (matriptase) was overexpressed in low malignant potential tumors and ovarian carcinomas relative to normal ovarian surface epithelium. We found that matriptase was expressed in the absence of its endogenous inhibitor, the hepatocyte growth factor activator inhibitor-1 (HAI-1) 1The abbreviations used are: HAI-1, hepatocyte growth factor activator inhibitor-1; LDL, low density lipoprotein; LDLR, LDL receptor; MMP, matrix metalloproteinase; mAb, monoclonal antibody.1The abbreviations used are: HAI-1, hepatocyte growth factor activator inhibitor-1; LDL, low density lipoprotein; LDLR, LDL receptor; MMP, matrix metalloproteinase; mAb, monoclonal antibody. (9Lin C.Y. Anders J. Johnson M. Dickson R.B. J. Biol. Chem. 1999; 274: 18237-18242Abstract Full Text Full Text PDF PubMed Scopus (230) Google Scholar), more frequently in late stage ovarian tumors, suggesting that matriptase proteolytic activity may be deregulated during tumor progression (20Oberst M.D. Johnson M.D. Dickson R.B. Lin C.Y. Singh B. Stewart M. Williams A. al Nafussi A. Smyth J.F. Gabra H. Sellar G.C. Clin. Cancer Res. 2002; 8: 1101-1107PubMed Google Scholar). In node-negative human breast tumors, high level expression of matriptase, and HAI-1, in addition to c-Met, were associated with poor patient outcome. Both c-Met and HAI-1 proved to be independent prognostic factors when compared with traditional breast cancer markers in multivariate analysis (21Kang J.Y. Dolled-Filhart M. Ocal I.T. Singh B. Lin C.Y. Dickson R.B. Rimm D.L. Camp R.L. Cancer Res. 2003; 63: 1101-1105PubMed Google Scholar). In addition to alterations in the levels of matriptase and HAI-1 proteins in primary tumors, deregulation of matriptase activation could contribute to cancer initiation, promotion, or progression. Indeed we have found that in breast cancer cells matriptase is constitutively activated (22Benaud C.M. Oberst M. Dickson R.B. Lin C.Y. Clin Exp. Metastasis. 2002; 19: 639-649Crossref PubMed Scopus (76) Google Scholar), in contrast to immortalized mammary epithelial cells, where the activation of matriptase is dependent on sphingosine 1-phosphate, a blood-borne phospholipid (23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar, 24Benaud C. Oberst M. Hobson J.P. Spiegel S. Dickson R.B. Lin C.Y. J. Biol. Chem. 2002; 277: 10539-10546Abstract Full Text Full Text PDF PubMed Scopus (77) Google Scholar). As with other typical serine proteases, matriptase is activated by cleavage at a canonical activation motif to convert a single chain zymogen to a two-chain active protease (23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar). Spontaneous activation of a bacterially expressed, purified serine protease domain of matriptase has been reported, and mutation of the serine residue of the active site triad abolished this activation (3Takeuchi T. Shuman M.A. Craik C.S. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 11054-11061Crossref PubMed Scopus (231) Google Scholar), demonstrating that matriptase activity is sufficient for activation of this truncated form of matriptase. For most serine proteases, activation-related cleavage is carried out by other active proteases. An alternative mechanism, transactivation, whereby a latent zymogen can interact with another zymogen resulting in transactivation of each zymogen, is believed to be relevant for some serine proteases, particularly for those proteases at the pinnacle of protease cascades. The best studied example of this mechanism is the activation of complement subcomponent C1r protease, where interaction of C1r/C1s tetramers with the C1q protein induces conformational changes that result in C1r protease transactivation (25Thielens N.M. Bersch B. Hernandez J.F. Arlaud G.J. Immunopharmacology. 1999; 42: 3-13Crossref PubMed Scopus (21) Google Scholar). Because the active site triad of serine proteases already may be well formed in the zymogen state (26Bode W. Schwager P. Huber R. J. Mol. Biol. 1978; 118: 99-112Crossref PubMed Scopus (245) Google Scholar), transactivation probably requires the interaction of two protease zymogen molecules with each other, and possibly with other proteins, to induce a conformational change in the substrate-binding pocket required for catalysis. In the current study, we first investigated whether the activational cleavage of matriptase could be carried out if the catalytic triad of the enzyme were modified to inactivate the enzyme. If the canonical cleavage of matriptase could still occur in cells transfected with matriptase mutated in its catalytic triad, then a mechanism of activation involving another cellular protease would be likely. We next examined whether noncatalytic domains of matriptase or HAI-1 are required in its activation. Cell Culture Conditions—BT549 human breast cancer cells were maintained in culture by growth in Iscove's minimal essential medium (Invitrogen) supplemented with 5% fetal bovine serum in a humidified chamber at 37 °C and 5% CO2. Monoclonal Antibodies—Human matriptase protein was detected using either the M32 or M84 monoclonal antibodies that recognize both the latent (one-chain) and activated (two-chain) forms of the protease or using the M69 monoclonal antibody that recognizes an epitope present only in the activated (two-chain) form of the enzyme (23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar, 24Benaud C. Oberst M. Hobson J.P. Spiegel S. Dickson R.B. Lin C.Y. J. Biol. Chem. 2002; 277: 10539-10546Abstract Full Text Full Text PDF PubMed Scopus (77) Google Scholar). Human HAI-1 was detected using the HAI-1-specific monoclonal antibody M19 (9Lin C.Y. Anders J. Johnson M. Dickson R.B. J. Biol. Chem. 1999; 274: 18237-18242Abstract Full Text Full Text PDF PubMed Scopus (230) Google Scholar). Constructs and Transfections—The cDNA clones for the full human matriptase coding sequence and the full human HAI-1 coding sequence in the vector pcDNA3.1 (Invitrogen) were used in transient transfections. These constructs were also used to make site-directed or deletion mutants of matriptase or HAI-1. For making site-directed mutations, the QuikChange® site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used with primers containing appropriate nucleotide changes according to the manufacturer's protocol. To create deletion mutants, pcDNA3.1 vectors containing the full-length human matriptase or human HAI-1 coding sequences were used as template in PCRs with Pfu Turbo® polymerase (Stratagene, La Jolla, CA) according to standard protocols with primers designed to anneal to each strand flanking the sequence to be deleted and polymerized in the direction opposite this sequence. PCR products were then blunt end-ligated using the Quick ligation kit (New England Biolabs, Beverly, MA) to create each deletion mutant. All site-directed mutations and deletion mutants were confirmed by DNA sequencing. For each mutant, at least two separately isolated expression constructs were isolated and tested in transient transfections. As with any study using point or deletion mutants, the potential exists for confounding results because of protein misfolding that could more generally disrupt protein function. In this work, although we did not formally test individual mutants for proper protein folding, we did make conservative amino acid changes with the point mutants and did not observe any significant changes in protein size or apparent stability of latent mutant matriptase or HAI-1 proteins that would suggest misfolded or unstable proteins. Transient transfection of human matriptase or human HAI-1 constructs (wild-type or mutant) was accomplished using FuGENE 6 transfection reagent (Roche Applied Science), according to the manufacturer's protocol. When doing co-transfections, the amount of DNA used with the transfection reagent was kept constant for each individual transfection by including empty vector pcDNA3.1 DNA, where appropriate. Western Blotting—Protein for Western blotting was prepared by the lysis of cells in RIPA buffer (0.1% Nonidet-P40, 0.5% sodium deoxycholate, 0.1% SDS in phosphate-buffered saline, with added protease inhibitors 1 μg/ml leupeptin, 1 μg/ml aprotinin, 1 μg/ml pepstatin A, and 100 μg/ml phenylmethylsulfonyl fluoride) after washing cells two times in phosphate-buffered saline. Insoluble debris was removed by centrifugation, and the protein concentration was measured using the BCA protein assay reagents (Pierce) according to the manufacturer's protocol. The lysates were diluted immediately in 2× sample buffer (100 mm Tris, pH 6.8, 16% sucrose (w/v), 3% SDS, 0.02% bromphenol blue) and stored at –80 °C prior to Western blotting. The sample buffer does not contain a reducing agent, and samples were not boiled prior to SDS-PAGE, because reducing agents destroy the epitopes recognized by the mAbs, and boiling disrupts matriptase-HAI-1 complexes. The proteins were resolved by 10% SDS-PAGE, transferred to Protran® nitrocellulose membranes (Schleicher & Schuell), and probed with monoclonal antibodies that recognize human matriptase (M32, M84, and M69), human HAI-1 (M19), or α tubulin, using the α tubulin-specific monoclonal antibody-2/DM1A (Lab Vision Corporation, Fremont, CA). The binding of the primary antibody was followed by recognition with a goat anti-mouse horseradish peroxidase-conjugated secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) and detected using the Western Lightening® Chemiluminescence Reagent Plus (PerkinElmer Life Sciences). The monoclonal antibodies recognizing human matriptase and human HAI-1 were generated against conformation-dependent epitopes in the respective proteins, and therefore samples were run under nonreducing SDS-PAGE conditions and were not boiled prior to electrophoresis to preserve the formation of complexes between activated matriptase and HAI-1. Matriptase contains an N-terminal transmembrane domain, followed by a SEA (sperm protein, enterokinase, and agrin) domain (31Kilpatrick, L. M., and Ellis, V. (2002) Biochemical Society Meeting 678, London, December 16–18, 2002, Abstract 20Google Scholar), two tandem CUB (C1r/s, Uegf, and bone morphogenic protein-1) domains, four tandem LDL receptor class A domains, and a catalytic serine protease domain. It also contains four putative asparagine-linked glycosylation sites, and a potential integrin-binding RGD motif is found in the first CUB domain. The protease also contains a consensus cleavage site in the SEA domain (GSVI) that is the target for a proteolytic processing cleavage in matriptase (31Kilpatrick, L. M., and Ellis, V. (2002) Biochemical Society Meeting 678, London, December 16–18, 2002, Abstract 20Google Scholar) and in the mouse orthologue, epithin (6Kim M.G. Chen C. Lyu M.S. Cho E.G. Park D. Kozak C. Schwartz R.H. Immunogenetics. 1999; 49: 420-428Crossref PubMed Scopus (123) Google Scholar). This cleavage converts matriptase and epithin to smaller forms that lack the N-terminal 149 amino acids containing the transmembrane domain, yet these processed forms are still tightly associated with the membrane. To investigate the roles of these domains and motifs in the activation of matriptase, we systematically constructed various mutants of human matriptase (hereafter referred to as simply matriptase) by site-directed or deletion mutagenesis (Fig. 1). We also constructed point and deletion mutants of the human HAI-1 (hereafter refereed to as simply HAI-1) in the LDL receptor class A domain to test the role of this domain within the inhibitor in matriptase activation (Fig. 1). Matriptase Activation Requires the Catalytic Activity of the Serine Protease Domain—BT549 human breast cancer cells do not endogenously express either matriptase or HAI-1, as assessed by anti-matriptase or anti-HAI-1 Western blotting (Fig. 2, A and C). Therefore, the detection of transfected matriptase and HAI-1 could not be confused with endogenous proteins. In cells transfected with both wild-type matriptase and HAI-1, both the uncomplexed, processed form of matriptase (70-kDa form) and the processed form, complexed to HAI-1 (120-kDa complex) were observed, in addition to the full-length matriptase that migrates at ∼95 kDa (Fig. 2A). This result indicates that matriptase had been activated when co-transfected with HAI-1, because only the activated form of the enzyme binds to the inhibitor HAI-1 to form the 120-kDa complex. In addition, we confirmed this observation by Western blotting with the M69 mAb, that recognizes only the two-chain, activated form of matriptase and not the one-chain, latent form of the protease (Fig. 2B). It should be pointed out that both the complexed and uncomplexed forms of matriptase were observed in M69 Western blots, and the ratio of these forms varies from experiment to experiment for unknown reasons (22Benaud C.M. Oberst M. Dickson R.B. Lin C.Y. Clin Exp. Metastasis. 2002; 19: 639-649Crossref PubMed Scopus (76) Google Scholar, 23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar, 24Benaud C. Oberst M. Hobson J.P. Spiegel S. Dickson R.B. Lin C.Y. J. Biol. Chem. 2002; 277: 10539-10546Abstract Full Text Full Text PDF PubMed Scopus (77) Google Scholar). When inactive, catalytic triad mutants of matriptase were expressed alone or with HAI-1, the activated form of the protease was not observed, as indicated by a lack of the formation of a 120-kDa matriptase-HAI-1 complex and lack of immunoreactivity with the M69 mAb (Fig. 2, A and B). This data indicates that matriptase activation in BT549 cells occurs by a transactivation mechanism, and is not likely due to the activity of other proteases. Glycosylation of the First CUB and Serine Protease Domains Is Required for Activation—In addition to mutations in the catalytic triad, other matriptase mutants were created to test the importance of noncatalytic domains and motifs in matriptase activation. Conservative point mutations were made in the putative glycosylation sites of matriptase (N109Q, N302Q, N485Q, and N772Q matriptase) to test their role in matriptase activation. The N772Q matriptase mutant profoundly inhibited the formation of the matriptase-HAI-1 complex (Fig. 3A) and reduced the M69 signal on Western blot (Fig. 3B). The N302Q matriptase mutant also dramatically reduced the formation of the matriptase-HAI-1 complex and M69 mAb immunoreactivity, although not quite as strongly as did the N772Q matriptase mutant (Fig. 3, A and B). These results strongly suggest that matriptase activation requires glycosylation of the serine protease domain and the first CUB domain. In contrast, the N109Q and N485Q matriptase glycosylation mutants did not reduce the matriptase-HAI-1 complex formation nor M69 mAb immunoreactivity (Fig. 3, A and B). Overall, these results strongly suggest that matriptase glycosylation can differentially influence the level of protease activation, with glycosylation of the serine protease domain and first CUB domain potentially being important for the activation process. Proteolytic Processing in the N-terminal SEA Domain of the Protease Is a Prerequisite for Matriptase Activation—Other point mutations made in matriptase included a putative proteolytic processing site within the SEA domain (G149N matriptase) important for the conversion of matriptase from a full-length 95-kDa protease to smaller molecular mass species that migrates at ∼70 kDa and an RGD motif (D251E matriptase) found in the first CUB domain (Fig. 1). In addition, each CUB domain was deleted individually (matriptase ΔCUB1 and matriptase ΔCUB2) and together (matriptase ΔCUB1&2) to test the function of these domains in matriptase activation. Mutation of the putative proteolytic processing site in the SEA domain at Gly-149 (G149N matriptase) resulted in the appearance of only nonprocessed, full-length matriptase, as evidenced by the shift in band migration from the processed 70-kDa form to the ∼95-kDa nonprocessed form (Fig. 3A). This is consistent with the inhibition of processing of matriptase when this site was mutated in A549 human lung carcinoma cells (31Kilpatrick, L. M., and Ellis, V. (2002) Biochemical Society Meeting 678, London, December 16–18, 2002, Abstract 20Google Scholar) and in the mouse orthologue of matriptase, epithin, when the homologous glycine at position 149 was mutated (6Kim M.G. Chen C. Lyu M.S. Cho E.G. Park D. Kozak C. Schwartz R.H. Immunogenetics. 1999; 49: 420-428Crossref PubMed Scopus (123) Google Scholar). When the N-terminal proteolytic processing of matriptase was inhibited, none of the full-length form was activated as evidenced by the lack of matriptase-HAI-1 complex formation (Fig. 3A) and a lack of M69 signal at 95 kDa on Western blot (Fig. 3B). This indicates that the N-terminal proteolytic processing is a prerequisite for enzyme activation. A low level of M69 signal was observed at 70 and 120 kDa for G149N matriptase, however, indicating that a small amount of this mutant was cleaved and subsequently activated. This result confirms that the N-terminal processed form was capable of activation, as expected. Mutation of the RGD sequence in matriptase did not affect matriptase-HAI-1 complex formation (Fig. 3A) nor the level of M69 signal on Western blot (Fig. 3B), indicating that this motif is not required for matriptase activation. To test the importance of the CUB domains in matriptase activation, the matriptase mutants lacking the first CUB domain (matriptase ΔCUB1), the second CUB domain (matriptase ΔCUB2), or both CUB domains simultaneously (matriptase ΔCUB1&2) were transfected into BT549 cells, together with HAI-1. Western blotting showed that deletion of either the CUB domain or both domains together inhibited the N-terminal processing of the protease, as indicated by the appearance of a predominant higher molecular mass form, corresponding to that of full-length matriptase with CUB domain deletions (Fig. 4A). Like the G149N matriptase mutant that prevented proteolytic processing of matriptase at the N terminus, the matriptase mutants containing single CUB deletions did not efficiently activate (Fig. 4, A and B). A low level of matriptase activation was observed for matriptaseΔCUB1, consistent with some of this mutant undergoing N-terminal processing and subsequent activation. Deletion of both CUB domains together resulted in more substantial matriptase activation, nearly equivalent to that of the wild-type matriptase (Fig. 4, A and B). For this mutant, more of the N-terminal processed matriptase was observed than for either matriptaseΔCUB1 or matriptaseΔCUB2. Activation Requires Intact LDL Receptor Class A Domains— Four LDL receptor class A domains are found in matriptase. This domain is an ∼40-amino acid-long structure with three pairs of disulfide linkages (27Brown M.S. Herz J. Goldstein J.L. Nature. 1997; 388: 629-630Crossref PubMed Scopus (146) Google Scholar) and is found in membrane receptors and many type II transmembrane serine protease members. The prototype structure of the LDL receptor class A domain is found in the LDL receptor itself, which contains seven such domains. The crystal structure of the fifth LDL receptor class A domain in the LDL receptor revealed that this domain contains six amino acids that coordinate a Ca2+ atom in an octahedral arrangement, termed the calcium cage (28Fass D. Blacklow S. Kim P.S. Berger J.M. Nature. 1997; 388: 691-693Crossref PubMed Scopus (299) Google Scholar). Point mutation at critical residues in the calcium cage potently inhibits ligand binding to this domain (29Esser V. Limbird L.E. Brown M.S. Goldstein J.L. Russell D.W. J. Biol. Chem. 1988; 263: 13282-13290Abstract Full Text PDF PubMed Google Scholar). Point mutations in the LDL receptor class A domains of matriptase were created at an aspartic acid residue in the fifth LDL receptor class A domain that inhibits the binding of this domain with ligand, without affecting the overall folding of the molecule or adjacent domains (29Esser V. Limbird L.E. Brown M.S. Goldstein J.L. Russell D.W. J. Biol. Chem. 1988; 263: 13282-13290Abstract Full Text PDF PubMed Google Scholar). Thus, separate point mutations were created in each of the matriptase LDL receptor class A domains individually (D483Y matriptase, D519Y matriptase, D555Y matriptase, and D598Y matriptase) and in all four domains simultaneously (D→ Y×4 matriptase). In addition, the four LDL receptor class A domains were deleted as a single unit (matriptase ΔLDLR). When transfected into BT549 cells with HAI-1, these mutants variably affected the activation of matriptase. Point mutation of each of the four matriptase LDL receptor class A domains (D482Y, D519Y, D555Y, and D598Y matriptase) or all four domains simultaneously (D→ Y×4 matriptase) inhibited the activation of matriptase, as shown by a lack of the 120-kDa matriptase-HAI-1 complex in M84 Western blotting (Fig. 5A) and by a lack of M69 mAb immunoreactivity (Fig. 5B). However, deletion of all four domains together had the opposite effect, leading to efficient activation of matriptase and formation of discrete higher mass complexes (Fig. 5, A and B). The LDL Receptor Class A Domain of HAI-1 Is Essential for Matriptase Activation—Because the activation of matriptase required the LDL receptor class A domain within matriptase and because this domain may be involved in protein-protein interactions, we asked whether a similar domain in HAI-1 was also essential for protease activation. Therefore, we constructed a point mutation at a critical Asp residue involved in calcium coordination in the calcium cage of the LDL receptor class A domain of HAI-1 (D349Y HAI-1) to investigate the function of this domain in matriptase activation (Fig. 1). We also created another mutant that contains a deletion of the entire LDL receptor class A domain (HAI-1 ΔLDLR). Loss of the LDL receptor class A domain function through point or deletion mutation dramatically inhibited the activation of cotransfected wild-type matriptase, as demonstrated by a lack of formation of the matriptase-HAI-1 complex (Fig. 6A) and by the lack of M69 mAb immunoreactivity (Fig. 6B). A summary of the effects of the mutational analysis of matriptase and HAI-1 with regard to matriptase activation is presented in Table I.Table IEffects of matriptase and HAI-1 mutations on the activation of matriptase in BT549 breast cancer cellsMutationDomain where mutation residesDegree of activationMatriptase mutants transfected with HAI-1Wild-type matriptase+++S805ATriad-H656ATriad-D711ATriad-D482YLDLR class A+D519YLDLR class A+D555YLDLR class A+D598YLDLR class A+D→Y×4LDLR class A-ΔLDLRLDLR class A+++N109QGlycosylation+++N302QGlycosylation+N485QGlycosylation+++N772QGlycosylation-G149NN-terminal processing-RGD251ERGD+++ΔCUB1CUB+ΔCUB2CUB-ΔCUB1&CUB2CUB++HAI-1 mutants transfected with matriptaseWild-type HAI-1+++D349YLDLR class A+ΔLDLRLDLR class A- Open table in a new tab Mutation of the catalytic triad revealed that matriptase activity was necessary for its own activation. This strongly suggests that matriptase activation occurs by a transactivation mechanism in BT549 breast cancer cells and is not due to the activity of other proteases. An autocatalytic mechanism whereby matriptase catalytically activates itself in an intramolecular fashion is excluded based upon the structural inaccessibility of the R614 activation cleavage site to the active site pocket. 2E. Madison, personal communication. This is the first demonstration to our knowledge that matriptase activation occurs by transactivation in the context of a cell and is similar to the autoactivation seen in vitro for the bacterially expressed, purified matriptase serine protease domain. Because the flanking sequence of the activation motif of matriptase matches the cleavage preference of matriptase (14Takeuchi T. Harris J.L. Huang W. Yan K.W. Coughlin S.R. Craik C.S. J. Biol. Chem. 2000; 275: 26333-26342Abstract Full Text Full Text PDF PubMed Scopus (386) Google Scholar), we could not rule out the possibility that the newly active matriptase activates latent matriptase. Thus, initial activation of matriptase could occur through transactivation, whereas subsequent activation could occur through both continued transactivation and cleavage of latent matriptase by already activated matriptase. The transactivation mechanism for matriptase stands in contrast to the activation mechanisms of most other serine proteases, which depend on the presence of other active proteases. In our previous studies (23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar, 24Benaud C. Oberst M. Hobson J.P. Spiegel S. Dickson R.B. Lin C.Y. J. Biol. Chem. 2002; 277: 10539-10546Abstract Full Text Full Text PDF PubMed Scopus (77) Google Scholar), sphingosine 1-phosphate was identified as a blood-borne, exogenous activator of matriptase in immortal mammary epithelial cells. Given the transactivation mechanism that may require protein-protein interactions and the activity of endogenous activators such as sphingosine 1-phosphate, the protease could reside at the pinnacle of a protease cascade that is initiated by nonproteolytic events. The cascade of protease activity that follows has not been fully characterized. However, our group and others have shown that matriptase can activate urokinase-type plasminogen activator, which in turn can activate plasminogen (13Lee S.L. Dickson R.B. Lin C.Y. J. Biol. Chem. 2000; 275: 36720-36725Abstract Full Text Full Text PDF PubMed Scopus (350) Google Scholar, 14Takeuchi T. Harris J.L. Huang W. Yan K.W. Coughlin S.R. Craik C.S. J. Biol. Chem. 2000; 275: 26333-26342Abstract Full Text Full Text PDF PubMed Scopus (386) Google Scholar). Using site-directed mutants of the matriptase LDL receptor class A domain, we found that this domain was important for matriptase activation. This could indicate that the LDL receptor class A domain is important for protein-protein interactions between matriptase and proteins involved in the activation process. This domain is important for the binding of apolipoprotein B-100 and apolipoprotein E by the LDL receptor, which contains seven such LDL receptor class A repeats (29Esser V. Limbird L.E. Brown M.S. Goldstein J.L. Russell D.W. J. Biol. Chem. 1988; 263: 13282-13290Abstract Full Text PDF PubMed Google Scholar). Any potential ligands that matriptase binds through this domain to facilitate activation await further characterization. It is interesting to speculate, however, that these domains may be involved in the oligomerization of matriptase in a transactivation complex, either as a homo-oligomeric complex or as a hetero-oligomeric complex with other proteins. In contrast to matriptase LDL receptor class A domain point mutants, a matriptase mutant containing a wholesale deletion of all four LDL receptor class A domains was activated to nearly the same extent as wild-type matriptase. This result may reflect the nature of the regulation afforded by the LDL receptor class A domain. The domain may serve a dual function: to serve as a binding domain for activators of the protease and to serve at other times as an autoinhibitory domain when activators of the protease are not present. In support of this idea, the purified matriptase catalytic domain that lacks the entire N-terminal part of matriptase containing the LDL receptor class A and CUB domains is autoactivated in vitro (3Takeuchi T. Shuman M.A. Craik C.S. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 11054-11061Crossref PubMed Scopus (231) Google Scholar), and removal of LDL receptor class A domains in cell-expressed matriptase may mimic this state of the protease. In addition to the role played by LDL receptor class A domains, the proteolytic processing of matriptase from a 95-kDa full-length form to lower molecular mass species also appears to be a prerequisite for matriptase activation. This was evidenced by the fact that a mutation in a proteolytic cleavage motif (GSIV) that prevents the cleavage at the SEA domain of matriptase greatly reduced the activation of matriptase. The mechanism whereby this N-terminal proteolytic processing may be involved in matriptase activation is unknown. It is plausible that upon release from its transmembrane anchor, matriptase undergoes a conformational change or reorientation on the membrane surface that makes the protease susceptible to transactivation cleavage at the canonical activation motif or to interaction with other proteins that facilitate transactivation. Deletion of a CUB domain also prevents proteolytic processing of matriptase to its smaller molecular mass forms, and this also resulted in poor activation of matriptase. Some matriptase was processed at the N terminus for each deletion mutant, and the activation of the processed form was in proportion to the amount of processed form seen on Western analysis. Thus, the deletion of a CUB domain is unlikely to prevent matriptase activation per se but instead may simply prevent the processing that is a prerequisite for the activation process. Interestingly, deletion of both CUB domains simultaneously resulted in significant matriptase activation nearly equal to that of wild-type matriptase. This phenomena is similar to that which occurs with deletion of all four LDL receptor class A domains simultaneously: CUB domains together may serve as autoinhibitory domains when not involved in an activation event, the simultaneous deletion of which reveals this function. Point mutation of matriptase also demonstrated that glycosylation of the first CUB domain (at Asn-302) and the serine protease domain (at Asn-772) is important for matriptase activation. The function of matriptase glycosylation in the activation process in our system remains an open question. It is possible that the modification of matriptase by glycosylation in the CUB domain and serine protease domains could provide structural information necessary for the protein-protein interactions required for activation. These results also suggest the intriguing possibility that the activity of matriptase may be differentially regulated by the glycosylation status of these sites. In addition to the domains and motifs within matriptase discussed above, we discovered that the single LDL receptor class A domain of HAI-1 was also required for matriptase activation. The domain may recruit the inhibitor to a matriptase-HAI-1 activation complex where HAI-1 may help to activate the protease prior to acting as a competitive inhibitor. Such a situation is reminiscent of the role that the tissue inhibitor of matrix metalloproteinase-2 plays in pro-matrix metalloproteinase-2 (proMMP-2 or progelatinase A) activation, whereby the inhibitor facilitates MMP-2 activation by the membrane-type 1 matrix metalloproteinase at a low molar ratio on cell surfaces and inhibits protease activation at higher stoichiometry (30Kinoshita T. Sato H. Okada A. Ohuchi E. Imai K. Okada Y. Seiki M. J. Biol. Chem. 1998; 273: 16098-16103Abstract Full Text Full Text PDF PubMed Scopus (258) Google Scholar). This probably occurs when tissue inhibitor of matrix metalloproteinase-2 forms a trimolecular complex with both proMMP-2 and membrane type 1 MMP, allowing proMMP-2 to be cleaved and activated by a nearby unbound (and uninhibited) membrane type 1 MMP. Similarly, HAI-1 may form oligomeric complexes with matriptase and/or other proteins through its LDL receptor class A domain to facilitate transactivation. Such a mechanism of protease activation may ensure that protease activity is tightly controlled by closely approximating activation to subsequent processes that eliminate protease activity such as inhibitor binding and the internalization and shedding of protease-inhibitor complexes. The fact that HAI-1 might play a role in matriptase activation suggests that HAI-1 may modulate matriptase activity by several mechanisms. If HAI-1 were required for the activation of matriptase, then the protease could not be activated in the absence of a mechanism to control the subsequent activity. As a competitive inhibitor of matriptase, HAI-1 could additionally serve to modulate protease activity by tightly binding matriptase once the protease has cleaved relevant substrates (23Benaud C. Dickson R.B. Lin C.Y. Eur. J Biochem. 2001; 268: 1439-1447Crossref PubMed Scopus (117) Google Scholar). Shedding of the protease/inhibitor pair or internalization and degradation could then serve to permanently quench protease activity. Thus, HAI-1 may have acquired multiple strategies by which it controls matriptase activity, a redundancy that may be required given the potentially destructive nature of unregulated cell surface proteolysis. In summary, these results suggest that matriptase activation in BT549 breast cancer cells occurs by a transactivation mechanism that further requires noncatalytic domains of both matriptase and HAI-1. We propose a model whereby the LDL receptor class A domains of matriptase and HAI-1 mediate the formation of an activation complex that induces the conformational changes that transiently allow one matriptase zymogen to permanently activate another. This process apparently requires removal of the N terminus of matriptase and proper glycosylation of both the serine protease domain and the first CUB domain. The formation of the activation complex may be regulated by exogenous activators such as sphingosine 1-phosphate in immortalized breast epithelial cells but is unresponsive to such signals in breast cancer cell lines where activation complexes may be constitutively formed (22Benaud C.M. Oberst M. Dickson R.B. Lin C.Y. Clin Exp. Metastasis. 2002; 19: 639-649Crossref PubMed Scopus (76) Google Scholar).