Matriptase-2, a Membrane-bound Mosaic Serine Proteinase Predominantly Expressed in Human Liver and Showing Degrading Activity against Extracellular Matrix Proteins*

We have identified and cloned a fetal liver cDNA encoding a new serine proteinase that has been called matriptase-2. This protein exhibits a domain organization similar to other members of an emerging family of membrane-bound serine proteinases known as type II transmembrane serine proteinases. Matriptase-2 contains a short cytoplasmic domain, a type II transmembrane sequence, a central region with several modular structural domains including two CUB (complement factor C1s/C1r, urchin embryonic growth factor,bone morphogenetic protein) domains and three low density lipoprotein receptor tandem repeats, and finally, a C-terminal catalytic domain with all typical features of serine proteinases. The human matriptase-2 gene maps to 22q12-q13, a location that differs from all type II transmembrane serine proteinase genes mapped to date. Immunofluorescence and Western blot analysis of COS-7 cells transfected with the isolated cDNA confirmed that matriptase-2 is anchored to the cell surface. Matriptase-2 was expressed in Escherichia coli, and the purified recombinant protein hydrolyzed synthetic substrates used for assaying serine proteinases and endogenous proteins such as type I collagen, fibronectin, and fibrinogen. Matriptase-2 could also activate single-chain urokinase plasminogen activator, albeit with low efficiency. These activities were abolished by inhibitors of serine proteinases but not by inhibitors of other classes of proteolytic enzymes. Northern blot analysis demonstrated that matriptase-2 transcripts are only detected at significant levels in both fetal and adult liver, suggesting that this novel serine proteinase may play a specialized role in matrix remodeling processes taking place in this tissue during development or in adult tissues.

We have identified and cloned a fetal liver cDNA encoding a new serine proteinase that has been called matriptase-2. This protein exhibits a domain organization similar to other members of an emerging family of membrane-bound serine proteinases known as type II transmembrane serine proteinases. Matriptase-2 contains a short cytoplasmic domain, a type II transmembrane sequence, a central region with several modular structural domains including two CUB (complement factor C1s/C1r, urchin embryonic growth factor, bone morphogenetic protein) domains and three low density lipoprotein receptor tandem repeats, and finally, a Cterminal catalytic domain with all typical features of serine proteinases. The human matriptase-2 gene maps to 22q12-q13, a location that differs from all type II transmembrane serine proteinase genes mapped to date. Immunofluorescence and Western blot analysis of COS-7 cells transfected with the isolated cDNA confirmed that matriptase-2 is anchored to the cell surface. Matriptase-2 was expressed in Escherichia coli, and the purified recombinant protein hydrolyzed synthetic substrates used for assaying serine proteinases and endogenous proteins such as type I collagen, fibronectin, and fibrinogen. Matriptase-2 could also activate single-chain urokinase plasminogen activator, albeit with low efficiency. These activities were abolished by inhibitors of serine proteinases but not by inhibitors of other classes of proteolytic enzymes. Northern blot analysis demonstrated that matriptase-2 transcripts are only detected at significant levels in both fetal and adult liver, suggesting that this novel serine proteinase may play a specialized role in matrix remodeling processes taking place in this tissue during development or in adult tissues.
Proteolytic enzymes play crucial roles in the development and maintenance of an organism as well as in a number of pathological conditions including the progression of malignant tumors (1). Most studies on cancer-associated proteinases have focused on matrix metalloproteinases (MMPs), 1 a family of zinc-dependent endopeptidases that collectively degrade all major protein components from extracellular matrix and basement membranes (2,3). However, enzymes from other catalytic classes such as cysteine, aspartyl, and serine proteinases have been also implicated in different aspects of tumor progression. Among them, an emerging group of membrane serine proteinases, called TTSPs and containing a complex organization of domains, have raised recent interest because of their potential ability to participate in matrix-degrading processes associated with cancer (reviewed in Ref. 4). To date, 11 distinct human TTSPs have been described and characterized at the amino acid sequence level. They include enteropeptidase, hepsin, human airway trypsin-like protease (HAT), corin, matriptase/MT-SP1, epitheliasin/TMPRSS2, TADG-12/TMPRSS3, TMPRSS4, MSPL (mosaic serine protease large form), spinesin/TMPRSS5, and DESC1 protease (differentially expressed squamous cell carcinoma gene 1) (4 -6). All of them share a number of structural features: a short N-terminal cytoplasmic domain, a type II transmembrane sequence, a central region of variable length containing modular structural domains, and a C-terminal catalytic region with all of the characteristic features of serine proteinases. TTSPs have been found in a wide variety of mammalian tissues as well as in other eukaryotic organisms including Drosophila melanogaster (7) and Xenopus laevis (8).
Although the physiological roles of most TTSPs are still unclear, there are some cases in which their participation in specific functions has been suggested or demonstrated. This is the case of enteropeptidase that is involved in the proteolytic activation of trypsinogen to trypsin, which subsequently activates other digestive enzymes such as chymotrypsinogen or procarboxypeptidases (9,10). Likewise, matriptase/MT-SP1 has been proposed to initiate signaling and proteolytic cascades through their ability to activate cell surface-associated proteins like pro-uPA and protease-activated receptor 2 (11). Matriptase has also been suggested to participate in the control of intestinal epithelial turnover by regulating the cell-substratum adhesion (12). Hepsin has been involved in mammalian cell growth, developmental processes such as blastocyst hatching, and initiation of blood coagulation (13)(14)(15). Corin, a TTSP family member isolated from human heart, has been found to act as an * This work was supported by Grant SAF00-0217 from Comisión Interministerial de Ciencia y Tecnología-Spain and European Union (QLG1-CT-2000-01131). The Instituto Universitario de Oncología was supported by Obra Social Cajastur-Asturias. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  in vitro activator of pro-atrial natriuretic peptide, a cardiac hormone essential for the regulation of blood pressure (16,17). HAT, originally isolated from the sputum of patients with chronic airway diseases, may be involved in the host defense system on the mucous membrane (18). Recently, a HAT-related protease isolated from rat tissues has been found to cleave pro-␥-melanotropin at the adrenal cortex, stimulating the mitogenic actions of this peptide (19). Spinesin, predominantly expressed at synapses, may play specific roles in neural functions (20). Finally, insertion of ␤-satellite repeats into the gene encoding TMPRSS3 causes a form of autosomal recessive deafness, suggesting a role for this protease in the development or maintenance of the inner ear or in the turnover of the protein contents of the perilymph and endolymph (21).
The expression of virtually all TTSPs characterized to date is widely deregulated during the development and progression of tumor processes. Thus, matriptase/MT-SP1 was originally identified in breast cancer cells and is highly expressed in breast, prostate and colorectal cancers (22)(23)(24)(25). Inhibition of this protease abolishes both primary tumor growth and metastasis in a murine model of prostate cancer (24,26), whereas stabilization of active matriptase through glycosylation by N-acetylglucosaminyltransferase V is associated with the prometastatic effects of this enzyme (27). Hepsin is overexpressed in ovarian and prostate carcinomas (28 -31), and its expression correlates inversely with measures of patient prognosis (32). Epitheliasin is also overexpressed in prostate carcinomas, and a mutated form of this protease has been found in a case of aggressive disease (33)(34)(35). TMPRSS3/TADG-12 is overexpressed in ovarian cancer (36), and TMPRSS4 is overexpressed in pancreatic cancer (37). Finally, the recently described DESC1 was identified as a consequence of its differential expression in squamous cell carcinoma of the head and neck (6).
These recent findings have stimulated the search for new TTSPs potentially associated with some of the proteolytically mediated processes taking place during normal or pathological conditions and especially during tumor progression. In this work, and as part of our studies on tumor proteinases, we have examined the possibility that additional members of this family of membrane proteinases could be produced by human tissues, with the discovery of a novel family member named matriptase-2. We describe the molecular cloning and complete nucleotide sequence of a cDNA coding for this protein and report an analysis of its expression in human tissues. We also report the production of recombinant matriptase-2 in Escherichia coli and perform an analysis of its enzymatic activity against synthetic and endogenous substrates. Finally, we demonstrate that matriptase-2 is bound to the cell membrane.

EXPERIMENTAL PROCEDURES
Materials-A human fetal liver cDNA library constructed in DR2 and Northern blots containing polyadenylated RNAs from different adult and fetal human tissues were from CLONTECH (Palo Alto, CA). Chemicals, reagents, and synthetic and macromolecular substrates for proteases, including fibronectin, fibrinogen, laminin, type I collagen, and type I gelatin, were obtained from Sigma. Single-chain uPA was purchased from Oncogene Research Products (Boston, MA). Recombinant gelatinases A and B were kindly provided by Dr. G. Murphy (University of East Anglia, Norwich, UK). Recombinant collagenase-3 (MMP-13) was produced as described previously (38). Restriction endonucleases and other reagents used for molecular cloning were from Roche Molecular Biochemicals. Synthetic oligonucleotides were prepared in an Applied Biosystems (Foster City, CA) model 392A DNA synthesizer. Double-stranded DNA probes were radiolabeled with [␣-32 P]dCTP (3000 Ci/mmol) purchased from Amersham Biosciences using a random-priming kit from the same company.
Bioinformatic Screening of the Human Genome and cDNA Cloning-The BLAST program was used to search public (www.ncbi.nlm.nih.gov) and private (www.celera.com) human genome data bases, looking for regions with sequence similarity to previously described TTSPs. After identification of a DNA contig in chromosome 22 encoding a region similar to the catalytic domain of matriptase/MT-SP1, we analyzed in this contig the possible presence of regions encoding the remaining domains characteristic of TTSPs. This approach allowed us to identify DNA regions potentially encoding a full-length sequence for a novel member of this family of serine proteinases. Then we designed specific oligonucleotides to PCR amplify a cDNA for this protein, using a panel of commercially available cDNA libraries (CLONTECH) and the Expand TM High Fidelity PCR system (Roche Molecular Biochemicals). The following oligonucleotides were used: matriptase-2f, 5Ј-AGGATGC-CCGTGGCCGAGGC, and matriptase-2r, 5Ј-AGGTGGGCCC TGCTTT-GCAG. All of the PCRs were performed in a GeneAmp 2400 PCR system from PerkinElmer Life Sciences for 40 cycles of denaturation (94°C, 15 s), annealing (64°C, 15 s), and extension (68°C, 60 s). After cloning of the amplified PCR products in pBSII, their identities were confirmed by nucleotide sequencing.
Nucleotide Sequence Analysis-DNA fragments of interest were sequenced by using the kit DR terminator TaqFS and the automatic DNA sequencer ABI-PRISM 310 (PerkinElmer Life Sciences). All of the nucleotides were identified in both strands. Computer analysis of DNA and protein sequences was performed with the GCG software package of the University of Wisconsin Genetics Computer Group.
Membrane Immunolocalization-Full-length matriptase-2 cDNA was subcloned into pcDNA3 expression vector. In addition, a 24-bp linker coding for the hemagglutinin (HA) epitope of human influenza virus was inserted at the end of the cDNA sequence encoding the C-terminal region of matriptase-2. COS-7 cells were transfected with 1 g of plasmid DNA (pcDNA3-matriptase-2-HA), using FuGENE 6 reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. About 48 h after transfection, the cells were fixed for 10 min in cold 4% paraformaldehyde in PBS, washed in PBS and incubated for 5 min in 0.2% Triton X-100 in PBS. Fluorescent detection was performed by incubating the slides with monoclonal antibody 12CA5 (Roche Molecular Biochemicals) against HA, followed by another incubation with goat anti-mouse fluoresceinated antibody. After washing in PBS, the slides were mounted and observed under fluorescence with a Zeiss axiophot equipped with a CCD camera (Photometrics).

Preparation of Cell Membrane Fractions and Western Blot
Analysis-COS-7 cells were transiently transfected with the pcDNA3matriptase-2-HA plasmid as described previously. The cells were scraped from the plates, and the membrane fractions were prepared essentially following the procedure described by Strongin et al. (39). The extracts were separated by SDS-PAGE, analyzed by Western blotting with an anti-HA monoclonal antibody, and detected with an enhanced chemiluminescence kit (Amersham Biosciences).
Expression and Purification of the Protease Domain of Human Matriptase-2-A 708-bp fragment of the matriptase-2 cDNA containing the entire catalytic domain was generated by PCR amplification with primers 5Ј-ATTGTTGGTGGAGCTGTGTCC and 5Ј-TCAGGTCACCAC-TTGCTGGAT using the full-length matriptase-2 cDNA in pBSII as template. The PCR amplification was performed for 25 cycles of denaturation (95°C, 10 s), annealing (55°C, 10 s), and extension (68°C, 1 min) using the Expand TM High Fidelity PCR system. The PCR-amplified product was then phosphorylated and ligated in the SmaI site of the PGEX-3X expression vector (Amersham Biosciences). The resulting expression vector was transformed into BL21(DE3)pLysS competent E. coli cells, and expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside (final concentration, 0.5 mM) followed by further incubation for 4 -6 h at 30°C. The cells were collected by centrifugation, washed, and resuspended in 0.05 volumes of PBS. Finally, the cells were lysed by sonication and centrifuged at 20,000 ϫ g for 20 min at 4°C. The soluble extract was treated with glutathione-Sepharose 4B (Amersham Biosciences), and the glutathione S-transferase (GST)-matriptase-2 fusion protein eluted with glutathione elution buffer (10 mM reduced glutathione in 50 mM Tris-HCl, pH 8.0), following the manufacturer's instructions. Finally, the purified GSTmatriptase-2 fusion protein was used for enzymatic assays.
Enzymatic Assays-Enzymatic activity of purified recombinant matriptase-2 was detected using the synthetic fluorescent substrates N-t-Boc-Gln-Ala-Arg-AMC, N-t-Boc-Gln-Gly-Arg-AMC, N-t-Boc-Ala-Pro-Ala-AMC, N-t-Boc-Val-Leu-Lys-AMC, and N-t-Boc-Ala-Phe-Lys-AMC. Routine assays were performed at 37°C at substrate concentrations of 100 M in an assay buffer of 50 mM Tris-HCl, 20 mM NaCl, pH 8.0, with a final concentration of Me 2 SO of 2.5%. The fluorometric measurements were made in a MPF-44A PerkinElmer spectrofluorometer ( ex ϭ 360 nm, em ϭ 460 nm). For the inhibition assays, matriptase-2 and inhibitors were preincubated for 15 min at 37°C, and then the incubations were performed at the same conditions as above with the different inhibitors. Cleavage of type I collagen, type I gelatin, type I laminin, fibronectin, and fibrinogen by recombinant matriptase-2 was followed by SDS-PAGE. All of the assays were performed in the above described assay buffer for 4 -12 h at 37°C. The experiments using type I collagen as a substrate were also performed at 28°C. The enzyme/substrate ratio (w/w) used in these experiments was 1/100. Finally, to examine the activation of other proteinases by matriptase-2, the proenzymes of human uPA, plasmin, gelatinase A, and gelatinase B were incubated with active matriptase-2 at a 1:100 w/w enzyme/substrate ratio. The incubations were performed in assay buffer without Me 2 SO for 4 -12 h at 37°C. The processing of the different precursor proteins was assessed by SDS-PAGE.
Homology Modeling-A three-dimensional model of the catalytic domain of matriptase-2 was calculated using Swiss-Model, a semiautomated modeling server (40), and analyzed with the Swiss-Pdb Viewer. Briefly, the amino acid sequence of the predicted catalytic domain of matriptase-2 was compared with the sequences of the macromolecules deposited in the Protein Data Bank to identify suitable templates. After ranking the possible templates by sequence similarity to matriptase-2, structural quality, and nonredundancy, we chose that of human matriptase. The template accession entry is 1EAX. The target sequence was automatically threaded over the structure of the template, built with ProMod II, and energy-minimized with Gromos96. The quality of the resulting models was verified automatically with WhatCheck and manually with Swiss-Pdb Viewer. Electrostatic analyses of the model were performed with MolMol (41). We added hydrogens and utilized a protein dielectric constant of 3. Partial atomic charges were taken from the Amber94 force field. The figures were modeled with MolMol and rendered with Megapov and POV-Ray (www.povray.org/).
Northern Blot Analysis-Nylon filters containing 2 g of poly(A) ϩ RNA of a wide variety of human tissues were prehybridized at 42°C for 3 h in 50% formamide, 5ϫ SSPE (1ϫ SSPE ϭ 150 mM NaCl, 10 mM NaH 2 PO 4 , 1 mM EDTA, pH 7.4), 10ϫ Denhardt's solution, 2% SDS, and 100 g/ml of denatured herring sperm DNA and then hybridized with a radiolabeled matriptase-2-specific probe 2.2 kb long, containing nucleotides 220 -2420 of the isolated cDNA. Hybridization was performed for 20 h under the same conditions used for prehybridization. The filters were washed with 0.1ϫ SSC, 0.1% SDS for 2 h at 50°C and exposed to autoradiography. RNA integrity and equal loading was assessed by hybridization with an actin probe.

Identification and Characterization of a Human Fetal Liver cDNA Encoding a New Membrane-bound Serine Proteinase-
To identify novel members of the TTSP family of membrane serine proteinases, we used the BLAST algorithm to screen the human genome data bases looking for DNA contigs containing sequences with significant similarity to previously described family members. This search led us to the identification of a contig located in chromosome 22 containing coding information for a putative new serine proteinase with a type II transmembrane domain. To generate a cDNA clone for this gene, we performed PCRs using a panel of human cDNA libraries, and specific oligonucleotides derived from the identified genomic sequence. A fragment of the expected size (ϳ2.5 kb) containing in-frame initiator and stop codons was amplified from a cDNA library prepared from human fetal liver. After cloning and sequencing the PCR-amplified product, we confirmed by conceptual translation that the generated sequence was distinct from that reported for all previously identified human TTSPs (Fig. 1A). Computer analysis of the full-length cDNA sequence revealed that it codes for a protein of 802 amino acids, with a calculated molecular weight of 88,901 and showing significant sequence similarity to all other human membrane serine proteinases belonging to the TTSP family. Further analysis of the predicted sequence indicated that the maximum percentage of identities (35%) was with matriptase/MT-SP1 and spinesin. The percentage of identities with other members of the TTSP family, such as TMPRSS2, TMPRSS3, TMPRSS4, DESC1, corin, MSPL, and HAT, was of about 30%.
An alignment of the deduced amino acid sequence from the isolated liver cDNA confirmed that the identified protein possesses all domains characteristic of TTSPs (Fig. 1B). Thus, close to the initiator methionine residue there is a hydrophobic domain spanning from positions 44 -66 that is not preceded by a recognizable signal sequence. Computer analysis using the TMHMM (transmembrane helices Markov model) program (available at www.cbs.dtu.dk) revealed that this domain is predicted to act as a type II membrane anchor sequence. The transmembrane domain is followed by a stem region containing two CUB domains (the first one substantially degenerate with respect to the consensus CUB) and three LDLR repeats. This stem region is very similar to that present in matriptase, which contains two CUB and four LDLR repeats. Finally, there is a catalytic domain located at the C-terminal region of the identified protein and showing about 45% identities with the equivalent domain of matriptase. This catalytic domain also contains all structural hallmarks of functional serine proteinases (Fig.  2). In fact, an alignment of this sequence with that of other members of this class of proteolytic enzymes allows identification of a prodomain region ending in a conserved Arg-Ile-Val-Gly-Gly motif that is highly conserved in serine proteinases and that contains the Arg-Ile bond that is cleaved for protease activation (Fig. 2). The sequence alignment also allows identification of the active site Ser residue as that present at position 753 within the conserved motif Gly-Asp-Ser-Gly-Gly. The His and Asp residues necessary for catalytic activity should be those present at positions 608 and 659, respectively. The sequence Ser-Trp-Gly proposed to interact with the side chains of serine proteinase substrates for proper orientation of the scissile bond is also present at the C-terminal region of the identified sequence (positions 773-775). The putative catalytic domain of this protein also contains the six conserved cysteine residues present in the catalytic region of TTSP family members and involved in the formation of three disulfide bonds (Cys 593 -Cys 609 , Cys 724 -Cys 738 , and Cys 749 -Cys 778 ). A fourth predicted disulfide bridge should form between Cys 679 of the catalytic domain and Cys 559 located at the prodomain. The formation of this predicted disulfide bond observed in many serine proteinases including matriptase would suggest that the active catalytic domain of the identified protein should still remain located at the cell surface even after cleavage at the activation site. Finally, analysis of consensus motifs present in the identified sequence revealed the presence of a potential phosphorylation site in the cytoplasmic tail (Ser-Lys-Arg at position 34) that may participate in the recruitment of intracellular proteins potentially involved in activation of signal transduction pathways. This analysis also revealed six potential sites of N-glycosylation (Asn-Xaa-Thr or Asn-Xaa-Ser) at positions 127, 175, 329, 424, 444, and 509. All of these structural features are also conserved in the amino acid sequence deduced from a mouse cDNA isolated as part of a large scale cDNA sequencing project (42). The protein encoded by this mouse cDNA (accession numbers AK004939 and BAB23684) likely corresponds to the mouse ortholog of the human serine proteinase identified in this work (Fig. 2). The percentages of identities between the human and mouse enzymes were 85.1% in nucleotides and 84.3% in amino acids.
In summary, and according to this structural analysis, we can conclude that the cloned human cDNA encodes a novel membrane serine proteinase that, according to its significant sequence similarity and overall domain organization with matriptase, we propose to call matriptase-2.
Membrane Localization of Matriptase-2-To provide experimental support to the proposal that matriptase-2 is a membrane-bound protease, we transfected COS-7 cells with pcDNA3-matriptase-2-HA, a construct containing the HA epitope at the C-terminal region of the enzyme. Transfected cells were then analyzed by immunofluorescence with a mouse monoclonal antibody (12CA5) specific for this viral epitope. As shown in Fig. 3A, a fluorescent pattern surrounding the cell membrane was clearly visualized in transfected cells expressing matriptase-2. Furthermore, we performed SDS-PAGE analysis of lysates from COS-7 cells transfected with the matriptase-2-HA construct, followed by Western blotting detection with anti-HA monoclonal antibody. As can be seen in Fig.  3B, matriptase-2 was detected in the membrane-enriched fractions but not in the soluble fraction. Taken together, these results provide strong experimental evidence that matriptase-2 is a membrane-anchored protein.
Production of Recombinant Matriptase-2 in E. coli and Analysis of Its Enzymatic Properties-To analyze the enzymatic activity of matriptase-2, we first expressed the catalytic domain of this mosaic membrane-bound protein in bacterial cells. To this purpose, a cDNA coding for the catalytic region was subcloned into the expression vector pGEX-3X, and the resulting plasmid was transformed into the E. coli bacterial strain BL21(DE3)pLysS. After induction of transformed bacteria with isopropyl-1-thio-␤-D-galactopyranoside, a protein band of the expected size (52 kDa) was detected by SDS-PAGE analysis of bacterial protein extracts (Fig. 4). This recombinant fusion protein was purified using glutathione-Sepharose chromatography as described previously (43) (Fig. 4). The soluble GSTmatriptase-2 fusion protein eluted from the affinity column was directly used for enzymatic analysis. We did not release the catalytic domain of matriptase-2 with Factor Xa, because any putative trace of this factor that could remain after the purification process would interfere with assays to analyze the enzymatic activity of a protein such as matriptase-2, which belongs to the same class of proteolytic enzymes as Factor Xa. However, and similar to the case of a fusion protein containing the catalytic domain of matriptase (26), the GST-matriptase-2 fusion protein was apparently autoactivated after incubation at 37°C in the course of the different activity assays. Thus, SDS-PAGE analysis of the recombinant fusion protein incubated for 12 h at 37°C showed the presence of an additional band of 26 kDa that likely corresponds to the catalytic domain of matriptase-2 after proteolytic release of the GST-moiety (data not shown). To analyze the substrate specificity of this recombinant matriptase-2, a series of synthetic quenched fluorescent peptides commonly used for assaying serine proteinases were used. As can be seen in Fig. 5A, the peptides N-t-Boc-Gln-Ala-Arg-AMC and N-t-Boc-Gln-Gly-Arg-AMC were hydrolyzed by matriptase-2 (12.1 and 18.3 nM AMC/min, respectively). Other peptides, including N-t-Boc-Ala-Phe-Lys-AMC, N-t-Boc-Ala-Pro-Ala-AMC, and N-t-Boc-Val-Leu-Lys-AMC, were not signif- M. The hydrolytic activity of matriptase-2 against this substrate was substantially abolished by a number of inhibitors of serine proteinases (phenylmethylsulfonyl fluoride, 4-(2-aminoethyl)-benzenesulfonyl fluoride, leupeptin, and aprotinin) but not by EDTA or E-64, inhibitors of metallo-and cysteine-proteinases, respectively (Fig. 5B). Tosyl-L-phenylalanine chloromethyl ketone was also a poor inhibitor of matriptase-2 ( Fig.   FIG. 2. Amino acid sequence comparisons of the catalytic domains of matriptase-2 and matriptase. The amino acid sequence of human matriptase-2 identified in this work is shown in bold. The sequences for human and mouse matriptase were extracted from the SwissProt data base, whereas the sequence BAB23684 corresponding to the putative mouse matriptase-2 was deduced from a cDNA sequence reported in (42). The multiple alignment was performed with the PILEUP program of the GCG package. Gaps are indicated by dots. Common residues to all sequences are shaded. The numbering corresponds to the sequence of human matriptase-2.  5B). The substrate specificity of the matriptase-2 catalytic domain is consistent with structural features of this proteinase. In fact, it is well established that the S1 specificity of serine proteinases is largely determined by the residue located 6 amino acids N-terminal to the active site Ser residue. This position is occupied by an Asp residue in matriptase-2, pointing to a cleavage specificity for substrates with Arg/Lys at the P1 position. The presence of a polar Gln residue (position 802) very close to the catalytic Ser is also consistent with specificity for basic residues.
After these results demonstrating that the isolated cDNA for matriptase-2 encodes a protein whose catalytic domain exhibits an enzymatic activity against synthetic peptides and an inhibitory profile characteristic of serine proteinases, we evaluated the possibility that matriptase-2 could also degrade a series of extracellular matrix and basement membrane protein components. To do this, a variety of proteins including type I collagen, laminin, fibronectin, fibrinogen, and gelatin were treated with purified recombinant matriptase-2, and the reactions were followed by SDS-PAGE analysis. As can be seen in Fig. 6, this enzyme was able to degrade protein substrates like fibronectin, fibrinogen, and type I collagen. In relation to the degrading activity of matriptase-2 on type I collagen, it is remarkable that this enzyme is not a bona fide fibrillar collagenase because its collagenolytic activity did not generate the typical 3 ⁄4 and 1 ⁄4 fragments produced by fibrillar collagenases, as illustrated for human collagenase-3 in the comparative analysis performed under the same assay conditions (Fig. 6). Also in support of this conclusion, experiments of type I collagen hydrolysis with recombinant matriptase-2 at 28°C revealed the absence of significant degrading activity on this substrate at this temperature (not shown). Therefore, it seems that the matriptase-2 catalytic domain used in this work is only capable of degrading type I collagen at conditions in which the triple helical collagen is partially denatured. The hydrolyzing activity of matriptase-2 on the different macromolecular substrates was blocked in all cases by phenylmethylsulfonyl fluoride (Fig. 6 and data not shown), providing additional support for the proposal that this enzyme is a serine proteinase. Finally, we examined the possibility that matriptase-2 could be implicated in the activation of other proteinases like pro-MMPs, pro-uPA, or plasminogen as part of a coordinate action within a proteolytic cascade. Interestingly, and similar to the case of matriptase, matriptase-2 was able to process, albeit with a low efficiency, the 55-kDa single-chain precursor of uPA generating a 33-kDa protein and other smaller fragments (Fig. 6). By contrast, matriptase-2 was not able to process to a significant extent progelatinase A (MMP-2), progelatinase B (MMP-9), and plasminogen (data not shown).
Homology Modeling of the Catalytic Domain of Matriptase-2-The amino acid sequence similarity between matriptase-2 and matriptase, together with the recent resolution of the three-dimensional structure of the catalytic domain of matriptase (44), opens the possibility to perform a computer modeling of the structure of matriptase-2. Fig. 7A shows an analysis of the molecular surface of the catalytic domain of human matriptase-2 compared with that corresponding to human matriptase. As can be seen, there is a significant degree of similarity between them around their catalytic site, both showing a deep negatively charged S1 pocket and a similar overall charge distribution (Fig. 7A). The occurrence of this deep S1 pocket in both enzymes is consistent with a somewhat relaxed specificity for the catalytic activity of matriptases when compared with other serine proteinases. An additional structural feature shared by matriptase and matriptase-2 may be deduced from the analysis of the loops surrounding the active site cleft and especially of that called "60 insertion loop" (44). Matriptase-2 also shows the presence of this loop, which is similar in length and exhibits a ␤-hairpin conformation similar to that of the corresponding loop present in matriptase and thrombin, although in this latter protein the loop is oriented differently (Fig. 7B) (44,45). The matriptase loop is stabilized through internal hydrogen bonds made by the carboxylate groups of two Asp residues: Asp 60 (A) and Asp 60 (B). Matriptase-2 has Glu and Asp at equivalent positions, suggesting that the stabilization mechanism of its ␤-hairpin loop must be similar to that of matriptase (44). Nevertheless, some interesting structural differences between matriptase and matriptase-2 can be also predicted after analysis of the generated models. First, matriptase has Ser at the position 190 (1EAX numbering), making Lys residues at P1 equally acceptable as Arg. However, matriptase-2 has an Ala at the same FIG. 6. Analysis of enzymatic activity of matriptase-2 on protein substrates. Fibronectin, type I gelatin, type I collagen, laminin, fibrinogen, and pro-uPA were incubated alone (Ϫ) or in the presence of recombinant matriptase-2 (ϩ) in 50 mM Tris-HCl, 20 mM NaCl, pH 8.0, for 12 h at 37°C. Type I collagen was also incubated with human collagenase-3 (MMP-13) in 50 mM Tris-HCl, pH 7.6, 10 mM CaCl 2 , and 100 mM NaCl, as a positive control of collagenolytic activity. In all cases, the enzyme/substrate ratio was of 1/100 (w/w). For inhibition assays, incubations with matriptase-2 were performed in the presence of 2 mM phenylmethylsulfonyl fluoride (PMSF).
position. This substitution suggests that matriptase-2 may preferentially accept substrates with Arg at P1 position. A second noticeable difference is that position 151 is occupied by a Gly residue in matriptase and by an Ile residue in matriptase-2. This allows the accommodation of bulkier P2Ј residues in matriptase than in matriptase-2 (Fig. 7A). Taken together, these structural differences open the possibility of developing selective inhibitors against both matriptases, an aspect that is of future interest because of the potential involvement of these proteases in human diseases, including cancer.
Analysis of Matriptase-2 Distribution in Human Tissues-To investigate the presence of matriptase-2 mRNA transcripts in human tissues, Northern blots containing poly(A) ϩ RNAs prepared from a variety of fetal tissues (brain, lung, liver, and kidney) and adult tissues (leukocytes, colon, small intestine, ovary, testis, prostate, thymus, spleen, pancreas, kidney, skeletal muscle, liver, lung, placenta, brain, and heart) were hybridized with the full-length cDNA isolated for matriptase-2. As shown in Fig. 8, a transcript of about 3.5 kb was exclusively detected in liver. The restricted expression of matriptase-2 is consistent with previous data indicating that most TTSP genes show a highly restricted expression pattern, suggesting that they may have tissue-specific functions (4). Thus, enteropeptidase expression is restricted to enterocytes of the proximal small intestine (9), corin is predominantly produced by cardiac myocytes (16), HAT is mainly expressed in trachea (18), matriptase in the gastrointestinal tract and prostate (26), TM-PRSS2 in prostate and colon (46), hepsin in liver and kidney (15), and DESC1 is an epithelial-specific protease, being fundamentally detected in testes and prostate (6). DISCUSSION Over the last years, there has been an increasing interest in the characterization of proteolytic processes localized at the cell surface (47,48). Most studies on membrane-associated proteolytic systems have focused on metalloproteinases, but very recently, an emerging family of membrane-bound serine proteinases known as TTSPs has received considerable attention because of their potential role in multiple normal and pathological conditions (4). In this work, we describe the finding of a new human serine protease belonging to this family, which has been tentatively called matriptase-2, to emphasize its relationship with matriptase, a matrix-degrading TTSP originally described in human breast carcinoma cells (23), although there are also clear structural and enzymatic differences between both enzymes. The strategy followed to identify matriptase-2 was based on a computer search of the presently available human genome sequences, looking for regions with similarity to previously characterized TTSP family members. After identification of a DNA sequence presumably encoding the catalytic domain of a new TTSP and PCR amplification experiments using fetal liver cDNA as template, a full-length cDNA coding for human matriptase-2 was finally isolated and characterized. Structural analysis of the identified sequence shows the presence of a series of protein domains characteristic of TTSP proteins, including a short cytoplasmic domain, a type II transmembrane sequence, a central region with several modular structural domains, and a C-terminal catalytic domain with all of the typical features of serine proteinases.
Consistent with its structural characteristics, immunofluorescence and Western blot analysis of COS-7 cells transfected with the isolated cDNA confirmed that matriptase-2 is anchored to the cell surface. In addition, functional analysis of the recombinant catalytic domain of matriptase-2 produced in E. coli provided additional evidence that the isolated cDNA codes for a catalytically active serine proteinase. In fact, the purified recombinant protein exhibits a significant proteolytic activity against fluorogenic substrates used for assaying the enzymatic properties of this class of proteinases. In addition, this degrading activity was abolished by inhibitors of serine proteinases but not by inhibitors of any other class of proteolytic enzymes. The substrate specificity of matriptase-2 against synthetic peptides is similar to that of matriptase, with N-t-Boc-Gln-Gly-Arg-AMC and N-t-Boc-Gln-Ala-Arg-AMC being the preferred substrates for matriptase-2 and matriptase, respectively (Ref. 49 and this work). Matriptase-2 also shares with matriptase the ability to degrade extracellular matrix components, suggesting that this novel protease may participate in some of the matrix-degrading processes occurring in both normal and pathological conditions, including cancer pro- gression. Likewise, the finding that matriptase-2, as matriptase, may activate single-chain uPA suggests that it could act as an initiator of the biologically important proteolytic cascades mediated by activated uPA. Nevertheless, it should be emphasized that matriptase-2 shows very limited uPA activating properties when compared with the rapid and potent activity of matriptase in this regard (11), thereby raising doubts about the in vivo relevance of matriptase-2 as a biological activator of uPA. On the other hand, the observation that matriptase-2 is a fibrinolysin opens the possibility that this enzyme may play a role in processes involving fibrin formation and degradation, such as angiogenesis, in a way similar to that proposed for other membrane-bound proteases including MT-MMPs (50). These findings also raise the possibility that members of the TTSP family of membrane-bound proteases could be part of the alternate proteolytic systems that allow cells to infiltrate fibrin matrices via a plasminogen-independent process in diverse physiological and pathological conditions (51). In any case, we would like to remark that these preliminary enzymatic studies performed with the bacterially produced catalytic domain of matriptase-2 do not likely reflect the optimal conditions of in vivo activity of this enzyme. The recombinant protein shows a low specific activity and lacks the ancillary noncatalytic domains that can strongly influence the substrate specificity and catalytic activity of TTSPs. Accordingly, further studies with full-length matriptase-2 produced in eukaryotic expression systems will be required to provide additional information about the nature of the diverse macromolecular substrates presumably targeted by this enzyme.
To further characterize the structure of the catalytic domain of matriptase-2, we performed a homology model for this protease domain. The predicted structure is very similar to that of the catalytic domain of matriptase (44), although the observed structural differences between both proteases could serve to guide the search for specific inhibitors of each enzyme (52). Besides the overall similarities between matriptase and matriptase-2, it is remarkable that this enzyme presents some characteristic features when compared with other TTSP family members. First, the number and organization of the modular repeats present in the stem region are unique to matriptase-2 among TTSPs, although they are similar to those found at the equivalent region of matriptase. Thus, matriptase-2 contains a total of five modular domains, two CUB domains, and three LDLR repeats, whereas matriptase also contains two CUB repeats but possesses one additional LDLR repeat. There are two TTSPs, corin and enteropeptidase, that exhibit a much more complex organization than matriptases in this region. Thus, corin contains 11 modular domains in its stem region, including eight LDLR repeats, two frizzled domains, and one scavenger receptor domain (15). Likewise, enteropeptidase contains two LDLR repeats, two CUB domains, one disulfide knotted domain, one MAM (meprin, A5 antigen, and receptor protein phosphatase ) domain, and one scavenger receptor domain (9). Other than corin and enteropeptidase, all of the remaining human TTSPs characterized to date exhibit a simpler structural organization than matriptase and matriptase-2 and only contain one or two modular repeats in their respective stem regions or even none of them, as is the case for hepsin (15). The functional role of the CUB and LDLR repeats of matriptase-2 is presently unknown, although they can be involved in mediating protein-protein or protein-ligand interactions as proposed for other proteins containing these modules (53,54). Another peculiarity of matriptase-2 is that the gene encoding this proteinase maps to chromosome 22q12-13, a location unique among all TTSP genes identified to date. Interestingly, several TTSP genes lie on the long arm of human chromosome 11; spinesin, TMPRSS4, and MSPL genes are clustered in 11q23, whereas the gene for matriptase is located at 11q25. Similarly, three TTSP genes are located on chromosome 21, TMPRSS2 and TMPRSS3 genes are located at 21q22, and the gene for enteropeptidase is located at 21p11. Likewise, three TTSP genes are located at chromosome 4, HAT and DESC1 are located at 4q13, and corin is located at 4p12. Finally, the hepsin gene maps to 19q13 in a region containing several genes encoding serine proteinases (55). It is worthwhile mentioning that the region containing the matriptase-2 gene is frequently altered in several human tumors, such as insulinomas, ependymomas, and breast and colorectal carcinomas (56 -59). Genetic lesions in the 22q13 region have been also linked to diverse diseases including schizophrenia susceptibility (60). It will be of future interest to examine the possibility that the matriptase-2 gene could be a direct target of some of these genetic abnormalities.
Finally, in this work, as a step to try to define the physiological role of matriptase-2, we have examined the distribution of this protein in human tissues. Similar to the case of most TTSPs, matriptase-2 expression in normal tissues is very restricted, being only detected in significant amounts in fetal and adult liver. This finding suggests a role for this enzyme in some of the matrix-remodeling processes occurring in this tissue during development or in adult life as proposed for other proteolytic enzymes overexpressed in this tissue (61,62). These putative physiological roles for matriptase-2 in liver may also imply the possibility that their potential substrates could be something other than extracellular matrix components. In support of this proposal, several studies have provided evidence of the existence of multiple and distinct substrates for other TTSP family members (4). Furthermore, the above mentioned structural peculiarities of matriptase-2, when compared with other TTSP proteins, could also be consistent with distinct catalytic properties for this novel enzyme. Finally, the identification of the putative murine ortholog of human matriptase-2 raises the possibility of generating mice deficient in this gene, as recently described for matriptase (63). These mutant mice could contribute to clarify the role of matriptase-2 in physiological processes.