We report the characterization of a novel serine protease of the chymotrypsin family, recently isolated by cDNA-representational difference analysis, as a gene overexpressed in pancreatic cancer. The 2.3-kb mRNA of the gene, named TMPRSS3, is strongly expressed in a subset of pancreatic cancer and various other cancer tissues, and its expression correlates with the metastatic potential of the clonal SUIT-2 pancreatic cancer cell lines. The deduced polypeptide sequence consists of 437 amino acids and exhibits all of the structural features characteristic of serine proteases with trypsin-like activity. TMPRSS3 is membrane bound with a NH2-terminal signal-anchor sequence and a glycosylated extracellular region containing the serine protease domain. Thus,TMPRSS3 is a novel membrane-bound serine protease overexpressed in cancer, which may be of importance for processes involved in metastasis formation and tumor invasion.
Proteases have been increasingly recognized as important factors in the pathophysiology of tumorous diseases. The proteolytic degradation of the extracellular matrix, which is an indispensable step in tumor invasion and metastasis, is mediated by members of the four major classes of endopeptidases, including serine, cysteine, aspartyl,and metalloproteases (1). In this highly complicated process, a cascade of events requiring a variety of proteases seems to be involved. Numerous reports have demonstrated an increased production of extracellular matrix degrading enzymes, including type IV collagenase (MMP-2), cathepsin B, cathepsin D, and serine proteases such as plasminogen activator in tumor cells (1). The proteolytic enzymes of the serine protease family exist as single-chain or double-chain zymogens activated by specific and limited proteolytic cleavage. They contain the three active-site amino acids histidine,aspartate, and serine, which participate in peptide bond hydrolysis. The geometric orientation of this catalytic triad is similar in different serine proteases, despite the fact that folding of the proteases may be different (2).
In the present study, we report the cloning and characterization of a novel serine protease identified in a recent cDNA-RDA4approach (3). This study was designed to isolate gene fragments highly overexpressed in pancreatic cancer compared with normal pancreas and chronic pancreatitis tissue. From the 16 gene fragments isolated in this study, we selected the 313-bp gene fragment RDA12 (GenBank accession no. U54603) for further characterization. Database comparison revealed a moderate homology to a number of serine proteases, indicating that RDA12 may be a fragment of a novel protease with cancer-specific expression.
Materials and Methods
Human tissue from patients with ductal adenocarcinoma of the pancreas(n = 13), carcinoma tissues of different origin, human pancreatic tissue from organ donors(n = 6), and chronic pancreatitis tissue(n = 6) was provided by the Hungarian Academy of Sciences (Budapest, Hungary) and the Department of Surgery of the University of Ulm. All tissue samples were obtained after approval by the local Ethics Committee.
The human pancreatic cancer cell lines were obtained from the following suppliers: PATU-8988S and PATU-8988T (German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany); PANC-1 and MIA-PaCa-2 (European Collection of Animal Cell Cultures, Salisbury,United Kingdom); HPAF (Metzgar, Durham, NC); Capan-1, Capan-2, and AsPC-1 (Cell Lines Service, Heidelberg, Germany); Patu II(Elsässer, Marburg, Germany); PC2 (Bülow, Mainz, Germany);SUIT-2 (S2-007, S2-013, S2-020, and S2-028; Iwamura, Miyazaki, Japan;Ref. 4); and SKPC2 and IMIM-PC2 (P. Real, IMIM,Barcelona, Spain).
Cloning of a New Serine Protease cDNA.
In a recent screen for differentially expressed genes in pancreatic carcinoma, the 313-bp gene fragment RDA12 (accession no. U54603) was isolated by cDNA-RDA (3); this fragment encodes the putative motif of a new serine protease. The RDA12 fragment was used to screen ∼20,000 clones of an oligo(dT)-primed cDNA library from a pancreatic cancer cell line by hybridization. Both strands of the longest cDNA clone, RDA12/2, were sequenced by primer walking. For stable transfection in mammalian cells, the cDNA clone RDA12/2 was cloned in sense and antisense orientation into the BamHI site of the mammalian expression vector pHβ-Apr1-neo(5). A COOH-terminal-tagged TMPRSS3 expression vector was constructed by insertion of a 1427-bp fragment (nucleotides 96–1522) containing the open reading frame of TMPRSS3into the BstXI site of the mammalian expression vector pcDNA6/V5/His B (Invitrogen, San Diego, CA).
Northern Blot Analyses.
The expression of TMPRSS3 was studied by hybridizations using Northern blots containing 30 μg each of total RNA from normal pancreas tissue, chronic pancreatitis tissue, different carcinoma tissues, and cell lines. The Northern blots containing RNA of different human tissues were purchased from Clontech (Heidelberg, Germany).
Cell Culture and Transfection.
For functional analysis of TMPRSS3, the S2-020 pancreatic cancer cell line, which expresses no endogenous TMPRSS3mRNA, was transfected with the TMPRSS3-pHβ-Apr1-neo construct in sense and antisense orientation using DMRIE-C (Life Technologies, Inc., Eggenstein, Germany). Several clones were picked that showed various degrees of stable TMPRSS3sense/antisense mRNA expression. Two of each sense and antisense clones were used for functional assays.
HEK-293 cells were plated at 1.5 × 106 cells/10-cm dish and grown overnight in DMEM supplemented with 10% FCS. Cells were transiently transfected with the TMPRSS3-pcDNA6/V5/His plasmid DNA by use of the calcium phosphate protocol.
Preparation of Cell Extracts and Subcellular Fractionation.
Forty-eight h after transient transfection with V5-tagged TMPRSS3 into HEK-293 cells, protein extracts were prepared by resuspending pelleted cells in 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 150 mm NaCl, 50 mm Tris-HCl (pH 7.2) supplemented with 5 μg/ml Aprotinin, 5 mm Pefabloc, and 10 μg/ml Pepstatin. For immunopurification of the epitope-tagged protein, cell lysates were incubated with V5 antibody conjugated to protein G-agarose beads at 4°C for 4 h on a shaker. The agarose beads were pelleted by centrifugation and washed twice with 150 mm NaCl, 5 mm EDTA, 50 mm Tris + 0.1% NP40. The washed pellets were resuspended in 150 mm NaCl, 5 mm EDTA, 50 mm Tris + 0.1% NP40 for PNGase F treatment.
Subcellular fractions were prepared from transiently transfected HEK-293 cells as reported previously (6). The plasma membrane-enriched fraction, which was prepared using sucrose density gradient centrifugation, the cytosolic fraction, and concentrated culture medium were studied by Western blot analysis.
For PNGase F treatment, immunopurified protein was incubated overnight with 2 units of PNGase F supplemented with 10 mm EDTA at 37°C. Inhibition of N- and mucin-like O-glycosylation was performed by cultivating TMPRSS3-expressing HEK-293 cells for 24 h in DMEM, 10% FCS containing either 2.5 μg/ml tunicamycin (7) or 2 mmphenyl-N-Acetyl-α-d-galactosaminide(8). Thereafter, cells were harvested for protein extraction.
Nude mouse experiments were done by injecting 2 × 106 S2-020 cells stably transfected with TMPRSS3 sense/antisense constructs, both s.c. and in the tail vein of female nu/nu mice. Five weeks after the tail vein injections, the lung, spleen, and liver were used for standard histological analysis to identify the presence or absence of metastatic lesions. Subcutaneous tumors were measured and used for histological analysis.
In vitro matrigel invasion assays were done by seeding 105 transfected cells in medium + 1%FCS in the upper chamber of Matrigel-coated 8-μm transwell plates. The lower chamber was filled with medium + 10% FCS. The number of invading cells adhering to the lower side of the porous membrane was counted after fixation with 4% paraformaldehyde and staining with methylene blue.
The proteolytic activity in TMPRSS3sense/antisense-transfected S2-020 cells and transiently transfected HEK-293 cells was determined fluorometrically in native lysates and lysates treated with enterokinase for activation, using oligopeptide substrates for elastase-like (Ala-Ala-Ala-Ala) and trypsin-like(Ile-Pro-Arg) serine proteases as described previously(9).
Chromosomal Mapping of the TMPRSS3 Gene Locus.
The chromosomal localization of TMPRSS3 was determined by screening the GeneBridge4 radiation hybrid panel (Research Genetics,Huntsville, AL), using the TMPRSS3-specific primers 5′-CATGTGGTGGGCATCGTTA-3′ and 5′-CCAGTTGAGATAGGCTGAG-3′.
Results and Discussion
The 313-bp fragment encoding the putative motif of a new serine protease isolated in a recent cDNA-RDA screen for genes differentially expressed in pancreatic cancer (3) was used to screen a pancreatic cancer cDNA library. Among 16 isolated homologous clones, a clone designated RDA12/2 contained the full-length sequence. The sequence of clone RDA12/2 comprised 2071 bp, including a 214-bp 5′untranslated region, an open reading frame of 1311 nucleotides, and a 546-bp 3′ untranslated region (Fig. 1). Translation of the open reading frame suggests that the cDNA codes for a putative polypeptide of 437 amino acids with an estimated molecular mass of 48.202 kDa. The NH2-terminal region of the hypothetical protein contains a putative signal-anchor sequence characteristic for group II integral membrane proteins. The highly hydrophobic region of 22 amino acids may serve as a transmembrane domain that is involved in anchoring the protease to the cell membrane. According to the charge difference rule(10), it can be assumed that the COOH terminus of the protein with its protease module is located on the extracellular surface.
Although the nucleotide sequence is unique, database comparisons of the amino acid sequence revealed a homology to a number of serine proteases. Thirty-five percent identity and ∼50% similarity was found to members of the serine protease family known as the human transmembrane proteases, TMPRSS1/hepsin (11) or TMPRSS2 (12). Thus, our new protease is the third member of a family of transmembrane-bound serine proteases. Consequently, this new gene was named TMPRSS3 for transmembrane protease, serine 3. Sequence homology was high in the domains containing the three principal active-site amino acids H245, D290, and S387, required for peptide bond hydrolysis. The arrangement of the catalytic residues in the linear sequence defines the membership of TMPRSS3 to the S1 family of the chymotrypsin clan SA of serine-type peptidases (2). The prototype of this family is chymotrypsin, and the three-dimensional structures of some of its members have already been resolved (12).
TMPRSS3 is predicted to cleave in a trypsin-like manner after lysine or arginine residues because it contains D381 at the base of the specificity pocket that binds the substrate(13). In addition, the novel protein shares considerable structural similarities of the TMPRSS family, including the putative NH2-terminal membrane anchor and the conserved cysteine residues, which by homology most likely form the disulfide bonds C196–C310,C230–C246,C356–C372, and C383–C410. Serine proteases are most commonly synthesized as inactive proenzymes, which are activated by extracellular, proteolytic removal of a propeptide. At the NH2-terminal part of the protease domain, TMPRSS3 contains the peptide sequence RVVGG, which is typical for the proteolytic activator site of many protease zymogens. The potential cleavage between R204 and V205 would result in a new terminal α-amino group, which forms a salt bridge with D386 and thereby leads to the assembly of the functional catalytic sites. Therefore, the activated form would consist of a non-protease and a protease subunit linked by a disulfide bond that most likely involves C196–C310. Whether this activation is mediated under physiological conditions by autocatalytic cleavage or other proteases is not known. The TMPRSS3 gene locus was localized to chromosome 11 at q23.3 between the markers D11S4362 and D11S4387 by use of a radiation hybrid panel.
As anticipated, an overexpression of the 2.3-kb transcript was found in 9 of 13 primary pancreatic carcinoma tissues (Fig. 2) and in 10 of 16 pancreatic carcinoma cell lines (not shown) by Northern blot analysis. Because TMPRSS3 was not expressed in normal pancreas (n = 6) and in chronic pancreatitis (n = 6) tissue samples,overexpression appears to be cancer-specific and not due to inflammatory alterations in the stroma. No clear correlation was found between the stage of pancreatic tumors and the expression of the protease (Table 1). Northern blot analyses with RNA from a small number of other tumor tissues revealed that TMPRSS3 overexpression is not restricted to pancreatic cancer, but can also be found in gastric(n = 4), colorectal (n = 7), and ampullary (n = 1) cancer. No expression was found in one tissue sample each of soft tissue sarcoma and breast cancer (Fig. 2). TMPRSS3 transcripts were not detectable in normal heart, brain, placenta, lung, liver, skeletal muscle, uterus, and adipose tissue. A weak signal was found in tissues of the normal gastrointestinal tract (esophagus, stomach, small intestine, colon) and in some tissues of the urogenital tract (kidney and bladder). Nevertheless, expression was much weaker than in the corresponding tumors (data not shown). Furthermore, we analyzed the expression of TMPRSS3 in the SUIT-2 clonal cell lines S2-007, S2-013, and S2-028 (4). These subclones of the human pancreatic cancer cell line SUIT-2 differ in their spontaneous metastatic potential after s.c. injection in nude mice. In this setting S2-007 regularly shows a high rate of metastases, whereas the other two cell lines show a lower rate (S2-013) or no metastases at all (S2-028). As shown in Fig. 2, the strength of TMPRSS3 expression correlated well to the metastatic potential of the SUIT-2 subclones,which may serve as an indication that this serine protease is associated with the promotion of metastasis.
The sequence of TMPRSS3 suggests that this novel serine protease contains a signal anchor characteristic for group II integral membrane proteins with a hydrophobic transmembrane domain (Fig. 3,a). According to the charge difference rule(10), the transmembrane domain (amino acids 32–53)anchors the protease to the cell membrane. Because of this anchorage,the NH2-terminal domain (amino acids 1–31) would appear to be located intracellularly, and the COOH-terminal region(amino acids 54–437), which contains the catalytic domain, would be located extracellularly (Fig. 3,b). The alleged subcellular localization of the protease was confirmed using a V5-tagged TMPRSS3 construct, which was transiently transfected into HEK-293 cells. Membrane fractionation and Western blotting with the corresponding anti-V5 antibody revealed a signal only in the plasma membrane-enriched fraction, whereas no tagged TMPRSS3 protein was detectable in the cytosol and in the culture medium (Fig. 4).
This experiment also uncovered post-translational modifications of TMPRSS3. Although the calculated theoretical molecular mass of the epitope-tagged fusion protein is 52 kDa, its size in a SDS-polyacrylamide gel is ∼68 kDa, suggesting the presence of potential carbohydrate moieties. The primary sequence of TMPRSS3 displays two consensus motifs for N-linked glycosylation(N-X-T/S) at N130 and N178. To confirm this N-glycosylation, epitope-tagged TMPRSS3 was expressed in HEK-293 cells, immunoprecipitated, and treated with PNGase F. This resulted in an increase in mobility on denaturing SDS-PAGE,demonstrating N-glycosylation of TMPRSS3 (Fig. 4). Cultivation of transfected HEK-293 cells in the presence of tunicamycin, an inhibitor of N-glycosylation, revealed the same mobility shift of TMPRSS3 to a molecular mass of 60 kDa. Phenyl-N-acetyl-α-d-galactosaminide,which inhibits mucin-like O-glycosylation, had no effect on the molecular mass (data not shown). The generation of recombinant proteases frequently has been shown to be difficult or impossible(14). Despite extensive and repeated efforts, we were unable to successfully generate recombinant protein in Escherichia coli and insect cells, possibly because TMPRSS3,as many other proteases, had a cytotoxic effect on transfected cells. Repeated efforts to generate peptide antisera failed as well (data not shown), and a TMPRSS3 antibody was therefore not available for further studies.
Whereas the established physiological role of the chymotrypsin family of secreted serine proteases is primarily in protein catabolism, the function of serine proteases of the TMPRSS family is of special interest. Although the function of TMPRSS2 remains unknown (12, 15), TMPRSS1, also known as hepsin, frequently is overexpressed in ovarian tumors and may therefore contribute to the invasive nature or growth capacity of ovarian tumor cells (16). Treatment of hepatoma cells with antihepsin antibodies or specific antisense oligonucleotides confirmed that hepsin plays an essential role in cell growth and maintenance of cell morphology (17). It has also been shown that hepsin can proteolytically activate human coagulation factor VII and thereby contribute to the activation of the coagulation cascade (18).
The correlation of TMPRSS3 expression with the metastatic potential of the SUIT-2 cell lines is a first indication that this new protease, in the same way as hepsin, may be involved in promoting metastasis formation and tumor invasion. To confirm this hypothesis in functional assays, stably transfected S2-020 cell lines were generated using the TMPRSS3 cDNA cloned in sense and antisense orientation into the pHβ-Apr1-neo vector. Several clones were generated showing variable degrees of TMPRSS3sense/antisense mRNA transcription. Two sense and two antisense clones were further characterized by s.c. injections in nude mice, in vitro Matrigel invasion assays, and biochemically for their capacity to hydrolyze substrates for trypsin and elastase. No significant differences could be observed between sense and antisense clones in any of the functional assays. There was no difference in tumor size and local invasiveness after s.c. injections, and there was no evidence of metastasis formation after tail vein injection with both sense and antisense cells. Similarly, we failed to show an effect on in vitro invasiveness and on proteolytic activity of native and enterokinase-treated lysates for a selection of serine protease substrates. Many factors may be responsible for the failure of TMPRSS3-transfected tumor cells to behave differently in these assay, including the necessity for a complex activation mechanism, processes that affect protein folding, or the absence of essential cofactors. Furthermore, although transiently transfected HEK-293 cells showed expression of the V5-tagged recombinant TMPRSS3 protein, we could not directly demonstrate expression of the protein in the transfected cells because we lacked a specific antibody. In the absence of final experimental proof, we can therefore only hypothesize, based on the structural characteristics and the expression pattern in cancer tissues and in the SUIT-2 subclones, that this new protease has a potential role for tumor progression, metastasis formation, and tumor invasion.
Proteases have an important function in the context of tumor growth,because they can break down the surrounding extracellular matrix components, they can pave the way for spreading tumor cells, and they can release and activate growth and angiogenic factors. Protease activity on the surface of tumor cells is required to allow malignant invasion through surrounding connective tissue, which is an important event in the multistep process of metastasis formation(19). Thus, it is conceivable that TMPRSS3 may contribute to the invasive and metastatic potential of tumor cells. In this context, cell surface proteases such as TMPRSS3 may function as an activator of other extracellular proteases or act directly by degrading the extracellular matrix surrounding the tumor cells. Furthermore,TMPRSS3, as shown for many other proteases, may participate in the activation of hormones or growth factors by proteolytic cleavage of inactive proforms. Because the biochemical events required for the activation of this novel serine protease are unknown and the specific substrates have not yet been identified, the precise role of TMPRSS3 in carcinogenesis remains to be elucidated.
We thank G. Adler for continual support, U. Lacher for excellent technical assistance, M. A. Hollingsworth for the pHβ-Apr1-neo vector, and F. Gansauge and G. Varga for providing human pancreatic tissue samples.
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.
This work was supported by grants from the Bundesministerium für Bildung und Forschung (01 GB9401), the European Community (BMH4-CT98-3085), and the Deutsche Forschungsgemeinschaft (SFB518, project B1; to T. M. G.).
The nucleotide sequence in this report has been submitted to the GenBank Data Library with accession no. AF179224.
The abbreviations used are: RDA,representational difference analysis; PNGase F,peptide-N-glycosidase F.
|Tissue sample .||TNM classification .|
|Tissue sample .||TNM classification .|