Matriptase-2 is a member of the TTSPs (type II transmembrane serine proteases), an emerging class of cell surface proteases involved in tissue homoeostasis and several human disorders. Matriptase-2 exhibits a domain organization similar to other TTSPs, with a cytoplasmic N-terminus, a transmembrane domain and an extracellular C-terminus containing the non-catalytic stem region and the protease domain. To gain further insight into the biochemical functions of matriptase-2, we characterized the subcellular localization of the monomeric and multimeric form and identified cell surface shedding as a defining point in its proteolytic processing. Using HEK (human embryonic kidney)-293 cells, stably transfected with cDNA encoding human matriptase-2, we demonstrate a cell membrane localization for the inactive single-chain zymogen. Membrane-associated matriptase-2 is highly N-glycosylated and occurs in monomeric, as well as multimeric, forms covalently linked by disulfide bonds. Furthermore, matriptase-2 undergoes shedding into the conditioned medium as an activated two-chain form containing the catalytic domain, which is cleaved at the canonical activation motif, but is linked to a released portion of the stem region via a conserved disulfide bond. Cleavage sites were identified by MS, sequencing and mutational analysis. Interestingly, cell surface shedding and activation of a matriptase-2 variant bearing a mutation at the active-site serine residue is dependent on the catalytic activity of co-expressed or co-incubated wild-type matriptase-2, indicating a transactivation and trans-shedding mechanism.
Matriptase-2 [also known as TMPRSS6 (transmembrane protease, serine 6)] was identified in 2002  and belongs to the TTSPs (type II transmembrane serine proteases), which represent a family of cell surface proteinases [2–8]. Structurally, this enzyme consists of a short cytoplasmic N-terminal tail, a transmembrane domain, a stem region containing a SEA (sea-urchin sperm protein, enteropeptidase and agrin) domain, two CUB (complement protein subcomponents C1r/C1s, urchin embryonic growth factor and bone morphogenetic protein 1) domains and three LDLRA (low-density-lipoprotein receptor class A) domains, and a C-terminal serine protease domain.
Matriptase-2 shares high similarity with matriptase [also known as MT-SP1 (membrane-type serine protease 1)], which has been extensively characterized. Matriptase is expressed by a number of epithelial cells and is overexpressed in different human cancers [3,9–13]. Substrates of matriptase are proteins located at the cell surface or in the extracellular matrix, and are of major importance in the complex process of tumour invasion and angiogenesis [14–17]. Interestingly, in contrast with matriptase, overexpression of matriptase-2 significantly suppresses breast and prostate cancer growth and reduced levels of matriptase-2 correlate with poor patient outcome [12,13].
Compared with matriptase, matriptase-2 has a more limited expression pattern. TMPRSS6 mRNA can be found predominantly in liver, suggesting tissue-specific functions [1,18]. Indeed, the role of matriptase-2 as a key regulator in iron homoeostasis was identified recently and described in two independent studies [18,19]. Mutations in matriptase-2 were directly linked to the inability to down-regulate high levels of hepcidin [18–20], the systemic regulator of iron homoeostasis, leading to IRIDA (iron-refractory iron deficiency anaemia). Matriptase-2 was described to suppress hepcidin expression [19,20] by cleaving the bone morphogenetic protein co-receptor hemojuvelin .
Matriptase is synthesized as an inactive single-chain membrane-bound polypeptide and undergoes complex proteolytic processing during zymogen activation [22–24]. A cleavage after Gly149, within a conserved motif of the SEA domain, splits matriptase into two fragments . However, this cleaved matriptase appears to remain membrane-associated via non-covalent linkage [23,25,26]. Transactivation, in which the intrinsic weak proteolytic activity from one zymogen molecule activates another, occurs afterwards by proteolytic cleavage between Arg614 and Val615, within the canonical activation motif Arg614-Val-Val-Gly-Gly618 converting the inactive single-chain zymogen into the disulfide-linked two-chain protease (the activated form) [23,27]. As a consequence of a conserved disulfide bond , the catalytic domain of matriptase remains attached to the propeptide after the activation cleavage, representing a general feature of TTSPs [28–30]. Lastly, matriptase is released from the cell surface via ectodomain shedding by proteolytic cleavage after Lys189 or Lys204 in a complex with its cognate inhibitor HAI-1 (hepatocyte growth factor activator inhibitor-1) [22,31].
In contrast with matriptase, the molecular mechanisms involved in the activation and proteolytic processing of matriptase-2 are largely unknown. Recent publications demonstrate that mutations in matriptase-2, affecting regions other than the catalytic domain, of patients with IRIDA may have a profound influence on the activity of the protease [32–34]. On the basis of the presence of a conserved disulfide bond linking the pro- and catalytic domains, the catalytic domain of matriptase-2 might also remain membrane-bound following activation. However, soluble forms of matriptase-2 were detected in the conditioned medium of transfected cells, but were not further specified [21,32,34,35].
In the present study, we demonstrate efficient cell surface shedding of matriptase-2 into the conditioned medium of transfected cells. Importantly, shed matriptase-2 was detected in a two-chain form that was highly active. In contrast, cell-associated matriptase-2 was predominantly present as an inactive single-chain zymogen. We were able to isolate the shed activated two-chain form and identified the cleavage site as being between Arg437 and Val438 (numbering is based on NCBI accession number CAC85953) within the second CUB domain of the non-catalytic stem region. This processing site was confirmed by mutational analysis, which revealed a second cleavage site after Arg404 important for the release of matriptase-2. Furthermore, mutation of the catalytic active Ser753 blocked shedding and activation of matriptase-2, whereas both processing events occurred in the presence of wild-type matriptase-2, indicating a trans-mechanism of enzyme activation and release from the cell surface.
Cell lines and culture conditions
HEK (human embryonic kidney cells)-293 , which do not express endogenous matriptase-2, and HepG2 cells (a human hepatocellular carcinoma cell line), which do express matriptase-2, were cultured in DMEM (Dulbecco's modified Eagle's medium), supplemented with 100 units/ml penicillin, 100 μg/ml streptomycin, 0.2 M glutamine (all from Invitrogen) and 10% (v/v) FBS (fetal bovine serum; PAA Laboratories) under a humidified atmosphere of 5% CO2 at 37 °C.
Cloning and transfection
Total mRNA was isolated from HepG2 cells expressing endogenous matriptase-2 with TRIzol® (Invitrogen) and transcribed into cDNA using the Omniscript reverse transcriptase kit (Qiagen) and oligo(dT) primer (Invitrogen) according to the manufacturer's instructions. For construction of the cloning vector pDrive-matriptase-2, TMPRSS6 cDNA was obtained by PCR using the primers matriptase-2-fw (5′-GCTCCCGGTACCATGCCCGTGGCCGAGGCCC-3′) and matriptase-2-rv (5′-GCTCCCCTCGAGGGTCACCACTTGCTGGATCC-3′) (Acc65I and XhoI recognitions sites are in bold and the TMPRSS6 cDNA is underlined) and ligated into the cloning vector pDrive (Qiagen). This vector was directly used to generate the expression plasmids pcDNA4-matriptase-2-Myc-HisA and pcDNA4-matriptase-2-HisA or it was subjected to site-directed mutagenesis with the Gene Tailor™ site-directed mutagenesis system (Invitrogen) to the manufacturer's instructions using primers S753A-fw (5′-AGGATGCCTGTCAGGGTGACGCAGGTGGTCCG-3′) and S753A-rv (5′-TCACCCTGACAGGCATCCTTCTTGCCCTTG-3′), R437E-fw (5′-CCCTCACCGGGCCCGGTGTGGAGGTGCACTATG-3′) and R437E-rv (5′-CACACCGGGCCCGGTGAGGGAGATCTGGGAG-3′), R404E-fw (5′-CAGGAGGCTGTGTGGCTTGGAGATCCTGCAGC-3′) and R404-rv (5′-CAAGCCACACAGCCTCCTGTTCTGGATCGT-3′) generating pDrivematriptase-2-S753A, pDrive-matriptase-2-R437E and pDrive-matriptase-2-R404E. After their cloning into the Acc65I/XhoI restriction sites of expression plasmids pcDNA4-Myc-HisA (Invitrogen), pcDNA4-HisA or pcDNA4-MycA, the TMPRSS6 cDNA fragments were confirmed by sequencing. The last two mentioned expression vectors were generated by AgeI/PmeI or respective BstBI restriction of pcDNA4-Myc-HisA to remove one of the tags. The resulting expression vectors pcDNA4-matriptase-2-Myc-HisA, pcDNA4-matriptase-2-S753A-Myc-HisA, pcDNA4-matriptase-2-R437E-Myc-HisA, pcDNA4-matriptase-2-R404E-Myc-HisA, pcDNA4-matriptase-2-HisA and pcDNA4-matriptase-2-S753A-MycA encode the unmutated or mutated matriptase-2 protein tagged at the C-terminus with a c-Myc and/or His6. The expression vector pcDNA4-matriptase-2-R404E-R437E-Myc-HisA was generated by restriction digest of pcDNA4-matriptase-2-R437E-Myc-HisA and pcDNA4-matriptase-2-R404E-Myc-His A with BglII followed by ligation.
HEK-293 cells were stably transfected with these pcDNA4-vectors harbouring unmutated or mutated TMPRSS6 cDNA or the empty vector pcDNA4-Myc-HisA by lipofection with Lipofectamine™ 2000 (Invitrogen) according to the manufacturer's instructions to generate cells expressing MT2WT–Myc-His6, MT2S753A–Myc–His6, MT2R437E–Myc–His6, MT2R404E–Myc–His6, MT2R404E/R437E–Myc–His6, MT2WT–His6, MT2S753A–Myc and mock (where MT2 is matriptase-2 and WT is wild-type). At 48 h after transfection, HEK-293 cells were selected with 100 μg/ml Zeocin (Cayla SAS). Single-cell clones were raised and selected for their ability to produce matriptase-2 protein by Western blot analysis as described below. For the generation of co-transfected MT2WT-His6/MT2S753A-Myc cells, singly transfected MT2S753A-Myc cells were transiently transfected with pcDNA4-matriptase-2-HisA. HEK-293 cells transfected with the empty vector (mock cells) were checked by PCR analysis after genomic DNA preparation (using the PureLink™ genomic DNA mini kit; Invitrogen).
Cells expressing MT2WT–Myc–His6, MT2S753A–Myc–His6 and mock cells were cultured on poly-L-lysine-coated coverslips for 2 days. Cells were washed twice with ice-cold PBS and fixed in 4% (w/v) paraformaldehyde in PBS for 30 min at room temperature (25 °C) followed by three washing steps with PBS. For permeabilization, cells were incubated with 0.1% Triton X-100 in PBS for 10 min at room temperature followed by two washing steps with PBS. Unspecific binding sites were blocked by incubation with 5% (w/v) BSA (Carl Roth) in PBS for 1 h at room temperature, followed by incubation with a monoclonal mouse anti-c-Myc antibody (clone 9E10 ; Sigma–Aldrich) diluted 1:1000 in 2.5% (w/v) BSA in PBS for 1 h at room temperature. After three washing steps with PBS, cells were incubated with species-specific Alexa Fluor® 594-conjugated secondary antibodies [goat anti-(mouse IgG); Invitrogen] diluted 1:500 in 2.5% (w/v) BSA in PBS and with 10 ng/ml DAPI (4′,6′-diamidino-2-phenylindole), for nuclear staining, for 1 h at room temperature. This was followed by three washing steps with PBS and a further washing step with distilled water. Cover slips were mounted with Immu-Mount™ (Shandon). For hardening, slides were stored overnight at 4 °C. Fluorescence signals were recorded with a Zeiss Axiovert 200 microscope equipped with a charge-coupled-device camera.
Cell fractionation and protein isolation
Cells expressing MT2WT–Myc–His6, MT2R437E–Myc–His6, MT2R404E–Myc–His6, MT2R404E/R437E–Myc–His6, MT2WT–His6, MT2S753A–Myc, MT2WT–His6/MT2S753A–Myc and mock cells were seeded into 25-cm2 flasks (5×105 cells/cm2) and cultured for 1 day in full medium. The medium was then replaced with FBS-free Opti-MEM® after two washing steps with PBS. Cells were cultured for another 2 days to increase protein concentration. After that, conditioned medium was collected and concentrated using Amicon Ultra-15 centrifugal filters (3000 Da molecular-mass cut-off) and directly used for Western blot analysis and enzyme activity measurements after protein quantification with Roti®-Nanoquant (Carl Roth).
For isolation of cellular membranes, HEK-293 cells were washed three times with PBS and scraped off with 1 ml of ice-cold buffer (20 mM Hepes, pH 7.5, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA and 1 mM EGTA) followed by homogenization via ten aspirations through a 23-gauge needle. Intact cells and nuclei were removed by centrifugation (2000 g for 5 min at 4 °C). The resulting supernatants were centrifuged for 1 h at 21000 g and 4 °C. Pellets containing cellular membranes were resuspended in 50 μl of buffer and the total protein level was quantified. For measurements of proteolytic activity, the pellet fractions were supplemented with 0.5% (v/v) NP-40 (Nonidet P40) to solubilize membrane proteins.
For analysis of the N-glycosylations of matriptase-2, samples were either obtained from insoluble cell fractions or conditioned medium and were treated with PNGase F (N-glycosidase F; New England Biolabs) according to the manufacturer's instructions.
For detection of matriptase-2, 10 μg of total protein from cellular membranes, cytosolic cell fractions and conditioned medium were diluted in non-reducing or reducing [supplemented with 2% (v/v) 2-mercaptoethanol] SDS-loading buffer and separated by SDS/PAGE on 10% (w/v) polyacrylamide gels. Separated proteins were transferred on to nitrocellulose membrane (Schleicher & Schuell). Transfer efficiency was determined by Ponceau S staining. Nitrocellulose membranes were blocked by incubation with 5% (w/v) non-fat dried milk (Carl Roth) in PBST (PBS with 0.2% Tween 20) for 1 h at room temperature. Antigens were detected by incubation with a monoclonal mouse anti-c-Myc antibody (diluted 1:5000) or a monoclonal mouse anti-His6 antibody (diluted 1:5000; Invitrogen) for 1 h at room temperature. After three washing steps with PBST, primary antibody was incubated with species-specific alkaline-phosphatase-conjugated secondary antibody diluted 1:10000 in PBST [goat anti-(mouse IgG); Calbiochem] followed by three washing steps with PBST and one washing step with PBS. The resulting antigen–antibody complexes were detected with NBT/BCIP (Nitro Tetrazolium Blue chloride/5-bromo-4-chloro-3-indolyl phosphate; Sigma–Aldrich).
Purification, MS and N-terminal sequencing of shed matriptase-2
The catalytic domain of matriptase-2 was isolated from the concentrated conditioned medium of cells expressing MT2WT–Myc–His6 cells by immunoaffinity chromatography using the ProFound™ c-Myc tag IP (immunoprecipitation)/co-IP kit (Pierce Biotechnology) according to the manufacturer's instructions. Purity of isolated matriptase-2 was confirmed by SDS/PAGE. Protein bands representing purified matriptase-2 were extracted from reduced SDS gels and analysed by LC (liquid chromatography)-MS/MS by the Bioanalytic Service Centre at the Centre of Molecular Medicine, Cologne, Germany. The N-terminal sequence of the shed form of matriptase-2 was determined after blotting on to PVDF membrane (Schleicher & Schuell) using Edman degradation (Proteome Factory).
Enzyme activity assay and kinetics
Proteolytic activity of matriptase-2 was measured with 1–20 μg of total protein from concentrated conditioned medium and membranes of transfected HEK-293 cells prepared as described above. Activity was assayed according to Kolp et al.  in pre-warmed Tris/saline buffer (50 mM Tris, pH 8.0, and 150 mM NaCl) at 37 °C by monitoring the release of p-nitroaniline from the chromogenic substrate N-Boc (t-butoxycarbonyl)-Gln-Ala-Arg-p-nitroanilide (400 μM; Bachem) at a wavelength of 405 nm for up to 20 min using a Cary 100 UV–visible spectrophotometer (Varian). Inhibition assays were performed either with the metalloprotease inhibitor EDTA (10 mM) or with the serine protease inhibitors aprotinin and leupeptin (Carl Roth). Kinetic analysis was performed using GraFit version 5.0.
RESULTS AND DISCUSSION
Expression of human matriptase-2 in HEK-293 cells
TTSPs are cell surface transmembrane proteins undergoing complex processing events. To examine subcellular localization of human matriptase-2, HEK-293 cells were stably transfected with cDNA encoding human wild-type matriptase-2 (MT2WT–Myc–His6) or the proteolytically inactive variant of matriptase-2 (MT2S753A–Myc–His6). Both expressed constructs contain a C-terminal c-Myc and His6 tag. Immunofluorescence analysis of permeabilized cells expressing MT2WT–Myc–His6 and MT2S753A–Myc–His6 revealed an intracellular localization for matriptase-2 (Figures 1A and 1C). To analyse the membrane localization of matriptase-2, non-permeabilized cells were used and a membrane staining was detected in cells expressing MT2WT–Myc–His6 and MT2S753A–Myc–His6 (Figures 1B and 1D). The level of matriptase-2 at the cell surface of cells expressing MT2S753A–Myc–His6 was lower (Figure 1B), which was confirmed by immunoblot analysis, comparing the signals for matriptase-2 in membrane fractions of cells expressing MT2WT–Myc–His6 and MT2S753A–Myc–His6 (Figure 2). These results indicate that the C-terminus of matriptase-2 has an extracellular localization, consistent with the predicted cell membrane orientation of TTSPs [28–30]. In contrast, no signal was detectable within HEK-293 cells transfected with pcDNA4-Myc-HisA (mock) confirming the specificity of the immunofluorescence signals (Figures 1E and 1F). These results are in agreement with results previously described [1,21,32] and were generated to prove the effectiveness of the cell model used for the investigation of human matriptase-2.
Subcellular localization of matriptase-2
Immunoblot analysis of matriptase-2 from cells expressing MT2WT–Myc–His6
Identification of membrane-bound and shed forms of matriptase-2
Membrane localization, as observed in immunofluorescence experiments, was confirmed by immunoblot analysis of isolated membranes derived from cells expressing MT2WT–Myc–His6. As shown in Figure 2(A), matriptase-2 was visualized in membrane fractions of cells expressing MT2WT–Myc–His6 as a single band of approx. 120 kDa (MT2) under reducing conditions. This is consistent with the calculated molecular mass of 88.9 kDa for human matriptase-2 with both tags, as the protein is known to be N-linked glycosylated ; treatment of the cellular membrane fraction with PNGase F resulted in a shift of the 120-kDa fragment to a protein band with a molecular mass similar to the calculated mass of matriptase-2 (Figures 2C and 2D, MT2). Under non-reducing conditions the same immunoreactive signal of approx. 120 kDa was detected in membranes of cells expressing MT2WT–Myc– His6 (Figure 2B, MT2). Furthermore, signals at >170 kDa (Figure 2B, MT2 complex), which were not present under reducing conditions, indicated that matriptase-2 exists in membranes as homo- or hetero-meric complexes, tethered by disulfide bonds. The first extracellular cysteine residue of matriptase-2, Cys141 (see Figure 4D), remains intramolecularly unpaired as suggested by computational simulations with murine matriptase-2  and thus may serve as a potential site for homo- or hetero-meric protein interactions via disulfide bonding. Matriptase was also described previously as existing in heteromeric complexes within cellular membranes but these interactions are non-covalent [15,27]. The protein interacting with matriptase was identified in membrane-bound complexes , as well as in complexes localized in the conditioned medium of breast cancer cells or in human milk as the cognate inhibitor HAI-1 . It is known that, besides HAI-1, HAI-2  and several serpins [42,43] can also form complexes with matriptase. Owing to the size of the multimeric matriptase-2 form, various interacting partners are possible, including, for example, matriptase-2 itself as a transactivator (see below), which was also assumed for matriptase [22,23,27].
It has been postulated that activation of TTSPs occurs via a proteolytic cleavage converting the single-chain zymogen into a two-chain activated form in which the catalytic domain is tethered by a disulfide bond to the cell-membrane-anchored non-catalytic N-terminus [28–30,44,45]. Therefore, after activation cleavage, the catalytic domain of matriptase-2 is expected to be attached to the propeptide fragment via the postulated disulfide bond formed by the two conserved cysteine residues Cys559 and Cys679, which are localized within the activation domain and the catalytic domain respectively. Although we used reducing conditions, we were not able to visualize the expected cleavage product, i.e. the catalytic domain, in membrane fractions, suggesting that matriptase-2 is present predominantly in an inactive single-chain form in cellular membranes. In contrast, conditioned medium from cells expressing MT2WT-Myc– His6 contained a matriptase-2 variant with a molecular mass of approx. 30 kDa, which is similar to the predicted molecular mass of the catalytic domain (Figure 2A, MT2 cat). This finding suggests that matriptase-2 is released from the cell surface as described previously for matriptase [22,31]. PNGase F treatment had no effect on the molecular mass of the 30-kDa fragment (Figure 2C, MT2 cat) consistent with the lack of consensus N-glycosylation sites in the protease domain . None of these signals were detectable in mock-transfected cells or in cytosolic fractions of transfected HEK-293 cells (Figure 2A), demonstrating the specificity of the anti-c-Myc antibody.
Selective activity of shed matriptase-2
To analyse the catalytic activities of cell-associated and released variants of matriptase-2, we determined the hydrolysis rate of a substrate for trypsin-like proteases, N-Boc-Gln-Ala-Arg-p-nitroanilide, which is also suitable for matriptase-2 [1,39], in isolated membranes and conditioned medium from transfected HEK-293 cells. Significant proteolytic activity was detectable in conditioned medium derived from cells expressing MT2WT–Myc–His6, correlating with increasing amounts of total protein content. In contrast, only a weak proteolytic activity was detected in the membrane fractions of these cells (Figure 3A).
Cell surface shedding of active matriptase-2
It was concluded that all measured activity resulted from the specific catalytic activity of overexpressed matriptase-2 in cells transfected with MT2WT–Myc–His6, as we did not detect any activity in the conditioned medium and membrane fractions from mock cells or cells expressing MT2S753A–Myc–His6, although matriptase-2 protein was detectable in the membrane fractions of cells expressing MT2S753A–Myc– His6 (Figure 3C). Furthermore, enzyme activity in membrane fractions and conditioned medium from cells expressing MT2WT–Myc–His6 was inhibited by the general serine protease inhibitors aprotinin and leupeptin, and there was also no significant inhibition of proteolytic activity by 2 mM EDTA, a general inhibitor of metalloproteases (results not shown).
In order to analyse whether the higher proteolytic activity in the conditioned medium from cells transfected with MT2WT–Myc–His6, compared with that of membrane fractions, is due to a higher content of matriptase-2, we assessed the amount of the enzyme in isolated membranes and conditioned medium by Western blotting. Matriptase-2 protein content did not differ significantly in membranes and conditioned medium (Figures 3B and 3C). Thus the higher proteolytic activity in the conditioned medium from transfected HEK-293 cells resulted from a higher specific activity of matriptase-2 (an average of 0.073 unit/mg of total protein in the conditioned medium compared with an average of 0.006 unit/mg of total protein in membrane fractions). When expressed in HEK-293 cells, matriptase also showed a higher specific activity in the conditioned medium than in membrane fractions . These results indicate that cellular membranes mainly contain the inactive variant of matriptase-2, correlating with the finding that no active two-chain matriptase-2 form was detectable in membranes by Western blotting. The detection of the weak proteolytic activity in membranes might be due to a weak intrinsic proteolytic activity of single-chain TTSPs , which is important for auto-activation (see below) and which has also been described for non-activated matriptase [23,44,47]. Moreover, the concentration of the activated enzyme in membranes might be too low for detection by immunoblotting, probably due to continuous and rapid cell surface shedding.
Shedding requires proteolytic processing within the stem region
The isolation of soluble forms of TTSPs [31,48–50] suggests that the extracellular domain of at least some of them may be released from the cell surface. In agreement with recently published results on non-specified shed forms of matriptase-2 in the conditioned medium from transiently transfected HeLa and HepG2 cells [21,32,34,35], we have shown that a trypsin-like proteolytic activity exerted by matriptase-2 was predominantly found in the conditioned medium. In order to characterize the released matriptase-2 form in the conditioned medium from cells expressing MT2WT–Myc–His6 in more detail, we performed Western blotting after non-reducing SDS/PAGE. Interestingly, under these conditions a 55-kDa fragment was detected (Figure 2B, MT2 shed) instead of the 30-kDa immunoreactive band seen under reducing conditions (Figure 2A, MT2 cat). These results suggest that the shed form of matriptase-2 represents an active two-chain form in which the catalytic domain of matriptase-2 is either linked to the non-catalytic N-terminal part (stem region) via the conserved disulfide bond, which was described previously for matriptase , or, alternatively, the catalytic domain is released in a dimeric form as proposed by Silvestri et al. .
To assess both possibilities, we isolated the released 55-kDa form of matriptase-2 from the conditioned medium of cells expressing MT2WT–Myc–His6 by the use of immunoaffinity chromatography. After binding of matriptase-2 to immobilized anti-c-Myc antibody and several washing steps, matriptase-2 was eluted and the purity was confirmed by SDS/PAGE. Coomassie Brillant Blue staining revealed the presence of a single protein fragment of 55 kDa after non-reducing SDS/PAGE (Figure 4A), whereas two fragments of approx. 45 and 30 kDa were visualized under reducing conditions (Figure 4B). These findings strongly suggest that the catalytic domain of matriptase-2 is shed into the surrounding medium bound to a part of the stem region, as described for matriptase . Furthermore, we analysed the two different fragments under reducing conditions by LC–MS/MS. Analysis of the 30-kDa fragment revealed peptides matching the catalytic domain of matriptase-2, with a sequence identity of approx. 60% (Figure 4C). MS analysis of the 45-kDa fragment indicated that this fragment consisted of peptides from the three LDLRA domains and the N-terminal part of the activation site. Therefore the shed 55-kDa fragment (MT2 shed), visualized under non-reducing conditions, represents the activated two-chain fragment in which the 30-kDa serine protease domain (MT2 cat) is linked to a 45-kDa part of the stem region via disulfide bonding (Figure 4D). The inconsistency of the molecular masses of these fragments (55 kDa, as the sum of the 30 and 45 kDa fragments) can be explained by the observation that under non-reducing conditions the proteins seem to have a lower molecular mass due to the specific folding of the proteins under these conditions, which has also been described for other TTSPs [15,31,51].
Purification and analysis of shed matriptase-2 from cells expressing MT2WT–Myc–His6
These findings were confirmed by one additional experiment. In contrast with published data , PNGase F treatment of the conditioned medium lowered the molecular mass of the released 55-kDa fragment under non-reducing conditions to approx. 36 kDa (Figure 2D, MT2 shed), indicating that this fragment comprised part of the stem region because only the stem region harbours possible N-glycosylation sites . The minor band of 30 kDa in the same lane represents the catalytic domain cleaved from the stem region because the samples were pretreated with buffer containing dithiothreitol prior to PNGase F addition (Figure 2D, MT2 cat). Taken together, the cleavage site for the release of matriptase-2 from the cell surface must be located within the stem region.
Identification of the shedding sequences
To determine the positions of the processing sites of matriptase-2, the 30-kDa and 45-kDa fragments isolated from the conditioned medium from transfected HEK-293 cells by immunoaffinity chromatography and subsequent reducing SDS/PAGE were each subjected to N-terminal amino acid sequencing. The 30-kDa fragment contains the amino acid sequence Ile568-Val-Gly-Gly-Ala572, representing the cleavage site between Arg567 and Ile568 within the activation motif, and the 45-kDa fragment starts with the amino acid sequence Val438-His-Tyr-Gly-Leu-Tyr-Asn-Gln445 at its N-terminus (Figure 4C). This reveals a cleavage site after Arg437 lying within the stem region at the C-terminal end of the second CUB domain, confirming the results obtained by MS. The 45-kDa fragment must be extensively N-glycosylated, as predicted by computational analysis, because this part only comprises 130 amino acids but has an apparent molecular mass of 45 kDa. Notably, both cleavages occur after arginine residues, consistent with the substrate preference of matriptase-2 for a basic amino acid in the P1 position [1,39], giving a first hint that matriptase-2 itself might be involved.
To verify that the peptide bond after Arg437 is indeed the cleavage site required for cell surface release of matriptase-2, we generated cells expressing MT2R437E–Myc–His6, a mutant variant of matriptase-2 in which Arg437 was replaced with a glutamate residue. As shown in Figure 5(A), the catalytic domain of the mutant variant of matriptase-2 was detected in the conditioned medium after reducing SDS/PAGE (MT2 cat), indicating that mutation of Arg437 did not prevent the shedding process. However, a fragment of approx. 57 kDa, larger than the shed form of wild-type matriptase-2, was visualized in the conditioned medium by immunoblotting after non-reducing SDS/PAGE (Figure 5B, MT2 shed), demonstrating that cleavage after amino acid at position 437 was inhibited if no arginine is present and that more than one shedding sequences are possible. This unknown cleavage site must be located at the N-terminal side of Arg437 because this position represents the final processing site leading to the smallest two-chain form of matriptase-2 (55 kDa) present in the conditioned medium. Because the shedding process is an autocatalytic process (see below), we generated HEK-293 cells that expressed matriptase-2 variants with mutations at different arginine residues N-terminal of Arg437. As shown in Figure 5(B), the conditioned medium of MT2R404E–Myc–His6 cells revealed the 55-kDa fragment (MT2 shed) representing the released form produced by cleavage after Arg437. Surprisingly, only a weak signal at 30 kDa was detectable in the conditioned medium from these cells under reducing conditions (Figure 5A, MT2 cat), indicating that activation was affected. In contrast, no fragments were found in the conditioned medium of the cells expressing the double-mutant MT2R404E/R437E–Myc–His6 protein under both conditions (Figures 5A and 5B). Altogether, these findings show that cell surface release of matriptase-2 in HEK-293 cells occurs by cleavage after Arg404 or Arg437, within the second CUB domain, which is in contrast with a recent report of Ramsay et al.  who assumed that the released form of matriptase-2 is generated by cleavage at Gly81, located within the SEA domain. Nevertheless, future investigations are needed in order to analyse whether N-terminal processing without cell surface release occurs, as described for matriptase, at the corresponding Gly81 in matriptase-2 [23,25,26]. Interestingly, the localization of the two shedding sequences of matriptase-2 differs from that described for matriptase. Matriptase can be shed by proteolytic cleavage after two lysine residues between the SEA domain and the first CUB domain , whereas matriptase-2 is released without the two CUB domains.
Immunoblot analysis of matriptase-2 variants affected in the cell surface release
Furthermore, the present study has indicated the sequence of the shedding and activation, as has been determined previously for matriptase ; because predominantly single-chain zymogen was found in membrane fractions of cells expressing MT2R404E/R437E–Myc–His6, shedding seems to be the first step in the proteolytic processing of matriptase-2. Therefore, apart from its intrinsic activity, matriptase-2 is inactive on the cell surface, as shown in Figure 3, and conversion of the single-chain zymogen into the two-chain protease might occur immediately after shedding. Then, the intrinsic activity of the zymogen must be sufficient for shedding and activation, as has been described for the activation of matriptase .
Shedding and activation requires catalytically active matriptase-2
In order to investigate the requirement of matriptase-2 activity for shedding and activation cleavage, i.e. whether both processing events are autocatalysed, we utilized HEK-293 cells transfected with cDNA encoding an inactive variant of matriptase-2, tagged with the c-Myc epitope, in which the catalytic triad Ser753 was replaced with an alanine residue. Immunoblot analysis revealed a 120-kDa fragment in isolated membrane fractions from cells expressing MT2S753A–Myc under both reducing (Figure 6A, MT2) and non-reducing conditions (results not shown) by the use of the anti-c-Myc antibody, as observed for wild-type matriptase-2. No derivatives of the mutated S753A variant were visible in the conditioned medium from cells expressing MT2S753A–Myc indicating that shedding was prevented. Additionally, activation was also inhibited, as the 30-kDa fragment was absent in the conditioned medium and membrane fractions from these cells (Figure 6A). Taking these results together, both processing events, i.e. shedding and activation cleavage, appear to be autocatalytic processes, consistent with previous results for autoactivation of matriptase-2 expressed in bacteria . In addition, matriptase underwent autoactivation cleavage [23,47], whereas for most other serine proteases the activation cleavage is facilitated by upstream proteases. This autoactivation mechanism allows matriptase and matriptase-2 to serve as an upstream protease.
Immunoblot analysis of catalytically inactive matriptase-2
To analyse whether the autoactivation and/or autoshedding processes are catalysed by the same (cis-mechanism) or by another (trans-mechanism) matriptase-2 molecule, we co-transfected HEK-293 cells with two vectors, one expressing His6-tagged wild-type matriptase-2 and one the c-Myc-tagged inactive S753A matriptase-2 variant. No fragments were detectable in the conditioned medium from singly transfected MT2S753A–Myc cells as noted above. However, a weak signal was visualized in the conditioned medium of cells co-transfected with MT2WT–His and MT2S753A–Myc using the anti-c-Myc antibody (Figure 6A, MT2 cat), indicating that the inactive S753A variant of matriptase-2 was shed into the surrounding medium. The released form was additionally processed by activation cleavage because the 30-kDa fragment (Figure 6A, MT2 cat) appeared in the medium under reducing conditions. Therefore shedding and activation cleavage of the c-Myc-tagged inactive S753A mutant form were catalysed by the His6-tagged wild-type matriptase-2, suggesting a trans-mechanism in which the interaction between matriptase-2 molecules leads to generation of a secreted two-chain variant of matriptase-2. Notably, this is not a proximity-driven process because activated S753A matriptase-2 molecules were also detectable in the conditioned medium from cells expressing MT2S753A–Myc after the addition of conditioned medium from MT2WT–His6 cells and incubation for 2 days (results not shown).
Expression of His6-tagged wild-type matriptase-2 in cells co-transfected with MT2WT–His6 and MT2S753A–Myc was confirmed by the use of the anti-His6 antibody (Figure 6B, MT2 and MT2 cat); note that owing to the different numbers of tags on the shed matriptase-2 fragments a shift of the molecular mass of the 30-kDa fragment was seen in the conditioned medium from different transfected cells (Figure 6B, MT2 cat). No immunoreactive signals were detected in membrane fractions or in the conditioned medium of singly transfected MT2WT–His6 cells using the anti-c-Myc antibody (Figure 6A), nor in those fractions derived from MT2S753A–Myc cells using the anti-His6 antibody (Figure 6B).
In conclusion, we have demonstrated for the first time that human matriptase-2 exists both as single monomers and as complexes in cellular membranes and characterized in detail the released portion of the protein in the extracellular space. We identified three proteolytic processing events for matriptase-2: two cleavages occur after Arg404 and Arg437, within the stem region, and are important for cell surface release of matriptase-2, and one cleavage occurs after residue Arg567, within the conserved activation site, converting the single-chain zymogen into the activated two-chain protease. Shedding seems to be the first step in the proteolytic processing of matriptase-2 because activation was prevented in HEK-293 cells mutant expressing a matriptase-2 variant which cannot be shed (MT2R404E/R437E–Myc–His6). This is consistent with the finding that matriptase-2 exerted its activity predominantly in the conditioned medium. Additionally, our results reveal that matriptase-2 undergoes shedding and activation via a trans-mechanism.
Marit Stirnberg designed and supervised the project, and conceived and performed most of the experiments. Eva Maurer contributed to the design, experimental work and data interpretation. Angelika Horstmeyer, Sonja Kolp, Kai Prager and Jochen Walter participated in the initial planning of the project and the experimental approach. Stefan Frank, Tobias Bald, Katharina Arenz and Andreas Janzer contributed to the characterization of mutant variants in HEK-293 cells and in the establishment of enzyme activity measurements. Patrick Wunderlich and Jochen Walter collaborated on the microscopic analyses. Marit Stirnberg and Michael Gütschow interpreted the results and wrote the paper. Michael Gütschow conceived the project.
We thank Janna Rudolph for assistance with the cloning experiments.
This work was supported by Deutsche Forschungsgemeinschaft [grant number SFB645]; and by the North Rhine-Westphalia International Graduate Research School Biotech-Pharma.