N-Glycosylation Is Required for Matriptase-2 Autoactivation and Ectodomain Shedding

Background: Matriptase-2 is a hepatic membrane serine protease that regulates hepcidin expression and iron metabolism.

Results: Mutations at specific N-glycosylation sites prevented matriptase-2 activation and ectodomain shedding.

Conclusion: N-Glycans play an important role in regulating matriptase-2 activity.

Significance: The results provide new insights into the mechanism underlying matriptase-2 expression, zymogen activation, and ectodomain shedding.

Abstract

Matriptase-2 is a hepatic membrane serine protease that regulates iron homeostasis. Defects in matriptase-2 cause iron deficiency anemia. In cells, matriptase-2 is synthesized as a zymogen. To date, how matriptase-2 expression and activation are regulated remains poorly understood. Here we expressed human matriptase-2 in HEK293 and hepatic BEL-7402, SMMC-7721, and QGY-7703 cells. By labeling cell surface proteins and Western analysis, we examined matriptase-2 cell surface expression, zymogen activation, and ectodomain shedding. Our results show that matriptase-2 was activated on the cell surface but not intracellularly. Activated matriptase-2 underwent ectodomain shedding, producing soluble fragments in the conditioned medium. By testing inactive mutants, R576A and S762A, we found that matriptase-2 activation and shedding were mediated by its own catalytic activity and that the one-chain form of matriptase-2 had little activity in ectodomain shedding. We made additional matriptase-2 mutants, N136Q, N184Q, N216Q, N338Q, N433Q, N453Q, and N518Q, in which each of the predicted N-glycosylation sites was mutated. All of these mutants were expressed on the cell surface. However, mutants N216Q, N453Q, and N518Q, but not the other mutants, had impaired zymogen activation and ectodomain shedding. Our results indicate that N-glycans at specific sites are critical for matriptase-2 activation. Together, these data provide new insights into the cell surface expression, zymogen activation, and ectodomain shedding of matriptase-2.

Introduction

Matriptase-2, also known as TMPRSS6, is a member of the type II transmembrane serine protease family (1–3). Matriptase-2 protein consists of a short N-terminal cytoplasmic tail, a transmembrane domain, an extracellular stem region containing a SEA domain, two CUB domains, three LDL receptor repeats, and a C-terminal trypsin-like protease domain (2, 3). Matriptase-2 is mainly expressed in the liver (3, 4). Low levels of matriptase-2 mRNA expression have been found in other tissues, such as the kidney, lung, brain, and uterus (3, 4). Matriptase-2 expression also was reported in breast and prostate cancers (5–7).

Matriptase-2 plays a key role in iron homeostasis by modulating hepcidin, a hepatic peptide hormone that binds to and down-regulates the iron transporter ferroportin. Low levels of ferroportin prevent iron absorption in the gut and iron release from macrophages. Studies show that matriptase-2 inhibits hepcidin expression by proteolytic cleavage of hemojuvelin, a bone morphogenetic protein co-receptor on the hepatocyte surface membrane that is required for hepcidin gene expression (8–14). The importance of matriptase-2 in iron metabolism has been demonstrated in animal models and patients. In mice, matriptase-2 deficiency led to high levels of serum hepcidin and iron-deficient anemia (15, 16). In humans, mutations in the TMPRSS6 gene encoding matriptase-2 have been identified in patients with iron-refractory iron deficiency anemia (IRIDA)⁴ (11, 17–21). In functional studies, the mutations found in IRIDA patients were shown to impair matriptase-2 biosynthesis or proteolytic activity (11, 17–25).

Matriptase-2 is synthesized as an inactive zymogen (2, 3). Proteolytic cleavage at a conserved activation site, Arg⁵⁷⁶-Ile⁵⁷⁷, is required to activate matriptase-2. In other type II transmembrane serine proteases, such as matriptase, hepsin, and corin (1, 26, 27), the activated protease domain fragment is expected to remain on the cell surface by a disulfide bond linking the protease domain and the membrane-bound propeptide region. To date, how matriptase-2 expression and activity are regulated is unclear. Studies have shown that matriptase-2 may undergo ectodomain shedding and endocytosis (13, 21, 28, 29). It has been suggested that the shedding may be the first step in the proteolytic processing of matriptase-2 and that this event may be mediated by the single-chain form of matriptase-2 on the cell surface that has weak intrinsic proteolytic activity (29). Once matriptase-2 is shed from the cell surface, soluble matriptase-2 fragments may undergo autoactivation (3, 29).

In this study, we expressed matriptase-2 in HEK293 cells and human hepatocellular carcinoma cells. By immunostaining, cell surface labeling, immunoprecipitation, and Western analysis, we examined matriptase-2 protein on the cell surface and in cell lysate and conditioned medium. Our results indicate that matriptase-2 was autoactivated on the cell surface but not inside the cell and that the one-chain form of matriptase-2 had little activity in ectodomain shedding. Moreover, by site-directed mutagenesis, we identified specific N-glycosylation sites in matriptase-2 that are crucial for zymogen activation and ectodomain shedding.

EXPERIMENTAL PROCEDURES

Cell Culture

Human embryonic kidney 293 (HEK293) cells and human hepatocellular carcinoma cell lines BEL-7402, SMMC-7721, and QGY-7703 were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum and 4 mm glutamine at 37 °C with 95% humidified air and 5% CO₂.

Matriptase-2 cDNA and Expression Plasmids

The full-length human matriptase-2 cDNA was synthesized from HepG2 mRNA by reverse transcription PCR and cloned into pcDNA3.1 expression vector (Invitrogen). The recombinant matriptase-2 protein expressed by this vector contained a C-terminal V5 tag for protein detection in immunostaining and Western blotting (30). The matriptase-2 activation cleavage site mutant R576A; catalytic active site mutant S762A; and N-glycosylation site mutants N136Q, N184Q, N216Q, N338Q, N433Q, N453Q, and N518Q were made by PCR-based site-directed mutagenesis using wild-type (WT) matriptase-2 plasmid as a template. All constructs were verified by DNA sequencing.

Cell Transfection, Immunoprecipitation, and Western Blotting

Plasmids expressing human matriptase-2 WT and mutants were transfected into HEK293 and hepatic cells using TurboFect transfection reagent (Thermo Scientific). After 5 h, the medium was replaced with serum-free Opti-MEM medium. The conditioned medium was collected after 36 h, and matriptase-2 proteins were immunoprecipitated with an anti-V5-antibody and protein A-Sepharose beads, as described previously (31). The cells were lysed in a lysis buffer containing 50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1% Nonidet P-40 (v/v), and a protease inhibitor mixture (1:100; Sigma). Matriptase-2 proteins were separated by 10% SDS-PAGE in the presence (reducing) or absence (non-reducing) of β-mercaptoethanol followed by Western blotting using a horseradish peroxidase (HRP)-labeled anti-V5 antibody. Western blots were incubated with a chemiluminescence substrate solution (EZ-ECL, Biological Industries) and exposed to x-ray films.

Immunostaining

HEK293 and BEL-7402 cells were transfected with plasmid expressing matriptase-2 and cultured in 6-well plates with glass coverslips. After 24–48 h, the cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) at room temperature for 30 min and incubated with 5% bovine serum albumin in PBS at 37 °C for 1 h. After washing with PBS, the cells were incubated with an anti-V5-antibody at 37 °C for 1 h. After washing, the cells were incubated with Alexa Fluor 488-conjugated goat anti-mouse IgG antibody (Invitrogen) at room temperature for 1 h. The coverslips were mounted in Fluomount-G reagent (Southern Biotech). Immunostained cells were examined, and images were taken with a laser scanning confocal microscope (Olympus FV500).

Effects of Protease Inhibitors on Matriptase-2 Shedding

To identify proteases involved in matriptase-2 shedding, protease inhibitors of different classes, including trypsin-like serine protease inhibitor (benzamidine, 2.5 mm), metalloproteinase inhibitor (GM6001, 50 μm), ADAM inhibitor (TAPI-1, 50 μm), and cysteine protease inhibitor (ALLM, 10 μm), were added to cultured HEK293 cells expressing WT matriptase-2 (32). The conditioned medium was collected after 36 h, and the cells were lysed. Matriptase-2 protein fragments in the conditioned medium and cell lysate were analyzed with immunoprecipitation and Western blotting, as described above.

Glycosidase Digestion

To examine carbohydrate contents on matriptase-2, transfected HEK293 and BEL-7402 cells expressing matriptase-2 were lysed with a buffer containing 50 mm Tris-HCl, pH 8.0, 150 mm NaCl, and 1% Nonidet P-40. Cell lysate (50 μg of proteins) was incubated in a denaturing buffer at 98 °C for 10 min. Peptide-N-glycosidase F and O-glycosidase (New England Biolabs) were then added, either individually or together, to the samples and incubated at 37 °C for 2 h according to the manufacturer's instructions. Protein samples were analyzed with SDS-PAGE and Western blotting using an HRP-conjugated anti-V5 antibody.

Cell Surface Protein Labeling

HEK293 and BEL-7402 cells expressing matriptase-2 were grown in culture. Cell surface proteins were labeled using methods described previously (30). Briefly, the cells were washed with PBS and labeled with cell membrane-impermeable sulfo-NHS-biotin (1 mmol/liter) (Pierce) at 4 °C for 5 min. The reaction was quenched by glycine (100 mmol/liter) in PBS, and the cells were lysed in a buffer containing 50 mmol/liter Tris-HCl, pH 8.0, 150 mmol/liter NaCl, 1% Triton X-100 (v/v), and a protease inhibitor mixture (1:100 dilution). Streptavidin-Sepharose beads (30 μl) were added into the cell lysate, and the mixture was rotated at 4 °C overnight. After washing, the beads were boiled in a Laemmli sample buffer. Proteins were analyzed with SDS-PAGE and Western blotting using an HRP-conjugated anti-V5 antibody.

Flow Cytometry

Procedures for flow cytometry to study cell surface proteins were based on methods published previously (30). Matriptase-2 WT and mutants were expressed in transfected HEK293 cells. The cells were isolated and incubated with a primary anti-V5 antibody and then a secondary antibody conjugated with FITC. Life gating was performed using pyridinium iodide (Sigma). Experiments were done using a flow cytometer (FACSCalibur, BD Biosciences), and data were analyzed with CellQuest software.

RESULTS

Localization of Matriptase-2 on HEK293 and BEL-7402 Cell Surface

We expressed human matriptase-2 in HEK293 cells and human hepatic cancer BEL-7402 cells. To verify the surface expression of matriptase-2 in these cells, immunostaining was performed using an antibody recognizing the V5 tag at the C terminus of recombinant matriptase-2. Under non-cell membrane-permeable conditions, positive staining was detected in both HEK293 and BEL-7402 cells transfected with matriptase-2-expressing plasmid (Fig. 1). No positive staining was detected in control HEK293 and BEL-7402 cells transfected with a vector (Fig. 1). The results confirmed the cell surface expression and the predicted membrane topology of matriptase-2 in HEK293 and BEL-7402 cells in our experiments.

FIGURE 1.

Cell surface expression of matriptase-2 in HEK293 and BEL-7402 cells. HEK293 (top panels) and BEL-7402 (bottom panels) cells were transfected with a matriptase-2 (MT2)-expressing plasmid (right panels) or a control vector (left panels). Cells were fixed with 4% paraformaldehyde and immunostained with an anti-V5 antibody and an Alexa Fluor 488-conjugated secondary antibody. Images were taken with a confocal microscope. Scale bar, 10 μm. Data are representative of at least three independent experiments.

Matriptase-2 Protein in Cell Lysate and Cultured Medium

We analyzed matriptase-2 protein in transfected HEK293 cells by Western blotting. Under non-reducing conditions, a major band of ∼120 kDa was found in cell lysate (Fig. 2A, left), consistent with the predicted molecular mass of matriptase-2 (2, 3). There was another weaker band of ∼190–200 kDa, probably a matriptase-2 complex or dimer. Under reducing conditions, only the ∼120 kDa band was detected (Fig. 2A, right). In the conditioned medium, a major band of ∼110 kDa appeared under non-reducing conditions (Fig. 2B, left). Under reducing conditions, four bands of ∼110, ∼90, ∼60, and ∼30 kDa, respectively, were found in the conditioned medium (Fig. 2B, right), which probably represented matriptase-2 fragments cleaved at different positions (Fig. 2C). In controls, no specific bands were detected in cell lysate or conditioned medium from vector-transfected HEK293 cells (Fig. 2, A and B).

FIGURE 2.

Analysis of matriptase-2 in lysate and conditioned medium from HEK293 and hepatic cells. Western analysis of matriptase-2 (MT2) in HEK293 cell lysate (A) and conditioned medium (B) under non-reducing (NR) and reducing (R) conditions. Black arrowhead, serine protease (SP) domain fragment in the conditioned medium. C, schematic illustration of possible proteolytic cleavage sites in matriptase-2. TM, transmembrane; SP, serine protease. D, Western analysis of matriptase-2 in lysates from BEL-7402 (BEL), SMMC-7721 (SMMC), and QGY-7703 (QGY) cells transfected with matriptase-2 (MT2)-expressing plasmid or control vector (vec). Western blotting was done under reducing conditions. E, detection of soluble matriptase-2 fragments in the conditioned medium under non-reducing (left) and reducing (right) conditions. Data are representative of at least three independent experiments.

Matriptase-2 is primarily of liver origin (3, 4). We were unable to study endogenous matriptase-2 in hepatocytes due to a lack of suitable antibodies. Instead, we expressed and analyzed recombinant matriptase-2 in three human hepatocellular carcinoma cell lines. In Western analysis under reducing conditions, a similar ∼120 kDa band was detected in lysates from BEL-7402, SMMC-7721, and QGY-7703 cells transfected with matriptase-2-expressing plasmid (Fig. 2D). In the conditioned medium, the ∼190–200 and ∼110 kDa bands were detected under non-reducing conditions (Fig. 2E, left), and the ∼110, ∼90, ∼60, and ∼30 kDa bands were detected under reducing conditions (Fig. 2E, right). The results indicate that matriptase-2 underwent similar proteolytic cleavages in HEK293 and the hepatic cells. The presence of the ∼30 kDa band, representing the cleaved protease domain fragment, suggests that some matriptase-2 molecules were activated in these cells.

Effects of Protease Inhibitors on Matriptase-2 Shedding

To examine the enzyme(s) responsible for matriptase-2 shedding, we tested different protease inhibitors in HEK293 and BEL-7402 cells expressing matriptase-2. Levels of soluble matriptase-2 fragments (∼190–200 and ∼110 kDa bands in HEK293 cells and ∼110 kDa band in BEL-7402 cells) were markedly reduced in cells treated with benzamidine but not the other protease inhibitors (Fig. 3). In controls, matriptase-2 protein levels in cell lysates were not significantly altered in the presence of the protease inhibitors (Fig. 3). These results indicate that a trypsin-like protease or proteases were responsible for matriptase-2 cleavage and shedding in both HEK293 and BEL-7402 cells.

FIGURE 3.

Effects of protease inhibitors on matriptase-2 shedding. HEK293 (left panels) and BEL-7402 (right panels) cells expressing matriptase-2 were incubated with vehicle, GM6001, ALLM, benzamidine (Benz), or TAPI-1. Vector (vec)-transfected HEK293 cells were included as controls. Soluble matriptase-2 fragments in the conditioned medium were analyzed by immunoprecipitation and Western blotting under non-reducing conditions. Matriptase-2 (MT2) protein levels in cell lysate were verified. GAPDH levels in cell lysate were used as a control. Data are representative of at least three independent experiments.

Matriptase-2 Autocleavage

Matriptase-2 is a trypsin-like serine protease. It is possible that the detected matriptase-2 fragments in the culture medium were derived from matriptase-2 autocleavage. To test this hypothesis, we made two inactive matriptase-2 mutants: R576A, in which the activation cleavage site was abolished, and S762A, in which the catalytic Ser was mutated (Fig. 4A). In transfected HEK293 and BEL-7402 cells, these two mutants and WT matriptase-2 were expressed at similar levels in cell lysate (Fig. 4B). However, no soluble matriptase-2 fragments were detected in the conditioned medium from the cells expressing R576A and S762A mutants (Fig. 4B, top panels). The results indicate that the generation of the soluble fragments depends on the catalytic activity of matriptase-2. The results also indicate that the mutant R576A, which was expected to remain as a one-chain form, had little activity in ectodomain shedding.

FIGURE 4.

Matriptase-2 shedding in R576A and S762A mutants. A, schematic illustration of matriptase-2 activation cleavage site. The conserved activation cleavage site Arg⁵⁷⁶-Ile⁵⁷⁷ (R576-I577) and catalytic residues His (H), Asp (D), and Ser⁷⁶² (S762) are shown. A disulfide bond linking the propeptide region and the protease domain is indicated. B, soluble matriptase-2 fragments in the conditioned medium from HEK293 (left) and BEL-7402 (right) cells expressing WT, R576A, and S762A mutants were analyzed by immunoprecipitation and Western blotting under non-reducing conditions. Matriptase-2 expression levels in cell lysate were verified. GAPDH levels in cell lysate were used as a control. C, immunostaining of matriptase-2 WT and mutants R576A and S762A in transfected HEK293 cells. Data are representative of at least three independent experiments. D, flow cytometric analysis of the cell surface expression of matriptase-2 WT and mutants R576A and S762A in transfected HEK293 cells. Percentages of matriptase-2-positive cells (mean ± S.D.) from six experiments are indicated.

We verified the cell surface expression of R576A and S762A mutants. In immunostaining, a similar cell membrane expression pattern was observed for matriptase-2 WT and the two mutants (Fig. 4C). By flow cytometry, higher numbers of matriptase-2-positive cells were found in HEK293 cells transfected with R576A and S762A plasmids compared with that with WT plasmid (37.1 ± 3.8 and 36.3 ± 4.1%, respectively; n = 6, both p values <0.05 versus 29.0 ± 3.1% in WT) (Fig. 4D). These results show that high levels of R576A and S762A mutant proteins were present on the cell surface, probably due to reduced ectodomain shedding.

Matriptase-2 Autoactivation on Cell Membrane

If matriptase-2 is capable of autocleavage, it is expected to be activated first. In previous studies, no matriptase-2 activation was detected on the cell membrane (29). Similarly, we did not detect the ∼30-kDa activation cleaved protease domain fragment in HEK293 and hepatic cells by Western analysis (Fig. 2, A and D). We then examined matriptase-2 on the cell membrane by labeling cell surface proteins followed by Western blotting. An ∼120 kDa band, representing matriptase-2 zymogen, and an ∼30 kDa band, representing the activation-cleaved protease domain fragment, were detected in HEK293 and BEL-7402 cells expressing WT matriptase-2 (Fig. 5, A and B, left panels). As predicted by the protein domain structure (Fig. 4A), the ∼30 kDa band was only detected under reducing and not non-reducing conditions (Fig. 5, A and B, left versus middle panels). In the cells expressing mutants R576A and S762A, the ∼30 kDa band was not detected (Fig. 5, A and B, left panels). The results indicate that matriptase-2 is autoactivated on the cell surface.

FIGURE 5.

Analysis of cell surface-labeled matriptase-2. Cell surface proteins in HEK293 (A) and BEL-7402 (B) expressing matriptase-2 were labeled with biotin and analyzed by Western blotting under reducing (R) (left panels) and non-reducing (NR) (middle panels) conditions. Black arrowheads, cleaved protease domain fragment. Matriptase-2 expression levels in cell lysates were verified (right panels). GAPDH was used as a control for cell surface protein labeling (bottom right panels). Data are representative of at least three independent experiments.

Effects of N-Glycosylation on Matriptase-2 Activation and Shedding

N-Glycans are shown to play a role in cell surface expression and zymogen activation in membrane-bound serine proteases (33–36). There are seven potential N-glycosylation sites on human matriptase-2 (Fig. 6A). To verify the presence of N-glycans, we expressed matriptase-2 in HEK293 and BEL-7402 cells and digested the cell lysates with peptide-N-glycosidase F and O-glycanase. In Western analysis, peptide-N-glycosidase F digestion reduced the apparent mass of matriptase-2 from ∼120 to ∼100 kDa (Fig. 6B). O-Glycanase digestion did not yield a noticeable reduction in the apparent mass (Fig. 6B). The results indicate that matriptase-2 expressed in HEK293 and hepatic BEL-7402 cells contained N-glycans but no detectable amounts of O-glycans.

FIGURE 6.

N-Glycosylation in matriptase-2. A, schematic illustration of seven potential N-glycosylation sites in matriptase-2. B, glycosidase digestion. Matriptase-2 protein in transfected HEK293 (left) and BEL-7402 (right) cells was digested with peptide-N-glycosidase F (F) and O-glycosidase (O), individually or together, and analyzed by Western blotting under reducing conditions. Data are representative of at least three independent experiments.

We next made seven matriptase-2 mutants (N136Q, N184Q, N216Q, N338Q, N433Q, N453Q, and N518Q), in which each of the predicted N-glycosylation sites was mutated. We expressed these mutants in BEL-7402 cells. In Western analysis of labeled cell surface proteins, similar levels of the ∼120 kDa zymogen band were detected in WT and all mutants (Fig. 7A, left). However, the ∼30-kDa protease domain band was detected only in WT and mutants N136Q, N184Q, N338Q and N433Q and not in mutants N216Q, N453Q, and N518Q. In cell lysate, levels of WT and all mutant proteins were similar, and the ∼30 kDa band was not detected (Fig. 7A, right). In the conditioned medium, similar patterns of soluble matriptase-2 fragments were found in WT and mutants N136Q, N184Q, N338Q, and N433Q (Fig. 7B). In contrast, levels of soluble matriptase-2 fragments were markedly reduced in cells expressing mutants N216Q, N453Q, and N518Q (Fig. 7B). These results indicate that N-glycans at specific sites are critical for matriptase-2 autoactivation and ectodomain shedding.

FIGURE 7.

Matriptase-2 activation and shedding in WT and N-glycosylation site mutants. Matriptase-2 WT and N-glycosylation site mutants were expressed in BEL-7402 cells. A, matriptase-2 proteins on the cell surface (left) and in lysate (right) were analyzed by Western blotting under reducing conditions. B, soluble matriptase-2 fragments in the conditioned medium were analyzed by immunoprecipitation and Western blotting under non-reducing (NR) and reducing (R) conditions. Black arrowheads, cleaved protease domain fragment. Matriptase-2 proteins in cell lysate were verified. GAPDH was included as a control. Data are representative of at least three independent experiments.

DISCUSSION

Matriptase-2 has been identified as a critical enzyme in regulating hepcidin expression and iron metabolism (8, 11, 14, 15). To date, how matriptase-2 expression and activation are regulated remains poorly understood. In this study, we expressed human matriptase-2 in HEK293 and hepatic cells and examined matriptase-2 activation and shedding. In Western analysis, matriptase-2 zymogen, as indicated by the ∼120 kDa band, was detected in HEK293 cell lysate. No matriptase-2 activation, as indicated by the cleaved ∼30-kDa protease domain fragment, was detected in HEK293 cell lysate (Fig. 2A). Similar results were found in three hepatic cell lines, BEL-7402, SMMC-7721, and QGY-7703 (Fig. 2D). These results indicate that little matriptase-2 activation occurred inside the cells examined in our study.

We then labeled cell membrane proteins and examined cell surface matriptase-2. By Western analysis, the ∼30 kDa band was detected in both HEK293 and hepatic BEL-7402 cells, indicating that matriptase-2 was activated on the cell surface (Fig. 5, A and B). Consistent with the predicted matriptase-2 structure, the ∼30 kDa band was absent when Western blot was done under non-reducing conditions because the cleaved protease domain is expected to attach to the propeptide region via a disulfide bond (Fig. 4A). The lack of the ∼30 kDa band in samples from mutant R576A, in which the activation cleavage site was abolished, also indicated that the ∼30 kDa band was the activation-cleaved protease domain fragment (Fig. 5, A and B). Because the ∼30-kDa fragment was not detected in samples from mutant S762A, in which the catalytic Ser was mutated, the results indicate that matriptase-2 activation in these cells depended on its own catalytic activity, an indication of autoactivation.

Matriptase-2 autoactivation was reported previously (3, 29). It was unclear, however, where matriptase-2 activation takes place because Western analysis did not detect the cleaved ∼30-kDa protease domain fragment in HEK293 cell lysate and membrane fractions (29). It was suggested that matriptase-2 may undergo shedding first before activation cleavage (29), which raises the question of how shedding may occur before matriptase-2 activation. One possible explanation is that the initial shedding may be mediated by one-chain matriptase-2 on the cell membrane that may have weak proteolytic activity (29). By labeling cell surface proteins, we were able to increase the sensitivity of the Western analysis and detected the cleaved ∼30-kDa protease domain fragment in HEK293 cells expressing WT matriptase-2 (Figs. 5 and 7). Thus, our data show, for the first time, that matriptase-2 was activated on the cell surface. Moreover, we did not detect soluble fragments in mutant R576A lacking the cleavage site (Fig. 4), suggesting that one-chain matriptase-2 had little shedding activity. Together, our results suggest a mechanism, in which matriptase-2 is autoactivated on the cell surface before ectodomain shedding may occur. The sequence of the events may be of biological importance. Matriptase-2 is believed to cleave hemojuvelin, a bone morphogenetic protein co-receptor on the hepatocyte cell membrane. If matriptase-2 is activated on the cell membrane, the activated enzyme is expected to cleave its physiological substrate more efficiently. If matriptase-2 is shed from the cell membrane before it is activated, as proposed previously (29), the shed matriptase-2 will no longer be membrane-bound and may not cleave hemojuvelin efficiently on the cell membrane.

Ectodomain shedding is an important cellular process in regulating membrane protein function (37–39). In Western analysis of matriptase-2 on the cell surface, the ∼30 kDa band represented only a small fraction of total matriptase-2 protein (Fig. 5, A and B), suggesting that once matriptase-2 is activated, it is quickly shed from the cell surface. In Western analysis, we found matriptase-2 fragments of ∼110, ∼90, ∼60, and ∼30 kDa, respectively (Fig. 2, B and E). In previous studies, soluble matriptase-2 fragments of similar sizes also were reported in cell culture medium (13, 21, 28, 29), indicating that proteolytic cleavage may occur at different sites in the matriptase-2 propeptide region (Fig. 2C). Most likely, the observed matriptase-2 cleavage was mediated by its own activity, because the shedding was inhibited by benzamidine but not other classes of protease inhibitors (Fig. 3), and, more importantly, no shedding was detected in inactive matriptase-2 mutants R576A and S762A (Fig. 4). It is known that proteases of the ADAM family are primary enzymes that shed many membrane proteins (38, 40). ADAM10, for example, was shown to shed corin, another type II transmembrane serine protease, in HEK293 and cardiomyocytes (41). In this regard, it is interesting to note that ADAMs did not appear to be involved in matriptase-2 shedding.

N-Glycosylation is a key post-translational process that facilitates protein folding, intracellular trafficking, secretion, and cell surface expression (42–45). In type II transmembrane serine proteases, N-glycans are important for the cell surface targeting and zymogen activation of corin, a protease that regulates salt-water balance and blood pressure (34, 46, 47). A similar role of N-glycans also was reported in matriptase that is important in epithelial function (1, 35, 36). In human matriptase-2, there are seven potential N-glycosylation sites (Fig. 6A). It was unknown if N-glycans on matriptase-2 are necessary for cell surface expression and zymogen activation. In this study, we verified the presence of N-glycans on matriptase-2 expressed in HEK293 and BEL-7402 cells. We found that peptide-N-glycosidase F, but not O-glycosidase, treatment reduced the apparent mass of matriptase-2, indicating that matriptase-2 contained significant amounts of N-glycans but little O-glycan.

By site-directed mutagenesis, we mutated each of the seven predicted N-glycosylation sites on matriptase-2. In transfected BEL-7402 cells, all seven N-glycosylation site mutants were expressed at similar levels on the cell surface (Fig. 7A), suggesting that N-glycosylation at these sites is not critical for matriptase-2 targeting to the cell surface. Interestingly, zymogen activation was markedly impaired in mutants N216Q, N453Q, and N518Q but not in mutants N136Q, N184Q, N338Q, and N433Q (Fig. 7A). Consistently, ectodomain shedding also was reduced significantly in mutants N216Q, N453Q, and N518Q (Fig. 7B). These results indicate that N-glycans at these three specific sites are required for matriptase-2 activation on the cell surface. Remarkably, all three of these N-glycosylation sites are located away from the activation cleavage site (Fig. 6A). In a previous study, mutations at two separate N-glycosylation sites in matriptase, one in the CUB domain in the propeptide region and another in the serine protease domain, also impaired matriptase activation (36). These data suggest that N-glycans at specific sites may be required to maintain a proper protein conformation required for the autoactivation of these membrane-bound proteases.

In summary, we studied matriptase-2 zymogen activation and ectodomain shedding in HEK293 and hepatic cells. Our data indicate that matriptase-2 is autoactivated on the cell surface and then undergoes proteolytic shedding, producing distinct soluble fragments. We also show that one-chain matriptase-2 lacking the activation cleavage site had little shedding activity. We further identified specific N-glycosylation sites that are critical for matriptase-2 activation and ectodomain shedding. Together, our results provide new insights into the cell surface expression, zymogen activation, and ectodomain shedding of matriptase-2, a protease critical in iron homeostasis and iron-deficient anemia.