The catalytic, stem, and transmembrane portions of matriptase-2 are required for suppressing the expression of the iron-regulatory hormone hepcidin

Matriptase-2 (MT2) is a type-II transmembrane, trypsin-like serine protease that is predominantly expressed in the liver. It is a key suppressor for the expression of hepatic hepcidin, an iron-regulatory hormone that is induced via the bone morphogenetic protein signaling pathway. A current model predicts that MT2 suppresses hepcidin expression by cleaving multiple components of the induction pathway. MT2 is synthesized as a zymogen that undergoes autocleavage for activation and shedding. However, the biologically active form of MT2 and, importantly, the contributions of different MT2 domains to its function are largely unknown. Here we examined the activities of truncated MT2 that were generated by site-directed mutagenesis or Gibson assembly master mix, and found that the stem region of MT2 determines the specificity and efficacy for substrate cleavage. The transmembrane domain allowed MT2 activation after reaching the plasma membrane, and the cytoplasmic domain facilitated these processes. Further in vivo rescue studies indicated that the entire extracellular and transmembrane domains of MT2 are required to correct the low-hemoglobin, low-serum iron, and high-hepcidin status in MT2−/− mice. Unlike in cell lines, no autocleavage of MT2 was detected in vivo in the liver, implying that MT2 may also function independently of its proteolytic activity. In conjunction with our previous studies implicating the cytoplasmic domain as an intracellular iron sensor, these observations reveal the importance of each MT2 domain for MT2-mediated substrate cleavage and for its biological function.

Matriptase-2 (MT2) 2 is a member of the matriptase subgroup that belongs to the type II transmembrane serine prote-ase family (1). It is encoded by the TMPRSS6 gene in humans and Tmprss6 gene in mice. For the sake of simplicity, MT2 will be used for the gene and MT2 for the protein throughout the text. MT2 is expressed predominantly in hepatocytes (1,2), and MT2 acts as an essential suppressor for the expression of hepcidin, an iron regulatory hormone that is also predominantly expressed in hepatocytes under nonpathological conditions (3,4). Mutations in MT2 in humans result in an inappropriately high-hepcidin expression, which leads to iron-refractory irondeficiency anemia (IRIDA) (3)(4)(5). Similar phenotypes are also reported in mouse models either with a global MT2 knockout (MT2 Ϫ/Ϫ ) or with a truncated MT2 that lacks the serine protease (S/P) catalytic domain (mask mice) (6 -9). Hepcidin inhibits iron efflux from duodenal epithelial cells, macrophages, and hepatocytes into the circulation by targeting the plasma membrane iron exporter, ferroportin, for degradation (10). Lack of hepcidin causes juvenile hemochromatosis, a severe form of iron overload (11,12). Under physiological conditions, hepcidin expression is regulated positively by body iron content to maintain iron homeostasis.
Hepatic hepcidin expression is induced via the bone morphogenetic protein (BMP) signaling pathway (13)(14)(15). BMP signaling is initiated upon the binding of BMP ligands to type-I and type-II BMP receptors on the cell surface. The BMP2/6 ligands that are responsible for the induction of hepcidin expression in hepatocytes are derived primarily from the adjacent hepatic endothelial cells (16 -19). There are multiple type-I BMP receptors (ALK1, ALK2, ALK3, and ALK6) and type-II BMP receptors (BMPR2, ActRIIA, and ActRIIB). Studies in animal models demonstrate that hepatocytes utilize a selective set of BMP receptors, including ALK2, ALK3, BMPR2 and ActRIIA, to induce hepcidin expression (20,21). Importantly, a normal range of hepcidin expression also requires the involvement of other plasma membrane proteins, including hemojuvelin (HJV), hemochromatosis protein (HFE), transferrin receptor-2 (TfR2), and neogenin (3). In the liver, HJV and TfR2 are exclusively expressed in hepatocytes. HFE is predominantly expressed in hepatocytes, and neogenin is also highly expressed in hepatocytes (22,23). Mutations in the HJV gene in humans markedly reduce hepcidin expression in the liver and result in juvenile hemochromatosis (24). Mutations in the HFE and TfR2 genes also decrease hepcidin expression and cause type-I hemochromatosis, the most common form of hereditary iron overload, and type-III hemochromatosis, respectively (12). All these defects have been documented in animal models. Knockout of these genes recapitulate the human mutation phenotypes, indicating that the mutations impart a lack of function.
MT2 is a trypsin-like type-II transmembrane serine protease composed of a short cytoplasmic domain, a transmembrane domain, and a large extracellular domain, which contains a membrane-proximal stem region, a predicted cleavage-activation site, and a catalytic domain (C-terminal S/P domain) (Fig.  1A) (1,2). Early studies suggested that MT2 suppresses hepcidin expression solely by cleaving HJV (8,25). Our recent data showed that MT2 suppresses hepcidin expression independently of Hjv in vivo and that it cleaves multiple components of the hepcidin-induction pathway, including BMP receptors (ALK2, ALK3, ActRIIA, and Bmpr2), Hfe, and to a lesser extent, Hjv and Tfr2, in vitro (9). Consistent with these latter observations, MT2 Ϫ/Ϫ mice have a lower, rather than higher, level of Hjv protein in the liver (26,27). Thus it is likely that MT2 suppresses hepcidin expression by cleaving and inactivating multiple components of the hepcidin-induction pathway.
MT2 is synthesized as a zymogen in the endoplasmic reticulum. Its activation is predicted to occur on the cell surface by autocleavage (1,28). MT2 also undergoes ectodomain shedding, and multiple sizes of soluble MT2 are detected in the conditioned medium (CM) of transfected cells (25,28,29). At least four functional forms of MT2 have been suggested, including an uncleaved full-length form, an activated form with the cleaved S/P domain attached to membrane-tethered stem region through disulfide bonds, the shed ectodomain where the entire ectodomain is released into the medium and the shed S/P domain (1,2,16,25,28,30). However, what forms of MT2 function in vivo and importantly how MT2 functions in vivo are still not known.
We are interested in what roles the different domains of MT2 play in the specificity and efficiency of substrate cleavage. The cytoplasmic domain of MT2 contains an endocytosis motif that mediates the internalization of cell-surface MT2 (31). Our previous studies indicated that this cytoplasmic domain also acts as an intracellular iron sensor to inhibit its degradation upon iron depletion (16,32). Stabilization of MT2 is expected to suppress hepcidin expression leading to increased iron availability to cells. The precise roles of the transmembrane domain and the stem region for the function of MT2 are less understood. A recent study shows that the CUB, LDLRA, and S/P domains are required for cleavage of HJV (33). MT2 interacts with HJV through its stem region (25). Numerous clinical studies report that the IRIDA-causing mutations in MT2 are located throughout the entire extracellular domain (34 -40). They all are lossof-function mutations. Interestingly, at least three IRIDA-causing mutants in the stem region of MT2, R271Q, T287N, and G442R, are reported to traffic to the cell surface and to cleave HJV similarly to WT MT2 (33,36,37). These observations sug-gest that the nonproteolytic activity mediated through its ectodomain plays an important role for the function of MT2.
In this study, we found that the stem region of MT2 determined the specificity and efficiency of substrate cleavage and that the transmembrane domain was essential for substrate cleavage at specific sites. In vivo studies showed that a functional MT2 required the stem region and transmembrane domain in addition to its S/P domain.

The stem region of MT2 is required for an efficient cleavage of substrates
We sought to explore the functions of the noncatalytic domains of MT2 by using cultured cells and MT2 Ϫ/Ϫ mice as models. We previously showed that MT2 cleaves multiple components of the hepcidin-induction pathway (9). Our pilot studies on two other membrane-anchored serine proteases, matriptase and prostasin, strongly indicated the critical roles of the nonproteolytic sequence of MT2 in the specific and efficient cleavage of substrates. 3 Matriptase is a close family member of MT2. Prostasin is a glycosylphosphatidylinositol (GPI)-anchored serine protease lacking a stem region (Fig. S1) (1,(41)(42)(43).
We generated four truncated forms of murine MT2 to determine the contributions of the stem region as well as transmembrane and cytoplasmic domains to the ability of MT2 to cleave its substrates. They are MT2 with deletion of the first 50 amino acids in the cytoplasmic domain (MT2-⌬Cyto), the entire MT2 ectodomain (MT2-Ecto), the MT2 S/P domain (MT2-S/P), and MT2 with deletion of the S/P domain (MT2-mask) (Fig. 1A). MT2-Ecto and MT2-S/P are the secreted forms. For the convenience of immunodetection, a FLAG/MYC epitope was added to these constructs as well as to most of other constructs used in this study ( Table 1). Addition of a C-terminal FLAG/MYC epitope to MT2 does not affect either its cleavage of substrates in transfected cells or the suppression of hepcidin expression in mice (9,16,25). All expressed proteins migrated on SDS-PAGE as predicted by their molecular masses (Fig. 1B). At least two distinct forms of MT2 were detected in the CM of cells expressing either full-length MT2 or MT2-⌬Cyto under reducing conditions (Fig. 1B, lanes 2, 3, and 6). The lower band at 34 kDa is consistent with the molecular mass of the S/P domain, because it corresponds to the band observed in the CM of cells expressing MT2-S/P (lanes 8 and 9). The higher molecular weight band is likely a shed fragment of the entire ectodomain. There were three distinct bands detectable in the CM of cells expressing MT2-Ecto, in which the top band corresponds to the uncleaved form (lane 7). Intriguingly, a distinct band was also detected in the CM of cells expressing mask-MT2 (lane 4), which suggests that another protease in the cells can cleave MT2. Together, these data indicate that MT2 undergoes active cleavage in cells.
To elucidate the roles of the MT2 stem region, we first compared the ability of MT2-Ecto and MT2-S/P to cleave BMP receptors. HEK293 cells were co-transfected with a constant amount of BMP receptor construct plasmid and increasing amounts of MT2-Ecto or MT2-S/P plasmid DNA. Full-length MT2 was included as a positive control, and enhanced GFP (EGFP) as a negative control. Because all BMP receptors have short extracellular domains at around 120 -150 amino acids in length and because of the lack of appropriate antibodies against their extracellular domains for a direct immunodetection of cleaved products, a decrease in molecular weight in Western blots using an anti-FLAG antibody was adopted as an indicator for MT2 cleavage. As shown in Fig. 2A, MT2-Ecto was able to cleave ALK3 similarly to full-length MT2 (lanes 3-5 versus 9 -11). In comparison, cleavage was detected only at the highest concentrations of MT2-S/P ( Fig. 2A, lanes 3-5 versus [12][13][14]. Biotinylation of the cell-surface proteins revealed that cells expressing MT2-Ecto had decreased cell-surface full-length ALK3 similarly to full-length MT2, and that the detected ALK3 bands were likely the cleaved products because of the mild band downshift (Fig. 2B, lanes 3 and 5). In contrast, expression of MT2-S/P only led to a small decrease in cell-surface ALK3 ( . Similarly, we also found that MT2-Ecto but not MT2-S/P, was able to cleave ALK2, ActRIIA, and Bmpr2 in cells and to decrease their levels at the cell surface as efficiently as full-length MT2 (Fig. S2, B-J). Together these results indicate that an efficient cleavage of BMP receptors by the S/P domain of MT2 requires its stem region.
We next examined the roles of the stem region in the cleavage of Hjv, Hfe, and Tfr2. Hjv was cleaved by both MT2 and the ubiquitously expressed furin (Fig. 3A) (44 -47). As predicted, a single furin-cleaved Hjv band migrating at ϳ45 kDa was detected in the CM when Hjv is expressed alone (Fig. 3B, lanes 1 and 2). Consistent with our previous studies (9), co-expression of full-length MT2 resulted in a decrease in furin-cleaved Hjv in CM with a concurrent detection of at least two MT2-
Similar to the cleavage of Hjv by MT2, MT2-Ecto was able to cleave Hfe and Tfr2 but it did not cleave these proteins with the same specificity/efficiency as full-length MT2. When Hfe was expressed alone, two major Hfe bands were evident (Fig. 3E,  lanes 1 and 2). The upper band represents the fully processed mature form that is mainly detected at the cell-surface (9). Expression of MT2 decreased the upper band, and led to the detection of a ϳ17 kDa Hfe fragment in cell lysates (Fig. 3E, lanes [3][4][5]. In comparison, MT2-Ecto decreased the fully processed Hfe upper band (55 kDa) in cell lysates and abolished cell-surface Hfe more effectively than full-length MT2 (Fig. 3, E, lanes 9 -11, and F, lane 6; Fig. S2K). However, compared with the ϳ17 kDa fragment of Hfe detected in cells expressing MT2, no cleaved products of Hfe were detected in cells that expressed MT2-Ecto (Fig. 3E, lanes 3-5 versus 9 -11). MT2-S/P failed to detectably cleave Hfe (Fig. 3, E, lanes 12-14, and F, lane 7). MT2-Ecto also markedly reduced the levels of both cellular and cell-surface Tfr2 (Fig. 3, G, lanes 9 -11, and H, lane 5). In contrast to a distinctly cleaved Tfr2 band in CM by fulllength MT2, no cleaved Tfr2 products were detectable by MT2-Ecto expression (Fig. 3G, lanes 3-5 versus 9 -11). Different from Hfe, MT2-S/P also decreased cellular and cellsurface Tfr2 to a lesser extent than MT2-Ecto ( Fig. 3, G, lanes 12-14, H, lane 6; Fig. S2L). Taken together, cleavage studies of MT2-Ecto and MT2-S/P indicate that the stem region of MT2 is indispensable for the proteolytic activity as well as for the efficient substrate cleavage. These results also suggest that membrane tethering is essential for MT2 to cleave substrates at specific residues.

Plasma membrane tethering is required for substrate-specific cleavage by MT2
To determine the roles of the transmembrane and cytoplasmic domains of MT2 for substrate cleavage, we examined the specificity of cleavage by MT2-⌬Cyto in HEK293 cells. In contrast to MT2-Ecto, MT2-⌬Cyto retained the capability to cleave Hjv, Hfe, and Tfr2 in a manner similar to MT2. Expression of MT2-⌬Cyto resulted in the detection of two distinct Hjv fragments in CM, which were also seen with the co-expression of MT2, but not MT2-Ecto (Fig. 3B, lanes 6 -8). We detected similar-sized cleavage products of Hfe in cell lysates and Tfr2 in CM with MT2-⌬Cyto and MT2 expression (Fig. 3, E, lanes 6 -8, and G, lanes 6 -8). Importantly, MT2-⌬Cyto was able to reduce the levels of Hfe and Tfr2 both in cell lysates and on the cell surface similarly to MT2 (Fig. 3, E-H). These results indicate that the transmembrane domain of MT2 is essential for the specificity of cleavage of Hjv, Hfe, and Tfr2, whereas the cytoplasmic domain is dispensable.
We also found that MT2-⌬Cyto cleaved ALK2, ALK3, ActRIIA, and Bmpr2 as potently as MT2 and MT2-Ecto by examination of decreased full-length forms in cell lysates and on the cell surface ( Fig. 2; Fig. S2, B-J). Because of the lack of appropriate antibodies for the BMP receptors and positioning of the epitope tags on the C terminus, we were unable to detect receptor fragments in the CM. Rather we had to rely on the changes in molecular weight of the receptors on the plasma membrane. The results do support the idea that the cytoplasmic domain of MT2 is dispensable for the cleavage. A, an equal amount of pCMV6-ALK3 (2 g) was co-transfected with an increasing amount of pCMV6-MT2 or MT2 truncates (0.25, 1.0, and 2.0 g), and decreasing amount of pEGFP-N1 (1.75, 1.0, and 0 g) into HEK293 cells in 12-well plates. Co-transfection with 2 g of pEGFP-N1 (GFP) was included as negative controls. After 48 h of transfection, ϳ150 g of cell lysate proteins was subjected to SDS-PAGE and immunodetection by using anti-FLAG (for MT2 and ALK3), ␤-actin, and GFP antibodies. Each panel was cropped from the same image. B, effects of truncated MT2 on cell-surface ALK3. pCMV6-ALK3 was co-transfected with an equal amount of pEGFP-N1 (GFP), pCMV9-MT2, ⌬Cyto, mask, Ecto, or S/P plasmid DNA into HEK293 cells. At about 48 h after transfection, cell-surface proteins were biotinylated at 4°C, followed by pulldown of the biotinylated proteins using streptavidin-agarose beads. The entire eluted cell-surface proteins and ϳ25% of input lysate were subjected to SDS-PAGE and immunodetection using anti-FLAG (for ALK3), NaK-ATPase, and TfR1 antibodies. n.s., nonspecific bands. All experiments were repeated at least three times with consistent results.

Trafficking of MT2 to the plasma membrane is required for MT2-induced cleavage of substrates
To seek insight into the mechanisms that cause the differences between MT2 and MT2-Ecto, we analyzed the cellular location of cleavage by taking advantage of two inhibitors, the hepatocyte growth factor-activator inhibitor-2 (HAI-2) and aprotinin. HAI-2 is a membrane-associated serine protease inhibitor that primarily resides intracellularly within the endo- Co-transfection of pCMV9-Hjv with truncated MT2 and pEGFP-N1 into HEK293 cells was performed as described above in the legends to Fig. 2A. At about 30 h post-transfection, medium was changed to Opti-MEM, 1% FCS. After another 18 h of incubation, about 150 g of cell lysate proteins and the TCA-precipitated proteins from ϳ600 l of the CM were subjected to SDS-PAGE and immunodetection by using anti-FLAG (for Hjv and MT2), ␤-actin, GPF, and HJV antibodies. Each panel was cropped from the same image. C, effects of truncated MT2 on cell-surface Hjv. HEK293 cells were co-transfected with pCMV9-Hjv, and pEGFP-N1 or pCMV6-MT2 or truncates at 1:1 ratios of plasmid DNA. At about 48 h after transfection, cell-surface proteins were biotinylated at 4°C, followed by pulldown of the biotinylated proteins using streptavidin-agarose beads. The eluted cell-surface proteins and ϳ10% of input lysate was subjected to SDS-PAGE and immunodetection by using anti-FLAG antibody. D, quantification of cell-surface Hjv bands in C. E, cleavage of Hfe by truncated MT2. HEK293 cells in 12-well plates were co-transfected with constant amounts of pCMV6-Hfe (1.33 g) and pJB-1-B2M (1.33 g), an increasing amount of pCMV6-MT2 or truncated MT2 (0.33, 0.67, and 1.33 g), and decreasing amount of pEGFP-N1 (1.0, 0.66, and 0 g). Co-transfection with 1.33 g of pEGFP-N1 (GFP) was included as negative controls. After 48 h of transfection, ϳ150 g of cell lysate proteins was subjected to SDS-PAGE and immunodetection by using anti-FLAG (for Hfe and MT2), ␤-actin, and GFP antibodies. Each panel was cropped from the same image. F, effects of truncated MT2 on cell-surface Hfe. HEK293 cells were co-transfected with pCMV6-Hfe, pJB-1-B2M, and pEGFP-N1 or pCMV6-MT2 or truncates at 1:1:1 ratios of plasmid DNA. At about 48 h after transfection, cell-surface proteins were biotinylated as described above in C. The eluted cell-surface proteins and ϳ30% of input lysate were subjected to SDS-PAGE and immunodetection by using anti-FLAG antibody. G, cleavage of Tfr2 by truncated MT2. Co-transfection was performed as described above for Hjv in B. At about 30 h post-transfection, medium was changed to Opti-MEM, 1% FCS. Analysis was performed after another 18 h of incubation. About 150 g of cell-lysate proteins and the TCA-precipitated proteins from ϳ600 l of the CM were subjected to SDS-PAGE and immunodetection using anti-FLAG (for MT2), Tfr2, ␤-actin, and GPF antibodies. Each panel was cropped from the same image. H, effects of truncated MT2 on cell-surface Tfr2. Co-transfection, biotinylation, and immunodetection were performed essentially the same as described above in C for Hjv, except that Tfr2 was detected by a rabbit antibody against mouse Tfr2 extracellular domain. Ctrl, HEK293 cells with no transfection. n.s., nonspecific bands. All experiments were repeated at least three times with consistent results.

Mechanistic studies of matriptase-2 activity
plasmic reticulum (48). Aprotinin is a soluble membrane-impermeable serine protease inhibitor consisting of 58 amino acids. Both inhibit the proteolytic activity of MT2 (47,49,50). We reasoned that if the substrate cleavage is inhibited by both inhibitors, it would indicate that the cleavage takes place after MT2 reaches the plasma membrane. Conversely, if the cleavage is blocked by HAI-2 but not by aprotinin, it would indicate that the cleavage occurs in the biosynthetic pathway. As expected, co-expression of HAI-2 completely prevented the cleavage of ALK2, ALK3, ActRIIA, Bmpr2, Hfe, Tfr2, and Hjv by MT2 and MT2-⌬Cyto (Fig. 4, A-G, lane 4). The inhibitory effects of HAI-2 on the cleavage by MT2-Ecto were less efficient (Fig. 4,   A-G, lane 4 versus 8). These data support HAI-2 as a potent inhibitor of MT2.
We examined the cellular location of substrate cleavage by full-length MT2 and MT2-⌬Cyto. Transfected cells were treated with aprotinin (ϳ10 M) for ϳ18 h. Aprotinin was able to largely block the cleavage of ALK2, ALK3, ActRIIA, Hjv, Hfe, and Tfr2 by full-length MT2 (Fig. 5, A-F, lane 5, and D, lane 7), which indicates that cleavage occurs after the substrates reach the plasma membrane. Consistent with our previous studies (46), the cleavage of Hjv by MT2 was not inhibited by the furin convertase inhibitor (FCI) (Fig. 5D, lane 6). In comparison, only a mild inhibitory effect was detected for MT2 cleavage of . pCMV6-ALK2, ALK3, ActRIIA, or Bmpr2 were co-transfected into HEK293 cells with pCMV6-MT2 or truncates and pCMV6-HAI-2 or pEGFP-N1 at 1:1:1 ratios of plasmid DNA. After 48 h of transfection, ϳ150 g of cell lysate proteins were subjected to SDS-PAGE and immunodetection using anti-FLAG (for MT2, HAI2, ALK2, ALK3, ActRIIA, and Bmpr2), ␤-actin, and GFP antibodies. E, HAI-2 inhibition of cleavage for Hfe. HEK293 cells were co-transfected with pCMV6-Hfe, pJB-1-B2M, pCMV6-MT2, or truncates, and HAI-2 or pEGFP-N1 at 1:1:1:1 ratios of plasmid DNA. Analysis was performed at 48 h post-transfection as described above in A-D. F, HAI-2 inhibition of cleavage for Tfr2. Co-transfection of Tfr2 with MT2 truncates and HAI-2 or pEGFP-N1 were performed essentially the same as described above for A-D. At about 30 h post-transfection, medium was changed to Opti-MEM, 1% FCS. Analysis was performed after another 18 h of incubation. About 150 g of cell-lysate proteins and the TCA-precipitated proteins from ϳ600 l of CM were subjected to SDS and immunodetection. Tfr2 was detected by using a rabbit anti-Tfr2 antibody. G, HAI-2 inhibition of cleavage for Hjv. Co-transfection of HEK293 cells with Hjv, truncated MT2, and HAI-2 was performed as described above in A-D. At about 30 h post-transfection, medium was changed to Opti-MEM, 1% FCS. After another 18 h of incubation, about 150 g of cell-lysate proteins and the TCA-precipitated proteins from ϳ600 l of CM were subjected to SDS-PAGE and immunodetection using anti-FLAG (for MT2, Hjv, and HAI-2), ␤-actin, GPF, and HJV antibodies. n.s., nonspecific bands. All experiments were repeated at least three times with consistent results.

Mechanistic studies of matriptase-2 activity
Bmpr2 (Fig. 5G, lane 5), suggesting that its cleavage predominantly takes place in the biosynthetic pathway. When the effects of aprotinin on MT2-⌬Cyto were compared with fulllength MT2, we detected similar extents of inhibition for the cleavage of ALK2, Hjv, and Hfe, a relatively milder inhibition for ALK3, ActRIIA, and Tfr2, and no obvious effect for Bmpr2 (Fig. 5, A-G; Fig. S3, A-G). These observations suggest that the cytoplasmic domain of MT2 facilitates the cleavage of ALK3, ActRIIA, Tfr2, and Bmpr2 at the cell surface or after retrograde trafficking.
In contrast to full-length MT2 and MT2-⌬Cyto, incubation with aprotinin failed to inhibit cleavage by MT2-Ecto for all the tested substrates (Fig. 5, A-G, lane 9, and D, lane 13; Fig. S3, A-G). Similar results were also obtained by using another membrane-impermeable serine protease inhibitor, leupeptin (data not shown). These data indicate that MT2-Ecto cleaves

Mechanistic studies of matriptase-2 activity
substrates primarily within cells before reaching the plasma membrane, which is distinct from the membrane-anchored full-length MT2 and MT2-⌬Cyto (Table S1). Taken together, these observations suggest that the transmembrane domain is essential for MT2 to cleave substrates after they traffic to the plasma membrane and that the cytoplasmic domain facilitates these processes.

Plasma membrane tethering is essential for MT2 suppression of hepcidin expression in vivo
To determine the contributions of different MT2 domains to its biological function, we expressed MT2-⌬Cyto, MT2-Ecto, and MT2-S/P in the liver of MT2 Ϫ/Ϫ mice by AAV8 vector for replacement studies. MT2 is predominantly expressed in the hepatocytes of liver. The AAV8-MT2 constructs contained a hepatocyte-specific promoter to mimic the endogenous expression of MT2. AAV8-MT2 and MT2-mask constructs were used as positive and negative controls, respectively. All MT2 Ϫ/Ϫ mice used in the studies had similar extents of iron deficiency and anemia with no obvious gender-related difference (9). Eight-week-old MT2 Ϫ/Ϫ mice of both genders were injected intraperitoneally with AAV8-MT2 or AAV8-MT2-⌬Cyto at ϳ8 ϫ 10 11 and ϳ4 ϫ 10 12 genome-particles per mouse, and AAV8-MT2-mask, AAV8-MT2-Ecto, or AAV8-MT2-S/P at ϳ4 ϫ 10 12 genome-particles per mouse, or PBS vehicle as control. Previous studies indicated that expression of either GFP or glutaryl CoA-dehydrogenase, which are unrelated to iron metabolism, by AAV8 vector in mice had no effects on iron homeostasis (47,51). Age-matched WT littermates were included as additional controls.
The AAV8-mediated expression levels of exogenously administered MT2 in the liver were analyzed by quantitative real-time RT-PCR (qRT-PCR) and Western blotting. The mRNA levels of full-length MT2 and MT2-⌬Cyto in the low dosage groups were comparable with those of WT mice, and about 10-and 6.3-fold higher in the high dosage groups, respectively (Fig. 6A). The mRNA levels for MT2-mask, MT2-Ecto, and MT2-S/P groups were all higher than WT counterparts. Endogenous MT2 protein was not readily detectable using the antibody that we generated against human MT2 so we put a FLAG/MYC epitope on the C terminus of the MT2 constructs. All MT2 forms were detected in the liver extracts at the expected molecular weights by Western blotting using an anti-FLAG antibody (Fig. 6B, lanes 5-18). The relative protein levels between the low and high dosage groups for MT2 and MT2-⌬Cyto were proportional to mRNA levels. We also measured the secreted MT2 in serum by Western blotting. Only MT2-S/P and uncleaved MT2-Ecto were detected (Fig. 6B, lower panels, lanes [13][14][15][16]. No MT2 was detected in the serum of other groups, suggesting that there was either no cleavage or the levels fell below the limit of detection. These are in contrast to the detection of cleaved forms in the CM of transfected cells (Fig.  1B). Because the same MT2 constructs were used for these studies, these observations suggest that the post-translational processing of MT2 might be different between the transfected cells and in vivo in the liver.
Functional analysis indicates that only MT2-⌬Cyto was able to rescue the defects of MT2 Ϫ/Ϫ mice as robustly as full-length MT2 (Fig. 6, C-E). Consistent with our previous studies (9), expression of full-length MT2 at both dosages was able to fully correct the low serum iron and high hepcidin expression status (Fig. 6, C and D), and greatly ameliorated the anemia by significantly increasing the levels of hemoglobin, hematocrit, mean cell volume, and mean corpuscular hemoglobin ( Table 2). The parallel decrease in Inhibitor of DNA binding-1 (Id1) mRNA with hepcidin mRNA (Fig. 6, D and E) strengthened the idea that MT2 suppresses hepcidin expression by inhibiting the BMP signaling as previously reported (8). Id1 is a direct downstream target of the BMP signaling. Compared with full-length MT2, similar results were seen when MT2-⌬Cyto were expressed (Fig. 6, C-E; Table 2). There was no significant reduction for the high RBC count in MT2 Ϫ/Ϫ mice upon expression of either full-length MT2 or MT2-⌬Cyto (Table 2). This is likely because the analysis was conducted after 3 weeks of viral administration and that the half-life of mouse RBC is about 22 days. These results indicate that the cytoplasmic domain of MT2 is dispensable for the suppression of hepcidin expression under these conditions.
In contrast to MT2-⌬Cyto, no significant rescue was observed when high levels of MT2-Ecto or MT2-S/P were expressed. Rather they behaved similarly to the negative control MT2-mask (Fig. 6, C-E; Table 2). These observations indicate that the transmembrane domain is essential for the biological function of MT2. Together, results from in vivo studies support the idea that both the transmembrane domain and the entire ectodomain constitute the minimal requirements for a biologically functional MT2.

Discussion
In this study, we investigated the contribution of different domains of MT2 to its proteolytic activity and ability to suppress both BMP-signaling and hepcidin expression. Our results  4G. E, effects of aprotinin on the cleavage of Hfe by truncated MT2. HEK293 cells were co-transfected with pCMV6-Hfe, pJB-1-B2M, and pEGFP-N1 or pCMV6-MT2 or truncates at 1:1:1 ratios of plasmid DNA. Aprotinin treatment and the immunodetection of Hfe, ␤-actin, and GFP in cell lysates were performed as described above in A-C and in the legend to Fig. 4E. F, effects of aprotinin on the cleavage of Tfr2 by truncated MT2. Co-transfection and immunodetection were performed essentially the same as described above for Hjv in D except that Tfr2 was detected by a rabbit anti-Tfr2 antibody. n.s., nonspecific bands. G, effects of aprotinin on the cleavage of Bmpr2 by truncated MT2. Co-transfection and immunodetection were performed essentially the same as described above for A-C. Ctrl, HEK293 cells with no transfection. All experiments were repeated at least three times with consistent results.

Mechanistic studies of matriptase-2 activity
indicate that the stem region determines the specificity and efficiency for substrate cleavage. The transmembrane domain allows the cleavage of substrates at specific residues at the cell surface. In vivo studies reveal that suppression of BMP-signaling and hepcidin expression requires both the entire ectodomain and the transmembrane domain of MT2.
By taking advantage of the differences between the MT2 constructs, we demonstrated a determinant role of the nonproteolytic sequence of MT2 for specific cleavage of BMP receptors, Hfe, Tfr2, and Hjv. We showed that the stem region of MT2 is required for efficient substrate cleavage (Figs. 2 and 3; Fig. S2). This is likely due to the possibility that the stem region is Figure 6. Only the intracellular domain of MT2 is dispensable for suppression of hepcidin expression in MT2 ؊/؊ mice. Eight-week old MT2 Ϫ/Ϫ mice on B6/129S mixed background were intraperitoneally injected with AAV8-MT2 or ⌬Cyto at ϳ8 ϫ 10 11 (ϩ) and ϳ4 ϫ 10 12 (ϩϩ) genome-particles per mouse, and AAV8-mask, Ecto, or S/P at ϳ4 ϫ 10 12 genome-particles per mouse, or PBS vehicle as control (Ϫ). Animals were euthanized for analysis after 3 weeks of injection. Age-matched WT mice on the same background were included as additional controls. Each group consists of at least 10 animals with similar numbers of male and female. A, qRT-PCR analysis of hepatic MT2 mRNA. B, representative images for Western blot analysis of introduced MT2 protein in the liver extracts (250 g of protein) and serum (100 l) using an anti-FLAG antibody. Two images with different intensities (top two panels) for the liver extracts were presented. ␤-Actin was used as a loading control. n.s., nonspecific band. C, serum iron (Fe) concentrations. D and E, qRT-PCR analysis of hepatic hepcidin and Id1 mRNA. All qRT-PCR results are expressed as the amount relative to that of ␤-actin for each sample. The mean Ϯ S.D. are presented.

Mechanistic studies of matriptase-2 activity
involved in the interaction with all substrates. MT2 interacts with HJV through this region (25). Consistent with this idea, the stem region of MT2 harbors numerous IRIDA-causing mutations (34,35). Our data also suggest that the shed forms of soluble MT2 lack physiological relevance to iron homeostasis. We detected at least two soluble forms of MT2 from transfected cells, which corresponds to the entire ectodomain and S/P domain, respectively (Fig. 1B). A previous study reports two additional cleaved products (28). In this study, we indeed detected a robust activity for soluble MT2 ectodomain and a minor activity for soluble S/P domain (Figs. 2 and 3; Fig. S2), which indicate that all shed forms of MT2 do possess proteolytic activities. However, compared with full-length MT2, the soluble MT2 ectodomain lacks the activity to cleave Hjv, Hfe, and Tfr2 at specific residues (Fig. 3), and more importantly it mainly cleaves substrates in the biosynthetic pathway (Figs. 4 and 5). These observations suggest that these soluble forms act differently from full-length MT2. Therefore, it is unlikely that the function of MT2 is mediated through the shed MT2. Rather the shedding of MT2 may represent the pathway of clearance from the cells.
Our studies indicate that MT2 tethering to the plasma membrane is essential for its cleavage specificity (Fig. 3). The different localization of activation between the MT2-Ecto and the full-length MT2 (Figs. 4 and 5) suggests that MT2 mainly cleaves substrates after reaching the plasma membrane, and imply that the tethering of MT2 to the membrane protects MT2 zymogen from being activated before it reaches the cell surface. This idea is supported by previous studies suggesting that the cleavage of HJV by MT2 either occurs at cell-surface cells (25) or requires retrograde trafficking from the plasma membrane (52). This specific activation process possesses physiological significance, because it would allow the modulation of MT2 activities by the factors in extracellular environment. These observations support a model that the transmembrane domain acts to ensure the activation and substrate cleavage at the specific sites of cells. Together with findings that the cytoplasmic domain of MT2 facilitates the cleavage processes in this study (Fig. 5), our data indicate that the specific and efficient cleavage of substrates at the precise sites of cells requires the entire sequence of MT2.
Expression of the same constructs of MT2 in the livers of MT2 Ϫ/Ϫ mice showed some differences in the behavior of MT2. The in vivo studies did not detect MT2 shedding in the liver. In contrast to transfected cells, no soluble MT2 was detectable in the serum when either full-length MT2 or MT2-⌬Cyto was expressed in the liver of MT2 Ϫ/Ϫ mice (Figs. 1B and   6B). These discrepancies are unlikely due to the use of HEK293 cells, a nonhepatoma cell line, for the in vitro study. A previous study shows similar post-translational processing of human MT2 between HEK293 and hepatoma cells (28). We found that the substrate cleavage by mouse MT2 in Hep3B cells (a hepatoma cell line) is similar to that detected in HEK293 cells. 3 Thus, it is likely that other environmental factors may contribute to the differential post-translational processing of MT2 between in vitro and in vivo systems. Alternatively, the hepatic MT2 may undergo similar autocleavage and shedding as seen in transfected cells. But the levels of cleaved MT2 fragments fall below the limit of immunodetection because of the high dilution into the entire blood volume and/or a rapid clearance from the circulation.
Our in vivo studies strongly indicate that the membraneassociated MT2s are the biologically active forms, which suppress BMP signaling and hepcidin expression in the liver ( Fig. 6; Table 2). There are two predicted membrane-associated forms, the uncleaved full-length MT2 zymogen and the activated form with the cleaved S/P domain attached to the membrane-tethered stem region through a disulfide bond. Previous studies implicate the latter as the most catalytically active form (25,28). However, our data do not rule out the possibilities that the MT2 zymogen is also catalytically active, because both MT1 and prostasin zymogens are functional in vivo (41,42,53). Nevertheless, our studies do support the idea that soluble forms of MT2 lack the biological activity, because neither MT2-Ecto nor MT2-S/P was able to rescue the defects of MT2 Ϫ/Ϫ mice. In transfected cells, we found that soluble MT2-Ecto was able to cleave substrates as robustly as full-length MT2 (Figs. 2 and 3; Fig. S2), but compared with membrane-associated MT2, it lacked specificity and efficiency in cleavage (Figs. [2][3][4][5]. Thus, we speculate that membrane association would allow MT2 to be activated after reaching the plasma membrane to specifically remove the components of the hepcidin-induction pathway. Alternatively, MT2 may act as a membrane-associated zymogen to suppress hepcidin expression by binding to the substrates, which inhibits their functions. These hypotheses remain to be tested in future studies. Finally, the finding that the cytoplasmic domain is dispensable for the suppression of hepcidin in this study (Fig. 6) is not in conflict with our previous observations showing that the cytoplasmic domain acts as an intracellular iron sensor to inhibit its degradation under iron depletion (16,32). The severe iron deficiency and anemia of MT2 Ϫ/Ϫ mice requires a maximal activity of MT2 to suppress hepcidin expression.

Mechanistic studies of matriptase-2 activity
In summary, this is the first study, to our knowledge, to systemically investigate the post-translational processing and biological functions of MT2 by using both the in vitro transient expression systems and the in vivo expression systems. Our studies indicate that the specific and efficient cleavage of substrates at the precise site of cells requires the entire sequence of MT2, and that the shed MT2 may represent the products of degradation. In vivo studies demonstrate that suppression of hepcidin expression by MT2 requires its transmembrane domain and entire ectodomain. Our findings have important significance for further elucidating the underlying mechanism by which MT2 suppresses hepcidin expression, and for characterizing the defects of IRIDA-causing mutations in different subdomains of MT2.
We generated three truncated murine MT2 constructs (Fig.  1A). MT2-⌬Cyto, which lacks the first 50 amino acids in the cytoplasmic domain, was generated by site-directed mutagenesis using pCMV6-MT2 as a template and the QuikChange kit (Stratagene). MT2-Ecto and MT2-S/P constructs were generated by addition of an Ig -chain leader sequence (SS) immediately before Phe-77 and Asp-567 residues, respectively, using pCMV6-MT2 as a template and PCR with the Gibson Assembly Master Mix (New England Biolabs). MT2-mask was generated in our previous studies (9). The sequences in all these constructs were verified by DNA sequencing.
AAV8-⌬Cyto-MT2, MT2-Ecto, and MT2-S/P constructs were generated by subcloning into an AAV8 construct containing a strong liver-specific promoter as described in our previous study (9,54). The liver-specific promoter is a combination of two copies of a human ␣ 1 -microglobulin/bikunin enhancer and the promoter from the human thyroid hormone-binding globulin gene. AAV8-MT2 and MT2-mask constructs were generated in our previous study (9). AAV8 viral vectors were produced at the Molecular Virology Support Core, OHSU.

Cell lines and transfection
HEK293 cells were obtained from ATCC and maintained in DMEM, 10% fetal calf serum (FCS). We used transient cotransfection with Lipofectamine 2000 (Invitrogen) and polyethyleneimine transfection reagent (Polysciences, Inc., 24765-1) to determine the cleavage of ALK2, ALK3, ActRIIA, Bmpr2, Hjv, Hfe, and Tfr2 by truncated MT2 with or without HAI-2. The detailed procedures were described in the figure legends. All cells were transfected with equal amounts of total plasmid DNA. After 48 h of transfection, cell lysates were collected, and protein concentrations were determined by Pierce BCA protein assay kit (Thermo). Lysate protein (ϳ150 g) was then subjected to SDS-PAGE. We used Ponceau S staining to ensure equal transfer of proteins onto the nitrocellulose membrane. ALK2, ALK3, ActRIIA, Bmpr2, Hjv, and Hfe were immunodetected by using an HRP-coupled mouse anti-FLAG M2 IgG, chemiluminescence, and the c600 Western blotting imaging system (Azure Biosystems, Inc.). Tfr2, EGPF, TfR1, and ␤-actin in cell lysates were probed with rabbit anti-mouse Tfr2 extracellular domain (number 25257), mouse anti-GFP (Pro-teinTECH), mouse anti-TfR1 (Invitrogen), and mouse anti-␤actin, respectively, immunodetected by using an Alexa Fluor 680 goat anti-rabbit secondary antibody (1:10,000; Invitrogen) or an Alexa Fluor 800 goat anti-mouse secondary antibody (1:10,000; Invitrogen), and visualized by using an Odyssey IR Imaging System (Licor). For most of the transfection studies, images with sequential probing for ␤-actin and GFP were presented.
To detect the cleaved Hjv and Tfr2 in CM, medium was changed to Opti-MEM, 1% FCS at ϳ30 h post-transfection. CM was collected after ϳ18 h of incubation. Proteins in CM were first precipitated by using 6% TCA, followed by resuspending the protein pellets with 1ϫ Laemmli buffer and subjecting to SDS-PAGE. Hjv in CM was probed with both an HRP-coupled mouse anti-FLAG M2 IgG and a rabbit anti-HJV antibody generated against the entire sequence of human HJV (46). The latter was able to detect the cleavage products of Hjv by both furin and MT2. Tfr2 in CM was detected as described above for the cell lysate Tfr2.
To test the effects of aprotinin (G-Biosciences) on MT2 cleavage of all substrates and FCI (Alexis, San Diego, CA) on MT2 cleavage of Hjv, aprotinin at 5 M and FCI at 1 M were added to the medium at ϳ30 h post-transfection. Analysis was performed after ϳ18 h of incubation as described above. The intensities of ALK2, ALK3, ActRIIA, Bmpr2, Hjv, Hfe, and Tfr2 bands in Western blots were quantified by using either Licor or ImageJ software.

Animal studies
All animal procedures were approved by the Oregon Health & Science University, Department of Comparative Medicine (OHSU DCM). We purchased heterozygous MT2 mutant mice on B6/129 background from Lexion/Mutant Mouse Resource and Research Center (9). Both WT and homozygous MT2 Ϫ/Ϫ mice were generated by breeding heterozygous MT2 mutant mice on mixed background. Animals were genotyped by PCR using genomic DNA from toe clips. All MT2 Ϫ/Ϫ mice used in the studies were within two generations of breeding and had similar extents of iron deficiency and anemia. We previously showed no obvious gender-related difference with respect to iron deficiency, anemia, and high hepatic hepcidin expression in MT2 Ϫ/Ϫ mice (9).

Mechanistic studies of matriptase-2 activity
We used MT2 Ϫ/Ϫ mice to examine the function of truncated MT2 in vivo. Eight-week-old MT2 Ϫ/Ϫ mice of both genders were intraperitoneally injected with AAV8-MT2, AAV8-MT2-⌬Cyto, AAV8-MT2-Ecto, AAV8-MT2-S/P, or AAV8-MT2mask at dosages indicated in the text and figure legends. Injection of PBS vehicle was included as control. Three weeks later, mice were euthanized for analysis. Blood was collected by cardiac puncture for serum iron and hematology analysis. Livers were rapidly removed, snap-frozen in liquid nitrogen, and then stored at Ϫ80°C for qRT-PCR and Western blotting. Age, gender, and background-matched WT littermates were included as additional controls. Each group consisted of at least 10 animals. The viral vector stocks were handled according to Biohazard Safety Level 2 guidelines published by NIH.

qRT-PCR
qRT-PCR analysis of ␤-actin, MT2, hepcidin, and Id1 transcripts in the liver were conducted as previously described (9,54). All primers were verified for linearity of amplification. The results are expressed as the amount relative to that of ␤-actin for each sample.

Blood parameters and serum iron assays
Blood parameters for MT2 Ϫ/Ϫ mice were measured using Hemavet 950 (Drew Scientific). Serum iron concentrations were measured using a serum iron/TIBC Reagent Set (Teco Diagnostics, Anaheim, CA).

Immunodetection of hepatic MT2 and serum MT2 in mice
About 250 g of liver extract proteins were separated by SDS-PAGE under reducing conditions. We used Ponceau S staining to ensure equal transfer of proteins onto the nitrocellulose membrane. The AAV8-introduced exogenous MT2 was probed directly by using an HRP-coupled mouse anti-FLAG M2 IgG (1:10,000; Sigma) and chemiluminescence (Super Signal, Pierce). ␤-Actin was immunodetected with mouse anti-␤actin (1:5,000; Sigma), immunodetected by using the corresponding secondary antibody, and visualized by Licor.
The secreted MT2 in serum (ϳ100 l) was first pulled down by using anti-FLAG MT2-agarose affinity gel (Sigma; A-1205). The eluates were then subjected to SDS-PAGE and immunodetection using an HRP-coupled mouse anti-FLAG M2 IgG.

Statistical analysis
One-way analysis of variance and Tukey's post-test were used to analyze the data from animal studies.