The Protease Inhibitor HAI-2, but Not HAI-1, Regulates Matriptase Activation and Shedding through Prostasin*

Background: Hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2 may regulate matriptase trafficking and activation. Results: Ablation of endogenous HAI-2, but not HAI-1, causes loss of epithelial matriptase in vivo and in vitro, due to uncontrolled prostasin-dependent activation and shedding. Conclusion: HAI-2, but not HAI-1, regulates matriptase cell surface expression level. Significance: The study identifies HAI-2 as a principal regulator of prostasin-mediated matriptase activation. The membrane-anchored serine proteases, matriptase and prostasin, and the membrane-anchored serine protease inhibitors, hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2, are critical effectors of epithelial development and postnatal epithelial homeostasis. Matriptase and prostasin form a reciprocal zymogen activation complex that results in the formation of active matriptase and prostasin that are targets for inhibition by HAI-1 and HAI-2. Conflicting data, however, have accumulated as to the existence of auxiliary functions for both HAI-1 and HAI-2 in regulating the intracellular trafficking and activation of matriptase. In this study, we, therefore, used genetically engineered mice to determine the effect of ablation of endogenous HAI-1 and endogenous HAI-2 on endogenous matriptase expression, subcellular localization, and activation in polarized intestinal epithelial cells. Whereas ablation of HAI-1 did not affect matriptase in epithelial cells of the small or large intestine, ablation of HAI-2 resulted in the loss of matriptase from both tissues. Gene silencing studies in intestinal Caco-2 cell monolayers revealed that this loss of cell-associated matriptase was mechanistically linked to accelerated activation and shedding of the protease caused by loss of prostasin regulation by HAI-2. Taken together, these data indicate that HAI-1 regulates the activity of activated matriptase, whereas HAI-2 has an essential role in regulating prostasin-dependent matriptase zymogen activation.

The membrane-anchored serine proteases, matriptase and prostasin, and the membrane-anchored serine protease inhibitors, hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2, are critical effectors of epithelial development and postnatal epithelial homeostasis. Matriptase and prostasin form a reciprocal zymogen activation complex that results in the formation of active matriptase and prostasin that are targets for inhibition by HAI-1 and HAI-2. Conflicting data, however, have accumulated as to the existence of auxiliary functions for both HAI-1 and HAI-2 in regulating the intracellular trafficking and activation of matriptase. In this study, we, therefore, used genetically engineered mice to determine the effect of ablation of endogenous HAI-1 and endogenous HAI-2 on endogenous matriptase expression, subcellular localization, and activation in polarized intestinal epithelial cells. Whereas ablation of HAI-1 did not affect matriptase in epithelial cells of the small or large intestine, ablation of HAI-2 resulted in the loss of matriptase from both tissues. Gene silencing studies in intestinal Caco-2 cell monolayers revealed that this loss of cell-associated matriptase was mechanistically linked to accelerated activation and shedding of the protease caused by loss of prostasin regulation by HAI-2. Taken together, these data indicate that HAI-1 regulates the activity of activated matriptase, whereas HAI-2 has an essential role in regulating prostasin-dependent matriptase zymogen activation.
Epithelial development and homeostasis depends on a complex system of membrane-anchored trypsin-like serine proteases, including the type II transmembrane serine protease, matriptase, the glycophosphatidylinositol (GPI) 2 -anchored serine protease, prostasin, and the two Kunitz-type transmembrane serine protease inhibitors, hepatocyte growth factor activator inhibitor (HAI)-1 and HAI-2 (also known as Spint1 and Spint2, respectively) (reviewed in Ref. 1).
Matriptase is a 95-kDa type II transmembrane glycoprotein composed of a cytoplasmic N-terminal peptide, a signal anchor that functions as a single-pass transmembrane domain, a sea urchin sperm protein, enteropeptidase, agrin (SEA) domain, two complement C1r/s urchin embryonic growth factor and bone morphogenetic protein-1 (CUB) domains, four low-density lipoprotein receptor class A (LDLA) domains, and a trypsin-like serine protease domain (2)(3)(4). The FSEG 149 -SVIA peptide bond in the SEA domain is spontaneously hydrolyzed during matriptase biosynthesis, but the newly formed N-and C-terminal halves of the protease are held together by strong non-covalent interactions within the SEA domain. This SEA domain auto-hydrolysis is a prerequisite for the subsequent activation site cleavage of matriptase after a conserved Arg 614 , in the serine protease domain within the amino acid sequence RQAR 614 -VVGG (5)(6)(7), which leads to the formation of the active protease. Activated matriptase subsequently is shed from the plasma membrane, facilitated by the formation of inhibitor complexes with HAI-1 (8,9).
Prostasin is a 40-kDa GPI-anchored glycoprotein. It is synthesized as a zymogen that is converted to an active protease by a proteolytic cleavage after Arg 44 within the amino acid sequence PQPR 44 -ITGG (10 -12).
HAI-1 and HAI-2 are closely related integral membrane proteins that are comprised of two extracellular Kunitz-type inhib-* This work was supported by the NIDCR, National Institutes of Health, Intra-itory domains that are followed by a C-terminal transmembrane domain. HAI-1 additionally contains a motif at N terminus with eight cysteines (MANEC) domain and a single LDLA domain located between the two Kunitz domains (13)(14)(15).
Activation of matriptase and prostasin in epithelial cells is accomplished by the formation of a reciprocal zymogen activation complex in which matriptase is activated by a non-catalytic allosteric interaction with the prostasin zymogen, and prostasin, in turn, is activated by a low intrinsic activity of the matriptase zymogen (16). Additional active matriptase may be generated by activation site cleavage of the matriptase zymogen by the newly activated prostasin, and additional active prostasin may be generated by activation site cleavage of the prostasin zymogen by the newly activated matriptase (16 -18).
HAI-1 was first purified from milk as a soluble protein that formed a stable complex with cell surface-shed activated matriptase, immediately suggesting a role of HAI-1 in regulating the membrane-anchored serine protease (19). Indeed, subsequent studies showed that stable protease-inhibitor protein complexes are formed between HAI-1 and activated matriptase in ex vivo systems with purified components, in epithelial cell cultures, and in organotypic cultures (20 -22). HAI-1 also was found to form stable inhibitor complexes with prostasin, suggesting a dual function in regulating the matriptase-prostasin system (20,22,23). Compatible with these biochemical observations, subsequent genetic epistasis analysis placed HAI-1 downstream from both matriptase and prostasin during development (24 -26). In fact, HAI-1 becomes dispensable for development and long-term survival of mice with low levels of either active matriptase or prostasin (27,28), suggesting that a principal role of the inhibitor is to restrict the activity of the matriptase-prostasin system. More recently, a similar role in regulating the matriptase-prostasin system has been ascribed to HAI-2, based on the ability of soluble recombinant HAI-2 to form high affinity inhibitor complexes with soluble recombinant matriptase, and on the genetic rescue of developmental defects in HAI-2-deficient animals that can be achieved by either loss of expression or by low-level expression of matriptase or prostasin (20,25,28).
In addition to the canonical role of HAI-1 and HAI-2 in restricting the activity of matriptase subsequent to its activation, both inhibitors also have been proposed to have unique functions in regulating the intracellular trafficking and activation of matriptase. Thus, HAI-1 is reported to interact with the matriptase zymogen already within the biosynthetic pathway to prevent its premature activation, to facilitate its transport to the cell surface, and even to induce its activation once located on the plasma membrane (6, 29 -32). Likewise, HAI-2 was recently suggested to be critical for the retention of active matriptase on the plasma membrane (33). A potentially confounding factor in these studies, however, is the frequent reliance on cell-based overexpression systems to dissect the mechanistic interactions of matriptase with HAI-1 and HAI-2. Moreover, discrepancies have been reported as to the necessity of HAI-1 for proper expression of matriptase, even within the same cell-based model system (33)(34)(35).
Cognizant of the considerable knowledge gaps regarding these putative non-traditional roles of HAI-1 and HAI-2 in matriptase function, herein we used a novel approach to analyze the functional relationship of the two inhibitors with the matriptase-prostasin system. Rather than relying on overexpression models, we used gene targeting and gene silencing to determine the effect of ablating endogenous HAI-1 and HAI-2 on endogenous matriptase cell surface localization, activation, and shedding in mouse intestinal epithelia and in intestinal epithelial cell monolayers. We find that loss of HAI-1 does not affect cell surface localization or abundance of matriptase in polarized epithelium of either the small intestine or the colon. In contrast, loss of HAI-2 causes a dramatic decrease in cell surface expression of matriptase in intestinal epithelia, which is mechanistically linked to increased prostasin-mediated activation and shedding.
Protein Extraction from Mouse Tissue-The tissues were dissected, snap frozen on dry ice, and stored at Ϫ80°C until homogenization. The tissues were homogenized in ice-cold lysis buffer containing 1% Triton X-100, 0.5% sodium-deoxycholate in phosphate-buffered saline (PBS) plus Proteinase Inhibitor Mixture (Sigma) and incubated on ice for 10 min. The lysates were centrifuged at 20,000 ϫ g for 20 min at 4°C to remove the tissue debris, and the supernatant was used for further analysis. The protein concentration was measured with standard BCA assay (Pierce).
Cell Culture-HEK293 cells were grown in Dulbecco's modified Eagles medium (DMEM) supplemented with 2 mM l-glutamine, 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Caco-2 cells (ATCC, Manassas, VA) were grown in DMEM supplemented with 2 mM L-glutamine, 20% fetal bovine serum, 1ϫ nonessential amino acids, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen) at 37°C in an atmosphere of 5% CO 2 . For all experiments, 2 ϫ 10 6 cells were seeded into 0.4 m pore-size, 24 mm Transwell filter chamber plates (Corning, Corning, NY) that allow sepa-rate access to the apical and basolateral plasma membrane. The tightness of filter-grown cells was assayed by filling the inner chamber to the brim with cell culture media and allowing it to equilibrate overnight. The cell culture medium was changed every other day.
Transfection of siRNA in Caco-2 Cells and Immunoprecipitation-Subconfluent Caco-2 cells (10 6 cells/transfection) were transfected in suspension with siRNA (SPINT2; HSS116435, HSS116436, HSS173811, PRSS8; HSS108633, HSS108631, controls; 46 -2000, 46 -2001, 46 -2002 (Invitrogen)) using Dharmafect1 (Thermo Scientific, Hudson, NH), following the manufacturer's instructions. For the siRNA, 3.3 l of 20 m RNA stock and 7 l of transfection agent in a final volume of 2 ml of serum-free medium were used. The transfected Caco-2 cells were seeded on 6-well Transwell filters. The medium was changed 12-24 h after transfection, and 72 h posttransfection, the cells were lysed in lysis buffer (1% Triton X-100, 0.5% sodium deoxycholate in PBS) plus Proteinase Inhibitor Mixture (Sigma) and were spun for 20 min at 20,000 ϫ g at 4°C. For immunoprecipitation the inhibitor mixture was excluded. Gammabind G-Sepharose (GE Healthcare, Upsala, Sweden) was prepared as instructed by the manufacturer. The lysates were pre-cleared with 40 l of Sepharose for 30 min at 4°C with end-over-end rotation where after the samples were spun 2 min at 1,500 ϫ g and the supernatant was transferred to a new tube containing 75 l of Sepharose and 20 g of polyclonal goat anti-human HAI-2 antibody (R&D Systems, AF1106). The immunoprecipitation was incubated for 5 h at 4°C with end-over-end rotation. After incubation the samples were spun at 2,000 ϫ g at 4°C to precipitate the affinity gel. The gel was subsequently washed three times with wash buffer I (25 mM Tris/HCl, pH 7.8, 500 mM NaCl, and 0.5% Triton X-100) followed by four washes with wash buffer II (10 mM Tris/HCl, pH 7.8, and 150 mM NaCl). The precipitated proteins were eluted by boiling in 2ϫ sample buffer for 7 min and analyzed by Western blot as described above.
Immunofluorescence Analysis of Mouse Intestine-Intestinal samples from mice were fixed in 4% paraformaldehyde in PBS for 24 h, embedded into paraffin, and sectioned. Tissue sections were cleared with xylene-substitute (Safe Clear, Fisher Scientific), rehydrated in a graded series of alcohols, and boiled in Reduced pH Retrieval Buffer (Bethyl, Montgomery, TX) for 20 min for antigen retrieval. The sections were blocked for 1 h in PBS containing 10% horse serum and incubated at 4°C overnight with sheep anti-human matriptase (AF3946, R&D Systems, Minneapolis, MN) or goat anti-mouse HAI-1 (AF1141, R&D Systems) and rabbit anti-mouse E-cadherin (Zymed Laboratories Inc., Invitrogen, Carlsbad, CA). The slides were washed 3 times in PBS and incubated at room temperature for 45 min with secondary antibodies, Alexa fluor 594-labeled donkey anti-sheep (Invitrogen) and FITC-labeled goat anti-rabbit (Zymed Laboratories Inc.). The tissue sections were washed three times for 5 min with PBS and mounted with VectaShield Hard Set Mounting Medium (Vector Laboratories Inc., Burlingame, CA). The samples were subjected to laser scanning confocal microscopy using the Leica TCS SP2 system and Zeiss LS700 system.
Immunofluorescence Analysis of Caco-2 Cells-Caco-2 cells grown on Transwell filters were fixed for 20 min in 4% parafor-maldehyde in PBS (Z-fix, Anatech, Battle Creek, MI) at room temperature. The following procedures were performed on ice unless otherwise specified. Cells were permeabilized with 0.05% Triton X-100 in PBS for 20 min. Unspecific staining was blocked with PBS containing 5% BSA (PBS/BSA) for 1 h. Cells were incubated with primary antibody diluted in PBS/BSA overnight at 4°C. The antibodies used for immunofluorescence were mouse monoclonal anti-human matriptase, M32, (39), and polyclonal rabbit anti-human occludin (71-1500, Zymed Laboratories Inc.). The next day the cells were washed three times in PBS, followed by incubation with relevant Alexa Fluorconjugated secondary antibodies (Invitrogen) for 1 h. The cells were finally washed 3ϫ in PBS and mounted with Vectashield Hard Set Mounting Medium (Vector Laboratories Inc.) and subjected to laser scanning confocal microscopy using the Leica TCS SP2 system and Zeiss LS700 system.
RT-PCR on Mouse Tissue-Intestines were snap-frozen in liquid nitrogen and ground to a fine powder with a mortar and pestle. Total RNA was prepared by extraction in Trizol reagent (Invitrogen), as recommended by the manufacturer. Reverse transcription and PCR amplification were performed using the RETROscript TM Kit (Ambion Inc., Austin, TX), as recommended by the manufacturer. First strand cDNA synthesis was performed using an Oligo dT primer. The following primer pairs were used for qPCR: St14; 5Ј-AGATCTTTCTGGAT-GCGTATGAGA-3Ј and 5Ј-GGACTTCATTGTACAGCA-GCTTCA-3Ј. The annealing temperature for these primer sets was 58°C. Expression levels were normalized to 18 S in each sample; 5Ј-GAACTGCGAATGGCTCATTAAA-3Ј and 5Ј-CCACAGTTATCCAAGTAGGAGAGGA-3Ј.
PAR-2 Activation Assay-The assay was performed essentially as described (16). HEK293 cells (250,000 cells/well) were plated in 24-well poly-L-lysine-coated plates and grown for 24 h. Cells were co-transfected with pSRE-firefly luciferase (50 ng) (40), pRL-Renilla luciferase (20 ng), pcDNA3.1 containing a full-length human protease-activated receptor (PAR)-2 cDNA (100 ng) (Missouri S&T cDNA Resource Center), pcDNA3.1 containing human matriptase (100 ng), pcDNA3.1 containing human HAI-1 (100 ng), pIRES2-EGFP human prostasin (100 ng) (18), pBig-HAI-2HA (100 ng) and doxycycline inducible rtTA2S-M2 Tet-transactivator, as indicated in the individual experiments (41). Lipofectamine (Invitrogen) was used as the transfection agent according to the manufacturer's instructions. The pBig-HAI-2HA plasmid was generated by inserting the full-length murine Spint2 cDNA (NP_001076017.1) with a synthetic DNA sequence encoding a C-terminal human influenza hemagglutinin (HA)-Tag (YPYDVPDYA) inserted just prior to the stop codon into a unique NotI site of the pBig transgene expression vector containing a Tet responsive element (TRE) (Clonetech, Mountain View, CA). The pSRE-firefly luciferase is a luciferase reporter gene under the control of a multimerized serum-response element. The transfection medium was changed after 18 h, and 48 h after transfection, the cells were serum-starved overnight while induced with Doxycycline (1 g/ml). Cells were lysed, and luciferase activity was determined using the Dual-Luciferase assay kit (Promega) according to the manufacturer's instructions. Luminescence was measured using a Wallac Victor2 1420 multilabel counter (PerkinElmer Life Sciences), and the serum-response element activity was determined as the ratio of firefly luciferase to Renilla luciferase light units.
Transfection of HAI-2 HA in HEK293T Cells and Immunoprecipitation-For the human pcDNA3.1 HAI-2 construct the HA-tag was positioned between amino acid (AA) 30 and 31 to preserve the N-terminal signal peptide (AA 1-27). The HA-tag was inserted in the HAI-2 pcDNA3.1 vector encoding fulllength human HAI-2 cDNA (inserted with EcoRI) with the use of QuickChange site-directed mutagenesis kit according to the protocol provided by the manufacturer (Stratagene, La Jolla, CA). The primers contained the HA-tag sequence (lowercase letters): HAI-2HA forward, 5Ј-tatccttatgacgtgcctgactatgccCGCAG-CATCCACGACTTC-3Ј, HAI-2HA reverse, 5Ј-CTGGCGGC-CGACCGAGAAtatccttatgacgtgcctgactatgcc-3Ј. HEK293T cells were grown to 95% confluency on poly-L-lysine-coated plates before they were co-transfected with human HAI-2 cDNA in the pcDNA3.1 vector (either with or without the HA tag), fulllength human matriptase in the pcDNA3.1 vector and fulllength human prostasin cDNA in the pIRES-EGFP vector. Lipofectamine (Invitrogen) was used according to the manufacturer's instructions, transfecting a total of 4 g DNA in a 6 well format. The cells were lysed 48 h post-transfection, and the lysates were spun at 20,000 ϫ g at 4°C for 20 min. A fraction of the supernatant was used for Western blot and the remaining was used for immunoprecipitation. The EZview Red anti-HA affinity gel (Sigma) was prepared according to the manufacturer's instructions. Each sample was added to 50 l affinity gel and incubated end-over-end for 5 h at 4°C. After incubation the samples were spun at 2,000 ϫ g at 4°C to precipitate the affinity gel. The gel was subsequently washed three times with wash buffer I (25 mM Tris/HCl, pH 7.8, 500 mM NaCl, and 0.5% Triton X-100) followed by four washes with wash buffer II (10 mM Tris/HCl, pH 7.8, and 150 mM NaCl). The precipitated proteins were analyzed by Western blot as described above.

Loss of Endogenous HAI-1 Does Not Affect Expression or Subcellular Localization of Endogenous Matriptase in Intestinal
Epithelium-We used Western blotting and immunofluorescence to directly determine the protein expression levels and the subcellular localization of endogenous matriptase in intestinal epithelia in the absence of endogenous HAI-1 and HAI-2. We chose to study mouse intestinal epithelium because matriptase, prostasin, HAI-1, and HAI-2 are all expressed in this tissue, because the matriptase-prostasin system is critical for intestinal epithelial homeostasis, and because matriptase is expressed at high levels and displays a distinct basolateral localization at steady state (20,(42)(43)(44). Since the validity of our experiments was critically dependent on the ability to specifically detect membrane-localized matriptase in intestinal epithelial cells, we first performed matriptase immunofluorescence analysis on wild type and matriptase-deficient intestines (Fig. 1). Immunofluorescence staining of wild type intestine using the polyclonal sheep anti-matriptase antibody, AF3946, resulted in strong staining of the basolateral membranes of epithelial cells of both the small intestine (Fig. 1, A-C) and the large intestine (Fig. 1, G-I), which co-localized with the cell-cell adhesion protein E-cadherin. This staining was absent in matriptase-deficient epithelial cells from both small (Fig. 1, D-F) and large ( Fig. 1, J-L) intestine, demonstrating that cell surface matriptase can be specifically detected in both tissues using this antibody, and that changes in the abundance and distribution of matriptase in intestinal epithelial cells can be readily determined.
We next used two separate approaches to directly determine the effect of ablation of endogenous HAI-1 from intestinal epithelial cells on the expression and subcellular localization of endogenous matriptase. First, we used young adult mice carrying a floxed and a null Spint1 allele and also carrying a tamoxifen-inducible Cre recombinase gene under the control of a ␤-actin promoter (Spint1 LoxP/Ϫ ;␤-actin-Cre-ERtm ϩ/0 mice). Littermate Spint1 LoxP/ϩ mice were used as controls. These mice were subjected to daily tamoxifen treatments over a period of 5 days to ablate HAI-1 from intestinal tissues. Intestinal tissue was analyzed for matriptase expression and subcellular localization 2 days after the final tamoxifen treatment to prevent the demise of epithelial tissue architecture associated with longterm HAI-1 ablation (36) from confounding the data interpretation. HAI-1 was successfully ablated from the intestine by the tamoxifen treatment, as shown by Western blotting (example in Fig. 2A, top panel, compare lanes 1-3 to lanes 4 and 5) and immunofluorescence (insets in Fig. 2, G and O, compare with insets in C and K, respectively) using HAI-1 antibodies. Inter-estingly, the matriptase protein level in the intestine was unaffected by the loss of HAI-1, as shown by similar levels of matriptase in protein extracts from intestinal tissues from HAI-1-sufficient and HAI-1-deficient mice analyzed by reducing SDS-PAGE and Western blotting using a matriptase antibody that recognizes an epitope located within the serine protease domain (example in Fig. 2A, second panel from top, lanes 1-3,  compare with lanes 4 and 5). Furthermore, the subcellular localization of matriptase was unaffected by the depletion of HAI-1 in epithelial cells of both the small (example in Fig. 2, F-I, compare with Fig. 2, B-E) and in large (Fig. 2, N-Q, compare with J-M) intestine, as analyzed by immunofluorescence and confocal microscopy. This absence of an effect of HAI-1 expression on matriptase levels and subcellular localization could not be explained by a compensatory up-regulation of HAI-2, as the level of expression of HAI-2 was unaffected by the presence or absence of HAI-1 (example in Fig. 2A, second panel  from bottom, compare lanes 1-3 to lanes 4 and 5). The predicted molecular mass of HAI-2 is 28 kDa but splice variants A and B have been reported as well as glycosylation of the protein, rendering two major bands around 32 and 28 kDa (14, 45).
HAI-1 is expressed in a high molar excess to matriptase in most epithelial cells (46). A potential caveat of the above experiment, therefore, was that residual HAI-1 could be expressed at levels that eluded detection by Western blot or by immunofluorescence, but were sufficient to allow for appropriate transport to the epithelial cell surface. To examine this possibility, our second approach was to investigate matriptase expression and subcellular localization in intestinal tissues of mice in which complete HAI-1 deficiency was compensated by a hypomorphic mutation in the prostasin gene (Spint1 Ϫ/Ϫ ;Prss8 fr/fr mice) using HAI-1-sufficient prostasin hypomorphic (Spint1 ϩ/ϩ ; Prss8 fr/fr ) littermates as controls (28,47). Again, intestinal matriptase protein expression levels were unaffected by the absence of HAI-1, as determined by Western blot (example in Fig. 3A, second panel from top, lanes 1-3, compare with lanes  4 -6). Likewise, the subcellular distribution of matriptase again was unaffected by the absence of HAI-1 in epithelial cells of both the small (example in Fig. 3, F-I, compare with B-E) and the large (example in Fig. 3, N-Q, compare with J-M) intestine. As above, the absence of an effect of HAI-1 ablation on matriptase levels and subcellular localization could not be explained  3 to lanes 4 -6). Also, no significant alteration in prostasin level was observed in the absence of HAI-1 (example in Fig. 3A, second panel from bottom, compare lanes 1-3 to 4 -6). Taken together, the above studies demonstrate that the level of expression and the subcellular localization of endogenous matriptase in intestinal cells of the small and large intestine are unaffected by the presence or absence of endogenous HAI-1.

Loss of Endogenous HAI-2 Reduces Expression of Endogenous Matriptase in Intestinal Epithelial
Cells-We next examined the effect of endogenous HAI-2 deficiency on the expression and subcellular localization of endogenous matriptase in intestinal epithelium. For this purpose, we used mice in which the embryonic lethality caused by complete HAI-2 deficiency is bypassed by the hypomorphic prostasin gene mutation (Spint2 Ϫ/Ϫ ;Prss8 fr/fr ) (28, 47) and we used HAI-2-sufficient prostasin hypomorphic (Spint2 ϩ/ϩ ;Prss8 fr/fr ) littermates as controls (example in Fig. 4A, top panel, compare lanes 1 and 2  with 3 and 4). Interestingly, the level of expression of endogenous intestinal epithelial matriptase was markedly decreased by the absence of HAI-2, as determined by reducing SDS-PAGE and Western blotting (example in Fig. 4A, second panel from   top, compare lanes 1 and 2 with lanes 3 and 4). As was the case for HAI-2, no compensatory increase in HAI-1 expression in HAI-2-deficient intestines was observed, although considerable genotype-independent fluctuations in HAI-1 expression levels were observed (example in Fig. 4A, third panel from top, compare lanes 1 and 2 with lanes 3 and 4). No significant increase of prostasin was observed (Fig. 4A, second panel from bottom) when compared with the level of tubulin that was used as a loading control (Fig. 4A, bottom panel). To determine if the decrease in the expression of matriptase protein in HAI-2-deficient intestines was due to regulation at the level of transcription or mRNA processing, we compared intestinal St14 mRNA in Spint2 Ϫ/Ϫ , Spint2 ϩ/Ϫ , and Spint2 ϩ/ϩ littermates (Fig. 4B). This analysis showed that St14 mRNA levels were not different in mice of the three genotypes, suggesting that endogenous HAI-2 regulates endogenous matriptase expression at the translational or post-translational level. Using immunofluorescence, we next analyzed the subcellular localization of matriptase in HAI-2-deficient intestinal epithelium. In agreement with the reduced abundance of matriptase revealed by Western blot, matriptase could not be detected on the basolateral membrane of HAI-2-deficient intestinal epithelial cells, with residual low matriptase expression found only in intracellular structures of both the small (example in Fig. 4, G-J, compare with  Fig. 4, O-R, compare with K-N) intestine, which were also observed in the control mice. Taken together, these findings suggest that endogenous HAI-2 is an important regulator of endogenous matriptase expression in intestinal epithelium.

C-F) and large (example in
Reduced HAI-2 Expression Enhances Matriptase Activation and Shedding in Epithelial Cell Monolayers-To mechanistically dissect the function of HAI-2 in polarized epithelium, we next silenced HAI-2 in intestinal epithelial Caco-2 cells (48) and studied the effect on matriptase expression. These cells were chosen because of their intestinal epithelial origin, because they express matriptase, prostasin, and HAI-2, and because they form matriptase and prostasin-dependent tight junctions with well-defined apical and basolateral membrane domains when cultured to confluency (33,42,49). Furthermore the Caco-2 cells allow easy detection of the various forms of matriptase in both lysates and conditioned media (Fig. 5A). Matriptase is synthesized as a 95 kDa protein that is processed in the SEA domain into the 70 kDa pro-form (zymogen). Upon activation, the 30 kDa serine protease domain of matriptase is cleaved but remains attached to the stem of the protease by disulfide bridges. The activated matriptase rapidly forms a 120 kDa complex with HAI-1 and is shed into the extracellular space in two different forms (Fig. 5A). HAI-2 expression was silenced using siRNA. To minimize the risk of off-target effects confounding data interpretation, we used three non-overlapping SPINT2 siRNAs for silencing of HAI-2, as well as a combination of the three siRNAs for each experiment. All siRNA treatments resulted in a significant decrease in HAI-2 expression when compared with paired control siRNAs (example in Fig. 5B, top panel, lanes 1-4, compare with lanes [5][6][7][8]. Similar to loss of HAI-2 in intestinal epithelium, silencing of HAI-2 in Caco-2 cells resulted in a dramatic decrease in cell-associated matriptase (example in Fig. 5B, second panel from top, lanes 1-4, compare with 5-8). We found that only the 70 kDa pro-form of matriptase in the cell was reduced, while the activated form of matriptase (as represented in the 30-kDa form in the Western blot of reducing SDS/PAGE gels) was unaffected by the HAI-2 silencing. A marked reduction of cell-associated matriptase-HAI-1 complex was also observed by the HAI-2 silencing, as revealed by a 120-kDa species in SDS/PAGE and Western blot of non-boiled and non-reduced samples (example in Fig. 5B,  third panel from top, lanes 1-4, compare with lanes 5-8). The level of prostasin was unaffected by the depletion of HAI-2 in the cell lysates (Fig. 5B, second panel from bottom, compare lanes 1-4  to 5-8). Importantly, the reduction of cell-associated matriptase, caused by HAI-2 silencing, was associated with a reciprocal increase in an activated soluble matriptase species shed into the conditioned medium. This was evidenced by large increases in activated matriptase, as represented in the 30-kDa form in the Western blot of reducing SDS/PAGE gels (example in Fig. 5C, top  panel, lanes 1-4, compare with lanes 5-8) and in the 110 kDa soluble matriptase-HAI-1 complex revealed by SDS/PAGE and Western blot of non-boiled and non-reduced samples (example in Fig. 5C, bottom panel, lanes 1-4, compare with lanes 5-8). Taken together, this data shows that a reduction in HAI-2 levels results in a shift of matriptase from an immature, unprocessed cell-associated zymogen toward a mature, activated and inhibitor-complexed cell surface-shed form.
To further investigate the effect of HAI-2 levels on matriptase processing, we next performed immunofluorescence studies of confluent Caco-2 monolayers. We first validated the capacity of the mouse monoclonal matriptase antibody, M32, to specifically detect matriptase in this cell-based system. As shown in Fig. 6, strong staining of the basolateral plasma membrane was observed in Caco-2 monolayers treated with control siRNA (Fig. 6F), but not in Caco-2 monolayers treated with matriptase-silencing siRNA (Fig. 6B). As previously reported by us and by others, matriptase co-localized with occludin on the basolateral plasma membrane when the Caco-2 cells were transfected with a control siRNA (example in Fig. 6, E-H) (9,42). However, as expected from the above biochemical analysis, matriptase expression was greatly diminished when HAI-2 was depleted using siRNA, with residual expression confined to intracellular vesicles and absence of plasma membrane co-localization with occludin (Fig. 6, I-L). We have previously reported that matriptase activation and HAI-1 complex formation is followed by endocytosis of this complex, which is consistent with the observations above (9).
Enhanced Matriptase Activation and Shedding Caused by Reduced HAI-2 Is Prostasin-dependent-To further investigate the role of HAI-2 in regulating matriptase post-translational processing, we used a previously established assay in which the activation of the matriptase target substrate, proteinase-acti- vated receptor(PAR)-2 (4, 17, 51, 52), is measured by its ability to induce transcription of a serum response element(SRE)luciferase reporter plasmid (16,28). HEK293 cells, which express low levels of endogenous matriptase and prostasin, were transfected with a PAR-2 expression plasmid and with a doxycycline-inducible HAI-2 expression plasmid (Fig. 7A). As described recently (16), when these cells additionally were transfected with matriptase and prostasin expression plasmids, robust PAR-2 activation was observed, as shown by increased luciferase activity (Fig. 7A, lane 1). As expected, PAR-2 activation by the matriptase-prostasin complex was inhibited by induction of HAI-2 expression by doxycycline (Fig. 7A, lane 2). Transfection of HEK293 cells with matriptase or prostasin expression plasmids, individually in each case, resulted in a small, but significant, activation of PAR-2 (Fig. 7A, lanes 3 and  5, respectively). Interestingly, however, whereas induction of HAI-2 expression reduced prostasin-mediated PAR-2 activation to baseline levels (Fig. 7A, lane 6), the induction of HAI-2 did not significantly reduce PAR-2 activation by matriptase (Fig. 7A, lane 4). Compatible with a role of HAI-2 in regulating prostasin, an HA-tagged version of the inhibitor co-immunoprecipitated prostasin (Fig. 7B, middle panels) as well as matriptase, however mainly in its 95 kDa pro-form (Fig. 7B, bottom panels), in transfected HEK293 cells. Furthermore, endogenous prostasin was efficiently immunoprecipitated with HAI-2 antibodies in control siRNA-treated Caco-2 cells, but not from HAI-2-silenced Caco-2 cells (Fig. 7C, compare lanes 1  and 4 with 3 and 6). Taken together, these data tentatively suggest that HAI-2 may regulate the matriptase-prostasin complex by interacting with prostasin. A central prediction from this hypothesis would be that the "runaway" activation and consumptive depletion of matriptase from intestinal epithelial cells that was caused by reduced HAI-2 expression would be attenuated by the simultaneous reduction in the expression of prostasin. To challenge this hypothesis, we again used the Caco-2 cells and siRNA to determine if the simultaneous silencing of endogenous HAI-2 and prostasin would rescue the increased activation and shedding of matriptase observed by the single silencing of endogenous HAI-2. Two independent siRNAs were used to silence the expression of prostasin, both resulting in a significant reduction in the expression of the protease (example in Fig. 7D, second  panel from top, compare lanes 5-8 with lanes 1-4). As above, silencing of HAI-2 alone markedly reduced the levels of the 70-kDa proform of matriptase (Fig. 7D, third panel from top,  compare lanes 3 and 4 to lanes 1 and 2) and increased the shedding of the activated form to the conditioned medium, as detected by an increase in the 30-kDa matriptase serine protease domain (Fig. 7E, top panel, lanes 3 and 4, compare with  lanes 1 and 2) and matriptase-HAI-1 complex formation (Fig.  7E, bottom panel, compare lanes 3 and 4 to lanes 1 and 2). When prostasin levels were decreased, no variation in the total level of matriptase was observed both in the cell lysates (Fig. 7D, third  panel from top, compare lanes 5-8 with lanes 1 and 2) and in the conditioned medium (Fig. 7E, top panel, compare lanes 5-8   with 1 and 2), however, a slight reduction of matriptase-HAI-1 complex formation was observed both in the cell lysates (  1 and 2). Interestingly, however, cells with simultaneous silencing of prostasin and HAI-2 displayed a markedly higher level of the 70-kDa matriptase proform than cells with silencing of HAI-2 alone (Fig. 7D, third panel from top and 7E, top panel, compare lanes 9 -12 with lanes 3 and 4). The decrease in matriptase-HAI-1 complex observed upon prostasin knock-down (Fig. 7D, second panel  from bottom, lanes 5-8 compare with lanes 1 and 2) was not rescued by the simultaneous knock-down of HAI-2 (Fig. 7D,  second panel from bottom, lanes 9 -12). To further characterize the rescue of matriptase expression levels by simultaneous depletion of both HAI-2 and prostasin, we next performed immunofluorescence studies of confluent Caco-2 monolayers. As described above, matriptase expression was greatly diminished when HAI-2 was depleted using siRNA (Fig. 7J). However, upon simultaneous depletion of prostasin (Fig. 7, O-Q), matriptase was once again observed on the basolateral plasma membrane, indistinguishable from the control cells (Fig. 7, F-H). Taken together, these data show that HAI-2 is an important regulator of matriptase that determines the rate of matriptase activation, matriptase-HAI-1 inhibitor complex formation, and matriptase shedding, by regulating the activity of prostasin.   1-4), prostasin (lanes 1-4) and either HAI-2 (lanes 1 and 3) or HAI-2 containing an HA-tag (lanes 2 and 4). The cells were lysed 48 h after transfection and a fraction of the lysate (lanes 1 and 2) 1 and 4), siRNA targeting prostasin (siPRSS8, lanes 2 and 5) or HAI-2 (siSPINT2, lanes 3 and 6). Three days posttransfection the cells were lysed and a fraction of the lysate (lysate, lanes 1-3) was analyzed by SDS-PAGE and Western blot using antibodies against HAI-2 (top panel) and prostasin (bottom panel). The remaining lysate was used for immunoprecipitation (IP, lanes 4 -6) with HAI-2-antibodies and the precipitated proteins were subjected to SDS-PAGE and Western blot as for the lysates. Prostasin was efficiently co-precipitated with HAI-2. The specific proteins are indicated on the right, and the molecular markers are indicated on the left (kDa). The data are representative of two similar experiments. D and E, siRNAs for silencing down the gene of HAI-2 (siSPINT2, lanes 3 and 4), two prostasin-silencing siRNAs (siPrss8 a and b, lanes 5-8) and a combination hereof (lanes 9 -12) where transfected into Caco-2 cells. Scrambled siRNA was used as control (siCTRL, lanes 1-2). All samples were loaded in duplicates.

DISCUSSION
The current study directly examined the role of endogenous HAI-1 and HAI-2 in regulating endogenous matriptase trafficking and activation in polarized epithelium of the large and small intestine of mice. We found no evidence that loss of HAI-1 affected the abundance, steady state localization or state of activation of matriptase in either epithelium, suggesting that the primary function of HAI-1 in intestinal tissues, as regards matriptase, is to form inhibitor complexes with the activated protease. It should be noted that an independent study recently reported that ablation of HAI-1 from the mouse colon resulted in a diffuse cytoplasmic mislocalization of matriptase (36). Importantly, however, whereas we studied matriptase localization following the acute, tamoxifen-induced ablation of HAI-1 from the intestine, the matriptase localization studies in the aforementioned study were performed only after the overall demise of the intestinal epithelial architecture, caused by longterm HAI-1 ablation, was manifest (36), which may indirectly have affected the subcellular localization of matriptase.
The absence of an effect of HAI-1 on matriptase localization in intestinal epithelium in vivo is incongruent with cell-based overexpression studies in which both the efficient trafficking to the cell surface and the activation of overexpressed matriptase was shown to require the simultaneous overexpression of HAI-1 (see Introduction for references). While it is not within the scope of the current study to explain why these cell-based overexpression systems differ from polarized epithelium in vivo in terms of their requirement for regulation of matriptase by HAI-1, we speculate that supraphysiological levels of HAI-1 may be required to prevent pathogenic matriptase autoactivation within the secretory pathway caused by supraphysiological levels of expression of the protease.
In stark contrast to HAI-1, ablation of HAI-2 from intestinal epithelium of small and large intestine resulted in a near complete loss of matriptase from both epithelia. By using gene silencing in intestinal epithelial monolayers, we found that loss of HAI-2 markedly increased the activation and shedding of matriptase from the plasma membrane. The conspicuous absence of matriptase from HAI-2-ablated intestinal epithelium, therefore, likely can be attributed to the rate of activation and associated shedding of matriptase exceeding its rate of synthesis in the absence of HAI-2. This scenario is analogous to both consumptive coagulopathy and C1-inhibitor deficiency, where the uncontrolled activation of, respectively, coagulationand complement system-associated serine proteases leads to a systemic depletion of their corresponding zymogens (54,55). An important question emanating from our study is where in the secretory pathway HAI-2 acts to regulate prostasin-mediated matriptase activation? Currently available antibodies do not allow for specific detection of endogenous HAI-2 (data not shown). However, previous studies have shown that an epitopetagged HAI-2 variant resides predominantly in an intracellular compartment when overexpressed in epithelial cells (33).
Our proposal that HAI-2 is essential for matriptase expression in intestinal epithelium by preventing the "consumptive depletion" of matriptase may also explain the etiology of the sodium-selective diarrhea that is associated with congenital HAI-2 deficiency in humans (56). The epithelial sodium channel ENaC is critical for sodium reabsorption in the distal colon, and the activity of ENaC in the distal colon is dependent on its proteolytic activation by the matriptase-prostasin system (53). It was, therefore, not intuitively obvious why HAI-2 deficiency is associated with decreased, rather than increased, sodium reabsorption. However, our model of HAI-2 deficiency-induced "consumptive depletion" of matriptase from colonic epithelium is fully compatible with this clinical manifestation of HAI-2-deficient individuals and is supported by a recent study that shows that HAI-2 can regulate prostasin-mediated ENaC activation (50).
Matriptase, prostasin, HAI-1, and HAI-2 are co-expressed in most developing and adult mammalian epithelia. It is important to underscore that the current study only provides evidence for how HAI-1 and HAI-2 regulate the matriptase-prostasin system in intestinal epithelium. Additional studies will be required to determine if similar regulation is observed in other simple, polarized epithelia, in transitional epithelium, and in multi-layered squamous epithelium.