Matrix metalloproteinase-7 induces homotypic tumor cell aggregation via proteolytic cleavage of the membrane-bound Kunitz-type inhibitor HAI-1

Matrix metalloproteinase-7 (MMP-7) plays important roles in tumor progression and metastasis. Our previous studies have demonstrated that MMP-7 binds to colon cancer cells via cell surface–bound cholesterol sulfate and induces significant cell aggregation by cleaving cell-surface protein(s). These aggregated cells exhibit a dramatically enhanced metastatic potential. However, the molecular mechanism inducing this cell–cell adhesion through the proteolytic action of MMP-7 remained to be clarified. Here, we explored MMP-7 substrates on the cell surface; the proteins on the cell surface were first biotinylated, and a labeled protein fragment specifically released from the cells after MMP-7 treatment was analyzed using LC-MS/MS. We found that hepatocyte growth factor activator inhibitor type 1 (HAI-1), a membrane-bound Kunitz-type serine protease inhibitor, is an MMP-7 substrate. We also found that the cell-bound MMP-7 cleaves HAI-1 mainly between Gly451 and Leu452 and thereby releases the extracellular region as soluble HAI-1 (sHAI-1). We further demonstrated that this sHAI-1 can induce cancer cell aggregation and determined that the HAI-1 region corresponding to amino acids 141–249, which does not include the serine protease inhibitor domain, has the cell aggregation–inducing activity. Interestingly, a cell-surface cholesterol sulfate-independent proteolytic action of MMP-7 is critical for the sHAI-1–mediated induction of cell aggregation, whereas cholesterol sulfate is needed for the MMP-7–catalyzed generation of sHAI-1. Considering that MMP-7–induced cancer cell aggregation is an important mechanism in cancer metastasis, we propose that sHAI-1 is an essential component of MMP-7–induced stimulation of cancer metastasis and may therefore represent a suitable target for antimetastatic therapeutic strategies.

Matrix metalloproteinases (MMPs) 2 make up a family of zinc-dependent endopeptidases capable of degrading protein components of extracellular matrix and play pivotal roles in tissue remodeling under physiological and pathological conditions, such as morphogenesis, angiogenesis, tissue repair, and tumor invasion (1)(2)(3)(4). MMP-7 is one of a few MMPs that are overexpressed by carcinoma cells rather than stromal cells (5,6). Among more than 20 MMPs, MMP-7 appears to be one of the most important MMPs in cancer metastasis, because expression of this MMP is correlated well with tumor malignancy and metastasis, especially with liver metastasis of colon cancers (7,8).
Our previous study demonstrated that MMP-7 binds to cellsurface cholesterol sulfate (CS) and acts as a membrane-associated protease, and the treatment of human colon carcinoma cells with active MMP-7 in vitro induces cell aggregation by cleaving cell-surface proteins (9). It has also been reported that the seven amino acid residues of MMP-7 are essential for the interaction with CS; a variant of MMP-7, named MMP-7 (29,33, 51, 55/M2)⌬C3, which has the critical internal four residues of MMP-7 replaced with the corresponding residues of MMP-2, and the C-terminal three residues deleted, lacks both affinity for CS and the cell aggregation-inducing activity (10).
Formation of cancer cell aggregation likely contributes to the survival of cancer cells in the circulation and is expected to play a key role in lodging the cells into the capillary vessel, thereby promoting hematogenous metastasis of cancers (11). It has also been suggested that cell-cell adhesion contributes to the maintenance of cancer stem cells (12), which have recently been hypothesized to represent the driving force behind tumor metastasis. The MMP-7-induced cancer cell aggregation actually enhances their metastatic potential in the nude mouse model (13).
The MMP-7-induced cell aggregation of colon carcinoma cells consists of two steps; initial loose cell aggregation is secondly converted to tight cell aggregation (13). Although it has been found that the latter step is mediated by E-cadherin, the mechanism of the initial cell aggregation has remained to be clarified.
In this study, we identified hepatocyte growth factor activator inhibitor type 1 (HAI-1) as a novel substrate of membranebound MMP-7, and we demonstrated that the fragment of HAI-1 released from colon cancer cells upon MMP-7 cleavage has the ability to induce cell aggregation. We also determined a region of HAI-1 essential for induction of the homotypic cell adhesion.

Identification of HAI-1 as a cell-surface protein cleaved by membrane-bound MMP-7
To explore the cell-surface proteins, which are released from WiDr human colon carcinoma cells by the CS-dependent proteolytic action of MMP-7, surface proteins of WiDr cells were first biotinylated. The surface protein-labeled cells were then treated with MMP-7 or MMP-7 (29,33,51,55/M2)⌬C3, the variant of MMP-7 lacking affinity for CS (10). When the proteins in the conditioned medium (CM) of the treated cells were analyzed by ligand blotting, using avidin-conjugated alkaline phosphatase as a probe, we found that several biotinylated protein fragments were released from the cells treated with MMP-7 or MMP-7 (29,33,51,55/M2)⌬C3, and a 44-kDa fragment was released only from the MMP-7-treated cells (Fig. 1A). The biotinylated proteins released from the MMP-7-treated cells were collected using an avidin-Sepharose column, which were then subjected to SDS-PAGE followed by Coomassie Brilliant Blue R-250 (CBB) staining. The protein band of the 44-kDa fragment was excised from the gel. LC-MS/MS analysis of the tryptic peptides of the 44-kDa fragment revealed that the fragment was derived from HAI-1, a type I membrane protein, suggesting that the extracellular region of HAI-1 is proteolytically released as "soluble HAI-1" (sHAI-1). When the CMs of the WiDr cells were analyzed by immunoblotting under non-reduced conditions, the HAI-1-derived 44-kDa fragment in the culture medium was indeed increased upon treatment of the cells with MMP-7 (Fig. 1B). The immunoreactive protein in the CM migrated as a 51-kDa protein under reduced conditions. The intramolecular disulfide bonds of sHAI-1 probably cause the difference of mobilities between reduced and non-reduced conditions. We found that HAI-1 in the lysate of WiDr cells also showed different mobilities in the immunoblot analysis under non-reduced (51 kDa) and reduced (60 kDa) conditions.

Binding of MMP-7 to CS is important for cleavage of HAI-1 localized in the raft region and determination of the peptide bond of HAI-1 cleaved by MMP-7
To examine whether cell-surface HAI-1 is shed by MMP-7 in a CS-dependent manner, MMP-7 (29,33,51,55/M2)⌬C3 and wild-type MMP-7 were compared for their abilities to release the soluble fragment of HAI-1. As shown in Fig. 2A, treatment of WiDr cells with the variant of MMP-7 led to a slight release of the 44-kDa HAI-1-derived fragment, but the level was almost the same as that released from the non-treated cells, suggesting that the variant of MMP-7 hardly sheds HAI-1. In contrast, wild-type MMP-7 effectively released the HAI-1 fragment. Furthermore, when WiDr cells were treated with methyl-␤-cyclodextrin (M␤-CD), release of the HAI-1 fragment by MMP-7catalyzed cleavage was decreased significantly (Fig. 2B). Therefore, it is likely that binding of MMP-7 to CS is crucial for the shedding of HAI-1. We next examined localization of HAI-1 on the cell membrane. We prepared the membrane fraction from Colo201 human colon carcinoma cells by the differential centrifugation method as described under "Experimental procedures," and the membrane fraction was solubilized with The CM was prepared from the incubated cells, and then the labeled proteins released in the CM were analyzed by SDS-PAGE under non-reduced conditions followed by ligand blotting (LB) with the avidin-conjugated alkaline phosphatase (AP-avidin) as a probe. The arrowhead represents the 44-kDa fragment released only from the MMP-7-treated cells. Ordinate, molecular mass in kDa. B, WiDr cells were incubated in serum-free medium without (Ϫ) or with (ϩ) 50 nM MMP-7 at 37°C for 2 h. Fragments of HAI-1 released into the culture medium were analyzed by immunoblotting (IB) under reduced (ϩ2ME) or non-reduced (Ϫ2ME) conditions with the anti-HAI-1 pAb. HAI-1 protein in the whole lysate of WiDr cells was also analyzed by the immunoblotting. The 44-kDa arrowhead and 51-kDa arrowhead represent the immunoreactive bands of non-reduced and reduced forms of sHAI-1, respectively. The 51-kDa arrow and 60-kDa arrow represent the immunoreactive bands of non-reduced and reduced forms of HAI-1, respectively. Ordinate, molecular mass in kDa; 2ME, 2-mercaptoethanol.

Shed HAI-1 fragment has cell aggregation-inducing activity
the non-ionic detergent Triton X-100 at 4°C. As shown in Fig.  2C, HAI-1 was mainly partitioned into the detergent-insoluble fraction when the membrane fraction prepared from the nontreated cells was analyzed. In contrast, HAI-1 was efficiently solubilized when the membrane fraction was prepared from M␤-CD-treated cells. Consistent with our previous study (9), when the membrane fraction prepared from Colo201 cells incubated with MMP-7 was analyzed, MMP-7 was also detected in the detergent-insoluble fraction, whereas this MMP did not bind to the M␤-CD-treated cells; hence, MMP-7 was detected neither in the detergent-insoluble fraction nor in the soluble fraction. These data suggest that both HAI-1 and MMP-7 are colocalized in distinctive microdomains such as rafts.
We next constructed a mammalian expression vector of N-terminally FLAG-tagged HAI-1 (named nFL-HAI-1) and stably-transfected it into DLD-1 human colon carcinoma cells. When the nFL-HAI-transfected cells were treated with MMP-7 and the resultant CM was analyzed by immunoblotting under reduced conditions (Fig. 2D), a 52-kDa fragment was at 37°C for 30 min, and then the cells were further incubated without (Ϫ) or with (ϩ) 50 nM MMP-7 at 37°C for 2 h. Fragments of HAI-1 protein released into the culture medium were analyzed by immunoblotting (IB) under non-reduced conditions with the anti-HAI-1 pAb. C, Colo201 cells were preincubated without (Ϫ) or with (ϩ) 10 mM M␤-CD at 37°C for 30 min, and then the cells were further incubated without (top two panels) or with (bottom two panels) 50 nM MMP-7 at 37°C for 3 h. The membrane fraction prepared from the incubated cells was dissolved in Triton X-100 at 4°C. HAI-1 and MMP-7 in the detergent-soluble (Sol.) or -insoluble (Insol.) fractions were detected by immunoblotting under non-reduced conditions. D, construction of nFL-HAI-1 is schematically represented. The numbers in the scheme indicate the position of amino acid residues. The number in parentheses represents the deduced molecular mass in Da of the polypeptide moiety of nFL-HAI-1. CHO represents the potential site of Asn-linked glycosylation (top). The nFL-HAI-transfected DLD-1 cells or the mock-transfected cells were treated with 50 nM MMP-7 at 37°C for 24 h. The resultant CM corresponding to 5 ϫ 10 5 mock-transfected cells or that corresponding to 1 ϫ 10 5 nFL-HAI-1-transfected cells was analyzed by immunoblotting (IB) under reduced conditions with the anti-FLAG M2 mAb or anti-HAI-1 pAb (bottom left). 52-kDa arrow and 51-kDa arrow represent the FLAG-tagged sHAI-1 and non-tagged sHAI-1, respectively. The nFL-HAI-1-transfected DLD-1 cells were treated without (ϪMMP-7) or with 50 nM MMP-7 (ϩMMP-7) at 37°C for the indicated length of time. The N-terminally tagged fragments of HAI-1 released into the medium were analyzed by immunoblotting under reduced conditions with the anti-FLAG M2 mAb (bottom right). 52-, 45-, and 38-kDa arrows represent the released FLAG-tagged fragments. E, nFL-HAI-1 transfected DLD-1 cells were treated with 50 nM MMP-7 at 37°C for 24 h, and CM was harvested from the cells. The N-terminally tagged fragments of HAI-1 released into the medium were collected with an anti-FLAG M2 mAb-conjugated agarose column, which were then subjected to SDS-PAGE under reduced conditions followed by CBB staining. Ordinate, molecular mass in kDa. Mass spectrometric analysis revealed that arginyl endopeptidase digestion of the 52-kDa protein yielded a peptide assigned to have the GISKKDVFG sequence, and Asp-N protease digestion of the 45-kDa protein yielded peptides assigned to have the DEAACEKYTSG and DEAACEKYTSGFDE sequences, which are deduced to be derived from the C termini of respective HAI-1 fragments. The putative MMP-7 cleavage sites in HAI-1 are also shown by arrowhead in the scheme in D. F, DLD-1 cells were transfected transiently with empty vector (Mo) or expression vector of the nFL-HAI-1 (WT), the single amino acid residue-substituted variant HAI-1 L452/G (variant 1, V1) or the triple amino acid residues-substituted variant nFL-HAI-1 F376/G, L379/G, L452/G (variant 2, V2). Forty eight hours after transfection, the cells were incubated without (ϪMMP-7) or with 50 nM MMP-7 (ϩMMP-7) at 37°C for 3 h. The CM and cell lysate prepared from the incubated cells were examined for their contents of FLAG-tagged proteins by the immunoblotting with the anti-FLAG M2 mAb. ␤-Actin in the cell lysate was also detected by immunoblotting and used as an internal loading control.

Shed HAI-1 fragment has cell aggregation-inducing activity
mainly detected by anti-HAI-1 pAb and anti-FLAG M2 mAb. When the CM of the MMP-7-treated mock-transfected DLD-1 cells was analyzed, a 51-kDa fragment was mainly detected by anti-HAI-1 pAb but not by anti-FLAG M2 mAb. These data suggest that non-tagged HAI-1 expressed endogenously in the mock-transfected cells and the nFL-HAI-1 in the expression vector-transfected cells are cleaved at the same specific site by MMP-7, and the fragments of which difference in molecular mass corresponding to that of FLAG tag moiety (ϳ1 kDa) are released. Therefore, it is likely that the 52-kDa soluble fragment is the FLAG-tagged form of the 51-kDa sHAI-1. The 51-kDa reduced form of non-tagged sHAI-1 corresponds to the 44-kDa non-reduced form of sHAI-1 in Fig. 1.
We also examined the time course of release of HAI-1 fragments from the nFL-HAI-1-transfected cells after MMP-7 treatment (Fig. 2D), and we found that in addition to the major 52-kDa fragment, minor 45-and 38-kDa fragments were released from the transfected cells after a 48-h incubation. Because the 38-kDa fragment was also released from the DLD-1 cells without MMP-7 treatment, it is likely that the 52-and 45-kDa HAI-1 fragments are generated by MMP-7-catalyzed cleavage. To determine the sites of HAI-1 cleaved by MMP-7, the FLAG-tagged HAI-1 fragments released into the medium were collected, using an anti-FLAG M2 mAb-conjugated agarose column, which were then subjected to SDS-PAGE under reduced conditions. The band of 52-or 45-kDa proteins ( Fig.  2E) was excised from the gel and subjected to arginyl endopeptidase or Asp-N digestion followed by LC-MS/MS analysis, respectively. We found that the arginyl endopeptidase digests of the 52-kDa protein contained a fragment corresponding to amino acid residues 443-451 of HAI-1, which does not include the arginine residue. The Asp-N digests of the 45-kDa protein contained fragments corresponding to amino acid residues 365-375 and that corresponding to residues 365-378 of HAI-1, both of which had C termini followed by non-aspartic acid residues in the HAI-1 sequence. These data suggest that MMP-7 cleaves HAI-1 mainly at the peptide bonds between Gly 451 and Leu 452 , and the peptide bonds corresponding to Gly 375 -Phe 376 and Glu 378 -Leu 379 in HAI-1 are slightly susceptible to MMP-7 cleavage. To verify this, we constructed expression vectors of a variant of nFL-HAI-1 that had Leu 452 of HAI-1 replaced with glycine (named HAI-1 L452/G), and another variant that had three residues Phe 376 , Leu 379 , and Leu 452 of HAI-1 replaced with glycine (named HAI-1 F376/G, L379/G, L452/G). The vectors of these variants or a wild-type FLAG-tagged HAI-1 were transiently transfected into DLD-1 cells. These transfectants were then treated with MMP-7, and the release of FLAG-tagged fragments was examined by immunoblotting. As shown in Fig.  2F, 52-and 45-kDa FLAG-tagged fragments were released from the wild-type HAI-1-transfected cells, whereas the 52-kDa fragment was not released from the HAI-1 L452/G-transfected cells. Neither the 52-kDa nor the 45-kDa FLAG-tagged fragment was released from the HAI-1 F376/G, L379/G, L452/Gtransfected cells. Therefore, it is likely that the Gly 375 -Phe 376 , Glu 378 -Leu 379 , and Gly 451 -Leu 452 bonds of HAI-1 are the sites of cleavage by MMP-7.

Induction of colon cancer cell aggregation by sHAI-1
It is known that several cell adhesion proteins, such as E-cadherin and integrins, work in a metal ion-dependent manner. We next examined whether metal ions are involved in cell aggregation induced by MMP-7. Consistent with our previous study (13), when the aggregated cells were freed from MMP-7 by the treatment of the cells with synthetic MMP inhibitor TAPI-1, and then dispersed into single cells by pipetting, these cells were re-aggregated during further incubation in the presence of TAPI-1. However, when the aggregated cells were treated both with TAPI-1 and EDTA, these cells were not re-aggregated during further incubation (Fig. 3A), suggesting that metal ions are required for the MMP-7-induced cell aggregation.
To examine whether the cleaved-HAI-1 fragments bind to the cell surface in a metal ion-dependent manner, Colo201 cells were incubated with MMP-7 and then washed with serum-free medium supplemented without or with 5 mM EDTA. As shown in Fig. 3B, the 44-kDa sHAI-1 (non-reduced form) generated by MMP-7 treatment was detected in the membrane fraction prepared from the cells without the EDTA treatment, whereas the cell-bound HAI-1 fragment was diminished by washing the cells with the EDTA-containing medium. Therefore, metal ions are also essential for the binding of the 44-kDa sHAI-1 to the cell surface. We also determined the ratio of amounts of HAI-1 and sHAI-1 in the membrane fraction and sHAI-1 in CM by comparing their band intensities of immunoblotting, and we found that the ratio of cell-associated sHAI-1/cell-associated HAI-1/sHAI-1 in CM was ϳ1:210:30, suggesting that small Colo201 cells were treated with 50 nM MMP-7 at 37°C for 3 h, and the membrane-bound MMP-7 was removed by treating the cells with 2 M TAPI-1. These cells were washed two times with serum-free medium supplemented without or with 5 mM EDTA and were then washed with serum-free medium. A, cells were further incubated in the serum-free medium containing 5 M TAPI-1 at 37°C for 3 h and photographed. Scale bar, 100 m. B, washed cells were homogenized and fractionated by centrifugation as described under "Experimental procedures." HAI-1-derived fragments in the membrane fraction were detected by immunoblotting under non-reduced conditions. The intact arrowhead and the soluble arrowhead represent the immunoreactive bands of HAI-1 and sHAI-1, respectively. Ordinate, molecular mass in kDa.

Shed HAI-1 fragment has cell aggregation-inducing activity
fraction of sHAI-1 is associated with the surface of the aggregated cells.
To examine whether sHAI-1 is involved in the MMP-7induced cell aggregation, Colo201 cells were treated with MMP-7 to allow the cells to form aggregation, and then the cell-associated MMP-7 and the cleaved-HAI-1 fragment were removed by sequential treatments with TAPI-1 and EDTA. These cells were dispersed by pipetting and then further incubated with or without recombinant sHAI-1. As shown in Fig.  4A, the cells were re-aggregated only in the presence of sHAI-1, suggesting that sHAI-1 has the ability to induce cell aggregation. The cell aggregation was enhanced in an sHAI-1 concentration-dependent manner (Fig. 4A). When the binding of exogenous sHAI-1 to the cell surface was examined by the fluorescence staining, using biotinylated sHAI-1 as a probe, the labeled protein was localized on the cell surface, including regions of intercellular contact, suggesting that sHAI-1 behaves as a cell-adhesion molecule. When the binding of biotinylated sHAI-1 to MMP-7-treated or non-treated cells was tested by the fluorescence staining, the extent of sHAI-1 bound to nontreated cells was lower than that bound to MMP-7-treated cells. The cell ELISA analysis also demonstrated that the MMP-7 treatment facilitated the binding of sHAI-1 to the cells (Fig. 4B). The binding of sHAI-1 to MMP-7-treated cells was metal ion-dependent, and the bound sHAI-1 was released upon the treatment of the cells with EDTA, as expected ( Fig. 4C).

HAI-1 expression is necessary for MMP-7-induced cell aggregation
To verify that HAI-1 expression is necessary for WiDr cells to be aggregated upon MMP-7 treatment, we prepared WiDr cells stably transfected with a short hairpin RNA (shRNA) targeting the hai-1 gene or non-targeting shRNA. The sHAI-1 was hardly released from WiDr cells of which the expression of HAI-1 was prevented by the shRNA (Fig. 5A). When the HAI-1 expression-prevented WiDr cells were treated with MMP-7, they were hardly aggregated (Fig. 5B). However, the MMP-7 induction of cell aggregation was restored (Fig. 5C) when the HAI-1 expression was rescued by transient transfection of the HAI-1 expression vector (Fig. 5D). These data strongly suggest that HAI-1 expression is necessary for the MMP-7-induced cell aggregation.

MMP-7 induces aggregation of HT1080 fibrosarcoma cells transfected with HAI-1
We found that human fibrosarcoma-derived HT1080 cells did not express HAI-1 (Fig. 6A), and they were hardly aggregated upon MMP-7 treatment under suspended cell culture conditions (Fig. 6B). When the MMP-7-induced aggregation

. Exogenously added sHAI-1 binds to cell surface in an MMP-7 treatment-and a metal ion-dependent manner and induces cell aggregation.
A, aggregation assay of Colo201 cells was performed in the absence (control) or presence of 50 nM sHAI-1, as described under "Experimental procedures," and the cells after a 5-h incubation were photographed. Scale bar, 100 m (left, top). In the aggregation assay, Colo201 cells treated with MMP-7 followed by TAPI-1 and EDTA were then washed and incubated without (control) or with 50 nM biotin-labeled sHAI-1 at room temperature for 1 h, and the labeled protein bound to Colo201 cells was visualized by staining with NeutrAvidin-FITC (left, bottom). Scale bar, 20 m. The aggregation assay was performed in the presence of indicated concentrations of sHAI-1, and percent aggregation was determined as described under "Experimental procedures" (right). B, 50 nM biotin-labeled sHAI-1 was incubated with MMP-7-treated or non-treated (Non-treated) Colo201 cells at room temperature for 1 h. The labeled protein bound to cells was visualized by fluorescent staining (left). Scale bar, 20 m. The indicated concentrations of biotin-labeled sHAI-1 was incubated with MMP-7-treated (OE) or non-treated (‚) Colo201 cells at 37°C for 1 h. The amount of the labeled sHAI-1 bound to the cell surface was measured by cell ELISA (right). C, indicated concentrations of biotin-labeled sHAI-1 was incubated with MMP-7-treated Colo201 cells in the presence (f) or absence (Ⅺ) of 5 mM EDTA at 37°C for 1 h. The amount of the labeled protein bound to the cells was measured by cell ELISA (top). Error bars in A-C represent mean Ϯ S.D.; n ϭ 3. The aggregated cells after a 5-h incubation with 50 nM sHAI-1 in A were harvested, washed, and then treated without (Ϫ) or with (ϩ) 5 mM EDTA. The cells were removed by centrifugation, and the content of sHAI-1 in the supernatant was analyzed by immunoblotting under reduced conditions (bottom).

Shed HAI-1 fragment has cell aggregation-inducing activity
of HT1080 cells stably transfected with HAI-1 was tested, they were significantly aggregated (Fig. 6B). To examine whether the MMP-7-catalyzed cleavage of HAI-1 is necessary for the cell aggregation, expression vectors of the MMP-7 cleavage-resistant HAI-1 variants HAI-1 L452/G and HAI-1 F376/G, L379/G, L452/G were transiently transfected HT1080 cells, and expression of HAI-1 and the two variants on the cell surface was examined by fluorescence-activated cell-sorting analysis. These transfectants were then treated with MMP-7, and the release of HAI-1 fragments was examined by immunoblotting. As shown in Fig. 6C, both the variants and wild-type HAI-1 were expressed on surface of HT1080 cells, and HAI-1 F376/G, L379/G, L452/G-transfected cells did not release any soluble fragment of HAI-1 upon MMP-7 treatment. When the MMP-7-induced aggregation of HT1080 cells expressing wild-type HAI-1 or HAI-1 F376/G, L379/G, L452/G were tested, the cells expressing the cleavage-resistant HAI-1 variant were hardly aggregated (Fig. 6D), suggesting that cleavage of HAI-1 is critical for the MMP-7-induced cell aggregation.

CS-independent proteolytic action of MMP-7 on the cell surface is necessary for the sHAI-1-mediated induction of cell aggregation
In the studies in Fig. 4, we also tested whether sHAI-1 induces aggregation of Colo201 cells previously treated without MMP-7, and we found that the non-treated cells were not aggregated even in the presence of sHAI-1 (data not shown). Therefore, it is likely that proteolytic action of MMP-7 on the cell surface, other than HAI-1 cleavage, is required for the sHAI-1-mediated induction of cell aggregation.
To examine whether the CS-dependent proteolytic action of MMP-7 is also needed for the sHAI-1-mediated cell aggregation, we first prepared the CM of WiDr cells, which were pretreated without or with M␤-CD and then incubated with MMP-7 under the suspended cell culture condition. We found that the WiDr cells pretreated without M␤-CD were aggregated but those pretreated with M␤-CD were not, and the M␤-CD treatment of the cells caused significant decrease of the The transfected cells in suspended conditions were incubated without (ϪMMP-7) or with 50 nM MMP-7 (ϩMMP-7), in poly-HEMA coated 35-mm dishes in serum-free medium supplemented with 0.5 mg/ml DNase I at 37°C for 4 h, and the cells were photographed. Scale bar, 100 m (top). The degree of cell aggregation was quantified. Error bars represent mean Ϯ S.D.; n ϭ 3 (bottom). D, 48 h after the transfection as described in C, the cell lysates were examined for their contents of HAI-1 proteins by the immunoblotting with an anti-HAI-1 pAb under reduced conditions. ␤-Actin in the cell lysate was also detected by immunoblotting and used as an internal loading control.

Shed HAI-1 fragment has cell aggregation-inducing activity
MMP-7-catalyzed release of sHAI-1, as expected (Fig. 7A). The CM from WiDr cells treated as described above was then incubated with HT1080 cells previously treated without or with M␤-CD followed by MMP-7 treatment also under the suspended cell culture condition. As shown in Fig. 7A, the CM from the aggregated WiDr cells, but not that from the M␤-CDtreated WiDr cells, induced aggregation of the MMP-7-treated HT1080 cells regardless of the pretreatment of the HT1080 cells with M␤-CD, suggesting that CS-independent proteolytic action of MMP-7 on the cell surface is needed to endow the potential of the cells to form aggregates in the presence of sHAI-1. These data also suggest that the concentration of endogenous sHAI-1 in the CM of MMP-7-treated WiDr cells, which was determined to be ϳ4 nM, is sufficient to induce the aggregation of WiDr and HT1080 cells, both of which are treated with MMP-7.
To further verify that CS is not required for the sHAI-1mediated induction of cell aggregation, HT1080 cells in the suspended condition were pretreated without or with M␤-CD, and then incubated with MMP-7 in the presence of recombinant sHAI-1. The cells were also incubated with MMP-7(29,33,51,55/M2)⌬C3 and recombinant sHAI-1. As shown in Fig. 7B, the HT1080 cells were aggregated upon the MMP-7 treatment in the presence of sHAI-1 regardless of their pretreatment with M␤-CD. The variant of MMP-7 lacking CSbinding ability also induced the cell aggregation, but the cells incubated with sHAI-1 alone were not aggregated even after a 2-h incubation. The aggregation of the cells treated with

Shed HAI-1 fragment has cell aggregation-inducing activity
M␤-CD followed by MMP-7 treatment was slightly faster than that treated with MMP-7 alone, and as compared with wildtype MMP-7, the variant of MMP-7 lacking the affinity for CS induced the cell aggregation more effectively (Fig. 7B). These data suggest that CS-independent action of MMP-7 on the cell surface is critical for the sHAI-1-mediated induction of cell aggregation.

Region of HAI-1 corresponding to amino acid residues 141-249 is essential for cell aggregation-inducing activity
To explore the region of HAI-1 essential for induction of the homotypic cell aggregation, we constructed mammalian expression vectors for various domains-deleted variants of sHAI-1 (Fig. 8A). These vectors were stably transfected into CHO cells (Fig. 8B), and each of the variants of sHAI-1 secreted from the transfected cells was purified to homogeneity. As shown in Fig. 8C, all of the variants lacking two Kunitz-type inhibitor domains (KD), the low-density lipoprotein receptor (LDLR)-like domain, or the motif at the N terminus with seven cysteines (MANSC) domain of HAI-1 induced the cell aggregation, suggesting that these domains are not essential for the cell aggregation-inducing activity. These data also suggest that the region of HAI-1 corresponding to Leu 141 -Tyr 249 is essential for the induction of homotypic cell adhesion. The sHAI-1 variants that do not include the region between Leu 141 and Tyr 249 , such as sHAI-1(245-465) and sHAI-1(⌬141-249), however, were not secreted from CHO cells (Fig. 8B); therefore, further analysis was infeasible.
We then constructed an Escherichia coli expression vector for the region corresponding to Leu 141 -Tyr 249 of HAI-1, named HAI-1(141-249), and the protein was expressed in E. coli, refolded, and purified to homogeneity (Fig. 8C). The HAI-1(141-249) showed significant cell aggregation-inducing activity (Fig. 8D), suggesting that this region of HAI-1 is sufficient for the induction of homotypic cell aggregation. When the binding of exogenous HAI-1(141-249) to the cell surface was examined by fluorescence staining, using biotinylated-HAI-1(141-249) as a probe, the labeled protein was localized on the cell surface (Fig. 8D).
When the time course of binding of biotin-labeled HAI-1(141-249) to the surface of MMP-7-treated Colo201 cells was examined (Fig. 9A), the recombinant HAI-1 fragment rapidly bound to the cells, and the amount of the fragment bound to cells reached a constant after a 30-min incubation. Neither degradation nor decrease of the cell-bound fragment during the 5-h incubation was observed. The amount of HAI-1 fragment bound to the MMP-7-treated cells was much higher than that bound to the non-treated cells, suggesting that MMP-7 modifies cell-surface protein(s) and facilitates the binding of the HAI-1 fragment to the cell surface.

Shed HAI-1 fragment has cell aggregation-inducing activity
Cell ELISA revealed that HAI-1(141-249) bound to the MMP-7-treated Colo201 cells in a concentration-dependent and saturable manner (Fig. 9B). The half-maximal binding was observed at ϳ30 nM HAI-1(141-249). We found that HAI-1(141-249) bound to Colo201 cells without MMP-7 treatment, in a concentrationdependent manner, and the half-maximal binding was observed also at 30 nM HAI-1(141-249). At the saturating concentrations of HAI-1(141-249), the amount of the protein bound to the MMP-7-treated cells was much higher than that bound to the nontreated cells, suggesting that the MMP-7 treatment leads to an increase of the sites for HAI-1(141-249) binding on the cell surface. The binding of HAI-1(141-249) to the cells was also metal ion-dependent (Fig. 9C).
To examine whether sHAI-1 and HAI-1(141-249) compete with each other to bind to common site on the cell surface, a fixed concentration of biotin-labeled sHAI-1 or HAI-1(141-249), and various concentrations of non-labeled HAI-1(141-249) were incubated with the MMP-7-treated Colo201 cells, and the amount of labeled sHAI-1 or labeled HAI-1(141-249) bound to the cells was measured by cell ELISA. As shown in Fig.  9D, non-labeled HAI-1(141-249) effectively inhibited the binding of the labeled HAI-1(141-249) to the cells, as expected. In contrast, HAI-1(141-249) partially inhibited the binding of the labeled sHAI-1 to the cells, suggesting that HAI-1(141-249) and sHAI-1 share, at least in part, a common binding site on the cells.

Discussion
In this study, we identified HAI-1, a type I membrane protein, as a novel substrate of membrane-bound MMP-7; the Twenty four hours after transfection, the cells were washed two times with serum-free medium, and the culture was continued in serum-free medium. After 24 h, the CMs (top panel) and the cell lysates (middle and bottom panels) were harvested and subjected to immunoblotting (IB) with anti-FLAG or anti-HAI-1 antibody. ␤-Actin in the cell lysate was also analyzed by immunoblotting. The Mo and numbers on top of the lanes represent the loaded samples prepared from the mock-transfectant and those from the cells transfected with the constructs corresponding to the numbers on the left of the schemes in A, respectively. NS represents non-specific bands. Ordinate, molecular mass in kDa. C, approximately 1 g each of the purified sHAI-1 variants was analyzed by SDS-PAGE followed by CBB staining. The numbers on top of the lanes correspond to the numbers of the constructs shown on the left of the schemes in A. Ordinate, molecular mass in kDa (top). The aggregation-inducing activities of indicated sHAI-1 variants (each 50 nM) were analyzed by the aggregation assay, and the degree of cell aggregation was quantified as described under "Experimental procedures." Error bars represent mean Ϯ S.D.; n ϭ 3; Student's t test: *, p Ͻ 0.001 (bottom). D, Colo201 cells treated with MMP-7 followed by TAPI-1 and EDTA and then washed were incubated without (control) or with 50 nM biotin-labeled HAI-1(141-249) at room temperature for 1 h, and the labeled protein bound to the cells was visualized by staining with NeutrAvidin-FITC. Scale bar, 20 m (left). The aggregation assay was performed, using the Colo201 cells treated with MMP-7 followed by the TAPI-1 and EDTA treatment, in the presence of indicated concentrations of native (F) or boiled (E) HAI-1(141-249), and percent aggregation was determined as described under "Experimental procedures" (right).

Shed HAI-1 fragment has cell aggregation-inducing activity
MMP-7-catalyzed cleavage of HAI-1 on the surfaces of human colon cancer cells led to release of a 51-kDa fragment covering the extracellular region of the membrane protein. We also demonstrated that the released HAI-1 fragment acts as a cell-adhesion molecule.
It has been reported that a variant of MMP-7 lacking affinity for CS does not induce the cell aggregation (10), and binding of MMP-7 to CS is essential for MMP-7-catalyzed modulation of cell-surface proteins (9). This study showed that the variant of MMP-7 lacking CS-binding ability failed to shed HAI-1. Our data further suggest that active MMP-7 and a part of HAI-1 are localized in the lipid raft region of the cell membrane. These data are consistent with the view that colocalization of MMP-7 and HAI-1 in the CS-containing lipid rafts facilitates the cleavage of HAI-1.
The extracellular region of HAI-1 consists of an N-terminal MANSC domain (14), a polycystic kidney disease (PKD)-like domain (15), KD1, LDLR-like domain, and KD2. The KD1 of HAI-1 has inhibitory activities against various trypsin-type serine proteases, whereas the KD2 does not (16). HAI-1 was initially identified as an inhibitor of hepatocyte growth factor activator, a trypsin-type serine protease (17); and further analyses revealed that it has potent inhibitory activities against matriptase, hepsin, plasmin, and trypsin (18 -21). The expression of HAI-1 is increased during tissue remodeling and inflammation (22,23), and it is thought to regulate activation of hepatocyte growth factor precursor. It has been reported that the extracellular domain of HAI-1 is cleaved at several sites, and the pattern of cleavage changes in the presence or absence of EDTA, suggesting that metalloproteinases are involved in the cleavage (24). A recent study has reported that membrane type-1 MMP (MT1-MMP) cleaves HAI-1 at a peptide bond between Gly 451 and Leu 452 in the membrane-proximal external region, and at a site between KD1 and LDLR domain (25). We showed that the former site of HAI-1 cleaved by MT1-MMP is also cleaved by cell-associated MMP-7. HAI-1 is not known as a metal ioncontaining protein; however, sHAI-1 binds to the cell surface in a metal ion-dependent manner, suggesting that metal ions stabilize the functional structure of sHAI-1 or that of unidentified sHAI-1 receptor(s).
This study for the first time revealed that sHAI-1, generated by MMP-7-catalyzed cleavage, binds to the cell surface and plays a role in homotypic cell aggregation. As the MMP-7 treatment led to an increase of the sHAI-1-binding capacity of the cells, MMP-7 may modify and activate an unidentified cellsurface receptor(s) of sHAI-1. This study also demonstrated that the CS-independent proteolytic action of MMP-7 on cell surface is critical for the sHAI-1-mediated induction of cell aggregation. Considering that the CS-dependent and the CS-independent actions of MMP-7 are essential for the generation of sHAI-1 and sHAI-1-mediated induction of cell aggregation, respectively, it seems likely that MMP-7 acts as a specific inducer of the cell aggregation due to having the dual activities. For instance, some metalloproteinases other than MMP-7 that can shed HAI-1 will not be able to induce the cell aggregation if they do not have the activity corresponding to the CS-independent action of MMP-7 on cell surface. Further studies are needed to clarify the detailed mechanism.
We determined a region of sHAI-1 essential for the cell aggregation-inducing activity; the region of HAI-1 corresponding to amino acid residues Leu 141 -Tyr 249 , including the PKD-like domain, had the activity. A previous study reported that polycystin-1, which is membrane protein having multiple PKD domains, forms homodimer via its PKD domains, thereby contributing to cell-cell adhesion (26,27). Because the concentration of HAI-1(141-249) required for half-maximal induction of cell aggregation was lower than that of sHAI-1, the cell aggregation-inducing activity of HAI-1(141-249) is likely higher than that of sHAI-1. A recent report suggests that the PKD-like domain interacts with the neighboring KD1 in HAI-1, thereby modulating the protease inhibitor activity (15). The inter-domain interaction may partially hamper the binding of the 141-249 region of sHAI-1 to cell-surface receptor(s), thereby lowering their affinity. In addition to the 141-249 region, some other region(s) of sHAI-1 is likely involved in the interaction with cell-surface molecules, because the binding of sHAI-1 to the cells was only partially competed by the HAI-1(141-249) fragment. Although contribution of the additional region(s) of sHAI-1 is currently unknown, our present data strongly suggest that interaction between the region of HAI-1(141-249) and its corresponding receptor(s) on the cell surface is directly involved in the induction of homotypic cell adhesion.
HAI-1 has been considered to down-regulate hepatocyte growth factor activity, thus suppressing cancer malignancy and metastasis. It has been reported that knockdown of HAI-1 induces epithelial to mesenchymal transition (28), and cleavage of HAI-1 by MT1-MMP induces invasive growth of oral squamous cell carcinoma cells via increasing proteolytic activity of matriptase (25). This study further suggests that cleavage of HAI-1 promotes cancer metastasis via production of the cell adhesion molecule. Therefore, it is likely that MMP-7 converts the cancer-suppressive molecule into a cancer-promoting one. Our finding also provides the potential to develop sHAI-1targeted novel anti-cancer drugs that block the MMP-7promoted cancer metastasis.

Construction of expression vector for HAI-1 variants or shRNA targeting HAI-1 gene
In this study, gene constructions were carried out using PCR with PrimeSTAR Max DNA polymerase. Oligonucleotide sequences used as primers and inserts are listed in Table 1.
To construct a mammalian expression vector for the C-terminally-tagged sHAI-1, PCR was first carried out, using a pair of primers pEAK EcoRIϩ and sHAI EcoRIϪ and the pEAK8-HAI-1 as a template. The primers having a 15-base overlapped sequence, including a mutagenic one, were designed in inverted tail-to-tail directions to amplify the cloning vector together with the extracellular region of the HAI-1 sequence and to introduce an EcoRI site in the C-terminal side of the part of HAI-1 sequence. The resultant PCR product having adhesive tails due to the overlapped sequence was used directly for transformation, according to the manufacturer's instruction. The resultant pEAK8-sHAI-1 vector was cleaved with EcoRI and ligated with annealed oligonucleotides cFLϩ and cFLϪ. The resultant pEAK8-sHAI-1/cFL vector was used for the following constructions. To fuse the sequence encoding sHAI-1 and that of the FLAG tag and to add a linker sequence consisting of three tandem glycine residues, PCR was carried out, using a pair of primers sHAIcFLjϩ and sHAIcFLjϪ, and the pEAK8-sHAI-1/ cFL as a template. The resultant pEAK8-sHAI-1-Gly 3 -cFL vector was used for expression of the recombinant protein. To construct the N-terminally truncated sHAI-1 variant sHAI-1(141-465) or sHAI-1(245-465), PCR was carried out using a pair of primers HAI 141-HindIIIϩ and cFL EcoRIϪ or HAI 245-HindIIIϩ and cFL EcoRIϪ, respectively, and the pEAK8-sHAI-1-Gly 3 -cFL as a template. The resultant PCR product was cleaved with HindIII and EcoRI and ligated into the pSecTagA also cleaved with HindIII and EcoRI, and then used for transformation. For the C-terminally truncated variant HAI-1(36 -306) or HAI-1(36 -249), PCR was carried out using a pair of primers HAI 306 cFLjϩ and HAI 306 cFLjϪ or HAI 249 cFLjϩ and HAI 249 cFLjϪ, respectively, and the pEAK8-sHAI-1-Gly 3 -cFL as a template. For the internal sequence-deleted variant sHAI-1⌬141-249, PCR was carried out using a pair of primers HAI ⌬141-249ϩ and HAI ⌬141-249Ϫ, and the pEAK8-sHAI-1-Gly 3 -cFL as a template. These PCR products having adhesive tails were used directly for transformation.
To construct an expression vector for the N-terminallytagged HAI-1, PCR was first carried out, using a pair of primers pSec nFLϩ and pSec nFLϪ, and the pSecTag2B as a template. The primers were designed to amplify the cloning vector so that the FLAG tag is fused to the C terminus of the Ig⌲ leader sequence encoded in the vector. The resultant pSec-nFL-Tag2 was amplified by PCR with a pair of primers pSec EcoRIϩ and nFLϪ. The resultant PCR product was cleaved with EcoRI and ligated with annealed oligonucleotides HAI 37-46ϩ and HAI 37-46Ϫ, encoding the amino acid sequence corresponding to the N-terminal 10 residues of HAI-1 mature protein with silent mutations, to replace a GC-rich DNA sequence in the part of HAI-1 cDNA. The resultant pSec-nFL-HAI-1(37-46) was amplified by PCR with a pair of primers pSec EcoRIϩ and HAI 46Ϫ, and the resultant PCR product was cleaved with EcoRI. A part of cDNAs encoding the amino acid sequence corresponding to 47-529 of HAI-1 was also amplified by PCR with a pair of primers HAI 47ϩ and HAI 529 EcoRIϪ, and the pEAK8-HAI-1 as a template, and the resulting PCR product was cleaved with EcoRI. These two PCR products both cleaved with EcoRI were combined and ligated. The resultant pSec-nFL-HAI-1 vector was used for expression of the recombinant protein. To replace the Leu 452 of HAI-1 with glycine, PCR was carried out using a pair of primers HAI L452/Gϩ and HAI L452/GϪ, and the pSec-nFL-HAI-1 as a template. To further replace the Phe 376 and the Leu 379 of HAI-1 with glycine, PCR was carried out using a pair of primers HAI F376,L379/Gϩ and HAI F376,L379/GϪ, and the pSec-nFL HAI-1 L452/G as a template.

Shed HAI-1 fragment has cell aggregation-inducing activity
To construct a mammalian expression vector for the shRNA targeting the hai-1 gene, a pair of oligonucleotides HAI shRNAϩ and HAI shRNAϪ were annealed and ligated with pBAsi-hU6 Neo DNA cleaved with BamHI and HindIII. To construct a vector for the non-targeting shRNA, a pair of oligonucleotides NT shRNAϩ and NT shRNAϪ were annealed and ligated with pBAsi-hU6 Neo DNA as described above.
To construct an E. coli expression vector for the region of HAI-1 corresponding to amino acid residues 141-249 with an N-terminal FLAG tag, PCR was first carried out, using a pair of primers pnFL1stϩ and pnFL1stϪ, and the pFLAG-N-APP-IP-MMP-2-cat-FLAG, which was constructed in the previous study (29), as a template. The resultant PCR product was further amplified by PCR with a pair of primers pnFL2ndϩ and pnFL2ndϪ. The PCR product having adhesive tails was used directly for transformation. The cloning vector together with the N-terminal FLAG tag region of the resultant pnFL-APP-IP-MMP-2cat-FLAG was amplified by PCR with a pair of primers pnFL EcoRϩ and pnFLϪ, and the resultant PCR product was cleaved with EcoRI. A part of cDNA encoding the amino acid sequence corresponding to 141-249 of HAI-1 was also amplified by PCR with a pair of primers HAI 141ϩ and HAI 249 EcoRIϪ, and the pEAK8-HAI-1 as a template, and the resultant PCR product was cleaved with EcoRI. These two PCR products both cleaved with EcoRI were combined and ligated. The resultant pnFL-HAI-1(141-249) vector was used for expression of the recombinant protein in E. coli.

Cell lines and culture conditions
Human colon carcinoma cell lines WiDr, DLD-1, Colo201, human fibrosarcoma cell line HT1080, and CHO cell line were obtained from the Japanese Cancer Resources Bank. They were maintained in DME/F12 medium supplemented with 10% FBS, penicillin G, and streptomycin sulfate at 37°C in a humidified atmosphere of 5% CO 2 and 95% air.

Biotinylation of cell-surface proteins and detection of biotinylated protein fragments
WiDr cells were rinsed two times with serum-free medium and treated with the biotinylation reagent EZ-Link Sulfo-NHS-LC-biotin (50 g/ml) diluted with 50 mM HEPES (pH 7.5), containing 150 mM NaCl at 37°C for 20 min. The reaction was terminated by adding 0.1 M glycine in PBS. The surface-biotinylated cells were washed two times with serum-free medium and incubated in serum-free medium at 37°C for 1 h. The cells were then treated with 50 nM MMP-7 in the serum-free medium at 37°C for 15 min. The culture supernatant collected from the incubated cells was load on a SoftLink TM soft release avidin resin column (0.5-ml bed volume) previously equilibrated with 50 mM Tris-HCl (pH 7.5) containing 150 mM NaCl, 5 mM EDTA, and 0.1% Nonidet P-40, and biotinylated protein fragments released from the cells were allowed to be adsorbed. The column was washed with the equilibration buffer, and the biotinylated proteins adsorbed were eluted with the equilibration buffer containing 10 mM biotin. The eluted sample was analyzed by SDS-PAGE followed by ligand blotting, using alkaline phosphatase-conjugated streptavidin as a ligand.

In-gel digestion and MS analysis
The protein bands were excised from CBB-R250 -stained gel, destained, washed, and subjected to in-gel digestion as described previously (30) with slight modifications as described below. Briefly, the gel pieces were washed three times with 50 mM ammonium bicarbonate (pH 8.0), 60% acetonitrile (ACN). After completely dried, the gel pieces were incubated with 100 l of 50 mM ammonium bicarbonate (pH 8.0) in the presence of 10 mM DTT and 0.2 M guanidine HCl at 60°C for 1 h and were subsequently alkylated with an equal volume of 50 mM ammonium bicarbonate (pH 8.0) containing in 108 mM iodoacetamide at 37°C for 30 min in the dark. Next, the gel pieces were washed with 50 mM ammonium bicarbonate (pH 8.0), 60% ACN for 70 min (with a buffer change every 10 min) to remove the excess salt. After the gel pieces were completely dried, in-gel digestion was performed using 100 ng of mass spectrometry grade trypsin (Promega, Madison, WI) in 50 mM ammonium bicarbonate (pH 8.0) at 37°C overnight. For MS analysis, the resulting peptides were desalted and enriched using a selfpacked SDB/C18 tip column (Stage tip). Stage tips were prepared by packing Empore SDB XD (3M, Tokyo, Japan) and Empore C18 (3M) into a 200-l pipette tip as described previously (31). First, the columns were washed with 100 l of 80% ACN, 0.1% TFA by centrifuging at 3000 ϫ g and then equilibrated with 100 l of A buffer (2% ACN, 0.1% TFA) by centrifuging at 3000 ϫ g. After sample-loading by centrifugation at 1000 ϫ g, flow-through samples were reloaded, centrifuged, and then washed twice with 100 l of A buffer and were eluted with 100 l of 30% ACN, 0.1% TFA and then with 100 l of 60% ACN, 0.1% TFA. Eluted samples were completely dried and stored at Ϫ20°C.
To identify peptides, peak lists were created using a Proteome Discoverer software version 1.4 (Thermo Fisher Scientific) and were searched against the human protein sequences in the UniProt Knowledgebase (UniProtKB/Swiss-Prot) database (version May, 2013; 538,849 entries) using MASCOT (version 2.4.1, Matrix Science, London, UK). The search parame-Shed HAI-1 fragment has cell aggregation-inducing activity ters were as follows: trypsin digestion with three missed cleavages permitted; variable modifications, protein N-terminal acetylation, oxidation of methionine, propionamidation of cysteine, and biotin of lysine and N terminus; peptide-mass tolerance for MS data, Ϯ5 ppm; and fragment mass tolerance, Ϯ0.5 Da. We used significance threshold (p Ͻ 0.05) as a cutoff to export results from the analysis by MASCOT. In addition, peptides that yielded a peptide ion score, which was greater than or equal to 30, were considered positive identifications.

Preparation and fractionation of cell membrane by the differential centrifugation method
Colo201 cells (1 ϫ 10 7 cells) were washed two times with serum-free medium, and then homogenized in 1 ml of 20 mM HEPES buffer (pH 7.5) containing 250 mM sucrose by a Potter-Elvehjem-type homogenizer. The homogenates were centrifuged at 800 ϫ g for 7 min to remove nuclei and cellular debris. The supernatant was centrifuged at 21,000 ϫ g for 30 min, and resultant precipitate was used as the cell membrane fraction. To separate detergent-soluble and -insoluble fractions, the membrane fraction was suspended in 20 mM Tris-HCl (pH 7.5) containing 150 mM NaCl (TBS) supplemented with 10 mM CaCl 2 and 1% non-ionic detergent Triton X-100, and then incubated at 4°C. After incubation, the sample was centrifuged at 21,000 ϫ g at 4°C for 15 min. The resultant precipitate and supernatant were dissolved in an SDS sample buffer consisting of 50 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, and 20 g/ml bromphenol blue and subjected to SDS-PAGE.

Knockdown of HAI-1
The vector for expression of shRNA targeting the hai-1 gene and non-targeting shRNA constructed as described above were transfected into WiDr cells using Lipofectamine LTX and Plus reagent according to the manufacturer's instructions. Stable transfectants were selected with G418. The selection was performed by culturing the cells for 3 weeks in DME/F12 medium containing 600 g/ml G418. After selection, the G418resistant cells were cultured DME/F12 medium containing 600 g/ml G418.

Fluorescence-activated cell-sorting analysis
The various expression vectors were transfected into HT1080 cells using Lipofectamine LTX and Plus reagent. Forty eight hours after transfection, the cells were prepared for cellsurface labeling of HAI-1 as follows: the cells were washed in PBS and then the cells were removed from the culture dish with trypsin and EDTA. The cells were washed twice with PBS supplemented with 0.02% EDTA (PBSE). After 5 min of centrifugation at 800 ϫ g, they were suspended in PBSE containing 3% BSA. Cells were counted, diluted, and aliquoted into 100-l fractions that contained 5 ϫ 10 5 cells. Each sample was then incubated on ice for 1 h with a 1:50 dilution of anti-HAI-1 pAb. After repeated washings, each cell pellet was resuspended in ice-cold PBSE containing 1% BSA. Cells were then labeled with a 1:1000 dilution of rabbit anti-goat FITC in the dark, on ice, for 30 min. After extensive washing in ice-cold PBSE containing 1% BSA, the cells were suspended in 1 ml of PBSE containing 1% BSA and subjected to flow cytometric analysis using a JSAN cell sorter (Bay Bioscience, Kobe, Japan).

Expression and preparation of sHAI-1 variants tagged with FLAG epitope
The various expression vectors were transfected into CHO cells using Lipofectamine LTX and Plus reagent according to the manufacturer's instructions. Stable transfectants expressing the recombinant protein were selected with puromycin or Zeocin. The selection was performed by culturing the cells for 3 weeks in DME/F12 medium containing 5 g/ml puromycin or 800 g/ml Zeocin. The transfected CHO cells were grown to confluence in 25 ml of the growth medium in 150-mm dishes. To prepare the CMs, cells were rinsed with PBS, and the culture was continued in 15 ml of serum-free medium for 24 h. After incubation, the CM was harvested, clarified by centrifugation, and stored at Ϫ40°C until used for purification of recombinant proteins.
The frozen CM was thawed and added with ammonium sulfate to make an 80% saturated ammonium sulfate solution and stirred at 4°C for 15 h. The sample was then centrifuged at 13,000 ϫ g at 4°C for 30 min. The resultant precipitates were dissolved in TBS and dialyzed extensively against the same buffer. After dialysis, the sample was clarified by centrifugation and then loaded onto an anti-FLAG M2 mAb-conjugated agarose column equilibrated previously with TBS. The column was washed with the equilibration buffer, and soluble variants of HAI-1 tagged with FLAG were eluted with 100 g/ml FLAG peptide dissolved in TBS. The eluted sample from the anti-FLAG antibody column was dialyzed against 50 mM HEPES (pH 7.5), containing 150 mM NaCl.

Expression and purification of HAI-1(141-249)
The E. coli expression vector pnFL-HAI-1(141-249) constructed as described above was used for transformation of E. coli strain DH5␣ competent cells. The transformant was cultured in 2ϫ YT medium (0.08% (w/v) tryptone, 0.5% (w/v) yeast extract, and 0.25% (w/v) NaCl) at 37°C, and the recombinant protein was induced by the addition of 1.0 mM isopropyl ␤-Dthiogalactopyranoside. After a 5-h induction, E. coli cells were broken in 50 mM Tris-HCl (pH 8.0) containing 50 mM NaCl and 5 mM EDTA by sonication, and the resultant inclusion bodies were collected by centrifugation. The inclusion bodies were solubilized in 50 mM Tris-HCl (pH 8.0) containing 6 M guanidine HCl and 100 mM DTT with gentle stirring at 25°C for 2 h. The solubilized sample was first clarified by centrifugation and then refolded by the rapid dilution method using a refolding buffer consisting of 1 M arginine, 50 mM Tris-HCl, 150 mM NaCl, and 5 mM CaCl 2 in which the pH was adjusted to 7.5. The refolded protein was dialyzed extensively against TBS, and concentrated using a Centriprep-10 centrifugal filter device (Merck Millipore Ltd., Darmstadt, Germany). After concentration, the recombinant protein was purified with an anti-FLAG antibody column as described above. To remove lipopolysaccharide potentially included in the sample, the fraction eluted from the anti-FLAG antibody column was loaded onto a polymyxin B-agarose column equilibrated previously with TBS containing 10 mM CaCl 2 , and the flow-through fraction was collected. The collected Shed HAI-1 fragment has cell aggregation-inducing activity sample from the polymyxin B-agarose column was dialyzed against 50 mM HEPES (pH 7.5) and 150 mM NaCl, containing 10 mM CaCl 2 .

SDS-PAGE and immunoblotting analysis
SDS-PAGE was performed on polyacrylamide gel under non-reduced or reduced conditions. In immunoblotting analysis, proteins separated by SDS-PAGE were transferred onto nitrocellulose or PVDF membranes and visualized by the ECL method (GE Healthcare, Buckinghamshire, UK).

Assay of cell aggregation-inducing activities of variants of sHAI-1
Colo201 cells were incubated with 50 nM MMP-7 at 37°C for 3 h and then with 2 M TAPI-1 and 5 mM EDTA in the serumfree DME/F12 medium. These cells were dispersed by pipetting and washed two times with PBS. The cell density was adjusted to 5 ϫ 10 5 cells/ml, and the cells were further incubated without or with variants of sHAI-1 (each 50 nM) in serum-free DME/F12 medium containing 5 M TAPI-1 and 0.01% BSA at 37°C for 5 h. The degree of cell aggregation was quantified by following equation: (aggregated cells/total cell) ϫ 100, where the aggregates formed by over four cells are defined as aggregated cells.

Fluorescence staining
Colo201 cells treated without or with 50 nM MMP-7 as described above were plated on poly-L-lysine-coated plastic plates, fixed with ice-cold acetone/methanol for 15 min, and washed three times with PBS. After blocking the non-specific binding sites with 3% BSA in PBS, the cells were incubated with 50 nM biotinylated sHAI-1 or 50 nM biotinylated HAI-1(141-249) in TBS containing 10 mM CaCl 2 for 1 h at room temperature. The cells were then incubated with FITC-conjugated NeutrAvidin protein at room temperature for 1 h. The fluorescence image was observed using a fluorescence microscope (KEYENCE, Osaka, Japan).

Cell ELISA for measuring sHAI-1 or HAI-1(141-249) bound to cell surface
Colo201 cells were treated without or with 50 nM MMP-7 at 37°C for 3 h, and cells were added with 2 M TAPI-1 and 5 mM EDTA. The cells were washed and inoculated at a density of 1 ϫ 10 3 cells per well of 96-well plate (Sumilon, Tokyo, Japan) in 100 l of serum-free DME/F12 medium, and incubated at 37°C for 1 h. After incubation, the cells were fixed by adding an equal volume of 50% (v/v) glutaraldehyde and incubated at room temperature for 20 min. The fixed cells were washed three times with PBS and incubated with 3% BSA in PBS at 4°C overnight to block the non-specific binding sites. After blocking, the cells were incubated with various concentrations of biotinylated-sHAI-1 or -HAI-1(141-249) in TBS containing 10 mM CaCl 2 at 37°C for 1 h. Biotinylated proteins bound to the cell surface were allowed to react with HRP-conjugated streptavidin (Vector Laboratories, Burlingame, CA) by incubating at 37°C for 30 min. The cells were washed three times and then added with 180 l of chromogenic reaction mixture consisting of 75 mM citrate/phosphate buffer (pH 5.0), containing 3.7 mM -phenylenediamine and 0.01% H 2 O 2 . The reaction was terminated by adding with 50 l of 2 M H 2 SO 4 , and the intensity of the color developed was measured at 485 nm.

Statistical analysis
All experiments were carried out independently at least three times. Comparisons between the two groups were performed using Student's t test, with p Ͻ 0.05 considered to be significant.