Hepatocyte Growth Factor Activator Inhibitor Type 1 Is a Suppressor of Intestinal Tumorigenesis

Published OnlineFirst February 27, 2013; DOI: 10.1158/0008-5472.CAN-12-3337

Cancer Tumor and Stem Cell Biology Research

Shinri Hoshiko1,2, Makiko Kawaguchi1, Tsuyoshi Fukushima1, Yukihiro Haruyama1, Kenji Yorita1, Hiroyuki Tanaka1, Motoharu Seiki3, Haruhiko Inatsu2, Kazuo Kitamura2, and Hiroaki Kataoka1

Abstract Hepatocyte growth factor activator inhibitor type 1 (HAI-1/SPINT1) is a membrane-bound serine protease inhibitor expressed on the surface of epithelial cells. Although HAI-1/SPINT1 is abundantly expressed in the intestinal epithelium, its role in intestinal tumorigenesis is not known. In this study, we investigated the role of Hai-1/Spint1 in intestinal tumorigenesis using mouse models. The membranous Hai-1/Spint1 immunoreactivity þ was decreased in murine ApcMin/ tumors and also in carcinogen (azoxymethane treatment followed by dextran sodium sulfate administration)-induced colon tumors compared with the adjacent non-neoplastic epithelium. The decreased immunoreactivity appeared to be due to sheddase activity of membrane-type 1 matrix metallo- þ protease. Then, we examined the effect of intestine-specific deletion of Spint1 gene on ApcMin/ mice. The loss of þ Hai-1/Spint1 significantly accelerated tumor formation in ApcMin/ mice and shortened their survival periods. þ Activation of HGF was enhanced in Hai-1/Spint1–deficient ApcMin/ intestine. Gene expression profiling revealed upregulation of the Wnt/b-catenin signaling circuit, claudin-2 expression, and angiogenesis not only in tumor þ tissue but also in the background mucosa without macroscopic tumors in Hai-1/Spint1–deficient ApcMin/ intestine. Intestinal deletion of Spint1 also enhanced the susceptibility to carcinogen-induced colon tumorigenicity of wild-type Apc mice. Our findings suggest that HAI-1/SPINT1 has a crucial role in suppressing intestinal tumorigenesis, which implies a novel link between epithelial cell surface serine protease inhibitors and protection from carcinogenic stimuli. Cancer Res; 73(8); 1–12. 2013 AACR.

Introduction complexes that restrict paracellular macromolecular ex- The intestinal mucosal surface area represents the most change between mucosal and systemic compartments has significant anatomic and physiologic portal for systemic expo- been studied extensively and the molecular components of sure to environmental toxins and microorganisms (1). The tight junction proteins described (1, 2). However, how the cell systemic compartment is protected from these exposures by surface proteolysis systems participate in maintaining epi- fi the mucosal epithelial barrier, with barrier dysfunction con- thelial barrier function remains incompletely de ned. Recent tributing to local and systemic diseases including inflamma- evidence indicates that normal functioning of the cell surface tory bowel disease, allergic disorders and malignancies (1, 2). serine protease matriptase is critically required for epithelial – Regarding the epithelial barrier, the role for the junctional integrity (3 6). Matriptase, encoded by the ST14 gene, belongs to the type II transmembrane serine protease superfamily and is expressed on the epithelial cell surface of most epithelial tissues (6). Moreover, recent studies indicate that its cognate Authors' Affiliations: 1Section of Oncopathology and Regenerative Biol- cell surface inhibitor, hepatocyte growth factor activator 2 ogy, Department of Pathology, Section of Circulatory and Body Fluid inhibitor (HAI), also has crucial roles in epithelial functions Regulation, Department of Internal Medicine, Faculty of Medicine, Univer- 3 – sity of Miyazaki, Miyazaki; and Division of Cancer Cell Research, Institute (7 10). of Medical Science, The University of Tokyo, Tokyo, Japan HAI belongs to the type I family of transmembrane serine Note: Supplementary data for this article are available at Cancer Research protease inhibitors that consists of 2 members: HAI type 1 Online (http://cancerres.aacrjournals.org/). (HAI-1)/serine protease inhibitor Kunitz type 1 (SPINT1) and HAI-2/SPINT2, which both have 2 extracellular Kunitz S. Hoshiko and M. Kawaguchi contributed equally to this work. domains and a transmembrane domain and are encoded by The whole raw data of microarray analysis for gene expression are available the SPINT1 and SPINT2 genes, respectively (11). Unlike HAI-2/ at GEO (accession no. GSE40856). SPINT2, HAI-1/SPINT1 also has an N-terminal cysteine-rich Corresponding Author: Hiroaki Kataoka, Section of Oncopathology and domain and a low-density lipoprotein receptor type A-like Regenerative Biology, Department of Pathology, Faculty of Medicine, University of Miyazaki, 5200 Kihara, Kiyotake, Miyazaki 889-1692, Japan. domain (11). HAI-1/SPINT1 is expressed in most epithelial Phone: 81-985-852809; Fax: 81-985-856003; E-mail, cells, with significant expression in enterocytes (8, 12). HAI-1/ [email protected] SPINT1 was initially discovered as an endogenous inhibitor of doi: 10.1158/0008-5472.CAN-12-3337 HGF activator, a serum-derived serine protease (13), but its 2013 American Association for Cancer Research. membrane association may also allow it to regulate cell surface

www.aacrjournals.org OF1

Hoshiko et al.

serine proteases. Indeed, ample evidence indicates that HAI-1/ lesion size was scored as follows: <1 mm, score 1; <2 mm, score SPINT1 is a cognate inhibitor of matriptase on epithelial cell 2; <3 mm, score 3; >3 mm, score 4. For the chemical carcino- fl fl fl fl surfaces (10, 14–17). Other TTSPs, such as hepsin, human genesis study, Spint1 ox/ ox/Vil-Cre mice and Spint1 ox/ ox airway trypsin-like protease, TMPRSS4 and TMPRSS13, as well control mice were used for combined treatment with azox- as the glycosylphosphatidylinositol-anchored cell surface ser- ymethane (AOM) and DSS (24). At 8 weeks of age all mice (male, fl fl fl fl ine protease prostasin, may also be regulated by HAI-1/SPINT1 n ¼ 18 and 23 for Spint1 ox/ ox/Vil-Cre and Spint1 ox/ ox, (6, 18–20). Therefore, HAI-1/SPINT1 must have a critical respectively) were treated with a single intraperitoneal injec- regulatory role in pericellular proteolysis of epithelial cells. In tion of AOM (Sigma-Aldrich; 10 mg/kg body weight). One week fact, HAI-1/SPINT1 has essential functions in placental tro- after the AOM injection, mice were exposed to 1% DSS (molec- phoblasts, epidermal keratinocytes, and hair follicle epithelium ular weight 36–50 kDa; MP Biomedicals) in the drinking water (7, 21). Recently, we reported that conditional deletion of the for 5 days and thereafter received no further treatment. All Spint1 gene in intestinal epithelial cells results in enhanced mice were sacrificed 20 weeks after the AOM treatment. For intestinal permeability and susceptibility to dextran sulfate histologic analysis, formalin-fixed, paraffin-embedded tissue sodium (DSS)-induced colitis (8). As intestinal epithelial integ- sections were stained with hematoxylin and eosin (H&E) or rity has an important protective role against mucosal inflam- processed for immunohistochemistry. mation and neoplastic progression (1, 2, 22), HAI-1/SPINT1 may have a significant role in intestinal tumorigenesis. More- Immunohistochemistry, immunofluorescence, over, all major physiologic HGF activators (matriptase, hepsin, immunoblotting, and gene expression analysis HGF activator) are sensitive to HAI-1/SPINT1, and previous The following antibodies were used: anti-human HAI-1 study revealed that colorectal cancer cells show decreased cell monoclonal antibody 1N7 (12), anti-mouse Hai-1 goat poly- surface HAI-1/SPINT1 immunoreactivity with enhanced HGF clonal antibody raised against recombinant Hai-1 lacking activation in the cancer tissue (23). However, little is known transmembrane and intracytoplasmic domains (R&D Sys- about how HAI-1/SPINT1 may participate in intestinal tems), anti-mouse/human Hai-1 rabbit polyclonal antibody tumorigenesis. raised against intracytoplasmic domain (Pro484-Leu507; In this study, we evaluated the role of HAI-1/SPINT1 in ref. 25), anti-mouse/human matriptase (AnaSpec), anti-mouse intestinal tumorigenesis using intestinal specific Hai-1/Spint1 Hgf (R&D Systems), anti-mouse b-catenin (Sigma-Aldrich), knockout mice (8) in 2 intestinal carcinogenesis mouse models: anti-phosphorylated c-MET (Y1235; ref. 26), and anti-mouse an Apc mutant mouse and 2-step chemical carcinogenesis in CD31 (AnaSpec) rabbit polyclonal antibodies. The protocol for mouse colon. Our results from both models confirm that immunohistochemistry of paraffin-embedded sections of sur- Hai-1/Spint1 has a critical role in suppressing intestinal gically resected colorectal tumors (4 cases) was approved by tumorigenesis. the ethical board of the Faculty of Medicine, University of Miyazaki (Miyazaki, Japan). The immunohistochemical methods used were described previously (12, 26, 27). Immunoflu- Materials and Methods orescence and immunoblot analyses were conducted as Animal experiments described (8). Preparation of Hgf-enriched tissue extract and All animal studies were reviewed and approved by the analysis of Hgf processing by immunoblot were conducted as Committee on the Ethics of Animal Experiments of the Uni- described previously (23). Reverse transcription (RT)-PCR and versity of Miyazaki Animal Research Committee, in accordance quantitative real-time RT-PCR (qRT-PCR) were carried out with international guidelines for biomedical research involving using the primer sequences and methods described in Sup- animals. All mice were housed in a pathogen-free barrier plementary Table S1. Microarray analysis was conducted as þ environment for the duration of the study. ApcMin/ mice were described previously (27) using a GeneChip Mouse Genome obtained from The Jackson Laboratory. Mice with intestinal 430 2.0 Array (Affymetrix). fl fl epithelial cell–specific deletion of the Spint1 gene (Spint1 ox/ ox/ Vil-Cre) were generated by interbreeding mice carrying Cell culture, gene knockdown, and forced expression fl fl Spint1/LoxP homozygous alleles (Spint1 ox/ ox) with Cre The MT1rev2 cell line was established from Mmp14-null transgenic mice under the control of the intestine-specific mice intestine as described previously (28). MT1rev2 cells carry þ villin promoter (8). ApcMin/ mice were then interbred an exogenous mouse Mmp14 gene whose expression is sup- fl fl þ with Spint1 ox/ ox/Vil-Cre mice, generating ApcMin/ mice pressed in the presence of doxycycline (1 mg/mL; Sigma) so that þ with intestine-specific deletion of the Spint1 gene (ApcMin/ / cells express MT1-MMP/Mmp14 in doxycycline-free medium. fl fl Spint1 ox/ ox/Vil-Cre). All mice had a C57BL/6 genetic back- To detect secreted Hai-1/Spint1, cells were washed 3 times ground and were fed a semipurified AIN-76 diet under specific with PBS and cultured in serum-free medium without doxy- þ pathogen-free condition. For ApcMin/ tumorigenesis assays, cycline for the indicated time period. The culture supernatant age- and sex-matched mice were selected to assess tumor was then concentrated in an Amicon-Ultra-4 (Millipore) and formation in the small and large intestine. At 15 weeks of age, used for immunoblotting þ fl fl all mice (male, n ¼ 19 and 15 for ApcMin/ /Spint1 ox/ ox/Vil-Cre þ fl fl and ApcMin/ /Spint1 ox/ ox, respectively) were sacrificed by Statistics diethyl ether inhalation, and the numbers and major diameters Statistical analysis was done using GraphPad Prism, version of intestinal polyps were determined. The polypoid tumor 5.04 (GraphPad Software Inc.).

OF2 Cancer Res; 73(8) April 15, 2013 Cancer Research

HAI-1 Suppresses Intestinal Tumorigenesis

þ fl fl Results ApcMin/ /Spint1 ox/ ox mice, which became apparent at 7 weeks of age (Supplementary Fig. S1) and significantly shorter Decreased membranous HAI-1/SPINT1 þ survival times (P ¼ 0.0009, log-rank test; Fig. 2B). The ApcMin/ immunoreactivity accompanies neoplastic progression Min/þ þ/ of intestinal epithelium mice with heterozygous Spint1 (Apc /Spint1 ) had an In accordance with previous observations (23, 25), the intermediatesurvivalcurve(Fig.2C).Themajorcauseofdeath was severe anemia probably due to increased bleeding from membranous immunoreactivity of HAI-1/SPINT1 was þ the intestinal tract, which was more evident in ApcMin/ / markedly reduced in colon adenocarcinoma cells compared fl fl Spint1 ox/ ox/Vil-Cre mice (mean hemoglobin concentration, to adjacent adenoma cells and non-neoplastic epithelium (Fig. Min/þ flox/flox 1A). This trend was also observed in a murine model of 5.1 g/100 mL) than in Apc /Spint1 mice (9.3 g/100 intestinal tumorigenesis mediated by mutation of the Apc mL) at 15 weeks of age (Supplementary Table S2). þ gene (ApcMin/ mice; Fig. 1B). The neoplastic epithelium also showed significantly decreased membranous Hai-1/Spint1 Loss of intestinal Hai-1/Spint1 accelerates tumor ApcMin/þ immunoreactivity compared with the adjacent non-neoplastic formation in mice epithelium. Notably, similar to human colon cancer tissue (25), We then analyzed the effect of intestinal epithelium Spint1 þ fi the Spint1 mRNA level was essentially preserved in ApcMin/ deletion on intestinal tumorigenesis. Mice were sacri ced at 15 tumor tissue (Fig. 1B), suggesting that the decreased immu- weeks of age, and the number of visible neoplastic polyps was – fi Min/þ noreactivity may be caused by enhanced shedding of mem- counted. Notably, intestinal Hai-1/Spint1 de cient Apc fi brane-bound Hai-1/Spint1 from the neoplastic cell surface. mice showed signi cantly enhanced tumor formation in the small intestine (Fig. 3A and B), a major site of tumor formation Given that HAI-1 shedding is reported to be mediated by þ in ApcMin/ mice but not in the colon (Fig. 3C). The mean metalloprotease activity (29) and in human oral carcinoma Min/þ flox/flox cells MT1-MMP/MMP14 is responsible for HAI-1/SPINT1 multiplicity (no. of tumors/mouse) in Apc /Spint1 / Vil-Cre male mice was 95.0, which was significantly higher than ectodomain shedding (30), we compared Mmp14 expression þ fl fl þ Min/ ox/ ox in tumor tissue with corresponding normal mucosa in ApcMin/ that of the control Apc /Spint1 male mice (38.5 per intestine and found that the Mmp14 mRNA level was mouse), and the difference was particularly evident in the significantly increased in tumor tissue (Fig. 1C). The effect of proximal small intestine (Fig. 3C). The sizes of tumors formed MT1-MMP/Mmp14 on Hai-1/Spint1 shedding was then exam- by the 15th week tended to be larger in the intestinal Spint1- ined using the immortalized mouse intestinal epithelial cell deleted mice than those in the control mice (Fig. 3D). Histo- fi line MT1rev2, which was established from Mmp14-knockout logically, signi cant morphologic differences were not mice and carries an inducible mouse Mmp14 gene that drives observed between Spint1-deleted tumors and control tumors exogenous Mmp14 expression under tetracycline off condi- (Fig. 3B). tions (28). In the MT1rev2 cells, MT1-MMP/Mmp14 expression did indeed result in enhanced Hai-1/Spint1 shedding into the Absence of Hai-1/Spint1 enhances activation of Hgf culture supernatant generating 58- and 42-kDa secreted forms We compared the mRNA levels of Hai-1/Spint1 and its as previously reported (ref. 30; Fig. 1D). In accordance with important target membrane proteases, matriptase, by qRT- PCR. While matriptase (St14) mRNA levels were not signifi- these observations, mature membrane form Hai-1 was þ þ Min/ markedly decreased in extracts of ApcMin/ tumors accompa- cantly altered in the intestinal tumor tissue of Apc mouse nying increased amounts of 58- and 42-kDa forms and further (Fig. 4A), its subcellular localization was altered in the tumor degraded fragments (Fig. 1E). Consequently, C-terminus frag- cells, showing a decreased membranous localization and ments smaller than 15 kDa were also increased in the tumor increased intracellular immunoreactivity with punctuate tissue (Fig. 1E). staining (Fig. 4A). It should be noted that in normal epithelium, matriptase localization was not altered even in the absence of Hai-1/Spint1 as long as the epithelial and villous architectures þ ApcMin/ mice deficient in intestinal Hai-1/Spint1 have were retained. shorter survival times HGF is secreted as an inactive proform (proHGF) and To test whether the loss of membrane-bound HAI-1/SPINT1 requires proteolytic activation to induce c-MET–mediated is simply an epiphenomenon of the neoplastic transformation signaling (11). As all major HGF-activating proteases (HGF process in intestinal epithelial cells or has a causal role in activator, matriptase, and hepsin) are inhibited by HAI-1/ carcinogenesis and progression, we tested the effect of Hai-1/ SPINT1 (11, 13), we examined the activation of proHgf in Spint1 defect on tumorigenesis in Apc mutant mice. By inter- Spint1-deleted mouse intestinal tissues in the early stage (10 fl fl þ breeding intestine-specific Spint1-deleted mice (Spint1 ox/ ox/ weeks old) of ApcMin/ tumorigenesis (n ¼ 4). Even at 10th þ þ Vil-Cre; ref. 8) with ApcMin/ mice, we generated ApcMin/ mice week, visible tumors are found in small intestine and the having a conditional deletion of the Spint1 gene in the intestinal number was apparently higher in Spint1-deleted intestine. As þ fl fl epithelium (ApcMin/ /Spint1 ox/ ox/Vil-Cre). Hai-1/Spint1 shown in Fig. 4B, processing of Hgf was enhanced in Spint1- þ immunoreactivity was lost in most intestinal epithelial cells deleted ApcMin/ tumor tissue compared with control tumor þ fl fl in ApcMin/ /Spint1 ox/ ox/Vil-Cre mice, although focal residual tissue. Notably, background mucosa without visible tumor immunoreactivity in a small group of cells was occasionally (hereafter nontumor mucosa) also showed enhanced Hgf þ observed (Fig. 2A). The intestinal Spint1-deleted ApcMin/ mice activation in response to Spint1 deletion. In accordance with showed decreased body weight gain compared with control these observations, the immunoreactivity of phosphorylated c-

www.aacrjournals.org Cancer Res; 73(8) April 15, 2013 OF3

Hoshiko et al.

Figure 1. Decreased membranous HAI-1/SPINT1 in intestinal tumor cells and its shedding by MT1-MMP/MMP14. A, immunohistochemistry of HAI-1/SPINT1 in human colon adenomas and carcinomas. Serial sections were stained with H&E and anti-HAI-1 antibody (bar, 100 mm). B, immunohistochemistry þ of Hai-1/Spint1 in intestinal tumors of ApcMin mice (bar, 50 mm). Data for qRT-PCR of Spint1 mRNA levels in tumor and corresponding nontumor mucosa tissue (n ¼ 3) are also indicated. C, enhanced MT1-MMP/Mmp14 expression in mouse intestinal tumor compared with the corresponding normal mucosa. Data for qRT-PCR of Mmp14 mRNA levels (n ¼ 5) are also indicated. D, doxycycline (Dox) off-induced expression of MT1-MMP/Mmp14 (RT-PCR) and enhanced shedding of Hai-1/Spint1 by MT1-MMP/Mmp14 (immunoblot) in MT1rev2 cells. sHai-1, secreted form Hai-1. Data in bar graphs are mean SD. , P ¼ 0.008; Mann–Whitney U test. E, decreased mature membrane form and increased cleaved fragments of Hai-1/Spint1 in ApcMin/þ tumor tissues. Arrow and arrowhead indicate fragments compatible with 58- and 42-kDa sHAI-1, respectively, observed in MT1-MMP/Mmp14-expressing MT1rev2 supernatant. , about 15-kDa C-terminus fragments of Hai-1/Spint1.

þ þ fl fl Met was increased in Spint1-deleted ApcMin/ tumor cells compared with those of control ApcMin/ /Spint1 ox/ ox mice compared with control cells (Fig. 4B). using microarray, and the expression of selected genes was further confirmed by qRT-PCR. The entire raw data set for the Wnt signaling circuit is augmented in response to microarray analyses is available at Gene Expression Omnibus þ intestinal deletion of Spint1 in ApcMin/ mice (GEO). Compared with control tumors from mice of matched þ Comprehensive gene expression profiles of intestinal muco- age, Hai-1/Spint1–deficient ApcMin/ tumors from 15-week-old þ fl fl sa and tumors from ApcMin/ /Spint1 ox/ ox/Vil-Cre mice were animals showed upregulation of 22 genes that were related to

OF4 Cancer Res; 73(8) April 15, 2013 Cancer Research

HAI-1 Suppresses Intestinal Tumorigenesis

the Wnt signal pathway, including ligands, receptors, tran- Intestinal Hai-1/Spint1 is required for resistance to scription factors, and downstream genes (Table 1). Among chemical carcinogen-induced colonic carcinogenesis these genes, 8 (Apcdd1, Prox1, Sox17, Wif1, Fzd3, Igfbp4, Ptk7, To test whether a similar phenomenon can be reproduced Tcf4) were also upregulated in the nontumor mucosa of Spint1- under other carcinogenic conditions without germline muta- þ þ deleted ApcMin/ intestine compared with control ApcMin/ tion of Apc, we examined the effect of Spint1 deletion on chem- intestine (Table 1). In addition to those 22 genes, 11 Wnt ical carcinogenesis of the colon using mice with a wild-type Apc fl fl fl fl signal–related genes, including Fzd6, Wnt11, Wnt5A, Ror2, and gene. Spint1 ox/ ox/Vil-Cre and control Spint1 ox/ ox mice Tcf7 (Tcf-1), were upregulated in nontumor mucosa in received an injection of AOM followed by administration of response to Spint1 deletion (Table 1). Notably, Wnt11, Apcdd1, 1% DSS in the drinking water for 5 days (Fig. 5A). This and Sox17 mRNA levels in the Spint1-deleted nontumor muco- carcinogenesis treatment was not efficient for the control fl fl sa were comparable with those in tumor tissue of control Spint1 ox/ ox mice, where only 17% of treated mice (4 of 23) þ ApcMin/ mice (Supplementary Fig. S2). Immunohistochemi- showed tumor formation (Fig. 5B). However, defects in epithe- cally, tiny foci of b-catenin nuclear translocation, suggesting lial Hai-1/Spint1 conferred significant susceptibility to AOM/ fl fl preneoplastic lesion or microadenoma, were frequently DSS-induced tumorigenesis, as 67% (12 of 18) of Spint1 ox/ ox/ observed in Spint1-deleted nontumor mucosa but not in Vil-Cre mice showed tumor formation in the colon with control nontumor mucosa (Fig. 4C). increased multiplicity compared with the control mice (Fig. 5B). The mean size of the formed tumors was similar between Gene expression signature suggests enhanced mucosal the 2 groups (3.73 and 3.58 mm for Spint1-deleted colon and þ fl fl permeability and angiogenesis in Spint1-deleted ApcMin/ control colon, respectively; Fig. 5B). The Spint1 ox/ ox/Vil-Cre intestine mice that had sham treatment did not show tumor formation. It þ The loss of Hai-1/Spint1 also induced upregulation of several should be noted that similar to the ApcMin/ mouse model genes possibly related to tumor progression and angiogenesis described above, the intestinal site of tumor formation was not (Supplementary Table S3). While MT1-MMP/Mmp14 was altered by the absence of Hai-1/Spint1 with predominant tumor þ upregulated in control ApcMin/ tumors compared with cor- formation occurring in the distal colon in both groups (Fig. 5C). responding nontumor mucosa (Fig. 1C), it was also upregu- While tumors in both groups showed differentiated tubular þ lated in Hai-1/Spint1–deficient ApcMin/ tumors (6.66-fold). adenocarcinoma, Spint1-deleted tumors showed less differen- Among these genes, upregulation of Flt1, Cdh13, Cdh5, and tiated features than control tumors (Fig. 5D). Tnfrsf12a strongly suggested enhanced angiogenesis in Spint1- deleted tumor and nontumor mucosa compared with control tumor and nontumor mucosa, respectively (Supplementary Discussion Table S3 and Fig. 4D; refs. 31–33). Furthermore, other genes Using mouse knockout technology, we explored the function indicating increased angiogenesis, such as Vegfb, Tek, Egfl7, of the membrane-associated serine protease inhibitor, Hai-1/ Epas1, Eng, Fgfr1, Ctgf, Pf4, Hmox1,andSrf,werealso Spint1, in intestinal carcinogenesis and found that Hai-1/ upregulated in Spint1-deleted nontumor mucosa, although Spint1 has a novel role as a suppressor of carcinogenesis in their upregulation was not apparent in tumor tissue (Sup- the intestinal tract. The absence of Hai-1/Spint1 in enterocytes plementary Table S4). Pml,atumorsuppressorwithanti- resulted in significantly accelerated tumor formation in þ angiogenic function (34), was decreased (Supplementary ApcMin/ mice and conferred susceptibility to chemical carci- Table S4 and Fig. 4D). In fact, CD31 immunostain revealed nogenesis in the colons of wild-type Apc mice. In accordance increased vascular density in Spint1-deleted tumors (P ¼ with these findings, cell surface Hai-1/Spint1 expression was þ 0.0005), which may result in enhanced bleeding (Supple- decreased in ApcMin/ tumor cells in mice, and importantly, mentary Fig. S3). Cldn2 was also upregulated in nontumor human colorectal cancer cells also showed decreased cell mucosa in response to Spint1 deletion as well as in tumor surface HAI-1/SPINT1 levels compared with adjacent normal tissues (Fig. 4D). Cldn2 encodes claudin-2 that enhances and adenoma epithelia. Our data also suggest that the tight junction permeability (4), which may allow increased decreased amount of membrane-associated Hai-1/Spint1 in permeation of genotoxic substances. In fact, several genes the intestinal tumor cells may be due to proteolytic ectodo- related to DNA damage and/or repair were upregulated in main shedding induced by MT1-MMP/Mmp14, which implies þ fl fl the nontumor mucosa of ApcMin/ /Spint1 ox/ ox/Vil-Cre a novel pathway for MT1-MMP/MMP14 contribution to early mice compared with that of control mice (Supplementary neoplastic progression of intestinal epithelium via the deple- Table S5). tion of membrane-anchored HAI-1/SPINT1. þ In contrast, Spint1-deleted small intestine without Apc It should be noted that in contrast to the mouse ApcMin/ fl fl mutation (Spint1 ox/ ox/Vil-Cre) did not show significantly tumor in this study, previous studies of human colon cancer upregulated expression of Cldn2, Flt1, Apcdd1, and Sox17, showed decreased SPINT1 mRNA in cancer tissues (23, 35). fl fl compared with control Spint1 ox/ ox mouse intestine, and only However, given the nearly complete loss of cell surface HAI-1/ Tnfrsf12a showed statistically significant upregulation (Sup- SPINT1 immunoreactivity, the reduction levels of SPINT1 plementary Fig. S2). Consequently, no tumor formation was mRNA were not sufficient in most cases: 20% and 40% reduc- fl fl observed in the intestine of Spint1 ox/ ox/Vil-Cre mice with tion in well-differentiated carcinomas and overall, respectively wild-type Apc gene during the observation period (15 weeks; (23, 35). In situ hybridization study also revealed a substantial data not shown). amount of SPINT1 mRNA in tumor cells that showed loss of cell

www.aacrjournals.org Cancer Res; 73(8) April 15, 2013 OF5

Hoshiko et al.

surface HAI-1/SPINT1 (25). Therefore, in human colon cancer, posttranslational enhanced ectodomain shedding might be we suggest that both decreased transcription and enhanced an important mechanism underlying the loss of cell surface ectodomain shedding contribute in a coordinated manner to HAI-1/SPINT1 in tumor cells. the observed signiﬁcant reduction of cell surface HAI-1/ Multiple mechanisms could potentially contribute to the SPINT1. Recent genome-wide analyses for mutated genes in suppressive effect of the membrane-associated Hai-1/Spint1 human colon cancer did not identify relevant mutation or on intestinal tumorigenicity, but so far, the evidence suggests signiﬁcant change of gene expression level of SPINT1, and only that its roles for epithelial integrity to maintain the barrier one point mutation, which results in R390H (R406H in HAI-1B, function and regulation of Hgf processing are responsible. A a common splicing variant), was found (36). Collectively, previous study indicated that although mice lacking intestinal

A Apc Min/+Spint1flox/flox Apc Min/+Spint1flox/floxVil-Cre

Apc Min/+ Spint1flox/flox

pSI dSI pLI dLI Hai-1 Matriptase β-Actin

ApcMin/+ Spint1flox/flox Vil-Cre pSI dSI pLI dLI Hai-1 matriptase β-Actin

B C 100 100 Min/+ +/+ ApcMin/+Spint1flox/flox Apc Spint1

80 80

60 60

40 40 Min/+ +/- Survival rate (%) Survival rate (%) Apc Spint1

20 20 Min/+ flox/flox Apc Spint1 Vil-Cre Min/+ flox/flox Apc Spint1 Vil-Cre

0 0 0 525351510 20 30 0 525351510 20 30 Age (wks) Age (wks)

þ Figure 2. Generation of intestinal Hai-1/Spint1-deficient ApcMin/ mice and their survival curves. A, Hai-1/Spint1 immunostaining of small intestine from control (ApcMin/þ/Spint1flox/flox) and intestinal epithelium–specific Spint1-knockout mice (ApcMin/þ/Spint1flox/flox/Vil-Cre). Focal residual Hai-1/Spint1 immunoreactivity is indicated by arrows. Bars, 100 (top) and 20 (bottom) mm. RT-PCR for Spint1 and matriptase mRNAs in proximal small intestine (pSI), distal small intestine (dSI), proximal large intestine (pLI), and distal large intestine (dLI) are also shown. B, Kaplan–Meier survival curves of ApcMin/þ/Spint1flox/flox/Vil- Cre (n ¼ 12, all males) and control mice (ApcMin/þ/Spint1flox/flox; n ¼ 10, all males). The Spint1-mutant mice showed shorter survival times (P ¼ 0.0009, log-rank test). C, survival curves of ApcMin/þ/Spint1flox/flox/Vil-Cre (n ¼ 20, 12 males and 8 females), ApcMin/þ/Spint1þ/ (Spint1 heterozygous; n ¼ 5, one male and 4 females), and ApcMin/þ/Spint1þ/þ (wild-type Spint1; n ¼ 20, 8 males and 12 females) mice. ApcMin/þ/Spint1flox/flox/Vil-Cre mice showed shorter survival times compared with ApcMin/þ/Spint1þ/ (P ¼ 0.0008, log-rank test) and ApcMin/þ/Spint1þ/þ (P < 0.0001, log-rank test) mice.

OF6 Cancer Res; 73(8) April 15, 2013 Cancer Research

HAI-1 Suppresses Intestinal Tumorigenesis

Hai-1/Spint1 showed normal development and growth, intes- HGF/c-MET in cancer progression have been established (38). tinal permeability was increased, and these mice had enhanced The enhanced Hgf activation may also be involved, at least þ susceptibility to DSS-induced colitis (8). In this study, Hai-1/ partly, in enhanced angiogenesis in ApcMin/ tumors (38). þ Spint1 loss induced acceleration of tumor formation in ApcMin/ The cross-talk between HGF/c-MET signaling and WNT/ mice or when mice received a carcinogenic insult but not in b-catenin signaling has been reported. The activated HGF mice with wild-type Apc or untreated mice. Therefore, the secreted by stromal myofibroblasts reportedly activates pivotal function of Hai-1/Spint1 likely emerges when epithelial b-catenin–dependent transcription and subsequent colon cells encounter pathologic stimuli and/or cellular stress due to cancer stem cell clonogenicity and also restores the cancer genetic abnormalities. Under adverse cellular conditions, the stem cell phenotype in more differentiated tumor cells (39). loss of Hai-1/Spint1 may result in increased epithelial perme- HGF-induced activation of WNT pathway via transcriptional ability that could amplify these pathologic stimuli and accel- activation of lymphoid enhancer–binding factor 1 (LEF1) was erate accumulation of genetic abnormalities. Upregulation of also reported in human cancer cells (40). Indeed, comprehen- claudin-2 expression would further enhance epithelial perme- sive gene expression profiling revealed that the Spint1-deleted þ þ ability in Spint1-deleted ApcMin/ mice, as this particular ApcMin/ intestine showed significant upregulation of proteins claudin correlates with epithelial leakiness (4, 5). Moreover, involved in the Wnt signaling circuit and enhanced angiogen- claudin-2 is reported to be involved in colon tumorigenesis esis even in nontumor mucosa, which would provide a cancer- (37). Enhanced activation of Hgf and its specific receptor c-Met promoting microenvironment. T-cell factor (Tcf) 4, part of the may also be an important mechanism for tumorigenesis Tcf4–Lef1 complex, was upregulated in Spint1-deleted mucosa. enhanced by Hai-1/Spint1 loss, as HAI-1/SPINT1 is a potent In addition, recent study suggested that Tcf7 (also known as inhibitor of major proHGF-activating proteases, HGF activa- Tcf-1) and Lef1 are obligate partners in the oncogenic activities tor, matriptase, and hepsin (6, 13, 16), and important roles of of b-catenin in Apc-mutant mice (41), so it is of interest that

Figure 3. Enhanced tumor formation in the small intestine of ApcMin/þ mice in response to Spint1 deletion. A, macroscopic appearance of small intestines from ApcMin/þ/Spint1flox/flox (Cont) and ApcMin/þ/Spint1flox/flox/Vil-Cre (KO) mice (bar, 1 cm). B, microscopic findings of intestinal tumors from ApcMin/þ/Spint1flox/flox (Cont) and ApcMin/þ/Spint1flox/flox/Vil-Cre (KO) mice. In whole intestine sections (top 2 panels, H&E), enhanced tumor formation is evident in KO mice (bar, 1 mm). Higher magnification photos of control and KO tumors (bottom left 2 panels, H&E) indicate severely dysplastic tubules of similar histology (bar, 20 mm). Immunohistochemistry for Hai-1/ Spint1 (bottom right 3 panels) shows absence of Hai-1/Spint1 in both normal epithelium and tumor cells of KO intestine. In control intestine, neoplastic epithelium (T) showed decreased immunoreactivity of cell surface Hai-1/Spint1. Bars represent 100 and 20 mm for low- and high- magnification images, respectively. C, number of visible polyps in the intestines (mean SEM). , P ¼ 0.0001; , P ¼ 0.0084 (Mann– Whitney U test). D, comparison of polyp size scores. The box shows the interquartile range, the whiskers the largest and smallest observed scores that are <1.5 box length from the end of the box, and the median is indicated by a bold vertical line. , P ¼ 0.0018 (Mann–Whitney U test).

www.aacrjournals.org Cancer Res; 73(8) April 15, 2013 OF7

Hoshiko et al.

Figure 4. Enhanced Hgf processing and upregulation of Wnt signal–related genes. A, qRT-PCR for Spint1 and matriptase (St14) and matriptase localization þ fl fl (immunofluorescence) in small intestine. Nontumor mucosa (Non-T) and tumor (T) from small intestine of ApcMin/ /Spint1 ox/ ox (C, control) or ApcMin/þ/Spint1flox/flox/Vil-Cre (KO, Spint1 deletion) mice were analyzed. qRT-PCR data are means SD (, P < 0.0001; NS, not significant; n ¼ 3). Green fluorescence, matriptase; blue, nuclei. N, non-neoplastic epithelium; T, neoplastic epithelium. Bar, 20 mm. B, processing of proHgf in nontumor mucosa (top) and tumor tissues (bottom; 10 weeks old). Representative immunoblot photos and bar graph data of relative processing rate (n ¼ 4) are shown. , P < 0.05. Immunostaining photos of phosphorylated c-Met in intestinal neoplastic lesions are also shown. Bar, 20 mm. C, immunostaining of b-catenin (b-cat) of nontumor mucosa (top) and tumor mucosa (bottom). Arrows indicate nuclear translocation of b-catenin in epithelial cells of nontumor mucosa from Spint1-deleted intestine and enlarged image is also shown as inset. In the last panel, phosphorylated c-Met (pMet) was also immunostained using a serial section. Bar, 20 mm. D, RT-PCR of angiogenesis-related genes and Cldn2 and qRT-PCR analysis of Cldn2 expression. , P ¼ 0.0008. P values were analyzed by Mann–Whitney U test.

OF8 Cancer Res; 73(8) April 15, 2013 Cancer Research

HAI-1 Suppresses Intestinal Tumorigenesis

Table 1. Effects of Spint1 deletion on Wnt signaling–related gene expression

Up- or downregulated (>2.0-fold) genes in response to Spint1 deletion Fold change

Annotation and gene symbol Entrez Gene ID Tumor Nontumor Adenomatosis polyposis coli downregulated 1, Apcdd1 494504 6.59 4.63 b-Catenin, Ctnnb1 12387 5.08 NA LIM domain binding 1, Ldb1 16825 4.37 NA Prospero-related homeobox 1, Prox1 19130 4.16 2.93 Frizzled homolog 2, Fzd2 57265 4.09 NA Protein inhibitor of activated STAT4, Pias4 59004 3.39 NA Casein kinase 1 delta, Csnk1d 104318 3.32 NA F-box and WD-40 domain protein 2, Fbxw2 30050 3.22 NA SRY-box containing gene 17, Sox17 20671 3.21 10.28 Wnt inhibitory factor 1, Wif1 24117 3.06 3.33 Casein kinase 1 gamma 2, Csnk1g2 103236 2.98 NA Casein kinase 2 alpha 1, Csnk2a1 12995 2.87 2.11 Frizzled homolog 3, Fzd3 14365 2.80 3.66 Low-density lipoprotein receptor–related protein 6, Lrp6 16974 2.60 NA Methyl-CpG binding domain protein 2, Mbd2 17191 2.42 NA Insulin-like growth factor–binding protein 4, Igfbp4 16010 2.38 4.97 Protein phosphatase 1A, alpha isoform, Ppm1a 19042 2.36 2.01 Protein tyrosine kinase 7, Ptk7 71461 2.35 14.34 Casein kinase 1 alpha 1, Csnk1a1 93687 2.26 NA Low-density lipoprotein receptor–related protein 5, Lrp5 16973 2.09 NA Transcription factor 4, Tcf4 21413 2.06 2.11 SUMO/sentrin-speciﬁc peptidase 2, Senp2 75826 2.01 NA Frizzled homolog 6, Fzd6 14368 NA 4.28 Naked cuticle 1 homolog, Nkd1 93960 NA 4.16 Dickkopf homolog 3, Dkk3 50781 NA 3.77 Wingless-type MMTV integration site 11, Wnt11 22411 NA 3.55 Receptor tyrosine kinase–like orphan receptor 2, Ror2 26564 NA 2.98 Wingless-type MMTV integration site 5A, Wnt5A 22418 NA 2.49 Casein kinase 1 epsilon, Csnk1e 27373 NA 2.48 SRY-box containing gene 4, Sox4 20677 NA 2.46 Transcription factor 7, Tcf7 (also known as Tcf-1) 21414 NA 2.41 Microtubule-actin crosslinking factor 1, Macf1 11426 NA 2.21 Lymphoid enhancer–binding factor 1, Lef1 16842 NA 2.02 Transducin-like enhancer of split 3, Tle3 21887 NA 2.28

NOTE: Bold, gene symbol.

Tcf7 gene expression was also upregulated in Spint1-deleted Spint1 deletion amplified Wnt signaling. Taken altogether, the mucosa. In accordance with these findings, Sox17 and Apcdd1 function of intestinal Hai-1/Spint1 as a tumor suppressor genes were markedly upregulated not only in tumor tissue but retards the formation of a tumor-promoting microenviron- also in the Hai-1/Spint1–deficient nontumor mucosa. Sox17 is ment under cancer-prone epithelial conditions and does not induced by Wnt signal activation in the early stage of gastro- act as a gatekeeper in carcinogenesis. This notion is supported intestinal tumorigenesis (42), and Apcdd1 is also regulated by by the fact that enhanced tumor formation was observed only the b-catenin/Tcf complex with its elevated expression con- in its proper site in each of the mouse models examined in this þ tributing to colorectal carcinogenesis (43). As these alterations study: small intestine in ApcMin/ mice or distal large intestine in gene expression in the nontumor mucosa were not observed in the AOM/DSS carcinogenesis model. in the Spint1-deleted intestine with wild-type Apc, microade- Matriptase is one of the most important cognate proteases nomas intermingling with normal epithelium may contribute of HAI-1/SPINT1 (14–17) and is critically required to maintain to the upregulation of these genes. However, the degrees of intestinal epithelial integrity (3–5). Therefore, deregulated upregulation of some genes, such as Apcdd1 and Sox17,in matriptase activity may be involved in the phenotypes Spint1-deleted nontumor mucosa were comparable with observed here for Hai-1/Spint1 deficiency. In fact, the ratio of þ tumor tissue itself from control ApcMin/ mice, indicating that matriptase to HAI-1/SPINT1 is higher in colorectal cancers

www.aacrjournals.org Cancer Res; 73(8) April 15, 2013 OF9

Hoshiko et al.

Figure 5. Effect of Spint1 deletion on AOM/DSS-induced colon carcinogenesis. A, AOM/DSS treatment protocol. B, tumor multiplicity and size. The number of fl fl fl fl visible polyps/mouse was counted and the maximum diameter of each polyp measured. cont, Spint1 ox/ ox; KO, Spint1 ox/ ox/Vil-Cre. Bold vertical line indicates the mean value. , P ¼ 0.0002 (Mann–Whitney U test). C, representative photographs of macroscopic findings of distal colon tissues from Spint1flox/flox (Cont) and Spint1flox/flox/Vil-Cre (KO) mice. Bar, 1 cm. D, histology (H&E) and Hai-1/Spint1 immunostaining of colon tumors. Photographs in top are tumor formed in Spint1flox/flox (Cont) mouse, showing decreased Hai-1/Spint1 immunoreactivity in tumor cells. Dotted line indicates boundary between tumor (T) and non-neoplastic epithelium (N). Middle, low-magnification histology (H&E) of Spint1flox/flox/Vil-Cre (KO) mouse tumor that lacks Hai-1/Spint1 expression (inset). Photographs in the bottom indicate comparative histology of cont and KO tumors. Tumor cells in KO mouse show decreased mucin production. Bar, 50 mm.

than in normal tissue (44) and cancer cells show increased of membrane-associated HAI-1/SPINT1 resulted in decreased matriptase activity (45). However, there is an apparent paradox membrane-associated matriptase and increased pericellular for the intestinal tumorigenesis in the matriptase:Hai-1/Spint1 matriptase activity (27, 30) that might activate pericellular balance. Matriptase hypomorphic mice showed enhanced molecules involved in tumor progression such as proHGF. In susceptibility to DSS colitis and increased claudin-2 expression the current study, intestinal tumor cells with decreased mem- in intestinal epithelial cells (4). Conditional deletion of the branous Hai-1/Spint1 showed abnormal matriptase subcellu- mouse matriptase gene (St14) in the intestinal tract impairs lar localization. However, in nonneoplastic epithelial cells, the colonic mucosal integrity and induces colonic cancer devel- cell surface localization pattern of matriptase was preserved in opment (3, 46). Therefore, similar to Hai-1/Spint1, matriptase Spint1-deleted cells as long as the epithelium retained a normal is a tumor suppressor in the intestinal tract. We hypothesize histologic architecture. In this regard, Hai-2/Spint2 is also that membranous Hai-1/Spint1 may be critical for "normal" expressed in the intestinal epithelium and may compensate matriptase function that is required for epithelial integrity, and for the loss of Hai-1/Spint1 in normal epithelial cells (8). loss of Hai-1/Spint1 may cause abnormal subcellular localiza- A potential limitation of this study is the question whether tion or enhanced release of matriptase under adverse cellular total absence of Hai-1/Spint1 in the Spint1-deleted intestinal conditions. In fact, HAI-1/SPINT1 is required for intracellular tumor cells exactly represents the pathobiology of the tumor matriptase trafﬁcking (15, 16, 47), and in tumor cells, loss cells with ectodomain shedding-induced loss of cell surface

OF10 Cancer Res; 73(8) April 15, 2013 Cancer Research

HAI-1 Suppresses Intestinal Tumorigenesis

Hai-1/Spint1. Although known functional domain of HAI-1/ Disclosure of Potential Conflicts of Interest SPINT1 is extracellular N-terminal Kunitz domain and no No potential conflicts of interest were disclosed. function has been reported for C-terminal intracytoplasmic domain, remnant C-terminus of HAI-1/SPINT1 after ectodo- Authors' Contributions main shedding may acquire unexpected biologic functions. Conception and design: H. Kataoka This possibility should be clarified in a future study. Another Development of methodology: S. Hoshiko, M. Kawaguchi, T. Fukushima fi Acquisition of data (provided animals, acquired and managed patients, important issue in this study that requires further clari cation provided facilities, etc.): S. Hoshiko, M. Kawaguchi, Y. Haruyama, M. Seiki is whether a novel protumorigenic protease regulated by Hai- Analysis and interpretation of data (e.g., statistical analysis, biostatistics, 1/Spint1 may exist in intestinal epithelial cells. Deregulated computational analysis): S. Hoshiko, M. Kawaguchi, K. Yorita, H. Tanaka, H. Kataoka activity of this putative protease in the absence of Hai-1/Spint1 Writing, review, and/or revision of the manuscript: S. Hoshiko, M. Kawa- may contribute to accelerated tumorigenesis. In this regard, guchi, H. Kataoka several Hai-1/Spint1–sensitive membrane serine proteases, Administrative, technical, or material support (i.e., reporting or orga- nizing data, constructing databases): T. Fukushima, M. Seiki, H. Kataoka such as TMPRSS13, TMPRSS4, hepsin, and prostasin, are also Study supervision: M. Seiki, H. Inatsu, K. Kitamura, H. Kataoka expressed in the intestinal tract (8). Prostasin activates epithelial sodium channel and protease-activated receptor 2 (6), which may be involved in intestinal tumorigenesis (48). While Acknowledgments The authors thank Yumiko Nomura and Ai Kanemaru for preparation of the biologic function of TMPRSS4 is poorly understood, this histologic specimens. TTSP does induce an invasive phenotype in cancer cells (20, 49). In conclusion, our data suggest that the epithelial cell Grant Support surface protease–inhibitor interaction may impact mucosal This work was supported by Grant-in-Aid for Scientific Research no. 24390099 (H. Kataoka) and no. 23790250 (M. Kawaguchi) from the Ministry of Education, carcinogenesis, and the loss of membranous HAI-1/SPINT1 in Science, Sports and Culture, Japan. epithelial cells confers susceptibility to tumorigenesis in the The costs of publication of this article were defrayed in part by the payment of intestinal tract. Further analysis of the molecular mechanisms page charges. This article must therefore be hereby marked advertisement in that underlie enhanced intestinal tumor formation in the accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Spint1-deleted mouse carcinogenesis models described in this Received August 22, 2012; revised January 15, 2013; accepted February 4, 2013; study may reveal novel targets for therapeutic intervention. published OnlineFirst February 27, 2013.

References 1. Turner JR. Intestinal mucosal barrier function in health and disease. activation of HGF/scatter factor. Cancer Metastasis Rev 2003;22: Nat Rev Immunol 2009;9:799–809. 223–36. 2. Turksen K, Troy TC. Junctions gone bad: claudins and loss of the 12. Kataoka H, Suganuma T, Shimomura T, Itoh H, Kitamura N, Nabe- barrier in cancer. Biochim Biophys Acta 2011;1816:73–9. shima K, et al. Distribution of hepatocyte growth factor activator 3. List K, Kosa P, Szabo R, Bey AL, Wang CB, Molinolo A, et al. Epithelial inhibitor type 1 (HAI-1) in human tissues. Cellular surface localization integrity is maintained by a matriptase-dependent proteolytic path- of HAI-1 in simple columnar epithelium and its modulated expression in way. Am J Pathol 2009;175:1453–63. injured and regenerative tissues. J Histochem Cytochem 1999;47: 4. Buzza MS, Netzel-Arnett S, Shea-Donohue T, Zhao A, Lin CY, List K, 673–82. et al. Membrane-anchored serine protease matriptase regulates epi- 13. Kataoka H, Kawaguchi M. Hepatocyte growth factor activator (HGFA): thelial barrier formation and permeability in the intestine. Proc Natl pathophysiological functions in vivo. FEBS J 2010;277:2230–7. Acad Sci U S A 2010;107:4200–5. 14. Lin CY, Anders J, Johnson M, Dickson RB. Puriﬁcation and charac- 5. Netzel-Arnett S, Buzza MS, Shea-Donohue T, Desilets A, Leduc R, terization of a complex containing matriptase and a Kunitz-type serine Fasano A, et al. Matriptase protects against experimental colitis and protease inhibitor from human milk. J Biol Chem 1999;274:18237–42. promotes intestinal barrier recovery. Inﬂamm Bowel Dis 2012; 15. Oberst MD, Chen LY, Kiyomiya K, Williams CA, Lee MS, Johnson MD, 18:1303–14. et al. HAI-1 regulates activation and expression of matriptase, a 6. Antalis TM, Bugge TH, Wu Q. Membrane-anchored serine proteases in membrane-bound serine protease. Am J Physiol Cell Physiol 2005; health and disease. Prog Mol Biol Transl Sci 2011;99:1–50. 289:C462–70. 7. Nagaike K, Kawaguchi M, Takeda N, Fukushima T, Sawaguchi A, 16. List K, Bugge TH, Szabo R. Matriptase: potent proteolysis on the cell Kohama K, et al. Defect of hepatocyte growth factor activator inhibitor surface. Mol Med 2006;12:1–7. type 1/serine protease inhibitor, Kunitz type 1 (Hai-1/Spint1) leads to 17. Szabo R, Kosa P, List K, Bugge TH. Loss of matriptase suppression ichthyosis-like condition and abnormal hair development in mice. Am J underlies spint1 mutation-associated ichthyosis and postnatal lethal- Pathol 2008;173:1464–75. ity. Am J Pathol 2009;174:2015–22. 8. Kawaguchi M, Takeda N, Hoshiko S, Yorita K, Baba T, Sawaguchi A, 18. Hashimoto T, Kato M, Shimomura T, Kitamura N. TMPRSS13, a type II et al. Membrane-bound serine protease inhibitor HAI-1 is required for transmembrane serine protease, is inhibited by hepatocyte growth maintenance of intestinal epithelial integrity. Am J Pathol 2011;179: factor activator inhibitor type 1 and activates pro-hepatocyte growth 1815–26. factor. FEBS J 2010;277:4888–900. 9. Heinz-Erian P, Muller T, Krabichler B, Schranz M, Becker C, Ruschen- 19. Kato M, Hashimoto T, Shimomura T, Kataoka H, Ohi H, Kitamura N. dorf F, et al. Mutations in SPINT2 cause a syndromic form of congenital Hepatocyte growth factor activator inhibitor type 1 inhibits protease sodium diarrhea. Am J Hum Genet 2009;84:188–96. activity and proteolytic activation of human airway trypsin-like prote- 10. Miller GS, List K. The matriptase-prostasin proteolytic cascade in ase. J Biochem (Tokyo) 2012;151:179–87. epithelial development and pathology. Cell Tissue Res 2013;351: 20. Cheng H, Fukushima T, Takahashi N, Tanaka H, Kataoka H. Hepato- 245–53 cyte growth factor activator inhibitor type 1 regulates epithelial to 11. Kataoka H, Miyata S, Uchinokura S, Itoh H. Roles of hepatocyte growth mesenchymal transition through membrane-bound serine protei- factor (HGF) activator and HGF activator inhibitor in the pericellular nases. Cancer Res 2009;69:1828–35.

www.aacrjournals.org Cancer Res; 73(8) April 15, 2013 OF11

Hoshiko et al.

21. Tanaka H, Nagaike K, Takeda N, Itoh H, Kohama K, Fukushima T, et al. 36. Seshargiri S, Stawiski EW, Durinck S, Modrusan Z, Storm EE, Conboy Hepatocyte growth factor activator inhibitor type 1 (HAI-1) is required CB, et al. Recurrent R-spondin fusions in colon cancer. Nature 2012; for branching morphogenesis in the chorioallantoic placenta. Mol Cell 488:660–4. Biol 2005;25:5687–98. 37. Dhawan P, Ahmad R, Chaturvedi R, Smith JJ, Midha R, Mittal MK, et al. 22. Yamauchi M, Lochhead P, Morikawa T, Huttenhower C, Chan AT, Claudin-2 expression increases tumorigenicity of colon cancer cells: Giovannucci E, et al. Colorectal cancer: a tale of two sides or a role of epidermal growth factor receptor activation. Oncogene continuum? Gut 2012;61:794–7. 2011;30:3234–47. 23. Kataoka H, Hamasuna R, Itoh H, Kitamura N, Koono M. Activation of 38. Gherardi E, Birchmeier W, Birchmeier C, Vande Woude G. Targeting hepatocyte growth factor/scatter factor in colorectal carcinoma. Can- MET in cancer: rationale and progress. Nat Rev Cancer 2012;12: cer Res 2000;60:6148–59. 89–103. 24. Suzuki R, Kohno H, Sugie S, Nakagama H, Tanaka T. Strain differences 39. Vermeulen L, De Sousa EMF, van der Heijden M, Cameron K, de Jong in the susceptibility to azoxymethane and dextran sodium sulfate- JH, Borovski T, et al. Wnt activity deﬁnes colon cancer stem cells and is induced colon carcinogenesis in mice. Carcinogenesis 2006;27:162–9. regulated by the microenvironment. Nat Cell Biol 2010;12:468–76. 25. Nagaike K, Kohama K, Uchiyama S, Tanaka H, Chijiiwa K, Itoh H, et al. 40. Huang FI, Chen YL, Chang CN, Yuan RH, Jeng YM. Hepatocyte growth Paradoxically enhanced immunoreactivity of hepatocyte growth factor factor activates Wnt pathway by transcriptional activation of LEF1 to activator inhibitor type 1 (HAI-1) in cancer cells at the invasion front. facilitate tumor invasion. Carcinogenesis 2012;33:1142–8. Cancer Sci 2004;95:728–35. 41. Angus-Hill ML, Elbert KM, Hidalgo J, Capecchi MR. T-cell factor 4 26. Inoue T, Kataoka H, Goto K, Nagaike K, Igami K, Naka D, et al. functions as a tumor suppressor whose disruption modulates colon Activation of c-Met (hepatocyte growth factor receptor) in human cell proliferation and tumorigenesis. Proc Natl Acad Sci U S A gastric cancer tissue. Cancer Sci 2004;95:803–8. 2011;108:4914–9. 27. Baba T, Kawaguchi M, Fukushima T, Sato Y, Orikawa H, Yorita K, et al. 42. Du YC, Oshima H, Oguma K, Kitamura T, Itadani H, Fujimura T, et al. Loss of membrane-bound serine protease inhibitor HAI-1 induces oral Induction and down-regulation of Sox17 and its possible roles during squamous cell carcinoma cells' invasiveness. J Pathol 2012;228: the course of gastrointestinal tumorigenesis. Gastroenterology 2009; 181–92. 137:1346–57. 28. Taniwaki K, Fukamachi H, Komori K, Ohtake Y, Nonaka T, Sakamoto T, 43. Takahashi M, Fujita M, Furukawa Y, Hamamoto R, Shimokawa T, Miwa et al. Stroma-derived matrix metalloproteinase (MMP)-2 promotes N, et al. Isolation of a novel human gene, APCDD1, as a direct target of membrane type 1-MMP-dependent tumor growth in mice. Cancer the beta-Catenin/T-cell factor 4 complex with probable involvement in Res 2007;67:4311–9. colorectal carcinogenesis. Cancer Res 2002;62:5651–6. 29. Kataoka H, Shimomura T, Kawaguchi T, Hamasuna R, Itoh H, Kitamura 44. Vogel LK, Saebo M, Skjelbred CF, Abell K, Pedersen ED, Vogel U, et al. N, et al. Hepatocyte growth factor activator inhibitor type 1 is a speciﬁc The ratio of matriptase/HAI-1 mRNA is higher in colorectal cancer cell surface binding protein of hepatocyte growth factor activator adenomas and carcinomas than corresponding tissue from control (HGFA) and regulates HGFA activity in the pericellular microenviron- individuals. BMC Cancer 2006;6:176. ment. J Biol Chem 2000;275:40453–62. 45. Darragh MR, Schneider EL, Lou J, Phojanakong PJ, Farady CJ, Marks 30. Domoto T, Takino T, Guo L, Sato H. Cleavage of hepatocyte growth JD, et al. Tumor detection by imaging proteolytic activity. Cancer Res factor activator inhibitor-1 by membrane-type MMP-1 activates 2010;70:1505–12. matriptase. Cancer Sci 2012;103:448–54. 46. Kosa P, Szabo R, Molinolo AA, Bugge TH. Suppression of Tumorige- 31. Wali RK, Roy HK, Kim YL, Liu Y, Koetsier JL, Kunte DP, et al. Increased nicity-14, encoding matriptase, is a critical suppressor of colitis and microvascular blood content is an early event in colon carcinogenesis. colitis-associated colon carcinogenesis. Oncogene 2012;31:3679–95. Gut 2005;54:654–60. 47. Godiksen S, Selzer-Plon J, Pedersen ED, Abell K, Rasmussen HB, 32. Berx G, van Roy F. Involvement of members of the cadherin super- Szabo R, et al. Hepatocyte growth factor activator inhibitor-1 has a family in cancer. Cold Spring Harb Perspect Biol 2009;1:a003129. complex subcellular itinerary. Biochem J 2008;413:251–9. 33. Burkly LC, Michaelson JS, Hahm K, Jakubowski A, Zheng TS. TWEAK- 48. Hirota CL, Moreau F, Iablokov V, Dicay M, Renaux B, Hollenberg MD, ing tissue remodeling by a multifunctional cytokine: role of TWEAK/ et al. Epidermal growth factor receptor transactivation is required for Fn14 pathway in health and disease. Cytokine 2007;40:1–16. proteinase-activated receptor-2-induced COX-2 expression in intes- 34. Bernardi R, Guernah I, Jin D, Grisendi S, Alimonti A, Teruya-Feldstein J, tinal epithelial cells. Am J Physiol Gastrointest Liver Physiol 2012;303: et al. PML inhibits HIF-1alpha translation and neoangiogenesis G111–9. through repression of mTOR. Nature 2006;442:779–85. 49. Jung H, Lee KP, Park SJ, Park JH, Jang YS, Choi SY, et al. TMPRSS4 35. Kataoka H, Uchino H, Denda K, Kitamura N, Itoh H, Tsubouchi H, et al. promotes invasion, migration and metastasis of human tumor cells by Evaluation of hepatocyte growth factor activator inhibitor expression in facilitating an epithelial-mesenchymal transition. Oncogene 2008;27: normal and malignant colonic mucosa. Cancer Lett 1998;128:219–27. 2635–47.

OF12 Cancer Res; 73(8) April 15, 2013 Cancer Research

Hepatocyte Growth Factor Activator Inhibitor Type 1 Is a Suppressor of Intestinal Tumorigenesis

Shinri Hoshiko, Makiko Kawaguchi, Tsuyoshi Fukushima, et al.

Cancer Res Published OnlineFirst February 27, 2013.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-12-3337

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2013/02/27/0008-5472.CAN-12-3337.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2013/04/11/0008-5472.CAN-12-3337. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.