The Antimetastatic Role of Thrombomodulin Expression in Islet Cell–Derived Tumors and Its Diagnostic Value

Vol. 10, 6179–6188, September 15, 2004 Clinical Cancer Research 6179

Satoshi Iino,1,2 Kazuhiro Abeyama,2 ically useful molecular marker not only for identifying Ko-ichi Kawahara,2 Munekazu Yamakuchi,2 ␤-cell–origin islet cell tumors (i.e., insulinomas) but also for Teruto Hashiguchi,2 Sumika Matsukita,3 predicting disease prognosis of islet cell tumors. Suguru Yonezawa,3 Shotaro Taniguchi,2 Masanori Nakata,2 Sonshin Takao,1 INTRODUCTION Takashi Aikou,1 and Ikuro Maruyama2 Neuroendocrine neoplasms arising primarily from the pan- Departments of 1Surgical Oncology and Digestive Surgery, creatic islets of Langerhans (also termed islet cell tumors) are 2Laboratory and Vascular Medicine, and 3Human Pathology, quite rare, with an overall incidence of only 1 to 1.5 per 100,000 Kagoshima University Graduate School of Medicine and Dental in the general population (1, 2). Up to 50% of all of the islet cell Science, Kagoshima, Japan tumors secrete biologically active peptides, causing their sys- temic clinical symptoms; thus, such tumors are categorized as ABSTRACT “functional” islet cell tumors (i.e., insulinoma, glucagonoma, gastrinoma, somatostatinoma and VIPoma). By contrast, other Islet cell tumors, endocrine neoplasm arising from pan- islet cell tumors showing no substantial secretion of the peptide creatic islets of Langerhans, are histologically difficult to hormones are categorized as “nonfunctional” islet cell tumor. diagnose as benign or malignant. Molecular markers are Clinically, it is very difficult to diagnose islet cell tumors as associated with the clinical characteristics that most of in- “benign or malignant” by using histologic examination alone. sulinoma are usually benign tumors, whereas other islet cell To determine their malignant potentials, clinical features of islet tumors are malignant but have not been identified. In this cell tumors (i.e., lymph node metastasis or distant metastasis) context, we newly found that an endothelial anticoagulant must also be taken into consideration. According to the past thrombomodulin was expressed in the normal islet ␤ cells studies concerning the metastatic capacities of islet cell tumors, and insulinoma, but not of other islet components or non- more than 90% of insulinoma cases were clinically benign, insulinoma islet cell tumors. Clinically, all of the subjects whereas 40 to 60% of other functional islet cell tumors (1–4) of the insulinoma group showed no metastasis (15 ؍ n) and 80 to 90% of nonfunctional islet cell tumors (2, 5) were together with thrombomodulin expression in the lesions, malignant. Thus, most insulinoma cases are usually considered whereas the other islet cell tumor groups showed a high to be benign tumors in the clinical situations. Nevertheless, incidence of metastasis (82%) and a low expression rate of molecular markers clinically associated with the metastatic po- thrombomodulin (6%). To examine the functional role of tentials of islet cell tumors have not yet been identified. thrombomodulin, especially regarding the clinical charac- Thrombomodulin has been identified as an endothelial teristics of islet cell tumors, we tested the effect of exogenous membrane protein (6), and the thrombomodulin-protein C path- thrombomodulin overexpression on cell adhesiveness and way has become well recognized for its essential anticoagulant/ proliferation using MIN6 insulinoma cell line. In cell-based antithrombotic properties. The association of thrombin and experiments, thrombomodulin overexpression reduced cell -2؉ thrombomodulin on the endothelial surface blocks the proco proliferation and enhanced Ca -independent cell aggrega- agulant properties of thrombin and redirects substrate specificity tion, possibly through direct interaction with neural cell toward the activation of plasma protein C (6–12) as well as adhesion molecule. Taken together, these results are sug- toward the inhibitor of fibrinolysis, thrombin-activatable fibrin- gesting that thrombomodulin may act as antimetastatic mol- olysis inhibitor (13, 14). Activated protein C exerts additional ecule of insulinomas. In addition, thrombomodulin is a clin- anticoagulant effects by inactivating coagulant cofactors Va and VIIIa (7, 8, 12). Thrombomodulin is also expressed at extra- vascular sites, such as in syncytiotrophoblasts in the placenta, in Received 12/16/03; revised 5/3/04; accepted 5/18/04. the epithelial tissues of gingiva, in the skin, lungs and digestive Grant support: Ministry of Education and Science of Japanese Gov- organs, and in the synovial lining cells (15–19). However, the ernment, Grant-in-Aid 13470324 and 14657627 (I. Maruyama) and functional role of thrombomodulin in the extra-vascular space, 13220016 (S. Yonezawa). apart from the possible limitation of thrombin generation at The costs of publication of this article were defrayed in part by the these sites, remains uncertain. Recent studies have shown that payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to thrombomodulin also attenuates inflammatory responses (20– indicate this fact. 22) and acts as an antimetastatic molecule against malignant Requests for reprints: Ikuro Maruyama, Department of Laboratory and tumor progression (23–27) beyond its anticoagulant activities. Vascular Medicine, Kagoshima University Graduate School of Medi- In the context of the antimetastatic properties of thrombo- cine and Dental Science, 8-35-1 Sakuragaoka, Kagoshima, 890-8520, Japan. Phone: 81-99-275-5437; Fax: 81-99-275-2629; E-mail: rinken@ modulin against tumor progression, we considered a possible m3.kufm.kagoshima-u.ac.jp. link between thrombomodulin and the benign character of in- ©2004 American Association for Cancer Research. sulinomas. In the present study, we first show the evidence that

thrombomodulin is expressed dominantly in cases of insulinoma viously (16), and antimouse thrombomodulin rat IgG FM/TM as well as in normal ␤ cells in the pancreatic islets, which antibody was generously provided by Dr. Sumi Imada (Meiji negatively correlates with the clinical incidence of tumor me- Institute of Health Science, Odawara, Japan). tastasis. Furthermore, the results shown herein also demonstrate Cells and Cell Cultures. Dr. Susumu Seino (Chiba for the first time that thrombomodulin functions as a regulator University, Chiba, Japan) kindly donated the islet ␤ cell- for cell adhesion and proliferation in vitro. derived MIN6 line (29, 30). Cells were maintained in DMEM (high glucose; Life Technologies, Inc., Grand Island, NY) MATERIALS AND METHODS supplemented with 15% heat-inactivated fetal bovine serum, ␮ Tissue Samples. Surgical specimens were obtained from 100 units/ml penicillin, 100 g/ml streptomycin sulfate, 31 patients (14 men and 17 women) with islet cell tumor, who and 0.5% 2-mercaptoethanol at 37°C (5% CO2). To ob- underwent tumor resection in the Department of Surgical On- tain thrombomodulin-overexpressing MIN6, we produced cology and Digestive Surgery (Kagoshima University Graduate pTracer-CMV/Bsd Vector (Invitrogen Corp., Carlsbad, CA) School of Medicine and Dental Science) between 1977 and containing cDNA of the entire precursor of thrombomodulin 2003. The age of the patients ranged from 18 to 79 years, with (gift from Asahi Chemical Industry Co., Ltd.). MIN6 was an average age of 42 years. Type of tumor were as follows: (a) transfected with the vector construct using LipofectAMINE 15 patients were diagnosed with insulinoma; (b) 10 patients Reagent (Invitrogen). were diagnosed with nonfunctional islet cell tumor; (c) 1 patient Islet Isolation. Male ICR mice were housed in the was diagnosed with glucagonoma; (c) 2 patients were diagnosed pathogen-free facility of the Animal Resource Center, Ka- with gastrinoma; (d) 2 patients were diagnosed with soma- goshima University. All experiments involving animals were tostatinoma; and (e) 1 patient was diagnosed with VIPoma. Of conducted following the guidelines of the NIH and with the the 31 patients, 7 showed liver metastasis and 6 had lymph node approval of the Institutional Animal Care and Use Committee. metastasis (Table 1). The islets were isolated from 8-week-old male ICR mice as Immunohistochemistry. Immunohistochemical staining described previously (31). was performed using an ABC kit (Vector Laboratories, Burl- FACS Analysis. Briefly, cell suspension was incubated ingame, CA), as described previously (17, 28). Briefly, paraffin- with rabbit antiserum against thrombomodulin (1:50) for 30 embedded tissue samples were reacted with each antibody, as minutes. The samples were subsequently washed and reacted indicated in the presence of 1% BSA for 12 hours at 4°C. with fluorescence-conjugated antirabbit IgG antibody for 30 Biotinylated IgG was used as a second antibody. minutes. Fluorescent measurements were then performed using Reagents and Antibodies. Recombinant human soluble FACS (EPICS Profile, Beckman Coulter Inc., Fullerton, CA). thrombomodulin spanning the extracellular domain of the pro- Western Blotting. Western blotting analysis was per- tein was generously provided by Asahi Chemical Industry Co., formed as described previously (32). Briefly, lysates were sep- Ltd. (Tokyo, Japan). Recombinant neural cell adhesion mole- arated by 8% SDS-PAGE, and the separated proteins were cule-IgG-Fc fusion protein and its control IgG were purchased transferred to a nitrocellulose membrane. The membranes were from R&D Systems Inc. (Minneapolis, MN). We also purchased incubated for 1 hour in blocking solution (1% BSA and 5% antiglucagon rabbit IgG (Novo, Newcastle, United Kingdom), skimmed milk in 25 mmol/L Tris-HCl buffer saline containing antisomatostatin rabbit IgG (Biomeda Corporation, Hayward, 0.02% Tween 20) followed by reaction with each first antibody CA), antiinsulin-guinea pig IgG (DakoCytomation Inc., Carpin- (1:500) for 1 hour. Horseradish peroxidase-conjugated IgG was teria, CA), antihuman-neural cell adhesion molecule mouse IgG used as a second antibody. The chemiluminescence of horse- (Zymed Laboratories, Inc., San Francisco, CA), antimouse neu- radish peroxidase was detected using an ECL system (Amer- ral cell adhesion molecule rat IgG (Chemicon International, Inc., sham Bioscience, Inc., Buckinghamshire, United Kingdom). Temecula, CA), rabbit anti-pan–cadherin whole antiserum Reverse Transcription-PCR. The cDNA converted (Sigma Chemical Co., St. Louis, MO), antihuman-E-cadherin from total RNA was synthesized using the cDNA Synthesis mouse IgG (TaKaRa Holdings Inc., Shiga, Japan), and 3-(4,5- System (Life Technologies, Inc.), and reverse transcription-PCR dimethylthiazole-2-yl-2, 5-diphenyltetrazolium bromide (MTT; was performed using the TaKaRa Ex Taq system. ϫ Dojindo Laboratory, Kumamoto, Japan). Rabbit antiserum Cell Proliferation Assay. MIN6 cells at a density of 1 3 against human-thrombomodulin was obtained as described pre- 10 cells/ml were suspended in culture medium and seeded in a 24-well culture plate. Cell viability was measured by MTT assay at 24 hours, 48 hours, and 72 hours after seeding. Cell Adhesion Assay. MIN6 cells at a density of 5 ϫ 105 cells/ml were suspended in HEPES buffer [5.4 mmol/L KCl, 136.9 Table 1 ICT cases used in the present study mmol/L NaCl, 0.34 mmol/L Na2HPO4.12H2O, 5.6 mmol/L glu- Case Meta cose, 9.7 mmol/L (pH 7.4) and 1 mmol/L CaCl2] and seeded in Insulinoma 15 0/15 type IV collagen coated, type I collagen coated, or plastic surface ء NFICT 10 9/10 24-well culture plates as indicated. Adherent cells (%) were meas- Glucagonoma 1 0/1 Gastrinoma 2 2/2 ured at 1 hour after seeding using a hemocytometer. Somatostatinoma 2 2/2 Cell Aggregation Assay. MIN6 cells at a density of 5 ϫ VIPoma 1 0/1 105 cells/ml were suspended in HEPES buffer [5.4 mmol/L KCl, Total patients 31 13/31 136.9 mmol/L NaCl, 0.34 mmol/L Na2HPO4.12H2O, 5.6 NFICT, nonfunctional islet cell tumor. mmol/L glucose, and 9.7 mmol/L (pH 7.4)] in the presence or ء

Fig. 1 Immunohistochemical patterns of thrombomodulin expression in normal islets. A, H&E; B, mouse control IgG; C, antiglucagon IgG; D, antiinsulin IgG; E, antisomatostatin IgG; F, antithrombomodulin IgG; G, anti-pan– cadherin IgG; and H, antineural cell adhesion molecule IgG. Note that thrombomodulin-positive cells and insulin-positive ␤ cells are located at the central part of the islet (D and F), whereas glucagon-positive ␣ cells and somatostatin-positive ␦ cells are located at the peripheral part of the islet. The expressions of pan-cadherin and neural cell adhesion molecule do not differ between ␤ and non-␤ cells.

absence of 1 mmol/L CaCl2. Subsequently, cells (1 ml of cells was calculated to examine the levels of cell aggregation suspension) were incubated in a 24-well culture dish at 37°C [i.e., cell aggregation (%) ϭ (total cell number Ϫ single cells)/ and shaken (80 rpm/minute) on a rotary shaker (EYELA Multi (total number of cells) ϫ 100 (%)]. Shaker MMS, Tokyo Rikakikai Co., Tokyo, Japan). Aliquots of Assays to Assess the Protein-Protein Interactions. To samples were taken at 0 minutes, 5 minutes, 10 minutes, and 15 examine the protein-protein interaction between neural cell adhe- minutes for a cell count assay, and the total number of single sion molecule and thrombomodulin, we used immunoprecipitation

Fig. 2 Thrombomodulin expression in islet cell tumors. A, immunohistochemical patterns of thrombomodulin in islet cell tumors. a and b, distribution of insulin (left) and thrombomodulin (right) in insulinoma. a, highly insulin-producible insulinoma. b, par- tially insulin-producible insulinoma. Note that whole tumor cells are thrombomodulin positive. c and d, nonfunctional islet cell tumors. c, thrombomodulin-positive nonfunctional islet cell tumor showing a par- tially thrombomodulin-positive area in the lesion (described in Table 2 as Case 16). The primary lesion was positive for thrombomodulin and the metastatic lesion was negative. d, thrombomodulin-negative nonfunctional islet cell tumor. Capillary vessels were positive for thrombomodulin expression, which served as an internal positive control for thrombomodulin staining. B, Western blotting analyses for protein levels of thrombomodulin (top panel) and ␤-actin (as an internal control; bottom panel). Human umbilical vascular endothelial cells were used as a positive control for thrombomodulin expression. C, a case of invasive insulinoma (Table 2, Case 15). (NFICT, nonfunctional islet cell tumor; HUVEC, Human umbilical vascular endothelial cell; TM, thrombomodulin)

method and pull-down assay. In cell-free pull-down assay, protein per pellet each) were lysed in immunoprecipitation buffer and G beads conjugated with neural cell adhesion molecule-immuno- centrifuged (10,000 rpm for 10 minutes at 4°C). Then the super- globulin [i.e., the pellet obtained after the coincubation of 50 ng natant samples were used for immunoprecipitation of neural cell neural cell adhesion molecule-immunoglobulin protein (R&D) and adhesion molecule protein expressed on MIN6. In the immunopre- 100 ␮L protein G agarose beads (Bio-Rad, Tokyo, Japan)] were cipitation procedure, the samples were incubated with protein G coincubated with 10 ng/100 ␮L recombinant human soluble- beads and antihuman-neural cell adhesion molecule mouse IgG for thrombomodulin in immunoprecipitation buffer [1% NP40, 0.2 1 hour. After washing the beads, the levels of bound thrombomodu- mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L DTT, and 25 lin protein (through the binding to neural cell adhesion molecule- mmol/L Tris-HCl (pH 7.5)] for 1 hour, and then the beads were antibody complex) were assayed by the Western blotting method. washed three times with immunoprecipitation buffer. The amount of bound thrombomodulin protein to the beaded neural cell adhesion molecule-immunoglobulin was assayed by the Western blot- RESULTS ting. To determine whether thrombomodulin could interact with Thrombomodulin Expression by Islet ␤-Cell and ␤-Cell neural cell adhesion molecule on the surface of MIN6, thrombo- Tumors. Islets of Langerhans are composed primarily of en- modulin-overexpressing or mock-transfected MIN6 cells (106 cells docrine ␣ cells, ␤ cells, ␦ cells, and pancreatic polypeptide cells,

Table 2 Characteristic profile of ICTs Case Tumor type TM (primary) Meta TM (meta) E-cadherin N-CAM 1 Insulinoma ϩϪ Ϫϩ 2 Insulinoma ϩϪ ϩϪ 3 Insulinoma ϩϪ ϩϪ 4 Insulinoma ϩϪ ϩϪ 5 Insulinoma ϩϪ ϩϩ 6 Insulinoma ϩϪ ϩϩ 7 Insulinoma ϩϪ ϩϩ 8 Insulinoma ϩϪ ϩϩ 9 Insulinoma ϩϪ ϩϩ 10 Insulinoma ϩϪ ϩϩ 11 Insulinoma ϩϪ ϩϩ 12 Insulinoma ϩϪ ϩϩ 13 Insulinoma ϩϪ ϩϪ 14 Insulinoma ϩϪ ϩϩ 15 Insulinoma (invasive type) ϩϪ ϩϪ 16 NFICT ϩ Liver Ϫϩϩ 17 NFICT ϪϪ ϪϪ 18 NFICT Ϫ Liver Ϫϩϩ 19 NFICT Ϫ Liver ϪϩϪ 20 NFICT Ϫ LN ϪϪϩ 21 NFICT Ϫ LN ϪϪϩ 22 NFICT Ϫ LN ϪϪϪ 23 NFICT Ϫ LN ϪϩϪ 24 NFICT Ϫ LN ϪϩϪ 25 NFICT Ϫ LN ϪϪϪ 26 Glucagonoma ϪϪ ϩϪ 27 Gastrinoma Ϫ Liver ϪϩϪ 28 Gastrinoma Ϫ Liver ϪϩϪ 29 Somatostatinoma Ϫ Liver Ϫϩϩ 30 Somatostatinoma Ϫ Liver Ϫϩϩ 31 VIPoma ϪϪ ϪϪ Abbreviations: TM, thrombomodulin; N-CAM, neural cell adhesion molecule; NFICT, nonfunctional islet cell tumor; LN, lymph node.

secreting glucagon, insulin, somatostatin, and pancreatic even in a case of insulinoma showing partial insulin production, polypeptide, respectively. In immunohistochemical analysis of thrombomodulin was strongly expressed in whole tumor lesion pancreas serial sections as shown in Fig. 1, insulin-positive ␤ (Fig. 2, A-b). Thus, we conclude that thrombomodulin could be cells, which appeared as major components of islets, were a useful molecular marker for identifying ␤ cell-derived tumors. located at the central part of the islet, surrounded by non-␤ cells Correlation between Thrombomodulin Expression and (i.e., glucagon-positive ␣ cells and somatostatin-positive ␦ the Metastatic Capacity of Islet Cell Tumors. Recently, cells). Compared with endocrine markers, thrombomodulin ex- thrombomodulin has been shown to act as an antimetastatic pression was dominantly localized on insulin-positive ␤ cells molecule against malignant tumor progression (23–27). Further- (Fig. 1, D and F), whereas the expressions of cadherin and more, most insulinoma cases but not other islet cell tumors neural cell adhesion molecule, currently considered as islet (noninsulinoma cases) are clinically considered to be benign adhesion molecules (33, 34), did not differ between ␤ and non-␤ tumors. Therefore, we considered a possible link between the cells (Fig. 1, G and H). These results suggested that thrombo- thrombomodulin expression pattern and the clinically benign modulin expression might be a good identifier of ␤ cells and character of ␤-cell tumors (insulinoma). As shown in Table 2, also led us to postulate that thrombomodulin might be an ap- any insulinoma cases examined in this study had no signs of plicable marker for the histologic diagnosis of insulinoma. To tumor metastasis. In contrast, 13 of the 16 noninsulinoma cases confirm this, we conducted immunohistochemical examination (81%) had distant tumor metastatic lesions. Regarding the cor- of 31 islet cell tumors [15 insulinoma and 16 noninsulinoma (1 relation between thrombomodulin expression and the incidence glucagonoma, 2 gastrinoma, 2 somatostatinoma, 1 VIPoma, and of distant metastasis, the pathology of thrombomodulin-positive 10 nonfunctional islet cell tumors)]. Expectedly, all of the islet cell tumors (15 insulinoma and 1 nonfunctional islet cell insulinoma cases (n ϭ 15) were positive for thrombomodulin tumor) exhibited a low incidence of metastasis (7%), whereas staining. In contrast, 15 of the 16 noninsulinoma cases (94%) that of thrombomodulin-negative islet cell tumors (9 nonfunc- were thrombomodulin negative. This immunohistochemical ev- tional islet cell tumor, 1 glucagonoma, 2 gastrinoma, 2 soma- idence was additionally confirmed by Western blot analysis tostatinoma, and 1 VIPoma) showed a high incidence of metas- (representative data are shown in Fig. 2B). On the basis of these tasis (80%). Furthermore, despite malignant morphology, a case results, the sensitivity, specificity, and accuracy of thrombo- of thrombomodulin-positive invasive insulinoma (shown as modulin staining for the diagnosis of ␤-cell tumors (i.e., insu- Case 15 in Table 2; Fig. 2C) had no substantial signs of linoma) were 100%, 93%, and 96%, respectively. Furthermore, metastasis. Thus, these results strongly implied that thrombo-

Fig. 3 Characteristics of thrombomodulin-negative and thrombomodulin-positive islet cell tumor models. A, reverse transcription-PCR analysis for thrombomodulin mRNA expression by MIN6. As a positive control for thrombomodulin expression, we used a mouse hemangioma cell line. The amplified PCR products (as indicated) were loaded on 2% agarose gel electrophoresis. Data shown are the intensity of expected bands (at 500 bp) as mRNA levels of thrombomodulin (Lanes 1–4) and ␤-actin (as an internal control; Lanes 5 and 6). To eliminate the issue of contamination by genome DNA, PCR products from RNA samples (i.e., without reverse transcriptase; RTϪ) were also examined (Lanes 3 and 4). The PCR products were cloned into T-vector and then sequenced to confirm the specificity of PCR procedure (data not shown). B, Western blotting analysis. The protein levels of thrombomodulin expression (top panel) and internal control ␤-actin (bottom panel) were examined in MIN6 cells and mouse islet. As a

positive control for thrombomodulin expression, we used mouse hemangioma cells. Although the Mr 75,000 bands (of the nonreduced form of thrombomodulin) were detected in all of the protein samples (i.e., obtained from hemangioma, MIN6 cells, and normal mouse islets) tested, the level of thrombomodulin expression by MIN6 was much lower than that of mouse islet tissue (top panel). C and D, characteristics of thrombomodulin- transfectant MIN6. Cells were transiently transfected with thrombomodulin-expression vector or control vector then examined for thrombomodulin expression and proliferation rate. C, FACS analysis of surface thrombomodulin expression. Cells transfected with thrombomodulin-expression vector (thrombomodulinϩMIN6) had a substantially increased surface thrombomodulin expression compared with the cells transfected with control vector (MockϩMIN6), which was also confirmed by Western blotting (data not shown). D, cell proliferation assay. MTT assay was performed as described in Materials and Methods. Bars, ϮSD. (TM, thrombomodulin)

modulin expression by islet cell tumors is negatively correlated normal mouse pancreatic islets (Fig. 3B), leading to consider with the clinical incidence of tumor metastasis, at least in the that MIN6 was a suitable model of low-thrombomodulin ex- cases of insulinoma. Although a case of noninsulinoma non- pressor (i.e., almost close to thrombomodulin-negative islet cell functional islet cell tumor (Case 16 in Table 2) was complicated tumor cells, characteristically) to test the effects of thrombo- with thrombomodulin-negative metastatic lesion in liver, its modulin overexpression. To obtain thrombomodulin-overex- immunohistochemical feature showed a focal thrombomodulin pressing MIN6 (thrombomodulin-MIN6) as a thrombomodulin- expression pattern in the primary lesion (Fig. 2, A-c). positive islet cell tumor model, MIN6 was transiently Impact of Thrombomodulin-Overexpression on Tumor transfected with thrombomodulin-expression vector construct Cell Activities Concerning Metastatic Capacity. To under- (Fig. 3, B and C). Thrombomodulin-MIN6 had a less prolifer- stand the mechanism of insulinoma-specific characteristics (i.e., ative character compared with control mock-transfectant (Fig. lower metastatic capacity), we examined the functional role of 3D), showing increased cell aggregation, morphologically (data thrombomodulin expression by insulinoma cells. For this pur- not shown). Thus, we speculated that thrombomodulin might act pose, we used a mouse immortalized ␤-cell line, MIN6, as an in as an antiproliferative regulator through modulating the func- vitro model of ␤-cell islet cell tumors. As shown in Fig. 3, A and tions of certain islet adhesion molecules, such as neural cell B, analyses of reverse transcription-PCR and Western blotting adhesion molecule, E-cadherin, and N-cadherin (33, 34). In- revealed that MIN6 expressed thrombomodulin at both mRNA deed, MIN6 expressed both cadherin and neural cell adhesion and protein levels. However, the protein level of thrombomodu- molecule at a protein level (Fig. 4A), which was not affected by lin expression by MIN6 was substantially less than that in thrombomodulin overexpression (data not shown). In this con-

Fig. 4 Impacts of thrombomodulin expression on the functions of islet cell adhesion molecules. A, endogenous expressions of cadherin and neural cell adhesion molecule by MIN6 cells. MIN6 cells were analyzed for the expression levels of cadherin and neural cell adhesion molecule by Western blotting using anti-pan–cadherin antibody or antimouse-neural cell adhesion molecule antibody. Mouse brain tissue was used as a positive control for the expressions of cadherin and neural cell adhesion molecule. Note that MIN6 showed only a Mr 140,000 isoform of neural cell adhesion molecule, neither showing the expressions of Mr 120,000 or Mr 180,000 iso- forms. B, cell aggregation assay (i.e., assay to assess “cell to cell” adhesion). To assess the cell-to-cell adhesion, transfectant cells were incubated as described in Materi- als and Methods for the indicated time in the presence (left) or absence (right) of 1 mmol/L Ca2ϩ. C, cell adhesion assay (i.e., assay to assess “cell to matrix” adhesion). Note that thrombomodulin overexpression substantially increased firm cell adhesion to basement membrane matrix type IV collagen but not to type I collagen or to the plastic surface in the presence of Ca2ϩ. D, ␤1-integrin expression level was not different between thrombomodulinϩMIN6 and MockϩMIN6. Bars, ϮSD. (N-CAM, neural cell adhesion molecule; TM, thrombomodulin)

text, neural cell adhesion molecule, an immunoglobulin-like neural cell adhesion molecule. A cell-free pull-down assay of adhesion molecule acting independently of Ca2ϩ, can increase neural cell adhesion molecule using recombinant proteins (i.e., ␤1-integrin activities (35), whereas cadherin is a Ca2ϩ- recombinant thrombomodulin and neural cell adhesion mole- dependent adhesion molecule. As shown in Fig. 4B, although no cule-immunoglobulin proteins) revealed a molecular interaction effects in the Ca2ϩ-present condition, thrombomodulin overex- between thrombomodulin and neural cell adhesion molecule pression substantially increased cell aggregation by MIN6 in the (Fig. 5A). The molecular interaction was competitively inhibited 2ϩ absence of Ca (Fig. 4B). Furthermore, thrombomodulin over- by the addition of excessive recombinant protein spanning NH2- expression also up-regulated the ability of firm adhesion by terminal lectin-like domain (Fig. 5A) but not COOH-terminal MIN6 to basement membrane matrix type IV collagen but not to epidermal growth factor-like domain of thrombomodulin pro- type I collagen or to the plastic surface without affecting the tein (data not shown). These additionally implied that thrombo- protein level of cellular ␤1-integrin (Fig. 4, C and D), suggest- modulin could interact with neural cell adhesion molecule ing that thrombomodulin might increase the function of ␤1- through the site of its lectin-like domain. Furthermore, the integrin in MIN6. Thus, these led us to hypothesize that throm- interaction between thrombomodulin and neural cell adhesion bomodulin might regulate neural cell adhesion molecule molecule was also evidenced physiologically on the cell surface function in the mechanism of thrombomodulin-mediated cellu- of thrombomodulin-overexpressing MIN6 by the immunopre- lar events. cipitate procedure of endogenous neural cell adhesion molecule Physiologic Interaction between Thrombomodulin and protein (Fig. 5B), suggesting that thrombomodulin might func- Neural Cell Adhesion Molecule. To validate the hypothesis, tion as a counter-ligand or an associate molecule of neural cell we examined whether thrombomodulin could directly bind to adhesion molecule to regulate the function of cell adhesion and

affected disease prognosis (26, 27). Additionally, all of the metastatic lesions shown in some cases of thrombomodulin- positive esophageal cancer (i.e., pathologically diagnosed as squamous cell carcinomas) were observed as thrombomodulin- negative or less thrombomodulin-expressed lesions than the primary lesions (26). As with squamous cell carcinomas, the study using a melanoma line also showed a less invasive and proliferative characteristics of high-thrombomodulin expressors (23, 24). These evidences also support our conclusions that thrombomodulin-positive islet cell tumors are likely to be clinically benign tumors in character and hypothesis concerning the diagnostic value of thrombomodulin expression in islet cell tumors as a clinical predictor of disease prognosis. Interestingly, thrombomodulin rather than neural cell adhesion molecule, currently recognized as an antimetastatic adhesion molecule in ␤-cell tumor (36), seemed to negatively correlate with the clinical incidence of islet cell tumor metastasis. For example, in our study, thrombomodulin was almost specifically expressed on only ␤-cell tumors (i.e., insulinoma), whereas neural cell adhesion molecule was expressed not only on ␤-cell tumors, but also on other types of islet cell tumors. Fig. 5 Molecular interaction between thrombomodulin and neural cell With respect to the mechanism of antimetastasis, thrombomodu- adhesion molecule. A, cell-free pull-down assay. Recombinant thrombomodulin protein and neural cell adhesion molecule-immunoglobulin- lin may regulate the function of neural cell adhesion molecule conjugated protein G were mixed and coincubated, as described in through the molecular interaction at the site of its lectin-like Materials and Methods. The amounts of bound thrombomodulin (top domain. In this context, the NH2-terminal lectin-like domain of panel, Lanes 2–5) and neural cell adhesion molecule-immunoglobulin thrombomodulin is structurally similar to human phagocytic (bottom panel, Lanes 2–4) to the beads were assayed by Western blotting. Thrombomodulin protein was coprecipitated with neural cell C1q receptor, which acts as an adhesion molecule for innate adhesion molecule-conjugated (Lane 3) but not IgG-conjugated beads immune host defense. The evidence is also suggesting that the (Lane 5). The binding of thrombomodulin to neural cell adhesion lectin-like domain of thrombomodulin may have an important molecule-conjugated beads was competitively inhibited in the presence role for cell-cell interaction (37, 38). Thus, we strongly put forth of excess NH -terminal lectin-like domain of thrombomodulin protein 2 the hypothesis that thrombomodulin may be an essential anti- (Lane 4). B, Interaction between overexpressed thrombomodulin and ␤ endogenous neural cell adhesion molecule physiologically on the sur- metastatic factor in -cell tumors, acting as a counter-ligand or face of MIN6. Using thrombomodulin-transfectant MIN6, the immuno- associate molecule for neural cell adhesion molecule. Regarding precipitation procedure as described in “Materials and Methods” was the more physiologic role, thrombomodulin may have a role for used to detect the interacted thrombomodulin with neural cell adhesion organizing islet structure, such as a central core of ␤-cell as- molecule on the cell surface. Thrombomodulin protein expressed by thrombomodulin-transfectant was coprecipitated by immunoprecipita- sembly surrounded by the three other endocrine cell types (i.e., tion of endogenous neural cell adhesion molecule (Lane 2). (rhs, re- classified as non-␤ cells) in the procedure of islet development combinant human soluble; TM, thrombomodulin; N-CAM, neural cell (39). In this context, it has been suggested that only differences adhesion molecule; LD, lectin-like domain; IP, immunoprecipitation) of intensity of cell adhesion molecules direct the sorting-out of intermixed embryonic cells and the spreading of the less cohesive cell population over the surface of cohesive cells (40). Neural cell adhesion molecule has been reported to play an proliferation, which may play an important role in the benign important role in islet construction (33), despite its expression ␤ and less-metastatic character of -cell tumors (i.e., insulinoma). pattern (i.e., also expressed by non-␤ cells as well as by ␤ cells). Therefore, regulation of neural cell adhesion molecule function DISCUSSION by thrombomodulin protein may contribute to the development Our study on the diagnostic value of thrombomodulin in of islets. the identification of ␤-cell tumors strongly indicated that throm- Likewise, an anticoagulant thrombomodulin, a coagulant bomodulin might be a clinically useful marker to define the tissue factor, was also expressed on islets, which might trigger origin of islet cell tumors. In addition, the present study also the activation of an extrinsic blood coagulation cascade by the demonstrated the clinical relationship between thrombomodulin binding between factor VII and tissue factor (41). The tissue expression and a low metastatic potential of insulinoma, the factor-mediated activation of the coagulation system leads to mechanism of which might be explained by the thrombomodu- thrombin generation and results in the triggering of both coag- lin-mediated various cellular events, including the induction of ulation and inflammatory response (42). In additional to its aggregation and the reduction of proliferation. Consistent with physiologic role, thrombomodulin as an anticoagulant molecule these conclusions, our previous studies have revealed that on normal islets might play a protective role for inhibiting thrombomodulin-negative squamous cell carcinomas showed a coagulation cascade (43), maintaining islet homeostasis through substantially higher rate of metastasis than thrombomodulin- the microcirculation of islets. positive squamous cell carcinomas, which also substantially In conclusion, the present study has demonstrated the first

evidence of thrombomodulin expression by ␤ cells and ␤-cell 19. Matsuyama T, Izumi Y, Shibatate K, et al. Expression and activity tumors, suggesting its potential clinical diagnostic value. Our of thrombomodulin in human gingival epithelium: in vivo and in vitro data also introduce a new antimetastatic mechanism through studies. J Periodontal Res 2000;35:146–57. thrombomodulin-mediated neural cell adhesion molecule acti- 20. Uchiba M, Okajima K, Murakami K, et al. Recombinant thrombomodulin prevents endotoxin-induced lung injury in rats by inhibiting vation. leukocyte activation. Am J Physiol 1996;271(3 Pt 1):L470–5. 21. Taoka Y, Okajima K, Uchiba M, Johno M. Neuroprotection by recombinant thrombomodulin. Thromb Haemostasis 2000;83:462–8. ACKNOWLEDGMENTS 22. Conway EM, Van de Wouwer M, Pollefeyt S, et al. The lectin-like The authors would like to thank Drs. Yoshito Ogura and Shigeho domain of thrombomodulin confers protection from neutrophil-medi- Maenohara for generously providing samples. ated tissue damage by suppressing adhesion molecule expression via nuclear factor kappaB and mitogen-activated protein kinase pathways. J Exp Med 2002;196:565–77. REFERENCES 23. Zhang Y, Weiler-Guettler H, Chen J, et al. Thrombomodulin mod- 1. Delcore R, Friesen SR. Gastrointestinal neuroendocrine tumors. ulates growth of tumor cells independent of its anti-coagulant activity. J Am Coll Surg 1994;178:187–211. J Clin Investig 1998;101:1301–9. 2. Thompson GB, van Heerden JA, Grant CS, Carney JA, Ilstrup DM. 24. Matsushita Y, Yoshiie K, Imamura Y, et al. A subcloned human Islet cell carcinomas of the pancreas: a twenty-year experience. Surgery esophageal squamous cell carcinoma cell line with low thrombomodulin (St Louis) 1988;104:1011–7. expression showed increased invasiveness compared with a high throm- 3. Perry RR, Vinik AI. Clinical review 72: diagnosis and management bomodulin-expressing clone-thrombomodulin as a possible candidate of functioning islet cell tumors. J Clin Endocrinol Metab 1995;80: for an adhesion molecule of squamou cell carcinoma. Cancer Lett 2273–8. 1998;127:195–201. 4. Norton JA. Neuroendocrine tumors of the pancreas and duodenum. 25. Ogawa H, Yonezawa S, Maruyama I, et al. Expression of throm- Curr Probl Surg 1994;31:77–156. bomodulin in squamous cell carcinoma of the lung: its relationship to lymph node metastasis and prognosis of the patients. Cancer Lett 5. Azimuddin K, Chamberlain RS. The surgical management of pan- 2000;149:95–103. creatic neuroendocrine tumors. Surg Clin North Am 2001;81:511–25. 26. Tezuka Y, Yonezawa S, Maruyama I, et al. Expression of throm- 6. Esmon CT, Owen GW. Identification of an endothelial cell cofactor bomodulin in esophageal squamous cell carcinoma and its relationship for thrombin-catalyzed activation of protein C. Proc Natl Acad Sci USA to lymph node metastasis. Cancer Res 1995;55:4196–200. 1981;78:2249–52. 27. Hamatake M, Ishida T, Mitsudomi T, Akazawa K, Sugimachi K. 7. Esmon CT, Esmon NL, Harris KW. Complex formation between Prognostic value and clinicopathological correlation of thrombomodulin thrombin and thrombomodulin inhibits both thrombin-catalyzed fibrin in squamous cell carcinoma of the human lung. Clin Cancer Res formation and factor V activation. J Biol Chem 1982;257:7944–7. 1996;2:763–6. 8. Maruyama I, Salem HH, Majerus PW. Coagulation factor Va binds 28. Hsu SM, Raine L, Fanger H. Use of avidin-biotin-peroxidase com- to human umbilical vein endothelial cells and accelerates protein C plex (ABC) in immunoperoxidase techniques: a comparison between activetion. J Clin Investig 1984;74:224–30. ABC and unlabeled antibody (PAP) procedures. J Histochem Cytochem 9. Esmon CT. The roles of protein C and thrombomodulin in the 1981;29:577–80. regulation of blood coagulation. J Biol Chem 1989;264:4743–6. 29. Miyazaki J, Araki K, Yamato E, et al. Establishment of a pancreatic 10. Suzuki K, Kusumoto H, Deyashiki Y, et al. Structure and expres- beta cell line that retains glucose-inducible insulin secretion: special sion of human thrombomodulin, a thrombin receptor on endothelium reference to expression of glucose transporter isolation. Endocrinology acting as a cofactor for protein C activation. EMBO J 1987;6:1891–7. 1990;127:126–32. 11. Zushi M, Gomi K, Yamamoto S, et al. The last three consecutive 30. Minami K, Yano H, Miki T, et al. Insulin secretion and differential epidermal growth factor-like structures of human thrombomodulin com- gene expression in glucose-responsive and -unresponsive MIN6 sub- prise the minimum functional domain for protein C-activating cofactor lines. Am J Physiol Endocrinol Metab 2000;279:773–81. activity and anticoagulant activity. J Biol Chem 1989;264:10351–3. 31. Gotoh M, Maki T, Kiyoizumi T, Satomi S, Monaco AP. An im- 12. Maruyama I. Recombinant thrombomodulin and activated protein C proved method for isolation of mouse pancreatic islet. Transplantation in the treatment of disseminated intravascular coagulation. Thromb (Baltimore) 1985;40:437–8. Haemostasis 1999;82:718–21. 32. Laemmli UK. Cleavage of structural proteins during the assembly 13. Bajzar L, Morser J, Nesheim M. TAFI, or plasma procarboxypep- of bacteriophage T4. Nature (Lond) 1970;227:680–5. tidase B, couples the coagulation and fibrinolytic cascades through the thrombin-thrombomodulin complex. J Biol Chem 1996;271:16603–8. 33. Esni F, Ta¨ljedal IB, Perl AK, et al. Neural cell adhesion molecule 14. Bajzar L, Nesheim ME, Tracy PB. The profibrinolytic effect of (neural cell adhesion molecule) is required for cell type segregation and activated protein C in clots formed from plasma is TAFI-dependent. ultrastructure in pancreatic islets. J Cell Biol 1999;144:325–37. Blood 1996;88:2093–100. 34. Rouiller DG, Cirulli V, Halban PA. Uvomorulin mediates calcium- 15. Maruyama I, Bell CE, Majerus PW. Thrombomodulin is found on dependent aggregation of islet cells, whereas calcium-independent cell endothelium of arteries, veins, capillaries and lymphatics, and on syn- adhesion molecules distinguish between islet cell types. Dev Biol 1991; cytiotrophoblast of human placenta. J Cell Biol 1985;101:363–71. 148:233–42. 16. Salem HH, Maruyama I, Ishii H, Majerus PW. Isolation and char- 35. Cavallaro U, Niedermeyer J, Fuxa M, Christofori G. neural cell acterization of thrombomodulin from human placenta. J Biol Chem adhesion molecule modulates tumour-cell adhesion to matrix by induc- 1984;259:12246–51. ing FGF-receptor signalling. Nat Cell Biol 2001;3:650–7. 17. Yonezawa S, Maruyama I, Sakae K, et al. Thrombomodulin as a 36. Perl AK, Dahl U, Wilgenbus P, et al. Reduced expression of neural marker for vascular tumors: comparative study with factor VIII and cell adhesion molecule induces metastatic dissemination of pancreatic Ulex europaeus I lectin. Am J Clin Pathol 1987;88:405–11. beta tumor cells. Nat Med 1999;5:286–91. 18. Tabata M, Yonezawa S, Sugihara K, Yamashita S, Maruyama I. 37. Dean YD, McGreal EP, Akatsu H, Gasque P. Molecular and The use of thrombomodulin to study epithelial cell differentiation in cellular properties of the rat AA4 antigen, a C-type lectin like neoplastic and non-neoplastic oral lesions. J Oral Pathol Med 1995;24: receptor with structural homology to thrombomodulin. J Biol Chem 443–9. 2000;257:34382–92.

38. Conway EM, Pollefeyt S, Collen D, Steiner-Mosonyi M. The amino 41. Moberg L, Johansson H, Lukinius A, et al. Production of tissue terminal lectin-like domain of thrombomodulin is required for consti- factor by pancreatic islet cells as a trigger of detrimental thrombotic tutive endocytosis. Blood 1997;89:652–61. reactions in clinical islet transplantation. Lancet 2002;360:2039–45. 39. Oric L, Unger RH. Functional subdivision of islets of Langerhans 42. Peppelenbosch MP, Spek CA. Type I diabetes: a role for tissue and possible role of D-cells. Lancet 1975;2:1243–4. factor in pancreatic islet transplantation? Lancet 2002;360:1999–2000. 40. Steinberg MS, Takeichi M. Experimental specification of cell sort- 43. Esmon CT. Protein C anticoagulant pathway and its role in con- ing, tissue spreading, and specific spatial patterning by quantitative differ- trolling microvascular thrombosis and inflammation. Crit Care Med ence in cadherin expression. Proc Natl Acad Sci USA 1994;91:206–9. 2001;29(7 Suppl):s48–51.

Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2004 American Association for Cancer Research. The Antimetastatic Role of Thrombomodulin Expression in Islet Cell-Derived Tumors and Its Diagnostic Value

Satoshi Iino, Kazuhiro Abeyama, Ko-ichi Kawahara, et al.

Clin Cancer Res 2004;10:6179-6188.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/10/18/6179

Cited articles This article cites 40 articles, 14 of which you can access for free at: http://clincancerres.aacrjournals.org/content/10/18/6179.full#ref-list-1

Citing articles This article has been cited by 4 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/10/18/6179.full#related-urls

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://clincancerres.aacrjournals.org/content/10/18/6179. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.