Suppression of urokinase-type plasminogen activator expression fromhuman ovarian cancer cells by urinary trypsin inhibitor

Hiroshi Kobayashi *, Mika Suzuki, Gung W. Sun, Yasuyuki Hirashima, Toshihiko TeraoDepartment of Obstetrics and Gynecology, Hamamatsu University School of Medicine, Handacho 3600, Hamamatsu,

Shizuoka 431-3192, Japan

Received 17 April 2000; accepted 5 July 2000

Abstract

Urinary trypsin inhibitor (UTI), a Kunitz-type protease inhibitor, efficiently inhibits tumor cell invasion and metastasis.We examined the effect of UTI on urokinase-type plasminogen activator (uPA) expression in ovarian cancer cell lines, HOC-I and HRA. By Northern blot, Western blot, ELISA, and zymographic analyses, we demonstrated that UTI inhibited theexpression of uPA mRNA and protein in these cells in a time- and dose-dependent manner, independent of whetherinduction was triggered by phorbol ester. Monoclonal antibody 4G12, which inhibits UTI binding to the cells, produced adose-dependent abrogation in UTI-mediated down-regulation of uPA expression. These data suggest that UTI significantlydown-regulates tumor cell uPA mRNA expression and protein secretion, and that UTI binding to the cells is necessary toexert the UTI's action. ß 2000 Elsevier Science B.V. All rights reserved.

Keywords: Northern blot; Ovarian cancer; Tumor invasion; Urokinase-type plasminogen activator; Urinary trypsin inhibitor;Zymography

1. Introduction

Urinary trypsin inhibitor (UTI), a Kunitz-typeprotease inhibitor, presents in amniotic £uid, urine,and to a lesser degree in serum [1]. UTI inhibitsvarious serine proteases such as trypsin, K-chymo-trypsin, plasmin, and granulocyte elastase, but noturokinase-type plasminogen activator (uPA) itself.In recent years, a number of studies have emergedinvestigating the e¡ects of UTI on the function ofinhibition of tumor cell invasion and metastasis [2^8]. UTI inhibits not only tumor cell invasion in an invitro assay [2,4^6] but also production of experimen-tal and spontaneous metastasis in an in vivo mouse

model [3,7]. We initially considered that the anti-in-vasive e¡ect is dependent on the antiplasmin activityof UTI, since UTI could act as a protease inhibitor[2].

uPA, a serine protease, has been demonstrated toconvert plasminogen into plasmin, which itself, and/or by activating matrix metalloproteinases, is able todegrade the extracellular matrix [9]. A high uPA con-tent in cancer tissue is associated with a poor prog-nosis [10,11], raising the possibility that uPA is a keyenzyme in extracellular matrix degradation. It is pos-sible that UTI may a¡ect tumor cell behavior bymodulating the expression of the components ofthe plasminogen-plasmin system. In the presentstudy, we focused on the e¡ect of UTI on uPA ex-pression in two ovarian cancer cell lines, HOC-I andHRA, both high uPA producers.

0167-4838 / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved.PII: S 0 1 6 7 - 4 8 3 8 ( 0 0 ) 0 0 1 7 3 - 4

* Corresponding author. Fax: +81-53-435-1626.

BBAPRO 36248 21-9-00

Biochimica et Biophysica Acta 1481 (2000) 310^316www.elsevier.com/locate/bba

2. Materials and methods

2.1. Cells and culture

Human ovarian cancer HOC-I (endometrioid ad-enocarcinoma; [12]) and HRA (serous adenocarcino-ma; [13]) cells were grown in monolayer cultures asdescribed. Stimulated cells were obtained by incuba-tion in the medium containing 100 nM phorbol-12-myristate-13-acetate 4-O-methyl ester (PMA; Sig-ma). Studies were performed by incubating the cul-tures at 37³C in RPMI-1640 containing the indicatedconcentration of UTI (0^5000 nM) and 10% FCS inthe presence or absence of 100 nM PMA for 24 h at37³C. After incubation, the conditioned medium andthe cells were separately collected, centrifuged at1000 rpm for 5 min at 4³C, and stored at 320³Cuntil assayed. The cell number was counted using ahemocytometer.

In a parallel experiment, cells were coincubatedwith the domain-speci¢c antibodies to UTI (mono-clonal antibody (mAb) 4G12 and anti-HI-8 anti-body; see [14]) in the presence of UTI or vehicle asindicated above. Antibodies were used at a ¢nal con-centration of 1, 10, and 100 nM. Mouse or rabbitpreimmune IgG was diluted at the same concentra-tion and used as a control.

2.2. Puri¢cation of UTI

Highly puri¢ed preparation of UTI was suppliedby Mochida Pharmaceutical Co., Tokyo, Japan.

2.3. Quanti¢cation of uPA by ELISA

uPA was quanti¢ed using a commercially availableenzyme-linked immunosorbent assay (ELISA) kit(TintElize, Sweden) according to the manufacturer'sinstructions. All quanti¢cations were done in tripli-cate.

2.4. Sodium dodecyl sulfate^polyacrylamide gelelectrophoresis (SDS^PAGE) and Western blot

Samples were analyzed by SDS^PAGE on a 12%acrylamide gel under nonreducing conditions and byWestern blot. A mouse monoclonal antibody to uPAwas obtained from American Diagnostica Inc.

Greenwich, CT (mAB #394 1:1000 dilution) andused as the primary antibody. Following incubationwith anti-uPA antibody, the paper was incubated inthe secondary antibody, goat anti-mouse IgG conju-gated to biotin (1:500 dilution; Dako, Copenhagen,Denmark). This paper was incubated in avidin^per-oxidase (1:500 dilution; Dako) and then analyzedusing an ECL detection system (Amersham, Tokyo).

2.5. Immunoprecipitation of uPA followed byzymography

Fifty Wl of protein A-Sepharose (Sigma) was incu-bated with culture medium (500 Wl) for 3 h at 4³C.The culture medium was centrifuged and the super-natant was obtained. Fifty Wl of anti-uPA polyclonalantibody (Cosmo Bio, Tokyo) coupled to CNBr-ac-tivated Sepharose 4B was incubated with the super-natants for 3 h at 4³C. The immobilized immuno-complexes were washed, and the ¢nal pellets wereresuspended in electrophoresis sample bu¡er and an-alyzed by a 10% SDS^PAGE containing 0.1% gelatinand 10 Wg/ml plasminogen. The gel was incubated at23³C for 2 h in the presence of 2.5% Triton X-100and then stained for protein with 0.25% Coomassieblue and photographed on a light box. Proteolysiswas detected as a white zone in a dark ¢eld.

2.6. Northern blot

Total RNA was isolated from cells by lysis in Tri-zol reagent (Gibco) according to the manufacturer'sinstructions; 10 Wg of RNA were separated in 1.2%agarose gels and blotted onto Hybond N� mem-branes. uPA mRNA was detected by a radioactivelylabeled uPA oligonucleotide probe. uPA cDNA wasprepared according to the instructions obtainedthrough American Type Culture Collection (Rock-ville, MD). Following hybridization with uPA, blotswere stripped and rehybridized with GAPDH as asemi-quantitative control by densitometry. Aftereach hybridization, the membranes were washedand exposed on Kodak BioMax MS-1 ¢lm at 370³C.

2.7. Statistical analysis

All experiments were performed using at leastthree di¡erent cell preparations. Data are presented

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316 311

as mean þ standard deviation (S.D.). All statisticalanalyses were performed using StatView for Macin-tosh. The Mann^Whitney U-test was used for thecomparisons between di¡erent groups. P less than0.05 was considered signi¢cant.

3. Results

3.1. E¡ect of unstimulated and PMA-stimulatedexpression by UTI on uPA protein level

We examined the time-course of PMA stimulationof uPA protein level in medium conditioned byHOC-I and HRA cells. Fig. 1 shows the data ofexperiments performed using HOC-I cells. The val-ues for HRA cells are very similar to those for HOC-I cells. Prolonged conditions (more than 7 days) sup-plemented with UTI did not reduce the viability of

the cells. Nonstimulated control cells produce ap-proximately 13 ng uPA protein/105 cells per 24 h.The addition of 100 nM PMA resulted in morethan 400% increase in the secretion of uPA protein.During the ¢rst 4 h, there was no di¡erence betweenthe PMA-stimulated cells and the control. Over thenext 4 h between 4 and 8 h, the average rate ofsecretion, as indicated by the slope, increased toV10 ng/105 cells per 4 h, which is about 2.5-foldgreater than control. This 4-h time lag suggeststhat PMA may induce uPA protein synthesis andnot the release of stored uPA protein.

To test whether continuous exposure to UTI caninduced inhibition of unstimulated and PMA-stimu-lated uPA secretion, HOC-I cells were treated withand without 100 nM PMA in the presence of 1.0 WMUTI for up to 24 h (Fig. 1). We demonstrated thatafter 8 h incubation UTI itself has signi¢cant e¡ecton uPA secretion by these cells, independent ofwhether induction was triggered by PMA.

The concentration-dependent increase in the uPA

Fig. 1. E¡ect of PMA on the time course of uPA secretion byHOC-I cells. HOC-I cells at 1U105 cells/well were unstimulated(a, E) and stimulated (b, F) with 100 nM PMA in the pres-ence (E, F) or absence (a, b) of 1 WM UTI. The concentrationof uPA in each culture medium was calculated by readingagainst a standard curve generated with human uPA and ex-pressed in nanograms of uPA/105 cells. Data are mean þ S.D.for quadruplicate samples from a representative experiment.For clarity of presentation, the results for HRA cells are notshown in this ¢gure. *P6 0.05 (b vs. F) ; �P6 0.05 (a vs. E).

Fig. 2. E¡ect of PMA on uPA secretion by HOC-I cells. HOC-I cells were cultured in RPMI-1640 medium supplemented withthe indicated concentrations of PMA and stimulated for 24 h,after which uPA levels were quanti¢ed by ELISA. Data aremean þ S.D. for triplicate samples from a representative experi-ment. Asterisks indicate values that are signi¢cantly di¡erentfrom control (PMA = 0 WM) by Mann^Whitney U-test(*P6 0.05).

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316312

secretion by the PMA stimulation was shown in Fig.2. The e¡ect of PMA on uPA secretion was detect-able at 10 nM and was maximal at 100 nM, where a350^400% increase in uPA secretion was observed.Higher concentrations of PMA (greater than 300nM) did not result in an increase in uPA secretion.

This observation was also apparent at the proteinlevel as demonstrated by Western blot experimentsusing uPA-speci¢c antibodies (data not shown). In aparallel experiment, the media were collected 4, 8, 12,18, and 24 h later in the presence or absence of 100nM PMA and analyzed by immunoprecipitation fol-lowed by zymography (Fig. 3), indicating that PMAstimulated caseinolytic activity in agreement with theincreased uPA antigen level.

The concentration-dependent decrease in the uPAsecretion by UTI in the PMA-stimulated HOC-I cellswas analyzed by ELISA (Fig. 4). The e¡ect of UTIon uPA secretion was detectable at 300 nM and wasmaximal at 1000 nM, where a 65% decrease in uPAsecretion was observed. This observation was againcon¢rmed by Western blot and zymography (datanot shown), demonstrating that UTI inhibited uPAproduction in the PMA-stimulated HOC-I cells in adose-dependent manner.

3.2. E¡ect of unstimulated and PMA-stimulatedexpression by UTI on uPA mRNA level

The e¡ect of UTI on unstimulated and PMA-

Fig. 4. The e¡ect of UTI on PMA-induced uPA secretion: ELI-SA analysis. HOC-I cells were treated for 24 h with PMA (100nM) in the presence of varying concentrations of UTI (0^3000nM) added at t = 0 h. At 24 h, culture media were aspiratedand uPA levels were quanti¢ed by ELISA. The data representthe mean þ S.D. of three experiments, each performed in tripli-cate. Asterisks indicate values that are signi¢cantly di¡erentfrom control (UTI = 0 WM) by Mann^Whitney U-test(*P6 0.05).

Fig. 3. Time course of uPA secretion in the presence or absence of PMA: immunoprecipitation followed by zymographic analysis.HOC-I cells were either untreated (left) or stimulated with 100 nM PMA (right) added at t = 0 h. At 4, 8, 12, 18, and 24 h, culturemedia were aspirated and uPA levels were quanti¢ed by immunoprecipitation (10% acrylamide gel) using anti-uPA antibody followedby zymography. A representative result of three independent experiments is shown. uPA, puri¢ed uPA 4 Wg/lane (lane 1) and 2 Wg/lane (lane 7). The major lytic zone was con¢rmed as uPA since the immunoprecipitation method using a speci¢c anti-uPA antibodywas used and it corresponds to the standard uPA marker.

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316 313

stimulated uPA mRNA expression was shown in Fig.5. UTI (s 100 nM) signi¢cantly inhibited uPAmRNA expression in unstimulated HOC-I cells.The cells were also treated with 100 nM PMA for24 h in the presence of various concentrations ofUTI (30^3000 nM). The PMA e¡ect on uPAmRNA expression was also inhibited by UTI in adose-dependent manner. Decreased uPA mRNA lev-el was detectable at 100^300 nM UTI, and 1000 and3000 nM UTI produced the strongest inhibition forunstimulated and PMA-stimulated uPA mRNA ex-pression.

3.3. UTI's action on uPA mRNA expression wasspeci¢c for the amino-terminus of UTI

It has been established that binding of UTI to thecells was speci¢c for the NH2-terminus of UTI [15].mAb 4G12 recognizes the NH2-terminal domain ofUTI [8]. It does not inhibit UTI active site but in-hibits UTI binding to the cells. polyclonal antibody

to the carboxyl-terminal domain of UTI, HI-8, doesnot recognize UTI-binding region. To determinewhether decreased expression of uPA mRNA and

Fig. 5. The e¡ect of UTI on unstimulated or PMA-induced levels of uPA mRNA. The e¡ect of PMA on uPA mRNA expression wasexamined in HOC-I cells. (A) UTI was added at t = 0 h. Cells were treated as indicated for 24 h and Northern analysis was performedas described in Section 2 using either the cDNA probe for human uPA or GAPDH. A representative blot is shown. (B) Mean inten-sities ( þ S.D.) normalized for loading for GAPDH (n = 4 independent studies).

Fig. 6. The e¡ect of anti-UTI antibodies on UTI-mediateddown-regulation of uPA mRNA in PMA-stimulated cells. mAb4G12 or anti-HI-8 antibody was added to PMA (100 nM)-stimulated HOC-I cells at t = 0 h in the absence or presence ofUTI (1000 nM). Cells were treated as indicated for 24 h andNorthern analysis was performed as described in Section 2 us-ing either the cDNA probe for human uPA or GAPDH. A rep-resentative result of three independent experiments is shown.

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316314

protein by UTI is depend on the UTI binding to thecells, the HOC-I cells were treated with UTI (1000nM) supplemented with PMA (100 nM) for 24 h inthe presence of various concentrations (1.0, 10, and100 Wg/ml) of the domain-speci¢c antibodies to UTI(4G12 or anti-HI-8 antibody) (Fig. 6). 4G12 pro-duced a dose-dependent suppression in UTI-medi-ated down-regulation of uPA mRNA expression inPMA-stimulated cells, whereas anti-HI-8 antibodyhad only a modest e¡ect at the gene level even athigher concentrations of 100 nM. This observationwas also apparent at the protein level as demon-strated by ELISA and Western blot experiments(data not shown).

4. Discussion

We have previously reported that (1) UTI treat-ment showed signi¢cantly reduced invasive andmetastatic potential in vitro and in vivo [2^8,14],that (2) UTI is able to reduce cell-associated proteaseactivity directly via inactivation of plasmin and tryp-sin, while UTI fails to inhibit uPA activity [2], that(3) UTI reduces the production of uPA at the proteinlevel in both leukemia cell line and human umbilicalvein endothelial cells treated with TNF-K [16], andthat (4) UTI is as e¡ective as PKC inhibitors, H7,calphostin C and staurosporin, in inhibiting uPA ex-pression by TNF-K [16]. Taken together, it is likelythat UTI suppresses tumor cell invasion and meta-stasis by mechanisms, possibly by reducing uPA pro-duction and by inhibiting PKC translocation [17]. Ithas not yet been elucidated, however, whether UTIinhibits uPA expression at the gene level.

There is evidence that local extracellular matrixdegradation mediated by the expression of the cell-associated proteases such as uPA and matrix metal-loproteinases is one of the initial steps in tumor cellinvasion and metastasis [18]. uPA is known to betransiently inducible in tumor cells by a variety ofstimuli such as mitogens, growth factors, transmem-brane receptor tyrosine kinases and cytoplasmic on-coprotein members of the Ras signal transductionpathway [19]. It is clearly important to understandhow the uPA expression in tumor cells is regulatedby UTI. These ¢ndings led us to investigate the e¡ectof UTI on tumor cell uPA mRNA expression.

Here we have used human ovarian cancer celllines, HOC-I and HRA, to determine whether UTIa¡ects the expression of uPA on the unstimulatedand PMA-stimulated cells in vitro. We found in thepresent study that UTI signi¢cantly down-regulatedtumor cell uPA mRNA expression and protein secre-tion, independent of whether induction was triggeredby PMA. Inhibition of uPA expression by UTI wasobserved at concentrations neither a¡ecting totalprotein synthesis nor being directly cytotoxic to thecells (data not shown). Concentrations (V300 nM)inhibiting 50% of uPA expression are similar to peakplasma levels that are reached after a single thera-peutic UTI administration [20]. Hence, UTI-medi-ated suppression of uPA expression will be of phar-macologic relevance.

We also found that its inhibitory e¡ect on uPAexpression and secretion appears to be conductedmainly through a UTI binding mechanism, sincemAb 4G12, but not anti-HI-8 antibody, e¤cientlyblocked the UTI-dependent down-regulation ofuPA expression. The best-characterized mechanismby which UTI in£uences the expression of uPA isvia direct binding of UTI to its binding protein(s)on the surface of tumor cells. The UTI binding sitesmay prevent UTI to di¡use away from the cell sur-face. This may result in the e¡ective inhibition orregulation of tumor cell-associated protease activity.The UTI binding sites may not only modulate theprotease activity at the cell surface but also triggerdirectly or indirectly intracellular signal transductionfollowed by regulation of uPA mRNA expression,protein synthesis, and secretion. It has been estab-lished that reduced expression of uPA suppressednot only the invasive potential, but also motilityand adhesion of the tumor cells [21]. To our knowl-edge, this is the ¢rst report showing that a Kunitz-type protease inhibitor, UTI, directly decreases tu-mor cell uPA expression at the gene level.

It is possible that in£ammatory cytokines such asIL-1 and TNF may be responsible for the UTI'se¡ect shown here. This statement is based on theevents reported for cytokine-induced upregulationof uPA expression [22] and UTI-mediated down-reg-ulation of cytokine production [23]. Thus one mayspeculate that UTI blocks uPA expression by down-regulating cytokine production, or that UTI serves toa process important in PKC-dependent protein ki-

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316 315

nase pathway in uPA expression. Taken togetherwith our previous experiments, it is likely that UTImay inhibit the PKC-signalling pathways down-stream of diacylglycerol by a mechanism, possiblyby interrupting the PKC translocation, which mayresult in the down-regulation of uPA mRNA expres-sion. The study on intracellular pathways involved inUTI modulation of uPA will enhance our under-standing of the role that UTI plays in uPA-mediatedcellular behaviors such as in£ammation, neovascula-rization, wound healing, leukocyte extravasation,and tumor cell invasion. Further studies will bealso required to determine whether UTI modulatesdown-regulation of cytokines and whether UTIcauses down-regulation of uPA expression via reduc-ing its half-life or enhancing internalization and deg-radation.

We conclude for the ¢rst time that, at least in thetwo cell lines analyzed here, uPA expression patternsare regulated by UTI both at the mRNA level and atthe protein level, and that the UTI binding to thecells is necessary to inhibit the uPA production inthe cells, independent of whether induction was trig-gered by phorbol ester. Our present results supportthe hypothesis that UTI-induced down-regulation ofuPA expression may contribute to its successful usein anti-metastatic medicine.

Acknowledgements

The authors thank Drs H. Morishita, K. Kato andT. Yamakawa (BioResearch Institute, MochidaPharmaceutical Co., Tokyo), Drs Y. Tanaka andT. Kondo (Chugai Pharmaceutical Co. Ltd., Tokyo),and Drs S. Miyauchi and M. Ikeda (Seikagaku Ko-gyo Co. Ltd., Tokyo) for their continuous and gen-erous support of our work.

References

[1] J.P. Salier, Trends Biochem. Sci. 15 (1990) 435^439.

[2] H. Kobayashi, H. Shinohara, H. Ohi, M. Sugimura, T. Ter-ao, M. Fujie, Clin. Exp. Metastasis 12 (1994) 117^128.

[3] H. Kobayashi, H. Shinohara, K. Takeuchi, M. Itoh, M.Fujie, M. Saitoh, T. Terao, Cancer Res. 54 (1994) 844^849.

[4] H. Kobayashi, M. Fujie, H. Shinohara, H. Ohi, M. Sugi-mura, T. Terao, Int. J. Cancer 57 (1994) 378^384.

[5] H. Kobayashi, J. Gotoh, Y. Hirashima, M. Fujie, D. Sugi-no, T. Terao, J. Biol. Chem. 270 (1995) 8361^8366.

[6] H. Kobayashi, J. Gotoh, N. Kanayama, Y. Hirashima, T.Terao, D. Sugino, Cancer Res. 55 (1995) 1847^1852.

[7] H. Kobayashi, H. Shinohara, J. Gotoh, M. Fujie, S. Fujish-iro, T. Terao, Br. J. Cancer 72 (1995) 1131^1137.

[8] H. Kobayashi, J. Gotoh, M. Fujie, T. Terao, J. Biol. Chem.269 (1994) 20642^20647.

[9] P. Mignatti, D.B. Rifkin, Physiol. Rev. 73 (1993) 161^195.[10] C. Duggan, T. Maguire, E. McDermott, N. O'Higgins, J.J.

Fennelly, M.J. Du¡y, Int. J. Cancer 61 (1995) 597^600.[11] J. Grondahl-Hansen, I.J. Christensen, C. Rosenquist, N.

Brunner, H.T. Mouridsen, K. Dano, M. Blichert-Toft, Can-cer Res. 53 (1993) 2513^2521.

[12] M. Maeda, T. Fujii, T. Terao, Acta Obstet. Gynaecol. Jap.39 (1987) 256^262. in Japanese

[13] Y. Kikuchi, I. Kizawa, K. Oomori, M. Miyauchi, T. Kita,M. Sugita, Y. Tenjin, K. Kato, Cancer Res. 47 (1987) 592^596.

[14] H. Kobayashi, D. Sugino, M.Y. She, H. Ohi, Y. Hirashima,H. Shinohara, M. Fujie, K. Shibata, T. Terao, Eur. J. Bio-chem. 253 (1998) 817^826.

[15] H. Kobayashi, Y. Hirashima, G.W. Sun, M. Fujie, K. Shi-bata, S. Tamotsu, K. Miura, D. Sugino, Y. Tanaka, S. Kon-do, T. Terao, Biochim. Biophys. Acta 1383 (1998) 253^268.

[16] H. Kobayashi, J. Gotoh, T. Terao, Eur. J. Cell Biol. 71(1996) 380^386.

[17] H. Kobayashi, Y. Hirashima, J. Gotoh, D. Sugino, T. Ter-ao, Int. J. Oncol. 12 (1998) 95^105.

[18] E.J. Bernhard, S.B. Gruber, R.J. Muschel, Proc. Natl. Acad.Sci. USA 91 (1994) 4293^4297.

[19] T. Watabe, K. Yoshida, M. Shindoh, M. Kaya, K. Fujika-wa, H. Sato, M. Seiki, S. Ishii, K. Fujinaga, Int. J. Cancer 77(1998) 128^137.

[20] H. Nishikawa, T. Nakabayashi, Y. Nakai, Y. Kurita, M.Fukuoka, T. Onoshi, T. Ogura, A. Sakuma, H. Niitani, E.Tsubura, Jpn. J. Cancer Chemother. 25 (1998) 97^109.

[21] C. Haeckel, S. Krueger, A. Roessner, Int. J. Cancer 77(1998) 153^160.

[22] A.C. de Bart, P.H. Quax, C.W. Lowik, J.H. Verheijen,J. Bone Miner. Res. 10 (1995) 1374^1384.

[23] Y. Futamura, S. Kajikawa, N. Kaga, Y. Shibutani, Obstet.Gynecol. 93 (1999) 100^108.

BBAPRO 36248 21-9-00

H. Kobayashi et al. / Biochimica et Biophysica Acta 1481 (2000) 310^316316