Oncogene (1998) 16, 3143 ± 3149  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc Frequent downregulation of the KAI1(CD82) metastasis suppressor protein in human cancer cell lines

A White, PW Lamb and JC Barrett

Laboratory of Molecular Carcinogenesis, National Institute of Environmental Health Sciences, National Institutes of Health, POB 12233, Research Triangle Park, North Carolina 27709, USA

KAI1 is a metastasis suppressor gene on human their expression could indicate which tumors may chromosome 11p11.2 that encodes a glycoprotein of metastasize. Several metastasis suppressor genes have the transmembrane four superfamily. Reduced KAI1 been mapped, but only a few cloned to date (Dear and expression associates with malignant progression of Ke€ord, 1990; Dong et al., 1996). human prostatic, lung and pancreatic cancers, but the Recently, we reported that the KAI1 gene, located role of KAI1 protein in the malignant progression of on human chromosome 11p11.2, is a metastasis other human cancers remains to be elucidated. We suppressor gene for prostate cancer (Dong et al., analysed KAI1 protein in normal and cancer cells of the 1995). When transfected into the highly metastatic prostate, ovary, bladder, endometrium, lung and Dunning rat subline AT6.1, KAI1 signi®cantly reduces melanocytes by Western blot to determine if KAI1 the metastatic ability of the rat cells. The study also may be involved in multiple cancers. We also showed that KAI1 mRNA expression is down- investigated the relationship of KAI1 expression and regulated in human cell lines derived from prostate two other transmembrane four superfamily proteins, cancer metastases when compared to normal prostate CD81 and CD9, in the cells. We found that KAI1 cells, suggesting a suppressor role for KAI1 in human protein was downregulated in 31/42 of the cancer cell prostate cancer. KAI1 is ubiquitously expressed by lines analysed. Alternatively, some ovarian, bladder and human lymphoid and epithelial cells, and also is endometrial cells had distinct, heterogeneous KAI1 expressed by some ®broblasts (Dong et al., 1995; protein band patterns in Western blots that were due Fukudome et al., 1992). Therefore, it is possible that primarily to N-linked glycosylation. Most of the cancer KAI1 is a metastasis suppressor gene for di€erent types cells expressed two other transmembrane four super- of human cancers. family proteins, CD81 and CD9. Downregulation of The potential importance of KAI1 as a metastasis KAI1 protein may be an indicator of metastatic suppressor gene has been implicated in three human potential in cancers of urogenital, gynecological, and cancers; prostatic, non-small cell lung, and pancreatic pulmonary origin and in melanomas. KAI1 may also carcinomas. Advanced prostate tumors show reduced have post-translational modi®cations speci®c to tissue KAI1 expression compared to normal prostate and type or malignant progression. benign prostatic hyperplasia, and the downregulation does not involve mutation or allelic loss of KAI1 Keywords: KAI1 (CD82); metastasis suppressor gene; (Dong et al., 1996). Both Gleason grade and clinical prostate cancer; transmembrane four superfamily stage of prostate cancers have an inverse correlation with the percentage of KAI1 positive cancer cells (Ueda et al., 1996). Prostate cancers with higher Gleason grades or that are more advanced clinically Introduction have reduced levels of KAI1 expression. Additionally, with non-small cell lung cancers, the survival rate of Metastasis is the major cause of death in cancer patients with KAI1-positive cancer was signi®cantly patients. The spread of tumor cells from a primary higher than the survival rate of patients with KAI1- tumor to secondary sites within the body is complex, negative cancer (Adachi et al., 1996). Finally, advanced involving processes such as migration, invasion, pancreatic tumor stages, in which metastases are adhesion, proliferation and angiogenesis. A primary present, have reduced KAI1 mRNA levels compared challenge of cancer biologists is to develop improved with earlier pancreatic tumor stages (Guo et al., 1996). methods to predict metastatic potential of tumor cells. KAI1 protein is a member of the transmembrane One approach to address this challenge may involve four superfamily (TM4SF) (Gaugitsch et al., 1991). the identi®cation of the genetic events controlling There are at least 15 TM4SF proteins. Members of the metastasis, which are not completely understood TM4SF are cell membrane proteins that contain four (Kohn, 1993). Metastasis suppressor genes, which are hydrophobic, presumably transmembrane, domains functionally inactivated as tumor cells acquire meta- and one large extracellular, hydrophilic domain that static ability, may have signi®cant prognostic value as often contains potential N-linked glycosylation sites. markers of tumor metastatic potential. Mutations in KAI1 is identical to the previously characterized metastasis suppressor genes or downregulation of the antigens R2, IA4, C33 and 4F9 and is designated CD82 by the clusters of di€erentiation (CD) nomen- clature (Engel and Tedder, 1994). KAI1(CD82) contains three potential N-linked glycosylation sites and is a glycoprotein (Dong et al., 1995; Fukudome et Correspondence: JC Barrett Received 28 August 1997; revised 20 January 1998; accepted 21 al., 1992; Lebel-Binay et al., 1994; Nojima et al., 1993). January 1998 TM4SF proteins associate with each other and with KAI1 protein in human cancer cells AWhiteet al 3144 other proteins such as CD4, CD8, Leu-13, CD19, Results CD20 and integrins (Angelisova et al., 1994; Berditch- evski et al., 1996; Rubinstein et al., 1996; Takahashi et KAI1 protein in normal prostate epithelial cells and al., 1990). KAI1 associates with CD81 in human B cell prostate carcinoma cell lines in culture and T cell lines (Imai and Yoshie, 1993) and associates with HLA class I heavy chain and MHC I in various B Normal prostate epithelial cells, PrEC 4428, an cell lines (Lagaudriere et al., 1997). The physiological immortal prostate cell line, PZHPV7, and four roles of TM4SF proteins are largely unknown, but the metastasis-derived prostate carcinoma cell lines were proteins are implicated in signal transduction, cell-cell assayed for KAI1 protein. The apparent molecular interactions, and cell-extracellular matrix interactions. weight of KAI1 in the normal prostate cells and These putative functions are consistent with a role in PZHPV7 cells ranged from 46 to 60 kilodaltons (kD) malignancy. (Figure 1a). The TSU-Prl carcinoma cells had KAI1 Three other TM4SF proteins, in addition to KAI1, protein that was comparable in apparent molecular are associated with malignancy. Antibodies against weight to the protein in normal prostate cells. The level CD81(TAPA1) induce an antiproliferative e€ect on of KAI1 protein was slightly reduced in TSU-1 cells some lymphoma and leukemia cell lines (Oren et al., compared to normal prostate gene epithelial cells. 1990). CD9(MRP-1) inhibits cell motility in Chinese However, signi®cant downregulation (loss of detectable hamster ovary cells, MAC10 human lung adenocarci- expression) of KAI1 protein occurred in three of the noma cells, and ARH77 human myeloma cells (Miyake four prostate carcinoma cell lines (Figure 1a, Table 1). et al., 1991) and reduces metastasis in BL6 mouse KAI1 RNA expression as measured by reverse melanoma cells (Ikeyama et al., 1993). Low CD9 transcriptase polymerase chain reaction analysis in expression correlates with breast cancer metastasis the prostate carcinoma cell lines, and the other types (Miyake et al., 1995). A study of skin-super®cial of cancer cell lines that will be described, was spreading melanomas and metastatic melanomas consistent with KAI1 protein expression (data not showed that CD63(ME491) expression associates with shown). Immuno¯uorescence staining for KAI1 in the early stages of melanoma progression and reduced normal prostate cells and TSU-Pr1 cells showed cell expression in advanced stages of melanoma (Atkinson membrane speci®c staining for KAI1. The cell staining et al., 1985). was heterogeneous and some subclones of TSU-Pr1 In this study, we used Western blot analysis to were positive for KAI1 expression and others were evaluate KAI1 protein expression in multiple types of negative (data not shown). human cancer cell lines and their respective normal We removed N-linked oligosaccharides from cellular cells as an initial study to determine if KAI1 may be proteins using N-glycanase to determine the contribu- involved in the malignant progression of di€erent tion of N-linked glycosylation to the KAI1 molecular types of cancer. Furthermore, we determined the weight. KAI1 protein in PrEC 4428 cells was a 28 kD relationship of KAI1, CD81 and CD9 protein band after removal of N-linked oligosaccharides expression in the cancer cell lines. Established cell (Figure 1b). The 28 kD, deglycosylated KAI1 protein lines are unique tools for studying malignancy but is consistent with the predicted molecular weight of also have limitations due to selection of cells in vitro. KAI1 based upon its amino acid sequence (Dong et al., Nevertheless, human cancer in culture is the starting 1995) and indicates that the broad molecular weight point for much of current cancer research. The majority of cells in tumor-derived cell lines maintain the di€erentiated state of the original tumor (Fusenig a c et al., 1991). Analyses of KAI1 expression in prostate b tumors and prostate tumor-derived cell lines have shown similarities; in both instances, KAI1 down- regulation associates with advanced stage tumors TSU-1 + N-glycanase DU145 PrEC PZHPV7 PrEC PrEC (Dong et al., 1995, 1996; Ueda et al., 1996). We LNCaP PC-3 found that KAI1 protein expression is reduced in TSU-Pr1 TSU-Pr1 + N-glycanase most of the cancer cell lines we examined in 97 — comparison to normal cells. Additionally, normal 97 — 66 — 66 — ovarian and bladder cells and some ovarian, bladder KAI1 66 — and endometrial carcinoma cell lines have KAI1 46 — 46 — protein with distinct, heterogeneous protein band 46 — patterns in Western blots that are most likely due to 30 — N-linked glycosylation. The majority of the cancer cell 30 — 30 — lines have normal CD81 and CD9 protein expression MRP 1 MRP 1 with downregulated KAI1 expression. Collectively, 21.5 — 21.5 — these data suggest that KAI1 downregulation may 21.5 — occur frequently and independently of related TM4SF proteins in malignant progression of various types of Figure 1 Western blot analysis of KAI1 and MRP-1 proteins in normal prostate cells and established prostate cell lines. (a) Cells human tumors and that the protein may contain post- assayed were normal prostate cells, PrEC 4428, immortalized translational modi®cations speci®c to tissue type or prostate cells, lane 2, and prostate carcinoma cell lines, lanes 3 ± 6. malignant progression. The prognostic value of KAI1 The Western blot was hybridized with both KAI1 and MRP-1 in cancers of urogenital and gynecological origin and antibodies. (b) Glycosylated KAI1 protein, lane 1, and deglycosylated KAI1 protein, lane 2, in PrEC 4428 cells. (c) malignant melanomas should be investigated to Glycosylated and deglycosylated KAI1 protein in TSU-Pr1 cells. address the role of KAI1 in these cancers. Size markers in kD are noted KAI1 protein in human cancer cells AWhiteet al 3145 range in KAI1 protein was primarily the result of protein was a discreet band at 45 kD (Figure 2b). various states of N-linked glycosylation. Similarly, KAI1 protein in SK-OV-3 ovarian carcinoma cells was deglycosylated KAI1 protein in TSU-Pr1 prostate also a discreet 45 kD band (data not shown), but it cells was 28 kD (Figure 1c). In all the cancer cell was detected only after extended exposure of the ECLTM types analysed in this study with detectable KAI1 reaction to X-ray ®lm because the protein expression expression, deglycosylated KAI1 protein was 28 kD was downregulated. In stark contrast to the normal (data not shown). cells, some of ovarian carcinoma cell lines had heterogeneous KAI1 protein patterns ranging from 35 ± 60 kD. Half (3/6) of the ovarian carcinoma cell KAI1 protein in malignant melanoma cell lines lines had downregulated KAI1 protein (Figure 2b, Normal human melanocytes, NHEM 2486, contained Table 1). the highest level of KAI1 protein of all the normal cells analysed (Figure 2a). However, the KAI1 protein still KAI1 protein in bladder carcinoma cell lines ranged between 46 and 60 kD. All ®ve of the melanoma cell lines examined originated from metastases and all The KAI1 protein pattern also di€ered in normal had decreased levels of KAI1 protein compared to the bladder cells compared to the other normal cells normal melanocytes (Figure 2a, Table 1). analysed, although the size range of KAI1 protein was similar, 46 ± 60 kD (Figure 2c). The normal bladder cell KAI1 protein pattern consisted of two KAI1 protein in ovarian carcinoma cell lines distinct series of bands instead of the continuous smear KAI1 protein in normal ovarian tissue was unique that was observed in normal prostate cells and compared to the other normal cells examined. The melanocytes. At least 90% of malignant bladder tumors are transitional cell carcinomas (Kakizoe et al., 1995). We analysed KAI1 protein in seven cell lines Table 1 KAI1, CD81, and CD9 protein expression in various derived from transitional cell carcinomas and one cell human cancer cell lines line derived from a transitional cell papilloma, RT-4. KAI1 CD81 CD9 KAI1 protein was downregulated in 38% (3/8) of the Cell type Cell line Origina proteinb protein protein bladder cell lines (Figure 2c, Table 1). The TCC-SUP cell line had increased KAI1 protein expression Prostate PC-3 M 7 + + compared to normal bladder cells. The KAI1 protein DU145 M 7 + + LNCaP.FCG M 7 + + pattern in half of the carcinoma cell lines had an TSU-Pr1 M + + + additional band at about 40 kD (Figure 2c). A Melanoma NZM1 M 7 7 7 representative Western blot for actin protein (Figure NZM2 M 7 7 7 2d) showed consistent loading of protein from the NZM3 M 7 + + NZM4 M 7 + 7 di€erent cell lines. HT144 M 7 7 + Ovarian BG-1 T 7 + + Gratch T + + + KAI1 protein in endometrial carcinoma cell lines Gupton T 7 7 7 KAI1 protein in normal endometrial cells was similar SW626 T + + + PA-1 M + + 7 to the protein in normal prostate epithelial cells (Figure SK-OV-3 M 7 + 7 3a). Normal endometrial epithelial cells were not Bladder HT1197 T 7 + + available for analysis. The endometrial carcinoma cell HT1376 T + + + lines originated from adenocarcinomas, with the J82 T + + 7 RT-4 T 7 7 + exception of the ANC3A cell line, which was derived SCaBer T + + + from a lymph node metastasis (Table 1). KAI1 protein T24 T + 7 7 was downregulated in 88% (7/8) of the carcinoma cell TCCSUP T 2+ + + lines compared to the normal endometrial cells. The UM-UC-3 T 7 7 + HHOUA, Ishikawa and KLE cells had the same Endometrial HEC1A T 7 + 7 HEC1B T 7 + + molecular weight KAI1 protein pattern, 50 ± 70 kD, HEC59 T 7 + 7 although they were derived from tumors in di€erent HHOUA T 7 + 7 patients (Figure 3a). One exception to the consistent HHUA T 7 + 7 correlation between KAI1 RNA and protein expression Ishikawa T 7 7 + KLE T + + + was the HEC59 endometrial carcinoma cell line. KAI1 ANC3A M 7 + + RNA was present in HEC59 cells (data not shown), Lung A549 T 7 + + but KAI1 protein was absent Figure 3. H1299 T 7 + + HTB57 T + + + HTB178 T 7 7 + KAI1 protein in lung carcinoma cell lines HTB182 T 7 + + 1848 M 7 + + We assessed KAI1 protein in three di€erent types of Calu-1 M 7 + + normal human lung cells and eleven cell lines derived HTB58 M 7 + + from non-small cell lung cancers. The normal cells were HTB174 M 7 + 2+ HTB177 M 7 + + mucosal epithelial cells, squamous epithelial cells, and HTB183 M 7 7 7 WI38 embryonic ®broblasts. The normal mucosal cells aOrigin of cell lines; T=primary tumor, M=metastasis. bProtein expressed half the level of KAI1 protein expressed in expression relative to normal control cells; 7, 5twofold; +, normal squamous cells (Figure 3b). The normal comparable; 2+, 4twofold mucosal cells were a control for the 1848 cancer cell KAI1 protein in human cancer cells AWhiteet al 3146

a b Gratch SW626 Gupton SKOV3 PA-1 Ovary NHEM NZM1 NZM2 NZM3 NZM4 HT144 Normal BG-1

97 — 97 —

66 — 66 —

46 — 46 —

30 — 30 —

c d J82 TCC-SUP Scaber Bladder T24 HT1197 Normal UM-UC-3 HT1376 RT-4

97 — WI38 HTB57 SK-OV-3 PA-1 HTB183 HTB183 NHEM NZM1 NZM2 NZM3

66 — 66 KAI1 46 —

30

30 —

TAPA 1 21.5 —

Figure 2 Western blot analysis of normal cells and carcinoma cell lines. (a) KAI1 protein in normal melanocytes, lane 1, and malignant melanoma cell lines, lanes 2 ± 6, showing reduced KAI1 expression in the melanoma cell lines. (b) KAI1 protein in normal ovary, lane 1, and ovarian carcinoma cell lines, lanes 2 ± 7. (c) KAI1 and CD81 proteins in normal bladder, lane 1, and bladder carcinoma cell lines, lanes 2 ± 9. CD81 is indicated by the arrow. (d) Representative Western blot of actin control in di€erent cell lines. Size markers in kD are noted

line. Both the normal mucosal cells and the 1848 cell potential correlation with KAI1 protein in malignant line had KAI1 protein that was 50 ± 60 kD. The normal progression. CD81 protein is not a glycoprotein (Oren squamous cells and WI38 ®broblasts had similar KAI1 et al., 1990) and had an apparent molecular weight of protein expression, 46 ± 60 kD, and were controls for 24 kD in the normal cells and cancer cell lines (Figures the rest of the lung cancer cell lines (Figure 3b and c). 2c and 3c). Of the four possible relationships between KAI1 protein was downregulated in 80% (4/5) of cell KAI1 and CD81 expression in the cancer cell lines, the lines derived from primary lung tumors and in 100% most prevalent correlation, 52% (22/42), was down- (6/6) of the cell lines derived from metastatic lung cells regulated KAI1 expression and normal CD81 protein (Figure 3c, Table 1). expression (Table 1). Both KAI1 and CD81 had normal expression in 24% (10/42) of the cancer cell lines, whereas both proteins were at reduced levels in CD9 and CD81 proteins in the cancer cell lines 21% (9/42) of the cell lines. Only one cell line, T24, Western blot analysis of two KAI1-related proteins, had normal KAI1 protein expression and reduced CD81 and CD9, was evaluated to determine any levels of CD81 protein (Figure 2c, Table 1). KAI1 protein in human cancer cells AWhiteet al 3147 Also the majority of the cancer cell lines, 50% (21/ Discussion 42), had downregulated KAI1 protein and normal CD9 protein expression (Table 1). CD9 is a glycoprotein As a metastasis suppressor protein, KAI1 could be (Miyake et al., 1995) and our Western blots showed two functionally inactivated by mutation, downregulation, protein bands at 25 and 28 KD (Figures 1a and 3c). In post-translational modi®cation, or altered interactions our analyses, the 25 kD protein band was usually more with associated proteins in tumor cells as they become intense than the 28 kD band, and after protein metastatic. This study shows that KAI1 protein deglycosylation only the 25 kD band was detected expression is frequently downregulated in various types (data not shown). The combination of downregulated of human cancer cell lines. Reduced KAI1 protein was KAI1 and CD9 proteins was found in 24% (10/42) of observed in 75% of prostate, 50% of ovarian, 38% of the cancer cell lines, and normal expression of both bladder, 88% of endometrial and 91% of lung proteins was found in 19% (8/42) of the cells. Only three carcinoma-derived cell lines and 100% of the malignant cell lines, J82, T24, and PA-1, had normal KAI1 protein melanoma-derived cell lines we examined. Our results expression and reduced CD9 protein. with the cancer cell lines are consistent with published Of the eight possible expression combinations of data on KAI1 protein expression in human prostate and KAI1, CD81 and CD9 proteins, the most prevalent, lung tumors. Advanced human prostate tumors have 36% (15/42), was downregulated KAI1 expression and reduced KAI1 protein levels (Dong et al., 1996; Ueda et normal expression of CD81 and CD9. Of the cancer cell al., 1996) and nearly 80% of primary, non-small cell lines examined 19% (8/42) had normal expression of all lung cancers had reduced KAI1 protein levels in a recent three TM4SF proteins, whereas about 10% (4/42) of the study (Adachi et al., 1996). Downregulation of KAI1 cell lines had reduced expression of all three proteins. occurs primarily at the transcriptional level. RNA and All of the other possible expression combinations were protein expression data directly correlates in the cancer found in the cancer cell lines, except for the cell lines with the exception of the HEC59 cell line, combination of normal KAI1 and CD9 protein which has normal KAI1 RNA expression but reduced expression and reduced CD81 expression. Four of the protein expression. In addition, KAI1 protein levels in cancer cell lines analysed, HTB183, Gupton, NZM1, the prostate cancer cell lines we analysed directly and NZM2, contained reduced levels for all three correlated with previously reported KAI1 mRNA levels TM4SF proteins (Table 1). Actin protein in those four in those cell lines (Dong et al., 1995). KAI1 down- cell lines was equivalent to positive controls (Figure 2d). regulation may be important in the malignant progres- sion of various solid tissue-derived tumors. KAI1 may be a useful target to study the e€ect of a b speci®c N-linked glycosylation alterations on meta- static potential. We know that KAI1 is a glycoprotein from studies of the protein in other contexts, C33, IA4 and 4F9 antigens, mainly in lymphocytes (Fukudome 96-0027 HHUA KLE Squamous Mucosal HEC1A HEC59 ANC3A HEC1B Ishikawa HHOUA et al., 1992; Lebel-Binay et al., 1994; Nojima et al., 97 — 97 — 1993). The heterogeneous molecular weight, 42 ± 100 kD, of KAI1 in normal and malignant B and T 66 — 66 — cells is attributed to N-linked glycosylation in those studies. KAI1 in normal and cancer cells in our study 46 — 46 — is within the size range reported for lymphocytes and N-linked deglycosylation of KAI1 indicates that the size range is due primarily to N-linked glycosylation. 30 — 30 — We show that KAI1 protein in melanocytes and normal cells of the prostate, ovary, bladder, endome- c trium, and lung has similar, heterogeneous molecular weights, 46 ± 60 kD. However, normal ovarian and AT6.1 W138 A549 HTB174 HTB177 HTB178 HTB182 HTB183 Calu1 H1299 HTB57 1848 bladder cells have KAI1 protein band patterns that suggests tissue speci®c modi®cations in glycosylation. Two normal cell strains we analysed, human prostate ®broblasts and umbilical vein endothelial cells 46 (HUVEC), have no detectable KAI1 protein (data KAI1 not shown). Furthermore, some of the ovarian, bladder, and endometrial carcinoma cell lines have 30 more heterogeneous molecular weights than their normal counterparts. Whether the increased hetero- MRP1 MRP1 geneity in KAI1 molecular weight is a consequence of TAPA1 or antecedent to malignancy remains to be determined. 21.5 Increased b1!6 branched, complex, N-linked oligo- saccharides in malignant cells is common and directly Figure 3 Western blot analysis of endometrial and lung cells. (a) associates with metastatic potential, and accordingly, KAI1 protein in normal endometrial ®broblasts, lane 1, and inhibition of N-linked glycosylation associates with endometrial carcinoma cell lines, lanes 2 ± 9, are shown. (b) KAI1 decreased cell invasion and metastasis (Dennis, 1991). protein in squamous lung epithelial cells, lane 1, and mucosal We did not address speci®c oligosaccharide structural lung epithelial cells, lane 2. (c) KAI1, MRP-1, and TAPA1 proteins in normal lung ®broblasts, lane 1, and non-small cell features of KAI1 in this study, but such studies in lung carcinoma-derived cell lines, lanes 2 ± 12 normal and cancer cells are in progress. KAI1 protein in human cancer cells AWhiteet al 3148 KAI1-related TM4SF proteins CD81 and CD9 are University School of Medicine) and grew in KSFM (Gibco expressed in the majority of the cancer cell lines assayed, BRL) supplemented with 5 ng/ml EGF and 50 mg/ml in contrast to KAI1 protein. The most common bovine pituitary extract (Gibco BRL). Dr Carrie Rinker- expression combination of the three proteins is down- Schae€er (University of Chicago) provided TSU-Pr1 regulated KAI1 and normal expression of CD81 and prostate carcinoma cells. Dr Cli€ord Rinehart (University of North Carolina at Chapel Hill) provided normal CD9. Compared to only 26% (11/42) of the cancer cell endometrial ®broblasts, 96 ± 0027. Normal human mela- lines with normal KAI1 expression, 76% (32/42) of the noctyes, NHEM 2486 (Clonetics), grew in MGM-3 cell lines have normal CD81 expression, and 69% (29/ Bulletkit (Clonetics). Dr Anton Jetten (National Institute 42) of the cell lines have normal CD9 expression. These of Environmental Health Sciences) provided normal results support studies in which CD81 and CD9 squamous lung epithelial cells and normal mucosal lung expression are detected in cell lines derived from epithelial cells. Dr Bruce Baguley (University of Auckland carcinoma, sarcoma, and melanoma cells (Berditch- School of Medicine, Auckland, New Zealand) provided evski et al., 1996; Oren et al., 1990) and another study malignant melanoma cell lines NZM1, NZM2, NZM3, and that indicated CD9 is expressed on almost all human NZM4 that grew in aMEM (Gibco BRL) supplemented solid tumor cell lines (Miyake et al., 1991). However, with 10% fetal bovine serum. Cell lines contributed by Dr H Sasaki (Jikei University School of Medicine, Tokyo, the expression arrangement among the three proteins is Japan) included BG-1, HEC59, HHOUA, HHUA and variable in the cell lines we examined, which implies that Ishikawa. We obtained all other cell lines from the co-regulation of the three proteins is unlikely. American Type Culture Collection (ATCC, Rockville, This study did not address whether KAI1 protein Maryland). Lung carcinoma cell lines were cultured in expression in the cancer cell lines correlates with media speci®ed by the ATCC. Dr Jack Taylor (National invasion (in vitro) or spontaneous metastasis (in vivo). Institute of Environmental Health Sciences) kindly In vitro invasion assays using tumor cell lines and their provided frozen, normal human bladder tissue. relevance to metastatic human tumors is open to question, and a review of the literature indicates that Western blot analysis most malignant human tumor cell lines are not The Western blot protocol used was previously described metastatic in experimental systems (Sharkey and Fogh, (Fukudome et al., 1992). Brie¯y, cell proteins were 1979). Perhaps the best way to establish a correlation solubilized in lysis bu€er (10 mM Tris pH 8.0, 150 mM between metastatic potential and KAI1 expression is NaCl, 3 mM MgCl2,0.5%NP±40,2mMphenylmethylsul- through retrospective studies with normal and benign fonly ¯uoride) when cell cultures reached 50 ± 90% tissues, primary tumors, and matched metastases. con¯uence. The proteins were normalized to 1 mg/ml, We think downregulation of KAI1 signi®es that a mixed with an equal volume of Laemmli's sample bu€er tumor cell can metastasize and thus KAI1 may be a without 2-mercaptoethanol, and boiled for 3 min. After potential prognostic marker. Allelic loss and mutation SDS-polyacrylamide gel electrophoresis (17.5%) of 40 mg rarely occur in the KAI1 gene in prostate carcinomas total protein per cell line, proteins were electrophoretically (Dong et al., 1996). So, it is probable that KAI1 transferred to Immobilon-P (Millipore) membrane. Mem- branes were stained with Ponceau S (Sigma) to assess equal mutations may not occur frequently in other cancer loading of proteins. We used C33 hybridoma supernatant types. Therefore, studies that focus on KAI1 down- (Dr Osamu Yoshie, Shionogi Institute for Medical Science, regulation and glycosylation alterations may be most Osaka, Japan) to detect KAI1 protein, 5A6 antibody (Dr informative concerning the value of KAI1 as a Shoshana Levy, Stanford University) to detect CD81 prognostic marker. In particular, our results suggest (TAPA1) protein, and CD9 antibody (Serotec) to detect that the role of KAI1 in malignant progression of CD9(MRP-1) protein. The ECLTM Western blotting gynecological and other urogenital cancers and malig- analysis system (Amersham) detected bound antibody. nant melanomas should be investigated along with Protein from normal human ovary was provided by Dr ongoing studies of prostatic, lung and pancreatic Bernard Weissman (University of North Carolina at cancers. In a current study, we are analysing KAI1 Chapel Hill). Each cell line was assayed in three independent experiments. Densitometer readings of the protein expression in human breast cancer cell lines, protein were made, and reduced expression was scored as primary breast tumors, and breast cancer metastases. cell lines with a reproducible 4twofold decrease in Preliminary results have shown KAI1 downregulation in expression of all forms of the protein. some breast tumors and alterations in the KAI1 protein pattern in other breast tumors. Primary ovarian tumors Protein deglycosylation have been analysed for KAI1 protein expression, and heterogeneous KAI1 protein patterns spanning a wide We used N-glycanase (Genzyme) and the protocol molecular weight range were observed in Western blots. recommended by the manufacturer to remove N-linked oligosaccharides from proteins. In the protocol, proteins were incubated with enzyme in the presence of 5% NP-40 overnight. Enzyme-treated proteins were mixed with Materials and methods sample bu€er and analysed by Western blotting as described. Cells and culture conditions All cells were of human origin and were maintained in a

378C humidi®ed chamber with 5% CO2. The cells grew in Acknowledgements DMEM/Ham's F12 medium (1:1) (Gibco BRL) supple- We thank Dr Cynthia Afshari for helpful discussion and mented with 10% fetal bovine serum (Summit), unless review of this manuscript. Thanks also to Dr John Roberts noted otherwise. Normal prostate epithelial cells, PrEC for thoughtful review of this manuscript and Drs Osamu 4428 (Clonetics), grew in PrEGM Bulletkit (Clonetics). Yoshie, Shoshana Levy, Bernard Weissman, Jin-Tang PZHPV7, an HPV18-immortalized prostate epithelial cell Dong, Cli€ord Rinehart, Anton Jetten, Jack Taylor and line, was a generous gift from Dr Donna Peehl (Stanford Donna Peehl for providing reagents. KAI1 protein in human cancer cells AWhiteet al 3149 References

Adachi M, Taki T, Ieki Y, Huang C, Higashiyama M and Imai T and Yoshie O. (1993). J. Immunol., 151, 6470 ± 6481. Miyake M. (1996). Cancer Res., 56, 1751 ± 1755. Kakizoe T, Fair WR, Smith PH, Algoba F, Ferrari P, Angelisova P, Hilgert I and Horejsi V. (1994). Immunoge- Grossman HB, Zirkali Z, Tsukamoto T and Tachibana M. netics, 39, 249 ± 256. (1995). Int. J. Urol., 2 suppl 2, 58 ± 63. Atkinson B, Ernst CS, Ghrist BFD, Ross AH, Clark WH, Kohn EC. (1993). Anticancer Res., 13, 2553 ± 2560. Herlyn M, Herlyn D, Maul G, Steplewski Z and Lagaudriere-Gesbert C, Lebel-Binay S, Wiertz E, Ploegh Koprowski H. (1985). Hybridoma, 4, 243 ± 255. HL, Fradelizi D, and Conjeaud H. (1997). J. Immunol., Berditchevski F, Zutter MM and Hemler ME. (1996). Mol. 158, 2790 ± 2797. Biol. Cell, 7, 193 ± 207. Lebel-Binay S, Gil ML, Lagaudriere C, Miloux B, Marchiol- Dear TN and Ke€ord RF. (1990). Mol. Aspects Med., 11, Fournigault C, Guillet-Mary A, Lopez M, Fradelizi D, 243 ± 324. and Conjeaud H. (1994). Cell. Immunol., 154, 468 ± 483. Dennis JW. (1991). Semin. Cancer Biol., 2, 411 ± 420. Miyake M, Koyama M, Seno M and Ikeyama S. (1991). J. Dong JT, Lamb PW, Rinker-Schae€er CW, Vukanovic J, Exp. Med., 174, 1347 ± 1354. Ichikawa T, Isaacs JT and Barrett JC. (1995). Science, 268, Miyake M, Nakano K, Ieki Y, Adachi M, Huang C, Itoi S, 884 ± 886. Koh T and Taki T. (1995). Cancer Res., 55, 4127 ± 4131. Dong JT, Suzuki H, Pin SS, Bova S, Schalken JA, Isaacs WB, Nojima Y, Hirose T, Tachibana K, Tanaka T, Shi L, Doshen Barrett JC and Isaacs JT. (1996). Cancer Res., 56, 4387 ± J, Freeman GJ, Schlossman SF, and Morimoto C. (1993). 4390. Cell. Immunol., 152, 249 ± 260. Engel P and Tedder TF. (1994). Leuk. Lymphoma, 13, 61 ± Oren R, Takahashi S, Doss C, Levy R and Levy S. (1990). 64. Mol. Cell. Biol., 10, 4007 ± 4015. Fukudome K, Furuse M, Imai T, Nishimura M, Takagi S, Rubinstein E, Naour FL, Lagaudriere-Gesbert C, Billard M, Hinuma Y and Yoshie O. (1992). J. Virol., 66, 1394 ± 1401. Conjeaud H, and Boucheix C. (1996). Eur. J. Immunol., Fusenig NE, Breitkreutz D and Boukamp P. (1991). Human 26, 2657 ± 2665. Cancer in Primary Culture. Masters JRW (ed.), Kluwer Sharkey FE and Fogh J. (1979). Int. J. Cancer, 24, 733 ± 738. Academic Publishers: Netherlands. pp. 55 ± 80. TakahashiS,DossC,LevyS,andLevyR.(1990).J. Gaugitsch HW, Hofer E, Huber NE, Schnabl E and Immunol., 145, 2207 ± 2213. Baumruker T. (1991). Eur. J. Immunol., 21, 377 ± 383. Ueda T, Ichikawa T, Tamaru J, Mikata A, Akakura K, GuoX,FriessH,GraberHU,KashiwagiM,Zimmermann Akimoto S, Imai T, Yoshie O, Shiraishi T, Yatani R, Ito H A, Korc M and Buchler MW. (1996). Cancer Res., 56, and Shimazaki J. (1996). Am. J. Pathol., 149, 1435 ± 1440. 4876 ± 4880. Ikeyama S, Koyama M, Yamaoko M, Sasada R and Miyake M. (1993). J. Exp. Med., 177, 1231 ± 1237.