Oncogene (2000) 19, 1288 ± 1296 ã 2000 Macmillan Publishers Ltd All rights reserved 0950 ± 9232/00 $15.00 www.nature.com/onc Prostate stem cell antigen (PSCA) expression increases with high gleason score, advanced stage and bone metastasis in

ZGu1,7, G Thomas1,2,7, J Yamashiro1, IP Shintaku2, F Dorey3, A Raitano4, ON Witte5, JW Said2, M Loda6 and RE Reiter*,1

1Department of Urology, University of California, Los Angeles, California, CA 90095, USA; 2Department of Pathology, University of California, Los Angeles, California, CA 90095, USA; 3Department of Orthopedics, University of California, Los Angeles, California, CA 90095, USA; 4Urogenesys, Inc., Santa Monica, California, CA 90404, USA; 5Department of Microbiology and Molecular Genetics, Molecular Biology Institute, Howard Hughes Medical Institute, University of California, Los Angeles, California, CA 90095, USA; 6Department of Adult Oncology, Dana Farber Cancer Institute, Boston, Massachusetts, MA 02115, USA

Prostate stem cell antigen (PSCA) is a recently de®ned Introduction homologue of the Thy-1/Ly-6 family of glycosylpho- sphatidylinositol (GPI)-anchored cell surface antigens. Prostate cancer is the most common cancer diagnosis PSCA mRNA is expressed in the basal cells of normal and the second leading cause of cancer-related death prostate and in more than 80% of prostate cancers. The in American men (Lalani et al., 1997). Although great purpose of the present study was to examine PSCA progress has been made in the diagnosis and manage- expression in clinical specimens of human ment of localized disease, there continues to be a need prostate cancer. Five monoclonal antibodies were raised for new diagnostic markers that can better discrimi- against a PSCA-GST fusion protein and screened for nate between indolent and aggressive variants of their ability to recognize PSCA on the cell surface of prostate cancer. There also continues to be a need human prostate cancer cells. Immunohistochemical for the identi®cation and characterization of potential analysis of PSCA expression was performed on new therapeutic targets on prostate cancer cells. paran-embedded sections from 25 normal tissues, 112 Current immunological approaches to the control primary prostate cancers and nine prostate cancers and treatment of recurrent and metastatic prostate metastatic to bone. The level of PSCA expression in cancer have been limited by a lack of prostate and prostate tumors was quanti®ed and compared with cancer-speci®c target antigens. Although a number of expression in adjacent normal glands. The antibodies prostate-speci®c have been identi®ed (i.e. detect PSCA expression on the cell surface of normal prostate speci®c antigen, prostatic acid phosphatase, and malignant prostate cells and distinguish three glandular kallikrein 2), the majority of these are extracellular epitopes on PSCA. Prostate and transi- secreted not ideally suited for many im- tional epithelium reacted strongly with PSCA. PSCA munological strategies, such as in vivo targeting by staining was also seen in placental trophoblasts, renal monoclonal antibodies. Prostate-speci®c membrane collecting ducts and neuroendocrine cells in the stomach antigen (PSMA) is the only cell surface antigen yet and colon. All other normal tissues tested were negative. reported which has relative prostate-speci®city (Fair et PSCA protein expression was identi®ed in 105/112 al., 1997; Murphy et al., 1998a; Silver et al., 1997). (94%) primary prostate tumors and 9/9 (100%) bone Monoclonal antibodies that recognize the extracellular metastases. The level of PSCA expression increased with domain of PSMA have recently been reported and are higher Gleason score (P=0.016), higher tumor stage currently being evaluated for use in prostate cancer (P=0.010) and progression to androgen-independence diagnosis and treatment (Liu et al., 1997; Murphy et (P=0.021). Intense, homogeneous staining was seen in al., 1998b). all nine bone metastases. PSCA is a cell surface protein We recently reported the identi®cation of prostate with limited expression in extraprostatic normal tissues. stem cell antigen (PSCA) (Reiter et al., 1998). PSCA is PSCA expression correlates with tumor stage, grade and a 123 amino acid glycoprotein with 30% identity to androgen independence and may have prognostic utility. stem cell antigen 2 (Sca 2), a cell surface marker of Because expression on the surface of prostate cancer immature thymic lymphocytes (Antica et al., 1997; cells increases with tumor progression, PSCA may be a Classon and Coverdale, 1994). Like Sca-2, PSCA is useful molecular target in advanced prostate cancer. attached to the cell membrane by a GPI anchor which Oncogene (2000) 19, 1288 ± 1296. can be cleaved by treatment with a GPI-speci®c phospholipase. Among 16 normal tissues examined, Keywords: prostate; neoplasm; prostate stem cell PSCA mRNA was expressed predominantly in prostate antigen (PSCA); antibody and placenta. mRNA in situ hybridization localized PSCA expression in normal prostate to the basal cell epithelium, the putative stem cell compartment of prostatic epithelium, suggesting that PSCA may be a *Correspondence: RE Reiter, Department of Urology, UCLA, 66- marker of prostate stem/progenitor cells. PSCA 134 Center for the Health Sciences, 10833 Le Conte Avenue, Los mRNA was also detected in greater than 80% of Angeles, CA 90095-1738 primary prostate cancers by in situ analysis. In 7Both authors contributed equally to this work Received 20 August 1999; revised 14 December 1999; accepted 20 addition, mRNA expression in cancers often appeared December 1999 to be stronger than in adjacent normal glands, raising Prostate stem cell antigen expression ZGuet al 1289 the possibility that PSCA may be overexpressed in neomycin-alone containing vector (LNCaP-neo) were some prostate tumors and might have some cancer both negative in both analyses. These results demon- speci®city. PSCA maps to 8q24.2, a strate that all ®ve mAbs react with PSCA on the cell region of genetic gain/ampli®cation in a large percen- surface of intact prostate cancer cells. tage of advanced prostate cancers (Cher et al., 1994). The location on PSCA of the epitopes recognized by These initial results supported PSCA as a potential the ®ve mAbs was determined by immunoblot analysis target for prostate cancer diagnosis and therapy. using three truncated PSCA-GST fusion proteins. In order to examine PSCA protein expression in vivo mABs 4A10, 2H9 and 3C5 recognize an epitope and validate it as a potential target for prostate cancer residing within the amino-terminal portion of PSCA diagnosis and/or therapy, we have isolated and (amino acids 21 ± 50); mAb 1G8 recognizes an epitope characterized a series of monoclonal antibodies in the middle of PSCA (amino acids 46 ± 85); and mAb directed against PSCA. These antibodies detect PSCA 3E6 reacts with the carboxyl-terminal portion of PSCA on the cell surface of nonpermeabilized cells, con®rm- (amino acids 85 ± 99). These results demonstrate that ing the cell surface localization of PSCA. Immunohis- the ®ve mAbs recognize at least three distinct epitopes tochemical analysis of a large series of primary tumors on PSCA. demonstrates that PSCA expression increases in more advanced stages and grades of prostate cancer. In Immunohistochemical staining of PSCA in a normal addition, PSCA is expressed strongly in all cases of prostate prostate cancer bone metastases examined. These results support PSCA as a potential diagnostic and/or PSCA mRNA localizes to a subset of basal cells in a therapeutic target in prostate cancer. normal prostate, suggesting that PSCA may be a cell surface marker for prostate stem/progenitor cells (Reiter et al., 1998). In order to test the possibility that PSCA protein may be a marker of basal cells, Results PSCA expression was examined immunohistochemi- cally in paran-embedded sections of normal prostate. Monoclonal antibodies directed against PSCA stain the mAbs 1G8 and 2H9 stained the cytoplasm of both cell surface of prostate cancer cells and recognize a basal and secretory cells, while mAb 3E6 reacted minimum of three epitopes on PSCA predominantly with basal cells (Figure 2). Atrophic Monoclonal antibodies (mAbs) were raised against a glands, which express basal cell cytokeratins, stained PSCA-GST fusion protein lacking both the amino and strongly with all three mAbs (O'Malley et al., 1990). In carboxyl terminal signal sequences of PSCA. Positive addition, prostatic neuroendocrine cells were positive fusions were selected by ELISA using the PSCA-GST (con®rmed by double-staining for chromogranin) (data fusion protein and GST alone. Out of 400 hybridomas not shown). mAbs 3C5 and 4A10 gave strong back- screened, 28 recognized the PSCA-GST fusion but not ground staining and/or nonspeci®c nuclear staining in GST alone. These fusions were screened secondarily by paran sections and were not used further. Therefore, ¯ow cytometry of nonpermeabilized 293T cells trans- although PSCA mRNA is detected speci®cally in basal fected with PSCA and mock transfected 293T cells. cells, PSCA protein is present in both epithelial cell Secondary screening by FACS was done in order to layers (i.e. basal and secretory) and the neuroendocrine select clones capable of recognizing PSCA on the cell cells of the prostate. surface, hypothesizing that these might later become useful for in vivo targeting applications. Seven positive Immunohistochemical analysis of normal tissues fusions were identi®ed in this manner, of which ®ve (mAbs 4A10, 1G8, 3E6, 3C5 and 2H9) were subcloned Our initial studies indicated that PSCA expression in and puri®ed. men was largely prostate-speci®c, with low levels of In order to determine the ability of mAbs 2H9, detectable mRNA in kidney and small intestine. PSCA 3E6, 1G8, 4A10 and 3C5 to recognize PSCA mRNA was also detected in placenta. The prostate- speci®cally on the cell surface of prostate cancer cells, speci®city of PSCA protein expression was tested by LNCaP cells transfected with PSCA (LNCaP-PSCA) immunohistochemical staining of 25 tissues using mAb and LAPC-9 cells were examined by ¯ow cytometry 1G8 (see Table 1). As predicted by the RNA analysis, and indirect immuno¯uorescence. LAPC-9 is a placenta was positive with all mAbs tested, with prostate cancer xenograft which expresses PSCA cytoplasmic staining detected in the trophoblasts mRNA, whereas the LNCaP prostate cancer cell line (Figure 3). In kidney, weak staining was detected in does not express endogenous PSCA. All ®ve mAbs the collecting ducts and some tubules, but not in were able to detect PSCA on the cell surface of glomeruli (Figure 3). Transitional epithelium of the nonpermeabilized LNCaP-PSCA and/or LAPC-9 cells bladder, which had not been examined previously at by ¯ow cytometry (Figure 1a) and immuno¯uores- the mRNA level, was also positive (Figure 3). In cence (Figure 1b). mAb 2H9 stained the cell surface of addition, there was staining of colonic and gastric LAPC-9 but not LNCaP-PSCA, suggesting that the neuroendocrine cells (con®rmed by doublestaining with epitope recognized by this mAb may be obscured in chromogranin) (Figure 3). Expression of PSCA by the the latter cells. All mAbs produced a punctate staining stomach and bladder were con®rmed by Northern blot pattern on the cell surface, which was most analysis (data not shown). These results demonstrate pronounced with mAbs 3E6, 3C5 and 4A10 (Figure that PSCA expression in men is restricted to prostate 1b). This pattern may re¯ect aggregation or clustering and a small set of nonprostatic tissues, most prominent of PSCA to regions of the cell surface. Mock- of which are bladder and neuroendocrine cells of the transfected LNCaP and LNCaP transfected with a stomach and colon.

Oncogene Prostate stem cell antigen expression ZGuet al 1290

Figure 1 Cell surface recognition of PSCA by anti-PSCA monoclonal antibodies. (a) Flow cytometric recognition of PSCA on the cell surface of nonpermeabilized LAPC-9 human prostate cancer cells using mAbs 1G8, 2H9, 3E6, 3C5 and 4A10. Staining was compared to an irrelevant isotype control antibody. (b) Immuno¯uorescent analysis demonstrating cell surface expression of PSCA in nonpermeabilized prostate cancer cells. LNCaP cells were stably transfected with PSCA and stained with mAbs 1G8, 3E6, 3C5 and 4A10. Negative controls included irrelevant isotype antibody and LNCaP cells transfected with control vector, all of which showed no staining even after prolonged exposures (not shown). mAb 2H9 also did not stain these cells, as described in the text, and is not shown

Figure 2 Immunohistochemical staining of a normal prostate. Paran-embedded sections of normal prostate were immunostained with anti-PSCA mAbs 1G8 and 3E6. mAb 1G8 reacts equally with both basal (arrowhead) and secretory (arrow) cells, whereas mAb 3E6 stains basal cells (arrowhead) more strongly than secretory cells (arrow). Speci®city of staining was tested by preincubation of mAbs 1G8 and 3E6 with excess GST-fusion protein and by staining with secondary antibody alone, both of which showed no staining (data not shown)

Oncogene Prostate stem cell antigen expression ZGuet al 1291 mAb 1G8 (Figure 4). As described in Materials and PSCA protein is expressed by a majority of localized methods, slides were assigned a composite score of 0 ± 9 prostate cancers and correlates with gleason score, pathologic stage and androgen-independence based on the percentage of positive cells and staining intensity. One hundred and ®ve of 112 (94%) tumors In our previous study, PSCA mRNA was expressed in stained positive for PSCA, of which 23 (21%) cancers *80% of tumors and appeared to be expressed more stained very strongly for PSCA (i.e. nine) and another highly in malignant than normal glands (Reiter et al., 70 (63%) had moderate staining (i.e. scores 3 ± 6) 1998). In order to determine if PSCA protein can be (Table 2). The remaining 19 (17%) tumors were weakly detected in prostate cancers and if PSCA protein levels positive (i.e. scores 1 ± 2; 11%) or negative (i.e. score 0; are increased in malignant compared with benign 6%). Using this semi-quantitative scoring method, the glands, 112 paran-embedded prostate cancer speci- level of PSCA expression correlated signi®cantly with mens were analysed by immunohistochemistry using both Gleason score (P=0.016) and the pathologic tumor stage (P=0.008). Thirty-two per cent of Gleason 8 ± 10 cancers had very strong staining com- Table 1 Immunohistochemical staining of human tissues with anti- pared to only 11% with Gleason scores 5 ± 6. Likewise, PSCA monoclonal antibody 1G8 40% of locally advanced and node positive tumors (i.e. Staining Tissue (cell type)* stages T3c-T4 and N+) expressed high levels of PSCA Positive Prostate (epithelium-basal, secretory and neuroendocrine versus 15% that were organ con®ned or had extra- cells), bladder (transitional epithelium), placenta (tro- capsular extension alone (i.e. T2-T3b) (Table 3). phoblasts), stomach (neuroendocrine cells), colon When PSCA immunostaining in the tumor was (neuroendocrine cells), kidney (collecting ducts) compared to that in adjacent normal glands, 40 Negative Testis, endometrium, liver, gallbladder, pancreas, small intestine, bone marrow, thymus, spleen, lymph node, (36%) tumors showed PSCA overexpression (i.e. breast expression in the tumor greater than normal). PSCA skeletal muscle, heart, brain, peripheral nerve, skin, overexpression correlated with Gleason score lung, thyroid, adrenal (P=0.008). Forty-eight per cent of Gleason 8 ± 10 *A minimum of two specimens from each tissue was examined tumors overexpressed PSCA compared with only 22%

Figure 3 Examples of PSCA immunostaining in extraprostatic normal tissues. Paran-embedded tissue specimens from placenta, bladder, colon and kidney were stained with anti-PSCA monoclonal antibodies. In the placenta, staining was seen in trophoblasts (arrow). In the bladder, staining was con®ned to the super®cial layer of transitional epithelium (arrow). Staining in the colon was seen in neuroendocrine cells (arrow) within colonic crypts and con®rmed by double-staining with chromogranin as shown (blue signi®es chromogranin and brown signi®es 1G8); no counterstain used). Staining in the kidney was seen in collecting ducts (arrow) and occasional distal tubules (note that endogenous biotin activity was blocked as described in Materials and methods)

Oncogene Prostate stem cell antigen expression ZGuet al 1292

Figure 4 Examples of PSCA immunostaining in primary prostate cancers. (a) A poorly di€erentiated (Gleason 4+4=8) tumor (arrow) with strong staining in all malignant cells (composite score=9, see text) and signi®cantly stronger expression in tumor when compared with adjacent normal glands, which in this case stain weakly or not at all (occasionally patchy staining of normal glands was noted) (arrowhead) (b) A moderately di€erentiated (Gleason 3+3=6) tumor (arrow) with moderate staining in all malignant cells (composite score=6, see text) and equivalent expression in tumor when compared with normal glands (arrowhead). (c)A poorly di€erentiated (Gleason 5+5=10) tumor (arrow) with strong, uniform staining. Note the strong staining of the invading tumor cells around blood vessels (arrowhead). (d) A cribriform carcinoma (arrow) which overexpresses PSCA when compared to adjacent normal glands which are weakly staining (arrowhead)

Table 2 Correlation of PSCA expression with gleason score using two independent scoring systems Intensity6frequency* Tumor vs normal**,*** Gleason score 0 (%) 1 ± 2 (%) 3 ± 6 (%) 9 (%) T5N (%) T=N (%) T4N (%)

2 ± 4 0 (0) 0 (0) 4 (100) 0 (0) 1 (25) 3 (75) 0 (0) 5 ± 6 3 (11) 4 (15) 17 (63) 3 (11) 5 (19) 16 (59) 6 (22) 7 2 (4) 3 (6) 33 (70) 9 (19) 2 (4) 27 (57) 18 (38) 8 ± 10 2 (6) 5 (15) 16 (47) 11 (32) 3 (4) 14 (42) 16 (48) *Kendall Tau coecient=0.081, P=0.016 for the comparison of a score of 9 with scores 0 ± 6. **Kendall Tau coecient=0.221; P=0.008 for the comparison of T4N with T=5N. ***One tumor could not be scored because no normal glands could be identi®ed

Table 3 Correlation of PSCA expression with pathologic tumor overexpressed PSCA compared with 28% of localized stage using two independent scoring systems cancers, although this did not reach statistical sig- Intensity6 ni®cance (P=0.067; Table 3). Six specimens were frequency* Tumor vs normal** obtained from patients treated prior to surgery with Tumor stage 0 ± 6 (%) 9 (%) T5=N (%) T4N (%) hormone ablation therapy. Five of six (83%) of these T1-T3b 62 (85) 11 (15) 52 (72) 21 (28) androgen-independent cancers overexpressed PSCA, T3c-T4, N+ 15 (60) 10 (40) 13 (52) 12 (48) compared with 33% of patients with androgen- *P=0.010 (Fisher's exact test) for the comparison of scores 0 ± 6 and dependent tumors (P=0.021). 9. **P=0.067 (Fisher's exact test) for the comparison of T5=N and These results demonstrate that PSCA protein is T4N expressed in a majority of cancers and that the level of PSCA expression (de®ned both semi-quantitatively and by comparison with adjacent normal glands) correlates of Gleason 5 ± 6 cancers (Table 2). Interestingly, four with grade, stage and progression to androgen- of the six Gleason 5 ± 6 tumors with PSCA over- independence. expression contained large foci of cribriform carcino- ma, a subtype of prostate adenocarcinoma which has Correlation of PSCA immunostaining and mRNA in situ been associated with worse prognosis and a high hybridization frequency of genetic aberrations (McNeal et al., 1986; Qian et al., 1997; Rubin et al., 1998). With regards to In 102 cases, we were able to compare the results of the pathologic stage, 48% of T3c-T4, N+ tumors PSCA immunostaining with our previous mRNA in

Oncogene Prostate stem cell antigen expression ZGuet al 1293 situ hybridization analysis (ISH) (Reiter et al., 1998). with decalci®cation bu€er. Although this treatment Areas of positive immunostaining were identical to increased background staining, it did not alter those seen by ISH in 92/102 (90.2%) specimens. epithelial reactivity signi®cantly, indicating that the Importantly, 12/15 cases with mRNA composite scores strong signal in the bone was unlikely to be caused of 0 ± 4 also had PSCA protein expression scores of 0 ± by the decalci®cation process (data not shown). These 4. However, in 21 cases with mRNA expression scores results demonstrate that PSCA is expressed strongly 6 ± 9 there was less or absent protein expression (i.e. and may be upregulated in prostate cancer bone scores 0 ± 4), suggesting that PSCA may be modi®ed metastases. posttranscriptionally or that the epitopes recognized by the monoclonal antibodies may be obscured in some tumors. Discussion

This study describes the ®rst characterization of PSCA PSCA protein is expressed strongly in prostate cancers protein expression using ®ve monoclonal antibodies metastatic to bone directed against PSCA. PSCA mAbs stain the cell Prostate cancer is unique among human tumors in its surface in a punctate manner, suggesting that PSCA propensity to metastasize preferentially to bone and may be localized to speci®c regions of the cell surface. to induce osteoblastic responses. Nine sections of GPI-anchored proteins are known to cluster in prostate cancer bone metastases were examined detergent-insoluble glycolipid-enriched microdomains immunohistochemically (Figure 5). All reacted in- (DIGS) of the cell surface (Varma and Mayor, 1998). tensely and uniformly (i.e. composite score of nine) These microdomains, which include caveolae and with mAb 1G8. In two instances, micrometastases sphingolipid-cholesterol rafts, are believed to play not readily detectable on hematoxylin and eosin critical roles in signal transduction and molecular sections could be seen after staining. In three cases, transport (Anderson, 1993; Friedrichson and Kurzcha- biopsy specimens from the primary tumors were lia, 1998; Hoessli and Robinson, 1998). Thy-1, a available for comparison. All were weakly positive homologue of PSCA, has previously been shown to for PSCA when compared with the matched bone transmit signals to src kinases through interaction in metastases, suggesting that PSCA expression was lipid-microdomains (Stefanova et al., 1991; Thomas increased in bone. To rule out the possibility that and Samuelson, 1992; Xavier et al., 1998). Additional the strong staining in bone was caused by the studies will be needed to determine whether PSCA decalci®cation process used to prepare bone sections, localizes preferentially in caveolae and/or cholesterol the three primary biopsy specimens were also treated rafts and whether the mAbs described herein can

Figure 5 PSCA immunostaining in prostate cancer bone metastases. Hematoxylin and eosin (a,c) and PSCA (b,d) staining of bony lesions from two patients with metastatic prostate cancer. In one patient (a and b), a single focus suspicious for metastasis (arrow) was identi®ed in the H and E section and con®rmed by intense staining with anti-PSCA monoclonal antibodies (arrow). The second patient (c and d) has a large focus of cancer on H and E (arrow), which stained uniformly and intensely for PSCA (arrow)

Oncogene Prostate stem cell antigen expression ZGuet al 1294 transmit signals through PSCA analogous to those seen study can recognize PSCA protein in LAPC-4 unless with Thy-1. the tumor is deglycosylated prior to immunoblotting Although PSCA mRNA localizes exclusively to basal (R Reiter, unpublished result) (Reiter et al., 1998). cells, the current results suggest that PSCA protein The cause(s) of PSCA overexpression is not known. may be present in both basal and secretory cells. PSCA maps distal to c-myc on chromosome 8q24.2. Similar di€erences between mRNA and protein Ampli®cation of c-myc correlates with locally ad- localization in prostate have been described for p27, vanced, cribriform, metastatic and androgen-indepen- PSMA and androgen receptor (Cordon-Cardo et al., dent cancers, similar to PSCA expression (Jenkins et 1998; Kawakami and Nakayama, 1997; Magi-Galluzzi al., 1997; Kokontis et al., 1994; Qian et al., 1997). C- and Loda, 1996; Magi-Galuzzi et al., 1997). One myc ampli®cation also correlates strongly with de- possible explanation for the presence of PSCA protein creased survival from prostate cancer (Sato et al., in secretory cells is that PSCA mRNA is transcribed in 1999). We have analysed a small number of locally basal progenitor cells, but that PSCA protein expres- advanced tumors with c-myc ampli®cation and found sion persists as basal cells di€erentiate into secretory that PSCA is co-ampli®ed with c-myc (Reiter et al., cells. Another possibility is that PSCA protein may be 2000). PSCA expression in these tumors was higher transferred from basal to secretory cells posttransala- than in adjacent normal glands. PSCA overexpression tionally. GPI-anchored proteins can incorporate into may be a potential cell surface marker of aggressive the plasma membranes of adjacent cells via their tumors with c-myc ampli®cation. glycolipid anchors (Arienti et al., 1997). Spermatozoa, One of the most intriguing results of the present for example, acquire GPI-anchored proteins from study was the consistent, intense staining seen in the epithelial cells lining the epididymis (Arienti et al., nine prostate cancer bone metastates. LAPC-9, a 1997). Finally, false-positive staining of secretory cells xenograft established from a bony metastasis, also cannot be absolutely excluded. Additional studies will stained intensely for PSCA. It is not known whether be necessary to understand fully the expression and PSCA expression is up-regulated in bone, whether cells natural history of PSCA in normal prostate epithelium. expressing high levels of PSCA may localize preferen- Although largely prostate-speci®c in men, PSCA is tially to bone, or whether PSCA expression simply also expressed in urothelium, colonic and gastric re¯ects the advanced nature of these tumors and/or the neuroendocrine cells and renal collecting ducts. The presence of PSCA ampli®cation (Alers et al., primary impetus for identifying prostate-speci®c cell 1997). In three patients, matched primary biopsy surface genes is the desire to develop more selective, specimens showed low levels of PSCA expression less toxic therapies. Expression of PSCA in advanced compared to the bony metastases, suggesting that prostate tumors appears signi®cantly stronger than the PSCA expression may be upregulated. However, areas staining seen in extraprostatic sites. Therapies directed of strong PSCA expression in the primary tumors of against PSCA may therefore be relatively selective for the three patients examined may have been missed cancer, much as Her-2/neu antibodies primarily target since only biopsies were available for analysis. Also, in breast cancers which overexpress Her-2/neu (Disis and two cases the primary tumor was sampled at least a Cheever, 1997). Preclinical and clinical trials will be decade prior to the bone metastasis, raising the necessary to assess potential toxicity related to possibility that clones expressing high levels of PSCA extraprostatic expression. within the primary could have developed subsequent to The present study supports our earlier observation the initial biopsy. Nevertheless, these results clearly that PSCA mRNA is overexpressed in a subset of demonstrate strong PSCA expression in bone metas- prostate cancers (Reiter et al., 1998). A new ®nding in tases, further supporting it as a potentially novel the present study was the correlation of PSCA protein diagnostic or therapeutic target for advanced disease. expression with tumor stage, grade and androgen- independence. Similar results were obtained regardless of whether we assessed absolute levels of PSCA Materials and methods expression semi-quantitatively (i.e. percentage of posi- tive cells and staining intensity) or compared expres- Generation and production of monoclonal antibodies sion in tumor and normal glands. The association of PSCA overexpression with tumor stage and grade BALB/c mice were immunized three times with a puri®ed PSCA-glutathione S-transferase (GST) fusion protein con- raises the possibility that PSCA may have diagnostic taining PSCA amino acids 21 ± 99. Spleen cells were fused utility in prostate cancer. It will be particularly with HL-1 myeloma cells using standard hybridoma interesting to see if PSCA protein expression correlates technique. Hybridomas that were positive for PSCA by with or can predict clinical outcome. Follow-up for our ELISA and FACS analysis (see Results) were subcloned. cohort was too short (i.e. mean follow-up *2 years Ascites ¯uid was produced in CB 17 scid/scid mice and with only two deaths) to provide meaningful survival monoclonal antibodies (mAbs) puri®ed using a protein G ®gures. anity column (Pharmacia Biotech, Piscataway, NJ, USA). In a number of instances, there was absent or weak PSCA mAb 1G8 was also produced in Cell-Pharm System immunostaining despite strong mRNA expression by in 100 as recommended by the manufacturer (Unisyn Technol- situ hybridization. Although it is possible that this may ogies, Hopkinton, MA, USA). re¯ect posttranscriptional modi®cation of PSCA, it is likely that the epitopes recognized by the mAbs are ELISA for hybridoma screening hidden in some cancers by aberrant glycosylation. For GST or PSCA-GST were immobilized on Reacti-Bind maleic example, we previously reported that the LAPC-4 anhydride-activated polystyrene plates (Pierce, Rockford, IL, xenograft tumor expresses large amounts of PSCA USA). Fifty ml of hybridoma media were added to each well mRNA; however, none of the mAbs described in this and incubated for 1 h at room temperature. Wells were

Oncogene Prostate stem cell antigen expression ZGuet al 1295 washed three times with 200 ml PBS containing 0.1% BSA swine IgG, all biotin conjugated. Slides were then incubated and 0.05% Tween 20 and incubated for 1 h with 100 ml anti- with strepavidin-peroxidase and antibody localization per- mouse IgG (1 : 4000) labeled with alkaline phosphatase formed using the diaminobenzidene reaction. All slides were (Promega, Madison, WI, USA). Plates were developed with stained with secondary antibody alone as a negative control. an alkaline phosphatase substrate (Bio-Rad, Hercules, CA, Speci®city of staining was also tested by incubating PSCA USA). mAb with excess PSCA GST-fusion protein prior to staining. Specimens obtained from Beth-Israel-Deaconess Medical Center were stained as previously described using an Cell culture automated Ventana NexES instrument (Ventana Medical LNCaP was obtained from ATCC and stably transfected Systems, Tucson, AZ, USA) (Magi-Galluzzi et al., 1997). with a pCDNA II (Invitrogen) expression vector containing Antigen retrieval was done by microwave for 15 min in PSCA or vector alone (Reiter et al., 1998). 293T cells EDTA, pH 8.05 at 750 W. mAbs puri®ed at a concentration transiently transfected with PSCA or vector alone were of *1 mg/ml of SCID ascites were used at the following prepared as described previously (Reiter et al., 1998). LAPC- concentrations: 1G8=1 : 20; 3E6=1 : 30; 2H9=1 : 50; 9 xenograft explants were propagated in PrEGM media 4A10=1 : 100; 3C5=1 : 100. Endogenous biotin activity was (Clonetics, San Diego, CA, USA) after digestion in 1% blocked using Ventana AB block. mAb 1G8 was produced in pronase for 18 min at room temperature (Whang et al., CellPharm System 100 and used at concentrations of 1 : 10 or 1998). Before FACS analysis, LAPC-9 cells were passed 1 : 20 depending on mAb titre. Positive controls included through a 40 mm cell strainer to obtain single cell suspen- LAPC-9 and LNCaP-PSCA and negative controls were sions. LNCaP and isotype matched irrelevant antibody. Primary biopsy specimens were available for three patients with bone metastases. To approximate conditions of decalci®cation, Immunofluorescence slides from these specimens were treated for 20 min in Decal- Cells were grown on glass coverslips coated with poly-L- Stat (Lengers, NY, USA) prior to staining with PSCA mAbs. lysine. Immuno¯uorescence assays were carried out on permeabilized and nonpermeabilized ®xed cells. For ®xation, Analysis of immunohistochemical staining cells were treated with 2% paraformaldehyde in PBS-CM (PBS, 100 mM CaCl2,1mM MgCl2) for 30 min in the dark, Slides of primary prostate tumors were read and scored by quenched with 50 mM NH4Cl in PBS-CM-BSA (PBS, 100 mM two pathologists (GV Thomas and M Loda) in a blinded CaCl2,1mM MgCl2, 0.2% BSA) for 10 min, and washed fashion. There was greater than 90% inter- and intra- twice with PBS-CM-BSA. For permeabilization, cells were observer agreement. Scoring was done by two independent treated additionally with PBS-CM-BSA-Saponin (0.075% methods. The ®rst method of scoring quanti®ed the level of saponin (Sigma) in PBS-CM-BSA) for 15 min at room PSCA expression in tumors. Scores of 0 ± 3 were assigned temperature. Primary mAb at 2 ± 5 mg/ml in PBS-CM-BSA according to the percentage of positive tumor cells (0=0%; (plus saponin in cases of permeabilization) was added for 1=525%; 2=25 ± 50%; 3=450%) and the intensity of 60 min and washed twice with PBS-CM-BSA. FITC-con- staining in tumor (0=0; 1=1+; 2=2+; 3=3+). The two jugated goat antimouse IgG antibody (1 : 500 diluted in PBS- scores were multiplied to give an overall score of 0 ± 9, of CM-BSA +/7 saponin; Southern Biotechnology, Birming- which 0 was considered negative, 1 ± 2 was considered weak, ham, AL, USA) was added for 30 min and washed three 3 ± 6 moderate and nine strong staining (Epstein et al., 1995; times with PBS-CM. Slides were mounted in vectashield Hanas et al., 1999; Sauvageot and Epstein, 1998). The (Vector Laboratory, Inc., Burlingame, CA, USA). mRNA in situ hybridizations reported previously were recorded in the same manner (Reiter et al., 1998). The second method of scoring examined the intensity of Flow cytometry staining of tumor glands compared to adjacent normal Cells (16106) were incubated for 30 min at 48C with 100 ml glands. Here, the intensity of staining in greater than 50% mAb at 2 mg/ml in PBS containing 2% fetal bovine serum or of the tumor was compared to the intensity of staining in hybridoma conditioned medium. After washing, cells were adjacent normal glands on the same slide. Staining in tumor stained with a 1 : 500 dilution of FITC-conjugated goat was scored as being greater, equal or less than in adjacent antimouse IgG (Southern Biotechnology, Birmingham, AL, normal glands. Staining which was greater in tumor than USA). Data was acquired on a FACScan (Becton Dickinson) normal was taken as evidence of PSCA overexpression. and analysed by using LYSIS II software. Statistics Immunohistochemistry Associations between PSCA expression and Gleason score, Normal formalin-®xed, paran-embedded tissue samples androgen dependence and pathologic stage were tested using were obtained from the Departments of Pathology at Beth- the Kendall Tau Correlation method or Fisher's Exact Test Israel-Deaconess Medical Center and UCLA. Primary radical as appropriate. and transurethral prostatectomy specimens were selected from a previously described database (Magi-Galluzzi et al., 1997). Bone metastases and matched primary biopsy speci- mens were obtained from the UCLA Department of Acknowledgments Pathology. Normal tissues were stained and scored indepen- We thank Igor Vivanco and Connie Lin for technical dently at two institutions in order to ensure reproducibility. assistance and Charles Sawyers and Doug Sa€ran for Specimens obtained from UCLA were stained using mod- critical reading of the manuscript. We also thank Jean B i®cations of an immunoperoxidase technique previously deKernion for his support. This work was supported by described (Said et al., 1998). Antigen retrieval was performed grants from CaPCURE (RE Reiter, ON Witte and M on paran sections using a commercial steamer and 0.01 M Loda), National Institutes of Health Grant K08 CA74169 citrate bu€er pH 6.0. After incubation with PSCA mAbs for (RE Reiter), the STOP Cancer Foundation (RE Reiter) 50 min (see below), slides were treated sequentially with and the Cancer Research Institute (RE Reiter and ON rabbit anti-mouse IgG, swine anti-rabbit IgG and rabbit anti- Witte).

Oncogene Prostate stem cell antigen expression ZGuet al 1296 References

Alers JC, Krijtenburg PJ, Rosenberg C, Hop WC, Verkerk Magi-Galluzi C, Mishra R, Fiorentino M, Montironi R, Yao AM,SchroderFH,vanderKwastTH,BosmanFTand H, Capodieci P, Wishnow K, Kaplan I, Stork PJS and van Dekken H. (1997). Lab. Invest., 77, 437 ± 448. Loda M. (1997). Lab. Invest., 76, 37 ± 43. Anderson R. (1993). Proc. Natl. Acad. Sci. USA, 90, 10909 ± Magi-Galuzzi C, Xu X, Hlatky L, Hahnfeldt P, Kaplan I, 10913. Hsiao P-w, Chang C and Loda M. (1997). Mod. Pathol., Antica M, Wu L and Scollay R. (1997). Immunol. Letts., 55, 10, 839. 47 ± 51. McNealJE,ReeseJH,RedwineEA,FreihaFSandStamey Arienti G, Carlini E, Verdacchi R, Cosmi EV and Palmerini TA. (1986). Cancer, 58, 1714 ± 1719. CA. (1997). Biochim. Biophys. Acta, 1336, 533 ± 538. Murphy GP, Elgamal AA, Su SL, Bostwick DG and Holmes Cher ML, MacGrogan D, Bookstein R, Brown JA, Jenkins EH. (1998a). Cancer, 83, 2259 ± 2269. RB and Jensen RH. (1994). Genes Chromo. Cancer, 11, Murphy GP, Greene TG, Tino WT, Boynton AL and 153 ± 162. Holmes EH. (1998b). J. Urol., 160, 2396 ± 2401. Classon BJ and Coverdale L. (1994). Proc. Natl. Acad. Sci. O'Malley FP, Grignon DJ and Shum DT. (1990). Virchows USA, 91, 5296 ± 5300. Arch. A. Pathol. Anat. Histopathol., 417, 191 ± 196. Cordon-Cardo C, Ko€ A, Drobnjak M, Capodieci P, Osman Qian J, Jenkins RB and Bostwick DG. (1997). Mod. Pathol., I, Millard SS, Gaudin PB, Fazzari M, Zhang ZF, 10, 1113 ± 1119. Massague J and Scher HI. (1998). J. Natl. Cancer Inst., Reiter R, Gu Z, Watabe T, Thomas G, Szigeti K, Davis E, 90, 1284 ± 1291. Wahl M, Nisitani S, Yamashiro J, LeBeau M, Loda M and Disis ML and Cheever MA. (1997). Adv. Cancer Res., 71, Witte O. (1998). Proc. Natl. Acad. Sci. USA, 95, 1735 ± 343 ± 371. 1740. Epstein JL, Carmichael M and Partin AW. (1995). Urology, ReiterRE,SatoI,ThomasG,QianJ,WatabeT,LodaM 45, 81 ± 86. and Jenkins RB. (2000). Genes Chrom. Cancer, 27, 95 ± Fair WR, Israeli RS and Heston WD. (1997). Prostate, 32, 100. 140 ± 148. Rubin MA, de La Taille A, Bagiella E, Olsson CA and Friedrichson T and Kurzchalia TV. (1998). Nature, 394, O'Toole KM. (1998). Am. J. Surg. Pathol., 22, 840 ± 848. 802 ± 805. Said JW, Pinkus JL, Shintaku IP, deVos S, Matsumura F, Hanas JS, Lerner MN, Lightfoot SA, Raczkowski C, Yamashiro S and Pinkus GS. (1998). Mod. Pathol., 11, Kastens DJ, Brackett DJ and Postier RG. (1999). Cancer, 1±5. 86, 756 ± 763. Sato K, Qian J, Lieber MM, Slezak JM, Bergstralh EJ and Hoessli DC and Robinson PJ. (1998). Trends Cell Biol., 8, Jenkins RB. (1999). J. Urol., 161, 60A. 87 ± 89. Sauvageot J and Epstein JI. (1998). Prostate, 34, 29 ± 33. Jenkins RB, Qian J, Lieber MM and Bostwick DG. (1997). Silver DA, Pellicer I, Fair WR, Heston WDW and Cordon- Cancer Res., 57, 524 ± 531. Cardo C. (1997). Clin. Cancer Res., 3, 81 ± 85. Kawakami M and Nakayama J. (1997). Cancer Res., 57, Stefanova I, Horejsi V, Ansotegui IJ, Knapp W and 2321 ± 2324. Stockinger H. (1991). Science, 254, 1016 ± 1019. Kokontis J, Takakura K, Hay N and Liao S. (1994). Cancer Thomas PM and Samuelson LE. (1992). J. Biol. Chem., 267, Res., 54, 1566 ± 1573. 12317 ± 12322. Lalani E-N, Laniado ME and Abel PD. (1997). Cancer Varma R and Mayor S. (1998). Nature, 394, 798 ± 801. Metas. Rev., 16, 29 ± 66. Whang YE, Wu X, Suzuki H, Reiter RE, Tran C, Vessella Liu H, Moy P, Kim S, Xia Y, Rajasekaran A, Navarro V, RL, Said JW, Isaacs WB and Sawyers CL. (1998). Proc. Knudsen B and Bander NH. (1997). Cancer Res., 57, Natl. Acad. Sci. USA, 95, 5246 ± 5250. 3629 ± 3634. Xavier R, Brennan T, Li Q, McCormack C and Seed B. Magi-Galluzzi C and Loda M. (1996). Eur. Urol., 30, 167 ± (1998). Immunity, 8, 723 ± 732. 176.

Oncogene