Decreased expression of keratinocyte growth factor receptor ... - Nature

Oncogene (1997) 14, 323 ± 330  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Decreased expression of keratinocyte growth factor receptor in a subset of human transitional cell bladder carcinomas Sixtina Gil Diez de Medina1,2, Dominique Chopin2, Ahmed El Marjou1, Annie DelouveÂe1, William J LaRochelle3, AndraÁs Hoznek2, Claude Abbou2, Stuart A Aaronson4, Jean Paul Thiery1 and FrancËois Radvanyi1 1

UMR 144, CNRS/Institut Curie, 26 rue d'Ulm 75248 Paris Cedex 05, France; 2Groupe d'Etude des Tumeurs Urologiques, CRCHM et Service d'Urologie CHU Henri Mondor, 94010 CreÂteil Cedex, France; 3National Cancer Institute, National Institute of Health, Bethesda, Maryland 20892, USA: 4Ruttenberg Cancer Center, Mount Sinai School of Medicine, 1 Gustav L. Levi Place, PO Box 11309, New York, NY 10029, USA

Growth factors and growth factor receptors are involved in tumor progression. The ®broblast growth factor receptor 2 gene encodes distinct isoforms. The isoforms which bind KGF (keratinocyte growth factor or FGF-7) are called KGF-R or FGFR2b. KGF-R is expressed in dierent epithelia and is involved in the control of epithelial-mesenchymal interactions. Expression of KGFR mRNA was examined in normal human bladder and transitional cell carcinoma of the bladder (TCC) by semi-quantitative RT ± PCR using TFIID and GAPDH as internal standards. In normal bladder, the KGF-R mRNA was detected in the urothelium but not in the underlying stroma. In TCCs, the level of KGF-R mRNA was generally either normal or low. Eighteen out of 54 TCCs had a KGF-R mRNA level below 30% of that found in normal urothelium. This decrease in KGF-R mRNA was not accompanied by an increase in BEK (FGFR2c) mRNA, the other major splice variant of the ®broblast growth factor receptor 2 gene. Expression of the KGF-R was also monitored by immunohistochemistry using a functional KGF-immunoglobulin chimera. The receptor was uniformly expressed throughout the normal urothelium except for the umbrella cells. Immunoreactivity for KGF-R was found to be negative in tumors with low levels of KGF-R mRNA, while the peritumoral normal urothelium was positive. Among patients with muscle invasive tumors, those exhibiting a low level of KGF-R mRNA had a signi®cantly higher proportion of cancer deaths. Our results suggest that decreased expression of KGF-R can be considered as a marker of tumor progression in muscle invasive TCCs. Keywords: bladder; ®broblast growth factor receptor; human; transitional cell carcinoma; tumor progression

Introduction It is dicult to predict the course of bladder cancer, because tumors with comparable clinicopathological characteristics frequently give rise to dierent outcomes. Epidemiological data support the idea that there are two major modes of presentation for TCC. A ®rst tumor group, de®ned as super®cial TCC, includes

Correspondence: F Radvanyi Received 5 July 1996; revised 12 September 1996; accepted 13 September 1996

Ta and T1 tumors of the TNM classi®cation (Spiessl et al., 1992). These tumors are treated in the ®rst instance by a conservative approach using endoscopy with or without intravesical therapy. The second group includes all muscle invasive tumors which are ®rst treated by radical therapy in the absence of detectable metastasis. There is a need for new biological markers as predictors of clinical evolution in order to tailor therapeutic aggressiveness for a given patient. In addition, understanding molecular events linked to tumor progression may provide a rationale for new therapeutic approaches. A large body of data support the idea that growth factors and their cognate receptors may be implicated at various stages during tumor progression. In the normal tissue, beside their role in proliferation, growth factors can also act as scatter factors, motogens and morphogens, and promote or maintain dierentiation. It is thus not surprising that growth factors and their cognate receptors can be both positive and negative regulators of tumor progression (Sporn and Roberts, 1988; Aaronson, 1991; Cross and Dexter, 1991). For example in breast cancer, overexpression of ERBB2 (also called neu) and in bladder cancer overexpression of ERBB1 (the epidermal growth factor receptor) have been correlated with a poor prognosis (Hynes and Stern, 1994; Mellon et al., 1995). By contrast, in neuroblastoma the loss of a receptor, trkA, one of the nerve growth factor receptors, has been shown to be associated with a poor outcome (Nakagawara et al., 1994). Loss of the TGFb receptors has been observed in various carcinomas as well (Markowitz et al., 1995). Much work has been devoted to the role of the ®broblast growth factor family and their receptors during development. To date, the FGF (®broblast growth factor) family consists of nine members (Basilico and Moscatelli, 1992; Tanaka et al., 1992; Miyamoto et al., 1993; Lorenzi et al., 1995). FGFs act via high-anity receptors. Four dierent FGF receptor genes have been identi®ed and designated FGFR1, FGFR2, FGFR3 and FGFR4 (Johnson and Williams, 1993). Splice variants of each gene can generate many dierent isoforms (Eiseman et al., 1991; Givol and Yayon, 1992). Splice variants of the FGFR2 gene can be divided into two groups depending on their anity for KGF (FGF-7): the KGF-R or FGFR2b has a high anity for KGF and possesses exon IIIb (Miki et al., 1991, 1992; Dell and Williams, 1992); BEK or FGFR2c has no detectable anity for KGF and contains the alternative IIIc exon. The KGF-R has

Decreased expression of KGF-R in a subset of TCCs S Gil Diez de Medina et al

324

been found only in epithelia, whereas its ligand, KGF, has been found only in stroma (Finch et al., 1989; Peters et al., 1992; Orr-Urtreger et al., 1993; LaRochelle et al., 1995; Finch et al., 1995). The KGF-R/KGF loop, which has been implicated in the growth and dierentiation of various epithelia, could play a role in tumor progression. In the prostate, a genitourinary epithelium, a switch from KGF-R (FGFR2b) to BEK (FGFR2c) has been found to be associated with stromal independence and tumor progression in a rat tumor model (Yan et al., 1993). Thus, we examined the expression of KGF-R in normal human urothelium and a variety of TCCs. As we observed a striking decrease in KGF-R expression in a signi®cant number of carcinomas, we have also examined if this decrease in expression was associated with an increase in the expression of the other splicing form of FGFR2, BEK and if it was correlated with a worse prognosis for such tumors.

respectively of the KGF-R mRNA level found in normal urothelium by RT ± PCR analysis, had undetectable levels of KGF-R mRNA by Northern blot analysis; samples #2 and 3 had normal levels of KGF-R mRNA using both techniques, whereas sample #6, which had a 4.5-fold increase in KGF-R mRNA by RT ± PCR, showed a similar increase in the Northern blot analysis. In order to determine whether the expression of the transcript correlated with the presence of the KGF-R

Results KGF-R expression in normal human urothelium The KGF-R has been reported to be expressed in several human epithelial tissues (LaRochelle et al., 1995). We investigated its presence in normal urothelium by means of semi-quantitative RT ± PCR. The urothelium, underlying connective tissue and muscle layer were separated surgically. Two dierent markers, cytokeratin 18 for the epithelium and desmin for muscle were used for con®rmation. As expected, cytokeratin 18 was found mainly in the urothelium, whereas desmin was found in connective tissue and more abundantly in muscle (Figure 1b and c). KGF-R mRNA was detected almost exclusively in the epithelial compartment (Figure 1a). It should be noted that the levels of KGF-R mRNA in the dierent urothelia were very similar (average value=1.88, s.d. 0.49). Low KGF-R expression in a subset of TCC We next analysed the level of KGF-R mRNA in TCCs (Figure 2). All the TCC samples used in this analysis were primarily composed of tumor cells: samples adjacent to the pieces used for RNA preparation were analysed histologically using hematoxylin/eosin staining and were found to contain more than 80% tumor cells. Among 54 dierent tumors analysed, only two had a level of KGF-R mRNA which was more than twice the level found in normal urothelium. The level of KGF-R mRNA in the other tumors was either normal or low. One third of the tumors contained levels of KGF-R mRNA less than 30% of the mean value found in normal urothelium. The results were similar when another internal standard (GAPDH) was used instead of TFIID (data not shown). One normal urothelium and ®ve tumors were compared for KGF-R mRNA by Northern blot analysis (Figure 3). A 4.5 kb mRNA identical to that found in normal urothelium was detected in several tumors. Quantitatively, the results obtained by Northern blot analysis were identical to those obtained using RT ± PCR. Samples #4 and 5, which had 1 and 7%

Figure 1 Expression of KGF-R, cytokeratin 18 and desmin mRNA in the normal bladder. Levels of the mRNAs for KGF-R, cytokeratin 18 (CK-18) and desmin were determined in normal urothelium, underlying connective tissue and smooth muscle by semi-quantitative RT ± PCR using TFIID as an internal standard. (a) KGF-R is expressed exclusively in the urothelium. (b) Cytokeratin 18 mRNA is expressed predominantly in the urothelium; detectable levels were measured in the underlying stroma. (c) Desmin mRNA is found in connective tissue and in the smooth muscle layers


protein, we applied an immunohistochemical method based on the speci®c binding of the KGF protein coupled to the Fc portion of an immunoglobulin

heavy chain (LaRochelle et al., 1995). As shown in Figure 4a, the expression of the KGF-R was detected at the cell surface of all the layers of the urothelium with a noticeable exception of the umbrella cells lining the lumen of the bladder. No immunostaining could be detected in the chorion or in the smooth muscle layers (Figure 4a). Controls immunostained with the Fc portion alone did not show any reactivity (data not shown). Twenty-®ve of the 54 tumors were available for immunohistochemical analysis of KGF-R protein. Figure 4b shows a stage Ta tumor which expressed the KGF-R uniformly with the exception of a few dierentiated luminal cells lacking any detectable immunoreactivity. A stage T3 tumor, which expressed a low level of KGF-R mRNA, was found to be negative for KGF-R immunodetection (Figure 4c). In contrast, the peritumoral urothelium from the same T3 tumor was positive (Figure 4d). Immunoreactivity for KGF-R was found to be

1

Figure 2 KGF-R mRNA in normal urothelium and in TCC. The level of KGF-R mRNA was measured in normal urothelia and in 54 dierent TCC. The analysis was performed by means of semi-quantitative RT ± PCR using TFIID as an internal standard. One third of the TCCs had a level of KGF-R mRNA below 30% of that found in normal urothelium

2

3

4

5

6

KGF-R ➝

— 4.5 kb

GAPDH ➝

— 1.4 kb

Figure 3 RNA blot analysis of KGF-R. KGF-R expression in normal urothelium (lane 1) and in several TCC (lanes 2 to 6) was determined by Northern blot using a riboprobe speci®c for the IIIb exon. As a control, the same blot was hybridized with a human GAPDH probe. The size of KGF-R mRNA was determined by comparison with RNA markers and ribosomal RNA (data not shown)

Figure 4 Immunohistochemical detection of KGF-R. KGF-R expression was probed with KGF-HFc. (a) Normal urothelium. KGF-R is expressed on all urothelial cells except on the umbrella cells which line the luminal surface. No detectable immunoreactivity is seen in the chorion and the smooth muscle wall. (b) A stage Ta tumor. KGF-R is expressed at the surface of the tumor cells with the exception of a few dierentiated luminal cells. (c) A stage T3 tumor. KGF-R was not detected in the tumor but was expressed in the peritumoral urothelium (d)

325


326

negative in tumors with low levels of KGF-R mRNA. In the tumors having a normal or high level of KGFR mRNA, there was a good correlation between mRNA and protein expression, although some tumors considered positive by RT ± PCR exhibited nonuniform staining (data not shown). Low level of BEK in urothelium and TCC Decreased expression of the KGF-R (FGFR2b) in carcinoma cells has been reported to be associated with increased expression of BEK (FGFR2c) (Yan et al., 1993; Savagner et al., 1994), an alternative splice variant of the same gene. Thus, we analysed the level of BEK mRNA in six TCCs expressing normal or high levels of KGF-R (group I TCC) and in six TCCs expressing low levels of KGF-R (group II TCC) (Figure 5a). BEK mRNA was present in the smooth muscle compartment, and was not present in normal urothelium (Figure 5b). It was not present in any of the TCCs, regardless of their KGF-R expression.

Correlation between survival and KGF-R mRNA levels in TCC The patients were subdivided into three groups of equal size according to the level of expression of KGF-R mRNA found in the tumors (18 patients per group) (see Materials and methods). There were no dierences with respect to survival between patients with high or intermediate levels of KGF-R (data not shown). By contrast, there was a signi®cant dierence between the group of patients with low levels of KGFR mRNA and the two other groups of patients (Figure 6a and b). Figure 6a shows that survival was signi®cantly lower among patients with a low level of KGF-R (P=0.013). When only patients with muscle invasive disease were considered, the dierence was still highly signi®cant (P=0.0095) (Figure 6b). In contrast,

a

b

Figure 5 Comparative expression of BEK in normal tissues and in TCC subdivided into two groups according to their levels of KGF-R mRNA. Levels of KGF-R (a) and BEK mRNAs (b) were determined in four dierent normal urothelia, two dierent smooth muscle samples and 12 dierent TCCs by semiquantitative RT ± PCR using TFIID as an internal standard. In group I TCC, the level of KGF-R mRNA was normal (®ve cases) or high (one case); in group II TCC, the level of KGF-R was low (six cases)

Figure 6 Kaplan and Meier survival curves for patients with TCC (a) or muscle invasive TCC (b). Patients were subdivided according to the KGF-R mRNA expression found in the tumors (see Materials and methods). Patients with weakly KGF-Rexpressing tumors (dashed lines) were compared to the other patients (plain lines). Tick marks indicate last follow-up evaluation. When all patients were considered median follow-up was 39 months, it was 32.5 months when only patients with muscle invasive TCC were considered


when only patients with super®cial disease were considered, there was no signi®cative dierence between those patients with low KGF-R levels and the others (data not shown). Although 40% of the G3 tumors and only 25% of the G1 and G2 tumors had a low level of KGF-R mRNA, there was no signi®cant correlation between KGF-R mRNA levels and grade or stage. Discussion Numerous studies have attempted to decipher the role of multifunctional growth factors and their receptors in tumor progression. Particularly striking is the case of KGF, a growth factor expressed by mesenchyme, which has been implicated as a mitogen, a morphogen and a dierentiation inducing agent in a wide variety of epithelia. We reasoned that any disturbance in the KGF/KGF-R paracrine loop could alter the behavior of malignant epithelia. It has been shown that this paracrine loop plays an important role in the morphogenesis and dierentiation of normal epithelia (Alarid et al., 1994; Peters et al., 1994; Werner et al., 1993). Interestingly, a non-malignant rat prostate tumor was able to dedierentiate and become malignant when transplanted into a syngenic host in the absence of stromal cells. The transition to the malignant state was accompanied by a switch from the KGF-R (FGFR2b) to the BEK (FGFR2c) isoform (Yan et al., 1993). In this study, we analysed the expression of the KGF receptor mRNA in normal urothelium and in TCC. KGF-R mRNA was expressed in the normal human urothelium, while little if any KGF-R mRNA was detected in connective tissue or smooth muscle. The identity of each compartment was con®rmed by using two markers: cytokeratin 18 for the epithelial compartment and desmin for the muscle compartment. We found consistently low but measurable amounts of cytokeratin 18 mRNA in the smooth muscle compartment (about 5% of the level found in the urothelium). Weak expression of cytokeratin 18 has already been reported in non-epithelial cells (Bader et al., 1988). Thus, urothelium is another example of a human epithelium which expresses the KGF-R (LaRochelle et al., 1995). Compared to normal urothelium, the KGF-R/ TFIID ratio was either normal or low in the transitional cell carcinomas. One third of the tumor samples had a KGF-R/TFIID level below 30% of the value found in normal urothelium. As the level of TFIID might have diered among the dierent samples, a fall in the KGF-R/TFIID ratio could also partly have been due to an increase in TFIID. This was very unlikely however, as the same tumors had low levels of KGF-R mRNA whether TFIID or GAPDH was used as an internal standard. Furthermore, when the level of KGF-R mRNA was determined in several tumors by Northern blotting, the results were similar to those found by RT ± PCR. Using a KGF-HFc chimera, which has been shown by means of immunohistochemistry to detect the KGF-R (La Rochelle et al., 1995), the normal bladder urothelium stained positively; moreover, there was a good correlation between the levels of KGF-R mRNA and KGF-R protein when the KGF-HFc immunostaining technique was applied to the tumor samples.

Several splice variants can be obtained from the FGFR2 gene and can be separated into two groups: one possessing the IIIb exon and having a high anity for KGF, and one containing the IIIc exon and having no anity for KGF. In our series of transitional cell carcinomas, 30% weakly expressed the IIIb exon. This could have been due to a switch from the splice forms expressing the IIIb exon to the forms expressing the IIIc exon, as observed in several tumor models (Yan et al., 1993; Savagner et al., 1994). Alternatively, it might have been due to decreased expression of the FGFR2 gene. In fact, regardless of IIIb exon expression, forms containing the IIIc exon were low or undetectable. Thus, the low level of forms containing the IIIb exon was due to a decreased FGFR2 gene expression. We found that low levels of KGF-R in TCC correlated with a poor prognosis, as judged by urothelial cancer death. This result could be interpreted in either of two ways. The KGF-R may simply be a dierentiation marker which is lost in the most aggressive bladder tumors, or it may play an active role in maintaining epithelial cells in a less aggressive phenotype. In line with this second possibility, Werner et al. (1993) suggested that FGF-Rs maintain the normal program of keratinocyte dierentiation in the suprabasal layer of the skin. If the KGF-R were to play the same role in the urothelium, then the FGFR2 gene could be considered as a tumor-suppressor gene in bladder carcinomas. The FGFR2 gene is located in 10q26 (Mattei et al., 1991; Dionne et al., 1992). Deletion of this region has been reported in 8.5% of transitional cell carcinomas of the bladder (Knowles et al., 1994). It is of interest to note that FGFR2 is a candidate tumor-suppressor gene in glioblastomas (Yamaguchi et al., 1994). In this latter situation, the alternative BEK (FGFR2c) isoform rather than KGF-R (FGFR2b), is expressed in normal tissues (astrocytes) and low-grade tumors. BEK expression is lost in the most advanced tumors (malignant astrocytomas), most of which have lost 10q25-qter (Fults and Pedone, 1993). In our series of TCCs, decreased expression of KGF-R correlated with a poor prognosis when all TCCs were taken into account. This was also true when muscle invasive TCCs were considered alone. However, there was no such correlation in the case of super®cial TCC. This would suggest that the transition from super®cial to muscle invasive carcinoma is not dependent on decreased expression of KGFR. The prognostic value of the decreased expression of KGF-R in TCC will need to be compared to that of other markers and clinico-pathological parameters in a larger series of tumors in order to establish its clinical relevance. In conclusion, the KGF-R signaling pathway can lead to mitogenesis, as is the case for basal cells of the skin (Guo et al., 1993; Werner et al., 1994), and progression of tumors, as in poorly dierentiated carcinomas of the stomach (Itoh et al., 1994; Ishii et al., 1995). Alternatively, the same receptor may be implicated in the control of cell dierentiation, as in the suprabasal cells of the skin (Werner et al., 1993), and in the maintenance of certain carcinomas in less aggressive forms (this study and Yan et al., 1993).

327


328

Materials and methods Tissue specimens Fifty-four tumors were obtained from 54 patients (11 women and 43 men) whose average age was 65.7 years (29 ± 89 years). Tumors were staged according to the TNM classi®cation (Spiessl et al., 1992) and graded according to criteria recommended by the World Health Organization (Mosto® et al., 1973). All of the tumors were TCC of the bladder. They consisted of 5 Ta, 22 T1, 8 T2, 14 T3, 4 T4 and one unknown stage; 12 were grade G1, 20 grade G2 and 22 grade G3. Twenty-seven patients had super®cial tumors and 26 had muscle invasive disease. Tissues were obtained during transurethral resection or radical cystectomy. Five normal urothelium samples were obtained from normal bladder obtained during organ procurement for transplantation on cadaveric donors. Fresh urothelial cells were scraped from the bladder wall. The lamina propria and muscle were dissected separately. A representative sample was taken from each tissue for histopathological and immunohistochemical assessment, and an adjacent piece was placed in liquid nitrogen prior to RNA extraction. RNA extraction RNA was prepared according to Chirgwin et al. (1979). Brie¯y, the frozen tissue sample was homogenized in 4 M guanidinium thiocyanate by means of a Ultraturrax T25 homogenizer (Janke & Kunkel, IKA-Labortechnik, Staufen, Germany) at maximum speed. After ultracentrifugation on a 5.7 M cesium chloride cushion, the RNA was extracted twice with phenol-chloroform. RNA was quanti®ed by spectrophotometry at 260 nm and its integrity was checked on agarose gel. RT ± PCR The level of mRNA was determined by semi-quantitative RT ± PCR, by comparison with an internal control. As internal control we used TFIID, an ubiquitous transcription factor (Peterson et al., 1990), or GAPDH (Hanauer and Mandel, 1984). cDNA was prepared from 1 mg of RNA from each sample, as described by Kandel et al. (1991). The polymerase chain reaction was performed according to Radvanyi et al. (1993), with 2 ml (1/50) of the reverse transcription product in a ®nal volume of 50 ml containing all four dNTPs (each at 100 mM), with 1 mM or 0.1 mM of each primer (a pair for the internal control TFIID or GAPDH, together with a pair speci®c for the factor analysed) and 2 mCi of a[32P]dCTP. A primer concentration of 0.1 mM was used only with the two primers used for the ampli®cation of desmin. The buer described by New England Biolabs (Beverly, MA, USA) (16.6 mM (NH4)2SO4, 67 mM Tris-HCl (pH 8.8), 6.7 mM MgCl2, 10 mM 2-mercaptoethanol, 170 mg/ml BSA and 10% glycerol) was used for the coampli®cation of GAPDH and desmin and for the coampli®cation of TFIID and KGF-R. Hi-Taq polymerase buer (Bioprobe, Montreuil sous Bois, France) was used for the other coampli®cations. The number of cycles was chosen to be in the exponential part of the two PCR reactions. Twenty-®ve cycles were used for the coampli®cation of KGF-R and TFIID, 25 cycles for BEK and TFIID, 22 cycles for CK-18 (cytokeratin 18) and TFIID, 19 cycles for desmin and GAPDH and 23 cycles for GAPDH and TFIID. The primer sequences are given: AGTGAAGAACAGT CCAGACTG and CCAGGAAATAACTCTGGCTCAT (TFIID); CTGCACCACCAACTGCTTAG and AGGTCCACCACTGACACGTT (GAPDH); GGTCAGAGACTGGAGCCATTA and GGCATTGTCCACAGTATTTGC (CK-18); CTGCAGGAGGAGATTCAGTT and GCGAGCTAGAGTAGCTGCAT (desmin); TTTCACTCTG-

CATGGTTGACA and GGTAGTCTGGGGAAGCTGTA (BEK); AGAAGTGCTGGCTCTGTTCA and CTCTTCCAGGCGCTTGCTGT (KGF-R). All primers are given in the sense 5' to 3' direction and were taken from the sequences of TFIID (Peterson et al., 1990); GAPDH (Hanauer and Mandel, 1984); CK-18 (Leube et al., 1986; Oshima et al., 1986); desmin (Li et al., 1989); BEK (Dionne et al., 1990); and KGF-R (Miki et al., 1992; Dell and Williams, 1992; Hattori et al., 1990). When the genomic sequence was known (GAPDH, CK-18, desmin, BEK, KGF-R), the two primers were chosen in two dierent exons (Ercolani et al., 1988; Kulesh and Oshima, 1989; Li et al., 1989; Johnson et al., 1991). The ampli®cation reactions were performed in a programmable thermocycler (Pharmacia LKB, Gene ATAQ-Controller, Uppsala, Sweden) with an initial cycle of 958C for 5 min prior to the addition of one unit of Hi-Taq DNA thermostable polymerase (Bioprobe). Each cycle was as follows: 948C for 1 min, 578C for 1 min and 728C for 80 s. These cycles were followed by a ®nal incubation step at 728C for 10 min. The PCR-ampli®ed products were loaded in duplicate and electrophoresed in 8% polyacrylamide gels, ®xed in 7% acetic acid and vacuumdried. Autoradiograms showed two bands corresponding to the coampli®ed fragments. The signals were quanti®ed with a Molecular Dynamics 300 PhosphorImager (Molecular Dynamics, Sunnyvale, CA, USA). Each measure was repeated in three independent PCR reactions and found to be identical within 10%. Little if any ampli®cation was observed when reverse transcriptase was omitted from the reverse transcription reaction. Northern blot A KGF-R-speci®c probe was prepared by cloning, in pGEM-3Zf+, the PCR product obtained after amplification of a normal urothelium cDNA, using the two oligos speci®c for KGF-R (sequences given above). The sequence of the clone was identical to the expected sequence (Miki et al., 1992; Dell and Williams, 1992; Hattori et al., 1990). This construct was used to prepare a 32P-labeled riboprobe. Total RNA (10 mg) from each sample was electrophoresed on a 1% formaldehyde agarose gel and transferred overnight to a Hybond N membrane (Amersham, Bucks, UK). After u.v. cross-linking, the membrane was prehybridized overnight at 658C in a solution containing 1% BSA, 0.2 M Na2HPO4/ H3PO4 pH 7.0, 45% formamide, 1 mM EDTA and 7% SDS. It was then hybridized for 16 h in the same solution with the [32P]riboprobe speci®c for KGF-R. The membrane was washed for 2 min and then 30 min at room temperature, in a solution containing 30 mM Na2HPO4/H3PO4 pH 7.0, 15% formamide, 0.1% SDS. Finally, the membrane was washed twice at 658C for 15 min in the same solution. A GAPDH probe was used as a control. This probe was prepared as described for the KGF-R probe and cloned in pGEM-3Zf+. The sequence of the clone was precisely as expected (Hanauer and Mandel, 1984). The plasmid was cut with SphI and EcoRI and ampli®ed using the two oligonucleotides speci®c for GAPDH. The ampli®ed product was puri®ed with a Microcon 30 (Amicon, Beverly, MA, USA) and quanti®ed with a spectro¯uorometer using the Hoechst 33258 reagent (Labarca and Paigen, 1980). Fifty nanograms of the puri®ed product were labeled by random priming (Random primed DNA labeling kit, Boehringer, Mannheim, Germany) and hybridized at 428C overnight with the membrane in a solution containing 50% formamide, 56 sodium saline phosphate EDTA buer (SSPE), 56 Denhardt, 0.5% sodium dodecyl sulfate (SDS) and 40 mg/ ml denatured salmon sperm DNA; the membrane was prehybridized in the same conditions for 8 h. The membrane was then washed for 2 min and 30 min, at room temperature, in a solution containing 26 SSPE and 0.1% SDS. It was


then washed at 558C for 30 min in a solution containing 16 SSPE and 0.1% SDS and ®nally at 558C for 30 min in a solution containing 0.16 SSPE and 0.1% SDS. Immunohistochemistry The construction of an expression vector for KGF-mouse immunoglobulin IgG1 heavy chain Fc domain (KGF-HFc) and the application of this ligand-immunoglobulin chimera to immunohistochemical detection of the KGF-R have been described (LaRochelle et al., 1995). Paran tissue sections (5 mm) were placed on Superfrost slides (Poly Labo, Strasbourg, France) and dewaxed. After rehydration, antigen unmasking of the paran sections was performed by placing the slides for 15 min in a solution heated at 958C containing 0.1 M citric acid buered to pH 6.0 with NaOH (Dirsch et al., in preparation). After one wash in phosphate-buered saline (PBS), the slides were placed for 30 min in PBS containing 1.5 M NaCl. After another wash in PBS the slides were placed in PBS containing 5% non-fat milk powder and 0.1% Tween 20 for 60 min (blocking step). Then the slides were incubated with KGF-HFc or control HFc in 5% milk, 0.1% Tween 20, 0.6 M NaCl, 10 mM sodium benzoate for 1 h at room temperature. After incubation, tissues were rinsed several times in PBS, 0.01% Tween 20 and further incubated with a goat anti-mouse antibody labeled with peroxidase. Immunoreactivity was visualized using 3,3 diaminobenzidine tetrachloride (Pierce Chemical Co, Rockford, IL, USA) as chromogen according to the manufacturer's protocol. Statistical analysis Patients were treated using standard practice and no patients under investigative protocol were included in this study. None of the patients had received preoperative chemotherapy or radiation therapy. In the analysis of

disease-speci®c survival, patients who died of any urothelial cancer were judged as deaths. All other clinical situations (alive with or without cancer or death of other causes) were considered as censored. All molecular and immunohistochemical analysis were completed, analysed and recorded blind to the clinical information. The survival interval was de®ned as the time between the date of initial treatment and the date of death or censoring. KGF-R expression was compared to disease-speci®c survival. The patients were subdivided into three groups of equal size according to the level of expression of KGF-R mRNA (18 patients per group). It appears that the patients belonging to the group with the lowest level of the receptor have a KGF-R level below 30% of the average value found in the normal urothelium. Survival curves were constructed using the Kaplan Meier method (Kaplan and Meier, 1958) and compared using the Log rank test (Mantel, 1966). The distribution of KGF-R expression was studied according to stage and grade. Distribution frequencies were analysed by means of the Chi-square test. BMDP software (Logisoft) was used for statistics and graphics.

Acknowledgements We thank Robert Dayton for excellent technical assistance, Jeanne-Marie Girault for preparing the GAPDH probe and Dominique Morineau for expert photographic assistance. We are grateful to Willem de Boer, Richard Asher and Bernard Asselain for helpful comments on the manuscript. This work has been supported in part by the Association Claude Bernard, UniversiteÂ Paris XII, DeÂleÂgation aÁ la Recherche Clinique AP-HP, CNRS, ARC 6455, Ligue contre le Cancer (National committee and committee of Paris), the Fondation de France and NIH grant 2RO1 CA 49417-06. S Gil Diez de Medina was the recipient of a fellowship from the Ligue contre le Cancer-ComiteÂ du Val de Marne.

References Aaronson SA. (1991). Science, 254, 1146 ± 1153. Alarid ET, Rubin JS, Young P, Chedid M, Ron D, Aaronson SA and Cunha GR. (1994). Proc. Natl. Acad. Sci. USA, 91, 1074 ± 1078. Bader BL, Jahn L and Franke WW. (1988). Eur. J. Cell. Biol., 47, 300 ± 319. Basilico C and Moscatelli D. (1992). Adv. Cancer Res., 59, 115 ± 165. Chirgwin JM, Przybyla AE, MacDonald RJ and Rutter WJ. (1979). Biochemistry, 18, 5294 ± 5299. Cross M and Dexter TM. (1991). Cell, 64, 271 ± 280. Dell KR and Williams LT. (1992). J. Biol. Chem., 267, 21225 ± 21229. Dionne CA, Crumley G, Bellot F, Kaplow JM, Searfoss G, Ruta M, Burgess WH, Jaye M and Schlessinger J. (1990). EMBO J., 9, 2685 ± 2692. Dionne CA, Modi WS, Crumley G, O'Brien SJ, Schlessinger J and Jaye M. (1992). Cytogenet. Cell. Genet., 60, 34 ± 36. Eisemann A, Ahn JA, Graziani G, Tronick SR and Ron D. (1991). Oncogene, 6, 1195 ± 1202. Ercolani L, Florence B, Denaro M and Alexander M. (1988). J. Biol. Chem., 263, 15335 ± 15341. Finch PW, Rubin JS, Miki T, Ron D and Aaronson SA. (1989). Science, 245, 752 ± 755. Finch PW, Cunha GR, Rubin JS, Wong J and Ron D. (1995). Dev. Dyn., 203, 223 ± 240. Fults D and Pedone C. (1993). Genes Chrom. Cancer, 7, 173 ± 177. Givol D and Yayon A. (1992). FASEB J., 6, 3362 ± 3369. Guo L, Yu QC and Fuchs E. (1993). EMBO J., 12, 973 ± 986.

Hanauer A and Mandel JL. (1984). EMBO J., 3, 2627 ± 2663. Hattori Y, Odagiri H, Nakatani H, Miyagawa K, Naito K, Sakamoto H, Katoh O, Yoshida T, Sugimura T and Terada M. (1990). Proc. Natl. Acad. Sci. USA, 87, 5983 ± 5987. Hynes NE and Stern DF. (1994). Biochim. Biophys. Acta., 1198, 165 ± 184. Ishii H, Yoshida T, Oh H, Yoshida S and Terada M. (1995). Mol. Cell. Biol., 15, 3664 ± 3671. Itoh H, Hattori Y, Sakamoto H, Ishii H, Kishi T, Sasaki H, Yoshida T, Koono M, Sugimura T and Terada M. (1994). Cancer Res., 54, 3237 ± 3241. Johnson DE, Lu J, Chen H, Werner S and Williams LT. (1991). Mol. Cell. Biol., 11, 4627 ± 4634. Johnson DE and Williams LT. (1993). Adv. Cancer Res., 60, 1 ± 41. Kandel J, Bossy-Wetzel E, Radvanyi F, Klagsbrun M, Folkman J and Hanahan D. (1991). Cell, 66, 1095 ± 1104. Kaplan G and Meier P. (1958). J. Am. Stat. Assoc., 53, 457 ± 481. Knowles MA, Elder PA, Williamson M, Cairns JP, Shaw ME and Law MG. (1994). Cancer Res., 54, 531 ± 538. Kulesh DA and Oshima RG. (1989). Genomics, 4, 339 ± 347. Labarca C and Paigen K. (1980). Anal. Biochem., 102, 344 ± 352. LaRochelle WJ, Dirsch OR, Finch PW, Cheon HG, May M, Marchese C, Pierce JH and Aaronson SA. (1995). J. Cell. Biol., 129, 357 ± 366. Leube RE, Bosch FX, Romano V, Zimbelmann R, Ho¯er H and Franke WW. (1986). Dierentiation, 33, 69 ± 85.

329


330

Li ZL, Lilienbaum A, Butler-Browne G and Paulin D. (1989). Gene, 78, 243 ± 254. Lorenzi MV, Long JE, Miki T and Aaronson SA. (1995). Oncogene, 10, 2051 ± 2055. Mantel N. (1966). Cancer Chemother. Rep., 50, 163 ± 170. Markowitz S, Wang J, Myero L, Parsons R, Sun L, Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW, Vogelstein B, Brattain M and Willson JKV. (1995). Science, 268, 1336 ± 1338. Mattei MG, Moreau A, Gesnel MC, Houssaint E and Breathnach R. (1991). Hum. Genet., 87, 84 ± 86. Mellon K, Wright C, Kelly P, Horne CH and Neal DE. (1995). J. Urol., 153, 919 ± 925. Miki T, Fleming TP, Bottaro DP, Rubin JS, Ron D and Aaronson SA. (1991). Science, 251, 72 ± 75. Miki T, Bottaro DP, Fleming TP, Smith CL, Burgess WH, Chan AML and Aaronson SA. (1992). Proc. Natl. Acad. Sci. USA, 89, 246 ± 250. Miyamoto M, Naruo KI, Seko C, Matsumoto S, Kondo T and Kurokawa T. (1993). Mol. Cell. Biol., 13, 4251 ± 4259. Mosto® FK, Sobin LH and Torloni H. (eds) (1973). International Histological Classi®cation of Tumors, No. 10: Histological Typing of Urinary Bladder Tumors. World Health Organization, Geneva. Nakagawara A, Arima-Nakagawara M, Azar CG, Scavarda NJ and Brodeur GM. (1994). Prog. Clin. Biol. Res., 385, 155 ± 161. Orr-Urtreger A, Bedford MT, Burakova T, Arman E, Zimmer Y, Yayon A, Givol D and Lonai P. (1993). Dev. Biol., 158, 475 ± 486. Oshima RG, Millan JL and Cecena G. (1986). Differentiation, 33, 61 ± 68.

Peters KG, Werner S, Chen G and Williams LT. (1992). Development, 114, 233 ± 243. Peters KG, Werner S, Liao X, Wert S, Whitsett J and Williams L. (1994). EMBO J., 13, 3296 ± 3301. Peterson MG, Tanese N, Pugh BF and Tjian R. (1990). Science, 248, 1625 ± 1630. Radvanyi F, Christgau S, Baekkeskov S, Jolicoeur C and Hanahan D. (1993). Mol. Cell. Biol., 13, 4223 ± 4232. Savagner P, Valles AM, Jouanneau J, Yamada KM and Thiery JP. (1994). Mol. Biol. Cell., 5, 851 ± 862. Spiessl B, Beahrs OH, Hermanek P, Hutter RVP, Scheibe O, Sobin LH and Wagner G. (1992). TNM atlas, Illustrated guide to the TNM/pTNM classi®cation of malignant tumors. 3d ed. 2d rev. UICC Sprinter-Verlag. (eds), pp. 214 ± 279. Sporn MB and Roberts AB. (1988). Nature, 332, 217 ± 219. Tanaka A, Miyamoto K, Minamino N, Takeda M, Sato B, Matsuo H and Matsumoto K. (1992). Proc. Natl. Acad. Sci. USA, 89, 8928 ± 8932. Werner S, Weinberg W, Liao X, Peters KG, Blessing M, Yuspa SH, Weiner RL and Williams LT. (1993). EMBO J., 12, 2635 ± 2643. Werner S, Smola H, Liao X, Longaker MT, Krieg T, Hofschneider PH and Williams LT. (1994). Science, 266, 819 ± 822. Yamaguchi F, Saya H, Bruner JM and Morrison RS. (1994). Proc. Natl. Acad. Sci. USA, 91, 484 ± 488. Yan G, Fukabori Y, McBride G, Nikolaropolous S and McKeehan WL. (1993). Mol. Cell. Biol., 13, 4513 ± 4522.