Tumour suppressive properties of ®broblast growth factor receptor 2-IIIb inhuman bladder cancer

David Ricol1, David Cappellen1, Ahmed El Marjou1, Sixtina Gil-Diez-de-Medina1,2,Jeanne-Marie Girault1, Teruhiko Yoshida3, Gilles Ferry4, Gordon Tucker4,Marie-France Poupon5, Dominique Chopin2, Jean Paul Thiery1 and FrancË ois Radvanyi*,1

1UMR 144, Centre National de la Recherche Scienti®que, Institut Curie, Section de Recherche, 26 rue d'Ulm, 75248 Paris Cedex05, France; 2Groupe d'Etude des Tumeurs Urologiques et Service d'Urologie, Centre Hospitalier Universitaire Henri Mondor,AP-HP, 94010 CreÂteil Cedex, France; 3Genetics Division, National Cancer Center Research Institute, Tokyo 104-0045, Japan;4Institut de Recherches Servier, Division de CanceÂrologie ExpeÂrimentale, 92150 Suresnes, France; 5UMR 147, Centre Nationalde la Recherche Scienti®que, Institut Curie, Section de Recherche, 26 rue d'Ulm, 75248 Paris Cedex 05, France

FGFRs (®broblast growth factor receptors) are encodedby four genes (FGFR1 ± 4). Alternative splicing results invarious receptor isoforms. The FGFR2-IIIb variant ispresent in a wide variety of epithelia, including thebladder epithelium. Recently, we have shown thatFGFR2-IIIb is downregulated in a subset of transitionalcell carcinomas of the bladder, and that this down-regulation is associated with a poor prognosis. Weinvestigated possible tumour suppressive properties ofFGFR2-IIIb by transfecting two human bladder tumourcell lines, J82 and T24, which have no endogenousFGFR2-IIIb expression, with FGFR2-IIIb cDNA. Nostable clones expressing FGFR2-IIIb were isolated withthe J82 cell line. For the T24 cell line, stabletransfectants expressing FGFR2-IIIb had reducedgrowth in vitro and formed fewer tumours in nude micewhich, in addition, grew more slowly. The potentialmechanisms leading to decreased FGFR2-IIIb mRNAlevels were also investigated. The 5' region of the humanFGFR2 gene was isolated and found to contain a CpGisland which was partially methylated in more than halfthe cell lines and tumours which do not express FGFR2-IIIb. No homozygous deletion was identi®ed in any ofthe tumours or cell lines with reduced levels of FGFR2-IIIb. Mutational analysis of the entire coding region ofFGFR2-IIIb at the transcript level was performed in 33bladder tumours. In addition to normal FGFR2-IIIbmRNA, abnormal transcripts were detected in twotumour samples. These abnormal mRNAs resulted fromexon skipping which a�ected the region encoding thekinase domain. Altogether, these results show thatFGFR2-IIIb has tumour growth suppressive propertiesin bladder carcinomas and suggest possible mechanismsof FGFR2 gene inactivation.

Keywords: bladder; ®broblast growth factor receptor;human; urothelial cell carcinoma; tumour suppressorgene

Introduction

Urothelial cell carcinoma is the fourth most commoncancer in men and the ninth most common cancer inwomen in Western countries. Bladder carcinomas areeither muscle-invasive or super®cial. Muscle-invasivetumours are associated with a high risk of metastasesand a poor prognosis, despite radical therapy.Unpredictably, super®cial tumours may recur, and in20% of cases they progress to muscle-invasive disease(Lamm and Torti, 1996).

In the search for new markers of bladder tumourprogression, we have previously reported that FGFR2-IIIb (®broblast growth factor receptor 2-IIIb) wasexpressed in the normal bladder urothelium and thatits expression was decreased in a subset of bladdercarcinomas, this change of expression being associatedwith a poor prognosis (Gil Diez de Medina et al.,1997). There are four genes (FGFR1, FGFR2, FGFR3and FGFR4) in mammals encoding FGF receptors(FGFRs). FGFRs are glycoproteins consisting of twoor three extracellular immunoglobulin (Ig)-like do-mains, an hydrophobic transmembrane region and acytoplasmic part containing a tyrosine kinase catalyticsite. Alternative mRNA splicing mechanisms producevarious receptor isoforms (Johnson and Williams,1993; McKeehan et al., 1998). Isoforms FGFR2-IIIband FGFR2-IIIc result from a mutually exclusivesplicing event in which the second half of thejuxtamembrane Ig-like domain of FGFR2 is encodedby either the 148-nucleotide IIIb exon or the 145-nucleotide IIIc exon (Miki et al., 1992). FGFR2-IIIb ispresent exclusively in epithelial cells and binds thestromal cell-derived factor FGF7/KGF with higha�nity (Finch et al., 1989; Rubin et al., 1989; Orr-Urtreger et al., 1993; LaRochelle et al., 1995).

Growth factors and their cognate receptors maypositively or negatively regulate tumour progression(Sporn and Roberts, 1988; Cross and Dexter, 1991;Aaronson, 1991). In many cases, FGFs and theirreceptors are thought to promote tumour progression.The upregulation of FGFs and FGFRs has beenreported in many cancers: FGFR1 for astrocytomas(Yamaguchi et al., 1994), FGF5, FGF7 and FGFR2-IIIb for pancreatic cancer (Kornmann et al., 1997;Ishiwata et al., 1998), FGF2 for melanomas (Halabanet al., 1987, 1988; Becker et al., 1989; Reed et al.,1994), and FGF3 in MMTV-induced mouse mammary

*Correspondence: F RadvanyiReceived 18 November 1998; revised 26 August 1999; accepted26 August 1999

Oncogene (1999) 18, 7234 ± 7243ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00

http://www.stockton-press.co.uk/onc

tumours (Muller et al., 1990; Shackleford et al., 1993).Several in vitro and in vivo investigations have shownthat FGF2 and FGFR1 are essential for melanomagrowth (Wang and Becker, 1997; Yayon et al., 1997).In haematopoietic malignancies, translocations invol-ving FGFR1 or FGFR3 have demonstrated that FGFreceptors may have oncogenic properties (Chesi et al.,1997; Xiao et al., 1998). In direct contrast to theirpositive e�ects on cell growth, invasiveness, motilityand angiogenesis, some data point to an inhibitorye�ect of FGFs and FGF receptors on the growth ofseveral tumour cell lines (Schweigerer et al., 1987; Fenget al., 1997; Fenig et al., 1997; Johnson et al., 1998;Matsubara et al., 1998). The mediation of inhibitorysignals by FGF receptors is a developing concept in the®eld of developmental biology. Germline mutations inhumans and knockout experiments in mice have shownthat FGFR3 negatively regulates bone growth (DeMoerlooze and Dickson, 1997; Colvin et al., 1996;Deng et al., 1996). There are very few reportssuggesting that FGFs and FGFRs have an inhibitoryrole in human cancer that is lost as the tumourprogresses. The loss of FGFR2-IIIc is closelyassociated with malignant progression in astrocyte-derived tumours (Yamaguchi et al., 1994). Recently, wehave shown that reduced expression of FGFR2-IIIb isassociated with a poor prognosis in transitional cellcarcinomas of the bladder (Gil Diez de Medina et al.,1997). We investigated whether FGFR2-IIIb hasinhibitory properties in bladder cancers by a series ofexperiments to determine the e�ects of FGFR2-IIIbexpression in human bladder tumour-derived cell lineslacking endogenous FGFR2-IIIb. We found thatFGFR2-IIIb reduced cell proliferation in vitro andtumour formation in nude mice. We also investigatedthe possible genetic and epigenetic mechanisms ofFGFR2 gene inactivation.

Results

Expression of FGFR2-IIIb in bladder tumour cell lines

The signi®cantly reduced level of FGFR2-IIIb in asubset of bladder tumours with poor prognosis (GilDiez de Medina et al., 1997) led us to examine itsexpression in bladder tumour cell lines. Very little orno FGFR2-IIIb mRNA was detected in nine of 17 celllines (53%). Seven cell lines expressed FGFR2-IIIbmRNA at similar levels to that of normal urothelium(Table 1). None of the 17 cell lines did expressFGFR2-IIIc, the other major splice variant of theFGFR2 gene (data not shown). Therefore, as observedin vivo (Gil Diez de Medina et al., 1997), the decreasein FGFR2-IIIb mRNA was not associated with anincrease in FGFR2-IIIc mRNA in vitro. In everysample in which protein and mRNA were both studied,there was agreement between the levels of mRNA andprotein.

Expression of FGFR2-IIIb in transfected clones

We assessed the e�ect of FGFR2-IIIb on the malignantphenotype, by transfecting the J82 and T24 cell lines,which lack endogenous FGFR2-IIIb expression, with amammalian expression vector containing the full-length

human FGFR2-IIIb cDNA under the control of thecytomegalovirus (CMV) promoter (pcDNAI/Neo-FGFR2-IIIb, see Materials and methods). The vectorwithout an insert was used as a control (pcDNAI/Neo).

J82 cells were selected by growth for 2 weeks onG418. Transfection with pcDNAI/Neo-FGFR2-IIIbgave almost half as many colonies as the vector alone(Table 2). None of the 13 FGFR2-IIIb-transfectedclones tested expressed the FGFR2-IIIb mRNA, asdetected by RT±PCR analysis. Eleven of the 13 cloneswere tested for the presence of the transfected FGFR2-IIIb cDNA by genomic PCR analysis. The FGFR2-IIIb cDNA was not ampli®ed from any of these cloneswhereas the neomycin resistance gene was detected inall cases, indicating that the transfected clonesselectively lost the FGFR2-IIIb transfected cDNAbut retained a functional neomycin resistance gene.Thus, FGFR2-IIIb expression seems to be incompa-tible with the growth of J82 cells.

For T24 cells, transfection with the FGFR2-IIIb-expressing plasmid gave slightly, but not signi®cantly,more G418 resistant clones than did transfection withthe control vector (Table 2). In six of the ten FGFR2-IIIb-transfected clones tested the receptor was detectedby semi-quantitative RT±PCR (data not shown). Thestable FGFR2-IIIb-expressing cells were similar inmorphology to vector-alone controls and untrans-fected parental cells (data not shown). Protein levelswere investigated by immunoblot analysis in threeclones expressing FGFR2-IIIb mRNA (T24K4,T24K6, T24K10), in two control clones (T24N2,T24N4), in the parental cell line (T24) and in normalurothelium. The polyclonal anti-FGFR2 antibodyrecognized a 94 kD species corresponding to FGFR2-IIIb (Figure 1a). For two of the three transfectedclones expressing FGFR2-IIIb, the amount of FGFR2-IIIb protein was similar to that of normal urothelium,whereas for the third, signi®cantly more FGFR2-IIIbprotein was present. As expected, the FGFR2 proteinwas not detected in vector control samples or theparental cell line. We determined the subcellular

Table 1 FGFR2-IIIb levels in human bladder carcinoma cell lines

Cell line FGFR2-IIIb/TBPa Proteinc

253JEJ138HCV-29J82JON53KK-47T24TCCSUPUM-UC3SD48647VHT-1376RT112RT4ScaBERVM-CUB-1VM-CUB-3Normal urothelium

00000.100000.51.933.35.14.94.13.5

1.9+0.4b

nd7nd7ndnd7ndndndndnd+++ndnd+

aLevel of FGFR2-IIIb mRNA determined by semi-quantitative RT-PCR using TBP as an internal standard. bMean+s.d. Level ofFGFR2-IIIb mRNA determined for ®ve di�erent normal urothelia.cProtein level determined by Western blot analysis; +, present; 7,absent; nd, not determined

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7235

location of the FGFR2-IIIb protein by confocalimmuno¯uorescence microscopy of the T24K10-trans-fected clone using the polyclonal anti-FGFR2 anti-body. The FGFR2-IIIb protein was present at the cellsurface, with some staining of cytoplasmic vesicularstructures also apparent (Figure 1b). We assessed thefunctionality of the transfected receptor, by determin-ing FGFR2-IIIb phosphorylation status in T24K10cells with or without KGF/FGF7 stimulation. Im-munoblot analysis with anti-phosphotyrosine antibodyshowed that KGF strongly stimulated FGFR2-IIIbtyrosine phosphorylation. Little or no phosphorylationwas observed in the absence of ligand (Figure 2).

Phenotype of stable FGFR2-IIIb-transfected clones

T24K4, T24K6 and T24K10 clones were used toinvestigate the e�ects of FGFR2-IIIb on cell prolifera-tion in vitro and tumour formation in nude mice.

The e�ects of FGFR2-IIIb on cell proliferation wereassessed by assaying 3H-thymidine incorporation.There was signi®cantly less thymidine incorporationin the FGFR2-IIIb expressing clones than in theparental cell line and the clones transfected with thevector only (T24N2, T24N4) (Figure 3). The largeste�ect (90% less) was observed with the high expresserclone (T24K10). No further decrease in thymidineincorporation was observed if one of two FGFR2-IIIbligands, FGF1 and FGF7, was added to the medium(data not shown).

We assessed the e�ects of FGFR2-IIIb expressionon the growth of tumours in vivo, by injecting theFGFR2-IIIb expressing clones (T24K6, T24K10) andcontrol clones (T24N2, T24N4) subcutaneously intothe ¯anks of female athymic nude mice. Mice injectedwith cells that did not express FGFR2-IIIb formedtumours within 5 weeks and mean tumour size wasabout 400 mm3 after 20 weeks. Both the FGFR2-IIIb-transfected cell lines, T24K6 and T24K10, causedsigni®cantly less tumour formation than the vector-only controls (T24N2 and T24N4) (Figure 4). Thenumber of tumours produced from the FGFR2-IIIb-expressing clones, T24K6 and T24K10, was noticeablysmaller (17 and 0% respectively) than the numberproduced from the control group T24N2, and T24N4(50 and 83% respectively) (Table 3). In addition, thetwo tumours resulting from injection with the FGFR2-IIIb-transfected clones grew much more slowly thanthose arising from the control vector transfectants,reaching a maximum size of 40 mm3 after 21 weeks,whereas control vector transfectants grew to a meansize of 500 mm3 in the same time. The two smalltumours produced by injection with the T24K6 clonewere examined histologically, but there was no obvious

di�erence between these tumours and those obtainedwith the T24N2 and T24N4 control clones (data notshown).

Table 2 Colony formation assay for bladder tumor cell lines transfected with FGFR2-IIIb

Mean number of colonies+s.d.a

J82 cell line T24 cell line

FGFR2-IIIb vector (pcDNAI/Neo-FGFR2-IIIb)Control vector (pcDNAI/Neo)Ratio of pcDNAI/Neo FGFR2-IIIb vs pcDNAI/NeoP valueb pcDNAI/Neo vs pcDNAI/Neo-FGFR2-IIIb

110+5214+60.51

50.0001

98+13189+200.52

50.006

222+37149+111.49NS

243+27156+461.56NS

aEach experiment was performed in triplicate. The results from two di�erent experiments are presented. bP values are for a two-tailed t-test; NS,di�erences not signi®cant

Figure 1 FGFR2-IIIb expression in stably transfected T24 celllines. (a) Immunoblot analysis of protein lysates from normalurothelium, untransfected parental T24 cells, T24 cells transfectedwith vector alone (T24N2 and T24N4), and T24 cells transfectedwith FGFR2-IIIb cDNA (T24K4, T24K6 and T24K10) using apolyclonal anti FGFR2 antibody. (b) Cellular location ofexogenous FGFR2-IIIb. FGFR2-IIIb was detected by confocalmicroscopy using a polyclonal anti-FGFR2 antibody in T24 cellsstably transfected with pcDNAI/Neo-FGFR2-IIIb (T24 K10)

Figure 2 Expression levels and phosphorylation of FGFR2-IIIbfollowing ligand stimulation in T24 transfected cells. Totalcellular lysate (1 mg) from T24 transfected either with vectoralone (T24N4) or with FGFR2-IIIb cDNA was immunoprecipi-tated with an anti-FGFR2 antiserum and immunoblotted withanti-phosphotyrosine antibody (aPTyr, upper blot). Whereindicated by a (+), cells were stimulated by incubation withKGF/FGF7 (50 ng/ml) for 20 min before harvesting. Total celllysates (50 mg) were also immunoblotted with a polyclonal anti-FGFR2 antibody (aFGFR2 lower panel)

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7236

Homozygous deletion analysis for the FGFR2 gene inhuman bladder cancer

The decreased expression of FGFR2-IIIb at bothmRNA and protein level in 30% of bladdercarcinomas (Gil Diez de Medina et al., 1997) and theabsence of FGFR2-IIIb mRNA in eight of 17 bladdertumour cell lines (Table 1) may be due to homozygousdeletion of the FGFR2 gene. We therefore used duplexPCR to study a series of seven bladder carcinoma celllines lacking FGFR2-IIIb and 25 primary bladdertumours including ten with low levels of FGFR2-IIIb.None of the cell lines or tumour samples investigatedshowed evidence of homozygous FGFR2 gene deletion.

CpG island methylation analysis of the FGFR2 gene inbladder tumour cell lines and primary bladdercarcinomas

Hypermethylation of normally unmethylated CpGislands in the promoter regions of genes correlateswith loss of transcription. The proximal promoterregion of the mouse FGFR2 gene has been character-ized and found to contain a CpG island (Avivi et al.,1992). A 181 bp probe from the 5' end of the humanFGFR2 cDNA (+172 to +353, based on Figure 5numbering) was used to isolate a human genomic PACclone. We sequenced 1 kb upstream from the known 5'end of the cDNA (Hattori et al., 1990; Dell andWilliams., 1992). Sequences homologous to those ofthe mouse promoter were identi®ed and are shown inFigure 5a. The 5' end of the human FGFR2 gene(7352 to 441) has a G+C content of 75% and a CpG/GpC ratio of 0.96, and is therefore a CpG island. Amethylation-sensitive restriction map of *1000 bpextending from the promoter region through exon 1of the FGFR2 gene is shown in Figure 5b. A PCR-based assay was used to analyse the methylation statusof this CpG island. Double digestion using a restrictionenzyme with ¯anking cut (AfIII) and methylation-sensitive enzymes (SmaI, HpaII or EagI) was followedby PCR ampli®cation of region 1 or 2. Normalurothelium is unmethylated at the sites tested withinthese two regions (Figure 6 and data not shown). Allcell lines (n=9) and primary tumours (n=6) expressingFGFR2-IIIb tested also had DNA that was unmethy-lated in these regions. In contrast, ®ve of the nine celllines and ®ve of the nine primary bladder tumours thatdid not express FGFR2-IIIb were partially methylated(Figure 6 and data not shown). These di�erences inmethylation were con®rmed by Southern blotting ofDNA isolated from bladder tumour cell lines (data notshown). The treatment of cell lines with 5-aza-2'-deoxycytidine did not reactivate the expression ofFGFR2 gene, as assessed by RT±PCR for cell linesin which the CpG island was methylated, whereas thistreatment clearly reactivated p16 expression in the T24cell line (data not shown).

Loss of heterozygosity of the FGFR2 locus in humanbladder cancer

The FGFR2 gene has been mapped to chromosomeregion 10q25.3-10q26 (Mattei et al., 1991). Allelic losson chromosome 10q occurs in about 50% of invasivebladder cancers (Cappellen et al., 1997). We used tworecently identi®ed microsatellites from the FGFR2 gene(Cappellen et al., submitted) to investigate the loss ofheterozygosity (LOH) at the FGFR2 locus in 35

Figure 4 E�ect of FGFR2-IIIb on tumour growth in nude mice.Two FGFR2-IIIb stably expressing clones (T24K6 and T24K10)were injected subcutaneously into nude mice (six mice per cell lineand two injections per mouse). Tumour growth was comparedwith that in mice injected with two non expressing clones (T24N2and T24N4) transfected with the control vector (six mice per cellline and two injections per mouse). Points, mean tumour volume;bars, s.e.

Figure 3 E�ect of FGFR2-IIIb on thymidine incorporation instably transfected T24 cell lines. 3H-thymidine incorporation,measured as described in `Materials and methods', was assessedfor the T24 parental cell line, two clones transfected with vectoralone (T24 N2 and T24 N4), and three clones transfected withFGFR2-IIIb cDNA (T24 K4, T24 K6 and T24 K10). The amountof FGFR2-IIIb in these clones is shown in Figure 1. Columns,mean values for quadruplicate wells; bars, s.d.

Table 3 Tumorigenicity of FGFR2-IIIb-expressing clones in nudemicea

Number of tumours

FGFR2-IIIb non-expressing clonesT24 N2T24 N4FGFR2-IIIb expressing clonesT24 K6T24 K10

6/1210/12

2/120/12

aNude mice were injected s.c. with 107 cells and tumour formationwas followed for 21 weeks

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7237

bladder carcinomas for which we already had dataconcerning FGFR2-IIIb expression data (Gil Diez deMedina et al., 1997). LOH occurred in nine of 33informative tumours. There was a decreased expressionof FGFR2-IIIb in nine of these 33 tumours with nostatistical correlation between these two factors(P=1.0, Fisher's exact test).

Screening for FGFR2-IIIb cDNA alterations in humanbladder cancer

The entire coding region of FGFR2-IIIb was analysedat the transcript level by SSCP in two tumour bladder

cell lines (SCaBER and 647V), 33 primary bladdercancer, one peritumoural urothelium and three normalurothelia. We detected aberrant SSCP patterns forthree of the 39 cDNAs examined, the SCaBER cell lineand two tumours (1512-6 and 1528-7). Sequencing ofthe DNA corresponding to the aberrant SSCP bandsderived from SCaBER showed that there had been a T-to-C substitution, resulting in an amino acid changefrom Met to Thr at codon 186 (Miki et al., 1992) in theextracellular domain of FGFR2-IIIb. No correspond-ing normal tissue is available for this cell line, so wecannot exclude the possibility that this nucleotidevariation is a polymorphism. Functional analysis is

Figure 5 5' region of the human FGFR2 gene. (a) Nucleotide sequence: the two arrows at positions +1 and +6 indicate thereported cDNA start of FGFR2 (Dell and Williams, 1992, and Hattori et al., 1990 respectively). The cDNA sequence is shown, withinterruptions under the gene sequence, ending with the initiator ATG (boxed). The gene sequence diverges from the cDNA sequenceat a splice site (AG/gt) followed by an intron sequence (small letters). Human homologies to the mouse promoter (Avivi et al., 1992)are indicated with dots. Sp1 binding sites are underlined. (b) Methylation-sensitive restriction maps and CpG nucleotide density. Thepositions of methylation-sensitive restriction enzyme sites (SmaI, HpaII, EagI) are indicated at the 5' end of the FGFR2 geneincluding those in exon 1. FGFR2 regions 1 and 2 analysed for their methylation status are shown at the top. The densities of CpGand GpC dinucleotides are indicated below the restriction map with vertical bars

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7238

required to determine the signi®cance of this change.The variant allele was the only allele detected in theSCaBER cell line and it was not detected among theother 38 cDNAs tested (data not shown). We carriedout DNA sequencing for the SSCP shifts in tumours1512-6 and 1528-7 and found that the mRNA specieswere aberrant, due to the deletion of a single exon,exon 13 for 1512-6 and 17 for 1528-7. These twodeletions a�ected the kinase domain of FGFR2-IIIb.The deletion in tumour 1512-6 removed 111 bp,resulting in a truncated protein lacking part of thekinase domain and the C-terminal end. The aberranttranscript in tumour 1528-7 had a 138 bp in-framedeletion. Both the aberrant and the wild-type mRNAswere present in these two tumours. We found that 2 ±10% of the mRNA present was aberrant. Abnormaltranscripts were not detected in any of the normal andperitumoral urothelia tested, including the peritumoralurothelium from patient 1528-7.

Discussion

FGFR2-IIIb has growth suppressive properties in humanbladder carcinomas

Cancer development is a multistage process that resultsfrom the accumulation of genetic and epigeneticchanges. From an ongoing interest in candidate geneswhose expression is altered during the course ofbladder carcinogenesis, we have recently found thatthe level of a particular isoform of the ®broblastgrowth factor receptor 2, FGFR2-IIIb, is decreased ina subset of bladder tumours, and that this down-regulation is associated with a poor prognosis (GilDiez de Medina et al., 1997). Consistent with this, ineight of 17 bladder tumour cell lines tested in thisstudy, no FGFR2-IIIb mRNA was detected. All thesedata suggested that FGFR2-IIIb could have inhibitoryfunctions in bladder cancers and that its inactivationcould be an important step in tumour progression. Toinvestigate the possible negative e�ects of FGFR2-IIIb

on human bladder tumour progression, FGFR2-IIIbcDNA was transfected into two bladder tumour celllines, J82 and T24, which lack endogenous FGFR2-IIIb expression. Transfection of the J82 cell line withFGFR2-IIIb cDNA decreased colony formation bytwofold. Moreover, all the stable transfectants that weobtained had no FGFR2-IIIb expression and had lostthe transfected FGFR2-IIIb cDNA. Thus, in this cellline, the expression of FGFR2-IIIb was incompatiblewith growth. Stable clones expressing FGFR2-IIIbwere obtained with the T24 cell line. The expression ofFGFR2-IIIb in T24 cells reduced growth in vitro andprevented tumour formation in nude mice. Thus, thiswork provides direct evidence that FGFR2-IIIbinhibits growth and tumour formation in humanbladder tumour cells. FGF-1 and FGF-7, twoFGFR2-IIIb ligands, had no e�ect on the growth ofFGFR2-IIIb transfectants. This may be because: (i)T24 and T24 transfectants express FGF-7, and to alesser extent, FGF-1 (data not shown) in su�cientquantities to create an autocrine loop with FGFR2-IIIb. The addition of exogenous FGF-1 or FGF-7would then have no further e�ect, or (ii) some of thee�ects of FGF receptors are ligand-independent (Greenet al., 1996). The fact that little or no phosphorylationwas observed in the absence of ligand does not supportthe ®rst hypothesis.

Mechanisms of FGFR2 gene inactivation

The inhibitory properties of FGFR2-IIIb and thefrequent loss of or decrease in FGFR2-IIIb expressionthat occurs in about 30% of primary bladder tumours(Gil Diez de Medina et al., 1997) and 50% of celllines, suggest that the inactivation of FGFR2-IIIb is acritical step in the development or progression ofbladder cancer. Several mechanisms may be respon-sible for FGFR2 gene silencing. We detected nohomozygous deletions in the cell lines and primarytumours analysed. We investigated the abnormalmethylation of the 5' CpG island of the FGFR2 geneas a potential inactivation mechanism for this gene.De novo methylation of the 5' CpG island of FGFR2was observed in 50% of tumours and cell lines withlittle or no FGFR2-IIIb expression, whereas the CpGisland was not methylated in FGFR2-IIIb expressingsamples. Treatment of cell lines in which the 5' CpGisland of FGFR2 was methylated with the demethylat-ing agent 5-aza-2'-deoxycytidine did not restoreFGFR2-IIIb expression (data not shown). Therefore,it is unclear whether FGFR2 methylation is the causeor consequence of FGFR2-IIIb downregulation. Weare currently exploring other mechanisms of FGFR2downregulation, such as the loss of a transactivator ofthe FGFR2 gene or altered chromatin structure in theFGFR2 region. Allelic loss on chromosome 10q, whereFGFR2 is located, frequently occurs in bladder cancer(Cappellen et al., 1997). We therefore looked forsomatic mutations of FGFR2-IIIb. The entire codingregion of FGFR2-IIIb was analysed at the transcriptlevel in our bladder tumour samples by SSCP analysis.Distinct banding patterns were observed with two of33 tumours (6%). These abnormal bands weretumour-speci®c and corresponded to transcriptslacking one exon (exon 13 or 17) encoding part ofthe kinase domain of FGFR2. We do not know

Figure 6 Methylation analysis of the FGFR2 gene in humanbladder carcinomas. A representative methylation-sensitive PCRanalysis of normal urothelium and neoplastic cells (RT112 andEJ138 cell lines and two examples of urothelial cell carcinoma) isshown. FGFR2+ are samples with FGFR2-IIIb expression,FGFR27 samples without. The various lanes represent PCRproducts from DNA samples digested with A¯II (A, cuttingoutside the region analysed) as a positive control; A¯II plus MspI(A+M, cutting within the region analysed, methylation-insensitive) as a negative control; A¯II plus SmaI (A+S), HpaII(A+H) or EagI (A+E) (all three cutting within the regionanalysed and methylation-sensitive)

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7239

whether this was due to aberrant alternative splicingor to a mutation resulting in exon skipping. Similaraberrant transcripts have been reported for the mdm 2gene (Sigalas et al., 1996) and the potential tumoursuppressor gene, FHIT (Ohta et al., 1996). The role ofthese aberrant transcripts in tumour progression is stilla matter of debate (Mao, 1998). The two aberrantFGFR2 transcripts encode FGFR2-IIIb without afunctional tyrosine kinase, which would act asdominant negative FGF receptors. The small propor-tion of aberrant transcripts in the tumours (less than10%) leaves open the possibility that this is of realimportance. It would be of value to determine whethera small proportion of the neoplastic cells expresspreferentially the aberrant transcripts, or whether allneoplastic cells express the aberrant transcripts, but ata low level. Tumour microdissection and cell linederivation should make it possible to determine whichof these alternatives applies.

Mechanisms of FGFR2-IIIb-mediated growth inhibition

The mechanisms underlying the inhibitory propertiesof FGFR2-IIIb are unknown. FGFR1 is not presentin the normal urothelium, but is expressed in manybladder carcinoma and cell lines, including T24 andJ82 (data not shown), suggesting that FGFR1 mayaccelerate tumour progression. In transfected cells,FGFR2-IIIb may heterodimerise with FGFR1,inhibiting the signal transduced by FGFR1, assuggested by Feng et al. (1997). Alternatively,FGFR2-IIIb may itself mediate an inhibition signal.Some FGF receptors induce the expression of the cellcycle inhibitor, p21/WAF1 (Su et al., 1997; Johnson etal., 1998). This was not the case for FGFR2-IIIb inT24 cells because there was no more p21/WAF1 in theFGFR2-IIIb expressing T24 clones than in the controlclones (data not shown).

Role of FGFR2-IIIb in other carcinomas

FGFR2-IIIb is expressed in various human epitheliaand is not expressed in many carcinoma-derived celllines (LaRochelle et al., 1995; Rubin et al., 1995;Carstens et al., 1997). This suggests that thedownregulation of FGFR2-IIIb in human carcinomais not restricted to the bladder. In a rat tumourmodel, loss of FGFR2-IIIb expression is associatedwith malignant progression (Yan et al., 1993).Furthermore, in this model, it has been recentlyshown that FGFR2-IIIb inhibits the growth ofmalignant cells both in vitro and in vivo (Feng etal., 1997, Matsubara et al., 1998). In human prostatetumour cell lines, Carstens et al. (1997) found thatthe loss of FGFR2-IIIb expression correlated withandrogen insensitivity. It is unknown whetherFGFR2-IIIb is downregulated in human prostatecancer samples. Unlike other carcinoma cell lines,many stomach carcinoma cell lines express FGFR2-IIIb, suggesting that, in these cancers, FGFR2-IIIbmay accelerate tumour progression. Stomach carcino-mas are the only tumours known in which a shorterform of FGFR2-IIIb is expressed, this shorter formbeing more oncogenic than the other forms ofFGFR2-IIIb in a 3T3 transformation assay (Itoh etal., 1994).

Conclusions

The downregulation of FGFR2-IIIb in a subset ofbladder carcinomas with poor prognosis led us toinvestigate the inhibitory properties of this receptorand the potential mechanisms of FGFR2 geneinactivation. FGFR2-IIIb inhibits cell growth both invitro and in vivo in bladder carcinomas. The FGFR2gene is methylated in a signi®cant proportion ofprimary tumours and cell lines not expressing thisgene. However FGFR2-IIIb cannot be de®ned as atrue suppressor gene because no clear inactivatingmutations have been identi®ed (Haber and Harlow,1997). In bladder carcinomas, FGFR2-IIIb behaveslike E-cadherin, which is functionally a tumoursuppressor gene, and the promoter of which issometimes methylated, but which has a low frequencyof mutation (Giroldi et al., 1994). The levels of E-cadherin and FGFR2-IIIb are related: tumours withlow levels of E-cadherin always have a low level ofFGFR2-IIIb, suggesting a common mechanism ofinactivation for these two genes (Gil Diez de Medinaet al., 1999). There are probably two classes of negativeregulators of tumor progression, those that arefunctionally and genetically tumour suppressor genesand those that meet only the functional criteria. Atumour suppressor gene satisfying the functionalcriteria in a given tumour (for example, E-cadherin inbladder carcinoma) may meet both the genetic andfunctional criteria in other types of tumour (E-cadherinin lobular breast carcinoma and in gastric carcinoma)(Guilford et al., 1998). Further work is required todetermine whether this is also the case for FGFR2.

Materials and methods

Tissue specimens

Normal urothelium (n=45) and transitional cell carcinomasof the bladder (n=33) were obtained as described elsewhere(Gil Diez de Medina et al., 1997). Tumour samples containedmore than 80% malignant cells, as determined by histologicalanalysis of adjacent sections.

Cell culture, constructs and transfection

Human bladder cell lines T24, J82, TCCSUP and SCaBERwere obtained from the ATCC (Rockville, MD, USA), EJ138and RT112 from the ECACC (Salisbury, UK). 647V, SD48and JON53 were provided by Professor D Chopin (Hoà pitalHenri Mondor, Cre teil, France). All cell lines were culturedin DMEM/F12+ (Dulbecco's Modi®ed Eagle Medium/F12containing 10% heat-inactivated foetal calf serum, 2 mM

glutamine, 100 units/ml penicillin, 100 mg/ml streptomycin,10 mM HEPES, 50 nM hydrocortisone, 5 mg/ml apotransfer-rin and 5 nM selenium). For demethylation analysis, 5-aza-2'-deoxycytidine (Sigma) was added to the medium atconcentrations of 0.3, 1 or 3 mM and cells were harvestedafter 3, 5 or 10 days of treatment. Lysates in 4 M

guanidinium thiocyanate of the bladder cell lines HCV-29,HT-1376, RT4, VM-CUB-1, VM-CUB-3, 253-J, KK47, UM-UC3 were kindly provided by Dr J Southgate (Leeds, UK).

The FGFR2-IIIb isoform used for transfection was K-samIIC1 (Itoh et al. 1994), the principal isoform in normalurothelium and bladder carcinomas (D Cappellen, unpub-lished results). The FGFR2-IIIb cDNA, in a cytomegaloviruspromoter-driven pcDNAI/Neo expression vector (Invitrogen,

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7240

San Diego, CA, USA) was provided by Drs M Terada and TYoshida (National Cancer Centre Research Institute, Tokyo,Japan) (Itoh et al. 1994). This plasmid is called pcDNAI/Neo-FGFR2-IIIb, and the control plasmid, pcDNAI/Neo.T24 and J82 cells (26105/35 mm tissue culture plate) weretransfected by a liposome-mediated method using LipofectA-MINE (Life Technologies). Three days after transfection,cells were transferred to 100 mm dishes ®lled with DMEM/F12+ containing Geneticin (G418; Life Technologies),400 mg/ml for T24 and 700 mg/ml for J82. Cell colonies onthe selection medium were stained using Di�-Quik stainingsolutions (Dade, Paris, France) and counted, or isolated andassessed for expression of FGFR2-IIIb.

Reverse transcription-PCR analysis

Total RNA was extracted using Trizol (Life Technologies) orcaesium chloride centrifugation and used as the template for®rst stand cDNA synthesis by random priming as previouslydescribed (Kandel et al. 1991, Radvanyi et al. 1993). Theamount of FGFR2-IIIb mRNA was determined by semi-quantitative radioactive RT±PCR, using TBP (TATA-binding protein) as an internal control, as previouslydescribed (Gil Diez de Medina et al., 1997). The number ofcycles was selected so as to be in the exponential part of thePCR reactions (25 cycles). The PCR-ampli®ed products weresubjected to electrophoresis in 8% polyacrylamide gels.Signals were quanti®ed with a Molecular Dynamics 300PhosphorImager (Molecular Dynamics, Sunnyvale, CA,USA). There was no ampli®cation if reverse transcriptasewas omitted from the reverse transcription reaction.

Immunodetection of FGFR2-IIIb

For Western blot analysis, cells were harvested at 80%con¯uence and lysed in 50 mM Tris-HCl (pH 7.4), 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM

EDTA, 1 mM EGTA, 1 mM Na3VO4, 50 mM NaF, 1 mg/mlaprotinin, 1 mg/ml leupeptin, 5 mM Pefabloc (BoehringerMannheim, Meylan, France). Protein concentration wasdetermined using the BioRad kit (BioRad, Ivry sur Seine,France). Protein (50 mg) of each extract was resolved by 10%sodium dodecylsulphate polyacrylamide gel electrophoresis(SDS ±PAGE), transferred to nitrocellulose membrane(Schleicher and Schuell, Dassel, Germany), and incubatedwith a rabbit polyclonal anti-FGFR2 antibody (Santa CruzBiotechnology, Santa Cruz, CA, USA). Bound antibodieswere detected with an anti-rabbit horse-radish peroxidase-conjugated secondary antibody (Amersham, Les Ulis,France), by enhanced chemiluminescence (ECL kit; Amer-sham).

For immuno¯uorescence analysis, cells were plated oncoverslips in 24-well plates at a density of 26105 cells/well inDMEM/F12+. They were allowed to attach for 48 h, ®xedin cold methanol/acetone (1/1), and then incubated for 1 h at378C with a rabbit polyclonal anti-FGFR2 antibody (SantaCruz). The cells were washed with 0.2% gelatin in phosphatebu�ered saline (PBS-gelatin) and incubated with biotin-conjugated donkey anti-rabbit secondary antibody (Amer-sham) for 1 h at room temperature. The coverslips werewashed and incubated for 45 min with Streptavidin-FITC(Amersham). The cells were washed, covered with mountingsolution and examined with a Leica confocal microscope.

Tyrosine phosphorylation of FGFR2-IIIb

The T24 cell stable transfectants grown in the presence ofG418 were washed once in PBS and incubated overnight at378C in serum-free DMEM/F12. Cells were left unstimulated

or were stimulated by incubation with 50 ng/ml KGF/FGF7for 20 min. Cells were lysed as described above. Lysates(1 mg) were immunoprecipitated with rabbit polyclonalantibody PK1-2, raised against a synthetic oligopeptidecorresponding to the 16 amino acids immediately down-stream from the signal peptide in the extracellular region ofFGFR2 (anti-FGFR2, 1 : 50 dilution) (Kishi et al., 1994).Immunocomplexes were collected by incubation with anti-rabbit IgG-agarose (Sigma). Immunoprecipitated proteinswere resolved by SDS±PAGE (10%) and immunoblottedwith the mouse anti-phosphotyrosine monoclonal antibody4G10 (1.5 mg/ml). Bound antibody was detected with an anti-mouse horse-radish peroxidase-conjugated secondary anti-body (Amersham), by enhanced chemiluminescence (ECL kit,Amersham).

Tritiated thymidine incorporation

Cells were seeded in DMEM/F12+ in 24-well plates at adensity of 26104 cells/well. They were incubated for 18 h at378C, washed twice with PBS Ca2+/Mg2+ and then incubatedfor 36 h in serum-free medium. 1% FCS was added and thecells were incubated for a further 18 h. Cells were labelled byincubation for 4 h with 1 mCi/well 3H-labelled thymidine(ICN, Orsay, France), washed twice with PBS Ca2+/Mg2+,®xed in cold 10% trichloracetic acid, and washed twice withwater. The cells were then dried, and were lysed in 0.1 M

sodium hydroxide. Incorporated thymidine was determinedusing a scintillation counter.

Tumour formation in nude mice

Six- to 8-week-old female Swiss nu/nu mice weighing 20 to25 g, bred in the animal facilities of the Curie Institute, Paris,France, were used for the assays. They were kept in speci®edpathogen-free conditions. Their care and housing were inaccordance with the institutional guidelines of the FrenchEthical Committee (MinisteÁre de l'Agriculture et de la ForeÃt,Direction de la Sante et de la Protection Animale, Paris,France) and were supervised by authorized investigators.Cells were suspended in DMEM/F12 at a concentration of5.107 cells/ml. Each nude mouse was injected subcutaneouslytwice in the dorsal ¯ank region (0.2 ml; 107 cells/site). Cellsfrom each clone were injected into six mice at two sites (12injections for each cell population). Tumour formation wasmonitored for up to 21 weeks. Tumours were measured withVernier calipers, and two perpendicular diameters were usedto estimate tumour volume by the formula ab2/2, where a isthe larger and b the smaller diameter. The animals weresacri®ced after 21 weeks. Half the tumours were frozen at7808C for RNA extraction, and the other half were used forhistological examination.

Loss of heterozygosity and homozygous deletion analysis of theFGFR2 gene

Genomic DNA from cell lines, normal tissue and thecorresponding tumour from the patient were prepared aspreviously described (Cappellen et al., 1997). The FGFR2gene contains two informative microsatellites (Cappellen etal., submitted). Loss of heterozygosity (LOH) at the FGFR2locus was determined as described, using these two markers(Cappellen et al., 1997). Homozygous deletions wereidenti®ed by a PCR assay. A fragment containing exon 8of the FGFR2 gene was coampli®ed with a genomicfragment of either GAPDH or PSA as control. TheGAPDH and PSA genes map to chromosomal regions thatare infrequently a�ected by allelic loss in bladdercarcinomas. A 202 bp FGFR2 fragment (FGFR2-F1/FGFR2-R1) was coampli®ed with a 136 bp GAPDH

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7241

fragment. A 344 bp FGFR2 fragment (FGFR2-F2/FGFR2-R1) was coampli®ed with a 255 bp PSA fragment (PSA-F/PSA-R). The primers used were asfollows: FGFR2-F1: 5'-AGAAGTGCTGGCTCTGTTCA-3'; FGFR2-R1: 5'-TCTC-CCAAAGCACCAAGTCT-3' GAPDH-F: 5'-TGGGGTGG-TGAATACCATGT-3'; GAPDH-R; 5'-AAGGCATGGCTG-CAACTGAA-3'; FGFR2-F2: 5'-GCCGCT-GTTTAGACG-TAATG-3'; FGFR2-R2: 5'-TCTCCCAAAGCACCAAGT-CT-3'; PSA-F: 5'-AGGCTGGGGCAGCATTGAAC-3';PSA-R: 5'-CACCTTCTGAGGGTGAACTTG-3' PCR wasperformed with 50 ng of genomic DNA and the number ofcycles was selected so as to be in the exponential part of thetwo PCR reactions (i.e. 23 cycles). PCR products wereanalysed as described in the reverse transcription-PCRsection.

Sequencing of the 5' region of the FGFR2 gene and methylationanalysis

Screening of a human PAC genomic library led to theisolation of a clone containing the 5' region of the FGFR2gene. Two kb of this clone were sequenced and the putativepromoter region was identi®ed by comparison with thepromoter sequence of the mouse gene (Avivi et al., 1992). Aquantitative PCR assay, adapted from that of Singer-Sam etal. (1990), based on the inability of certain restrictionenzymes to cut methylated DNA, was used to assess themethylation status of the CpG island of the 5' region of theFGFR2 gene. DNA (1 mg) from each tissue sample or cell linewas cut with ten units of A¯II and ten units of MspI(methylation-insensitive) or SacII, CfoI, EagI, SmaI or HpaII(all ®ve methylation-sensitive) for 16 h at 378C (258C forSmaI). Restriction enzymes were then heat-inactivated.Digested DNA (50 ng) was ampli®ed with primers corre-sponding to sequences ¯anking region 1 from the putativepromoter of the FGFR2 gene (F-5'-AGTCACTCAGC-GAGCCGTTGG-3' and R-5'-GGGTCACGCCCCCACT-GAAAC-3') or ¯anking region 2, from exon 1 (F-5'-CGAGCAAAGTTTGGTGGAGG-3' and R-5'-AGCA-ACTCCAAACGCAGAAGAGT-3') (Figure 6).

The reagents were as described above, except that 5%glycerol and 5% DMSO were added to facilitate amplifica-tion of the GC-rich fragment. The linear range of this PCR-assay was determined by plotting cycle curves and 23 cycles

were performed for the quantitative analyses. There was anegative control for ampli®cation in each independentexperiment, with the template DNA omitted from the PCRreaction.

SSCP analysis

cDNA, obtained as described above (see `Reverse transcrip-tion-PCR analysis' section) was ampli®ed by PCR using a setof 15 overlapping primer pairs covering the entire codingregion of FGFR2-IIIb. Thirty-®ve cycles of radioactive PCRwere performed using 10 ng of reverse-transcribed RNA inthe presence of [a33P]dCTP. The PCR products were analysedas described elsewhere (Ahomadegbe et al., 1995). Aberrantlymigrating bands were excised and reampli®ed with the sameset of primers. The PCR products were sequenced on bothstrands, using a cycle-sequencing kit (Pharmacia). Thesequencing primers used were the same as those used forPCR±SSCP.

AbbreviationsFGFR2-IIIb, ®broblast growth factor receptor 2-IIIb;PCR, polymerase chain reaction

AcknowledgementsWe would like to thank Dr Yutaka Hattori and DrMasaaki Terada, for providing the K-samIIC1/FGFR2-IIIb cDNA, Dr Jennifer Southgate for the gift of celllysates, Dr Marie-Aude LefreÁ re Belda for the interpretationof the histology, Dr Jean-Charles Ahomadegbe, MichelBarrois and Dr He leÁ ne Feracci for helpful advice and DrJacqueline Jouanneau for critical reading of the manu-script. This work was supported by the Association ClaudeBernard, Universite Paris XII, AP-HP (PHRC no:AOA94015), CNRS, Ligue Contre le Cancer (Comite deParis), the GEFLUC and NIH grant 2RO1 CA 49417-06.D Ricol was awarded a fellowship from the MinisteÁ re del'Enseignement Supe rieur et de la Recherche, D Cappellenfrom ARC and S Gil Diez de Medina from the LigueContre le Cancer-Comite du Val de Marne.

References

Aaronson SA. (1991). Science, 254, 1146 ± 1153.Ahomadegbe JC, Barrois M, Fogel S, Le Bihan ML, Douc-

Rasy S, Duvillard P, Armand JP and Riou G. (1995).Oncogene, 10, 1217 ± 1227.

Avivi A, Skorecki K, Yayon A and Givol D. (1992).Oncogene, 7, 1957 ± 1962.

Becker D, Meier CB and Herlyn M. (1989). EMBO J., 8,3685 ± 3691.

Cappellen D, Gil Diez de Medina S, Chopin D, Thiery JPand Radvanyi F. (1997). Oncogene, 14, 3059 ± 3066.

Carstens RP, Eaton JV, Krigman HR, Walther PJ andGarcia-Blanco MA. (1997). Oncogene, 15, 3059 ± 3065.

Chesi M, Nardini E, Brents LA, Schrock E, Ried T, KuehlWM and Bergsagel PL. (1997). Nature Genet., 16, 260 ±264.

Colvin JS, Bohne BA, Harding GW, McEwen DG andOrnitz DM. (1996). Nature Genet., 12, 390 ± 397.

Cross M and Dexter TM. (1991). Cell, 64, 271 ± 280.Dell KR and Williams LT. (1992). J. Biol. Chem., 267,

21225 ± 21229.De Moerlooze L and Dickson C. (1997). Curr. Opin. Genet.

Dev., 7, 378 ± 385.Deng C, Wynshaw-Boris A, Zhou F, Kuo A and Leder P.

(1996). Cell, 84, 911 ± 921.

Feng S, Wang F, Matsubara A, Kan M and McKeehan WL.(1997). Cancer Res., 57, 5369 ± 5378.

Fenig E, Wieder R, Paglin S, Wang H, Persaud R,Haimovitz-Friedman A, Fuks Z and Yahalom J. (1997).Clin. Cancer Res., 3, 135 ± 142.

Finch PW, Rubin JS, Miki T, Ron D and Aaronson SA.(1989). Science, 245, 752 ± 755.

Gil Diez de Medina S, Chopin D, El Marjou A, Delouve e A,La Rochelle WJ, Hoznek A, Abbou C, Aaronson SA,Thiery JP and Radvanyi F. (1997). Oncogene, 14, 323 ±330.

Gil Diez de Medina S, Popov Z, Chopin DK, Southgate J,Tucker GC, Delouve e A, Thiery JP and Radvanyi F.(1999). Oncogene, in press.

Giroldi LA, Bringuier PP and Schalken JA. (1994). InvasionMetastasis, 14, 71 ± 81.

Green PJ, Walsh FS and Doherty P. (1996). Bioessays, 18,639 ± 646.

Guilford P, Hopkins J, Harraway J, McLeod M, McLeod N,Harawira P, Taite H, Scoular R, Miller A and Reeve AE.(1998). Nature, 392, 402 ± 405.

Haber D and Harlow E. (1997). Nature Genet., 16, 320 ± 322.Halaban R, Ghosh S and Baird A. (1987). In Vitro Cell. Dev.

Biol., 23, 47 ± 52.

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7242

Halaban R, Kwon BS, Ghosh S, Delli Bovi P and Baird A.(1988). Oncogene Res., 3, 177 ± 186.

Hattori Y, Odagiri H, Nakatani H, Miyagawa K, Naito K,Sakamoto H, Katoh O, Yoshida T, Sugimura T andTerada M. (1990). Proc. Natl. Acad. Sci. USA, 87, 5983 ±5987.

Ishiwata T, Friess H, Buchler MW, Lopez ME and Korc M.(1998). Am. J. Pathol., 153, 213 ± 222.

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 and Williams LT. (1993). Adv. Cancer Res., 60,1 ± 41.

Johnson MR, Valentine C, Basilico C and Mansukhani A.(1998). Oncogene, 16, 2647 ± 2656.

Kandel J, Bossy-Wetzel E, Radvanyi F, Klagsbrun M,Folkman J and Hanahan D. (1991). Cell, 66, 1095 ± 1104.

Kishi T, Yoshida T and Terada M. (1994). Biochem. Biophys.Res. Commun., 202, 1387 ± 1394.

Kornmann M, Ishiwata T, Beger HG and Korc M. (1997).Oncogene, 15, 1417 ± 1424.

Lamm DL and Torti M. (1996). CA Cancer J. Clin., 46, 93 ±112.

LaRochelle WJ, Dirsch OR, Finch PW, Cheon H-G, MayM,Marchese C, Pierce JH and Aaronson SA. (1995). J. Cell.Biol., 129, 357 ± 366.

Mao L. (1998). J. Nat. Cancer Inst., 90, 412 ± 414.Mattei MG, Moreau A, Gesnel MC, Houssaint E and

Breathnach R. (1991). Hum. Genet., 87, 84 ± 86.Matsubara A, Kan M, Feng S and McKeehan WL. (1998).

Cancer Res., 58, 1509 ± 1514.McKeehan WL, Wang F and Kan M. (1998). Prog. Nucleic

Acid Res. Mol. Biol., 59, 135 ± 176.Miki T, Bottaro DP, Fleming TP, Smith CL, Burgess WH,

Chan AM and Aaronson SA. (1992). Proc. Natl. Acad.Sci. USA, 89, 246 ± 250.

Muller WJ, Lee FS, Dickson C, Peters G, Pattengale P andLeder P. (1990). EMBO J., 9, 907 ± 913.

Ohta M, Inoue H, Cotticelli MG, Kastury K, Ba�a R,Palazzo J, Siprashvili Z, Mori M, McCue P, Druck T,Croce CM and Huebner K. (1996). Cell, 84, 587 ± 597.

Orr-Urtreger A, Bedford MT, Burakova T, Arman E,Zimmer Y, Yayon A, Givol D and Lonai P. (1993). Dev.Biol., 158, 475 ± 486.

Radvanyi F, Christgau S, Baekkeskov S, Jolicoeur C andHanahan D. (1993). Mol. Cell. Biol., 13, 4223 ± 4232.

Reed JA, McNutt NS and Albino AP. (1994). Am. J. Pathol.,144, 329 ± 336.

Rubin JS, Osada H, Finch PW, Taylor WG, Rudiko� S andAaronson SA. (1989). Proc. Natl. Acad. Sci. USA, 86,802 ± 806.

Rubin JS, Bottaro DP, Chedid M, Miki T, Ron D, CunhaGR and Finch PW. (1995). Epithelial-MesenchymalInteraction in Cancer. Goldberg D. and Rosen E. (eds).BirkhaÈ user Verlag: Basel/Switzerland, pp. 191 ± 214.

Shackleford GM, MacArthur CA, Kwan HC and VarmusHE. (1993). Proc. Natl. Acad. Sci. USA, 90, 740 ± 744.

Schweigerer L, Neufeld G and Gospodarowitcz D. (1987). J.Clin. Invest., 80, 1516 ± 1520.

Sigalas I, Calvert AH, Anderson JJ, Neal DE and Lunec J.(1996). Nature Med., 2, 912 ± 917.

Singer-Sam J, Grant M, LeBon JM, Okuyama K, ChapmanV, Monk M and Riggs AD. (1990). Mol. Cell. Biol., 10,4987 ± 4989.

Sporn MB and Roberts AB. (1988). Nature, 332, 217 ± 219.Su WS, Kitagawa M, Xue N, Xie B, Garofalo S, Cho J, Deng

C, Horton WA and Fu XY. (1997). Nature. 386, 288 ± 291.Wang Y and Becker D. (1997). Nature Med., 3, 887 ± 893.Xiao S, Nalabolu SR, Aster JC, Ma J, Abruzzo L, Ja�e ES,

Stone R, Weissman SM, Hudson TJ and Fletcher JA.(1998). Nature Genet., 18, 84 ± 87.

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 andMcKeehan WL. (1993). Mol. Cell. Biol., 13, 4513 ± 4522.

Yayon A, Ma YS, Safran M, Klagsbrun M and Halaban R.(1997). Oncogene, 14, 2999 ± 3009.

Role of FGFR2-IIIb in bladder tumoursD Ricol et al

7243