Germinal activating mutations of FGFR3 are responsible for several forms of dwarfism due to the inhibitory effect of FGFR3 on bone growth. Surprisingly, identical somatic activating mutations have been found at the somatic level in tumours: at high frequency in benign epithelial tumours (seborrheic keratosis, urothelial papilloma) and in low-grade, low-stage urothelial carcinomas, and at a lower frequency in other types of urothelial carcinoma, in cervix carcinoma, and in haematological cancer, multiple myeloma. FGFR3 exists as two isoforms, FGFR3b and FGFR3c, differs in ligand specificity and tissue expression. FGFR3b is the main form in epithelial cells and derived tumours, whereas FGFR3c is the main form in mesenchyme-derived cells and multiple myeloma. Several lines of evidence suggest that mutated FGFR3c has transforming properties. Although mutated FGFR3b is mostly found in benign epithelial tumours or carcinomas of low malignant potential, we present evidence here that mutated FGFR3b is oncogenic. All bladder tumours presenting FGFR3 mutations expressed this receptor more strongly than normal urothelium or non-mutated tumours. NIH-3T3 cells transfected with a mutated form of FGFR3b—FGFR3b-S249C, the most common mutation in bladder tumours—presented a spindle-cell morphology, grew in soft agar and gave rise to tumours when xenografted into nude mice. We identified one line of 17 bladder cell lines tested (MGH-U3) that expressed a mutated form of FGFR3b, FGFR3b-Y375C. We showed using siRNA and SU5402, an FGFR inhibitor, that the tumour properties of MGH-U3 depended on mutated receptor activity. Thus, in two different models, mutated FGFR3b presents oncogenic properties.
FGFR3 (fibroblast growth factor receptor 3) belongs to a family of structurally related tyrosine kinase receptors (FGFR1–4) involved in many aspects of embryogenesis and tissue homeostasis. These receptors regulate various biological processes, including proliferation, differentiation, migration and apoptosis ( 1 , 2 ). They are also involved in pathological conditions and many studies have stressed their importance in developmental genetic diseases and cancer ( 3 , 4 ). FGFRs have an extracellular ligand-binding domain composed of two or three immunoglobulin-like domains, a transmembrane region and a cytoplasmic domain with tyrosine kinase activity. Ligand binding to the extracellular domain leads to FGFR dimerization, inducing receptor activation by transphosphorylation of intracellular tyrosines. The intracellular domain then interacts with, and phosphorylates various intracellular signalling proteins ( 5 ).
The FGFR3 gene is composed of 19 exons. Exclusive alternative splicing events between exons 8 and 9, encoding two different variants of the second half of the juxtamembrane Ig-like domain, generate two isoforms: FGFR3b and FGFR3c ( 2 , 6 , 7 ). This alternative mRNA splicing is tissue-specific: FGFR3b is mainly expressed in epithelial cells whereas FGFR3c is predominantly found in cells derived from the mesenchyme, including chondrocytes. This exclusive splicing determines which ligand-binding region is present and therefore influences ligand specificity ( 2 , 8 , 9 ).
Specific germline point mutations in various domains of FGFR3 are associated with autosomal dominant human skeletal disorders. These disorders include craniosynostoses and chondrodysplasias, such as (in increasing order of clinical severity) hypochondroplasia, achondroplasia, severe achondroplasia with developmental delay and acanthosis nigricans (SADDAN), and thanatophoric dysplasia (TD)—a lethal form of dwarfism ( 10 , 11 ). All these mutations generate receptors auto-activated by various mechanisms, depending on the position and nature of the nucleotide substitution ( 4 , 10 ). Biochemical analyses have established that the severity of the phenotype developed depends on the type of FGFR3 mutation and correlates with the degree of phosphorylation of the receptor ( 12 – 14 ). The most severe forms of dwarfism, TD and SADDAN, are caused by two different types of mutation: mutations to cysteine (R248C, S249C G370C, S371C, Y373C) in the extracellular region, which mediates receptor dimerization through intermolecular disulphide bond formation, and mutations affecting the kinase domain (K650E, K650M), which stimulate enzymatic activity, presumably by stabilizing the non-inhibitory conformation of the kinase regulatory loop. In contrast, achondroplasia, which is non-lethal and associated with less severe dwarfism, results almost exclusively from a G380R substitution in the transmembrane domain ( 12 – 14 ).
The association of germline FGFR3 activating mutations with different forms of dwarfism in human, in vitro studies in murine chondrocytes and loss- and gain-of-function studies in mouse models have provided evidence that activated FGFR3c inhibits normal bone growth. It seems to exert its inhibitory effects by restricting chondrocyte proliferation and modifying differentiation by triggering premature apoptosis in the epiphysal growth plate ( 4 ).
In sharp contrast with this inhibitory role in bone growth, mutated FGFR3 has also been suggested to play an oncogenic role, as the same activating point mutations causing chondrodysplasia are present in various types of tumour. Indeed, the same activating point mutations that are responsible for the severe forms of dwarfism, TD and SADDAN, have been identified in a haematological cancer, multiple myeloma, and carcinomas of the bladder and cervix ( 15 – 17 ). The frequency of FGFR3 mutations is low in multiple myelomas and cervix carcinomas (<2% of tumours present these mutations), but much higher in bladder carcinomas (∼50%) ( 18 ). The transforming properties of mutated FGFR3c, the main FGFR3 isoform expressed in multiple myeloma, have been well documented. Mutated FGFR3c (FGFR3c-K650E and FGFR3c-Y373C) can induce the malignant transformation of NIH-3T3 fibroblasts ( 19 – 21 ). In addition, FGFR3c-K650E has been shown to produce lymphoid cancers in mice ( 22 ). In contrast, only one mutated form of FGFR3b, FGFR3b-G384D, has been studied ( 21 ). In this study FGFR3b-G384D did not transform NIH-3T3. However, we could not definitely conclude that the b-form has no transforming properties as FGFR3c G382D, the c-form of the FGFR3b-G384D mutation, was also shown to be non-transforming ( 21 ). Some doubt has been cast on the transformation potential of mutated FGFR3b because (i) a mutated FGFR3b, FGFR3b-S249C, targeted to the mouse epidermis induced only benign tumours, with no sign of malignancy ( 23 ); (ii) FGFR3b mutations occur at high frequency (40%) in seborrheic keratosis, the most common benign epidermal tumour in humans, which has no malignant potential ( 23 ). In addition, FGFR3 mutations are not found in malignant epidermal tumours, basal cell carcinomas and spindle-cell carcinomas ( 23 , 24 ); (iii) in bladder carcinomas, FGFR3b mutations are mainly found in low-stage, low-grade tumours (TaG1-G2), which rarely progress ( 25 , 26 ). Furthermore, these mutations are also found at high frequency (75%) in urothelial papilloma, a benign tumour with no malignant potential ( 27 ). Hence, the exact role of mutated FGFR3b in epithelial tumour progression is far from understood. Given the high frequency of FGFR3b mutation in bladder carcinomas and benign skin tumours, we felt it was important to investigate further the transforming properties of mutated FGFR3b.
In this study, we first pointed out that FGFR3 mutations were associated with receptor overexpression in bladder carcinoma. We then demonstrated that FGFR3b-S249C, the most common mutated form of FGFR3 in bladder tumours, was able to transform NIH-3T3 cells, inducing their anchorage-independent growth and tumour formation when injected subcutaneously into nude mice. Finally, we identified a bladder epithelial cell line expressing a mutated form of FGFR3b, FGFR3b-Y375C, and demonstrated in this cell line that FGFR3 inhibition, using a FGFR kinase inhibitor, or FGFR3 depletion, using FGFR3 siRNA, decreased cell growth and inhibited anchorage-independent growth. All these results are consistent with mutated FGFR3b having the transforming properties classically attributed to oncogenes.
Materials and methods
Normal urothelium ( n = 5) and transitional cell carcinomas of the bladder ( n = 77) were obtained as described elsewhere ( 28 ). Tumours were staged according to the TNM classification ( 29 ) and graded according to World Health Organization recommendations. The 77 tumour samples used for RNA analysis contained more than 80% malignant cells, as determined by histological analysis of adjacent sections. The tumours studied belong to a group of tumours previously analysed for FGFR3 mutations ( 25 ).
RNA extraction and RT–PCR analysis
Total RNA was extracted by caesium chloride centrifugation for tumour samples ( 30 ) and with Trizol® for cell-line samples. It was used as a template for first-strand cDNA synthesis by random priming, as described previously ( 31 , 32 ). The amount of FGFR3b mRNA was determined by semi-quantitative radioactive RT–PCR, using TBP (TATA-binding protein) as an internal control, as described previously ( 28 ). The primers used were 5′-AGTGAAGAACAGTCCAGACTG-3′ and 5′-CCAGGAAATAACTCTGGCTCAT-3′ for FGFR3b, 5′-AGTGAAGAACAGTCCAGACTG-3′ and 5′-CCAGGAAATAACTCTGGCTCAT-3′ for TBP. In each case, the primers were designed to bind two different exons, preventing errors in mRNA quantification due to genomic DNA contamination. The number of cycles was selected so as to be in the exponential part of the PCR (25 cycles). The PCR products were subjected to electrophoresis in 8% polyacrylamide gels. Signals were quantified with a Molecular Dynamics 300 PhosphorImager (Molecular Dynamics, Sunnyvale, CA, USA).
The coding sequence of the human FGFR3b cDNA was amplified by PCR from a normal urothelium, using the primers 5′-CGGGGCTGCCTGAGGAC-3′ and 5′-TGCTAGGGACCCCTCACATT-3′. The PCR product was inserted into a cytomegalovirus promoter-driven pcDNA/Neo expression vector (Invitrogen, San Diego, CA, USA). The S249C mutation was subsequently introduced into this FGFR3b cDNA as follows: a 729 bp region of the FGFR3b cDNA carrying the S249C mutation was amplified by PCR from bladder carcinoma, using the primers 5′-CGTCGTGGAGAACAAGTTTGGCAG-3′ and 5′-CCGAGACAGCTCCCATTTG-3′. The XhoI–Aor5H1 fragment containing the mutation was excised from the PCR product and inserted in place of the corresponding sequence in the non-mutated FGFR3b construct. All the segments were checked by sequencing both strands.
Cell culture and transfection
The mouse fibroblastic cell line NIH-3T3 (kindly provided by Dr S. Bellusci) was cultured in DMEM with 10% newborn calf serum, 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. The human bladder carcinoma cell lines T24 and MGH-U3 ( 33 ) were cultured in DMEM/F12 with 10% fetal calf serum (FCS) and 2 mM glutamine.
For stable transfection, NIH-3T3 cells were washed three times in serum-free medium and resuspended at a density of 10 8 cells/ml. We transferred 250 μl of this suspension to a Gene Pulser cuvette (Bio-Rad Laboratories, Hercules, CA, USA) and added 20 μg of the appropriate plasmid. Cells were pulsed at 500 μF and 250 V for 10 ms, then left for 10 min at room temperature and carefully transferred to 3–5 ml of DMEM/10% FCS in a 55 mm culture dish. After 2 days, we added G418 (400 μg/ml), and drug selection was continued for 2 weeks. Single colonies were ring-cloned and expanded.
For siRNA transfection, MGH-U3 cells were seeded at a density of 8000 cells/cm 2 in 6-well or 24-well plates and transfected with 200 nM siRNA, using Oligofectamine® reagent (Invitrogen, Cergy Pontoise, France) according to the manufacturer's protocol. FGFR3 siRNAs and nonsense siRNA used as a control were chemically synthesized (Ambion, Huntingdon, UK). The sequences of the sense strands were as follows: control siRNA, 5′-GGCAAGAUUCUUCUCGUUGTT-3′; FGFR3 siRNA1, 5′-GCCUUUACCUUUUAUGCAATT-3′; and FGFR3 siRNA2, 5′-GGGAAGCCGUGAAUUCAGUTT-3′. We counted cells 72 h after transfection (24-well plates) or lysed them with lysis buffer (6-well plates).
For immunoprecipitation, MGH-U3 or transfected NIH-3T3 cell lysates containing 1 mg of protein were incubated with protein A–agarose-coupled anti-FGFR3 antibodies (C-terminal polyclonal antibodies; Sigma, Saint-Quentin Fallavier, France). The immunoprecipitated proteins were then used for immunoblotting.
Transfected NIH-3T3 proteins immunoprecipitated with an anti-FGFR3 antibody, prepared as described above, or proteins from siRNA-transfected MGH-U3 cells were subjected to SDS–PAGE in a 7.5% polyacrylamide gel, transferred onto a PVDF membrane and immunoblotted with anti-FGFR3 antibodies (Sigma) and horseradish peroxidase-conjugated goat anti-rabbit IgG or with a horseradish peroxidase-conjugated anti-phosphotyrosine antibody (RC20; Cell Signalling Technology, Ozyme, Montigny Le Bretonneux, France). Binding was detected by enhanced chemiluminescence (Amersham, Saclay, France) according to the kit manufacturer's instructions.
Thymidine incorporation assay
NIH-3T3 cells (2 × 10 4 ) were seeded in 24-well plates containing DMEM supplemented with 10% NCS. Cells were incubated for 24 h and then serum-starved for 24 h. Cells were stimulated by incubation with 10% FCS for further 18 h. We then added 1 μCi of [ 3 H]thymidine, incubated the plates for a further 6 h, fixed the cells in 10% trichloroacetic acid, washed them with water and lysed them with 0.1 N NaOH. Total incorporated radioactivity was counted with a micro-beta scintillation counter (LKB; Perkin Elmer Life Sciences, Courtaboeuf, France).
Soft agar assay
For NIH-3T3, we added 30 000 transfected cells, in triplicate, to each well of a 12-well plate containing DMEM supplemented with 10% NCS and solidified with agar. Cells expressing the wild-type FGFR3 were cultured in the presence or absence of 20 ng/ml FGF1 + 50 μg/ml heparin in agar and in culture medium. The same amounts of FGF1 and heparin were added to the 0.5 ml of culture medium twice weekly. For MGH-U3 and the depletion experiments, we added 20 000 non-transfected or siRNA-transfected cells, to triplicate wells of 12-well plates containing MEM supplemented with 10% FCS and agar. For MGH-U3 and the inhibition experiments, cells were cultured in the presence or absence of 40 μM SU5402, a FGFR tyrosine kinase inhibitor ( 34 ), in agar and in culture medium. The same amount of inhibitor was added to the culture medium weekly.
The plates were incubated for 21 days, and colonies with diameters greater than 50 μm were then scored as positive, using a phase-contrast microscope equipped with a measuring grid.
Tumour formation in nude mice
Six- to eight-week-old female Swiss nu/nu mice weighing 20–25 g were raised in the animal facilities of the Curie Institute in specified pathogen-free conditions. Their care and housing were in accordance with the institutional guidelines of the French National Ethics Committee (Ministère de l'Agriculture et de la Forêt, Direction de la Santé et de la Protection Animale, Paris, France) and were supervised by authorized investigators. Each nude mouse was injected subcutaneously in each flank (dorsal region) with 8 × 10 5 cells/site. Cells from each clone were injected into five mice (resulting in ten injections for each cell population). Tumour formation was monitored for up to 20 days and the size of tumours was measured with Vernier calipers: two perpendicular diameters were used to estimate tumour volumes by the formula ab2 /2, where a is the largest and b the smallest diameter.
Genomic DNA and cDNA from MGH-U3, RT4 and RT112 cells were sequenced directly, on both strands, with the Big Dye Terminator Kit. The sequences of the primers used are available on request.
Strong FGFR3b expression in bladder tumours is associated with the presence of FGFR3 point mutations
Genetic studies have shown a high frequency of FGFR3 activating mutations in bladder tumours. We searched for a possible association between FGFR3 expression level and the presence of FGFR3 mutation.
We used semi-quantitative RT–PCR to analyse FGFR3b mRNA levels in a set of 77 tumours and in 5 normal urothelium samples ( Figure 1 ). The stage, grade and mutational FGFR3 status of the tumours, which had been the object of a previous study ( 25 ), are summarized in Table I . As previously reported, the non-mutated tumours were of a higher grade and higher stage than the tumours with FGFR3b mutations ( Table I ). FGFR3 mutation was detected in 31 tumours (22/31, 71% had the S249C mutation), the other 46 tumours having no detectable FGFR3 mutation. FGFR3b mRNA levels were very similar in the five normal samples (mean = 6.3, SD = 1.8), but much more variable in tumour tissues. Tumours with FGFR3 mutations had FGFR3 mRNA levels significantly higher than those encountered in normal bladder epithelium (mean = 13.6, SD = 7.4) ( P < 0.0001). The level of FGFR3 expression in mutated tumours did not depend on the nature of the mutations or the stage of tumours (data not shown). In bladder tumours with no detected FGFR3 mutations, the overall level of FGFR3b expression did not differ significantly from that observed in normal tissues (mean = 4.4, SD = 6.1) ( P = 0.13), but FGFR3b was undetectable or present at very low levels in half these tumours (23 out of 46). The RT–PCR results were confirmed by northern blots of two normal urothelia and 10 representative bladder tumours (data not shown). Thus, in bladder carcinoma, strong FGFR3 expression is correlated with the presence of a mutated FGFR3 gene.
| Stage || Grade || FGFR3 wild-type || FGFR3 mutant || FGFR3-S249C |
| Stage || Grade || FGFR3 wild-type || FGFR3 mutant || FGFR3-S249C |
The S249C mutation leads to the ligand-independent phosphorylation of FGFR3
We studied the oncogenic properties of FGFR3b-S249C, the most common mutation in bladder carcinoma, by stably transfecting NIH-3T3 cells with an expression vector encoding wild-type FGFR3 or the mutant FGFR3b-S249C, or with the vector alone as a negative control. The various clones obtained were tested for FGFR3 expression by semi-quantitative RT–PCR and western blotting with an anti-FGFR3 antibody (data not shown). None of the control clones transfected with the vector alone expressed FGFR3b and various levels of expression were observed for clones expressing wild-type FGFR3 or FGFR3b-S249C. We selected two control clones (Neo1.5 and Neo2.1), two clones expressing mutated FGFR3 (S249C1.1 and S249C1.2) and two clones expressing wild-type FGFR3 (R3b1.1 and R3b1.3). The last four of these clones were chosen as they had similar levels of FGFR3 protein ( Figure 2 ). We used immunoblotting with an anti-phosphotyrosine antibody following FGFR3 immunoprecipitation to check that mutated FGFR3b was phosphorylated in the absence of exogenous ligand in S249C1.1 and S249C1.2 cells ( Figure 2 ), as expected based on potential to form a disulphide bond in the extracellular domain of the mutated receptor. In contrast, wild-type FGFR3b showed no activation in R3b1.1 and R3b1.3 cells ( Figure 2 ), but presented the same level of phosphorylation as mutated FGFR3b in the presence of one of its ligands, FGF1 (data not shown).
FGFR3b-S249C transforms NIH-3T3 cells in vitro
We then investigated whether FGFR3b-S249C expression induced the transformation of NIH-3T3 cells. Morphological analysis of the different NIH-3T3 clones showed major changes in the two FGFR3b-S249C-expressing clones whereas cells expressing wild-type FGFR3b, even in presence of FGF1, had a phenotype identical to that of control cells ( Figure 3A and data not shown). The FGFR3b-S249C-expressing cells acquired an elongated, refractile spindle-shaped morphology, reminiscent of transformed cells.
We performed a thymidine incorporation assay with near-confluent cells, for control cells (clones Neo1.5 and Neo2.1), and NIH-3T3 cells expressing FGFR3-S249C (clones S249C1.1 and S249C1.2) or wild-type FGFR3 (clones R3b1.1 and R3b13) ( Figure 3B ). At this state of confluence, FGFR3-S249C-expressing cells grew three to four times more rapidly than control or FGFR3-expressing cells, suggesting that FGFR3-S249C expression conferred a higher proliferation rate and a loss of contact inhibition, a characteristic feature of transformed cells.
We then investigated whether FGFR3b-S249C expression induced the cell anchorage-independent growth of NIH-3T3 cells. We evaluated the ability of control cells (clones Neo1.5 and Neo2.1), and of NIH-3T3 cells expressing wild-type FGFR3 (clones R3b1.1 and R3b1.3) in the presence or absence of FGF1, and of NIH-3T3 cells expressing FGFR3b-S249C (clones S249C1.1 and S249C 1.2) to form colonies in soft agar ( Figure 3C ). The two FGFR3b-S249C-expressing clones formed anchorage-independent growing colonies after 3 weeks, whereas few colonies were observed with wild-type FGFR3b-expressing clones, with and without FGF1 stimulation, or with control clones ( Figure 3C ).
FGFR3b-S249C expression induces NIH-3T3 tumorigenicity in nude mice
Our results clearly demonstrated the transforming activity of mutated FGFR3 in NIH-3T3 cells in vitro . We, therefore, investigated the in vivo oncogenic properties of the mutated receptor by studying the ability of FGFR3b-S249C-expressing cells to induce tumour formation in nude mice following xenografting. Control cells (clones Neo1.5 and Neo2.1), and NIH-3T3 cells expressing wild-type FGFR3b (clones R3b1.1 and R3b1.3) or FGFR3b-S249C (clones S249C1.1 and S249C1.2) were injected subcutaneously into nude mouse flanks and tumour growth was measured twice per week over 20 days ( Figure 4 ). At the end of this period, all the animals injected with FGFR3b-S249C-expressing NIH-3T3 cells presented rapidly growing tumours, whereas none of the mice injected with control cells or R3b1.3 cells had developed tumours and some animals injected with R3b1.1 cells had developed smaller tumours ( Figure 4 ).
No effect on proliferation in vitro and tumorigenicity in vivo in the T24 bladder tumour cell line upon transfection of FGFR3b-S249C
Having demonstrated that FGFR3b-S249C presented all the properties classically attributed to an oncogene in NIH-3T3 cells, we investigated whether this observation was of biological significance in a bladder epithelial cell line. We therefore studied the effect of transfection with a construct encoding mutated FGFR3b in the T24 bladder tumour-derived cell line, which does not express endogenous FGFR3 and is weakly tumorigenic (it forms very slowly growing tumours when xenografted into nude mice).
The T24 cell line was stably transfected with an expression vector encoding wild-type FGFR3b or the mutant FGFR3b-S249C or with the vector alone as a negative control. We selected two control clones, two FGFR3b-S249C-expressing clones and two wild-type FGFR3b-expressing clones. No morphological change and no change in [ 3 H]thymidine incorporation was observed following FGFR3b-S249C expression in control clones, FGFR3- and FGFR3-S249C-expressing clones (data not shown). We evaluated the potential oncogenic properties of the mutated receptor in vivo by comparing the ability of FGFR3b-S249C-T24-expressing cells and T24 control cells to induce tumour formation in nude mice following xenografting, as previously described with NIH-3T3 cells. After 10 weeks, all the mice had developed small tumours but no significant difference between tumour volume was observed with the four different clones (two control clones and two FGFR3b-S249C-expressing clones; data not shown).
The bladder tumour cell line MGH-U3 expresses a mutated form of FGFR3b, FGFR3b-Y375C, which is required for its transforming properties
Hence, in contrast to what was observed in NIH-3T3 cells, we were unable to demonstrate an oncogenic role for FGFR3b-S249C in T24 cells. Rather than transfecting a line not expressing the receptor with a mutated FGFR3, we used an alternative strategy to investigate the possible involvement of mutated FGFR3b in bladder epithelial cell tumorigenesis. We first searched for bladder carcinoma cell lines expressing mutated FGFR3b and then evaluated the role of the mutated receptor in cell transformation, using an FGFR inhibitor, SU5402, or FGFR3 siRNA.
We first measured FGFR3b expression levels in 16 cell lines in addition to T24 (EJ-138, HCV-29, HT-1197, HT-1376, JON53, J82, MGH-U3, RT4, RT112, SCaBER, SD48, UM-UC3, VM-CUB-1, VM-CUB-3, 253-J, 647V) and found three cell lines expressing FGFR3b (MGH-U3, RT4, RT112). We then sequenced FGFR3 cDNA and genomic DNA in these three FGFR3-expressing bladder carcinoma cell lines. Only one cell line, MGH-U3, presented a FGFR3 mutation, FGFR3b-Y375C, no wild-type FGFR3 sequence being detected ( Figure 5A ). The FGFR3b isoform expressed in epithelial cells contains two more amino acids that the FGFR3c isoform expressed in mesenchyme tissues. The FGFR3b-Y375C mutation is therefore equivalent to the FGFR3c-Y373C mutation. The Y375/373C mutation, like the S249C mutation, is thought to induce disulphide bond formation by introducing an additional cysteine in the extracellular domain of FGFR3, thereby causing constitutive activation of the receptor. We first checked the phosphorylation status of mutated FGFR3 in MGH-U3. As expected, mutated FGFR3 was phosphorylated in the absence of exogenous ligand ( Figure 5B ).
We investigated the role of the mutated receptor in MGH-U3 by evaluating the consequences of FGFR3 inhibition or depletion for MGH-U3 cell proliferation and anchorage-independent growth, using a FGFR tyrosine kinase inhibitor, SU5402 ( 34 ) or FGFR3 siRNA. We first checked that, as previously reported for FGFR2 ( 35 ), SU5402 did inhibit FGFR3 phosphorylation in the absence of exogenous ligand in MGH-U3 cells ( Figure 5B ). We then examined the ability of two different FGFR3 siRNAs to reduce endogenous levels of FGFR3 protein in MGH-U3 cells. Immunoblot analysis showed a dose-dependent manner, with complete inhibition observed with 200 nM siRNA1 and 80–90% inhibition observed with 200 nM siRNA2 (data not shown and Figure 5C ). The effects on cell proliferation of the FGFR3 tyrosine kinase inhibition with SU5402 or decrease in FGFR3 expression following siRNA treatment are shown in Figure 6A . Cell counts demonstrated 40–60% growth inhibition in bladder cancer cells, in both cases. Finally, we assessed the effect of SU5402 or FGFR3 siRNA treatment on the ability of MGH-U3 cells to form colonies on soft agar. siRNAs-transfected cells or SU5402-treated cells formed one-fifth to one-third as many colonies as control cells ( Figure 6B ). Thus, decreases in FGFR3 expression or activity lead to the inhibition of cell growth and cell transformation in MGH-U3 bladder cancer cells in vitro .
The identification of FGFR3b mutations in carcinomas (bladder and cervix) strongly suggested oncogenic properties of the FGFR3b mutated receptor. However, no functional studies have examined the role of mutated FGFR3b in cell transformation. In this work, we clearly demonstrated the transforming properties of mutated FGFR3b in both fibroblastic NIH-3T3 cells and the bladder carcinoma-derived cell line, MGH-U3. Indeed, the expression of a mutated form of FGFR3b, FGFR3b-S249C, in NIH-3T3 cells induced the transformation of these cells: transfected cells acquired a spindle-cell morphology, a higher proliferation rate, the ability to form colonies on soft agar and to give rise to tumours when xenografted into nude mice. Consistent with what has been shown for FGFR3c ( 20 ), the expression of wild-type FGFR3, even when activated by one of its ligands, did not lead to the transformation of NIH-3T3 cells, suggesting that the two receptors (mutated and ligand-stimulated) may be able to activate different transduction pathways accounting for the observed differences in transforming activity. The transforming properties of the mutated receptor, as demonstrated in an immortalized fibroblastic cell line, are certainly of biological relevance as they were also demonstrated in a bladder tumour cell line, MGH-U3, expressing a mutated FGFR3b, FGFR3b-Y375C. The inactivation of the receptor's enzyme activity with an FGFR tyrosine kinase inhibitor, SU5402, or the down-regulation of FGFR3 expression with an siRNA against FGFR3, strongly inhibited the proliferation and growth of MGH-U3 cells in the absence of attachment. In contrast, FGFR3b-S249C expression in bladder carcinoma-derived T24 cells did not increase tumorigenic potential in vitro or in vivo . There are two possible explanations for the lack of transforming activity of FGFR3b-S249C in the T24 cell line. The necessary elements of the mutated FGFR3b downstream signalling pathways may be absent in the T24 cell line, leading to a lack of effect of introduced mutations in FGFR3b. Alternatively, the signalling pathway activated by mutated FGFRb may already be activated in the T24 cell line, with no additional activation observed following the introduction of mutated FGFR3b. The fact that HRAS is mutated in T24 cell line ( 36 ), is in favour of this second possibility as it has recently been shown that FGFR3 and HRAS mutations are mutually exclusive events in bladder tumours ( 37 ), suggesting that HRAS is in the same pathway as mutated FGFR3. Rather than introducing mutated FGFR3b in a bladder carcinoma cell line, it would be interesting to introduce mutated FGFR3b in a normal bladder cell line, such as NHU cells ( 38 ). However, it is not clear that the expression of mutated FGFR3b alone would be sufficient to transform normal urothelial cells. Additional events might be necessary for mutated FGFR3b to induce epithelial cell transformation.
We identified FGFR3b expression in only 3 of the 17 cell lines examined. This proportion is very small, given that FGFR3 expression is detected in more than 50% of bladder carcinomas tested. The low level of FGFR3 expression in bladder cell lines may be due to a loss of expression of FGFR3 upon the culture in vitro of bladder epithelial cells. Indeed, we observed that the in vitro culture of normal urethral or bladder epithelial cells leads to lower levels of FGFR3 expression than observed in vivo (data not shown). This may explain why, despite the demonstration of FGFR3b transforming activities in immortalized fibroblastic cell lines, NIH-3T3, and the bladder carcinoma-derived cell line, FGFR3b mutations are found mainly in tumours with no malignant potential or in tumours of low malignant potential. Indeed, FGFR3b expression may require the differentiation of urothelial cells and, as the progression of a mutated tumour requires the loss of a certain degree of differentiation, this would be incompatible with FGFR3 expression in most cases.
The predominant association of FGFR3 mutations in epithelial cells with tumours with no malignant potential (seborrheic keratoses and bladder papillomas) or tumours with low malignant potential (low-grade Ta tumours) suggests that FGFR3 mutations may be involved at a very early stage of carcinogenesis and that these mutations are not required at later stages. However, as high levels of FGFR3 expression are observed in both low-stage, low-grade bladder tumours and in all invasive tumours with mutated receptors, it seems reasonable to assume that the mutated receptor continues to be important for tumour survival, even at later stages of tumour progression. This observation is important as it clearly identifies mutated FGFR3 as an important therapeutic target for invasive tumours with mutated receptors.
We thank Jennifer Southgate for providing the human bladder cell line extracts used for RNA and DNA preparation. This work was supported by the CNRS, the Institut Curie and the Ligue Nationale Contre le Cancer - Comité d'lle de France.
Conflict of Interest Statement : None declared.
UMR 144, CNRS—Institut Curie, 26 rue d'Ulm, 75248 Paris Cedex 05, France, 1INSERM 0337 and Service d'Urologie, Centre Hospitalier Universitaire Henri Mondor, AP-HP, Université Paris XII, 94010 Créteil Cedex, France