Targeting the Extracellular Domain of Fibroblast Growth Factor Receptor 3 with Human Single-Chain Fv Antibodies Inhibits Bladder Carcinoma Cell Line Proliferation

The fibroblast growth factors (FGF) represent one of the largest families of polypeptide growth and differentiation factors for mesodermal and epithelial cells (1). FGFs have been implicated in multiple biological processes during embryo development, wound healing, hematopoiesis, and angiogenesis. In addition, FGFs have been shown to increase the invasiveness of a variety of tumors from prostate, bladder, kidney, breast, and pancreas (2–5).

Although more than 20 FGFs with different effects on various cells have been reported, only five FGF receptors (FGFR) have been described (6, 7). These receptors share 55% to 72% homology at the protein level. The FGFR structure consists of an extracellular ligand-binding domain, a transmembrane domain, and an intracellular kinase domain. The ligand-binding domain contains three different immunoglobulin-like domains (called immunoglobulin I, immunoglobulin II, and immunoglobulin III). The ligand domain is followed by a single transmembrane domain and a cytoplasmic split tyrosine kinase domain in the intracellular domain. Alternative splicing of the mRNAs corresponding to FGFR1-3 gives place to two isoforms, α and β. Only isoform α, which contains the three immunoglobulin domains, has been found for FGFR3. Alternative splicing of FGFR3 transcripts involving the COOH-terminal domain of the immunoglobulin III domain yields two isoforms, IIIb and IIIc. The two variants show different binding affinities: the IIIc variant is more promiscuous and binds a wide range of FGFs, including FGF1, FGF2, FGF4, and FGF9; on the other hand, the IIIb variant preferentially binds FGF1 and, at a lower extent, FGF8 and FGF9 (8). After binding to FGFRs in the presence of heparin, FGFs induce receptor dimerization, resulting in autophosphorylation of the intracellular kinase domain and a downstream activation of intracellular signaling cascades (9). Following activation of the ligand receptor complex, various signal transduction pathways are initiated by FGFs: elevation of intracellular calcium levels, induction of mitogen-activated protein kinase and protein kinase C pathways, stimulation of adenylate cyclase, and induction of the proto-oncogenes c-myc and c-fos (6).

Specific mutations have been found in FGFR3 that lead to the activation of its tyrosine kinase activities and to different syndromes related with bone development, multiple myeloma (10, 11), cervical carcinomas (12, 13), and bladder carcinomas (12, 14–17). There are recent observations that FGFR signaling is absolutely necessary for the survival of prostate cancer cells in vitro (18). Recently, FGFR3 has been proposed as a possible therapeutic target for multiple myeloma (10, 19). Although the presence of activating mutations is well documented, little was known about the expression of FGFR3 in tumoral tissues. Recently, after gene expression analysis by using gene chip technology, it was found that FGFR3 was overexpressed in tumoral samples corresponding to stages pT_a, pT₁ (fold change >8), and pT₂ (fold change >2) of bladder carcinoma when compared with the normal counterparts (20). The gene expression levels were confirmed at the protein level by Western blot and immunohistochemical analysis. In fact, this protein overproduction seems to occur more frequently in transitional carcinomas than activating mutations (20). All these data together suggest that FGFR3 might constitute an attractive target for therapy in urologic tumors. Bladder cancer is the second most common malignancy of the genitourinary tract. About 40% to 50% of the bladder carcinoma showed FGFR3-activating gene mutations (17, 21); the incidence was significantly bigger (80%) in superficial tumors than in invasive tumors.

With the increasing interest in FGFR3 as a therapeutic target in different neoplasias and the recent findings about its overexpression in transitional cell carcinoma, we have started the development of human antibodies for therapeutic purposes by phage display. Phage antibody display is one of the best alternatives for the production of human antibodies for research, clinical, and therapeutic applications (for reviews on this topic, see refs. 22–24). However, FGFR3 is a very elusive molecule for antibody production given the high homology between FGFR3 from mouse and human species (92%). Only recently, and by using a very large commercial Fab library with a size of 2.1 × 10¹⁰, it has been described the production of Fab fragments against a nondisclosed FGFR3 isoform (25). In our experiments, we have used two available public libraries of single-chain Fv (scFv) antibodies, Tomlinson I + J (MRC Geneservices, Cambridge, United Kingdom), which have been previously described (26). The size of both libraries is around 1.4 × 10⁸. scFvs penetrate tumors better than immunoglobulin Gs (IgG) and Fabs, are cleared more rapidly, and have greater specificity. In this report, we have developed a collection of human scFv antibodies specific for FGFR3α isoform IIIc. These antibodies reacted with the bladder carcinoma cell line RT112 by fluorescence-activated cell sorting (FACS) analysis and inhibited cellular proliferation, showing promise for further therapeutic application.

Cell lines, proteins, and antibodies. RT112 cells were obtained from the German collection of Microorganisms and Cell Cultures (DSMZ; Braunschweig, Germany). Human embryonic kidney (HEK293T) and RT112 cells were cultured in RPMI 1640 (Sigma, St. Louis, MO) supplemented with 10% fetal bovine serum and antibiotics.

The recombinant human FGFR3α(IIIc)/Fc and FGFR1α(IIIc)/Fc chimera were purchased from R&D Systems (Minneapolis, MN) and ReliaTech (Braunschweig, Germany). These proteins consist of the isoforms IIIc of the extracellular domains of FGFR3α or FGFR1α, respectively, fused to the Fc region of human IgG1. FGF9, FGF1, and epidermal growth factor (EGF) were provided by ReliaTech.

Mouse anti-FGFR3 monoclonal antibody (mAb) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Goat anti-human IgG (Fc specific), mouse anti–c-myc mAb (clone 9E10), and antitubulin were from Sigma. Horseradish peroxidase–conjugated anti–c-myc (clone 9E10) was purchased from Roche (Mannheim, Germany). Anti-6xHis was provided by NeoMarkers (Westinghouse Dr, Fremont, CA). Anti-M13 and horseradish peroxidase–conjugated anti-M13 mAbs were obtained from GE Healthcare. FITC-conjugated rabbit anti-mouse IgG was from DAKO (Glostrup, Denmark). R-phycoerythrin goat anti-mouse IgG was purchased from Molecular Probes (Eugene, OR). Mouse IgG TrueBlot was provided by eBioscience (San Diego, CA).

FGFR3 cDNA cloning and cell transfection. For cellular expression of recombinant FGFR3s, the extracellular domain of the protein was fused to the COOH-terminal Fc region of human IgG1. Vector pcDNA3.1-Fc, which contains the human Fc region, was used to clone the cDNA fragment encoding the extracellular domain of FGFR3-IIIc (amino acids 1-370) to generate pcDNA3.1-FGFR3(IIIc)WT-Fc. This plasmid was used as template to prepare the mutant FGFR3 S249C by site-directed mutagenesis using the primer 5′-CCCGCCTGCAGGATGGGCCGGTGCGGGCAGCGC-3′. The resulting plasmid, pcDNA3.1-FGFR3(IIIc)S249C-Fc, was sequenced to confirm the presence of the S249C mutation. pcDNA3.1-FGFR3(IIIc)WT-Fc and pcDNA3.1-FGFR3(IIIc)S249C-Fc were transfected into HEK293T cells with FuGENE 6 (Roche) and the expression and reactivity of the wild-type and mutant proteins were analyzed 2 days after transfection by immunoblotting with anti-FGFR3 mAb and by FACS with the specific scFvs as described below.

Selection of fibroblast growth factor receptor 3–specific single-chain Fv antibodies from phagemid libraries. The human scFv libraries Tomlinson I + J, helper phage KM13, and Escherichia coli TG1 and HB2151 were obtained from MRC Geneservices. Both libraries were grown separately and the phages mixed 1:1 for selection. Phage selection and scFv production were done on microtiter plates (Maxisorb, Nunc, Rockilde, Denmark) according to the scheme shown in Fig. 1. In the first round of selection, wells of the microplates were coated with 1 μg of FGFR3 or human IgG in PBS at 4°C overnight. Next, wells were washed with PBS and blocked with 2% MPBS (skimmed milk in PBS) for 2 hours. Then, 10¹¹ phages in 2% MPBS were incubated first for 1 hour with human IgG to remove the reactivity against the Fc fragment of the recombinant antigen, and then for 2 hours with FGFR3-coated wells. Plates were washed 10 times with PBS-0.1% Tween 20 (20 times for the subsequent rounds) and treated with 100 μL of trypsin (Sigma) for elution of the bound phages. Elution and production of phages for other rounds of panning was carried out essentially as described by Goletz et al. (26).

In other panning experiments and for a major Fc blocking efficiency, phage selection was carried out by mixing first the phage suspension with 50 μg of human IgG in 2% MPBS. After 1-hour incubation at 37°C, this mixture was directly added to FGFR3-coated wells for 2 hours at room temperature.

Phage ELISA. Flexible microtiter plates (Falcon, BD Biosciences, Franklin Lakes, NJ) were coated overnight with 0.3 μg of FGFR3, FGFR1, or human IgG in PBS. After washing thrice with PBS, plates were blocked with 2% MPBS for 2 hours at room temperature. Following additional washing, different dilutions of the purified phage suspension coming from the selection steps, in the presence of 2% MPBS, were incubated for 1 hour at room temperature. After washing, a 1:5,000 dilution of anti-M13 labeled with peroxidase in 2% MPBS was added for 1 hour and the color was developed with 3,3′,5,5′-tetramethylbenzidine substrate (Sigma). The reaction was stopped after 10 minutes by adding 1 mol/L sulfuric acid and the absorbance was read at 450 nm.

Monoclonal phage ELISA. Usually groups of 96 colonies from the second or third round of selection were picked either manually or by using an automatic picker (Q-pix, Genetix, Hampshire, United Kingdom) and grown in 100 μL of 2× TY medium containing 100 μg/mL ampicillin and 1% glucose in a 96-well plate (Sarstedt, Numbrecht, Germany) at 37°C. The next day, cultures were diluted 1:100 in the same medium and shaken for 2 hours at 37°C. Then, 25 μL of 2× TY medium containing 10⁹ KM13 helper phages were added and the culture continued for 1 hour. Bacterial cells were pelleted, resuspended in 2× TY with 100 μg ampicillin and 50 μg/mL kanamycin, and grown overnight at 30°C. Finally, the culture was centrifuged and 50 μL of the supernatant were used for the monoclonal phage ELISA in the same conditions detailed above.

Sequence analysis. Phagemid DNAs from individual colonies were amplified by PCR with the primers LMB3 5′-CAGGAAACAGCTATGAC-3′ and pHENseq 5′-CTATGCGGCCCCATTCA-3′. PCR products were purified with the QIAquick PCR purification kit (Qiagen, Hilden, Germany) and directly sequenced by using fluorescent dideoxy chain terminators and by automated sequencing (ABI7002, Applied Biosystems, Foster City, CA).

Production and purification of soluble single-chain Fv antibody and ELISA. The procedure for E. coli induction was identical to that described above except that at the exponential phase (A₆₀₀ = 0.9), E. coli HB2151 cells were induced with 2× TY containing 100 μg/mL ampicillin and 1 mmol/L isopropyl-l-thio-β-d-galactopyranoside and incubated at 30°C overnight. For the preparation of periplasmic extracts, bacteria were harvested and resuspended in 1:50 of the culture volume of ice-cold TES buffer [0.5 mol/L sucrose in 200 mmol/L Tris-HCl (pH 8.0), 0.5 mmol/L EDTA], containing 20 μg/mL benzamidine (Sigma) and 10 μg/mL soybean trypsin inhibitor (Sigma), and rapidly diluted in 0.2× TES. Cells were incubated on ice for 30 minutes and centrifuged at 16,250 × g for 10 minutes at 4°C. The supernatant containing the periplasmic fraction was dialyzed overnight at 4°C against 50 mmol/L Tris-HCl (pH 7.0) and 300 mmol/L NaCl and subjected to affinity chromatography using TALON resin (BD Biosciences) following the instructions of the manufacturer. Fractions were analyzed by SDS-PAGE and Coomassie blue staining. For further purification and to separate monomers from dimers and larger multimers, scFv-containing fractions obtained from immobilized metal affinity chromatography were pooled, dialyzed against PBS, applied to a HiLoad 26/60 Superdex 75 pg column equilibrated with the same buffer, and run in an ÄKTA-FPLC (GE Healthcare) at a flow rate of 0.75 mL/min. Fractions were analyzed by SDS-PAGE and those containing the monomeric form were pooled and concentrated using Vivaspin concentrators (Vivascience, Hannover, Germany).

Immunoprecipitation-Western blot analysis. About 2 × 10⁸ RT112 cells for each experiment were washed with ice-cold PBS and scraped in ice-cold lysis buffer [50 mmol/L Tris-HCl (pH 7.4), 300 mmol/L NaCl, 5 mmol/L EDTA, 1% Triton X-100] supplemented with 1× protease inhibitor cocktail (Roche). The lysates were incubated for 30 minutes on ice and centrifuged at 16,000 × g for 15 minutes at 4°C. In parallel, a precipitation complex of antibodies was made by incubating 300 μL of protein G-Sepharose (GE Healthcare) with 1 μL of anti–c-myc (Sigma) for 90 minutes at 4°C. After three washes with lysis buffer, 50 μg of the purified scFvs were added and left to bind for 2 hours at 4°C followed by three washes with lysis buffer. For immunoprecipitation, the cell lysates were added to the preformed scFv precipitation complex and were incubated overnight at 4°C, pelleted by centrifugation, and washed thrice with lysis buffer. Proteins were released with 0.1 mol/L glycine-HCl (pH 2.7), precipitated with trichloroacetic acid (Sigma), and loaded onto 7.5 % SDS-PAGE for electrophoresis. They were then transferred to Hybond-C membrane (GE Healthcare) and incubated with anti-FGFR3 mAb overnight at 4°C. The filter was washed four times and bound antibodies were detected with Mouse IgG TrueBlot and SuperSignal Substrate (Pierce, Rockford, IL).

Surface plasmon resonance analyses. The binding kinetics of soluble scFvs to FGFR3 was measured at 25°C on a Biacore X (Biacore, Uppsala, Sweden). For immobilization of the FGFR3-Fc, goat anti-human IgG (Fc fragment) antibodies (Sigma) were immobilized on CM5 chips (Biacore) by using the amino coupling kit (Biacore). For each experiment, 1 μg of purified FGFR3-Fc was first captured and then the different purified scFvs were added at various concentrations ranging from 500 to 10 nmol/L. After each experiment, the surface was regenerated with 0.1 mol/L glycine-HCl (pH 2.5). Sensorgrams were obtained at each concentration and evaluated using the BIA Evaluation 3.0 program to determine the rate constants K_on and K_off to calculate the K_d. Experiments were made in duplicate.

To determine the binding sites of the different scFvs, we carried out a competitive binding assay; 0.5 μg FGFR3-Fc was immobilized and then a 1 μmol/L solution of each scFv was injected followed by the injection of the same amount of the competitor antibody. If there was an increase in the resonance units of the sensorgram, then the binding sites of both scFvs were different. If there was no difference, then the binding sites were the same or overlapping.

The competitions for binding to FGFR3 between the ligand FGF9, FGF1, or control EGF and scFvs were also monitored by surface plasmon resonance. After capturing FGFR3 on the sensor chip, 1 μmol/L FGF9, FGF1, or EGF and 100 μg/mL heparin were injected until saturation of binding. Then, 1 μmol/L of each scFv was subsequently injected and the binding curve of the antibody to the FGFR3-FGF ligand complex was compared with the binding to FGFR3 alone obtained with each scFv at the same concentration.

Flow cytometry and confocal microscopy analyses. Phages and soluble scFvs were used for flow cytometry. Phages were prepared from 50 mL of culture, concentrated by polyethylene glycol precipitation, and resuspended in 2 mL of PBS. To determine binding of phage scFv to native FGFR3 molecules, RT112 cells were gently detached from plastic plates by using a nonenzymatic cell dissociation solution (Sigma) followed by washing with PBS. Approximately, 1 × 10⁶ cells and 10¹¹ phage particles from fresh preparations were incubated in 2% MPBS for 1 hour on ice. Bound phages were detected using anti-M13 diluted 1:100 in 1% bovine serum albumin/PBS (1 hour on ice) and R-phycoerythrin goat anti-mouse diluted 1:100 in 1% bovine serum albumin/PBS (30 minutes on ice). After each incubation step, cells were washed twice with 1% bovine serum albumin/PBS. Finally, cells were resuspended in 500 μL of PBS and immediately analyzed in a FACScalibur with CellQuestPro software (BD Biosciences).

For soluble scFvs, RT112 or transfected HEK293T cells (2 × 10⁵) were incubated with the corresponding scFv at a concentration of 50 μg/mL in FACS buffer (PBS containing 3% fetal bovine serum and 0.1% sodium azide) for 30 minutes at room temperature. After washing with FACS buffer, cells were fixed with 4% paraformaldehyde (Merck, Darmstadt, Germany) in PBS for 10 minutes. Bound scFvs were detected with anti-6xHis tag antibody diluted 1:50 in FACS buffer for 30 minutes at room temperature and then FITC-labeled anti-mouse antibody diluted 1:50 in FACS buffer. After each incubation step, cells were washed with FACS buffer. Cells were finally resuspended in 500 μL PBS and analyzed in a FACScalibur.

Labeled RT112 cells were also examined with an ultraspectral confocal microscopy system (TCSSP2-AOBS-UV, Leica-Microsystems, Mannheim, Germany) after nucleus counterstaining with 4′,6-diamidino-2-phenylindole (Molecular Probes). Images were acquired with a 63× oil immersion objective using the Leica Confocal Software.

Cell proliferation assays. RT-112 cells were plated onto 96-well tissue culture plates at a density of 2 × 10³ cells per well in RPMI 1640 supplemented with 10% fetal bovine serum and antibiotics in duplicate. Once attached, cells were washed with PBS and cultured with different concentrations of anti-FGFR3 scFvs (ranging from 0.02 to 2 μmol/L) in RPMI 1640 containing 25 ng/mL FGF9, FGF1, EGF, or FGF9 plus 0.5 μg/mL of recombinant FGFR3, together with 50 μg/mL heparin (Sigma) and antibiotics for 48 hours at 37°C. Cell proliferation was scored by staining cells with the chromogenic dye 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide. After removal of the medium, 100 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide reagent (Sigma) were added to the culture wells at a final concentration of 1 mg/mL in RPMI 1640 and the cells were further incubated for 1 hour at 37°C. Then, 100 μL of DMSO (Sigma) were added to each well. Absorbance was read at 570 nm. All the experiments were done at least thrice. Cell viability is represented as the ratio of absorbance given by scFv-treated cells over absorbance of nontreated cells, expressed as a percentage.

Selection of fibroblast growth factor receptor 3–specific single-chain Fv antibodies. Two different phage selection procedures were carried out by using different blocking conditions. Selection process A was made by using a preblocking step with IgG-coated wells. In selection B, phages were mixed with the IgG for blocking. Three rounds of panning were done on each selection phase. Results are shown in Table 1. In total, 288 colonies were randomly picked and tested by ELISA after the different selection procedures, yielding 15 FGFR3-specific antibodies. The final selection efficiency for the blocking process A was 1.5% and for process B was 3.1%, showing that the inclusion of a high concentration of blocking molecule in the phage suspension is more efficient than a previous absorption of the phage on IgG-coated wells.

Characterization of the selected single-chain Fv antibodies. After DNA sequencing, we found that 6 of 15 scFv sequences specific for FGFR3 were unique and different (Fig. 2). Two of the six clones (3A and 1D) contained a stop codon within the sequence; they were identical except for a mutation in the complementary-determining region (CDR) 2 of the V_L region (DxA). The other four scFvs were different from each other. All the mutations were concentrated on the hypervariable regions CDR2 and CDR3. CDR3 was highly variable in both V_H and V_L, but CDR2 was mainly variable in V_H.

All the scFvs were expressed at high levels as soluble molecules. For scFv purification, the periplasmic fraction was purified by standard immobilized metal affinity chromatography procedures (Fig. 3A) followed by gel filtration on a Superdex 75 column (Fig. 3B). This step allowed for a good resolution and separation of monomers, dimers, and larger-size molecules. The purification of monomers is strictly necessary for determination of the kinetic variables of binding to the antigen.

To assess the specificity of the selected antibodies, the periplasmic scFvs were tested in ELISA with FGFR3 and the highly homologous (>62%) FGFR1. The results of the ELISA analysis are shown in Fig. 4. Although different from clone to clone, there was a significant reactivity of all the scFvs against FGFR3. In contrast, the reactivity against FGFR1 was marginal in all the cases and about the same magnitude as the background response against human IgG. Given the lower reactivity of the 3A and 1D clones, besides the presence of stop codons in the scFv gene, they were discarded for further analysis.

Fibroblast growth factor receptor-3 binding affinity of purified single-chain Fv antibodies. Monomeric forms of scFv were used for determination of the affinity constants by surface plasmon resonance on immobilized FGFR3 using a Biacore instrument. The results for K_on, K_off, and K_d determinations are shown in Table 2. The affinities oscillated between 40.9 and 12.4 nmol/L, indicating a moderate to high affinity, which is rather good if we consider the medium size and complexity of the original libraries (1.6 × 10⁸). However, the K_off of two of the scFvs, 3C and 7D, are relatively poor: 16.28 seconds and 1.36 minutes, respectively.

Single-chain Fv antibody and fibroblast growth factor competition analysis. To distinguish different binding sites, we did a competitive binding analysis by using FGFR3 immobilized on a sensorchip and consecutive binding of the two antibodies that we wanted to study by surface plasmon resonance. The first antibody was passed in a saturating amount, in such a way that signal could increase only if the binding sites of the antibodies were different. The results are shown in Fig. 5A. In summary, 2D and 3B should recognize similar or overlapping epitopes as they inhibit binding with each other. 7D and 3C must recognize different epitopes. However, the 7D epitope accessibility can be partially blocked by 2D and by previous binding of 3B, but not in the opposite way (data not shown).

The ability of the natural ligands FGF9 and FGF1 to block the binding of scFvs was also tested by Biacore. As a negative control, we used EGF. The results for 3C and 7D are shown in Fig. 5B. Previous binding of FGF9 to FGFR3 significantly blocks the union of the two scFvs, indicating that FGF9 and the scFvs compete for the same binding site. In contrast, FGF1 was very effective for blocking the binding of 7D, but not the binding of 3C. No blocking effect was noticed for an irrelevant growth factor such as EGF.

Reactivity of single-chain Fv antibodies to fibroblast growth factor receptor 3–expressing RT112 cells. The binding of scFvs to native FGFR3 expressed by RT112 bladder carcinoma cells was assessed by flow cytometry using freshly prepared phages or soluble scFv. Either six scFv phages or four soluble scFvs were used for the cytometry experiments. They stained RT112 cells albeit at various levels. Flow cytometric analyses with phages displaying the six different scFvs revealed that all of them reacted at a similar significant level with viable bladder carcinoma RT112 cells (Fig. 6A). When purified scFv fragments corresponding to 3B, 3C, 2D, and 7D were used, two of them, 3C and 7D, gave an intense immunofluorescence staining with a significant fraction of viable RT112 cells. Some cells remained unstained, indicating some heterogeneity at the cellular level. 2D and 3B stained a lower percentage of cells, suggesting a different accessibility to the epitopes by the scFvs. The analysis of the 3C- and 7D-stained cells by confocal microscopy revealed the characteristic staining pattern of membrane proteins (Fig. 6A).

None of the selected scFvs was functional by Western blot assays (data not shown). Therefore, to confirm the specific binding to FGFR3, an immunoprecipitation in nondenaturing conditions followed by Western blot assay was carried out with two purified scFvs. Results are shown in Fig. 6B. Immunoprecipitation of RT112 cell extracts with scFvs 3C and 7D yielded a prominent band with the correct size that was recognized by a FGFR3-specific mAb, confirming the ability of these two scFvs to recognize soluble and native FGFR3.

Reactivity of single-chain Fv antibodies to fibroblast growth factor receptor 3 containing activating mutations. Because FGFR3 is frequently mutated in bladder cancer, we decided to test the binding ability of these two scFvs to a mutant form of FGFR3 containing a Cys in position 249 instead of a Ser. This activating mutation has been reported as the most frequent in bladder cancer (27). We prepared this mutant by site-directed mutagenesis using PCR with a modified oligonucleotide. The mutation was confirmed by DNA sequencing (Fig. 7A). The correct expression of the mutant form after transfection of HEK293T cells was assessed by Western blot analysis (Fig. 7B). The mutant displayed similar levels of expression when compared with the wild-type form. Tubulin expression was used for normalization of the loaded protein. Flow cytometric analysis showed that both scFvs, 3C and 7D, were able to recognize HEK293T cells transfected with the FGFR3 S249C mutant even more strongly than cells transfected with the wild-type form (Fig. 7C).

Blocking of RT112 proliferation. We examined the efficacy of the scFvs in inhibiting tumor cell proliferation in vitro. The RT112 tumor cell line was used in this study because it expresses significant levels of FGFR3 (20). They were cultured in the presence of FGF9 and heparin and in the absence of FCS. Three different concentrations of scFv were added to the cells for 48 hours. In two cases, 3C and 7D, there was a significant inhibition of the proliferation that was dependent on the scFv concentration added to the cultures (Fig. 8A). The highest inhibition was achieved at 2 μmol/L, but it was also significant at 0.2 μmol/L. Therefore, anti-FGFR3 scFvs 3C and 7D significantly inhibited FGFR3-mediated proliferation of RT112 cells. 2D inhibition was at a much lower degree and 3B did not cause any inhibitory effect.

A further confirmation of the specificity of the blocking was made by the addition of different growth factors. The scFv inhibition in the presence of different growth factors is shown in Fig. 8B. The inhibitory effect was factor specific. Both 3C and 7D inhibited the proliferation induced by FGF9, but only 7D was able to block FGF1. None of them was able to block the proliferation induced by EGF. Moreover, the addition of soluble FGFR3 to the cultures abolished the scFv blocking capability. These results correlate well with the competition analysis, flow cytometry, and the accessibility of the epitopes on the cell membrane. Finally, from the morphologic point of view, treatment with 3C and 7D scFvs resulted in changes in cellular morphology and slower growth (Fig. 8C), with a more prominent vacuolization in the case of 7D.

FGFR3 is a receptor tyrosine kinase implicated in cell growth and tumorigenesis through activation of different signal transduction pathways such as extracellular signal-regulated kinases 1 and 2, phosphatidylinositol 3 kinase, and phospholipase c-γ pathways (9, 28, 29). Overexpression of receptor tyrosine kinase in tumors is generally associated with poor prognosis (30). Apart from bladder cancer, FGFR3 is mutated and overexpressed in a variety of cancers and, as a result, particular attention has been given to FGFR3 for therapeutic use. Blocking of the transduction pathways used by FGFR3 could be an appropriate way to inhibit cell growth and differentiation induced by this molecule. Thus, FGFR3 constitutes an excellent target for effective cancer intervention. In a previous report (20), it has been shown that murine mAbs specific for FGFR3 provoked an antiproliferative effect in the RT112 cell line. However, the use of mouse mAbs for human cancer therapy is not realistic due to the strong human anti-mouse antibody response, which has made almost impossible any therapeutic approach based on mouse mAbs. Currently available alternatives include “chimerization” and “humanization” of mouse antibodies. Both approaches are labor- and time-intensive and are unpredictable procedures.

We have directly prepared human antibodies from a human phage display library using a soluble chimeric molecule composed of the extracellular domain FGFR3α(IIIc) coupled to the Fc region of human IgG1 as antigen for antibody selection. Therefore, the specific reactivity obtained against FGFR3 must be targeted to the extracellular domains. After ELISA selection, we obtained six unique FGFR3-specific scFv sequences. The specificity of the antibodies was confirmed by immunoprecipitation, Biacore analysis, and flow cytometry. None of the selected scFvs was able to recognize FGFR3 by immunoblot analysis (data not shown). These results suggest that the recognized epitopes are conformational, nonlinear, and susceptible to denaturation, probably due to the use of soluble antigen for the panning procedures. These findings make difficult the identification of the amino acid sequence corresponding to these epitopes and its position into the FGFR3 molecule. Competition experiments have shown that the antibodies recognize different but, in some cases, overlapping epitopes. 2D and 3B probably recognize the same epitope.

Biacore analysis was used for the determination of the kinetic constants of the scFv binding to the antigen. One of the problems associated with phage display is related to the generally low affinity shown by these scFv. As previously described, for selection of antigens immobilized on polystyrene plates, the resulting scFvs are a mixture of monomers and dimers (31). The resulting homodimers show a significant increase in apparent affinity (avidity; ref. 32). Due to this increase in avidity, phages displaying mixtures of monomeric and dimeric forms will be preferentially selected. To determine the kinetic constants of the scFvs, it was necessary to separate the monomers by gel filtration. In our case, the affinities of the selected scFvs oscillated between 40.9 and 12.4 nmol/L, indicating moderately high affinities, which is rather good if we consider the low size and complexity of the original libraries (1.6 × 10⁸) and compare very favorably with the affinities obtained by using a much larger library (2 × 10¹⁰) of Fab fragments, which were approximately in the same range or 1 order of magnitude better (25). In terms of application to cancer therapy, the K_d of these scFvs might prove appropriate taking into account that recent reports have shown that a very high affinity can restrict the tumor penetration of the scFvs (33). Currently available antibodies for therapeutic use have a K_d range of 10 to 0.1 nmol/L (34).

Flow cytometric analysis confirmed the specific binding of the 3C and 7D scFvs to the native membrane FGFR3. More interestingly, 3C and 7D were able to recognize even better a mutant FGFR3 containing one of the most abundant activating mutations present in bladder cancer, S249C, stressing the ability of these two scFvs to bind tumoral tissues. This enhanced recognition could be explained by the natural tendency to dimerize the mutant forms, which might stabilize the antibody binding.

Previous results have shown that secreted soluble, extracellular ligand-binding domains of the native FGFRs play an inhibitory effect on cell proliferation, mopping up the ligand binding to the receptor (35). The antibody fragments may mimic this capability of the secreted extracellular domains. In fact, several results presented in this study suggest that the mechanism of RT112 cell growth inhibition by 3C and 7D scFvs is probably due to a direct block of FGF-FGFR3 interaction, resulting in the inhibition of FGFR3 activation. This is based on the observation that 3C and 7D specifically bind to both soluble and cell-surface expressed FGFR3 on RT112 cells. Moreover, 3C and 7D were blocked by FGF9 from binding to FGFR3, but only 7D competed with FGF1. This result indicates the capacity of these antibodies to use the same binding site as FGF9, blocking its natural effector function. Furthermore, 3C and 7D inhibit FGF9-induced proliferation in a significant and concentration-dependent manner. Both 7D and 3C have a very rapid off rate (1.22 × 10⁻² and 6.14 × 10⁻², respectively), which might lead to a very short association with the tumoral cells due to the rapid turnover and internalization. A reduction in K_off is typically the major kinetic mechanism resulting in higher affinity. Thus, it would be convenient to improve this affinity to slow the off rate. Because of the relatively poor K_off of 3C and 7D when compared with other scFvs, it is very likely that an improvement in the affinity of these molecules might suppose an improvement in their blocking ability. The 3C and 7D scFv sequences will serve as a template for affinity maturation studies according to well-established technologies such as complementary-determining region mutagenesis (36) or somatic hypermutation mimicry (37).

There are a number of important human or humanized antibodies in clinical trials against different molecules implicated in cancer progression (38). Antibodies are showing a great promise in cancer treatment. Compared with small molecules, antibody therapy may provide higher specificity to tumoral cells and better safety profile. The moderate affinity of the scFv fragments is sufficient to react only with a population of cells expressing significant levels of FGFR3, but not towards cell populations that express only low levels of the receptor. The scFvs could be used as therapeutic agents either alone or combined with other molecules such as immunotoxins or radionuclides that can selectively kill the tumoral cells displaying FGFR3, as previously shown for angiogenic molecules and growth factor receptors, such as vascular endothelial growth factor or erbB2 (neu/HER2; ref. 24), or other overexpressed proteins (39). The small size of the scFvs considerably increases their tumor penetration capabilities (33, 40) and pharmacokinetic properties (41). Moreover, the use of antibody fragments confers a high specificity to the inhibitory effect, which allows the discrimination between the different FGFRs. This is particularly relevant given the opposite effects they play. Thus, whereas FGFR3 presents an oncogenic activity, FGFR2 shows a tumor suppressor activity (42).

Apart from its overexpression in bladder or cervical cancer, dysregulated FGFR3 has been show to play a major role in multiple myeloma tumorigenesis and has already been tested as a molecular target for therapy (19). Specific inhibitors of FGFR3 induced cell cycle arrest and apoptosis in human myeloma cells. It will be interesting to test the antiproliferative effect of these antibodies on multiple myeloma cell lines. Our data, together with previous results of gene expression analysis and polyclonal anti-FGFR3 antibodies, support further evaluation of 3C and 7D scFv antibodies as antitumor agents. Previous findings (20) support that FGFR3 inhibition should be introduced at early disease stages (when FGFR3 overexpression is more abundant) before progressive loss of expression.

We thank Arancha Garcia, Diego Megías, and Maria Montoya of the Centro Nacional de Investigaciones Oncológicas confocal and cytometry unit for their help and collaboration in the flow cytometric and confocal microscopy analysis. We thank Silvia Romero for her excellent technical assistance and Giovanna Roncador for her help in the transfection experiments.