Antitumor activity of fibroblast growth factor receptor 3–specific immunotoxins in a xenograft mouse model of bladder carcinoma is mediated by apoptosis

Bladder carcinoma is the fourth leading cause of cancer in men and the ninth most frequent cancer among women in the United States and Europe (1, 2). Among patients with primary bladder tumors, 75% to 85% have been found to suffer from superficial pT_a or pT₁ transitional cell carcinoma. Systemic chemotherapy preceded or followed by cystectomy and urinary diversion is the standard treatment of patients. However, the overall response rates vary between 39% and 65%, with complete response in 15% to 25%. The median survival for this patient population is ∼16 months (3). Systemic conventional treatment presents several limitations because standard treatment is frequently accompanied by significant toxicity. This is particularly problematic because the patient population is frequently elderly with poor renal function. Clearly, rationally designed, molecularly targeted approaches may provide an opportunity for the development of new, potent therapeutic agents with an improved toxicity profile.

Monoclonal antibody (mAb) therapy has revolutionized cancer treatments and several mAbs are now approved for clinical use against some types of cancer (see ref. 4 for a review). The most developed class of targeted cytotoxic treatments is based on tumor ligands or antibodies that are linked to an antiproliferative component composed of bacterial or plant protein toxins (see ref. 5 for a review). Among them, the gelonin toxin is a 29-kDa ribosome-inactivating single-chain N-glycosidase, with a potency and mechanism of action similar to ricin-A chain (6). Ongoing clinical trials with gelonin-based agents suggest that the recombinant gelonin toxin (rGel) does not appear to generate capillary leak syndrome or hepatotoxicity (7) as observed with other toxins, such as ricin A chain or Pseudomonas exotoxin-based agents. Immunotoxins containing rGel have been shown to be very effective in specifically killing tumor cells in vitro and in vivo when fused to a ligand or antibody cell-targeting fragment (8–11).

Overexpression of fibroblast growth factor (FGF) receptor 3 (FGFR3) may play an important role in the molecular pathogenesis of bladder (12) and other types of cancers, such as multiple myeloma (13) and hepatocarcinoma (14). Ligand-independent FGFR3-activating mutations, equivalent to the mutations responsible for several skeletal anomalies, are present in ∼40% to 60% of bladder tumors (15, 16). Interestingly, the majority of these tumors are of low stage and grade, and it has been reported that there exists a correlation between the presence of FGFR3 mutation and a favorable clinical outcome and low recurrence (17, 18). In contrast, the mutation rate was found to decrease gradually among non-muscle-invasive tumors as grade increases (19) and this suggests that superficial low-grade/papillary and invasive high-grade tumors progress via different molecular pathways (20). Also, overexpression of FGFR3 mRNA and protein levels has been described in pT_a, pT₁, and pT₂ stage carcinomas when compared with normal adjacent bladder specimens and this appears to occur more frequently in transitional carcinomas than activating mutations (21).

On binding with FGF and heparin, FGFR3 undergoes dimerization and transphosphorylation at the intracellular kinase domain, resulting in a downstream signaling for proliferation and differentiation. Different isoforms of FGFR3, such as FGFR3 IIIb and IIIc, show different specificity for FGF ligands. Expression of FGFR3 IIIb appears to be associated with epithelial cells and binds FGF1 preferentially and FGF8 and FGF9 to a lower extent (22, 23). In contrast, expression of the IIIc variant was generally associated with mesenchymal-derived cell lineages and binds a wide range of FGFs. Other isoforms have been described, some of which result in frame shifts and early termination in exon 10 region, where others showed deleted domains but still in-frame (24, 25). Characterization of these variants has led to the hypothesis that aberrant splicing of FGFR3 mRNA may result in a selective advantage for cancer cells. Anomalous receptor kinase activation (26) or defective receptor degradation (27) both due to an inappropriate expression or mutation results in many human disorders as achondroplasia, thanotophoric dysplasia, multiple myeloma, cervical carcinoma, hepatocellular carcinoma, bladder carcinoma, and other epithelial carcinomas. FGFR3 seems to mediate opposite signals in those syndromes, acting as a negative regulator of growth in bone (28) and as an oncogene in the different tumor types.

Therefore, FGFR3 appears to be a biologically interesting and potentially important target for therapeutic application in bladder cancer, as it has been reported for multiple myeloma (29–32). As a result of this increasing interest, we have previously isolated and characterized two FGFR3-specific human single-chain Fv (scFv) antibody fragments, designated as 3C and 7D, which effectively inhibited the in vitro proliferation of RT112 tumoral cells in a dose- and FGF-dependent manner (33). Moreover, competition analyses by surface plasmon resonance showed that 3C competed with FGF9 but not with FGF1 for the binding site. In contrast, FGF1 and FGF9 blocked the binding of 7D to FGFR3 (33).

In the current study, we have generated fusion constructs using the anti-scFv antibody 3C containing rGel toxin (3C-rGel). Then, we evaluated and characterized the inhibitory activity of the scFv 3C on the growth of RT112-derived human bladder carcinoma in an orthotopic model in severe combined immunodeficient mice. Our results show that the immunotoxin 3C-rGel is an effective tumor inhibitor, being able to significantly delay tumor growth, suggesting that this construct might be an excellent candidate for clinical development in the treatment of bladder transitional carcinoma.

Human RT112 cells (German Collection of Microorganisms and Cell Cultures) were cultured in RPMI 1640 (Sigma) supplemented with 10% fetal bovine serum (Invitrogen) and 100 units/mL penicillin/streptomycin. Human RT4 cells [kindly provided by Dr. F. Real (Epithelial Carcinogenesis Group, Centro Nacional de Investigaciones Oncológicas)] were cultured in DMEM (Invitrogen) with 10% fetal bovine serum and 100 units/mL penicillin/streptomycin.

FGFR3α IIIc/Fc and FGF9 were provided by ReliaTech. Goat anti-human IgG (Fc specific) was provided by Sigma. Rabbit anti-FGFR3 polyclonal (C-15) and mouse anti-FGFR3 monoclonal (B-9) antibodies were obtained from Santa Cruz Biotechnology. Rat anti-CD34 mAb was from BD Pharmingen. Rabbit anti-Ki-67 mAb (clone SP6) was purchased from Master Diagnostica. Rabbit anti-caspase-3-active polyclonal antibody was from R&D Systems. Mouse anti-α-tubulin mAb was from the Monoclonal Antibodies Unit, Centro Nacional de Investigaciones Oncológicas. Rabbit anti-gelonin antisera were obtained from the Veterinary Medicine Core Facility at M. D. Anderson Cancer Center. Horseradish peroxidase (HRP)–conjugated goat anti-rabbit IgG, biotin-conjugated rabbit anti-rat, and HRP-streptavidin were obtained from DAKO. HRP-conjugated goat anti-mouse IgG was from Sigma. Cy3-coupled anti-rabbit IgG was from Jackson ImmunoResearch. Alexa Fluor 488 donkey anti-rabbit IgG was provided by Molecular Probes.

The scFvαFGFR3 3C was expressed in Escherichia coli HB2151 cells and purified by IMAC and gel filtration as described previously (33). The cDNA encoding 3C and rGel were fused together by using the splice overlap extension PCR method as described previously (9). To construct scFvαFGFR3#3C/rGel (3C/rGel), an upstream fragment encoding the restriction enzymes BspHI was introduced and amplified from the NH₂-terminal portion of the scFvαFGFR3 gene using FGF-For (5′-GGCGGTGGCTCCGTCATGACGGTGCAGCTGTTGGAGTCT-3′) and FGF SSG-Link Bac (5′-CGCACTACCCGAGCCGCTGGACCGTTTGATTTCCACCTT-3′). An adjoining downstream fragment encoding a 7–amino acid (SSGSGSA) linker and a restriction enzyme HindIII was amplified from the rGel gene using the primers FGF SSG-Link For (5′-GCTCGGGTAGTGCGGGCCTGGACACCGTGAGC-3′) and GEL-H3 Bac (5′-ACCGCATTAAGCTTATCATTTAGGATCTTTATCGAC-3′). The upstream and downstream PCR fragments were then reassembled as a full-length scFvαFGFR3/rGel fusion gene encoding the fusion protein by an additional PCR step using the pair of primers FGF-For and GEL-H3 Bac flanking the 5′- and 3′-ends. The final PCR fragment was purified and cleaved with BspHI and HindIII restriction endonucleases and then cloned into the Novagen expression vector pET-32a cloning sites NcoI and HindIII. For the scFvαFGFR3#3CF/rGel (3CF/rGel), an 18–amino acid flexible linker (Linker 218: GSTSGSGKPGSGEGSTKG) was first inserted between the rearranged heavy and light chains of the 3C scFv DNA using the oligonucleotide primers 3CF Link-For (5′- GGTAAGCCCGGGAGTGGTGAGGGTAGTACTAAGGGTGAGGTGCAGCTGTTGGAGTCT-3′) and 3CF Link-Bac (5′- ACCCTCACCACTCCCGGGCTTACCACTGCCGCTGGTGGAGCCCCGTTTGATTTCCACCTT-3′). Then, a downstream fragment encoding a G₄S linker was amplified between the scFv and the rGel gene using oligonucleotide primers 3CF-For (5′-GGCGGTGGCTCCGTCATGACGGACATCCAGATGACCCAG-3′) and Xombac2 (5′-GCTCGTGTCGACCTCGAGTCATTATTTAGGATCTTTATC-3′). The upstream and downstream PCR fragments were then reassembled and cloned into the pET-32a sites NcoI and XhoI.

The final scFvαFGFR3/rGel fusion gene encoding each construct (3C/rGel and 3CF/rGel; Fig. 1A) were then cloned into DH5α bacterial cells (Novagen) for plasmid propagation. The DNA sequences were verified by DNA sequencing before the plasmids were transformed into E. coli strain Origami (DE3) pLysS (Novagen) for expression of the fusion protein. The expressed protein was first purified by IMAC using chelating resin charged with cobalt chloride. After cleavage with recombinant enterokinase (Novagen) to remove the Trx and His tags, the final fusion protein was purified by affinity chromatography on a Blue Sepharose Fast Flow column as described previously (8, 9).

A BIAcore X apparatus (BIAcore) was used for analyzing specificity and affinity of immunotoxins. Assays to measure immunotoxin binding to FGFR3-Fc used a CM5 carboxymethyl-dextran sensor chip with anti-human IgG (Fc fragment) antibodies immobilized via amine coupling. Purified FGFR3-Fc (1 μg) was captured for each cycle. 3C, immunotoxins, and rGel at varying concentrations from 500 to 10 nmol/L were then sequentially injected and the corresponding response (in resonance units) was recorded at the end of injection. Binding interaction was done in HBS-EP buffer [10 mmol/L HEPES (pH 7.4) containing 150 mmol/L NaCl, 3 mmol/L EDTA, 0.005% surfactant P20; BIAcore] at a flow rate of 20 μL/min at 25°C. The surface was regenerated back to the baseline level of immobilized anti-human IgG (Fc fragment) antibodies with 0.1 mol/L glycine-HCl (pH 2.5). Global analyses were done using BiaEvaluation 3.0 program.

Cell growth was assessed using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay as reported previously (33). Briefly, 2 × 10³ RT112 cells were seeded into flat-bottomed 96-well plates in triplicate and allowed to adhere overnight. They were then treated with the intended doses of scFv or immunotoxins ranging from 0.02 to 2 μmol/L. After 3 days of incubation at 37°C in 5% CO₂, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma) was added to a final concentration of 1 mg/mL and the cells were incubated for a further 1 h under the same conditions. Then, DMSO (Sigma) was added for reaction and the absorbance at 570 nm was measured with a microplate reader (Bio-Rad Laboratories). Percent control refers to the percentage of viable cells in treated wells compared with that of control (untreated) cells.

Visualization of immunotoxin internalization was done by confocal microscopy. RT112 cells (2 × 10⁴) were allowed to adhere on eight-well chamber slides (Nunc) at 37°C overnight in 5% CO₂. The cells were treated with 50 nmol/L per well of immunotoxins or rGel prepared in complete DMEM for 1 h. Proteins bound to the cell surface were stripped by 10-min incubation with 0.1 mol/L glycine (pH 2.5) in 0.5 mol/L NaCl followed by neutralization with 0.5 mol/L Tris-HCl (pH 7.4) for 5 min. After a brief wash with PBS, cells were fixed with 4% (w/v) paraformaldehyde in PBS for 10 min at room temperature, washed extensively with PBS, and permeabilized in 0.5% Triton X-100 for 10 min. Then, slides were incubated with 3% bovine serum albumin in PBS for 1 h at 37°C and subjected to two 1-h incubations at room temperature with rabbit anti-rGel polyclonal antibodies diluted 1:500 in 0.2% bovine serum albumin/PBS/0.1% Tween 20 and with a 1:100 dilution of Cy3-coupled anti-rabbit IgG in TBS. Nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI; Molecular Probes) and slides were mounted with Mowiol 4-88 (Calbiochem). Finally, fluorescence was examined with an ultraspectral confocal microscopy system (Leica TCSSP2-AOBS-UV, Leica Microsystems) and images were acquired with the accompanying software package.

Cells were lysed in lysis buffer [10 mmol/L Tris-HCl (pH 7.2), 150 mmol/L NaCl₂, 5 mmol/L EDTA, 0.1% SDS, 1,0% Triton-X 100, 1% sodium deoxycholate, and complete protease inhibitor mix (Roche Molecular Biochemicals)] and clarified by centrifugation at 12,000 × g for 15 min. Protein concentrations were determined using the 2-D Quant kit (GE Healthcare). Protein (50 μg/lane) was loaded on NuPage 4% to 12% Bis-Tris gels (Invitrogen) and electrophoretically transferred to nitrocellulose Hybond-ECL membranes (GE Healthcare). After blocking, blots were first probed with mouse anti-FGFR3 mAb and then with HRP-conjugated goat anti-mouse IgG. Proteins were visualized by ECL detection system (GE Healthcare). Blots were stripped with Re-Blot Plus mild stripping solution (Chemicon) for 15 min before reprobing with mouse anti-α-tubulin mAb as a loading control.

A FlexiTube GeneSolution pool for FGFR3 (Entrez gene ID: 2261) containing four different small interfering RNA (siRNA) duplexes (Qiagen) was transfected in RT112 cells seeded in six-well dishes (1 × 10⁵ cells in each well) using Oligofectamine (Invitrogen) according to the manufacturer's protocol. A nontargeting siRNA pool (Negative Control siRNA; Qiagen) was similarly transfected as a negative control. After 24 h of siRNA delivery, FGFR3-specific siRNA [si(FGFR3)]–transfected and nonspecific siRNA [si(nonspecific)]–transfected cultures were assayed for FGFR3 content at mRNA and protein levels. Relative amounts of FGFR3 mRNA were determined by the Taqman real-time PCR method using the probe-primer mix for human FGFR3 (Hs00179829_m1; Applied Biosystems) on ABI 7900HT thermocycler (Applied Biosystems) using GUSβ for an endogenous control. The real-time PCR was conducted in triplicate and quantifications were done using absolute quantification (standard curve) module of 7900HT system software version 2.2.2. After normalizing to GUSβ, the expression value of FGFR3 transcripts from [si(FGFR3)]-transfected cells was calculated relative to that from [si(nonspecific)]-transfected cells, which was taken as 100%. FGFR3 protein levels were analyzed at 24 h post-transfection by Western blot using mouse anti-FGFR3 mAb on lysates of siRNA-transfected cells as described above. For cytotoxicity studies on FGFR3-silenced RT112 cells, cells were rinsed after 24 h of siRNA transfection and fresh medium, containing various concentration of 3C/rGel or free rGel, was added for 72 h. Proliferation was measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay as described.

RT112 cells were plated overnight into 35-mm tissue culture dishes and treated with 0.2 μmol/L 3C, immunotoxins, or rGel in complete RPMI 1640 at various times (24, 48, and 72 h). After that, both adherent and floating cells were collected and apoptosis was measured with FITC-conjugated Annexin V antibody and propidium iodide (PI) using Annexin V-FITC Apoptosis Detection kit I (BD PharMingen) according to the instructions from the manufacturer. Positive cells were detected by fluorescence-activated cell sorting on a FACSCalibur flow cytometer (BD Bioscience) using CellQuest Pro software (BD Bioscience). RT4 cells were treated with 0.2 μmol/L 3C-rGel or rGel alone in complete DMEM at 24, 48, and 72 h. Apoptosis was analyzed following the same described protocol.

Severe combined immunodeficient female mice (6-8 weeks old) with an average weight of 20 g were treated following international ethical guidelines. Log-phase RT112 transitional carcinoma cells were injected s.c. (1 × 10⁶ cells per mouse in 0.2 mL PBS) into the right flank and tumor size was measured periodically. Treatments were started when tumors reached a noticeable size (≈30 mm³) by days 5 to 7 after tumor implantation and mice were randomized to control and treatment groups. 3C/rGel and 3CF/rGel fusion toxins were administered at a total dose of 50 mg/kg and 3C and rGel at a total dose of 25 mg/kg i.v. via tail vein, in alternating days for a total of 2 weeks. Control animals received only saline vehicle following an identical schedule. Tumor sizes were monitored at least twice a week and each animal was euthanized using a CO₂ chamber when its neoplasm reached a predetermined endpoint size (1,500 mm³) in accordance with the Guidelines for Humane Endpoints for Animal Used in Biomedical Research.

After 3-week treatment, mice were sacrificed and tumors were excised and fixed in 10% buffered formalin (Sigma), dehydrated through ethanols and xylene, and embedded in paraffin for staining with H&E or immunohistochemical characterization. Proliferative activity of tumor cells was assessed using a rabbit anti-Ki-67 mAb and apoptotic cells were detected with a rabbit anti-human/mouse caspase-3-active and a HRP-conjugated anti-rabbit IgG. Ariol Imaging System (Applied Imaging Corp.) was used to evaluate the percentage of Ki-67 and caspase-3-active-positive tumor cells in relation to all tumor cells in the same section. Positive staining is calculated by applying two thresholds with one recognizing blue background (hematoxylin-stained) cells and other recognizing brown positive cells. Neovascularization was measured by CD34 staining of endothelial cells. For detection of FGFR3 protein, a rabbit polyclonal anti-human antibody was used. A biotinylated anti-rat secondary antibody and HRP-conjugated streptavidin were also used in this detection assay. In all sections, positive cells were visualized using 3,3-diaminobenzidine tetrahydrochloride plus (DAKO) as a chromogen and were counterstained with hematoxylin.

Immunofluorescence staining was also done for FGFR3 detection. Formalin-fixed and paraffin-embedded specimens were cut into 3-mm-thick sections, deparaffinized, and preincubated with fetal bovine serum to prevent nonspecific binding. The sections were incubated at room temperature for 30 min with primary rabbit polyclonal antibody to human FGFR3 (10 μg/mL) followed by incubation with Alexa Fluor 488–conjugated donkey anti-rabbit IgG (20 μg/mL) at 37°C for 20 min after being washed three times with PBS and 0.05% Tween 20. Nuclei were counterstained with DAPI (Molecular Probes) at 1:1,000 dilution and slides were mounted with Mowiol 4-88 (Calbiochem). Visualizations were done as above.

All statistical analyses were done with Microsoft Excel software. Data are presented as mean ± SD. P values were obtained using two-tailed paired t tests with 95% confidence interval for evaluation of the statistical significance compared with the controls. P values <0.05 were considered statistically significant.

The immunotoxin constructs were purified to homogeneity as assessed by SDS-PAGE (Fig. 1B). To confirm that the fusion constructs retained the FGFR3-binding ability, surface plasmon resonance experiments on immobilized FGFR3 were done. The resulting surface plasmon resonance binding curves are shown in Fig. 1C and the affinity constants (K_d) of the purified fusion proteins were compared with that of the corresponding parental scFv. The 3C/rGel was able to bind FGFR3 ligand with a similar affinity to 3C (18.3 versus 25.0 nmol/L). However, the other immunotoxin tested, 3CF/rGel, exhibited ∼2.5- and 3.6-fold lower affinity than those of the parental (65.2 versus 25.0 nmol/L) and the 3C/rGel (65.2 versus 18.3 nmol/L), respectively, suggesting that FGFR3 binding by 3CF/rGel was affected by the swapping of the V_L and V_H domains. As expected, rGel alone failed to interact with FGFR3.

The specific cytotoxic activities of immunotoxins were assessed on target RT112 cells in log-phase growth. Both immunotoxins tested were found to be specifically cytotoxic in a dose-dependent manner on the target cells (Fig. 2A). The IC₅₀ values for 3C/rGel and 3CF/rGel were ∼200 nmol/L compared with ∼1,500 nmol/L for rGel.

To show that 3C-rGel has negligible activity toward cells that do not express FGFR3, RT112 cells where FGFR3 expression was knocked down were prepared by using a pool of FGFR3-specific siRNA. Effective blocking of FGFR3 expression at the mRNA and protein level was confirmed by quantitative PCR and Western blot analysis, respectively (Fig. 2B). Cytotoxicity of 3C-rGel on FGFR3-silenced RT112 cells was significantly lower (IC₅₀, 600 nmol/L) than in control cells (Fig. 2C).

To study the internalization of the rGel immunotoxins, immunofluorescence staining with a rGel-specific antibody was followed by confocal microscopy on RT112 cells (Fig. 2D). Cells treated with 3C/rGel exhibited a massive punctuate staining pattern indicative of considerable internalization. 3CF/rGel-treated cells showed also cytosolic staining but in a much less extent, suggesting a lower efficient delivery of rGel to the cytoplasm. Cells treated with rGel displayed a lower intracellular content and thereby show that only a small amount of the free toxin becomes nonspecifically internalized. Therefore, 3C/rGel was found to be very efficient in binding FGFR3 on the tumor cell surface and in delivering rGel to the cell cytoplasm.

To investigate whether the mechanism of immunotoxin-induced cell death occurs through induction of apoptosis, we used Annexin V and PI staining. Cells positive for Annexin V labeling without PI staining were considered to be in early apoptosis and cells stained for Annexin V and PI in late apoptosis/necrosis. RT112 cells were exposed to 0.2 μmol/L 3C/rGel and 3CF/rGel, the IC₅₀ dose, or an equivalent molar dose of 3C and rGel for 24, 48, and 72 h (Fig. 3A). Cells treated with 3C/rGel were found to bind more FITC-conjugated Annexin V than cells treated with 3C, 3CF/rGel, rGel alone, or saline. Quantification of cell death after 3 days showed that 50% of 3C/rGel-treated cells and 40% of 3C-treated cells stained positive for Annexin V in comparison with ∼17% of control cells, untreated, and treated with rGel. 3CF/rGel exhibited reduced apoptotic activity because slight signs of apoptosis only appeared after 72-h incubation.

To confirm the apoptotic activity of 3C-rGel in other bladder tumoral cell line beyond RT112, we used RT4 cells, which also overexpress FGFR3 although at a lower level than RT112 (Fig. 3B). The apoptosis was equivalent in RT4 cells to that obtained in RT112 cells (Fig. 3C). These results in two different bladder cell lines confirm that the process of cell death induced by the immunotoxin is apoptotic.

RT112-derived tumor growth was significantly decreased when treated with 3C/rGel (Fig. 4A). After 20 days of treatment, the mean tumor size of the 3C/rGel-treated group was reduced by 70% and 55% relative to the control group treated with saline or rGel, respectively. Treatment with 3C scFv alone resulted in a moderate reduction of tumor growth (39%) compared with saline-treated mice. In contrast, the 3CF/rGel showed no significant effects on tumor growth. rGel-treated mice exhibited a 29% reduction in tumor size.

Treatment with 3C/rGel also resulted in an improved survival rate when compared with the other groups (Fig. 4B). Forty-one days after tumor injection, all the 3C/rGel-treated animals were alive, whereas only a 25% of 3C-treated mice survived. A 30% (3 of 10) of mice achieved long-term survival (>60 days) with 3C/rGel treatment.

To study the possible mechanisms underlying tumor growth retardation, we analyzed tumors taken 25 days after cell implantation from mice treated either with 50 mg/kg 3C/rGel or saline. Histopathologic analysis revealed, in both groups, well-delimitated, noninfiltrating tumors into the s.c. adipose tissue (Fig. 5A). They showed the characteristic pattern of the urothelial (transitional cell) carcinoma, with fronds covered by a thickened cellular epithelium in which cells maintained overall polarity or definite disturbance with variable nuclear enlargement and hyperchromasia (grades 2 and 3). A papillary pattern was usually preserved and it was not uncommon to see focal glandular differentiation. Large areas of individual cell necrosis were easier to find in the 3C/rGel-treated group (Fig. 5A, 3C/rGel).

We next studied whether immunotoxin treatment alters the angiogenic, proliferative, or apoptotic status of tumors by staining for endothelial cell marker CD34, Ki-67, and caspase-3-active, respectively. No significant differences were shown in tumor vascularity as determined by CD34 staining from mice treated or not treated with 3C/rGel (Fig. 5B). No significant differences were seen in proliferation marker Ki-67 staining between the control and the treated groups, 22.41 ± 5.29% and 18.91 ± 3.87% of Ki-67-positive cells, respectively (Fig. 5C). The percentage of apoptotic cells increased ∼4-fold (P < 0.0001) in the 3C/rGel-treated group (39.6 ± 8.72%) compared with that in the control group (9.78 ± 6.23%; Fig. 5D). If the relation of proliferation to apoptosis was considered, control mice exhibited a 5-fold higher ratio than the treated animal (2.29 versus 0.47) mainly due to the induction of apoptosis. Taken together, these results suggest that the growth-inhibitory activity of 3C/rGel was mediated through an increase in apoptotic cell death.

Subcellular distribution of FGFR3 varied between treated and control groups (Fig. 6A). In the control animals, FGFR3 showed mainly a cytoplasmic pattern of expression in the tumor with a characteristic distribution around the vascular cores. However, a nuclear staining of FGFR3 occurred in tumors from immunotoxin-treated mice together with a cytoplasmic staining similar to that observed in the control sections. To obtain a confirmation of this protein shuttling, we did double immunofluorescence staining with anti-FGFR3 and DAPI. In the 3C/rGel-treated tumor (Fig. 6B), anti-FGFR3 gave a dense nuclear staining, which colocalized with DAPI staining, with exclusion from the nucleoli, indicating that the localization of most FGFR3 was predominantly nuclear. In contrast, the control sections showed prominent cytoplasmic staining of FGFR3 but no stain within the nuclei. These results corroborate the observation that FGFR3 is translocated from the cytoplasm to the nucleus in xenograft tumors treated with 3C/rGel.

The effect of FGFR3 overexpression as a predictor of therapeutic response or overall survival has not been examined. Moreover, in our xenograft model that mimics quite well the morphology of a transitional bladder tumor, we observed significant overexpression of FGFR3 in the tumor after immunostaining with a FGFR3-specific antibody. This overexpression would explain the efficient antiproliferative effects of the immunotoxin. A recent report described the inhibitory activity of another anti-FGR3 antibody on multiple myeloma cells (34), although, in contrast to bladder cancer, no overexpression of FGFR3 has been described for multiple myeloma.

The high affinity and specificity of FGF receptors have been used previously as targets for delivery of adenovirus to tumor cells (35) using basic FGF (FGF2) as the targeting ligand. Moreover, chemical conjugates of FGF2 and the plant toxin saporin showed an impressive antiproliferative activity against human bladder cancer cell lines (36) by acting on its receptor FGFR1. However, there are not previously described examples of antibodies targeting the external domain of the FGFR3 used as therapeutic approaches in bladder tumors. Here, we propose the use of lower-molecular-weight scFv conjugated to immunotoxins as therapeutic agents. The use of scFv improves the penetration of the antibody into tumors and simplifies the genetic coupling of the toxin to the molecule. scFv are also potentially less immunogenic and show more rapid clearance kinetics from the circulation.

We have used immunotoxins based on rGel, which have shown improved stability and reduced toxicity compared with other toxins (37). The rGel was fused to an FGFR3-specific scFv for studies of in vitro and in vivo inhibition of bladder carcinoma cell proliferation. It is known that the mechanism of action of rGel is based on protein synthesis-inhibitory activity, which does not need an intracellular cleavage of the toxin (9). Previous reports have shown the ability of rGel to provoke a cytotoxic effect when coupled to other molecules or scFv (9, 38–40). Here, the mechanism of action of the immunotoxin was analyzed in tissue culture and after the in vivo experiments. The immunotoxin was effectively internalized after cell binding. The internalization and cytotoxicity of the immunotoxins seem to be FGFR3 specific as shown in FGFR3-silenced cells. This internalization was followed by a significant increase in cellular apoptosis. The apoptosis was visualized in cell lines overexpressing FGFR3, such as RT112 or RT4 cells. Selection of RT4 cells was based in a previous report of expression levels of FGFR3 in bladder cancer cell lines (41).

We observed a substantial antiproliferative effect of the 3C-rGel scFv on RT112-induced xenografts. The 3C-rGel immunotoxin was much more effective than the 3C scFv alone or the V_L-V_H reversed 3CF-rGel immunotoxin. Although 3C-rGel and 3CF-rGel showed comparable cytotoxicities in vitro, they did not the same in vivo. This difference could be attributed to the different affinity effect on in vivo tumor targeting. Affinity properties regulate the penetration and quantitative delivery of antitumor scFv molecules to tumor tissue (42). Although high affinities may impair efficient tumor penetration by a “binding site barrier” effect (43), affinities below a threshold will not permit a durable binding to the tumor antigen, as it might be the case for 3CF-rGel. Thus, the lower activity of 3CF-rGel could be attributed to a weaker affinity of the scFv after the reversion of the variable domains. This result suggests that further improvements in inhibiting the in vivo tumor growth may be obtained by identifying the optimal affinity of the 3C scFv component. This improvement would require an affinity maturation process according to published protocols (44, 45).

A surprising finding from this study was the observation that tumors from immunotoxin-treated mice displayed a translocation of FGFR3 from the cellular membrane to the nucleus of tumor cells. A similar effect has been described previously in human breast cancer models (46), suggesting that the translocation of FGFR3 from the plasma membrane to the perinuclear membrane is governed by FGF1 signaling. In glial cells and gliomas, ligand activation has been shown to cause redistribution of FGFR1 to a perinuclear location and an increase in cell proliferation (47, 48). FGFR3 IIIb, the isoform present in epithelial bladder cells, has a restricted binding capacity to only FGF9 and FGF1 (22). Previously, we showed that binding of scFv 3C to FGFR3 was able to compete with FGF9 but not with FGF1 binding and that this effect was probably due to the higher affinity of this ligand (23). This supports the hypothesis that FGF1 binding to the receptor would result in translocation of FGFR3 to the nucleus. Because FGF1 may have a higher affinity for the FGFR3 than the scFv antibody we used, generation of higher-affinity variants of scFv 3C to effectively compete with this ligand binding may induce a higher antiproliferative activity in the xenograft model.

In summary, we have shown that a fusion construct composed of an anti-FGFR3 single-chain antibody and the toxin rGel results in a construct that is capable of specific delivery of the rGel component to the cytoplasm of bladder cancer cells expressing FGFR3. Tumor xenografts from immunotoxin-treated mice showed a larger proportion of tumor cells undergoing apoptosis than saline-treated controls. A translocation of FGFR3 to the nucleus was also observed in tumors from immunotoxin-treated mice. This translocation to the nucleus might keep cell proliferation as a mechanism of escape. These data support a potential use of FGFR3-specific fusion toxins for bladder cancer therapy.

We thank Silvia Romero and the technical staff of the Confocal Unit and the Comparative Pathology Unit of the Centro Nacional de Investigaciones Oncológicas for technical assistance.