Human bladder RT4, RT112, 5637 were maintained in RPMI 1640 supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine and antibiotics (100 U/mL penicillin and 100 µg/mL streptomycine-sulphate); breast MDA-MB 468 and glioblastoma U87 cancer cell lines as well as skin fibroblast cells were maintained in DMEM supplemented with 10% FCS, 2 mM L-glutamine and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycine-sulphate). Murine MB49 bladder cancer cells were cultured in DMEM, supplemented with 10% FCS, 2 mM L-glutamine, antibiotics (100 U/mL penicillin and 100 µg/mL streptomycine-sulphate) and 1 mM sodium pyruvate.

MB49 Luc cells stably expressing luciferase were generated by transduction with a 3^rd generation lentiviral vector carrying the luciferase gene. pLenti PGK V5-LUC Neo (w623-2) was a gift from Eric Campeau, University of Massachusetts Medical School, Worcester, Massachusetts, US (Addgene plasmid # 21471). For lentivirus production, a monolayer of HEK293T cells, cultured in 10 cm² dish, were incubated with the following mixture: transfer vector (10 μg), packaging vector Δr 8.74 (6.5 μg), Env VSV-G vector (3.5 µg), REV vector (2.5 μg) in 450 μl double distilled water, 50 μl calcium chloride (2.5 M) and 500 μl Hank’s buffered saline (2-fold). Sixteen hours later, the medium was replaced with culture medium and 24 hours later the medium was collected and 0.22 µm-filtered to recover virus particles. Virus particles were then used to transduce MB49 cells. Infected cells were then cultured in presence of G418 antibiotic (0.5 mg/ml) for fifteen days.

SAP fused with ACDCRGDCFCG or CGGSGG at its N-terminus were prepared by GenScript (New Jersey, USA). The nucleotide sequences were obtained from the corresponding amino acid sequences of saporin S and optimized for the expression in E.coli. A GGSSRSS sequence was interposed between ACDCRGDCFC and SAP as a spacer and a 6xHis tag was added at the C-terminus to allow the purification by affinity chromatography. The whole encoding sequences were inserted into the pET22b(+) vector (Novagen), forming the pET22b(+)-RGD-SAP (5′-NdeI- ACDCRGDCFCG-GGSSRSS-SAP-HHHHHH-EcoRI-3′) and the pET22b(+)-CYS-SAP (5′-NdeI-CGGSGG-SAP-HHHHHH-EcoRI-3′) expression vectors. Ligation products were used to transformed the E.coli strain DH5alpha (Invitrogen).

The expression of RGD-SAP and CYS-SAP in transformed BL21(DE3) E.coli cells (Novagen) was induced for 3 hours at 37°C with 0.1 mM IPTG. Bacterial pellet from 1 L culture was resuspended in 15 ml of 50 mM Tris-HCl, pH 7.5, supplemented with a cocktail of protease inhibitor (Sigma-Aldrich). Soon after, 10 mM of imidazole, lysozyme (250 µg/ml) and DNAse (20 µg/ml) were added. The bacterial solution was then incubated on ice for 45 min, subjected to 3 cycles of sonication (using a UW3100 Bandelin sonicator operating at 60% power; 2 min cycle duration with 1 sec pulse and 1 sec pause), and centrifuged at 4°C for 25 min at 10000 x g. Soluble RGD-SAP and CYS-SAP contained in the supernatant were then purified by metal chelate affinity chromatography using a HisTrap HP 5 ml column (GE Healthcare Life Sciences) equilibrated in Tris-HCl 50 mM pH 7.5, 300 mM NaCl supplemented with 10 mM imidazole (and 5 mM DTT for CYS-SAP) operated with the AktaPurifier10 FPLC system. To elute proteins, imidazole concentration was increased step by step up to 500 mM in 20 column volumes (CV). The fractions containing the target proteins were dialyzed against 50 mM Tris-HCl, pH 7.5 (CYS-SAP) or 50 mM bicine, pH 8.2 (RGD-SAP) at 4°C for 16 hours. A cation exchange chromatography on HiTrap SP Sepharose FF column (GE Healthcare Life Sciences) was then performed for further purification, using a 20 CV gradient up to 1 M NaCl for protein elution. The proteins were then concentrated using 10 KDa cutoff Amicon centrifugal filters (Millipore-Sigma) and dialyzed against PBS with slide A lyzer dialysis cassette (Thermo Fisher). All solutions used in purification steps were prepared with sterile and endotoxin-free water (S.A.L.F. Laboratorio Farmacologico SpA, Bergamo, Italy). Protein concentration was measured using the BCA Protein Assay DC^™ Kit (BioRad). Protein purity and identity was checked by SDS-PAGE and Western blotting. Both CYS-SAP and RGD-SAP showed comparable yields ranging from 0.6 to 1.2 mg/l of bacterial culture.

Cells were washed twice with cold PBS, collected by scraping and centrifuged 5 min at 300 g. Cells were lysed for 30 min on ice in ice-cold buffer (50 mM Tris-HCl, pH 7.5, containing 150 mM NaCl, 2 mM NaF, 1 mM EDTA, 1 mM EGTA, 1 mM Na₃VO₄,1 mM PMSF, 75 mU/ml aprotinin (Sigma-Aldrich), 1% TritonX-100 and a proteinase inhibitor cocktail^® (Sigma-Aldrich). Cell lysates were centrifuged at 10000 x g at 4°C for 10 min and the supernatants were recovered and quantified for total protein content. Equal amounts of cell protein extracts were separated by SDS-PAGE under reducing conditions unless stated otherwise. For western blot analysis, proteins were transferred onto a nitrocellulose membrane, incubated with 5% non-fat powdered milk in TBS-T (0.5% Tween-20) for 1 hour and then with the following antibodies: anti-saporin anti-serum (rabbit, 1:5000), anti-caspase 3 (rabbit, 1:2000, clone E87, Abcam), anti-beta actin (mouse, 1:10000, clone AC-15, Sigma-Aldrich). The antibody binding was detected using a secondary horseradish peroxidase conjugated antibodies (donkey anti-mouse/rabbit IgG HRP-linked, GE Healthcare) and an enhanced chemiluminescent (ECL, Merck Millipore).

Seed SAP used as positive control for Western blot analysis was purchased from Advanced Targeting Systems, which had purified it from the seeds of the Soapwort plant (Saponaria officinalis).

Cultured cell lines were detached by TripLE Express (Gibco) to preserve receptor integrity, washed with PBS containing 1% FCS and incubated with PE-conjugated Ab specific for human αvβ3 and αvβ5 (R&D System) and FITC-conjugated Ab specific for human αvβ6 integrins (NovusBio). For receptor detection, cells were incubated with the fluorescently labelled Ab at 4°C for 30 min. Stained cells were resuspended in 100 μL of PBS containing 1% FCS. Samples were run through an Accuri™ flow cytometer (BD Biosciences). All data were analysed by FCS Express and expressed as relative fluorescence intensity (RFI), calculated as follows: mean fluorescence intensity after mAb staining/mean fluorescence intensity after isotype-negative control staining. Analysis was done on 20,000 gated events acquired per sample.

Cultured cell lines (5 x 10³ cells/well) were seeded in 96 wells plates and incubated for 72 h with various amounts of RGD-SAP or CYS-SAP at 37°C, 5% CO₂. Cell viability was then quantified by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide staining (MTT) (working solution 0.5 µg/ml). After 1 h of incubation, the supernatants were removed, the formazan crystals were dissolved with dimethyl sulfoxide and the absorbance at 570 nm was measured using a microtiter plate reader. Competitive experiments were performed as described above using 100 nM of RGD-SAP or 1000 nM of CYS-SAP in the presence of 5000 nM of ACDCRGDCFCG peptide for 48 and 72 hours.

To induce caspase 3 activation, cells were treated with 30 nM RGD-SAP or 2 mM DTT for 48 and 72 hours.

Studies on animal models were approved by the Institutional Animal Care and Use Committee (Institutional Animal Care and Use Committee, IACUC) and performed according to the prescribed guidelines. C57BL/6 female mice (7 weeks old, Charles River, Calco, Italy) were challenged with subcutaneous injection in the left flank of 3-5 x 105 MB49 living cells; 5 days later, 4-6 mice/group were intravenously administered with various doses of RGD-SAP or CYS-SAP diluted in sodium chloride (0.9%, i.v., 200 µl). Tumor growth was estimated by calculating the volume using the formula r1 × r2 × r3 × 4/3 π, where r1 and r2 are the longitudinal and lateral radii, and r3 is the thickness of tumors protruding from the surface of normal skin. Animals were euthanized when tumors reached 10 mm in diameter, became ulcerated or a 15% animal body weight loss was measured.

Orthotopic syngeneic tumor were developed according to the procedure described by Loskog et al. (27). Briefly, C57BL/6 female mice were anesthetized with ketamine/xylazine (80/10 mg/kg) and catheterized (PE50 catheter, BD Biosciences) using lubricated catheters with 2.5% lidocaine-containing gel (Luan). To enhance tumor engraftment, 100 µl of poly-L-lysine (0.1 mg/ml, mw 70000–150000, Sigma Aldrich), was injected transurethrally into the bladder and left in place for 30 min, then bladder was washed with PBS and instilled with 5x104 MB49 Luc diluted in serum-free medium (100 µl/mice), 30 min later the catheters were removed. On day 7, mice were treated with mitomycin C (MMC) alone (n=10), administered transurethrally and kept in the bladder for 1 hour (50 µg in 100 µl of PBS, every 4 days for 2 times), or combined with RGD-SAP (n=5) or RGD-SAP alone (i.v., 200 µl, every 5 days, for 3 times) (n=10), starting at day 9. Control group of mice was treated with vehicle (sodium chloride, 0.9%, i.v., 200 µl) (n=10). Orthotopic tumor engraftment and growth was monitored once a week by in vivo bioluminescence imaging (IVIS). Tumor growth was estimated by acquiring the bioluminescence signal (BLI) and it expressed in total photon flux. Mice were euthanized when the BLI intensity suddenly drop, due to irreversible necrosis and accompanied with hematuria or animal lethargy.

Blood samples were collected from the retroorbital plexus of anesthetized mice using 4% isoflurane at the end of each experiment or before animal sacrifice. Blood samples were left at room temperature for at least 30 min before being processed and then centrifuged (800 x g, 10 min) for serum separation. Serum albumin, aspartate transaminase, alanine transaminase, creatinine and urea were determined by using an automated analyzer (ScilVet ABC plus and Idexx Procyte analyzers) according to the manufacturers’ instructions. Standard controls were run before each determination.

All in vitro experiments were performed at least in triplicate. Mouse experiments were performed using at least 4 mice per group. When appropriate, statistical significance was determined using a 2-tailed Student’s t test. For tumor growth analyses, we performed one-way ANOVA statistical analysis. Survival curves were compared using the log rank test. Tests symbols mean: *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

RGD-SAP, consisting of RGD-4C fused to the N-terminus of SAP was produced in E.coli cells by recombinant DNA technology. In parallel we have also produced a SAP variant with a Cys residue in place of the RGD-4C domain (CYS-SAP) ( Figures 1A, B ).

To facilitate their purification and to promote endosomal escape of the toxin into recipient cells, both products were genetically engineered to express a histidine tag at the C-terminus (28). Since SAP can inactivate prokaryotic ribosomes, their production in E.coli was induced with IPTG for only 3 h. Western blot analysis of the purified RGD-SAP, performed under reducing and non-reducing conditions, showed a single band of 30 kDa as expected for monomers ( Figure 1C ), suggesting that inter-chain disulfide bonds between the ACDCRGDCFCG domains of different molecules were not formed. In contrast, two bands corresponding to monomers and dimers were detected in CYS-SAP, suggesting that an inter-chain disulfide bond was formed between the N-terminal cysteines of this protein ( Figure 1C ).

RGD-4C can recognize αv-integrins with different affinities (29). Integrins, such as αvβ3, αvβ5 and αvβ6, are present on a variety of tumor cells (24–26, 30). Therefore, to identify tumor cells that could be exploited as targets to validate the targeting properties of RGD-SAP in vitro, we characterized the surface expression of integrins by various cancer cell lines, including U87 glioblastoma cells, RT4, RT112 and 5637 bladder cancer cells, MDA-MB-468 breast cancer cells and normal fibroblasts, by flow cytometry. The results showed that U87 cells express high levels of αvβ3, but not of αvβ5 and αvβ6, whereas RT4, RT112 and 5637 and MDA-MB-468 cells showed a moderate-high positivity for αvβ5 and αvβ6, but not of αvβ3. Normal fibroblasts expressed none of these integrins ( Figure 2A ).

To assess whether the RGD domain can increase the cytotoxic effects of saporin against cancer cells, we then tested the cytotoxic effects of RGD-SAP and CYS-SAP on these cell lines. A stronger cytotoxic effect of RGD-SAP, compared to CYS-SAP, was observed with all cancer cells, but not with normal fibroblasts ( Figure 2B ). Considering that fibroblasts do not express αvβ3 and αvβ5, i.e. integrins known to be recognized by RGD-4C, these data suggest that the stronger cytotoxic effects of RGD-SAP against cancer cells was mediated by an RGD-dependent targeting mechanism.

To verify this hypothesis, we performed competition experiments with the free RGD-4C peptide on 5637 cell line, selectively sensitive to RGD-SAP and representative of a human muscle-invasive model of bladder cancer. To this end, cells were treated with 0.1 µM RGD-SAP or 1 µM CYS-SAP, concentrations reflecting the different sensitivity of cells towards the two toxins, in the presence or absence of an excess of free RGD-4C. As expected, RGD-4C significantly decreased the activity of RGD-SAP, but not that of CYS-SAP ( Figure 3A ). This result lends support to the hypothesis that indeed the RGD domain of RGD-SAP contribute to the cytotoxic activity of this conjugate by a receptor-mediated targeting mechanism, likely involving integrins.

It is well known that SAP induces cell apoptosis. Thus, we then investigated the activation of programmed cell death by analyzing caspase 3 in cells treated with RGD-SAP. To this aim, 5637 and MDA-MB-468 epithelial cells were incubated with RGD-SAP or DTT, a positive control, for 48 and 72 h. As shown in Figure 3B , caspase 3 activation was detectable as cleaved form in all samples treated with the toxin. These and the above results indicate that fusing the RGD moiety to SAP enhances the delivery and uptake of the toxin.

The anti-tumor activity of RGD-SAP and CYS-SAP were then investigated using C57BL/6 mice bearing subcutaneous murine MB49 urothelial carcinoma cells. A preliminary experiment, performed in vitro, showed that RGD-SAP could kill these cells ( Figure 4A ).

Thus, mice were treated, systemically, with 1 mg/kg of RGD-SAP or CYS-SAP at day 5. The treatment was repeated three times every five days ( Figure 4B ). RGD-SAP could reduce tumor growth after the second administration, leading to marked reduction of the tumor volume in all mice and complete eradication in one mouse ( Figures 4C–E and Supplementary Figure 1A ). On the contrary, CYS-SAP did not show any effect on tumor growth, suggesting that the therapeutic effect of RGD-SAP crucially involved the RGD domain. However, the dose of RGD-SAP used in this experiment also caused loss of body weight at day 19 ( Figure 4D ) and severe necrosis in the tails where the drug was injected.

To determine the minimal effective, non-toxic dose of RGD-SAP tumor-bearing mice were treated with 0.75, 0.5, 0.25 mg/kg of RGD-SAP at days 5, 10, 15 after tumor implantation, as described above. The doses of 0.75 and 0.5, but not 0.25 mg/kg, caused a significant delay of tumor growth, pointing to a dose-dependent effect ( Figure 5A and Supplementary Figure 1B ), without causing loss of body weight ( Figures 5B–D ).

The toxicological effects RGD-SAP was further investigated. To this aim, we collected blood samples at the end of each experiment or before animal sacrifice and analyzed biochemical parameters of liver and kidney toxicity (albumin, alanine transaminase, aspartate transaminase for liver toxicity, and creatinine and urea for kidney toxicity). As shown in Figure 6 , RGD-SAP systemic toxicity was dose dependent and apparent only with the highest dose used (1 mg/kg). In fact, significantly higher values of AST and CHE were measured in mice treated with 1 mg/kg RGD-SAP, but not with the lower doses.

The anti-tumor efficacy of RGD-SAP was then investigated in an orthotopic model of urothelial carcinoma. To this aim, we genetically engineered MB49 cells to express luciferase (MB49-luc). MB49-luc cells were then orthotopically implanted into immunocompetent C57BL/6 mice. Tumor engraftment and growth, as monitored by in vivo bioluminescence imaging, occurred in 100% of mice in 5-7 days after cells inoculation. This tumor model resembles advanced bladder cancer and is characterized by high proliferation rate accompanied by protrusion of the mass into the bladder lumen, obstruction of urethra, hematuria and necrosis in the tumor core a few days upon engraftment (27). These features can lead to an inadequate drug delivery to the tumor mass (27). To overcome this limitation, we decided to use mitomycin C (MMC) as a tool to slow down the tumor growth and delay necrosis formation, thereby allowing the toxin to reach tumor cells and exert its specific effect. MMC is one of the most widely used agents for preventing recurrence of superficial bladder cancer in clinics, usually administered intravesically after transurethral resection of cancer lesions (31–33). Of note, MB49 bladder cancer cells were extremely sensitive to MMC (IC50 ~2 μg/ml) ( Figure 7A ).

At the time of tumor detection (day 7 after cells infusion into the bladder) two experimental groups were treated with MMC through transurethral administration. A second dose of MMC was given after four days. In between, mice received a first dose of RGD-SAP (systemically, 0.5 mg/kg) or vehicle, which was repeated for three times ( Figure 7B ). One group of mice was treated with RGD-SAP alone, as control. As previously demonstrated (34), in this model, after an initial increase, the bioluminescent signal drops in a time dependent manner, due to the reduced light emission from extended necrotic and hemorrhagic areas compared with vital tumor areas. In the control group the bioluminescent signal increased, reaching the maximum value after 11 days, and then started decreasing. Although a similar behavior was observed after treatment with MMC or RGD-SAP alone, the combination of these drugs greatly limited the increase of the signal ( Figure 7C ), which implies a relevant anti-tumor effect. Accordingly, a significant effect on mice survival was observed in the group of mice treated with RGD-SAP/MMC combination ( Figures 7E, F ), as 80% of mice were still alive when the experiment ended after 23 days. Histochemical analysis of the tumors, explanted after mouse scarification, showed that only one mouse treated with the combined therapy had a necrotic tumor, while all the others did not display any necrotic area, suggesting that this treatment consistently reduced the tumor growth. Furthermore, no evidence of toxicity was observed, judging from the lack of significant changes of body weight ( Figure 7D ), as well as of albumin, alanine transaminase or urea levels in the bloodstream ( Figure 8 ).