The Thymidine Phosphorylase Inhibitor 5 -O-Tritylinosine (KIN59) Is an Antiangiogenic Multitarget Fibroblast Growth Factor-2

Published OnlineFirst February 1, 2012; DOI: 10.1158/1535-7163.MCT-11-0738

Molecular Cancer Therapeutic Discovery Therapeutics

The Thymidine Phosphorylase Inhibitor 50-O-Tritylinosine (KIN59) Is an Antiangiogenic Multitarget Fibroblast Growth Factor-2 Antagonist

Sandra Liekens1, Annelies Bronckaers1, Mirella Belleri3, Antonella Bugatti3, Rebecca Sienaert1, Domenico Ribatti4, Beatrice Nico4, Alba Gigante5, Elena Casanova5, Ghislain Opdenakker2, María-Jesus Perez-P erez 5, Jan Balzarini1, and Marco Presta3

Abstract 50-O-Tritylinosine (KIN59) is an allosteric inhibitor of the angiogenic enzyme thymidine phosphorylase. Previous observations showed the capacity of KIN59 to abrogate thymidine phosphorylase–induced as well as developmental angiogenesis in the chicken chorioallantoic membrane (CAM) assay. Here, we show that KIN59 also inhibits the angiogenic response triggered by fibroblast growth factor-2 (FGF2) but not by VEGF in the CAM assay. Immunohistochemical and reverse transcriptase PCR analyses revealed that the expression of laminin, the major proteoglycan of the basement membrane of blood vessels, is downregulated by KIN59 administration in control as well as in thymidine phosphorylase- or FGF2-treated CAMs, but not in CAMs treated with VEGF. Also, KIN59 abrogated FGF2-induced endothelial cell proliferation, FGF receptor activation, and Akt signaling in vitro with no effect on VEGF-stimulated biologic responses. Accordingly, KIN59 inhibited the binding of FGF2 to FGF receptor-1 (FGFR1), thus preventing the formation of productive heparan sulphate proteoglycan/FGF2/FGFR1 ternary complexes, without affecting heparin interaction. In keeping with these observations, systemic administration of KIN59 inhibited the growth and neovascularization of subcutaneous tumors induced by FGF2-transformed endothelial cells injected in immunodeficient nude mice. Taken together, the data indicate that the thymidine phosphorylase inhibitor KIN59 is endowed with a significant FGF2 antagonist activity, thus representing a promising lead compound for the design of multitargeted antiangiogenic cancer drugs. Mol Cancer Ther; 11(4); 817–29. 2012 AACR.

Introduction Blood vessel formation is a multistep process, which Angiogenesis is the process by which new blood involves (i) the degradation of the basement membrane capillaries are generated from preexisting blood vessels. and extracellular matrix (ECM) surrounding the existing It is an essential and well-regulated process during vessel, (ii) migration, (iii) proliferation of the endothelial embryogenesis, wound healing, and in the female repro- cells into the perivascular space, (iv) the formation of new ductive cycle. However, angiogenesis is also associated vascular loops, with (v) the development of tight junc- with numerous pathologic processes (1–3) and is consid- tions, and (vi) the deposition of a new basement mem- ered one of the hallmarks of cancer, being indispensable brane and ECM. These different steps are regulated by a for tumor growth and metastasis (4). Therefore, the tumor multitude of inhibitory and stimulatory signals, including vasculature is an attractive target for anticancer therapy. growth factors, chemokines, adhesion molecules, and enzymes (2, 3). Among them, the VEGF family and cog- nate tyrosine kinase receptors VEGFR1–3 have repre- Authors' Affiliations: 1Laboratory of Virology and Chemotherapy, and sented a central focus in tumor angiogenesis research 2Laboratory of Immunobiology, Rega Institute for Medical Research, KU Leuven, Leuven, Belgium; 3Unit of General Pathology and Immunology, (5, 6), and the VEGF/VEGFR system has been extensively Department of Biomedical Sciences and Biotechnology, University of investigated as a target for antineoplastic therapy (7). In Brescia, Brescia; 4Section of Anatomy and Histology, Department of Basic 2004, the anti-VEGF monoclonal antibody bevacizumab Medical Sciences, University of Bari Medical School, Bari, Italy; and 5Instituto de Química Medica, IQM-CSIC, Madrid, Spain was approved as first-line therapy in combination with standard 5-fluorouracil–based chemotherapy in patients Current address for A. Bronckaers: Laboratory of Histology, Biomedical Research Institute (BIOMED), Hasselt University, Diepenbeek, Belgium. with advanced colorectal cancer. Since then, agents that target the activity of VEGFRs, such as the tyrosine kinase Corresponding Author: Sandra Liekens, Rega Institute for Medical Research, Minderbroedersstraat 10 blok x-bus 1030, Leuven B-3000, inhibitors sunitinib and sorafenib, have been approved for Belgium. Phone: 32-16-337355; Fax: 32-16-337340; E-mail: use in selected cancers (8). [email protected] Even though VEGF plays a central role in switching on a doi: 10.1158/1535-7163.MCT-11-0738 proangiogenic phenotype in most tumors, neoplastic, 2012 American Association for Cancer Research. stromal, and infiltrating cells may produce a plethora of

www.aacrjournals.org 817

Liekens et al.

different proangiogenic factors during tumor progres- Our data indicate that, in addition to thymidine phos- sion, recurrences, and metastases (9). Thus, VEGF/ phorylase, KIN59 inhibits angiogenesis stimulated by VEGFR antagonists may not be effective in the treatment FGF2, without affecting VEGF activity. The FGF2 antag- of all cancers at all stages. Moreover, when the activity of onist activity of KIN59 results from its capacity to interfere VEGF is suppressed for a long period of time, resistance to with the formation of a productive HSPG/FGF2/FGFR VEGF/VEGFR2 blockade may occur (10, 11) because of ternary complex, required for receptor activation and the expression of other angiogenic factors, including biologic activity. In addition, the compound inhibited the fibroblast growth factor-2 (FGF2) and thymidine phos- growth of FGF2-overexpressing vascular tumors in mice. phorylase (2, 9, 12). These observations call for the devel- To our knowledge, KIN59 represents the first compound opment of novel antiangiogenic compounds, possibly that combines thymidine phosphorylase and FGF2 antag- characterized by a multitargeted mechanism of action. onist activity. Thus, KIN59 might represent a promising FGF2 is one of the best characterized members of a large lead compound for the design of a new class of multi- superfamily of structurally related cytokines that have targeted antiangiogenic cancer drugs. potent mitogenic and angiogenic properties (13). Their biologic activities depend on the binding to high-affinity Materials and Methods tyrosine kinase FGF receptors (FGFR) and low-affinity Compounds heparan sulphate proteoglycans (HSPG). The latter inter- KIN59 (Fig. 1A) was synthesized as described previ- action prevents the proteolytic degradation of FGF2 in the ously (23, 29). extracellular environment and induces growth factor oligomerization, which facilitates FGFR dimerization and subsequent transactivation (13). As such, various com- Cell cultures pounds, such as soluble heparin, polysulfonated com- Fetal bovine aortic endothelial GM7373 cells (30) were pounds, and FGF2-binding peptides, which abrogate the obtained from the National Institute of General Medical interaction of FGF2 with HSPGs and/or FGFRs, have been Sciences (Bethesda, MD) and Human Genetic Mutant shown to inhibit the biologic activity of FGF2 and to Cell Repository (Institute for Medical Research, Camden, possess antitumor activity in vivo (14–18). NJ) in 1990. No authentication was done since then. These Thymidine phosphorylase is an intracellular enzyme cells were used to generate FGFR1-overexpressing that catalyzes the conversion of thymidine and inorganic (GM7373-FGFR1; ref. 14) and VEGFR2-overexpressing phosphate into thymine and a-2-deoxy-D-ribose-1-phos- (GM7373-VEGFR2; ref. 31) cells in our laboratory. phate. Thymidine phosphorylase is upregulated in sev- BALB/c mouse aortic endothelial 22106 cells (MAEC) eral tumors, and high thymidine phosphorylase expres- and brain microvascular endothelial 10027 cells (MBEC) sion is associated with increased angiogenesis, tumor cell were obtained from R. Auerbach (University of Wiscon- metastasis, and poor prognosis (19). The enzymatic activ- sin, Madison, WI) in 1995 and were characterized by us ity of thymidine phosphorylase, and in particular, the (expression of endothelial markers, including oxLDL dephosphorylated reaction product 2-deoxy-D-ribose, uptake, CD31, and KDR expression in 1997 as described; was found to be essential for its angiogenic effect (20, ref. 32). No further authentication was done since then. 21). In contrast to FGF2 or VEGF, no receptor for thymi- MAECs were used to generate FGF2-overexpressing dine phosphorylase (or 2-deoxy-D-ribose) has been iden- FGF2-T-MAE cells in our laboratory. Wild-type Chinese tified. Moreover, thymidine phosphorylase is not a hamster ovary (CHO)-K1 cells and A745 CHO cell growth factor. It is hypothesized that 2-deoxy-D-ribose, mutants that were used in our laboratory to generate which can diffuse out of the cell, directly activates integ- FGFR1-overexpressing CHO A745 flg-1A cells (33) were a b a b rins v 3 and 5 1 and focal adhesion kinase on endothe- provided by J.D. Esko (University of California, La Jolla, lial cells, leading to cell migration (22). CA) in 2000. No authentication was done since then except We previously showed that the nucleoside derivative to confirm routinely that CHO A745 cell mutants do not 50-O-tritylinosine (KIN59) inhibits the catalytic activity express HSPGs. GM7373 cells and MBECs were main- of thymidine phosphorylase (23). Accordingly, KIN59 tained in Dulbecco’s Modified Eagle’s Medium (DMEM; completely prevented thymidine phosphorylase– Invitrogen) supplemented with 10% FBS (Biochrom induced angiogenesis in the chicken chorioallantoic mem- AG) and 10 mmol/L HEPES (Invitrogen). CHO-K1 cells brane (CAM) assay. Moreover, the compound was far and CHO A745 flg-1A were grown in Ham’s F-12 more potent than other thymidine phosphorylase inhibi- medium supplemented with 10% FBS. All cells were tors and also inhibited basal, developmental vasculariza- grown at 37C in a humidified incubator with a gas phase tion of the CAM (i.e., in the absence of exogenously added of 5% CO2. angiogenic factors; refs. 23–25). This suggests that KIN59 FGF2-T-MAE cells, which express high levels of the might also target angiogenic molecules other than thymi- Mr 18,000, Mr 22,000, and Mr 24,000 molecular weight dine phosphorylase. Because VEGF and FGF2 have been isoforms of FGF2 (34), were transfected with the firefly shown to be expressed in the developing CAM (26–28), we luciferase–expressing vector pGL4.50 (Promega), using studied the effect of KIN59 on the angiogenic activity of Nanojuice (Merck) according to the manufacturer’s in- these growth factors. structions. Single-cell clones, stably expressing luciferase,

818 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

A B Control Vehicle KIN59 O

N NH

O N N O

HO OH * Figure 1. Effect of KIN59 on KIN59 * endogenous angiogenesis and * neovascularization induced by thymidine phosphorylase (TP), FGF2, or VEGF in the CAM gelatine C Control KIN59 sponge assay. A, chemical structure of KIN59. B, at day 9, gelatin sponges containing either 5 units thymidine phosphorylase, 500 ng FGF2, 500 ng VEGF, or PBS (basal angiogenesis) alone or with KIN59 (125 nmol, 10 mL) or vehicle (40% cremophor in PBS, PBS 10 mL) were placed on the CAM. Two days later, the sponges were processed as described (see Materials and Methods) and the 500 µm 500 µm number of blood vessels was determined. The experiments were repeated 3 times with at least 5 eggs per group. Results are expressed as mean SEM. , P < 0.001 compared with control and vehicle. C, histologic TP examination (H&E staining) of CAM sponges treated with PBS, thymidine phosphorylase, FGF2, or VEGF in the absence (left) or presence (right) of 125 nmol KIN59. Numerous blood 500 µm 500 µm vessels developed in the sponges treated with thymidine phosphorylase, VEGF, or FGF2 compared with sponges containing PBS. CAMs stimulated with thymidine phosphorylase or FGF2 FGF2 and treated with 125 nmol of KIN59 are characterized by the absence of blood vessels, whereas no inhibition of angiogenesis can be observed in CAMs treated with VEGF and KIN59. 500 µm 500 µm

VEGF

500 µm 500 µm

were established, and the brightest clone (F2T-luc2.9) was inducing properties as the parent cells, but were not selected for further experiments. F2T-luc2.9 cells were further characterized. maintained in DMEM, supplemented with 10 mmol/L HEPES, 10% FBS, 500 mg/mL geneticin (Invitrogen), Cell proliferation assays and 500 mg/mL hygromycin (Invitrogen). These cells GM7373 cells were seeded in 24-well plates at 60,000 showed comparable growth, angiogenic and tumor- cells per cm2. After overnight incubation at 37C, the cells

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 819

Liekens et al.

were incubated for 24 hours in fresh medium containing Recombinant human sFGFR1(IIIc)/Fc chimera (5 mg/mL 0.4% FBS, 30 ng/mL FGF2 or VEGF, and different con- in 10 mmol/L sodium acetate, pH 3.0) was allowed centrations of KIN59. At the end of the incubation period, to react with a flow cell of a CM5 sensor chip. These cells were trypsinized and counted. In a second set of experimental conditions allowed the immobilization of experiments, cells were treated with 30 ng/mL of FGF2 6.183 RU, equal to 0.34 pmol/mm2 of sFGFR1(IIIc)/Fc in the presence of KIN59. After 24 hours, cells were chimera (18). Similar results were obtained for immo- washed and incubated with fresh medium containing bilization of the irrelevant Noggin-Fc protein chimera, 10% FBS for another 24 hours. Next, cells were trypsinized here used as a negative control and for blank subtrac- and counted. tion. In both experimental models, FGF2 (50 nmol/L) MBECs were seeded in 24-well plates at 25,000 cells was injected in the presence of increasing concentra- per cm2. After overnight incubation at 37C, the cells tions of the compound under test in HBS-EP buffer were incubated in fresh medium containing 0.4% FBS, (0.01 mol/L HEPES, pH 7.4 plus 0.005% surfactant 30 ng/mL FGF2 or VEGF, and different concentrations P20, 0.15 mol/L NaCl, 3 mmol/L EDTA) for 4 minutes of KIN59. After 48 hours, cells were trypsinized and to allow the association of FGF2 with the immobilized counted. ligand. The response (in RU) was recorded as a function F2T-luc2.9 cells were seeded in 48-well plates at 10,000 of time. cells per cm2. After 24 hours, the medium was replaced, and increasing concentrations of KIN59 were added. The Western blot analysis cell cultures were incubated for 3 days, trypsinized, and Serum-starved FGFR1-overexpressing GM7373 cells counted in a Coulter counter analyzer (Analis). (GM7373-FGFR1 cells; ref. 14) and VEGFR2-overexpressing GM7373 cells (GM7373-VEGFR2 cells; ref. 31) FGF2-mediated cell–cell adhesion assay were incubated with 60 mmol/L KIN59 or vehicle for 30 This was conducted as described previously (35). Brief- minutes, after which 10 ng/mL FGF2 or 30 ng/mL ly, wild-type CHO-K1 cells were seeded in 24-well plates VEGF was added. After 10 minutes of stimulation, at 150,000 cells per cm2. After 24 hours, cell monolayers cells were lysed and 50 mg aliquots of the cell extracts were washed with PBS and incubated with 3% glutaral- were analyzed by SDS-PAGE followed by Western dehyde in PBS for 2 hours at 4C. Fixation was stopped blotting with anti-P-FGFR1 antibody (Santa Cruz Bio- with 0.1 mol/L glycine, and cells were washed extensively technology), anti-p-Akt antibody (Cell Signaling Tech- with PBS. Then, CHO A745 flg-1A cells (50,000 cells nology), and anti-p-VEGFR2 antibody (R&D Systems). per cm2) were added to CHO-K1 monolayers in serum- In all the experiments, uniform loading of the gels free medium plus 10 mmol/L EDTA with or without was confirmed using anti-FAK antibody (Santa Cruz 30 ng/mL of FGF2 in the absence or presence of increasing Biotechnology). concentrations of KIN59. After 2 hours of incubation at 37C, unattached cells were removed by washing twice CAM gelatin sponge assay with PBS, and CHO A745 flg-1A cells bound to the CHO- The in vivo CAM gelatin sponge model was conduct- K1 monolayer were counted under an inverted micro- ed as described by Ribatti and colleagues (37) with scope at 125 magnification. Data are expressed as the slight modifications. Fertilized chicken eggs were incu- mean of the cell counts of 3 microscopic fields chosen at bated for 3 days at 37Catconstanthumidity.After random. All experiments were carried out in triplicate and removal of 3 mL of albumin to detach the developing repeated twice with similar results. CAM from the shell, a square window was opened in the eggshell, exposing the CAM. The window was BIAcore-binding assay sealed with cellophane tape, and the eggs were return- The ability of KIN59 to inhibit the binding of FGF2 to ed to the incubator. At day 9 of embryonic development, immobilized heparin or to FGFR1 was investigated using 2-mm3 sterilized gelatin sponges (Gelfoam, Upjohn) surface plasmon resonance (SPR) technology (BIAcore X containing the test compounds were placed on top of apparatus; GE Healthcare). the CAM. These sponges contained either thymidine Heparin was immobilized to a BIAcore sensor chip phosphorylase (5 mL containing 5 U; Sigma), FGF2 m m as described previously (36). Briefly, a CM3 sensor chip (500 ng in 5 L; Sigma), or VEGF165 (500 ng in 5 L; previously activated with 50 mL of a mixture contain- R&D systems) alone or with KIN59 (10 mL containing ing 0.2 mol/L 1-ethyl-3-(3-diaminopropyl)-carbodiimide 125 nmol in 40% cremophor-PBS) or 40% cremophor- hydrochloride (EDC) and 0.05 mol/L N-hydroxysucci- PBS (10 mL). Forty-eight hours later,thespongeswere nimide (NHS) was coated with streptavidin. Heparin was processed for zymography, reverse transcriptase PCR biotinylated at its reducing end and immobilized onto the (RT-PCR), histologic, or immunohistochemical analy- streptavidin-coated sensor chip. These experimental con- sis. Forty percent of cremophor [cremophor EL; Sigma; ditions allowed the immobilization of 80 resonance units instead of dimethyl sulfoxide (DMSO)] was used to (RU) equal to 5.8 fmol/mm2 of heparin. A sensor chip dissolve KIN59, because slight toxicity was observed coated with streptavidin alone was used as a reference in sponges treated with DMSO. Cremophor is a deriv- and for blank subtraction. ative of castor oil and ethylene oxide.

820 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

Histologic and immunohistologic analysis of the Primer (Invitrogen) according to the manufacturer’s sponges instructions. CAMs were fixed in ovo in 4% paraformaldehyde in Quantitative real-time PCR was carried out with an PBS for 1 hour, dehydrated in graded ethanol, embed- ABI Prism 7000 sequence detection system (Applied ded in paraffin, and serially sectioned at 7 mm(4mmfor Biosystems). Primers were designed according to the chick- immunohistochemistry) along a plane parallel to their en sequences of the A1 chain of laminin (50-GCCTAGTG- free surface. For histology, sections were stained with CTGATGATCC-30 and 50-GCCACCAACAGAGATGTCC- hematoxylin and eosin (H&E). To determine the num- 30) and the housekeeping genes hypoxanthine phosphori- ber of blood vessels, pictures of the CAM tissue around bosyltransferase (HPRT,50-TGGCGATGATGAACAAGGT- the gelatin sponges at a 100 magnification were taken 30 and 50-GCTACAATGTGGTGTCCTCCC-30), glyceralde- (with the Zeiss Imager Z.1 microscope, Zeiss), and the hyde-3-phosphate dehydrogenase (GAPDH,50-GGAGTCAA- blood vessels on these pictures were counted in a CGGATTTGGCC-30 and 50-TTTGCCAGAGAGGACGGC- double-blind fashion. The mean blood vessel counts 30), and b-actin (50-ATGGCTCCGGTATGTGCAA-30 and SD were determined for each analysis. For statistical 50-TGTC-TTTCTGGCCCATACCAA-30). analyses, the P values were determined using the Stu- All samples were analyzed in triplicate in a final dent t test, and P values < 0.05 were considered volume of 25 mL containing SYBR green PCR Master significant. Mix 30 (Eurogentec) containing 300 nmol/L of each After rehydration of the tissue sections, the slides primer (Invitrogen) and 40 ng of cDNA template. The were depleted of their endogenous peroxidase by incu- thermal cycle parameters used were 2 minutes at 50C, bation in 3% H2O2 for 20 minutes. The slides were 10 minutes at 95 C followed by 40 cycles of 15 seconds at washed in PBS for 15 minutes and incubated in 1% 95Cand60secondsat60C. After 40 cycles, samples trypsin (in 0.1% CaCl2)at37C for 30 minutes for wererunusingthedissociationprotocoltodetectany antigen retrieval. Next, the tissue sections were incu- nonspecific amplification. The results were evaluated by bated overnight with mouse anti-laminin antibody use of SDS software (Applied Biosystems). For each (L8271; Sigma) in PBS containing 1% bovine serum sample, the relative amount of laminin mRNA was albumin to prevent nonspecific binding. After 3 washes determined and normalized to 3 housekeeping genes b DD of 5 minutes in PBS, the slides were sequentially incu- (GAPDH, HPRT,and -actin)usingthe Ct method bated with a biotin-labeled anti-mouse secondary (38). The calibrator used was a control gelatin sponge antibody (Vector) for 60 minutes and with streptavi- incubated with cremophor. For statistical analyses, the din-peroxidase conjugate (Vector) for 30 minutes. Sec- Student t test was used, and P < 0.05 was considered tions were washed in PBS and counterstained with significant. Mayer’s hematoxylin for 10 minutes and mounted in buffered glycerin (glycergel; Dako). In the negative Tumor growth in mice controls, no primary antibodies were applied. Eight-week-old female, athymic, nude nu/nu mice, To quantify the staining, pictures of 0.45 mm2 were weighing about 25 g, were inoculated subcutaneously taken at 100 magnification with a Zeiss Imager Z.1 with 200 mLofserum-freeDMEMcontaining2 106 microscope. For each condition, at least 9 samples F2T-luc2.9 cells. KIN59 treatment was started after (3 sponges from 3 independent experiments) were taken, 24 hours and continued till the end of the experiment. and 2 sections per sample were stained and photo- The compound was administered subcutaneously at graphed. The pictures were always taken of a CAM area 15 mg/kg, twice daily (once daily during the weekend) adjacent to the sponge. Next, the stained area was calcu- at a site distant from the tumor (inoculation) site. The lated using Axiovision software. Staining was rated as a mice were imaged at regular time intervals to assess percentage of the photographed area. The mean values the growth of the luciferase-positive tumor cells. Before SD were calculated per sample and per group of samples. imaging, the mice were injected i.p. with 200 mLofa For statistical analyses, the Student t test was used, and P 40 mg/mL luciferin solution in PBS. Images were values < 0.05 were considered significant. captured by the IVIS Spectrum Imaging System (Caliper Life Sciences) and analyzed using the LivingImage Real-time RT-PCR Software (Caliper Life Sciences). Tumor size was mon- The CAM gelatin sponges were incubated in 700 mL itored by means of a caliper, and the tumor volume was RLT buffer (Qiagen) and homogenized as described calculated with the following formula: tumor volume earlier. After centrifugation for 20 minutes at 4Cand (mm3) ¼ 0.5 a b2,wherea is the longest diameter 13,000 rpm, the cleared supernatant was used for total and b is the shortest diameter. At the end of the experi- RNA isolation with the RNeasy Mini Kit (Qiagen). To ment, the tumors were excised, weighed, and proces- eliminate potential genomic DNA contamination, the sed for histologic analysis. For statistical analyses, the samples were treated with DNase I (Roche). Two hun- Student t test was used, and P < 0.05 was considered dred nanograms of total RNA was reverse-transcribed significant. All studies were done in compliance with to cDNA by means of Moloney murine leukemia virus the ethical guidelines for animal welfare of the KU reverse transcriptase (Invitrogen) and Oligo(dT)18 Leuven, Leuven, Belgium.

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 821

Liekens et al.

Histologic and immunohistochemical analysis of the (125 nmoles/sponge) strongly inhibited neovasculariza- tumors tion induced by thymidine phosphorylase or FGF2 (86% Tumors and organs were fixed in 4% paraformaldehyde and 84% inhibition, respectively; P < 0.001). In contrast, for 24 hours, rehydrated, and embedded in paraffin. the angiogenic activity of VEGF was not affected, thus Deparaffinized sections were subsequently stained with indicating that this growth factor was able to overcome H&E and examined microscopically. also the inhibitory effect exerted by KIN59 on basal angio- For immunohistochemical analysis of F2T-luc2.9 tu- genesis (Fig. 1B and C). Addition of vehicle (40% cremo- mors, deparaffinized sections were rehydrated, micro- phor in PBS) had no influence on basal angiogenesis or on waved in 10 mmol/L citrate buffer (pH 6.0) to retrieve the angiogenic properties of the stimulatory factors. the antigens, and incubated in 0.5% H2O2 for 30 minutes to Blood microvessels consist of endothelial cells sur- quench endogenous peroxidase activity. Next, the slides rounded by a basement membrane, a 3-dimensional net- were incubated for 20 minutes with 3% goat serum to work of interconnecting glycoproteins and collagens block nonspecific binding sites, washed 3 times in PBS, essential for proper endothelial cell function (39, 40). and incubated with rabbit anti-CD31 antibody (Abcam) or Because laminin is the major endothelial cell basement anti-laminin antibody (Abcam) for 1 hour. After washing membrane glycoprotein (41), we investigated whether the with PBS, the slides were incubated with peroxidase- antiangiogenic effect of KIN59 in the CAM assay was labeled goat anti-rabbit secondary antibody (Dako) for correlated to changes in laminin expression. Immunohis- 30 minutes and subsequently visualized using the Dako tochemistry of CAM implants showed that laminin Envision System (Dako). Finally, the sections were coun- immunoreactivity, present in the basement membrane of terstained with hematoxylin and mounted. blood vessels, epithelium and ECM of untreated samples, To analyze the tumor vasculature, 2 representative is abolished by KIN59 treatment (Fig. 2A and B). KIN59 pictures were taken from each tumor at 40 magnification also significantly reduced the levels of laminin in sponges and the areas of the highly vascular, vascular, and avas- treated with thymidine phosphorylase or FGF2 but not in cular zone were calculated with the item FEI Nikon sponges treated with VEGF (Fig. 2A and B). However, Software (Nikon). To quantify laminin staining, 2 sections KIN59 did not affect type I collagen, type IV collagen, or per sample were stained and for each section, 2 double- fibronectin immunoreactivity both in basal and angiogen- blind pictures were photographed. Next, the stained ic factor–stimulated CAMs, thus confirming the specific- brown immunopositive area was calculated using Axio- ity of the effect (data not shown). vision software (Zeiss). Staining was rated as a percentage On this basis, we investigated the effect of KIN59 on of the photographed area. For statistical analyses, the laminin mRNA expression in CAM implants using quan- Student t test was used, and P < 0.05 was considered titative real-time RT-PCR analysis. For each sample, the significant. relative levels of laminin mRNA were determined and normalized to 3 housekeeping genes (GAPDH, HPRT, and b-actin). KIN59 caused a 10-fold reduction in laminin Results mRNA levels in CAMs treated with thymidine phosphor- KIN59 inhibits FGF2-induced but not VEGF- ylase or FGF2, but it did not affect laminin expression in mediated neovascularization in the CAM assay VEGF-containing implants (Fig. 2C). At variance, gelatin KIN59 inhibits basal and thymidine phosphorylase– zymography on CAM extracts did not show any effect of induced angiogenesis in the CAM assay (23). The modest KIN59 on the activity of the laminin degrading matrix anti–thymidine phosphorylase activity of KIN59 could metalloproteinase (MMP)-2 (42) both in basal and angio- not explain its potent antiangiogenic activity. Therefore, genic factor–stimulated CAMs (data not shown), whereas we investigated the effect of KIN59 on angiogenesis MMP-9 could not be detected in the CAM extracts (see induced by FGF2 and VEGF, both expressed in the devel- also refs. 42, 43). oping CAM (26–28). The CAM gelatin sponge assay was Taken together, these results indicate that KIN59 selec- applied (37) because this technique allows histologic and tively suppresses developmentally regulated and thymi- immunohistochemical analysis of newly formed blood dine phosphorylase- or FGF2-triggered blood vessel for- capillaries and the surrounding ECM. mation and laminin deposition in the chick embryo CAM First, increasing amounts of KIN59 (10–250 nmol) were with no effect on VEGF-induced angiogenesis. applied onto the CAM in the absence of any exogenously added angiogenic factor. The maximal antiangiogenic KIN59 inhibits FGF2- but not VEGF-stimulated effect was obtained with 125 nmoles per sponge (data endothelial cell growth not shown), and this concentration was used in further FGF2 and VEGF, but not thymidine phosphorylase, experiments. Next, the effect of KIN59 on thymidine are potent endothelial cell mitogens. On this basis, we phosphorylase-, FGF2- and VEGF-induced angiogenesis evaluated the effect of KIN59 on endothelial cell prolifer- was evaluated. As shown in Fig. 1, thymidine phosphor- ation induced by FGF2 or VEGF. As shown in Fig. 3A, ylase, FGF2, and VEGF induced, respectively, a 2.4-, 2.5-, KIN59 inhibited FGF2-induced proliferation of bovine and 2.1-fold increase in CAM vascularization when macrovascular endothelial GM7373 cells in a dose-depen- < compared with PBS-treated CAMs (P 0.001). KIN59 dent manner with a 50% inhibitory concentration (IC50)of

822 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

A C

* * * * * *

100 µm 100 µm

Figure 2. Effect of KIN59 on the expression of laminin in the CAM. A, at day 9, gelatin sponges containing either 5 units thymidine phosphorylase (TP), 500 ng FGF2, 500 ng VEGF, or PBS with KIN59 (125 nmol) or vehicle (40% cremophor in PBS, 10 mL) were placed on the CAM. Two days later, the sponges were processed for immunohistochemical evaluation. The area immunoreactive for laminin was quantified as described in Material and Methods. The experiments were repeated 3 times with at least 3 eggs per group. Results are expressed as mean SEM. , P < 0.001 compared with vehicle. B, immunohistologic detection of the distribution of laminin in sponges treated with thymidine phosphorylase and vehicle (left) or 125 nmol KIN59 (right). C, real-time RT-PCR analysis for laminin in gelatin sponges. mRNA was purified from CAM sponges containing 5 units of thymidine phosphorylase, FGF2 (500 ng), VEGF (500 ng), or PBS with either KIN59 (125 nmol, 10 mL) or vehicle (40% cremophor in PBS, 10 mL) and real-time RT-PCR was carried out as described. For each sample, the relative amount of laminin mRNA was determined and normalized to 3 housekeeping genes (GAPDH, HPRT, and b-actin). The experiments were repeated 3 times with at least 3 eggs per group. Results are expressed as mean SEM. , P < 0.001. KIN59 significantly decreased the mRNA and protein level of laminin in sponges treated with PBS, thymidine phosphorylase, and FGF2, whereas it had no effect on the laminin expression in VEGF-treated sponges.

5.8 mmol/L, a potency 10 times higher than that shown to adhesion model was used in which the growth factor ¼ inhibit the proliferation induced by 10% FBS (IC50 63 mediates the interaction in trans of HSPG-deficient CHO mmol/L). In agreement with the CAM data, KIN59 (at cells overexpressing FGFR1 to a monolayer of HSPG- 30 mmol/L) did not affect the growth-promoting activity bearing CHO-K1 cells (35). As shown in Fig. 4A, KIN59 of VEGF in GM7373 cells (Fig. 3B). Similar results were exerts a significant inhibitory activity on FGF2-mediated ¼ m obtained when KIN59 was tested on murine microvascular cell–cell adhesion in this assay (IC50 10 mol/L), indi- MBECs (data notshown). Interestingly,theinhibitory effect cating that the compound is able to interfere with HSPG/ exerted by KIN59 on FGF2-induced GM7373 cell prolifer- FGF2/FGFR1 complex formation. ation seems to be cytostatic (i.e., fully reversible after fresh On this basis, SPR was exploited to assess whether this medium replacement after 24 hours treatment with the effect was because of the ability of KIN59 to interfere with compound) for KIN59 concentrations up to 30 mmol/L and FGF2/HSPG and/or FGF2/FGFR1 interaction (18). As cytotoxic only at higher concentrations (data not shown). shown in Fig. 4B, KIN59 does not affect the binding of free Accordingly, KIN59 inhibited FGFR1 phosphorylation FGF2 to heparin immobilized on a BIAcore sensor chip and Akt activation triggered by FGF2 in FGFR1-overex- whereas it significantly inhibits the interaction of the pressing GM7373-FGFR1 cells, exerting only minor effects growth factor to an immobilized sFGFR1(IIIc)/Fc chimera. on VEGF-mediated VEGFR2 phosphorylation and Akt Thus, the FGF2 antagonist activity of KIN59 results from activation in GM7373 cells overexpressing VEGFR2 its ability to inhibit the interaction of FGF2 with its high- (Fig. 3C). affinity tyrosine kinase receptor. As a consequence, formation of the ternary HSPG/FGF2/FGFR1 complex is ham- KIN59 inhibits the interaction of FGF2 with FGFR1 pered, resulting in the inhibition of FGF2 biologic activity. FGF2 exerts its biologic activity by leading to the formation of productive HSPG/FGF/FGFR ternary com- KIN59 inhibits the growth of FGF2-overexpressing plexes in target cells (13). To investigate whether KIN59 tumors may inhibit the angiogenic activity of FGF2 by preventing To investigate the ability of KIN59 to affect FGF2- the formation of this complex, a FGF2-dependent cell–cell dependent tumor growth in vivo, we took advantage of

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 823

Liekens et al.

Figure 3. Effect of KIN59 on endothelial cell proliferation. A, GM7373 cells were treated with FGF2 (30 ng/mL, *) or 10% FBS (*) in the absence or presence of different concentrations of KIN59. Cells were trypsinized and counted after 24 hours. Data are expressed as percentage of cell proliferation in the absence of compound. The experiment was carried out in duplicate and was repeated twice with similar results. B, GM7373 cells were treated with FGF2 (30 ng/mL), VEGF (30 ng/mL), or 10% FBS in the absence or presence of 30 mmol/L KIN59. Cells were trypsinized and counted after 24 hours. Data (mean SEM) are expressed as percentage of cell proliferation in the absence of compound. , P < 0.05. C, serum- starved GM7373-FGFR1 and GM7373-VEGFR2 cells were pretreated with 60 mmol/L KIN59 or vehicle for 30 minutes. Next 10 ng/mL FGF2 or 30 ng/mL VEGF were added. After 10 minutes of stimulation, cells were lysed and 50 mg aliquots of the cell extracts were analyzed by Western blotting with anti-p-FGFR1, anti-p-VEGFR2, and anti-p-Akt antibodies. Uniform loading of the gels was conﬁrmed using anti-FAK antibodies. FAK, focal adhesion kinase.

a highly tumorigenic FGF2-overexpressing mouse aortic the rate of tumor growth when compared with vehicle, as endothelial cell line (FGF2-T-MAE cells) generated in our assessed both by bioluminescence and tumor volume laboratory (34). These cells are characterized by a sus- measurements. This was confirmed by measurement of tained rate of growth in vitro and by the capacity to tumor weight at sacrifice (0.41 0.07 g vs. 0.23 0.03 g for generate opportunistic vascular lesions in nude mice as vehicle- and KIN59-treated animals, respectively; P < a consequence of an autocrine loop of stimulation trig- 0.05; Fig. 5C). No reduction in body weight was observed gered by the release of the overexpressed growth factor during the course of treatment. Also, histologic analysis of (44). Accordingly, different FGF2 antagonists inhibit the the different organs did not reveal any signs of toxicity growth of FGF2-T-MAE cells in vitro and in vivo (14, 15, 17, (not shown). 18). On this basis, as a tool for further in vivo experimen- Histologically, F2T-luc2.9 tumors consisted of a highly tation, FGF2-T-MAE cells were stably transfected with vascularized rim of spindle-shaped cells surrounding an firefly luciferase, thus generating F2T-luc2.9 cells. area with lower blood vessel density; the tumor core In vitro, KIN59 inhibited the proliferation of F2T-luc2.9 instead was poorly vascularized but consisted of viable m m cells with an IC50 of 44 9 mol/L and 65 4 mol/L tumor cells without any evidence of necrosis or apoptosis when cells were grown in 0.4% or 10% FBS, respectively. (Fig. 6A). KIN59-treated tumors, even though smaller in Next, F2T-luc2.9 cells were injected subcutaneously into size, showed a similar tissue organization. However, the flanks of female nu/nu mice and monitored for 3 immunohistochemical analysis for CD31 (Fig. 6B) showed weeks using bioluminescent imaging and caliper mea- that the avascular area of the lesions was more extended surements. Subcutaneous treatment (at a site distant from when compared with vehicle-treated tumors (62% 6% the inoculation site) with KIN59 was started 24 hours after vs. 24% 5% of the tumor section area for KIN59- and injection of the cells. Because of the low solubility of vehicle-treated animals, respectively; P < 0.05). Moreover, KIN59, the compound was administered in a mixture of in control sections, laminin was highly expressed in the DMSO/cremophor in PBS twice daily at 15 mg/kg. Con- ECM, predominantly at the tumor edge and surrounding trol mice received equal volumes of the vehicle. As shown the blood vessels (Fig. 6C). However, in KIN59-treated in Fig. 5A and B, KIN59 caused a significant inhibition in tumors, laminin was expressed at significantly reduced

824 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

phorylase (23). Although the anti–thymidine phosphor- A ylase activity of KIN59 was rather modest, the compound completely blocked thymidine phosphorylase–induced as well as endogenous angiogenesis in the chick embryo CAM assay, suggesting that this compound might also affect the activity of other proangiogenic factors expressed in the developing CAM, including FGF2 (26, 27). Indeed, previous observations had shown the capacity of a neu- tralizing anti-FGF2 antibody to fully block developmental vascularization of this embryonic adnexum (28). In this study, we show that KIN59 fully inhibits the KIN59 (µmol/L) angiogenic activity exerted by FGF2 exogenously added B to the chick embryo CAM via a gelatin sponge implant. 430 FGF2 alone Under the same experimental conditions, the compound Heparin chip + KIN59 (30 µmol/L) had no effect on the neovascular response elicited by + KIN59 (100 µmol/L) 310 VEGF. In keeping with these observations, KIN59 also inhibited FGF2-induced proliferation of endothelial 190 cells of different origin whereas it had no effect on VEGF-stimulated endothelial cell growth. Also, KIN59 suppressed FGFR1 phosphorylation and Akt activation 70 triggered by FGF2, whereas the compound did not inhi-

0 bit VEGFR2 activation and signaling triggered by VEGF. RU Thus, the FGF2 antagonist activity of KIN59 may explain its capacity to inhibit developmental vascularization of FGFR1 chip the chick embryo CAM. 250 FGF2 alone The ability of KIN59 to antagonize FGF2 activity was rather surprising. KIN59 is an atypical inhibitor of thy- 150 midine phosphorylase (19), an intracellular enzyme for + KIN59 (30 µmol/L) which no receptor has been identified. Unlike all previ- 50 ously described thymidine phosphorylase inhibitors, + KIN59 (100 µmol/L) KIN59 does not interact with the thymidine- or phos- 0 phate-binding site of thymidine phosphorylase, indicating that the compound binds to an allosteric site in the 700 150 230 310 enzyme. Computational studies on the thymidine phos- Time (s) phorylase/KIN59 complex identified a possible allosteric site involving amino acid residue Asp203 of human thy- Figure 4. Effect of KIN59 on formation of the HSPG/FGF2/FGFR1 complex. midine phosphorylase (45). At variance with thymidine A, HSPG-deficient FGFR1-transfected CHO cells were added to wild-type phosphorylase, FGF2 is an extracellular growth factor CHO-K1 monolayers in serum-free medium with 30 ng/mL FGF2 in the whose biologic activity is mediated by the activation of presence of increasing concentrations of KIN59. After 2 hours of incubation tyrosine kinase FGFRs on the cell surface (13, 46). FGF2 at 37C, the bound cells were counted under an inverted microscope. The experiment was carried out in duplicate and was repeated twice with similar also binds to HSPGs that act as low-affinity coreceptors, results. B, sensorgram overlay showing the binding of FGF2 to heparin- facilitating FGFR dimerization and signaling. Here, we coated (top) or sFGFR1(IIIc)/Fc chimera–coated (bottom) BIAcore sensor show that KIN59 abolishes the formation of this produc- chips in the absence (FGF2 alone) or presence of KIN59. The response (in tive HSPG/FGF2/FGFR1 ternary complex as a conse- RU) was recorded as a function of time. quence of its ability to prevent FGF2/FGFR1 interaction, as shown by SPR analysis in which the compound inhibits levels (10.7% 1.6% of the tumor section area for control the interaction of FGF2 with the extracellular domain of vs. 1.1% 0.4% for KIN59-treated tumors, P < 0.01; Fig. FGFR1 immobilized to the sensor chip. Under the same 6D) and was only detected at a small rim at the border of experimental conditions, the compound does not interfere the tumors and not in the basement membrane of the with the binding of the growth factor to immobilized endothelial cells (Fig. 6C). Together, these data suggest heparin, closely related in structure to heparan sulfate. that inhibition of tumor vascularization and laminin Thus, KIN59 seems to differ from various negatively deposition contribute to the antitumor activity of KIN59. charged FGF2 inhibitors identified so far, including polysulfonated and polysulfated molecules, able to bind the basic domain of FGF2, thereby preventing its interaction Discussion with HSPGs (15). Indeed, in contrast to these compounds The nucleoside derivative KIN59 was first described as (14), KIN59 is uncharged and does not protect the growth an inhibitor of the angiogenic enzyme thymidine phos- factor from proteolytic digestion (data not shown).

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 825

Liekens et al.

Figure 5. Antitumor activity of KIN59. A, nude nu/nu mice were A B inoculated subcutaneously with 6 Control Control 2 10 F2T-luc2.9 cells. Animals KIN59 KIN59 were treated subcutaneously at a site distant from the tumor site with 15 mg/kg KIN59, twice daily from day 2 till the end of the experiment. ** At regular time intervals, the mice * were imaged to assess the growth * * of the luciferase-positive tumor cells. Data are the mean ( SEM) of 2 independent experiments (n ¼ 18). , P < 0.05. B, tumor size C was also monitored by means of a caliper, and the tumor volume was calculated with the following formula: tumor volume (mm3) ¼ 0.5 a b2, where a is the longest diameter and b is the shortest diameter. Data are expressed as mean SEM. , P < 0.05. C, box Control KIN59 and whisker plots showing the tumor weight (after excision) at the Control KIN59 end of the experiment in control and KIN59-treated mice.

Recently, we have shown that the pattern recognition ysis showed a 10-fold reduction in laminin mRNA, which receptor long pentraxin-3 (PTX3) and the PTX3-derived indicates that KIN59 inhibits laminin biosynthesis by the N-acetylated pentapeptide ARPCA (in single letter code) CAM cells. In contrast, the compound did not change the bind FGF2 with high affinity and specificity, thus pre- expression of other ECM components, such as type IV venting HSPG/FGF2/FGFR1 ternary complex formation collagen, type I collagen, and fibronectin (data not and angiogenesis (18, 47). This effect seems to be because shown). Still, laminin deficiency alone may alter the firm- of the ability of these molecules to bind the receptor- ness and architecture of the basement membrane, which binding region in the FGF2 protein, thereby hampering may impact signal transduction by connecting proteins, FGF2/FGFR1 interaction, without affecting heparin bind- such as integrins. Lack of the ability to provide survival ing (18). However, KIN59 does not prevent the binding of signals may then lead to instability of the newly formed PTX3 or ARPCA to FGF2, as assessed by SPR analysis, vessels, as observed in the presence of KIN59 (23, 40). suggesting that KIN59 inhibits FGF2/FGFR1 binding by FGF2 has been implicated in tumor growth/vascular- interacting with a new, as yet unidentified, site in the ization and in tumor evasion from anti-VEGF therapies growth factor or its receptor (data not shown). Further (reviewed in ref. 49). On this basis, to investigate the experiments are required to fully characterize the molec- impact of KIN59 on FGF2-dependent tumor growth and ular bases of the FGF2 antagonist activity of KIN59. vascularization, we used a firefly luciferase–transfected In agreement with the angiogenesis data, immunohis- clone (F2T-luc2.9 cells) derived from the highly tumori- tochemical analysis showed a significant reduction in the genic FGF2-overexpressing FGF2-T-MAE cells (34). These expression of laminin, a crucial component of the capillary cells generate opportunistic vascular lesions when basement membrane, in CAM sponges treated with PBS, injected subcutaneously in nude mice as a consequence thymidine phosphorylase, and FGF2, but not in sponges of an autocrine loop of stimulation triggered by the release treated with VEGF. The basement membrane, consisting of the overexpressed growth factor, thus representing mainly of laminin and type IV collagen, provides struc- a valid model to study FGF2-driven tumor growth and tural stability to the vessels and supports key signals that vascularization (44). As observed for other FGF2 antago- regulate endothelial migration, proliferation, tube forma- nists (44, 47, 50), KIN59 inhibited the growth of FGF2-T- tion, and survival (39, 48). Remodeling of the basement MAE cell–derived F2T-luc2.9 cells in vitro and significant- membrane and ECM, through cycles of proteolytic deg- ly affected their growth in vivo. Also, KIN59-treated radation (e.g., by MMPs) and deposition of ECM compo- tumors showed a significant decrease in vascularity of nents is essential for the angiogenic process. KIN59 did the lesions and displayed strongly reduced laminin not change the expression or activity of MMP-2, whereas expression in the basement membrane and ECM, suggest- MMP-9 could not be detected in the CAM extracts. Thus, ing that inhibition of angiogenesis and modulation of the although it cannot be ruled out that other chick embryo ECM may, at least in part, be responsible for the antitumor proteases (43) are affected by KIN59, increased gelatinase activity of KIN59. activity does not seem to be responsible for the absence of Despite its interesting spectrum of activities, at laminin in KIN59-treated CAMs. Instead, RT-PCR anal- least 2 aspects of KIN59 properties deserve further

826 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

A Control B * Control av v hv KIN59 av vhv

c v * 500 µm 500 µm Avascular Vascular Highly vascular KIN59

v hv av

500 µm 50 µm 50 µm 50 µm

C Control KIN59 D

50 µm 50 µm *

Control KIN59

50 µm 50 µm

Figure 6. Immunohistochemical analysis of KIN59-treated tumors. A, H&E staining shows the presence of 3 distinct areas in F2T-luc2.9–induced tumors: a highly vascularized (hv) outer rim, surrounding an area with lower blood vessel density (vascular area, v), and a largely avascular (av) inner core. Both control and KIN59-treated tumors show a comparable tissue organization. However, KIN59-treated tumors contain a significantly larger avascular core than the vehicle-treated tumors, as measured by CD31 staining of the blood vessels (B). Data are presented as mean SEM, n ¼ 3. , P < 0.05 compared with control. Representative photographs of the different tumor areas stained with anti-CD31 (brown) are shown. C and D, immunohistochemical staining for laminin in control and KIN59-treated tumors (C). In control tumors, laminin is detected in the ECM (top; predominantly at the tumor edge) and in the basement membrane of the endothelial cells (bottom, white arrowheads). In KIN59-treated tumors, laminin is only present in a small rim at the tumor border (top, arrows) and not surrounding the blood vessels (white arrowheads and bottom). D, the immunopositive area was quantified as described in Materials and Methods. KIN59 significantly decreased the expression of laminin. Values are expressed as mean SEM. n ¼ 3. , P < 0.01. discussion: (i) the narrow therapeutic window of active compounds. Fortunately, various approaches may the compound due, at least in part, to its low solu- be applied to increase the solubility of highly lipophilic bility and (ii) its incapacity to affect VEGF-induced compounds. In particular, we managed to increase the neovascularization. water solubility (and oral bioavailability) of poorly solu- Because of its low solubility, KIN59 could only be ble amine-containing parent drugs up to 1,000-fold by administered at 15 mg/kg, which is why the compound using a dipeptidyl peptidase IV (DPPIV/CD26)-based was injected twice daily (subcutaneously, at a site distant prodrug approach (51, 52). A similar approach is in from the cell inoculation site). Low aqueous solubility is a progress in our laboratory to generate more soluble KIN59 common problem in the development of biologically prodrugs.

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 827

Liekens et al.

The second point, the lack of anti-VEGF activity of and causes the degradation of newly formed blood vessels KIN59, may not necessarily represent a disadvantage. In (23). As such, KIN59 might be a promising lead compound fact, KIN59 may be administered together with an anti- for the design of novel multitargeted anticancer drugs. VEGF drug or used to treat tumors that produce FGF2 and/or thymidine phosphorylase as their predominant Disclosure of Potential Conflicts of Interest angiogenic proteins. Moreover, clinical studies with No potential conflicts of interest were disclosed. drugs that target VEGF or VEGF-induced signaling, such as the VEGF-antibody bevacizumab (Avastin; Genen- Acknowledgments tech/Roche) and the kinase inhibitors sorafenib and The authors thank E. Meyen, S. Noppen, R. Ronca, I. Van Aelst, and P. sunitinib showed that the benefits from these antiangio- Vervaeke for excellent technical assistance. They also thank Prof. L. Arckens for providing histologic assistance. genic therapies are at best temporary and usually followed by the development of tumor resistance (53). Grant Support Tumor resistance may be caused by circumvention of the This research was supported by grants from the "Fonds voor angiogenic blockade by activation and/or upregulation Wetenschappelijk Onderzoek (FWO) Vlaanderen" (grant no. G.0486.08) of alternative proangiogenic pathways, including thymi- and the Centers of Excellence of the Katholieke Universiteit Leuven (PF 10/018) to S. Liekens and J. Balzarini; and grants from MIUR, AIRC (grant dine phosphorylase and FGF2 (2, 9, 12). Thus, KIN59 may no. 10396), Fondazione Berlucchi, CARIPLO (grant 2008-2264 and NOBEL be used as a second-line approach after evasion from Project) to M. Presta. A. Gigante acknowledges the CSIC and the TSR (Fondo Social Europeo) for a JAE-PREDOC contract. anti-VEGF/VEGFR therapies has occurred. The costs of publication of this article were defrayed in part by the In conclusion, this study shows that KIN59 is a potent payment of page charges. This article must therefore be hereby marked inhibitor of angiogenesis with antagonist activity against advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. at least 2 angiogenic factors, namely, thymidine phosphorylase and FGF2. KIN59 also inhibits the deposition of Received September 15, 2011; revised January 5, 2012; accepted January the essential basement membrane component laminin 26, 2012; published OnlineFirst February 1, 2012.

References 1. Carmeliet P. Angiogenesis in life, disease and medicine. Nature and mitogenic activity by heparin-mimicking polysulfonated com- 2005;438:932–6. pounds. Mol Pharmacol 1999;56:204–13. 2. Folkman J. Angiogenesis: an organizing principle for drug discovery? 15. Presta M, Leali D, Stabile H, Ronca R, Camozzi M, Coco L, et al. Nat Rev Drug Discov 2007;6:273–86. Heparin derivatives as angiogenesis inhibitors. Curr Pharm Des 3. Chung AS, Lee J, Ferrara N. Targeting the tumour vasculature: insights 2003;9:553–66. from physiological angiogenesis. Nat Rev Cancer 2010;10:505–14. 16. Rusnati M, Presta M. Fibroblast growth factors/fibroblast growth 4. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. factor receptors as targets for the development of anti-angiogenesis Cell 2011;144:646–74. strategies. Curr Pharm Des 2007;13:2025–44. 5. Lohela M, Bry M, Tammela T, Alitalo K. VEGFs and receptors involved 17. Wang Y, Becker D. Antisense targeting of basic fibroblast growth in angiogenesis versus lymphangiogenesis. Curr Opin Cell Biol factor and fibroblast growth factor receptor-1 in human melanomas 2009;21:154–65. blocks intratumoral angiogenesis and tumor growth. Nat Med 6. Takahashi H, Shibuya M. The vascular endothelial growth factor 1997;3:887–93. (VEGF)/VEGF receptor system and its role under physiological and 18. Leali D, Bianchi R, BugattiA, Nicoli S, Mitola S, Ragona L,et al. Fibroblast pathological conditions. Clin Sci (Lond) 2005;109:227–41. growth factor 2-antagonist activity of a long-pentraxin 3-derived anti- 7. Ellis LM, Hicklin DJ. VEGF-targeted therapy: mechanisms of antiangiogenic pentapeptide. J Cell Mol Med 2010;14:2109–21. tumour activity. Nat Rev Cancer 2008;8:579–91. 19. Bronckaers A, Gago F, Balzarini J, Liekens S. The dual role of thymidine 8. Brower V. Antiangiogenesis research is booming, as questions and phosphorylase in cancer development and chemotherapy. Med Res studies proliferate. J Natl Cancer Inst 2009;101:780–1. Rev 2009;29:903–53. 9. Relf M, LeJeune S, Scott PA, Fox S, Smith K, Leek R, et al. Expression 20. Haraguchi M, Miyadera K, Uemura K, Sumizawa T, Furukawa T, of the angiogenic factors vascular endothelial cell growth factor, acidic Yamada K, et al. Angiogenic activity of enzymes. Nature 1994;368:198. and basic fibroblast growth factor, tumor growth factor beta-1, plate- 21. Miyadera K, Sumizawa T, Haraguchi M, Yoshida H, Konstanty W, let-derived endothelial cell growth factor, placenta growth factor, and Yamada Y, et al. Role of thymidine phosphorylase activity in the pleiotrophin in human primary breast cancer and its relation to angio- angiogenic effect of platelet derived endothelial cell growth factor/ genesis. Cancer Res 1997;57:963–9. thymidine phosphorylase. Cancer Res 1995;55:1687–90. 10. Dempke WC, Heinemann V. Resistance to EGF-R (erbB-1) and VEGF- 22. Hotchkiss KA, Ashton AW, Schwartz EL. Thymidine phosphorylase R modulating agents. Eur J Cancer 2009;45:1117–28. and 2-deoxyribose stimulate human endothelial cell migration by 11. Casanovas O, Hicklin DJ, Bergers G, Hanahan D. Drug resistance by specific activation of the integrins alpha 5 beta 1 and alpha V beta evasion of antiangiogenic targeting of VEGF signaling in late-stage 3. J Biol Chem 2003;278:19272–9. pancreatic islet tumors. Cancer Cell 2005;8:299–309. 23. Liekens S, Hernandez AI, Ribatti D, De Clercq E, Camarasa MJ, Perez- 12. Dorrell MI, Aguilar E, Scheppke L, Barnett FH, Friedlander M. Com- Perez MJ, et al. The nucleoside derivative 50-O-trityl-inosine (KIN59) bination angiostatic therapy completely inhibits ocular and tumor suppresses thymidine phosphorylase-triggered angiogenesis via a angiogenesis. Proc Natl Acad Sci U S A 2007;104:967–72. noncompetitive mechanism of action. J Biol Chem 2004;279: 13. Presta M, Dell'Era P, Mitola S, Moroni E, Ronca R, Rusnati M. Fibro- 29598–605. blast growth factor/fibroblast growth factor receptor system in angio- 24. Liekens S, Balzarini J, Hernandez AI, De Clercq E, Priego EM, Camar- genesis. Cytokine Growth Factor Rev 2005;16:159–78. asa MJ, et al. Thymidine phosphorylase is noncompetitively inhibited 14. Liekens S, Leali D, Neyts J, Esnouf R, Rusnati M, Dell'Era P, et al. by 50-O-trityl-inosine (KIN59) and related compounds. Nucleosides Modulation of fibroblast growth factor-2 receptor binding, signaling, Nucleotides Nucleic Acids 2006;25:975–80.

828 Mol Cancer Ther; 11(4) April 2012 Molecular Cancer Therapeutics

FGF2 Antagonist and Antitumor Activity of KIN59

25. Liekens S, Bilsen F, De Clercq E, Priego EM, Camarasa MJ, Perez- 40. Sorokin L. The impact of the extracellular matrix on inflammation. Nat Perez MJ, et al. Anti-angiogenic activity of a novel multi-substrate Rev Immunol 2010;10:712–23. analogue inhibitor of thymidine phosphorylase. FEBS Lett 2002;510: 41. Hallmann R, Horn N, Selg M, Wendler O, Pausch F, Sorokin LM. 83–8. Expression and function of laminins in the embryonic and mature 26. Ribatti D, Presta M. The role of fibroblast growth factor-2 in the vasculature. Physiol Rev 2005;85:979–1000. vascularization of the chick embryo chorioallantoic membrane. J Cell 42. Ribatti D, Nico B, Vacca A, Iurlaro M, Roncali L. Temporal expression of Mol Med 2002;6:439–46. the matrix metalloproteinase MMP-2 correlates with fibronectin immu- 27. Ribatti D, Cruz A, Nico B, Favier J, Vacca A, Corsi P, et al. In situ noreactivity during the development of the vascular system in the chick hybridization and immunogold localization of vascular endothelial embryo chorioallantoic membrane. J Anat 1999;195:39–44. growth factor receptor-2 on the pericytes of the chick chorioallantoic 43. Hahn-Dantona EA, Aimes RT, Quigley JP. The isolation, characteri- membrane. Cytokine 2002;17:262–5. zation, and molecular cloning of a 75-kDa gelatinase B-like enzyme, a 28. Ribatti D, Urbinati C, Nico B, Rusnati M, Roncali L, Presta M. Endog- member of the matrix metalloproteinase (MMP) family. An avian enous basic fibroblast growth factor is implicated in the vascularization enzyme that is MMP-9-like in its cell expression pattern but diverges of the chick embryo chorioallantoic membrane. Dev Biol 1995;170: from mammalian gelatinase B in sequence and biochemical proper- 39–49. ties. J Biol Chem 2000;275:40827–38. 29. Casanova E, Perez-Perez MJ, Kappe CO. Microwave-assisted 44. Liekens S, Neyts J, De Clercq E, Verbeken E, Ribatti D, Presta M. selective 50-O-trityl protection of inosine derivatives. Synlett 2007; Inhibition of fibroblast growth factor-2-induced vascular tumor forma- 1733–5. tion by the acyclic nucleoside phosphonate cidofovir. Cancer Res 30. Grinspan JB, Mueller SN, Levine EM. Bovine endothelial cells trans- 2001;61:5057–64. formed in vitro by benzo(a)pyrene. J Cell Physiol 1983;114:328–38. 45. El Omari K, Bronckaers A, Liekens S, Perez-Perez MJ, Balzarini J, 31. Chiodelli P, Mitola S, Ravelli C, Oreste P, Rusnati M, Presta M. Heparan Stammers DK. Structural basis for non-competitive product inhibition sulfate proteoglycans mediate the angiogenic activity of the vascular in human thymidine phosphorylase: implications for drug design. endothelial growth factor receptor-2 agonist gremlin. Arterioscler Biochem J 2006;399:199–204. Thromb Vasc Biol 2011;31:e116–27. 46. Olsson AK, Dimberg A, Kreuger J, Claesson-Welsh L. VEGF receptor 32. Bastaki M, Nelli EE, Dell'Era P, Rusnati M, Molinari-Tosatti MP, Parolini signalling - in control of vascular function. Nat Rev Mol Cell Biol S, et al. Basic fibroblast growth factor-induced angiogenic phenotype 2006;7:359–71. in mouse endothelium. A study of aortic and microvascular endothelial 47. Rusnati M, Camozzi M, Moroni E, Bottazzi B, Peri G, Indraccolo S, et al. cell lines. Arterioscler Thromb Vasc Biol 1997;17:454–64. Selective recognition of fibroblast growth factor-2 by the long pen- 33. Esko JD. Genetic analysis of proteoglycan structure, function and traxin PTX3 inhibits angiogenesis. Blood 2004;104:92–9. metabolism. Curr Opin Cell Biol 1991;3:805–16. 48. Eble JA, Niland S. The extracellular matrix of blood vessels. Curr Pharm 34. Gualandris A, Rusnati M, Belleri M, Nelli EE, Bastaki M, Molinari-Tosatti Des 2009;15:1385–400. MP, et al. Basic fibroblast growth factor overexpression in endothelial 49. Alessi P, Leali D, Camozzi M, Cantelmo A, Albini A, Presta M. Anti- cells: an autocrine mechanism for angiogenesis and angioproliferative FGF2 approaches as a strategy to compensate resistance to anti- diseases. Cell Growth Differ 1996;7:147–60. VEGF therapy: long-pentraxin 3 as a novel antiangiogenic FGF2- 35. Leali D, Belleri M, Urbinati C, Coltrini D, Oreste P, Zoppetti G, et al. antagonist. Eur Cytokine Netw 2009;20:225–34. Fibroblast growth factor-2 antagonist activity and angiostatic capacity 50. Sola F, Gualandris A, Belleri M, Giuliani R, Coltrini D, Bastaki M, et al. of sulfated Escherichia coli K5 polysaccharide derivatives. J Biol Chem Endothelial cells overexpressing basic fibroblast growth factor (FGF-2) 2001;276:37900–8. induce vascular tumors in immunodeficient mice. Angiogenesis 36. Bugatti A, Urbinati C, Ravelli C, De Clercq E, Liekens S, Rusnati M. 1997;1:102–16. Heparin-mimicking sulfonic acid polymers as multitarget inhibitors of 51. Garcia-Aparicio C, Bonache MC, De Meester I, San Felix A, Balzarini J, human immunodeficiency virus type 1 Tat and gp120 proteins. Anti- Camarasa MJ, et al. Design and discovery of a novel dipeptidyl- microb Agents Chemother 2007;51:2337–45. peptidase IV (CD26)-based prodrug approach. J Med Chem 37. Ribatti D, Nico B, Vacca A, Presta M. The gelatin sponge-chorioal- 2006;49:5339–51. lantoic membrane assay. Nat Protoc 2006;1:85–91. 52. Diez-Torrubia A, Balzarini J, Andrei G, Snoeck R, De Meester I, 38. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the com- Camarasa MJ, et al. Dipeptidyl peptidase IV dependent water-soluble parative C(T) method. Nat Protoc 2008;3:1101–8. prodrugs of highly lipophilic bicyclic nucleoside analogues. J Med 39. Davis GE, Senger DR. Endothelial extracellular matrix: biosynthesis, Chem 2011;54:1927–42. remodeling, and functions during vascular morphogenesis and neo- 53. Bergers G, Hanahan D. Modes of resistance to anti-angiogenic ther- vessel stabilization. Circ Res 2005;97:1093–107. apy. Nat Rev Cancer 2008;8:592–603.

www.aacrjournals.org Mol Cancer Ther; 11(4) April 2012 829

The Thymidine Phosphorylase Inhibitor 5′-O-Tritylinosine (KIN59) Is an Antiangiogenic Multitarget Fibroblast Growth Factor-2 Antagonist

Sandra Liekens, Annelies Bronckaers, Mirella Belleri, et al.

Mol Cancer Ther 2012;11:817-829. Published OnlineFirst February 1, 2012.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-11-0738

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2012/02/01/1535-7163.MCT-11-0738.DC1

Cited articles This article cites 52 articles, 16 of which you can access for free at: http://mct.aacrjournals.org/content/11/4/817.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/11/4/817.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/11/4/817. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.