Zoledronate Sensitizes Endothelial Cells to Tumor Necrosis Factor-induced Programmed Cell Death

Bisphosphonates are potent inhibitors of osteoclast function widely used to treat conditions of excessive bone resorption, including tumor bone metastases. Recent evidence indicates that bisphosphonates have direct cytotoxic activity on tumor cells and suppress angiogenesis, but the associated molecular events have not been fully characterized. In this study we investigated the effects of zoledronate, a nitrogen-containing bisphosphonate, and clodronate, a non-nitrogen-containing bisphosphonate, on human umbilical vein endothelial cell (HUVEC) adhesion, migration, and survival, three events essential for angiogenesis. Zoledronate inhibited HUVEC adhesion mediated by integrin αVβ3, but not α5β1, blocked migration and disrupted established focal adhesions and actin stress fibers without modifying cell surface integrin expression level or affinity. Zoledronate treatment slightly decreased HUVEC viability and strongly enhanced tumor necrosis factor (TNF)-induced cell death. HUVEC treated with zoledronate and TNF died without evidence of enhanced annexin-V binding, chromatin condensation, or nuclear fragmentation and caspase dependence. Zoledronate inhibited sustained phosphorylation of focal adhesion kinase (FAK) and in combination with TNF, with and without interferon (IFN) γ, of protein kinase B (PKB/Akt). Constitutive active PKB/Akt protected HUVEC from death induced by zoledronate and TNF/IFNγ. Phosphorylation of c-Src and activation of NF-κB were not affected by zoledronate. Clodronate had no effect on HUVEC adhesion, migration, and survival nor did it enhanced TNF cytotoxicity. Taken together these data demonstrate that zoledronate sensitizes endothelial cells to TNF-induced, caspase-independent programmed cell death and point to the FAK-PKB/Akt pathway as a novel zoledronate target. These results have potential implications to the clinical use of zoledronate as an anti-angiogenic or anti-cancer agent.

Bisphosphonates are pyrophosphate analogues in which the oxygen is replaced by a carbon atom with various side chains (1). They efficiently accumulate in bone due to their high affinity for calcium. Bisphosphonates are potent inhibitors of bone resorption used for the treatment of diseases characterized by excessive bone loss, such as Paget's disease, osteoporosis, and osteolytic tumor bone metastases (2)(3)(4). Bisphosphonates inhibit osteoclast activity at multiple levels; they prevent differentiation of macrophages into osteoclasts, block the activity of mature osteoclasts, and induce osteoclast apoptosis (5,6). The exact molecular mechanisms of action of bisphosphonates are only partially understood and appear to differ among different families of bisphosphonates (6,7). Non-nitrogen containing bisphosphonates (e.g. clodronate and etidronate) can be incorporated into non-hydrolyzable ATP analogs, which accumulate intracellularly and thereby suppress ATP-dependent enzymes (8). Nitrogen-containing bisphosphonates (e.g. zoledronate and pamidronate) inhibit critical enzymes of the mevalonate pathway, in particular farnesyl diphosphate synthase, required for the synthesis of farnesyl diphosphate and geranylgeranyl diphosphate, and thereby suppress prenylation of small GT-Pases essential for many cellular functions (9 -12).
Clinical and experimental evidence indicates that bisphosphonates suppress progression of bone metastases, and recent observations suggest that this effect may be independent of the inhibition of bone resorption (13,14). Tumor progression and metastasis formation are critically dependent on tumor angiogenesis (15). Anti-angiogenic treatments suppress tumor progression in animal models, and many anti-angiogenic substances are currently being tested in clinical trials for their therapeutic efficacy against human cancer (16). Recent evidence indicates that zoledronate possesses anti-angiogenic activities. Zoledronate was shown to inhibit serum-, fibroblast growth factor-2-, or vascular endothelial growth factor-stimulated proliferation and to modulate adhesion and migration of cultured endothelial cells (17). These effects were paralleled by a reduced vascular sprouting in the aortic ring and chicken chorioallantoic membrane (CAM) angiogenesis assays and by the suppression of fibroblast growth factor-2-induced angiogenesis in mice (17). Furthermore, zoledronate suppressed testosterone-stimulated vascular re-growth in the ventral prostate of castrated rats (18).
Osteoclast-mediated bone resorption and angiogenesis are critically dependent on integrin-mediated cell adhesion and signaling (19 -21). Integrins are ␣␤ heterodimeric cell surface complexes acting as the main receptors for extracellular matrix proteins with bi-directional signaling activity (21,22). Ligand binding function is tightly regulated by cell signaling events, and in turn, ligated integrins activate multiple signaling pathways essential for cell migration, proliferation, and survival. Integrin ␣ V ␤ 3 is highly expressed in osteoclasts and is strongly up-regulated in angiogenic endothelial cells (23,24). Pharmacological antagonists of ␣ V ␤ 3 efficiently inhibit bone resorption (25) and angiogenesis, including tumor angiogenesis, in many different animal models (21,24). Recent results demonstrate, however, that developmental and tumor angiogenesis can proceed in the absence of ␣ V ␤ 3 integrin, thus demonstrating that additional integrins contribute to angiogenesis (26,27).
We raised the hypothesis that bisphosphonates may exert their anti-angiogenic effects, at least in part, by interfering with vascular integrin function or signaling. To test this hypothesis, we exposed cultured human umbilical vein endothelial cells (HUVEC) 1 to zoledronate, a nitrogen-containing bisphosphonate, and clodronate, a non-nitrogen containing bisphosphonate, and tested the effects on integrin-dependent cell adhesion and migration and on cell survival.
Cell Culture, Electroporation, and Adenoviral Infection-HUVEC were prepared and cultured as previously described (28). For electroporation, sub-confluent HUVEC were collected and incubated on ice 5 min with 20 g of the specific plasmid DNA in serum-free M199 medium and electroporated with a Gene Pulser (Bio-Rad). HUVEC were resuspended in complete medium and cultured for 36 h before use in the experiments. Electroporation efficiency (between 70 and 80%) was routinely assessed by FACS® analysis of enhanced green fluorescent protein expression from a co-transfected pEGFP-C1 plasmid (Clontech, La Jolla, CA). HUVEC infected with Ad⌬NI-kB or AdLacZ (multiplicity of infection 100) were used 24 h post-infection.
Cell Adhesion Assays-Maxisorp II Nunc enzyme-linked immunosorbent assay plates (Roskilde, Denmark) were coated with 10 g/ml fibronectin, vitronectin, or gelatin overnight at 4°C and blocked with bovine serum albumin for 2 h at room temperature. Assays were done as previously described (28). Briefly, untreated cells or cells treated for 24 h with clodronate, zoledronate, or EDTA (100 M each) were collected and seeded in serum-free M199 medium at a concentration of 10 4 cells/well. After 1 h at 37°C wells were gently washed with phosphatebuffered saline, and attached cells were fixed in 4% paraformaldehyde (Fluka Chemie, Buchs, Switzerland) and stained with 0.5% crystal violet. Absorbance of each well was read at 620 nm in a plate reader (Packard Spectra Count, Meriden, CT). Results are expressed as mean value of triplicate determinations Ϯ S.D.
Migration Assay-Confluent HUVEC monolayers were cultured in complete medium alone or in the presence of clodronate, zoledronate, or EDTA (100 M). After 16 h, wells were washed with serum-free medium, and one wound per well was applied with a plastic tip. Cells were further cultured in serum free human endothelial basal growth medium (Invitrogen) in the absence or presence of clodronate, zoledronate, or EDTA (100 M). Ten hours later, cultures were fixed in 4% paraformaldehyde and stained with 0.5% crystal violet. Migration was assessed by taking pictures of three different regions of each wound and counting the number of cells migrated inside of the wound area. Results are expressed as the mean of the number of migrated cells/1000 m 2 wound Ϯ S.D.
Cell Viability Assays-HUVEC (10 4 cells/well) were plated on fibronectin-coated (10 g/ml) Maxisorp II Nunc enzyme-linked immunosorbent assay plates in complete medium alone or in the presence of zoledronate, clodronate, or EDTA at the indicated concentrations. Eight hours later TNF (200 ng/ml or as otherwise indicated) or medium alone was added to the wells. Sixteen hours later viability was determined by MTT assay (29) or by staining of adherent cells with crystal violet. Stained cells were lysed in Me 2 SO (for MTT reading) or in 0.1 M sodium citrate, 50% ethanol (for crystal violet reading). Absorbance of each well was measured at 620 nm (MTT) or 570 nm (crystal violet) in a plate reader (Packard Spectra Count). Results are expressed as mean value of triplicate determinations Ϯ S.D.
Immunofluorescence-HUVEC were cultured in complete medium or in medium supplemented with clodronate, zoledronate, or EDTA (100 M each) for 24 h and fixed in 4% parafomaldehyde for 10 min at room temperature. After permeabilization with 0.1% Triton X-100 (Sigma) cells were blocked in 1% bovine serum albumin, washed, and incubated for 1 h with an anti-paxillin mAb (5 g/ml). After washing, cells were incubated with a Cy3-conjugated goat anti-mouse (8 g/ml). To stain the actin fibers, permeabilized cells were incubated with phalloidin-FITC.
Flow Cytometry-HUVEC were collected by trypsinization, washed, and incubated with anti-integrin or anti-ICAM-1 antibodies (1 g/ml) for 1 h at 4°C. After washing, cells were incubated with a FITC-labeled antiserum for 30 min at 4°C. Samples were analyzed with a FACScan II® and Cell Quest® software (BD Biosciences). For annexin-V binding, HUVEC were harvested as above and sequentially incubated with FITC-conjugated annexin-V (1:40 dilutions) and propidium iodide (1 g/ml) following the manufacturer's recommendations (Apotech). For detection of DNA degradation, control and treated adherent and nonadherent HUVEC were collected as above, resuspended in 70% ice-cold ethanol under vortex, and incubated for 2 h at Ϫ20°C. Cells were recovered by centrifugation and resuspended in phosphate-buffered saline. 50 g/ml RNase A (Roche Applied Science) was added, and samples were incubated at room temperature for 5 min before staining with propidium iodide (PI, 50 g/ml) for 30 min at 37°C. Stained cells were analyzed with a FACScan II® and Cell Quest® software.
Caspase Activity Assay-10 6  Hoechst 33258 Staining-HUVEC were cultured for 16 h in complete medium alone or in medium supplemented with EDTA, clodronate, or zoledronate (100 M each) with and without ZVAD (50 M). Cells were fixed in 4% paraformaldehyde and 4% sucrose, permeabilized with 0.1% Triton X-100, phosphate-buffered saline, and then incubated with 2.5 g/l Hoechst 33258 in 0.1% Triton X-100, phosphate-buffered saline for 5 min. Stained cells were washed, and nuclei were viewed by epifluorescence microscopy (Axioskop, Carl Zeiss AG, Zü rich) equipped with a CCD camera (Photonic Science Milham, UK) at 360 nm.

Zoledronate Inhibits ␣ V ␤ 3 -mediated HUVEC Adhesion-
Bisphosphonates have been reported to decrease the adhesion of tumor cells to bone matrices (30,31) and to modulate endothelial cell adhesion to vitronectin (17). Whether bisphosphonates may differentially modulate adhesion mediated by different integrins was not reported. HUVEC adhesion to gelatin and vitronectin fully depends on integrin ␣ V ␤ 3 , whereas HUVEC adhesion to fibronectin is predominantly mediated by integrin ␣ 5 ␤ 1 , with only a minor contribution from ␣ V ␤ 3 (Fig.  1A). To test whether bisphosphonates may selectively interfere with ␣ V ␤ 3 -mediated adhesion, we cultured HUVEC for 24 h in medium alone or in medium supplemented with zoledronate, clodronate, or EDTA (100 M each) and assessed their ability to attach to fibronectin, vitronectin, and gelatin in a short term adhesion assay. Because bisphosphonates chelate cations and integrin function is modulated by divalent cations (e.g. Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ ), EDTA treatment was included in this and the successive experiments to exclude nonspecific effects due to cation chelation. Zoledronate treatment inhibited adhesion to vitronectin and gelatin by 50 and 80%, respectively, but did not affect adhesion to fibronectin (Fig. 1B). Pretreatment with clodronate or EDTA (100 M each) had no effect on HUVEC adhesion (Fig. 1B).
Zoledronate Inhibits HUVEC Migration-To test for the effects of zelodronate and clodronate on cell migration, HUVEC were cultured until they reached confluence and then supple-mented with EDTA, zelodronate, or clodronate for 16 h. A "scratch" wound was then made with a plastic tip, and 10 h later wells were fixed, and migration was quantified (32). Zoledronate treatment suppressed cell migration by more than 90%, whereas treatment with EDTA or clodronate had no effect (Fig. 2).
Zoledronate Does Not Affect Cell Surface Integrin Expression or Affinity-Decreased HUVEC adhesion and migration caused by zoledronate could be due to a reduced cell surface expression or a decreased affinity of integrins. To test for these possibilities, the total cell surface level of ␤ 1 and ␣ V ␤ 3 integrins and LIBS expression (33) on ␤ 1 and ␤ 3 subunits were determined. HUVEC cultured for 24 h in medium alone or in the presence of EDTA, clodronate, or zoledronate (100 M each) were first stained with anti-␣ V ␤ 3 or -␤ 1 integrin mAbs and analyzed by flow cytometry. Exposure to zoledronate had no effect on ␤ 1 or ␣ V ␤ 3 integrin expression levels ( Fig. 3A and data not shown). Thereafter, control or treated HUVEC were stained with mAb HUTS-21 and LIBS-1, which specifically bind to the high affinity conformation of ␤ 1 or ␤ 3 integrin subunits, respectively (34,35), to test whether zoledronate suppressed integrin affinity maturation. Zoledronate-treated HUVEC had ␤ 1 and ␤ 3 LIBS-staining profiles identical to control or clodronate-treated HUVEC ( Fig. 3B and data not shown). MnCl 2 , which is used to promote integrin affinity maturation (36), induced LIBS expression regardless of HUVEC pretreatment (Fig. 3B). MnCl 2 did not modify total ␤ 1 and ␣ V ␤ 3 integrin levels (Fig. 3A). Taken together, these experiments demonstrated that zoledronate, but not clodronate, inhibit ␣ V ␤ 3 -mediated HUVEC attachment to vitronectin and gelatin and block HUVEC migration without affecting ␣ V ␤ 3 and ␤ 1 integrin cell surface expression or affinity.
Zoledronate Disrupts Focal Adhesions and Actin Stress Fibers-Stable integrin-dependent cell adhesion and efficient cell migration are critically dependent on integrin post-receptor events, including focal adhesion assembly and actin cytoskeleton rearrangement (37). To test whether zoledronate interferes with these events, HUVEC were cultured for 24 h in medium alone or in the presence of EDTA, clodronate, or zoledronate and then stained for paxillin, to visualize focal adhesions, and for F-actin, to visualize the actin cytoskeleton. Untreated and EDTA-and clodronate-treated HUVEC showed the typical focal paxillin staining and the fibrillar actin cytoskeleton, consistent with focal adhesion assembly and actin stress fiber formation. In contrast, in zoledronate-treated HUVEC the focal paxillin staining was completely lost, and the F-actin cytoskeleton was nearly completely dissolved (Fig. 4A).
Zoledronate Treatment Inhibits Phosphorylation of Focal Adhesion Kinase-FAK is a critical regulator of focal adhesion formation, maintenance, and turnover (38). FAK is activated by auto-phosphorylation in response to integrin ligation and by c-Src-mediated phosphorylation (39). To analyze the effect of zoledronate on FAK phosphorylation, HUVEC were cultured for 24 h in the absence or presence of EDTA, clodronate, or zoledronate (100 M each), and FAK phosphorylation was analyzed by Western blotting using a phospho-FAK-specific antibody. Zoledronate, but not clodronate or EDTA treatment, efficiently suppressed FAK phosphorylation without altering total FAK protein levels (Fig. 4B). Phosphorylation of c-Src was not diminished by zoledronate, suggesting that decreased FAK phosphorylation was not due to suppressed c-Src activation (Fig. 4B). These experiments demonstrated that zoledronate disrupted focal adhesions and actin stress fibers in stably adherent cells and inhibited FAK phosphorylation, three events downstream of integrin ligation.
Zoledronate Reduces HUVEC Viability and Enhances TNF Cytotoxic Activity-FAK-dependent signaling events promote cell survival and protect cells against death induced by stress, cytotoxic drugs, and TNF (40 -42). To test whether zoledronate treatment had an effect on cell viability, HUVEC were cultured in medium alone or in the presence of bisphosphonates, and cell viability was determined after 24 h. Zoledronate treatment caused a dose-dependent decrease in HUVEC viability, with a 30% cell loss at 100 M, whereas EDTA and clodronate had no effect (Fig. 5A). Zoledronate-treated cells had a retracted cytoplasm, an elongated shape with long protrusions at the ends (Fig. 5B). Because HUVEC are normally resistant to TNFinduced death (29,43), we decided to test whether zoledronate may sensitize HUVEC to TNF cytotoxicity. HUVEC were cultured for 8 h in the absence or presence of bisphosphonates and then exposed to TNF at doses ranging from 0.02 ng/ml up to 200 ng/ml, and viability was determined 24 h later. Treatment with TNF alone had a minimal effect on viability. Zoledronate pretreatment, however, induced a dose-dependent decrease in viability of cells exposed to TNF (Fig. 5C). Zoledronate/TNFtreated cells showed a severely altered morphology, with many cells rounding up and retracting from the substrate (Fig. 5D). Pretreatment with EDTA or clodronate did not enhance TNF cytotoxicity. Strikingly, a 6 -8-h pretreatment with zoledronate was necessary to sensitize HUVEC to TNF cytotoxicity. Simultaneous exposure to zoledronate and TNF did not result in decreased HUVEC viability (data not shown).
To further document this synergistic effect, we determined the appearance of cells with fragmented (sub-G 1 ) DNA content in cultures exposed to increasing doses of TNF alone or in combination with zoledronate (100 M). Untreated cultures had few cells with sub-G 1 DNA (Ϸ3-4%), and treatment with TNF alone had little effect (Ϸ7-8% at 200 ng/ml) (Fig. 5E). Zoledronate treatment alone caused a significant increase in the fraction of cells with sub-G 1 DNA content (Ϸ22%), whereas combined treatment with zoledronate and TNF resulted in more than 40% of the cells with fragmented DNA already at TNF doses as low as 2 ng/ml. Increase of the TNF dose up to 200 ng/ml or the addition of IFN-␥, an enhancer of TNF toxicity (44), did not further increase the fraction of cells with sub-G 1 DNA (Fig. 5E). Treatment with clodronate alone or in combination with TNF with and without IFN␥, however, did not cause detectable internucleosomal DNA fragmentation, as analyzed by DNA agarose gel electrophoresis (data not shown). As a third way to document this synergistic cytotoxic effect, control and treated cultures were fixed and stained with crystal violet, and the proportion of attached cells was revealed by optical density reading. Zoledronate treatment alone caused an ϳ25-30% cell loss at 100 M, whereas combined zoledronate and TNF treatment resulted in up to 70% cell loss (Fig. 5, F and  G). Untreated and EDTA-or clodronate-pretreated HUVEC remained well attached even in the presence of TNF (Fig. 5, F  and G). From these experiments we concluded that zoledronate induced a dose-dependent death of HUVEC and effectively sensitized HUVEC to TNF cytotoxic activity.
Zoledronate Does Not Inhibit TNF-induced NF-B Activation-Activation of the transcription factor NF-B promotes survival of cells exposed to TNF (45,46). Zoledronate was recently shown to inhibit the osteoclastogenic activity of RANKL, an inducer of NF-B activation, through the enhanced production of osteoprotogerin, an endogenous RANKL antagonist (47,48). To test whether zoledronate suppressed NF-B activation, we analyzed the phosphorylation state and degradation of I-B, the inhibitory subunit of the cytoplasmic NF-B complex (49) grading I-B in response to TNF stimulation (Fig. 6A). To demonstrate that NF-B was transcriptionally active in zoledronate-treated cells, we measured cell surface expression of ICAM-1, an event dependent on NF-B transcriptional activity (50). HUVEC were cultured for 16 h in the presence of EDTA, clodronate, or zoledronate and then exposed to TNF for 6 h. Cell surface expression of ICAM-1 was determined by immunostaining and flow cytometry analysis. Zoledronate treatment did not inhibit TNF-induced expression of ICAM-1 (Fig. 6B). As a control, inhibition of NF-B activation by HUVEC infection with a recombinant adenovirus expressing a dominant negative I-B mutant (DN-I-B) fully prevented ICAM-1 up-regulation in response to TNF (Fig. 6C). These results demonstrate that decreased viability of HUVEC exposed to zoledronate and TNF was not due to deficient NF-B activation.
Zoledronate Inhibits Sustained Activation of Protein Kinase B-The serine/threonine PKB/Akt promotes cell survival and protects cells against death induced by various cytotoxic stimuli, including TNF (51)(52)(53). Both TNF and cell adhesion activate PKB/Akt through FAK and phosphatidylinositol 3-kinase (53,54). To tested whether zoledronate treatment suppressed activation of PKB/Akt, HUVEC were cultured for 16 h in the presence of EDTA, clodronate, and zoledronate (100 M each), serum-starved for 2 h, and then exposed to TNF (200 ng/ml), and the phosphorylation state of PKB/Akt was determined up to 60 min. Zoledronate pretreatment did not prevent TNF- show that combined treatment with zoledronate and TNF causes severe cell rounding. E, flow cytometry analysis of DNA content of HUVEC cultured for 8 h in the absence (Control) or presence of zoledronate (100 M) and exposed to TNF with and without IFN␥ for additional 16 h as indicated. DNA content was determined by propidium iodide staining of permeabilized cells (n ϭ 3). F, HUVEC were exposed to EDTA, clodronate, or zoledronate (100 M) for 8 h and further cultured alone or in combination with TNF (0.02-200 ng/ml) for 16 additional hours. Wells were rinsed, fixed, and stained by crystal violet. Results represent the mean of duplicate values Ϯ S.D. of the optical density reading (n ϭ 3). G, micrographs of cultures of the same experiment (EDTA, clodronate, or zoledronate, 100 M each; TNF, 200 ng/ml) .  Bars in panels A, D, and G, 100 m. induced short term PKB/Akt phosphorylation (Fig. 7A). Similarly, induction of PKB/Akt activation upon HUVEC replating on fibronectin was not affected by zoledronate alone or in combination with TNF (data not shown). Next we monitored the long term effect of zoledronate and TNF treatment on PKB/Akt phosphorylation. In this experiment control and bisphosphonate-treated HUVEC had a strong basal level of PKB/Akt phosphorylation due to the presence of serum (required by the long term kinetics of the experiment). The addition of TNF/IFN␥ to zoledronate-pretreated HUVEC caused a decreased in PKB/Akt phosphorylation at 6 h (Fig. 7C). Zoledronate-pretreated HUVEC had undetectable levels of phospho-PKB/Akt up to 16 h after the addition of TNF. Zoledronate caused a mild decrease, whereas TNF caused a slight increase in phospho-PKB/Akt (Fig. 7C). Wortmannin, a pharmacological inhibitor of phosphatidylinositol 3-kinase, suppressed basal and TNF-induced PKB/Akt phosphorylation, thus demonstrating the phosphatidylinositol 3-kinase dependence of TNF-mediated PKB/Akt phosphorylation (Fig. 7C). FAK phosphorylation was efficiently suppressed in HUVEC treated with zoledronate alone or in combination with TNF, whereas clodronate had no effect.
Constitutive Active PKB/Akt Protects HUVEC against Death Induced by Zoledronate with and without TNF and INF␥-To assess whether suppression of sustained activation of PKB/Akt was responsible for the decreased viability of HUVEC exposed to zoledronate with and without TNF, we transiently expressed constitutive active or wild type PKB/Akt in HUVEC. Transfectants were pretreated with clodronate or zoledronate for 8 h and exposed to TNF/IFN␥ for an additional 24 h. Cell death was determined by measuring the fraction of cells with sub-G 1 DNA. Expression of constitutive active PKB/Akt suppressed cell death induced by zoledronate with and without TNF/IFN␥ by more than 65% (by a transfection efficiency of 70%) (Fig.  7E). Expression of wild-type PKB/Akt had no protective effect. Taken together these results demonstrate that HUVEC death after combined zoledronate and TNF with and without IFN␥ treatment was associated with suppressed sustained phosphorylation of FAK and PKB/Akt and that constitutive active PKB/Akt protected HUVEC against death induced by zoledronate with and without TNF/IFN␥.
Combined Zoledronate and TNF Treatment Induces Caspase-independent Programmed Cell Death-TNF and bisphosphonates, including zoledronate, can induce caspasedependent as well as caspase-independent programmed cell death (PCD) (55)(56)(57)(58)(59)(60)(61)(62). To test whether HUVEC death induced by the combined exposure to zoledronate and TNF was due to apoptotic or non-apoptotic PCD, we analyzed cellular and molecular changes associated with apoptosis; that is, exposure of phosphatidylserines, chromatin condensation, nuclear fragmentation, and caspase activation. First, control and bisphosphonates-pretreated HUVEC were exposed to TNF for 20 h, harvested, and stained with FITC-annexin-V (to detect phosphatidylserine exposure in early apoptotic cells) and propidium iodide (to detect increased membrane permeability in late apoptotic/necrotic cells) (56,63). Zoledronate treatment did not increase the frequency of PI ϩ cells, whereas it induced a small increase in annexin-V ϩ /PI Ϫ cells. TNF treatment alone did not significantly increased the number of annexin-V ϩ /PI Ϫ or PI ϩ cells (Fig. 8A). Combined zoledronate and TNF treatment resulted in a nearly 3-fold rise in PI ϩ cells (5.9 -15.6%), with only a minor rise in annexin-V ϩ /PI Ϫ cells (Fig. 8A). Second, we looked for evidence of nuclear fragmentation or chromatin condensation by Hoechst 33258 staining. No increase in condensed or fragmented nuclei was observed at 6 or 24 h after the addition of TNF to pretreated HUVEC (Fig. 8B and not shown). Consistent with these morphological results, no internucleosomal DNA fragmentation ("DNA laddering") was observed by DNA-agarose gel electrophoresis (data not shown). Third, we tested for caspase activity using a fluorescent caspase-3 substrate (Ac-DEVD-amidomethylcoumarin) conversion assay (64). HUVEC were pretreated with bisphosphonates for 8 h and exposed to TNF for 16 h. Combined treatment with zoledronate and TNF resulted in a 2-3-fold increase in signal intensity, whereas treatment with zoledronate or TNF alone or combined exposure to clodronate and TNF had no effect (Fig. 8C). The pan-caspase inhibitor ZVAD completely suppressed the appearance of this caspase-3 like activity. Because this enzymatic assay is also sensitive to caspase 1, 4, 7, and 8 activity (manufacturer's data sheet), we directly looked for evidence of cleavage of caspase 3 and of its substrate PARP by Western blotting. No caspase-3 and PARP cleavage were observed in HUVEC treated with TNF and zoledronate (Fig. 8D). We then determined whether the caspase activity detected in the fluorogenic assay was nevertheless required for PCD induced by zoledronate and TNF. The addition of ZVAD at doses that fully suppressed caspase enzymatic activity did not protect against cell death induced by zoledronate and TNF treatment (Fig. 8E). Taken together, these results demonstrate that HUVEC treated by zoledronate and TNF died by caspase-independent PCD.

DISCUSSION
There is increasing clinical and experimental evidence indicating that bisphosphonates may exert anti-cancer effects through the direct inhibition of tumor cell proliferation, survival, migration and invasion, and the suppression of angiogenesis. Here we have investigated the effects of zoledronate and clodronate on endothelial cell adhesion, migration, and survival and report the following. 1) Zoledronate inhibited ␣ V ␤ 3 -, but not ␣ 5 ␤ 1 -, mediated HUVEC adhesion and suppressed HUVEC migration. These effects were not due to changes in integrin cell surface level or affinity. 2) Zoledronate disrupted established focal adhesions and actin stress fibers in stably adherent HUVEC, and this effect was associated with suppression of FAK phosphorylation. 3) Zoledronate treatment potently sensitized HUVEC to TNF-induced death. This effect was associated with the suppression of sustained phosphorylation of FAK and PKB/Akt, whereas activation of NF-B was not inhibited. 4) HUVEC treated with zoledronate and TNF died in the absence of caspase activation, chromatin condensation and nuclear fragmentation. 5) Clodronate had no effect on HUVEC adhesion, migration, and survival nor did it sensitize HUVEC to TNF-induced death. Taken together these data demonstrate that zoledronate interferes with endothelial cell adhesion and migration by perturbing integrin-dependent to clodronate or zoledronate and for 16 additional hours to TNF/IFN␥ (200/300 ng/ml) as indicated. Cells were analyzed for DNA content by PI staining and flow cytometry. The numbers above the histograms give the percent of cells with sub-G 1 DNA content (n ϭ 3). were serum-starved for 2 h and exposed to TNF (200 ng/ml). Phospho-PKB (p-PKB) and total PKB/Akt levels were determined by Western blotting before (t ϭ 0) and 7.5, 15, 30, 60 min after the addition of TNF (n ϭ 2). B and C, effect on sustained PKB/Akt activation. Untreated (Ctrl) and zoledronate (Zol)-and clodronate (Clo)treated HUVEC cultures (100 M for 24 h) were exposed to TNF/IFN␥ (200/300 ng/ml) (B) or to TNF alone (200 ng/ml) (C) as indicated, and PKB/Akt phosphorylation was determined before (t ϭ 0) and 2, 4, 6 (B), or 16 h (C) later. Wortmannin (Wort, 100 nM) was added during TNF treatment (n ϭ 3). D, effect on FAK activation. HUVEC treated as in panel C were tested for phospho-FAK and total FAK levels (n ϭ 3). E, constitutive active PKB/Akt protect HUVEC. Plasmid encoding for the wild type (wtPKB) or a constitutive active (mpPKB) form of PKB/Akt were electroporated into HUVEC. After 36 h, cells were exposed for 8 h Several bisphosphonates such as ibandronate, alendronate, pamidronate, or clodronate were shown to inhibit attachment of tumor cells to extracellular matrix proteins or to bone sections (30,31,65). Zoledronate was reported to enhance endothelial cell adhesion and migration on vitronectin at low doses (1 and 3 M) and to inhibit adhesion and migration at high doses (100 M) (17). Our results extend these findings by demonstrating that the zoledronate can differentially affect adhesion mediated by different integrins (i.e. ␣ V ␤ 3 versus ␣ 5 ␤ 1 ). Furthermore, zoledronate disrupted established focal adhesions and actin stress fibers and suppressed phosphorylation of FAK in stably adherent cells. Alendronate was recently reported to inhibit adhesion, migration, focal adhesion, and actin stress fibers assembly in ovarian cancer cell lines by attenuating the activation of Rho and phosphorylation of FAK, paxillin, and myosin light chain when cells were stimulated with lysophosphatidic acid, a Rho agonist (66). The fact that c-Srcphosphorylation was not affected in our system suggests the zoledronate effect on FAK is independent of c-Src. The phosphotyrosine phosphatases SHP-2 and PTP1D were reported to dephosphorylate FAK (67)(68)(69)(70). Rapid and transient dephosphorylation of FAK is necessary for focal adhesion remodeling and cell migration (38) and can be induced by a number of physiological stimuli such as activation of the EphA2 kinase (67), signaling from the IGF-I and insulin receptors (71,72), and stimulation with estradiol (73). Taken together the lack of down-regulation of integrin expression or affinity, the decreased FAK phosphorylation, the disassembly of focal adhesions, and the disruption of the actin cytoskeleton are consistent with zoledronate targeting integrin-postreceptor events. How zoledronate selectively suppress ␣ V ␤ 3 function and whether the reduced FAK phosphorylation caused by zoledronate is due to increased dephosphorylation or reduced phosphorylation remain to be elucidated.
The main novel finding reported here is that zoledronate sensitized endothelial cells to TNF-induced caspase-independent PCD and that this effect involved suppression of sustained PKB/Akt activation. TNF is often considered as a paradigm pro-apoptotic factor, but most cells, including endothelial cells, are resistant to TNF-induced death (29,74). TNF can induce two distinct forms of PCD: (i) "classical" apoptosis, characterized by caspase-dependent chromatin condensation and fragmentation, membrane blebbing, and generation of apoptotic bodies, and (ii) apoptosis-like or necrosis-like caspase-independent PCD, characterized by absent or marginal chromatin condensation, lack of nuclear fragmentation, and disruption of membrane integrity (56). We concluded that HUVEC died by caspase-independent (non-apoptotic) cell death based on three groups of observations. First, HUVEC exposed to zoledronate and TNF died in the absence of detectable nuclear fragmentation, chromatin condensation and internucleosomal DNA degradation whereas they had a 3-fold rise in PI uptake, a sign of increased membrane permeability typical of necrotic cell death. The appearance of a cell population with sub-G 1 DNA content or a positive TUNEL assay (17), alone, are not conclusive for apoptosis, since DNA cleavage into large (i.e. 50 kb) fragments occurs during non-apoptotic/caspase-independent death (56). Also, annexin-V binding, which is often considered as a sign of apoptosis, may occur during necrosis-like PCD when the cell membrane loses its integrity and intracellular phosphatidylserine residues becomes accessible to labeled annexin-V (56). Second, caspase activation was not necessary for PCD of treated HUVEC. Consistently, there was no detectable cleavage of caspase-3, the main mediator of caspase-dependent TNF induced apoptosis, nor of its substrate PARP (75). Some enzymatic caspase activity, however, was detected in zoledronate/ TNF-treated HUVEC (Fig. 8C), but this activity was not required for PCD, since the pan-caspase inhibitor ZVAD fully suppressed it but did not prevent PCD (Fig. 8E). Third, zoledronate-pretreated and TNF-stimulated HUVEC had sustained activation of NF-B. Activation of NF-B suppresses TNF-induced apoptosis by promoting the expression of antiapoptotic proteins (75), such as NDED or c-Flip, which inhibit activation of caspase-8 (76,77). Accordingly, pharmacological or genetic inhibition of NF-B activation results in TNF-mediated apoptosis (45,46).
The FAK-PKB/Akt pathway has emerged as an important pathway promoting cell survival and conferring protection against death induced by cytotoxic drugs and death factors, including TNF (78,79). For example, staurosporine caused a rapid FAK dephosphorylation and dissociation from focal adhesions resulting in endothelial cell apoptosis (80). Renal epithelial cell apoptosis caused by the nephrotoxic agent dichlorovinylcysteine was associated with FAK dephosphorylation, loss of focal contacts and FAK cleavage (81). Insulin-mediated activation of PKB/Akt protected endothelial cells against death induced by TNF (82). PKB␣/Akt1-null mouse embryo fibroblasts were more susceptible to cell death induced by TNF and anti-Fas antibodies (83). Pharmacological inhibition of phosphatidylinositol-3 kinase, a downstream target of FAK and an essential activator of PKB/Akt, sensitized endothelial cells to TNF cytotoxicity (52). In our experiments, zoledronate alone efficiently suppressed sustained FAK phosphorylation ( Fig. 4  and 7), and partially suppressed PKB/Akt phosphorylation (Fig. 7C). In combination with TNF with and without IFN␥, however, it potently suppressed PKB/Akt phosphorylation (Fig.  7, B and C). Constitutive active PKB/Akt protected HUVEC against death induced by zoledronate alone or in combination with TNF with and without IFN␥ (Fig. 7E). So far the cytotoxic and pro-apoptotic activity of nitrogen-containing bisphosphonates observed on cancer cells was associated with their ability to inhibit enzymes in the mevalonate pathway, and to suppress protein prenylation (4,14). For example, the apoptosis of human myeloma cells induced by incadronate was prevented by exogenous geranylgeraniol and farnesol (60). Zoledronate induced caspase-3-dependent apoptosis of breast and prostate cancer cells in association with impaired Ras membrane localization, a prenylation-dependent event (84,85). The exact mechanism by which loss of protein prenylation caused caspasedependent apoptosis has not been fully elucidated. particular interest. The first one concerns the mechanisms by which zoledronate suppresses phosphorylation of FAK and PKB/Akt (alone or in combination with TNF) and whether it is related to the inhibition of protein prenylation. A second one relates to the mechanisms by which zoledronate and TNFtreated endothelial cells die. One likely possibility is that suppression of PKB/Akt signaling results in the destabilization of the mitochondrial membrane, resulting in the production of reactive oxygen species or release of apoptosis-inducing factor (56). Alternatively, apoptoses signal-regulating kinase-1/c-Jun NH 2 -terminal kinase-mediated death should also be considered (86). A third important question is whether zoledronate can sensitize the tumor vasculature to the cytotoxic activity of TNF. Administration of high doses of TNF and chemotherapy through the isolated limb perfusion technique to patients with advanced cancers of the limbs causes selective disruption of the tumor vasculature, rapid induction of tumor necrosis, and high rates of complete anti-tumor responses (87). This effect involves the selective suppression of integrin ␣ V ␤ 3 function on angiogenic endothelial cells and induction of endothelial cell death (43). Quiescent endothelial cells and normal vessels are not killed by this treatment. So far the systemic use of TNF as an anti-cancer agent is not possible because of the appearance of severe side effects before therapeutic (anti-tumor vasculature) TNF levels are reached (87). If indeed zoledronate selectively sensitizes the tumor vessels to the cytotoxic activity of TNF as it does to HUVEC, this may allow lowering of the doses of TNF needed to induce death of angiogenic endothelial cells and may, thus, open the way to systemic TNF therapy. We are currently performing in vivo experiments to test this hypothesis. It should be noted that zoledronate was recently shown to synergistically augment the cytotoxic activities of paclitaxel on human breast cancer cells and of dexamethasone on myeloma cells in vitro (88,89) and of STI571 (imanitib) on chronic myelogenous leukemia cells in vivo (90).
In conclusion, we have demonstrated that zoledronate sensitizes endothelial cells to TNF-induced, caspase-independent programmed cell death and have provided original evidence that the FAK-PKB/Akt pathway is a zoledronate target. Future experiments will tell us about the relevance of these observations to human cancer treatment.