Canstatin Acts on Endothelial and Tumor Cells Via Mitochondrial Damage Initiated Through Interaction with Avb3 and Avb5 Integrin

Research Article

Canstatin Acts on Endothelial and Tumor Cells via Mitochondrial

Damage Initiated through Interaction with AvB3 and AvB5 Integrins Claire Magnon,1 Ariane Galaup,1 Brian Mullan,1 Vale´rie Rouffiac,2 Jean-Michel Bidart,3 Frank Griscelli,1 Paule Opolon,1 and Michel Perricaudet 1

1UMR 8121 Laboratoire de vectorologie et transfert de ge`nes, 2Laboratoire d’imagerie du petit animal, and 3De´partement de biologie clinique, Institut Gustave Roussy, Villejuif cedex, France

Abstract NC1 globular domain fragments of a-chains of human collagen Canstatin, the noncollagenous domain of collagen type IV represent a new class of antiangiogenic molecules that differ A-chains, belongs to a series of collagen-derived angiogenic fundamentally in structure to the plasminogen-derived group of inhibitors. We have elucidated the functional receptors and antiangiogenic proteins such as angiostatin kringles 1 to 3 (K1-3; intracellular signaling induced by canstatin that explain its refs. 8–11). Thus, we decided to investigate the relative pharma- strong antitumor efficacy in vivo. For this purpose, we cologic properties of one protein from each therapeutic class: generated a canstatin-human serum albumin (CanHSA) fusion canstatin and K1-3. protein, employing the HSA moiety as an expression tag. We In this study, to improve the pharmacokinetic properties of show that CanHSA triggers a crucial mitochondrial apoptotic canstatin, we investigated in situ the anticancer properties of a mechanism through procaspase-9 cleavage in both endothe- canstatin-human serum albumin (CanHSA) fusion protein by adenoviral-mediated gene transfer, which efficiently yields high lial and tumor cells, which is mediated through cross-talk levels of transgene expression in many different cell types (12). To between AvB3- and AvB5-integrin receptors. As a point of reference, we employed the first three kringle domains of this end, we constructed a recombinant adenovirus (AdCanHSA) angiostatin (K1-3), fused with HSA, which, in contrast to encoding canstatin fused with HSA to decrease in vivo clearance of fused protein by increasing its molecular weight (13). The HSA CanHSA, act only on endothelial cells through AvB3-integrin receptor–mediated activation of caspase-8 alone, without moiety in the fusion protein carries the added advantage of acting ensuing mitochondrial damage. Taken together, these results as a tag which allows the expression of canstatin to be analyzed provide insights into how canstatin might exert its strong in vitro and in vivo. To evaluate the antitumor potential of CanHSA, we compared its features both in vitro and in vivo with anticancer effect. (Cancer Res 2005; 65(10): 4353-61) that of adenoviral-expressed unfused canstatin (AdCan) and K1-3, either fused (AdK1-3HSA) or unfused (AdK1-3) with HSA (11, 14). Introduction Importantly, to characterize the antitumor properties of Type IV collagen extracellular matrix (ECM) plays a functional CanHSA compared with those of K1-3HSA, we focused our study role in angiogenesis, the process whereby new capillaries sprout on the mechanism of action of each protein. Given that angiostatin from existing vessel. Proteolytic remodeling of the ECM not only and canstatin both induce apoptosis in endothelial cells (7, 11, degrades barriers that obstruct vascular cell migration but also 15, 16), we asked (a) whether tumor cells as well as endothelial exposes cryptic sites within collagen IV that promote novel cells could be a direct target of canstatin via an integrin-dependent integrin-ligand interactions required for angiogenesis (1). The type interaction, and (b) whether the signaling pathways leading to IV collagen supramolecular structure is composed of a-chains, induction of apoptosis are different following canstatin or K1-3 each containing a noncollagenous (NC1) domain at the COOH treatment. To this end, we explored cell cycle progression and the terminus which provides the mechanical stability of the ECM activities of caspases within tumor and endothelial cells. structure (2, 3). ECM is a potential therapeutic target to regulate neovasculari- Materials and Methods zation in cancerous diseases as angiogenesis is required for neoplastic tumor growth (4). Systemic injection of the recombinant Construction of the recombinant adenoviruses. AdCan and a a2NC1 domain monomer of type IV collagen, which has been AdCanHSA express the NC1 domain of the human 2 chain of type IV collagen, also called canstatin (GenBank accession GI 81011723), either named canstatin, inhibits the in vivo growth of xenografted tumors fused or unfused to HSA moiety. Both canstatin and CanHSA are secreted following daily injection over a period of 15 days as compared with through a 24-residue prepro signal sequence from HSA (GenBank accession placebo-treated mice. In vitro, the antiproliferative and antimi- E01107), which was inserted into the NH2 terminus of both proteins. AdK1- gratory effects of canstatin seem to be due to the induction of 3 and AdK1-3HSA express respectively the angiostatin kringles 1 to 3 (K1-3) apoptotic events through disruption of the FAK signaling pathway and K1-3 conjugated with HSA. A recombinant adenovirus not containing a and induction of Fas ligand expression (5–7). transgene (AdCO1) was used as a negative control in this study (11, 14). Cell lines. MDA-MB-231 and human umbilical vascular endothelial cells (HUVEC) have been previously described (17). PC-3 cells (ATCC ref. CRL 1435) were grown in F-12K supplemented with 10% fetal bovine serum (FBS), 1 mmol/L L-glutamine, and 1.5 g/L sodium bicarbonate (Invitrogen, Requests for reprints: Claire Magnon, UMR 8121, PR2, Institut Gustave Roussy, 39 Carlsbad, CA). rue Camille Desmoulins, 94805 Villejuif cedex, France. Phone: 33-1-42115075; Fax: 33-1- 42115245; E-mail: [email protected]. Anticanstatin antibodies. To generate a polyclonal antibody against I 2005 American Association for Cancer Research. canstatin, a peptide spanning the NH2-terminal region of canstatin www.aacrjournals.org 4353 Cancer Res 2005; 65: (10). May 15, 2005

(VSIGYLLVKHSQTDQEPMC) was synthesized by conventional solid-phase Roche Applied Science, Basel, Switzerland) was used to detect apoptotic methodology using an Applied Biosystems Model 431A peptide synthesizer. cells within paraffin sections. Five fields per animal were digitized (200 Â). The purity of the peptide was ascertained by high-performance liquid The number of stained cells per field was counted and mean values with chromatography and its identity verified by mass spectrometry. The their SD were calculated. synthetic peptide was conjugated to keyhole limpet hemocyanin, and two Immunofluorescence. HUVEC and MDA-MB-231 cells were infected rabbits were immunized by two intradermal injections of the synthetic with adenoviruses 24 hours following seeding of cells. Forty-eight hours peptide-carrier conjugate. Antisera were verified by ELISA for their capacity later, cells were washed, fixed in PBS/3% formaldehyde, washed twice with to react with the canstatin synthetic peptide. Antibodies were purified by PBS, and incubated overnight at 4jC in PBS/0.5% FBS. CanHSA and K1- affinity chromatography. 3HSA proteins were detected using the mouse anti-HSA mAb described Western blot analysis of canstatin, CanHSA, K1-3, and K1-3HSA above. After washing thrice with PBS/0.5% FBS, cells were incubated with a expressions in vitro. For canstatin and CanHSA expression analysis, cell fluorescein-conjugated secondary antibody (Jackson, West Grove, PA; ref. culture supernatants of MDA-MB-231 were collected 96 hours after 111-095-144). adenoviral infection of cell monolayers. Samples (8 Ag) were incubated Flow cytometry. For flow cytometric analysis, cells were harvested with the rabbit anticanstatin antibody and were revealed with a peroxidase- following adenoviral infection [coupled, or not, with small inhibitory RNA conjugated donkey antirabbit antibody (Amersham, Buckinghamshire, (siRNA) transfection] for 48 hours. Cells were stained with fluorescein- United Kingdom). conjugated mAbs (10 Ag/mL) directed against specific integrins: LM609 h h For CanHSA and K1-3HSA expression analysis, samples were incubated (anti-av 3), P1F6 (anti-av 5), and AMF7 (anti-av). Normal mouse with a mouse anti-HSA antibody (Biovalley, Marne la Valle´e, France; ref. fluorescein-conjugated IgG1 was used as a negative control antibody. ab7793). For K1-3 and K1-3HSA expression analysis, membranes were For cell cycle studies, HUVEC and MDA-MB-231 cells were collected at incubated with an antihuman plasminogen monoclonal antibody (mAb) 96 hours postinfection. Cells were suspended in PBS/0.1% glucose and then A1D12 (11). Both of these antibody complexes were revealed by a goat fixed with 70% ethanol. After centrifugation, the supernatants were antimouse immunoglobulin G1 (IgG1)-horseradish peroxidase secondary discarded and cellular DNA was stained in a solution containing 100 Ag/mL antibody (SouthernBiotech, Birmingham, AL; ref. 1070-05). of RNase A and 50 Ag/mL of propidium iodide. For canstatin and CanHSA stability analysis in vivo, MDA-MB-231 Determination of caspase-3, -8, and -9 activities. HUVEC, MDA-MB- xenografted tumors were dissected following intratumoral adenoviral 231, and PC-3 cells were harvested 72 hours after infection with adenovirus, injection. Tumor protein extracts (20 Ag) were assessed for canstatin and with or without caspase-3 and -9 inhibitors, during the infection period CanHSA expression as described above. (R&D Systems, Minneapolis, MN; ref. FMK004, FMK008). Caspase-3, -8, and -9 Quantification of CanHSA and K1-3HSA in vivo. For quantification of activities were determined by caspase-3, -8, and -9 colorimetric assays (R&D CanHSA and K1-3HSA by ELISA, wells were coated with a mouse anti-HSA Systems). Staurosporine, which induces the mitochondrial caspase-dependent mAb described above. After washing, wells were incubated with mice sera apoptotic pathway, was used as an apoptotic positive control (50 nmol/L; at the appropriate dilution or a HSA solution standard (Sigma-Aldrich, St. Oncogene, Darmstadt, Germany). A B B Louis, MO; ref. A8763). Wells were washed and incubated with a rabbit anti- v, 3, and 5 integrin gene silencing in HUVEC cells. siRNAs were HSA polyclonal antibody (Sigma-Aldrich; ref. A0433). Antigen-antibody purchased as designed, synthesized, purified, and annealed oligos from complexes were revealed with an alkaline phosphatase–conjugated goat Ambion (Huntingdon, United Kingdom). The target sequence names for h h antimouse IgG antibody (Promega, Madison, WI; H + L) and development human 3, 5, and av integrins are ITGB3_1, ITGB5_1, and ITGAV_1. Cells at with an alkaline phosphatase kit (Bio-Rad, Hercules, CA). Absorbance at 50% confluence were treated with 50 nmol/L siRNA in Oligofectamine 405 nm was converted to HSA concentration (ng/mL) via a purified HSA reagent (Invitrogen) according to the instructions of the manufacturer. A calibration curve. nonsilencing siRNA (Ambion), transfected under the same conditions as Endothelial cell proliferation, migration assays, and contrast- anti-integrin siRNAs, was used as control. Four hours following siRNA enhanced color Doppler high-frequency ultrasonography of tumor transfection, cells were infected with AdCO1, AdK1-3HSA, or AdCanHSA. neovascularization. Experiments were done as described (17). Expression of integrins was analyzed by flow cytometry 48 hours after Immunohistology and quantification of the intratumoral micro- adenoviral transduction as described above. Caspase-8 and -9 activities were vessels. We sacrificed six mice per group at an early stage (day 8) following determined following infection of cells for 72 hours as described above. intratumoral injection of each adenovirus before the appearance of Statistical analysis. Each experiment was done twice. Statistical necrosis. Paraffin sections of PC-3 tumors were incubated to quantify the analysis was done using the Student’s t test (unilateral and unpaired). number of intratumoral vessels, as described (17). For each animal (n = 6), a representative murine CD31-immunostained histologic sample was Results analyzed using a Zeiss Axiophot microscope (Zeiss, Oberkochen, Germany) and a PCO digital camera. Eight fields per animal were digitized (100 Â), In vitro and in vivo characterization of recombinant excluding fibrotic areas ascertained by HES staining of parallel sections. adenoviruses encoding canstatin, CanHSA, K1-3, and K1- Images were then analyzed with custom software run on a Linux-Mandrake 3HSA. Experiments were done to verify expression and secretion of 8.2 workstation. canstatin and CanHSA in vitro and in vivo. In vivo curative and preventive treatment. Human tumors were In vitro, each protein was detected at the expected MW of 24 induced by s.c. injection of 4 Â 106 MDA-MB-231 or 5 Â 106 PC-3 cells into and 89 kDa, respectively, from AdCan- and AdCanHSA-infected cell the dorsa of nude mice. When tumors had reached 3 to 4 mm in diameter, supernatants at 96 hours postinfection (Fig. 1A, 1). No signal was recombinant adenoviruses (n = 6 for each group) were injected into tumors observed from AdCO1-infected supernatants. Similar experiments 9 [2 Â 10 plaque-forming unit (pfu) per injection, for a total of two injections were carried out for AdK1-3 and AdK1-3HSA (Fig. 1A, 3). There separated by a 48-hour interval]. was judged to be no quantitative difference in the maximal level Â 9 Adenoviruses (5 10 pfu) were injected systemically 24 hours before expression of all four proteins through the hybridization signal to s.c. implantation of 4 Â 106 MDA-MB-231 cells (n = 8 for AdCO1, n = 11 for the shared HSA tag (Fig. 1A, 2). AdK1-3 and AdK1-3HSA, n = 12 for AdCan, and n = 15 for AdCanHSA). In vivo, the circulating levels of expressed proteins in blood For each experiment, blood samples were collected at days 8, 15, and 20 to quantify antiangiogenic proteins by ELISA. samples of MDA-MB-231 xenografted mice, following intra- Immunohistology and apoptotic cell quantification (terminal tumoral or systemic injection of AdCanHSA and AdK1-3HSA, deoxyribonucleotidyl transferase–mediated dUTP nick end labeling). were quantified by ELISA (Fig. 1B). CanHSA and K1-3HSA An in situ cell death detection kit employing the terminal deoxyribonu- showed a similar expression pattern in the sera of mice after cleotidyl transferase–mediated dUTP nick end labeling method (TUNEL; both local and systemic adenoviral injection.

Cancer Res 2005; 65: (10). May 15, 2005 4354 www.aacrjournals.org

Figure 1. Assessment of the expression level in vitro and in vivo of canstatin, CanHSA, K1-3, and K1-3HSA proteins. A, Western blotting of adenoviral infected MDA-MB-231 cell culture supernatants confirms that expression levels of canstatin, CanHSA, K1-3, and K1-3HSA are similar. B, quantification of CanHSA and K1-3HSA by ELISA against the HSA moiety of the fusion proteins in the sera of mice at several time points following two intratumoral injections of the appropriate adenovirus (1) or systemic injection (2). C, Western blotting of protein extracts of adenoviral-infected xenografted MDA-MB-231 tumors show the increased in vivo stability of CanHSA (1; for 9 days) as compared with unfused canstatin (2).

To determine whether fusion with HSA affects the in vivo However, microvessels remained numerous and relatively thick in stability of canstatin, Western blot analysis of MDA-MB-231 tumor contrast to the vessels observed in the canstatin-treated samples. protein extracts was done at days 2, 5, and 9 following intra- More precisely, injections of AdCan and AdCanHSA decreased the tumoral injection of AdCan, AdCanHSA, or AdCO1. CanHSA was area of vascularization (88% on average; P <10À6 as compared the only protein for which a consistent level of expression was with AdCO1). The results are also significant as compared with the detected for 9 days. No signal was observed for unfused canstatin angiostatin-treated group (P < 0.01). at day 9 (Fig. 1C). Thus, fusion with HSA improves the intra- All tumor vessels are not equal in their ability to provide oxygen tumoral bioavailability of canstatin. and nutrients to the tumor cells they support (18). The Canstatin has a more prominent antiangiogenic effect than quantification of intratumoral vessels by contrast-enhanced color angiostatin. The proliferation of HUVEC cells was more strongly Doppler high-frequency ultrasonography analysis enabled us to inhibited by AdCan or AdCanHSA (60% on average) than by AdK1- follow the efficacy of treatment over the treatment period by 3 or AdK1-3HSA (40% on average; P =10À6 as compared with focusing specifically on the functional vessels in the tumor. At the AdCO1). Following in vitro infection at different multiplicities of end of the treatment period (day 42), capillary vessels were infection, canstatin exhibited a dose-dependent inhibition of observed to invade tumors only in the AdCO1-, AdK1-3–, and endothelial cell proliferation (data not shown). Similarly, canstatin AdK1-3HSA–treated groups. No intratumoral vessels were detected and CanHSA arrested the migration of HUVEC cells (80%; P = 0.02 following injection of AdCan and AdCanHSA (Fig. 2C). The as compared with AdCO1) more strongly than observed with K1-3 thickness of vessel walls (CD31 staining) treated with K1-3 or or K1-3HSA (not significant; Fig. 2A). K1-3HSA at day 8 was thinner by visual inspection than for the Having shown in vitro that the inherent activities of canstatin CO1 slice (Fig. 2B), which, however, seemed to contain as many and K1-3 were not disrupted by fusion with HSA (Fig. 2A), their functional vessels as the thicker nontreated vessels (Fig. 2C). antiangiogenic effects were quantified in vivo by anti- Interestingly, CanHSA induced a strong and long-lasting inhibition CD31 immunohistochemistry following injection of viruses of the penetration of functional vessels into the tumor mass as into PC-3 tumors. As shown in Fig. 2B, the level of intratumoral compared with unfused canstatin (90% and 60%, respectively; P < vascularization was reduced following canstatin treatment. The 0.01 as compared with AdCO1; Fig. 2C). thickness of capillary microvessel walls in the K1-3–treated group Thus, unfused canstatin has potent antiangiogenic features was thinner than those observed in the CO1-treated group. in vitro and in vivo, but we show its fusion with HSA undisputably www.aacrjournals.org 4355 Cancer Res 2005; 65: (10). May 15, 2005

Figure 2. Characterization of the in vitro and in vivo antiangiogenic effects of canstatin and CanHSA. A, canstatin- and CanHSA-induced inhibition of HUVEC proliferation (1) and migration (2) compared with that of K1-3 and K1-3HSA. Endothelial cell viability was assessed at 96 hours postinfection by 3-(4 c,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide assay. Percentage inhibition was calculated with reference to the control adenovirus (AdCO1). Representative of two separate experiments. B, tumor vessel endothelial cells (arrows) were immunostained with a murine anti-CD31 antibody following intratumoral injection of AdCO1- (1), AdK1-3– (2), AdK1-3HSA– (3), AdCan- (4), or AdCanHSA-injected (5) PC-3–derived xenografted tumors (100Â). 6, columns, mean of the area of tumor vascularization for each animal; bars, SD. C, color Doppler high-frequency ultrasonography of MDA-MB-231–derived tumors (boxed), following intratumoral injection of the adenoviruses, listed above, in living mice was used to estimate the number of functional intratumoral capillary vessels (white arrows) once per week for 6 weeks (1). Columns, mean number of vessels for each group of five mice for a given observation point (week); bars, SD. Representative images of K1-3HSA (2) and CanHSA (3) at day 42 (other data are not shown). Color Doppler high-frequency ultrasonography was not done for PC-3– derived tumors due to the aggressive evolution of this tumor (with respect to the slower growing MDA-MB-231–derived tumors), which prevented pragmatic observation of functional intratumoral capillary vessels over the short tumor growth period. improves its efficacy in vivo in inhibiting intratumoral vasculari- antitumor effect of the CanHSA conjugate as compared with that zation in the weeks following treatment. of free canstatin in vivo. CanHSA exhibits a stronger in vivo antitumor effect than CanHSA induces an apoptotic mechanism in both endothe- canstatin, K1-3, or K1-3HSA. To evaluate whether the strong lial and tumoral cells. We showed in situ by TUNEL analysis that antiangiogenic activity of CanHSA was sufficient to broadly affect CanHSA induced tumor cell apoptosis in MDA-MB-231 xenografts tumor viability, we followed the growth of xenografted tumor cells on nude mice (Fig. 3B). The results showed a marked increase after curative and prophylactic injection of the recombinant of apoptotic cells in the AdCanHSA-injected tumors 30 days adenoviruses. postinfection (11 F 4 cells/field, P <10À5, for CanHSA as compared Adenoviruses were injected directly into MDA-MB-231 xeno- with 1.7 F 0.9 cells/field observed for the AdCO1-treated group). grafted tumors. Following expression of canstatin and CanHSA, The results obtained for CanHSA are again significant when tumor growth was inhibited by 40% (not significant) and 98% (P < compared with those obtained for the angiostatin-treated group 0.003), respectively, as compared with AdCO1. Moreover, CanHSA (P <10À4) and the canstatin-treated group (P <10À3). exhibited better efficacy with 20% total tumor regression than To determine whether CanHSA acts directly against tumor cells, AdCO1- or AdCan-treated group. Similarly, PC-3 tumor growth was or indirectly by inducing tumor hypoxia, nuclear DNA fragmen- inhibited by CanHSA with better efficacy as compared with tation, a hallmark of cellular apoptosis, was assessed by propidium canstatin and CO1 (72%; P =10À4 at day 19). The antitumor effect iodide staining of AdCanHSA-infected MDA-MB-231 or HUVEC of CanHSA is still significant as compared with K1-3 or K1-3HSA cells. CanHSA induced apoptosis in 31% of tumor cells whereas no (P < 0.04) for both tumor models (Fig. 3A). apoptotic cells were observed after treatment with AdK1-3HSA or To evaluate the efficacy of CanHSA in a prophylactic protocol, we AdCO1 (Fig. 3C). Moreover, CanHSA treatment leads to apoptosis systemically injected each recombinant virus 24 hours before s.c. in 42% of the endothelial cells whereas only 17% of K1-3HSA– injection of MDA-MB-231 cells in mice. After 8 weeks of treatment, treated cells were observed in apoptotic (sub-G1) phase. No 77% inhibition of tumor growth was observed with CanHSA as apoptotic cells were detected in the CO1-treated population. compared with CO1- (P = 0.003) and canstatin-treated groups (P = The strong antiangiogenic and antitumor effects of canstatin 0.01; Fig. 3A). These results were also significant as compared with seem to be due to a direct apoptotic mechanism in endothelial and K1-3 (P = 0.005) and K1-3HSA (P = 0.006; data not shown). tumoral cells. The long-lasting effect of CanHSA in inhibiting tumor penetra- CanHSA interacts with integrins on the endothelial and tion by intratumoral vessels is also correlated with the stronger tumor cell surface. To determine whether CanHSA exhibited a

Cancer Res 2005; 65: (10). May 15, 2005 4356 www.aacrjournals.org

h direct antitumor effect, we infected MDA-MB-231 and HUVEC cells (FACS) analysis was done using fluorescein-conjugated anti-av 3 in vitro with adenoviruses. Then, cells were fixed, but not or anti-avh5 mAbs on AdCO1- or AdCanHSA-infected cells. permeabilized, and the presence of CanHSA and K1-3HSA was Initially, we characterized the presence of avh3 and avh5 on detected by immunostaining. Endothelial cells infected by HUVEC and MDA-MB-231 cells following infection with mock AdCanHSA or AdK1-3HSA were clearly labeled compared with adenovirus. avh3 and avh5 integrins were detected at high AdCO1-infected HUVEC cells (Fig. 4A). On the contrary, only MDA- levels on endothelial cells. MDA-MB-231 cells express a low h h MB-231 cells infected by AdCanHSA were labeled. Thus, CanHSA level of av 3 and a high level of av 5, supporting the results of seems to localize at the extracellular surface of both HUVEC and Wong et al. (21). After expression of CanHSA by infected cells, MDA-MB-231 cells. we repeated the FACS analysis to assess the level of integrins Endothelial cell migration and adhesion to the ECM are bound to CanHSA, which would be inaccessible to anti-integrins h h h h mediated by integrins, especially av 3 and av 5 (19, 20), which mAbs. Fifty percent to 66% of the av, 3,or 5 integrins on the also play a role in tumor progression (21). To identify the HUVEC and MDA-MB-231 cells were not bound to mAbs after receptor that binds CanHSA, fluorescence-activated cell sorting expression of CanHSA (Fig. 4B). On the contrary, expression of

Figure 3. Tumor growth inhibition and detection of apoptosis following infection with AdCan and AdCanHSA. A, MDA-MB-231– (1) or PC-3–derived (2) tumor size on nude mice, following two intratumoral injections of the adenoviruses listed above, was measured once a week. Points, mean of six mice as compared with the AdCO1-treated group; bars, SE. Asterisks, statistical significance. 3, inhibition of tumor growth by CanHSA in comparison with the canstatin-, K1-3–, and K1-3HSA–treated groups highlights significant tumor regression mediated by CanHSA as compared with other treatment options versus AdCO1. 4, s.c. injected MDA-MB-231–derived tumor growth over f8 weeks following systemic injection of AdCO1, AdCan, and AdCanHSA. B, TUNEL staining was done on adenoviral-treated MDA-MB-231 tumors (day 30 postinfection) as previously described in A1. The proportion of apoptotic cells (arrows) per field (five fields per animal) were quantified within sections (200Â) from AdCO1-, AdK1-3–, AdK1-3HSA–, AdCan-, or AdCanHSA-injected tumors. Representative fields of AdCO1 (1), AdK1-3HSA (2), and AdCanHSA (3). 4, columns, mean of two mice for each treated group; bars, SD. C, cell cycle analysis. HUVEC and MDA-MB-231 cells were infected with AdCanHSA (1 and 4), AdK1-3HSA (2 and 5), or AdCO1 (3 and 6). At 96 hours postinfection, cells were fixed with 70% ethanol overnight and chromatin was stained with propidium iodide. FACS analysis was done twice for each group. www.aacrjournals.org 4357 Cancer Res 2005; 65: (10). May 15, 2005

Figure 4. Determination of CanHSA binding to HUVEC and MDA-MB-231 cell surface. A, HUVEC and MDA-MB-231 cells were infected with AdCanHSA (1 and 2), AdK1-3HSA (3 and 4), or AdCO1 (5 and 6) at 800 pfu for 48 hours. Cells were fixed but not permeabilized and incubated with a mouse anti-HSA mAb followed by an anti-mouse FITC-labeled antibody. Note the localization of CanHSA at the surface of HUVEC cells (1) as well as tumor cells (2). Each experiment was done twice. B, to determine which av integrins are associated with CanHSA or K1-3HSA, FACS competition analysis was done on AdCO1- (blue line), AdCanHSA- (green line), or AdK1-3HSA–infected (orange line) HUVEC (1-4) or MDA-MB-231 (5 and 6) cells using anti-avh3 (1, 3,and5) or anti-avh5 (2, 4, and 6) FITC-conjugated mAbs. FACS analysis included a mouse FITC-conjugated IgG1 as negative control (red line). Y-axis, events (cell number); X-axis, fluorescence intensity.

K1-3HSA impeded binding of anti-avh3 mAb only on the endothelial pathways in endothelial cells through the activation of caspase- cell surface. For each cell type, normal fluorescein-conjugated 8 and caspase-9 leading to cleavage of procaspase-3 (7). K1-3 isotype control antibody IgG1 was used as a negative control, activates only the mitochondrial-independent pathway involving confirming that the observed signals were specific. cleavage of caspase-8 and -3. To confirm that CanHSA binds to avh3 and avh5 integrins at the CanHSA-treated MDA-MB-231 cells showed an activation of the cell surface, immunoprecipitations were done from AdCanHSA- or mitochondrial caspase-9–dependent apoptotic process (increase AdCO1-infected cell lysates with mAbs against a h or a h . Western of 1.7-fold; P = 0.04 as compared with AdCO1; Fig. 5A). v 3 v 5 h h blotting revealed bands characteristic of CanHSA from immuno- Additionally, PC-3 prostate tumor cells express av 3 and av 5 precipitated AdCanHSA-infected endothelial and tumor cellular integrins, and activation of caspase-9 was observed following extracts following hybridization with anti-HSA (data not shown). CanHSA or staurosporine treatment of PC-3 (P <0.01as Taken together, these results show that CanHSA interacts with at compared with AdCO1; Fig. 5A). No CanHSA-induced caspase- h h 8 activation was detected in either tumor cell type (data not least two kinds of integrins, av 3 and av 5, on the surface of both human endothelial and tumor cells, whereas K1-3HSA binds a h shown). v 3 Caspase-dependent apoptotic events are mediated by integrins only on the membrane of HUVEC cells. Thus, canstatin binding of CanHSA to A , B , and B integrins. To further inhibits the broad capacity of endothelial and tumor cells to adhere v 3 5 characterize the functional receptors mediating the apoptotic and spread by blocking ECM-a -integrin interactions (22). v signaling pathways for CanHSA, we used anti-integrin siRNAs to CanHSA-induced apoptosis is sequential to mitochondrial decrease a , h ,orh expression in HUVEC cells to inhibit binding damage both in endothelial and tumor cells. Antagonists of v 3 5 of CanHSA to the cell surface. Incubation of AdCO1-infected a h integrins have been shown to participate in inhibition of cell v 3 HUVEC cells for 48 hours with anti-h , anti-h , or anti-a siRNAs proliferation and migration as well as apoptosis (23). 3 5 v decreased the level of expression of each integrin population on To determine whether canstatin induces caspase-dependent the cell surface by 50% on average (Fig. 5B). Identical transfection apoptotic pathways, caspase-3, -8, and -9 activities in HUVEC and with a random siRNA bearing no homology with any integrin MDA-MB-231 cellular extracts were quantified. CanHSA treatment genes had no effect on integrin expression. of HUVEC cells led to a 4-fold increase in caspase-9 activity as h h The specific inhibition of av-, 3-, or 5-integrin expression compared with CO1-treated cells (P = 0.03). K1-3HSA–treated cells significantly reduced caspase-9 activity following CanHSA treat- exhibited no significant activation of caspase-9 (Fig. 5A). Caspase- ment (decrease of 1.8- to 3.8-fold; P < 0.01 as compared with 8 activity was increased less strongly (1.5-fold) by either canstatin transfection of siRNA control oligo coupled with AdCanHSA (P = 0.024 as compared with AdCO1) or angiostatin. In addition, infection; Fig. 5C). canstatin induced a 3.5-fold increase in caspase-3 activity as As previously shown, K1-3HSA increases caspase-8 activity alone compared with CO1 treatment (P = 0.03), whereas angiostatin (by 2-fold; Fig. 5A). Inhibition of av-orh3-integrin expression increased this level by only 2.5-fold. Each result was done by impeded the activation of caspase-8 following K1-3HSA, but not h comparison to staurosporine, a known inducer of caspase- CanHSA, infection (Fig. 5D). Thus, binding of K1-3HSA with av 3 dependent apoptosis. The specificity of caspase cleavage was shown integrins induces a caspase-8–dependent apoptotic pathway, by addition of selective caspase-9 or -3 inhibitors to the cell medium. whereas interaction of CanHSA with avh3 and avh5 integrins Our results show that canstatin induced two distinct apoptotic selectively triggers the caspase-9–dependent pathway.

Cancer Res 2005; 65: (10). May 15, 2005 4358 www.aacrjournals.org

Discussion The soluble NC1 domain fragment of the a2(IV) chain of collagen, called canstatin, is known to potently inhibit tumor- associated angiogenesis (5–7). Initially, we investigated the in vivo antiangiogenic properties and antitumor efficacy of canstatin using HSA fusion to increase the bioavailability of canstatin following adenoviral gene transfer. This study sought to evaluate the anticancer efficacy of CanHSA compared to that of the K1-3HSA as a reference molecule belonging to the plasminogen-derived class of proteins (10, 11). In vivo, CanHSA strongly inhibited the vascularization of tumors through a stable and long-lasting effect. Our approach provides a solution to maintain an in situ level of canstatin sufficient to strongly impede the growth of cancer and disrupt implantation of cancer cells and, as such, decrease the number of therapeutic injections required and thus, potentially, adherence to the treatment regime. Recently, therapeutic gene transfer became more commercially attractive following the debut of adenoviral- mediated p53 gene transfer for head and neck squamous cancers, called Gendicine (Sibiono GeneTech, Shenzhen, China). Next, we attempted to delineate the mechanism of action of CanHSA to explain its improved efficacy as compared with K1- 3HSA. The exact mechanisms by which a(NC1) domains, in particular canstatin (a2(IV)NC1), inhibit tumor angiogenesis are not completely understood. We show here that canstatin binds to the endothelial and tumor cell surface in an integrin- dependent manner. Integrins mediate adhesion to matrix proteins of both endothelial and tumor cells (19–21). Remodeling of the ECM promotes novel integrin-ligand interactions required for angiogenesis (1). As such, CanHSA could compete with collagen IV ECM for cell surface integrin binding and reverse the proliferative and migratory effects induced by cell-ECM inter- h h actions. Thus, av 3, and av 5 integrins would seem to mediate the antiangiogenic and direct antitumor properties of CanHSA. Just two other proteins are known to have a direct effect on tumor cells through interaction with integrins: endostatin (a5h1) h and tumstatin (av 3), which are both also collagen-derived NC1 domain fragments (24–26). Interestingly, Tarui et al. (27) also showed that K1-3 binds only to avh3 integrins on endothelial cells. Integrin receptors are known to cross-talk with other receptors and their associated pathways (22). Angiogenesis proceeds through two cytokine-dependent pathways: one depending on avh3 that is largely induced by basic fibroblast growth factor and vascular endothelial growth factor (VEGF), and another one which h is potentiated by av 5 and VEGF alone (19, 20). As such, dual antagonists of both a h and a h , such as canstatin, as we have Figure 5. Determination of CanHSA-induced apoptotic pathway and the key v 3 v 5 receptor(s) through siRNA-mediated silencing of av, h3,orh5 expression. established here, would have a definite therapeutic advantage by A, HUVEC (1-3), MDA-MB-231 (4), or PC-3 (5) cells were infected with AdCO1, blocking the two tumor-induced angiogenesis pathways. AdK1-3HSA, or AdCanHSA for 48 and 72 hours, respectively. Whole cell lysates were mixed with reaction buffer and caspase colorimetric substrate as Here, we investigated the potential role of integrins as regulators recommended by R&D Systems. A Bradford assay was done to normalize of apoptosis (23, 28). We have shown that in vitro CanHSA strongly absorbances. Each experiment was done thrice. Columns, mean; bars, SD. To inhibited the migration and proliferation of HUVEC cells. More- show expression of the canstatin-adhesive proteins, avh3 and avh5 integrins, on the PC-3 cell surface, FACS analysis was done on noninfected PC-3 cells over, our present findings show that these events are mediated by h h h h (blue line) using anti-av 3 (6) or anti-av 5 (7) FITC-conjugated mAbs as an upstream event involving canstatin binding to av 3 and av 5 compared with IgG1 FITC-conjugated antibody (red line). B, to assess the integrins, which has not yet been reported. Integrin receptor suppression of integrin expression using siRNA, FACS analysis was done on AdCO1-infected HUVEC cells which were transfected with random siRNA engagement and subsequent clustering lead to transduction of h h (1-3; blue line), or 3 (1), 5 (2), or av siRNA (3; green line). A mouse positional cues from the ECM to intracellular signaling phosphor- FITC-conjugated IgG1 was included as a negative control (red line). Y-axis, events (cell number); X-axis, fluorescence intensity. To determine the integrin ylation cascades that promote cell survival, such as the activation receptor(s) that promote caspase-9 (C) or caspase-8 (D) activities following of focal adhesion kinase/phosphatidylinositol 3-kinase (FAK/PI3K) CanHSA infection, HUVEC cells were pretransfected with each siRNA and pathway (23, 29). Phosphorylated FAK promotes integrin-mediated then infected with either AdCO1, AdK1-3HSA, or AdCanHSA for 72 hours. Caspase activities were quantified as described previously in A. Each cell proliferation and migration events (29). When cell-ECM experiment was done twice. interactions are lost, cells undergo a type of apoptosis called www.aacrjournals.org 4359 Cancer Res 2005; 65: (10). May 15, 2005

caspase-9 as compared with K1-3 treatment of cells. This pathway is triggered through both avh3 and avh5 integrins. Both canstatin and K1-3 induce the mitochondria-independent pathway with a similar increase in caspase-8 activity. Our findings show that canstatin initiates two apoptotic pathways including activation of caspase-8 and -9, both initiators of the downstream apoptotic process leading to activation of caspase-3. Canstatin-activated caspase-8, by down-regulation of Flip levels and up-regulation of Fas/Fas ligand, triggers not only cell death directly through caspase-3 activation but also indirectly through mitochondrial damage via activation of caspase-9 within the apoptosome (7, 31, 32). On the other hand, phosphorylated FAK/PI3K is known to inactivate the mitochondrial apoptotic pathway by inhibition of caspase-9 (30). So, canstatin directly activates procaspase-9 through inhibition of the FAK/PI3K pathway and amplifies the Fas-dependent pathway in mitochondria (ref. 29; Fig. 6). This dual mechanism explains how canstatin strongly enhances caspase-9 activity and therefore the number of cells entering apoptosis. Mitochondrial disruption and caspase-9 activation can engage in a circular self-amplification loop that accelerates the apoptotic process. Mitochondrial damage seems to be the critical point of no return of the cell death process (33). Several studies have shown that angiostatin inhibits endothelial cell function by increasing apoptosis (11, 15, 16). However, the moderate reduction in K1-3–treated endothelial cell number seen in our proliferation assays is due to a single activation of caspase- 8 without mitochondrial damage following specific interaction of K1-3HSA with avh3. Recently a study has described that unligated avh3 integrins (but not h5) or soluble antagonists of this receptor initiate endothelial apoptosis, called integrin-mediated death, by recruitment of caspase-8 to the plasma membrane independent of interaction of Fas/Fas ligand with the death domain (34). However, CanHSA-induced caspase-8 activity was not disrupted following av or h3 gene silencing, as we have established following treatment with K1-3HSA. In this case, canstatin would act only via one of two possible mechanisms: Fas-mediated pathway in caspase-8 activity; the second possible mechanism (i.e., integrin-mediated death leading to caspase-8 activity) being abrogated due to siRNA knockdown of avh3 (Fig. 6). Collectively, it would seem that the involvement of distinct Figure 6. Schematic illustration of distinct signaling pathways induced by apoptotic pathways is correlated with profound cell death and thus canstatin and angiostatin in endothelial or tumoral cells. A, canstatin binds to avh3 and avh5, whereas angiostatin (K1-3) binds to avh3 alone on endothelial a strong antitumor effect. cell surface. Canstatin initiates two apoptotic pathways, involving activation In our study, we report that canstatin inhibits tumor develop- of caspase-8 and -9, leading to activation of caspase-3. Canstatin activates procaspase-9 not only directly through inhibition of the FAK/PI3K/Akt pathway ment in vivo and enhances apoptotic induction in tumor cells (1) but also by amplifying indirectly the mitochondrial pathway through through generation of active caspase-9. No caspase-dependent h Fas-dependent caspase-8 activation (2). In contrast, K1-3 binding to av 3 alone apoptotic mechanism was observed in K1-3–treated tumor cells, as leads to activation of procaspase-8 with no effect on apoptotic mitochondrial events (3). B, canstatin activates only the mitochondrial pathway in tumoral cells. described previously by Lucas et al. As confirmed by other authors, fragments derived from NC1 domains, like tumstatin, could act as a specific antitumor antagonist of avh3 integrins in melanoma anoı¨kis (30). The FAK pathway is activated by most integrins. progression (26). Although both endostatin and tumstatin display differential In conclusion, CanHSA is a powerful anticancer molecule which integrin binding characteristics, they both inhibit phosphorylation triggers two distinct programmed cell death pathways in endothe- of FAK, thus leading individually to distinct downstream signaling lial cells which converge on mitochondrial disruption: the av- events (24). Recently, Panka and Meier (7) have shown that integrin-FAK/PI3K/caspase-9 pathway and the Fas/Fas ligand/ canstatin inhibits the phosphorylation of FAK, interfering with caspase-8/caspase-9 cascade. These two pathways result in downstream signaling events. Collectively, these data suggest that amplification of mitochondrial-linked apoptotic events and activa- h h the binding of canstatin to av 3 and av 5 integrins, shown here, tion of caspase-3, the central executioner of the apoptotic process. could mediate FAK inactivation, leading to suppression of cell Our study argues for the delivery in situ of therapeutic gene spreading and migration and thus apoptosis. with mitochondrial-dependent proapoptotic properties to im- To further document endothelial cell-specific apoptosis, our prove antiangiogenic and antitumor features of such selective results show that canstatin induces a profound activation of dual avh3 and avh5 agonists fused with a carrier protein.

Cancer Res 2005; 65: (10). May 15, 2005 4360 www.aacrjournals.org

Acknowledgments le Cancer. C. Magnon was financed by the Institut National de la Sante´ et de la Recherche Me´dicale. Received 9/30/2004; revised 2/10/2005; accepted 3/9/2005. The costs of publication of this article were defrayed in part by the payment of page Grant support: Centre National de la Recherche Scientifique, the Institut Gustave charges. This article must therefore be hereby marked advertisement in accordance Roussy, the ligue nationale contre le cancer, and the Association pour la Recherche sur with 18 U.S.C. Section 1734 solely to indicate this fact.

References mouse skeletal muscles and heart. J Clin Invest 1992; inducing apoptosis of angiogenic blood vessels. Cell 90:626–30. 1994;79:1157–64. 1. Xu J, Rodriguez D, Petitclerc E, et al. Proteolytic 13. Capon DJ, Chamow SM, Mordenti, et al. Designing 24. Sudhakar A, Sugimoto H, Yang C, Lively J, Zeisberg M, exposure of a cryptic site within collagen IV is required CD4 immunoadhesins for AIDS therapy. Nature Kalluri R. Human Tumstatin and human endostatin for angiogenesis and tumor growth in vivo. J Cell Biol 1989;337:525–31. exhibit distinct antiangiogenic activities mediated by 2001;154:1069–79. h h 14. Bouquet C, Frau E, Opolon P, et al. Systemic av 3 and a5 1 integrins. Proc Natl Acad Sci U S A 2. Timpl R, Wiedemann H, Van Delden V, Furthmayr H, administration of a recombinant adenovirus encoding 2003;100:4766–71. Ku¨hn K. A network for the organization of type IV a HSA-angiostatin kringle 1-3 conjugate inhibits MDA- 25. Dkhissi F, Lu H, Soria C, et al. Endostatin exhibits a collagen molecules in basements membranes. FEBS MB-231 tumor growth and metastatization in a direct antitumor effect in addition to its antiangiogenic Lett 1981;120:203–11. transgenic model of spontaneous eye cancer. Mol activity in colon cancer cells. Hum Gene Ther 2003;14: 3. Sundaramoorthy M, Meiyappan M, Todd P, Hudson BG. Ther 2002;7:174–84. 997–1008. Crystal structure of NC1 domains. J Biol Chem 2002;277: 15. Lucas R, Holmgren L, Garcia I, et al. Multiple forms 26. Floquet N, Pasco S, Ramon L, et al. The antitumor 31142–53. of angiostatin induce apoptosis in endothelial cells. properties of the a3(IV)-(185-203) peptide from the NC1 4. Folkman J. Anti-angiogenesis: new concept for Blood 1998;92:4730–41. domain of type IV collagen (Tumstatin) are conformation- therapy of solid tumor. Ann Surg 1972;175:409–16. 16. Claesson-Welsh L, Welsh M, Ito N, et al. Angiostatin dependent. J Biol Chem 2004;279:2091–100. 5. Petitclerc E, Boutaud A, Prestayko, et al. New induces endothelial cell apoptosis and activation of 27. Tarui T, Miles LA, Takada Y. Specific interaction of functions for non-collagenous domains of human focal adhesion kinase independently of the integrin- angiostatin with integrin avh3 in endothelial cells. J Biol collagen type IV. J Biol Chem 2000;275:8051–61. binding motif RGD. Proc Natl Acad Sci U S A 1998;95: Chem 2001;276:39562–8. 6. Kamphaus GD, Colorado PC, Panka DJ, et al. Canstatin, 5579–83. 28. Hynes RO. A reevalution of integrins as regulators of a novel matrix-derived inhibitor of angiogenesis and 17. Galaup A, Magnon C, Rouffiac C, et al. Full kringles angiogenesis. Nat Med 2002;8:918–21. tumor growth. J Biol Chem 2000;275:1209–15. of plasminogen (AA 1-566) mediates complete regres- 29. Chen H, Guan JL. Association of focal adhesion 7. Panka DJ, Mier JW. Canstatin inhibits Akt activation sion of breast tumors in mice. Gene Ther. In kinase with its potential sustrate phosphatidylinositol and induces Fas-dependant apoptosis in endothelial press 2005. 3-kinase. Proc Natl Acad Sci U S A 1994;91:10148–52. cells. J Biol Chem 2003;278:37632–6. 18. Hlatky L, Hahnfeldt P, Folkman J. Clinical applica- 30. Frisch SM, Vuori K, Ruoslahti E, Chan-Hui PY. 8. Maeshima Y, Colorado PC, Torre A, et al. Distinct tion of antiangiogenic therapy: microvessel density, Control of adhesion-dependent cell survival by focal antitumor properties of a type IV collagen domain what it does and doesn’t tell us. J Natl Cancer Inst adhesion kinase. J Cell Biol 1996;134:793–9. derived from basement membrane. J Biol Chem 2000; 2002;94:883–93. 31. Aoudjit F, Vuori K. Matrix attachment regulates 275:21340–8. 19. Friedlander M, Brooks PC, Shaffer RW, Kincaid CM, Fas-induced apoptosis in endothelial cells : a role for 9. O’Reilly MS, Boehm T, Shing Y, et al. Endostatin: an Varner JA, Cheresh DA. Definition of two angiogenic c-Flip and implications for anoı¨kis. J Cell Biol 2001; endogenous inhibitor of angiogenesis and tumor pathways by distinct av integrins. Science 1995;270: 152:633–43. growth. Cell 1997;88:277–85. 1500–2. 32. Li H, Zhu H, Xu CJ, Yuan J. Cleavage of BID by 10. Cao Y, Ji RW, Davidson D, et al. Kringle domains of 20. Eliceiri BP, Cheresh DA. Role of av integrins during caspase 8 mediates the mitochondrial damage in the human angiostatin. J Biol Chem 1996;271:29461–7. angiogenesis. Cancer J 2000;S245–9. Fas pathway of apoptosis. Cell 1998;94:491–501. 11. Griscelli F, Li H, Bennaceur-Griscelli A, et al. 21. Wong NC, Mueller BM, Barbas CF, et al. av integrins 33. Green D, Kroemer G. The central executioners of Angiostatin gene transfer: inhibition of tumor growth mediate adhesion and migration of breast carcinoma apoptosis: caspases or mitochondria? Trends Cell Biol in vivo by blockage of endothelial cell proliferation cell lines. Clin Exp Metastasis 1998;16:50–61. 1998;8:267–71. associated with a mitosis arrest. Proc Natl Acad Sci 22. Eliceiri BP, Cheresh DA. Adhesion events in angio- 34. Stupack DG, Puente XS, Boutsaboualoy S, Storgard U S A 1998;95:6367–72. genesis. Curr Opin Cell Biol 2001;13:563–8. CM, Cheresh DA. Apoptosis of adherent cells by 12. Stratford-Perricaudet LD, Makeh I, Perricaudet M, 23. Brooks PC, Montgomery AMP, Rosenfeld M, et al. recruitment of caspase-8 to unligated integrins. J Cell h Briand P. Widespread long-term gene transfer to Integrin av 3 antagonists promote tumor regression by Biol 2001;155:459–70.

www.aacrjournals.org 4361 Cancer Res 2005; 65: (10). May 15, 2005

Efficacy of Canstatin and Mechanism of Action In the article on the efficacy of canstatin and mechanism of action in the May 15, 2005 issue of Cancer Research (1), Ce´line Bouquet should have appeared in the list of authors. The list of authors should have read as the following: Claire Magnon, Ariane Galaup, Brian Mullan, Vale´rie Rouffiac, Ce´line Bouquet, Jean-Michel Bidart, Frank Griscelli, Paule Opolon, and Michel Perricaudet. Dr. Bouquet’s affiliation is UMR 8121 Laboratoire de vectorologie et transfert de ge`nes, Institut Gustave Roussy, Villejuif cedex, France.

1. Magnon C, Galaup A, Mullan B, Rouffiac V, Bidart JM, Griscelli F, Opolon P, Perricaudet M. Canstatin acts on endothelial and tumor cells via mitochondrial damage initiated through interaction with avh3 and avh5 integrins. Cancer Res 2005;65:4353–61.

www.aacrjournals.org 5989 Cancer Res 2005; 65: (13). July 1, 2005 Canstatin Acts on Endothelial and Tumor Cells via Mitochondrial Damage Initiated through Interaction with αvβ 3 and αvβ5 Integrins

Claire Magnon, Ariane Galaup, Brian Mullan, et al.

Cancer Res 2005;65:4353-4361.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/65/10/4353

Cited articles This article cites 30 articles, 18 of which you can access for free at: http://cancerres.aacrjournals.org/content/65/10/4353.full#ref-list-1

Citing articles This article has been cited by 15 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/65/10/4353.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 Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/65/10/4353. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.