Therapeutic Expression of an Anti-Death Receptor 5 Single-Chain Fixed-Variable Region Prevents Tumor Growth in Mice

Research Article

Juan Shi,1 Yanxin Liu,1 Yong Zheng,2 Yabin Guo,1 Jinchun Zhang,1 Pik-to Cheung,3 Ruian Xu,1,4,5 and Dexian Zheng1

1National Laboratory of Medical Molecular Biology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China; 2Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, Ontario, Canada; 3Department of Pediatrics and Adolescent Medicine and 4Gene Therapy Laboratory, University of Hong Kong, Hong Kong, China; and 5Institute of Molecular Medicine, Huaqiao University, Quanzhou, China

Abstract recombinant soluble TRAIL (sTRAIL) are also shown to cause The clinical use of the single-chain fixed-variable (scFv) apoptosis in normal cells (especially in hepatocytes), hampering its fragments of recombinant monoclonal antibodies as credible clinical use for cancer therapy (8, 9). As an alternative, agonistic alternatives for classic therapeutic antibodies has two monoclonal antibodies (mAb) specifically targeting DR5 are raised limitations: rapid blood clearance and inefficient local for the purpose of cancer treatment (10, 11). We previously expression of functional molecules. In attempt to address reported a novel mouse anti-human DR5 mAb, AD5-10, which these issues, we have developed a novel gene therapy protocol stimulated apoptosis in several tumor cell lines in vitro in the absence of cross-linking and showed strong tumoricidal activity in which the anti-death receptor 5(DR5)scFv fragments were either in vitro expressed in several tumor cell lines, or in vivo in vivo with negligible cytotoxicity toward normal cells (12). expressed in mice, using recombinant adeno-associated virus A critical drawback of using mouse antibodies on patients, (rAAV)–mediated gene transfer. Viral transduction using the however, is associated with the human immunoreactions to murine rAAV-S3C construct, which encodes a scFv molecule (S3C scFv) antibodies, which makes long-term use of this therapy infeasible specific to DR5, led to stable expression in tumor cell lines and for patients with chronic progressive conditions such as cancers showed apoptosis-inducing activity in vitro, which could be (13). The immunogenicity of rodent mAbs sometimes can be inhibited by recombinant DR5but not by DR4. A single i.m. successfully reduced, but in a laborious and time-consuming way, injection of rAAV-S3C virus in nude mice resulted in stable by antibody chimerization and humanization using protein- expression of DR5-binding S3C scFv proteins in mouse sera for engineering technologies. In addition, due to the large size of at least 240 days. Moreover, the expression of S3C scFv was intact antibodies, they are difficult to penetrate into solid tumors, associated with significant suppression of tumor growth and further limiting their applications in immunotherapy (14). the increase of tumor cell apoptosis in previously established Recently, the development of small recombinant antibody s.c. human lung LTEP-sml and liver Hep3B tumor xenografts. fragments as high-affinity therapeutic reagents with reduced (Cancer Res 2006; 66(24): 11946-53) immunogenicity has come under the spotlight (15–17). A popular format of engineered recombinant antibody fragment is the single- Introduction chain fixed-variable (scFv) molecule, in which the VH and VL regions of the parental antibody are joined by a polypeptide linker. Death receptor 5 [DR5, or tumor necrosis factor (TNF)–related The scFv fragment retains the target specificity and antigen- apoptosis-inducing ligand (TRAIL) receptor 2], a member of the binding affinity of the intact antibody, whereas it can be genetically TNF receptor superfamily, is emerging as a favorable target for designed and produced in large quantity by ectopically expressing the design of anticancer agents in recent years (1). When triggered both VH and VL regions from a single cDNA in cells (18, 19). by its ligand, DR5 undergoes oligomerization, mediated by its Moreover, due to its smaller size, the scFv molecule shows cytoplasmic death domains, and propagates apoptotic signaling improved pharmacokinetics in tumor penetration and is better cascades through nucleating the formation of death-inducing tolerated by the host immune system (20, 21). The length of the signaling complexes (2–5). DR5 can be specifically activated by linker peptide seems to play an important role in affecting the TNF-related apoptosis-inducing ligand (TRAIL), which also acti- function of scFv by determining whether the scFv molecules exist vates another functional death receptor, DR4 (or TRAIL receptor 1). as monovalent monomers, or autoassemble into divalent dimmers Upon triggering, DR5 selectively induces cell death in a wide variety (diabodies) and trivalent trimers (tribodies; refs. 22–29). of tumor cells both in vitro and in vivo under different Despite the many advantages of using scFv molecules for experimental settings (4–7). However, at least certain versions of immunotherapy, especially for chronic diseases, the treatment efficacy, however, is often compromised by the rapid blood clearance of infused scFv proteins and the difficulty in maintaining Note: Supplementary data for this article are available at Cancer Research Online high local concentrations of operative molecules at the target site (http://cancerres.aacrjournals.org/). J. Shi and Y. Liu contributed equally to this work. (13). To overcome these problems, techniques to in vivo express Requests for reprints: Dexian Zheng, National Laboratory of Medical Molecular therapeutic antibodies or their fragments have been developed. Biology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences and Among them, recombinant adeno-associated virus (rAAV)–based Peking Union Medical College, 5 Dong Dan San Tiao, Beijing 100005, China. Phone: 86- 10-6529-6409; Fax: 86-10-6510-5102; E-mail: [email protected] or Ruian Xu, vectors are recently shown to be a credible gene carrier for in vivo Institute of Molecular Medicine, Huagiao University, Quanzhuo, China. E-mail: gene expression in a variety of gene therapy applications. The [email protected] I2006 American Association for Cancer Research. advantages of using rAAV-based vectors include long-term trans- doi:10.1158/0008-5472.CAN-06-1227 gene expression and low immunogenicity (30).

Cancer Res 2006; 66: (24). December 15, 2006 11946 www.aacrjournals.org

In this study, we developed a gene therapy protocol in which our placed at 4jC in the dark for 0.5 hour before flow cytometry analysis. The previously reported mouse anti-human DR5 mAb, AD5-10, was propidium iodide fluorescence of individual nuclei was measured in the red fluorescence and the data were registered in a logarithmic scale. engineered into the scFv format and introduced into s.c. human 4 lung LTEP-sml and liver Hep3B tumor xenografts for in vivo At least 10 cells of each sample were analyzed. Apoptotic nuclei appeared as a hypodiploid DNA peak before the G phase of cell cycle, which expression via rAAV vector–mediated gene transfer. Xenografts i.m. 1 was easily distinguished from hyperdiploid DNA peak of cells at the G injected with rAAV-S3C, a rAAV carrying a cDNA that encodes the 1 phase or later at the G2 and S phases. Cell viability was quantified by anti-DR5 scFv (S3C scFv) molecule, exhibited sustained serum level 3-(4,5-dimethyl-thiazole-2-yl)-2,5-biphenyl tetrazolium (MTT) assay (Sigma, A of DR5-binding S3C scFv (100–200 g/mL) for at least 240 days. The St. Louis, MO; ref. 32). virally expressed scFv antibody fragments stimulated apoptosis in Membrane protein extraction. The cell membrane fraction was tumor cells and inhibited tumor growth in mice. obtained by homogenization and differential centrifugation. Briefly, cells were homogenized and sonicated on ice in homogenization buffer [20 mmol/L Tris-HCl (pH 7.4), 10% glycerol, 4 mmol/L EDTA, 2 mmol/L Materials and Methods EGTA, 10 mmol/L leupeptin, 1 mmol/L phenylmethylsulfonyl fluoride]. The Animals. All surgical procedures and care administered to the animals homogenate was spun at 86,000 Â g at 4jC for 45 minutes. The pellet were approved by the University of Hong Kong Ethics Committee and done (membrane fraction) was resuspended in homogenization buffer containing according to institutional guidelines. BALB/c male nude mice, 4 to 5 weeks 1% Triton X-100 and left on ice for 30 minutes, followed by centrifugation at old, were housed at a constant temperature and supplied with laboratory 14,000 Â g at 4jC for 30 minutes. The supernatant from this spin was the chow and water ad libitum on a 12 hours:12 hours light/dark cycle. solubilized membrane fraction. Cells. Human lung cancer cell line A549 (lung adenocarcinoma), human Detection of DR5by flow cytometry assay. Phycoerythrin-labeled anti- liver cancer cell lines Hep3B and HepG2, and human colon cancer cell line human DR5 antibody (R&D Systems, Minneapolis, MN) was used to detect HCT116 were originated from the American Type Culture Collection the DR5 expression level on the cell surface. The assay was carried out (Rockville, MD). Human liver cancer cell lines SMMC-7721 and BEL7402 according to the manufacturer’s instruction. Briefly, 105 cells were added were supplied by the Institute of Cell and Biochemistry, Chinese Academy of into phycoerythrin-conjugated antibodies after first blocking with human Sciences (Shanghai, China). Human small lung cancer cell line LTEP-sml IgG and incubated for 30 minutes at 4jC. After washing twice with PBS, cell was kindly provided by Professor Xiaoguang Chen (Institute of Materia surface expression of TRAIL receptors was determined by flow cytometry Medica, Chinese Academy of Medical Sciences, Beijing, China). These cells using 488-nm-wavelength laser excitation. The phycoerythrin-labeled were cultured at 37jC in RPMI 1640, F12, or DMEM (Life Technologies, mouse IgG2B (R&D Systems) was used as control. Grand Island, NY). The primary human hepatocytes (PHH), kindly provided Analysis of transgene expression. The expressions and activities of by the Department of Surgery, University of Hong Kong, were maintained in transgenes were verified by using standard techniques, including Western Waymouth’s MB-752/1 medium (Life Technologies) supplemented with blot analysis, immunohistochemical staining, and terminal deoxynucleo- 2 mmol/L glutamine, 150 nmol/L insulin, 10% FCS, 100 units/mL penicillin, tidyl transferase–mediated dUTP nick end labeling (TUNEL) assay, as 100 Ag/mL streptomycin, and 100 nmol/L dexamethasone, and were plated previously described (32). The antibodies against XIAP, cIAP1, caspase-3, at a density of 5 Â 104 cells/cm2. and caspase-8 were from Santa Cruz Biotechnology (Santa Cruz, CA). Construction of the recombinant rAAV-S0C, rAAV-S1C, rAAV-S3C, S3C scFv concentration determination. The S3C scFv concentration in rAAV-SIRESC vectors, and rAAV vectors encoding sTRAIL and the medium and mouse sera was determined by ELISA. The purified enhanced green fluorescent protein. The CAG promoter, the inserted recombinant S3C scFv protein expressed in Escherichia coli was used as genes, and polyadenylic acid sequences were inserted between the inverted standard. Briefly, 96-well plates were coated for 2 hours at 37jC with terminal repeats of an AAV plasmid using appropriate restriction enzymes. 200 AL/well of recombinant S3C protein (1 ng/mL–10 Ag/mL, 1:5 serial A woodchuck hepatitis posttranscriptional regulatory element (WPRE), dilution), or medium and mouse sera prediluted (1:10–1:500) with which was shown to boost transcription, was inserted into the constructs 0.05 mol/L Na2CO3/NaHCO3 (pH 9.6). After washing with PBS, the plates immediately after the expression cassette (31). The cDNA of AD5-10 was were blocked with the blocking buffer (0.05% Tween 20 and 5% nonfat j j used as the template for the scFv fragments. VH and VL sequences of milk in PBS) at 37 C for 0.5 hour, and followed by incubation at 37 C for AD5-10 were cloned and linked with a linker sequence to form a scFv 2 hours with an anti-AD5-10 antibodies raised in rabbit using a synthetic

‘‘gene’’ in the VH-linker-VL orientation. S0C was the scFv construct encod- high variable region peptide of AD5-10 heavy chain. After washing, ing the VH and VL domains that are directly linked. In S1C and S3C, the the horseradish peroxidase (HRP)–conjugate goat anti-rabbit IgG and VH domain was separated from the VL domain by a linker sequence 1,2-benzenediamine (0.4 mg/mL, Sigma) was sequentially added and the encoding a polypeptide containing 5 amino acids (GGGGS) and 15 amino color was allowed to develop at room temperature for 0.5 hour. The acids (GGGGSGGGGSGGGGS), respectively. In the SIRESC construct, absorbance was measured at 490 nm on a SPECTRAMAX 340 microplate sequence encoding VH domain was separated from VL domain by an reader (Molecular Devices, Menlo Park, CA). internal ribosome entry site to ensure the expression of these two domains Analysis of DR5binding by S3C scFv. S3C scFv expressed in HEK293 from the same RNA template. The scFv constructs with different linkers cells transduced with the rAAV-S3C vector or in the sera of the mice were subcloned into the AAV plasmid and subsequently packaged and injected with rAAV2/1-S3C were detected. To evaluate binding activity of purified. Recombinant AAV vectors encoding sTRAIL (rAAV-sTRAIL) or S3C to human DR5, the plates were coated with 200 ng/well of recombinant enhanced green fluorescent protein (rAAV-eGFP) were constructed as DR5 protein (12), blocked with 5% nonfat dry milk, and incubated with previously described (32). various concentrations of S3C scFv or AD5-10. After sequential incubation Production of rAAV particles. AAV serotype 2 (AAV2) was used for with rabbit anti-AD5-10 antibodies and HRP-conjugated goat anti- in vitro studies. For in vivo experiments, however, a pseudotyping strategy rabbit antibodies, the bindings were visualized by adding the peroxidase was used to produce rAAV particles packaged with AAV1 capsid proteins substrate. The absorbance was measured at 490 nm on a microplate reader. (AAV2/1). This is because AAV1 is more efficient in muscle transduction The data were analyzed by using GraphPad Prism software (GraphPad and evokes less host immune reaction. Infective rAAV particles were Software, San Diego, CA). generated by a three-plasmid, helper virus–free packaging method (32). The The S3C scFv binding to DR5 was also determined by coimmunopre- titer of rAAVs was quantified by real-time PCR analysis. cipitation. Medium from HEK293 cell culture transduced with rAAV-S3C for Apoptosis and viability determination. Cell apoptosis was deter- 4 days were incubated at 37jC for 3 hours with 100 ng/mL homemade mined by propidium iodide staining and flow cytometry assay. In brief, recombinant human DR5. Anti-AD5-10 antibody and protein A Sepharose cells were harvested by trypsinization at the indicated hours after were then added and the coprecipitated proteins were resolved with SDS- treatment and stained with 50 Ag/mL propidium iodide. The tubes were PAGE and probed for DR5. www.aacrjournals.org 11947 Cancer Res 2006; 66: (24). December 15, 2006

For immunoprecipitation of native DR5 in cells, HCT116 cells were incubated with the medium of rAAV-S3C–transduced HEK293 cells. Cells were lysed in the buffer [30 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% NP40, 10% glycerol] supplemented with protease inhibitor cocktail (Roche, Mannheim, Germany). After 15 minutes of incubation on ice, the lysates were centrifuged at 20,000 Â g and 4jC for 15 minutes. Approximately 700 to 800 Ag of total protein were incubated with the anti-AD5-10 antibody followed by immunoprecipitation with magnetic protein A. The coprecipitated proteins were subjected to SDS-PAGE and DR5 immunoblotting analysis. S.c. tumor xenografts and assessment of tumor growth. The Hep3B cells (1 Â 106 in 100 AL PBS) or LTEP-sml cells (5 Â 106 in 100 AL PBS) were injected in the s.c. tissue of the right dorsal flank of 6-week-old male nude mice. Tumor growth was monitored twice a week by measuring tumor size with calipers. Tumor volume (V) was calculated using the formula V = 1/2 Â length Â (width)2. When the tumors had reached f50 mm3 in size, animals were divided into two groups (n = 6–10) and the indicated rAAV vectors were administered by i.m. injection. Group 1 received rAAV2/ 1-S3C at a dose of 1 Â 1011 genome equivalents. Group 2 received empty rAAV2/1 at a dose of 1 Â 1011 genome equivalents. Experiments were terminated when tumors in the control group showed signs of necrosis. Statistical analysis. Results were expressed as mean values F SD. Student’s t test was used to evaluate statistical significance. P < 0.05 was considered significant. Figure 1. Construction and expressions of rAAV-scFv constructs. A, schematic illustration of anti-DR5 scFv–expressing cassettes. SP, signal peptide. G4S, linker with the amino acid sequence GGGGS, where G is glycine and S is serine. B, Western blot analysis of scFv expression in rAAV-scFv–infected HEK293 Results cells. Cells were infected with the indicated rAAV constructs at a MOI of 5 Â 104 Construction and expression of scFv fragments. The scFv and were collected 72 hours later. Cell lysates were subjected to SDS-PAGE and probed for scFv fragments using anti-AD5-10 antibodies. h-actin was also constructs were generated by using the cDNA encoding the VH and probed with specific antibodies for its use as the loading control. VL chains of AD5-10 as the template. As previously mentioned, the length of the linker may affect the function of the scFv molecules. We thus constructed four AD5-10 scFv cDNAs, in which the VH and showed stronger inhibitory activity than rAAV-sTRAIL in Hep3B VL encoding sequences were either directly linked (S0C scFv and LTEP-sml cells (P < 0.01, 96 hours), suggesting that the activity cDNA) or separated by oligonucleotides encoding for different of rAAV-S3C may be dependent on cell type. The cytotoxicity of lengths of linker peptides (S1C, S3C, and SIRESC scFv cDNAs). The rAAV-S3C seemed to be specific for tumor cell lines because the S1C and S3C scFv cDNAs expressed S1C and S3C scFv molecules viability of the PHHs were unaffected by virus infection (Fig. 2A). with a 5-amino-acid (GGGGS) and a 15-amino-acid linker Because the parental antibody of S3C scFv, AD5-10, had been (GGGGSGGGGSGGGGS), respectively. In SIRESC scFv cDNAs, an reported to induce apoptosis in tumor cell lines in a DR5- internal ribosome entry site element was inserted between VH and dependent manner, we then analyzed the DR5 expression in the VL cDNAs to promote the expression of VH and VL domain membrane fraction of cell lysates prepared from above cell lines by independently (Fig. 1A). The scFv cDNA fragments were then immunoblotting. As shown in Fig. 2B, all tumor cell lines expressed cloned into a rAAV vector. The recombinant virus was prepared by the DR5 protein, although at different levels. Interestingly, no a three-plasmid, helper virus–free packaging method (32). The membrane DR5 was detectable in PHH cells (Fig. 2B), and resulting viruses were added into HEK293 cell cultures for forced overexpression of DR5 in PHH cells (Fig. 2C) failed to infection. Western blot using an antibody specific for VH domain sensitize PHH cells to S3C scFv–induced cell death (Fig. 2D) even of AD5-10 showed that the rAAV scFv constructs were expressed at at high MOI. a similar level in infected cells (Fig. 1B). To determine whether the S3C scFv–mediated cell killing Cytotoxicity of the S3C scFv fragments among various involves the function of DR5, we carried out a competitive tumor cell lines. The cytotoxic activities of these scFv fragments experiment in vitro where increasing concentrations of recombi- were tested in several tumor-derived cell lines. Each cell line was nant DR5 or DR4 proteins were added into rAAV-S3C– or rAAV- infected with equal numbers of viral particles [multiplicity of eGFP–infected LTEP-sml cell culture. Relative cell viability was infection (MOI) = 5 Â 104] for 24, 48, or 96 hours. At each time determined by the MTT assay. As shown in Fig. 2E, recombinant point, the viability of infected cells was determined by MTT assay. DR5 protein, but not DR4, protected LTEP-sml cells from the As shown in Fig. 2A, transduced expression of the rAAV-S3C vector cytotoxicity of rAAV-S3C. A similar result was observed in rAAV- resulted in cell growth inhibition in most cell lines tested, including S3C–infected Hep3B cell culture (data not shown), suggesting the the human lung cancer cell lines A549 and LTEP-sml, liver cancer involvement of DR5 in S3C scFv–induced cytotoxicity. cell lines HepG2 and Hep3B, and colon cancer cell line HCT116. Collectively, these data showed that the cytotoxicity of As a comparison, the extent of viability reductions induced by engineered AD5-10 scFv fragments toward tumor cell lines might the rAAV-S0C, rAAV-S1C, or rAAV-SIRESC transduction resembled be affected by length of the linker peptide. The rAAV-S3C that of the rAAV-eGFP infection in these cells (P < 0.05 in A549 cell construct, which encoded an anti-DR5 scFv fragment with a line and P < 0.01 in other four tumor cell lines, 96 hours). Similar 15-amino-acid linker, was able to induce cell death via DR5 in a results were observed in human liver cancer cell lines SMMC 7721 time-dependent manner in several cancer cell lines, but not in and BEL 7402 (data not shown). Notably, the rAAV-S3C construct normal human hepatocytes.

Cancer Res 2006; 66: (24). December 15, 2006 11948 www.aacrjournals.org

S3C scFv induced tumor cell apoptosis via a classic The virally expressed S3C scFv fragments were released in a apoptosis pathway. To determine if the cytotoxicity induced by soluble form into the animal sera, which can be readily determined rAAV-S3C–mediated gene expression was due to apoptosis, by ELISA, with a peak level at 133.37 F 26.33 Ag/mL 60 days after propidium iodide staining and flow cytometry analysis were done injection (Fig. 4C). With only a slight decline after 60 days, the in rAAV-S3C–infected Hep3B and LTEP-sml cells. The hypodiploid serum level of scFv seemed to be stable and was maintained at DNA peaks, which indicate cell apoptosis, were seen in rAAV- approximately 100 to 120 Ag/mL throughout the 8 months’ study. sTRAIL– and rAAV-S3C–infected cells (Fig. 3A); the apoptosis rate Taken together, these results showed that i.m. injection of of LTEP-sml cells 72 hours after transduction of rAAV-S3C was rAAV2/1-S3C virus induced stable expression of S3C scFv frag- 30.37 F 8.14%, which was significantly higher than rAAV-sTRAIL– ments in systemic circulation. infected LTEP-sml cells (P = 0.0034 versus rAAV-eGFP, P = 0.0089 Secreted S3C scFvs efficiently bound to DR5. Next, we tested versus rAAV-sTRAIL), and the apoptosis rate of rAAV-S3C–infected if the virally expressed S3C scFv fragments in the medium of Hep3B cells was 27.91 F 4.95% (P = 0.0017 versus rAAV-eGFP, infected cells and the sera of transduced mice maintained its P = 0.027 versus rAAV-sTRAIL). Stimulation of DR5 usually induces affinity toward its antigen. Immobilized DR5 protein in a 96-well the classic apoptosis pathway characterized by caspase-3 and plate was incubated with the medium of rAAV-S3C–infected caspase-8 activation, as well as simultaneous down-regulation of HEK293 cells or the sera of rAAV-S3C–transduced mice. The the inhibitor of apoptosis protein XIAP and cIAP1. Indeed, these binding affinity between the two proteins was measured by ELISA. were observed in tumor cell lines expressing S3C scFv fragments The results showed that serum S3C scFv efficiently bound to its Â À9 (Fig. 3B). These data showed that the cytotoxicity of S3C scFv in antigen (Kd =4 10 mol/L). When compared with that of the À10 tumor cell lines was due to the activation of the classic apoptosis intact antibody (Kd =10 mol/L), the DR5 binding affinity of the pathway mediated by death receptors. antibody fragment was slightly lower, but a little higher than that of Muscular expression of S3C scFv in mice. To express the S3C the S3C scFvs secreted to the medium of virus-infected HEK293 Â À9 scFv in mice, a pseudotyped version of rAAV2/1 was created by cells (Kd =3.3 10 mol/L; Fig. 5A). To further confirm the using rep from AAV2 and cap from AAV1. The rAAV2/1 vector was binding of S3C scFv to DR5, recombinant DR5 protein was added selected for its low immunogenicity and efficient transduction in into medium prepared from rAAV-S3C virus–infected HEK293 to muscle cells (33, 34). The rAAV2/1 virus was used for in vivo induce the formation of S3C scFv-DR5 complex, which was, in turn, delivery of S3C scFv via i.m. injection into BALB/c nude mice. coimmunoprecipitated using the anti-AD5-10 antibody. As shown As shown in Fig. 4A, the expression of S3C scFv was easily in Fig. 5B, the DR5 was coimmunoprecipitated with the antibody detectable in isolated mouse muscle sections of 50 days after fragments only in rAAV-S3C–transduced cells. To show the injection of 1 Â 1011 genome equivalents of the rAAV2/1-S3C virus. recognition of native membrane DR5 by secreted S3C scFv Immunoblotting analysis further confirmed S3C scFv expression antibody fragments, the cell culture medium from rAAV-S3C– in vivo in injected muscle tissues (Fig. 4B). transduced HEK293 cells were added into HCT116 or the PHH

Figure 2. S3C scFv–stimulated tumor cell death involved DR5 function. A, cytotoxicities of different rAAV-scFv constructs on various tumor cell lines. The A549, LTEP-sml, Hep3B, HepG2, and HCT116 cells were infected with rAAV-scFv virus at a MOI of 5 Â 104 for the indicated time. Cell viability was determined by MTT assay. Columns, mean of three independent experiments with three replicates; bars, SD. *, P < 0.05; **, P < 0.01. B, immunoblot analysis of membrane DR5 expression in various tumor cells. Membrane fractions containing equal amounts of total protein were prepared from the whole-cell lysates and were subjected to SDS-PAGE, followed by immunoblotting with an anti-DR5 antibody. C, surface DR5 expression in the PHH cells. PHHs were transfected with pRK-DR5 expression vector and stained with an anti-DR5 antibody followed by flow cytometry analysis. Black filled area, pRK-DR5–transfected PHHs. Open area, wild-type PHHs (DR5 negative). Gray filled area, negative control (murine IgG2B). D, unsusceptibility of DR5-positive PHHs to S3C scFv–mediated apoptosis. pRK-DR5–transfected PHH cells were treated with the indicated virus at MOIs of 1 Â 104 to 1 Â 106. Cell viability was determined by MTT assay. Columns, mean of three independent experiments with three replicates; bars, SD. E, DR5, but not DR4, blocked the cytotoxicity of S3C scFv by competition. Indicated concentrations of recombinant DR5 or DR4 proteins were added into rAAV-S3C– or rAAV-eGFP–infected LTEP-sml cells. Relative cell viability was determined by MTT assay. Representative of three independent experiments. www.aacrjournals.org 11949 Cancer Res 2006; 66: (24). December 15, 2006

reduced the mean tumor volume from 3,061 F 306 mm3, as observed in empty vector–injected mice, to 1,788 F 297 mm3 (P < 0.01), accounting for a 42% decrease in tumor growth (Fig. 6B). To determine whether treatment with rAAV2/1-S3C led to apoptosis in tumor cells, TUNEL assays were carried out in lung tumor sections acquired from rAAV2/1-S3C– or empty vector– injected mouse xenografts. Results showed that the expression of the S3C scFv antibody fragment was associated with an accumulation of apoptotic cells within the tumors, which was largely absent in control mice (Fig. 6C). We also examined the histopathologic changes in lung, liver, spleen, and kidney of the animals used in this study to investigate the cytotoxicity of S3C scFvs toward normal tissues. No obvious lesions were found in any of the organs (data not shown). Moreover, the rAAV2/1-S3C–injected mice did not show any sign of systemic toxicity as estimated by body weight, gross appearance, and behavior.

Discussion In the present study, we developed a novel therapeutic strategy by expressing an anti-DR5 scFv antibody fragment, S3C scFv, in vivo through rAAV-based, vector-mediated gene transfer. We Figure 3. rAAV-S3C–mediated gene expression induced apoptosis in tumor showed that this scFv fragment was stably expressed in mice by cell lines. A, the Hep3B and LTEP-sml cells were infected by the indicated virus at a MOI of 5 Â 104 and analyzed 72 hours after infection by propidium iodide staining and flow cytometry. B, caspase activation and XIAP and cIAP1 inhibition in the virus-infected cells. LTEP-sml and Hep3B cells were infected with rAAV-S3C at a MOI of 5 Â 104 and harvested at the indicated time points. Approximately 50 Ag of total protein in the cell lysates were resolved by SDS-PAGE. Caspase-3 and caspase-8 activation and XIAP and cIAP-1 inhibition were detected by immunoblotting using corresponding antibodies. h-actin was used as the loading control. cultures. S3C scFv were then immunoprecipitated with the anti- AD5-10 antibody. Coimmunoprecipitants were resolved on SDS- PAGE and probed for DR5. As shown in Fig. 5C, the native DR5 was able to form a stable complex with S3C scFv in HCT116 cells. The S3C-DR5 complex could not be found in PHH cells, which express undetectable level of membrane DR5. These results showed that the secreted S3C scFv antibody fragment showed similar binding affinity as its parental antibody toward recombinant and native DR5 protein. S3C scFv suppressed the growth of s.c. tumor xenografts by inducing tumor cell apoptosis. To analyze the therapeutic potential of the S3C scFv fragments, we established a tumor model in BALB/c nude mice by s.c. inoculation of human lung cancer LTEP-sml cells (5 Â 106 per injection) in the right dorsal flank of the mice. When the tumors had reached a size of f50 mm3, animals were divided into two groups (n = 6) and received an i.m. injection of 1 Â 1011 particles of rAAV2/1-S3C or empty rAAV2/1 as control. The lung tumor volume was measured twice a week over a period of f50 days. The result showed that the mean F 3 tumor volume, which was 455 86 mm in the control group, was Figure 4. S3C scFv expression in mice. A, muscular expression of S3C scFv. reduced to 249 F 62 mm3 in mice injected with rAAV2/1-S3C, The BALB/c nude mice were given a single i.m. injection of 1 Â 1011 particles demonstrating a substantial 45% decrease (P < 0.01) in tumor of rAAV2/1-S3C or empty rAAV2/1 vector. Muscle sections were prepared 50 days after injection and were used for immunohistochemical staining using growth (Fig. 6A). a rabbit anti-AD5-10 antibody followed by HRP-conjugated goat anti-rabbit In another xenograft model, we s.c. implanted 1 Â 106 of human antibody, and visualized by 3,3¶-diaminobenzidine staining. Magnification, Â100. B, whole-cell lysates were prepared from muscles of virus-infected nude mice. liver cancer Hep3B cells into the mice to establish a liver tumor The expression of S3C scFv protein were determined by immunoblotting xenograft model. Again, the animals were divided into two groups using anti-AD5-10 antibodies. Each lane corresponds to an individual mouse. (n = 8–10) and the effect of rAAV2/1-S3C on tumor growth C, level of secreted S3C scFv in mouse serum. Sera from rAAV-S3C scFv–infected ( w ) or empty vector–infected (n) mice were collected at the suppression was similarly evaluated as in the human lung cancer indicated times. The level of S3C scFv in sera were determined by ELISA xenograft model. The result showed that injection of rAAV2/1-S3C using anti-AD5-10 antibodies. Points, mean of three mice; bars, SD.

Cancer Res 2006; 66: (24). December 15, 2006 11950 www.aacrjournals.org

27–29). In our study, the S3C scFv antibody fragment, which has a linker of 15 residues and is presumed to be monovalent, showed the highest apoptosis-inducing activity in several tumor cell lines, which can be blocked by soluble DR5 but not by DR4. This result is consistent with our previous observation that the parental AD5-10 mAb did not require secondary cross-linking for DR5 triggering (12). We speculate that the underlying mechanism may be involved in antibody-induced specific conformational changes that are sufficient to stimulate receptor aggregation. However, further studies are required to confirm this hypothesis. DR5 was widely expressed in a variety of tumor cells, including melanoma cell lines and nearly all non–small cell lung cancer tumors (35, 36). Several studies had shown that TRAIL sensitivity was positively related to the level of death receptors (37–39). However, that effect might be limited to certain cell types. For example, cell lines such as LTEP-sml, Hep3B, and certain

Figure 5. S3C scFv formed complex with DR5. A, binding affinity of S3C scFv to DR5. Recombinant human DR5 protein of indicated concentrations was immobilized on a 96-well plate. The plate was then incubated with the AD5-10 antibody, culture medium from rAAV-S3C–infected HEK293 cells, or serum from mice infected with the virus for 50 days. Binding affinities of AD5-10 in vitro and in vivo secreted S3C scFv to DR5 were measured by ELISA using anti-AD5-10 antibodies. The absorbance was measured at 490 nm on a microplate reader. Representative of three independent experiments. B, the cell culture medium of HEK293 cells infected with rAAV-S3C or rAAV-eGFP virus for 4 days were collected and incubated with (+) or without (À) recombinant human DR5. Immunoprecipitation (IP) was done by adding anti-AD5-10 antibodies and protein-A Sepharose beads. The coimmunoprecipitated proteins were subjected to SDS-PAGE and immunoblotting, and probed for DR5. C, recognition of native membrane DR5 by secreted S3C scFv antibody fragments. The cell culture medium of rAAV-S3C– or rAAV-eGFP–transduced HEK293 cells were added into HCT116 and PHH cultures. Endogenous S3C-DR5 complex were then immunoprecipitated with anti-AD5-10 antibodies from cell lysates. Coimmunoprecipitants were resolved on SDS-PAGE and probed for DR5. i.m. injection and had tumor suppression activity in two s.c. human tumor xenograft models. The scFv antibody fragments became recently known as promising alternatives for intact antibodies for immunotherapy. Figure 6. I.m. injection of rAAV2/1-S3C suppressed the growth of s.c. tumors in The linker region that connects V and V chains may affect the mice. A, LTEP-sml were injected in the s.c. tissue of the right dorsal flank of H L nude mice. When the tumors had reached a size of f50 mm3, animals were activity of scFvs. A linker of 15 residues or more usually renders divided into two groups and i.m. injected with 1 Â 1011 particles of rAAV2/1-S3C the scFv monomeric with similar antigen-binding affinity as the ( w ) or empty rAAV2/1 (n). Tumor volumes were determined as described in Materials and Methods. Points, mean of the tumor volume in six mice; bars, SD. parental antibody (22, 23). Reducing linker length leads to B, tumor growth in the Hep3B tumor–bearing mice injected with either autoassembly of scFv monomers to form functional divalent rAAV2/1-S3C ( w ) or the empty rAAV2/1 vector (n). Points, mean of the tumor dimmers (diabodies; refs. 24–26). When the linker is completely volume in 8 to 10 mice; bars, SD. C, increased tumor cell apoptosis in S3C scFv–treated mice. LTEP-sml lung tumor sections were prepared from S3C removed, scFv molecules exist as trimers (tribodies), which can be scFv–treated mice and control mice. The apoptosis of tumor cells in these Â either trivalent or nonfunctional (noncognate VH/VL pair; refs. sections were determined by TUNEL staining (green). Magnification, 400. www.aacrjournals.org 11951 Cancer Res 2006; 66: (24). December 15, 2006

Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2006 American Association for Cancer Research. Cancer Research melanoma cell lines were relatively resistant to sTRAIL despite biopsies at the site of injection. As shown in Supplementary Fig. SA, moderate receptor expression and/or absence of decoy receptors the DNA fragment encoding for WPRE was not detectable in liver, (40). In this study, although all of the five tumor cell lines expressed heart, spleen, lung, and kidney. Meanwhile, the expression of S3C different levels of surface DR5, their sensitivities toward S3C scFv– scFv mRNA in the muscle was only found in the rAAV-S3C virus– induced cell death were not proportional to the level of DR5 transduced mice, as determined by reverse transcription-PCR expression. Therefore, cells may incorporate additional regulatory analysis (Supplementary Fig. SB). The absence of virus sequence in signaling pathways to control DR5 receptor functions in a cell tissues other than muscle was consistent with the established type–specific manner. Indeed, multiple layers of regulation have properties of AAV2/1-based vectors. Therefore, the high muscle- been suggested during TRAIL receptor activation (41–43). specific expression of S3C scFv achieved in vivo via rAAV2/1- The PHH cells did not express detectable membrane DR5 and mediated gene transfer would make this vector potentially suitable were resistant to the killing by S3C scFvs. However, forced for human use. overexpression of DR5 in PHH cells failed to sensitize PHH cells The apoptotic effect of S3C scFv antibody fragments on tumor to S3C scFv–induced cell death. As previously discussed, although cells in vivo and in vitro can only be described as moderate, overexpression of death receptors often enhances apoptosis in cells although the concentrations of soluble S3C scFv protein secreted (44), it is not always the case (40). For example, despite similar into the culture medium and serum were relatively high. One surface expression of TRAIL receptor, primary and transformed possible explanation might be the relatively low affinity of the keratinocytes exhibited marked difference in TRAIL sensitivity (45). fragment compared with the parental antibody. Protein engineer- Thus, sensitivity to receptor stimulation could be determined by ing on AD5-10 is now underway in our laboratory to improve the the concerted regulations of multiple mechanisms including the binding affinity of both intact antibody and its fragments, which level and availability of DR5, the balance between decoy receptors may enhance the apoptosis-inducing activity of the antibodies. In and death receptors, and distinct intracellular signaling pathways addition, increased therapeutic efficacy can be expected by (40–43). Moreover, the mechanism underlying AD5-10–mediated combined use of rAAV-S3C virus with conventional therapeutics, cell death pathway is not yet fully characterized and requires such as chemotherapeutic agents and irradiation. further analysis in details. Nevertheless, the resistance of PHH to This study took advantage of our specific anti-DR5 (AD5-10) S3C scFv–mediated cytotoxicity provided initial data for the safety mAb and combined the scFv fragment technique with rAAV of using rAAV-S3C in gene therapy. vector–mediated gene transfer to achieve tumor suppression in We also showed that the i.m. injection of rAAV2/1-S3C virus in mice, providing a novel strategy with significant clinical prospect tumor xenografts led to stable serum secretion of S3C scFv for gene therapy. fragments that induced apoptosis of implanted tumors. Systemic gene delivery carries the risk of gene delivery to other tissues such Acknowledgments as liver. Therefore, it is highly desirable to achieve tissue-specific gene transfer and expression. In our study, we used the rAAV2/1 Received 4/4/2006; revised 9/8/2006; accepted 10/13/2006. Grant support: State High-tech Research and Development Program grants vector system, which has been shown to be more selective and safe 2003AA216092 and 2002AA216015, Natural Science Foundation of China grants for muscle transduction via i.m. administration (46). Transduction 30210103903 and 30571687, and State Key Basic Research Program of China. The costs of publication of this article were defrayed in part by the payment of page of rAAV2/1-S3C was confirmed by PCR amplification of a portion charges. This article must therefore be hereby marked advertisement in accordance of the WPRE fragment from genomic DNA isolated from muscle with 18 U.S.C. Section 1734 solely to indicate this fact.

References normal human hepatocytes by tumor necrosis factor- 16. Casey JL, Napier MP, King DJ, et al. Tumor targeting related apoptosis-inducing ligand. Nat Med 2000;6:564–7. of humanized cross-linked divalent-Fab¶ antibody frag- 1. French LE, Tschopp J. Protein-based therapeutic 9. Nitsch R, Bechmann I, Deisz RA, et al. Human brain- ments: a clinical phase I/II study. Br J Cancer 2002;86: approaches targeting death receptors. Cell Death Differ cell death induced by tumour-necrosis-factor-related 1401–10. 2003;10:117–23. apoptosis-inducing ligand (TRAIL). Lancet 2000;356: 17. Hudson PJ. Recombinant antibody constructs in 2. Sheridan JP, Marsters SA, Pitti RM, et al. Control of 827–8. cancer therapy. Curr Opin Immunol 1999;11:548–57. TRAIL-induced apoptosis by a family of signaling and 10. Mori E, Thomas M, Motoki K, et al. Human normal 18. Arndt MA, Krauss J, Vu BK, et al. A dimeric decoy receptors. Science 1997;277:818–21. hepatocytes are susceptible to apoptosis signal mediat- angiogenin immunofusion protein mediates selective 3. Chaudhary PM, Eby M, Jasmin A, Bookwalter A, ed by both TRAIL-R1 and TRAIL-R2. Cell Death Differ toxicity toward CD22+ tumor cells. J Immunother 2005; Murray J, Hood L. Death receptor 5, a new member of 2004;11:203–7. 28:245–51. the TNFR family, and DR4 induce FADD-dependent 11. Ichikawa K, Liu W, Zhao L, et al. Tumoricidal activity 19. Bermer E, Kuijlen J, Samplonius D, Walczak H, de Leij apoptosis and activate the NF-nB pathway. Immunity of a novel anti-human DR5 monoclonal antibody L, Helfrich W. Target cell-restricted and -enhanced 1997;7:821–30. without hepatocyte cytotoxicity. Nat Med 2001;7:954–60. apoptosis induction by a scFv: sTRAIL fusion protein 4. Kischkel FC, Lawrence DA, Chuntharapai A, Schow P, 12. Guo Y, Chen C, Zheng Y, et al. A novel anti-human with specificity for the pancarcinoma-associated anti- Kim KJ, Ashkenazi A. Apo2L/TRAIL-dependent recruit- DR5 monoclonal antibody with tumoricidal activity gen EGP2. Int J Cancer 2004;109:281–90. ment of endogenous FADD and caspase-8 to death induces caspase-dependent and caspase-independent 20. Goel A, Baranowska-Kortylewicz J, Hinrichs SH, et al. receptor 4 and 5. Immunity 2000;12:611–20. cell death. J Biol Chem 2005;280:41940–52. 99mTc-labeled divalent and tetravalent CC49 single- 5. Suliman A, Lam A, Datta R, Srivastava RK. Intracel- 13. Hudson PJ, Souriau C. Engineered antibodies. Nat chain Fvs: novel imaging agents for rapid in vivo lular mechanisms of TRAIL: apoptosis through mito- Med 2003;9:129–34. localization of human colon carcinoma. J Nucl Med chondrial-dependent and -independent pathways. 14. Jain RK, Baxter LT. Mechanisms of heterogeneous 2001;42:1519–27. Oncogene 2001;20:2122–33. distribution of monoclonal antibodies and other macro- 21. Le Gall F, Reusch U, Moldenhauer G, Little M, 6. Vincent H, Claire PP, Marion T, et al. Sensitivity of molecules in tumors: significance of elevated interstitial Kipriyanov SM. Immunosuppressive properties of anti- prostate cells to TRAIL-induced apoptosis increases pressure. Cancer Res 1988;48:7022–32. CD3 single-chain Fv and diabody. J Immunol Methods with tumor progression: DR5 and caspase 8 are key 15. Tahtis K, Lee FT, Smyth FE, et al. Biodistribution 2004;285:111–27. players. Prostate 2006;66:987–95. properties of indium-111-labeled C-functionalized 22. Lev A, Noy R, Oved K, et al. Tumor-specific Ab- 7. Walczak H, Miller RE, Ariail K, et al. Tumoricidal trans-cyclohexyldiethylenetriaminepentaacetic acid mediated targeting of MHC-peptide complexes induces activity of tumor necrosis factor-related apoptosis- humanized 3S193 diabody and F(ab¶)2 constructs in a regression of human tumor xenografts in vivo. Proc Natl inducing ligand in vivo. Nat Med 1999;5:157–63. breast carcinoma xenograft model. Clin Cancer Res Acad Sci U S A 2004;101:9051–6. 8. Jo M, Kim TH, Seol DW, et al. Apoptosis induced in 2001;7:1061–72. 23. Kuan Ct, Reist CJ, Foulon CF, et al. 125I-labeled

Cancer Res 2006; 66: (24). December 15, 2006 11952 www.aacrjournals.org

anti-epidermal growth factor receptor-vIII single-chain adeno-associated virus vectors with the woodchuck necrosis factor-related apoptosis-inducing ligand recep- Fv exhibits specific and high-level targeting of glioma hepatitis virus posttranscriptional regulatory element: tor 1 (TRAIL-R1) and receptor 2 (TRAIL-R2) genes in xenografts. Clin Cancer Res 1999;5:1539–49. implications for gene therapy. Hum Gene Ther 1999;10: metastatic breast cancers. Cancer Res 2001;61:4942–6. 24. Huang BC, Foote LJ, Lankford TK, Davern SM, 2295–305. 40. Zhang XD, Franco A, Myers K, Gray C, Nguyen T, McKeown CK, Kennel SJ. A diabody that dissociates to 32. Shi J, Zheng D, Liu Y, et al. Overexpression of soluble Hersey P. Relation of TNF-related apoptosis-inducing monomer forms at low concentration: effects on TRAIL induces apoptosis in human lung adenocarcino- ligand (TRAIL) receptor and FLICE- inhibitory protein binding activity and tumor targeting. Biochem Biophys ma and inhibits growth of tumor xenografts in nude expression to TRAIL-induced apoptosis of melanoma. Res Commun 2005;327:999–1005. mice. Cancer Res 2005;65:1687–92. Cancer Res 1999;59:2747–53. 25. Kortt AA, Dolezal O, Power BE, Hudson PJ. Dimeric 33. Chao H, Liu Y, Rabinowitz J, Li C, Samulski RJ, Walsh 41. Fulda S, Meyer E, Debatin KM. Inhibition of TRAIL- and trimeric antibodies: high avidity scFvs for cancer CE. Several log increase in therapeutic transgene induced apoptosis by Bcl-2 overexpression. Oncogene targeting. Biomol Eng 2001;18:95–108. delivery by distinct adeno-associated viral serotype 2002;21:2283–94. 26. Wu AM, Chen W, Raubitschek A. Tumor localization vectors. Mol Ther 2000;2:619–23. 42. Hietakangas V, Poukkula M, Heiskanen KM, Karvinen of anti-CEA single-chain Fvs: improved targeting by 34. Gao GP, Alvira MR, Wang L, Calcedo R, Johnston J, JT, Sistonen L, Eriksson JE. Erythroid differentiation non-covalent dimers. Immunotechnology 1996;2:21–36. Wilson JM. Novel adeno-associated viruses from rhesus sensitizes K562 leukemia cells to TRAIL-induced 27. Iliades P, Kortt AA, Hudson PJ. Triabodies: single monkeys as vectors for human gene therapy. Proc Natl apoptosis by downregulation of c-FLIP. Mol Cell Biol chain Fv fragments without a linker form trivalent Acad Sci U S A 2002;99:11854–9. 2003;23:1278–91. trimers. FEBS Lett 1997;409:437–41. 35. Spierings DC, de Vries EG, Timens W, Groen HJ, 43. Trauzold A, Wermann H, Arlt A, et al. CD95 and 28. Kortt AA, Lah M, Oddie GW, et al. Single-chain Fv Boezen HM, de Jong S. Expression of TRAIL and TRAIL TRAIL receptor-mediated activation of protein kinase C fragments of anti-neuraminidase antibody NC10 con- death receptors in stage III non-small cell lung cancer and NF-nB contributes to apoptosis resistance in ductal taining five- and ten-residue linkers form dimers and tumors. Clin Cancer Res 2003;9:3397–405. pancreatic adenocarcinoma cells. Oncogene 2001;20: with zero-residue linker a trimer. Protein Eng 1997;10: 36. Kurbanov BM, Geilen CC, Fecker LF, Orfanos CE, 4258–69. 423–33. Eberle J. Efficient TRAIL-R1/DR4-mediated apoptosis in 44. Song JH, Bellail A, Tse MC, Yong VW, Hao C. Human 29. Pei XY, Holliger P, Murzin AG, Williams RL. The 2.0-A melanoma cells by tumor necrosis factor-related astrocytes are resistant to Fas ligand and tumor necrosis resolution crystal structure of a trimeric antibody apoptosis-inducing ligand (TRAIL). J Invest Dermatol factor-related apoptosis-inducing ligand-induced apo- fragment with noncognate VH-VL domain pairs shows 2005;125:1010–9. ptosis. J Neurosci 2006;26:3299–308. a rearrangement of VH CDR3. Proc Natl Acad Sci U S A 37. Pai SI, Wu GS, Ozoren N, et al. Rare loss-of-function 45. Leverkus M, Neumann M, Mengling T, et al. 1997;94:9637–42. mutation of a death receptor gene in head and neck Regulation of tumor necrosis factor-related apoptosis- 30. During MJ, Symes CW, Lawlor PA, et al. An oral cancer. Cancer Res 1998;58:3513–8. inducing ligand sensitivity in primary and transformed vaccine against NMDAR1 with efficacy in experimental 38. Lee SH, Shin MS, Kim HS, et al. Alterations of the human keratinocytes. Cancer Res 2000;60:553–9. stroke and epilepsy. Science 2000;287:1453–60. DR5/TRAIL receptor 2 gene in non-small cell lung 46. Wang Z, Zhu T, Qiao C, et al. Adeno-associated virus 31. Loeb JE, Cordier WS, Harris ME, Weitzman MD, cancers. Cancer Res 1999;59:5683–6. serotype 8 efficiently delivers genes to muscle and heart. Hope TJ. Enhanced expression of transgenes from 39. Shin MS, Kim HS, Lee SH, et al. Mutations of tumor Nat Biotechnol 2005;23:321–8.

www.aacrjournals.org 11953 Cancer Res 2006; 66: (24). December 15, 2006

Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2006 American Association for Cancer Research. Therapeutic Expression of an Anti-Death Receptor 5 Single-Chain Fixed-Variable Region Prevents Tumor Growth in Mice

Juan Shi, Yanxin Liu, Yong Zheng, et al.

Cancer Res 2006;66:11946-11953.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/66/24/11946

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2006/12/15/66.24.11946.DC1

Cited articles This article cites 45 articles, 19 of which you can access for free at: http://cancerres.aacrjournals.org/content/66/24/11946.full#ref-list-1

Citing articles This article has been cited by 2 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/66/24/11946.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/66/24/11946. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.