[CANCER RESEARCH 64, 46994702, July 15, 2004]

Advances in Brief

Impaired Angiogenesis, Delayed Wound Healing and Retarded Tumor Growth inPerlecan Heparan Sulfate-Deficient Mice

Zhongjun Zhou,1,2 Jianming Wang,2,4 Renhai Cao,3 Hiroyuki Morita,2 Raija Soininen,2,5 Kui Ming Chan,1Baohua Liu,1 Yihai Cao,3 and Karl Tryggvason21Department of Biochemistry, Faculty of Medicine, University of Hong Kong, Hong Kong; 2Division of Matrix Biology, Department of Medical Biochemistry and Biophysics, and3Microbiology and Tumorbiology Center, Karolinska Institute, Stockholm, Sweden; 4Institute of Molecular Medicine, Department of Medicine, University of California at SanDiego, La Jolla, California; and 5Department of Medical Biochemistry and Molecular Biology, University of Oulu, Oulu, Finland

Abstract

Perlecan, a modular proteoglycan carrying primary heparan sulfate(HS) side chains, is a major component of blood vessel basement mem-branes. It sequesters growth factors such as fibroblast growth factor 2(FGF-2) and regulates the ligand-receptor interactions on the cell surface,and thus it has been implicated in the control of angiogenesis. Bothstimulatory and inhibitory effects of perlecan on FGF-2 signaling havebeen reported. To understand the in vivo function of HS carried byperlecan, the perlecan gene heparan sulfate proteoglycan 2 (Hspg2) wasmutated in mouse by gene targeting. The HS at the NH2 terminus ofperlecan was removed while the core protein remained intact. PerlecanHS-deficient (Hspg23/3) mice survived embryonic development andwere apparently healthy as adults. However, mutant mice exhibited sig-nificantly delayed wound healing, retarded FGF-2-induced tumor growth,and defective angiogenesis. In the mouse corneal angiogenesis model,FGF-2-induced neovascularization was significantly impaired inHspg23/3 mutant mice. Our results suggest that HS in perlecan posi-tively regulates the angiogenesis in vivo.

Introduction

Basic fibroblast growth factor (FGF-2) is a member of a largefamily of growth factors that regulate the growth and differentiation ofa number of cell types. FGF-2 is a key regulator of angiogenesis andis involved in a number of pathophysiological processes such aswound repair and tumor growth (1). The binding of FGF-2 to itsreceptors requires the participation of heparan sulfate (HS) or heparin(1). HS proteoglycans (HSPGs) are a group of modular proteins thatcarry HS side chains. According to their relationship with cells,HSPGs are divided into two major groups, membrane-associatedHSPGs (syndecans 14 and glypican 16) and extracellular HSPGs(perlecan and agrin; Ref. 2). Studies of perlecan expression in micerevealed its early expression in the heart primordium and large bloodvessels and subsequently in a number of mesenchymal tissues (3). Thefact that expression of perlecan is always prominent in the endothelialcell basement membranes of all vascularized organs (4) indicates itscrucial role in vasculogenesis and angiogenesis. Perlecan is not onlya structural protein, but is also able to bind a number of growth factorsand modulate their activities (4, 5). The expression pattern of perlecanparallels the distribution of FGF-2 in various basement membranes,suggesting its involvement in FGF-2 signaling (4). Although many

studies demonstrated that the perlecan is a key regulator in ligand-receptor binding, the in vivo role for HS in perlecan molecules duringangiogenesis has not yet been studied.

The mouse Hspg2 gene encodes the large perlecan core protein.Hspg2 null mice were embryonic lethal, due to the disruption ofbasement membrane structure (5). To study the potential in vivo roleof perlecan HS side chains, a 45-bp in-frame deletion was made to theHspg2 gene to remove all three HS chain attachment sites from thedomain I, resulting in the HS chains deficiency in a specific HSPG-perlecan HS deficiency. Hspg23/3 mice survived embryonic devel-opment. Adult mice appear healthy and fertile and are grossly indis-tinguishable from their littermate controls. However, a carefulexamination showed that Hspg23/3 mice had defects in lens devel-opment (6). The Hspg23/3 mice provided the unique model to studythe in vivo roles for specific HSPG HS chains in FGF-2 signaling andangiogenesis. The purpose of this study was to address the followingquestions: (a) Is angiogenesis in physiological and pathological con-ditions affected in Hspg23/3 mice? (b) What is the role for perlecanHS in FGF-2-induced angiogenesis? Here, we show that deficiency inperlecan HS does not affect the structure or composition of the bloodvessel basement membranes. We found that perlecan HS promotedangiogenesis in vivo, because the removal of perlecan HS side chainsled to impaired FGF-2-mediated angiogenesis.

Materials and Methods

Animals. The generation and characterization of Hspg23/3 mice has beendescribed elsewhere (6). In brief, exon 3 of mouse perlecan gene, Hspg2,which is 45 bp in size, was removed in-frame. After homologous recombina-tion in embryonic stem cells, the mutated allele was transmitted through thegerm line. Heterozygous mutant mice were back-crossed to C57BL/6 mice forseven generations. Homozygous mutants (Hspg23/3) were obtained throughmating between heterozygous animals. All animal experiments have beenapproved by the Stockholm Animal Ethics Committee.

Skin Wound-Healing Assay. The skin wound-healing assay was per-formed essentially as described previously (7). Two 8-mm full-thickness punchwounds were created on the dorsal surface of each of the 6-week-old mice witha dermal biopsy punch (Miltex GmbH, Ludwigshafen, Germany). Twenty-fourh after the surgery, the wounds were recorded as initial condition. The woundswere then recorded daily on transparent parafilm for 9 days. All recordingswere scanned, and the wound areas were calculated with Scion Image soft-ware.

Tumor Experiments. K1000, the NIH3T3 cells stably transfected withFGF-2 cDNA with a signal peptide (8), were maintained in DMEM suppliedwith 10% fetal bovine serum and 100 IU/ml penicillin-streptomycin (LifeTechnologies, Inc.). Before inoculation, cells were trypsinized, washed, andsuspended in PBS. Cells of 5 106 were inoculated s.c. on the backs ofnewborn mice. Tumors were dissected and weighed at designated days afterthe inoculation.

Cornea Micropocket Assay. The corneal micropocket assay was per-formed as described previously (9). In brief, a corneal micropocket was createdon one eye of each of the 6-week-old Hspg23/3 mice and controls. A

Received 3/5/04; revised 5/4/04; accepted 5/28/04.Grant support: University of Hong Kong Basic Research Seeds Fund, Swedish

Cancer Foundation, Novo-Nordisk Foundation, the Swedish Medical Council, and EUGrant QLGI-CT-01131. J. Wang was a recipient of a postdoctoral fellowship fromKarolinska Institute.

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

Requests for reprints: Zhongjun Zhou, Department of Biochemistry, Faculty ofMedicine, University of Hong Kong, 21 Sassoon Road, Hong Kong. Phone: 852-28199542; Fax: 852-28551254; E-mail: zhongjun@hkucc.hku.hk.

4699

Research. on February 24, 2015. 2004 American Association for Cancercancerres.aacrjournals.org Downloaded from

micropellet containing 80 ng of human recombinant FGF-2 (Invitrogen) wasimplanted into each pocket. Corneal neovascularization was examined 5 daysafter the operation.

Immunohistochemistry. Immunohistochemical staining was performedon cryostat sections. Fresh tumor samples and skin wound samples werecollected and snap-frozen in OCT compound (Sakura, Finetek, the Nether-lands). Cryostat sections (10 m thick) were fixed in 1% paraformaldehydeand stained with biotin antimouse CD31 (PharMingen) or rat monoclonalantibodies against perlecan (NeoMarkers), nidogen (NeoMarkers), laminin 4[a kind gift from Rupert Timpl (Max-Planck Institute for Biochemistry, Mar-tinsried, Germany)], and collagen IV 1 chain.

Measurement of Blood Vessel Density. To measure the blood vesseldensity, tumors of comparable size from control and Hspg23/3 mice wereused. Tumors from control mice were dissected 7 days after the inoculation,and tumors from Hspg23/3 mice were dissected 10 days after the inoculation.Cryostat sections (10 m thick) were immunostained with CD31 antibody.Five microscopic fields were randomly selected from each tumor section andrecorded with a CCD camera. In wound-healing experiments, skin woundsamples were harvested at day 5, and the sections were counterstained withhematoxyline after CD31 immunostaining. Two or three microscopic fields ofgranulation tissue were chosen from each section. The number of blood vesselswas counted in sections of three animals from each group, and the density wascalculated using Scion Image software.

Cell Surface Binding of 125I-FGF-2. Binding of 125I-FGF-2 to murineembryonic fibroblasts (MEF) cells was determined as described previously(10), with the following modifications. MEF cells were seeded at a density of2 105/well in 48-well plates and cultured until confluence in full medium.After washing with 1 PBS, cells were then incubated for an additional 24 hin serum-free DMEM for conditioned medium collection. A binding solutionwas prepared by mixing DMEM medium with various concentrations of125I-FGF-2 and cooling to 4C for 10 min. Confluent MEF cells were washedthree times with ice-cold DMEM. The binding solution was added to the cellsand incubated at 4C for 2.5 h. The binding medium was then removed, and thecells were gently washed three times with ice-cold DMEM. The FGF-2 boundto cell surface was extracted with ice-cold 2 M NaCl/20 mM sodium acetate (pH4.0) for 15 min, and the salt extracts were counted in a counter. Nonspecificbinding was considered as the value obtained in the presence of a 100-foldexcess of nonradioactive FGF-2. The value was then normalized by the cellnumber.

Statistics. Data were analyzed with the two-tailed Students t test.

Results

Delayed Wound Healing in Hspg23/3 Mice. Angiogenesis is anessential step in the wound-healing process. New blood vessels arerequired for the supply of oxygen and nutrients, for wound debrisremoval, and for granulation formation. FGF-2-mediated angiogene-sis plays critical role in wound healing (1). To investigate the role ofperlecan HS in physiological angiogenesis, we examined wound re-pair in mice. Six-week-old Hspg23/3 mice and their sex- and age-matched wild-type control mice from the same background were usedin this study. Two full skin thickness punch wounds were created onthe backs of these mice (Fig. 1A). Twenty-four h later, the wound areawas first recorded and regarded as the initial wound. The wound areawas then recorded daily, and the healing process was quantified as thepercentage of the initial wound area. The wound-healing process wassignificantly delayed in mutant mice (Fig. 1, A and B), with significantdifferences found between days 3 and 9 (Fig. 1B). From day 3 aftersurgery, the formation of granulation at the edge of the wound bedwas observed both in wild-type and in Hspg23/3 mice. However, theextent of development of granulation tissue was less in Hspg23/3mice and appeared to be poorly vascularized (Fig. 1C). To quantifythe vascularity, wound tissue was sampled at day 5 after the surgery,and sections were immunostained with the endothelial marker CD31.The number of blood vessels was counted. We found that blood vesseldensity in granulation tissue of Hspg23/3 mice was significantlylower (182 21/mm2, mean SE) in comparison with that inwild-type controls (287 25/mm2, mean SE, P 0.001; Fig. 1D).No obvious difference in the gross vascular structure was observedbetween mutants and controls.

Impaired FGF-2-Induced Tumor Angiogenesis and TumorGrowth in Hspg23/3 Mice. Tumor growth is also angiogenesisdependent. We therefore studied tumor growth and tumor angiogen-esis in Hspg23/3 mice to further address the role of perlecan HS inangiogenesis. Basic FGF-transformed cells K1000 were used in this

Fig. 1. Delayed wound healing, reduced granulation tissue formation, and vascular density in Hspg23/3 mice. Two punch wounds were created on the back of each Hspg23/3and control mouse. A, skin wounds in Hspg23/3 and sex- and age-matched control mice 4 days after surgery, showing the difference in wound sizes. B, quantification of thewound-healing process. Wound areas were measure from days 2 to 9 in seven Hspg23/3 and nine control mice (mean SD). The wound size at 24 h after surgery was set as initialwound (100%). , P 0.05; , P 0.01; , P 0.001. C, CD31 immunostaining, showing granulation tissue formation in the wound bed 5 days after surgery. D, quantificationof blood vessel density in granulation tissue. The number of blood vessels was counted in three Hspg23/3 and three control mice (mean SE); , P 0.01.

4700

PERLECAN AND ANGIOGENESIS

Research. on February 24, 2015. 2004 American Association for Cancercancerres.aacrjournals.org Downloaded from

experiment. K1000 cells of 5 106 were inoculated s.c. on the backsof newborn Hspg23/3 and control mice. Ten days after inoculation,tumors were dissected and weighed. Eight litters of newborn mice,amounting to 21 mutant and 37 control individuals, were used in thisstudy. The tumor mass from Hspg23/3 mice weighed 0.47 0.10 g(mean SE) and was significantly less than that of their littermatecontrols (0.79 0.17 g, mean SE, P 0.01; Fig. 2A). To elucidatethe retarded tumor growth in Hspg23/3 mice, we evaluated bloodvessel formation in tumors. Tumors from control mice were sampledat day 7, and tumors from mutant mice were sampled at day 10, so thetumors from controls and mutants were comparable in size. Tumorsections were stained with CD31 (Fig. 2, C and D), and the numbersof blood vessels were counted in three mutants and three controls,using Scion Image software. Tumors from mutant mice showed a

significantly reduced blood vessel density of 98 14/mm2, comparedwith 142 30/mm2 in controls (mean SE, P 0.05; Fig. 2, BD).

Perlecan HS Depletion Did Not Alter the Integrity of the BloodVessel Basement Membranes. The integrity of the blood vesselbasement membranes may also affect angiogenesis (11). We thereforecarefully examined the morphology of blood vessels in tumor tissuesand in granulation tissues. No blood vessel deformities, i.e., dilation,constriction, or signs of blood cell leakage, were observed in tumorsfrom Hspg23/3 mice. To examine whether the loss of perlecan HSaltered the composition of major basement membrane components inblood vessels, we immunostained tumor sections with antibodiesagainst perlecan (Fig. 2, E and F), laminin 4 (Fig. 2, G and H),collagen IV (Fig. 2, I and J), and nidogen (data not shown). Noobvious differences in signal intensities or staining patterns wereobserved between Hspg23/3 and control mice.

Impaired FGF-2-Induced Corneal Angiogenesis in Hspg23/3Mice. To further investigate the impaired angiogenesis in Hspg23/3mice, we used the corneal micropocket assay to study FGF-2-inducedangiogenesis in vivo. An FGF-2-containing pellet was implanted intothe cornea of each 6-week-old mouse. Three animals of each genotypegroup (Hspg23/3, heterozygous, and control) were used in theseexperiments. Five days after pellet implantation, blood vessel forma-tion was evident in all mice. No differences were found betweenwild-type and heterozygous mice. However, all Hspg23/3 miceshowed a severely impaired angiogenesis (Fig. 3).

Reduced FGF-2 Binding to Perlecan HS-Deficient Cells. Tounderstand the mechanism of the impaired angiogenesis observed inHspg23/3 mice, we isolated MEFs from Hspg23/3 mice andwild-type mice. The role of perlecan HS in the ligand-receptor inter-action was investigated in vitro. 125I-FGF-2 of different concentra-tions was incubated for 2.5 h at 4C with wild-type MEFs andHspg23/3 MEFs. After washing, the cell surface-bound 125I-FGF-2was extracted and measured by counter. At the range of 520 ng/ml,wild-type MEFs exhibited a significant higher binding efficiency thanHspg23/3 cells (P 0.05), suggesting that in wild-type cells, theremaining wild-type perlecan HS on cell surface facilitates the bind-ing of growth factor to cell surface (Fig. 4). These results indicatedthat perlecan with HS, but not the perlecan core protein, couldfacilitate the growth factor-receptor binding and promote cellularresponse. Loss of HS side chains in the perlecan protein compromisedthe ligand-receptor interaction in Hspg23/3 cells.

Discussion

Genetic approaches have been applied to understand the in vivoroles of HSPGs. Mouse models lacking functional HS chains havebeen generated through the disruption of enzymes involved in HSbiosynthesis (12). Diverse phenotypes were observed in theseanimal models with limited consistency. Knockout mice of severalmembrane-associated HSPGs have also been produced. These

Fig. 3. Corneal micropocket assay. Impaired FGF-2-induced angiogenesis in aHspg23/3 mouse compared with that in a wild-type control mouse at day 5 after growthfactor implantation.

Fig. 2. Tumor growth in newborn Hspg23/3 and control mice. A, tumor weight from21 Hspg23/3 and 37 littermate control mice (mean SE); , P 0.01. B, quantifi-cation of vascular density in K1000 tumors from three Hspg23/3 and three littermatecontrols (mean SE); , P 0.05. CD31 immunostaining of K1000 tumor sectionsshowing the vascular density in tumors of similar size from a Hspg23/3 mouse at day10 (C) and a littermate control at day 7 (D). Immunostaining of tumor sections withantibodies against perlecan (E and F), laminin 4 (G and H), and collagen IV 1 chain(I and J). Perlecan is prominent in endothelial basement membranes. Vascular morphol-ogy and the intensity of staining were comparable between Hspg23/3 mice and controls.A reduction in vascular density was found in Hspg23/3 mice.

4701

PERLECAN AND ANGIOGENESIS

Research. on February 24, 2015. 2004 American Association for Cancercancerres.aacrjournals.org Downloaded from

knockout mice provide valuable models to understand the functionof a specific HSPG in signaling (2). A complete knockout ofperlecan core protein caused embryonic lethality due to the ruptureof basement membranes in the heart and abnormal cartilage devel-opment, suggesting a critical role for this protein in the mainte-nance of cartilage and basement membrane integrity (6, 13). Be-cause the signaling functions of HSPGs are mediated by their HSchains (14), mice lacking HS side chains while maintaining thestructural role of perlecan core protein provide an unique model tostudy the signaling function of HS chains of a specific HSPG invivo. We have shown in this study that perlecan HS played animportant role in the regulation of wound healing and tumorgrowth. Loss of perlecan HS resulted in impaired angiogenesis,delayed wound healing and retarded tumor growth. Our data sug-gest that perlecan regulates FGF-2 signaling and angiogenesisthrough its HS side chains. To our knowledge, this is the firstanimal model demonstrating the in vivo function of HS side chainsin a specific HSPG.

A number of studies have been carried out to understand thebiochemical and biological roles of perlecan in growth factor signal-ing (1417). Several functions of perlecan HS have been proposedbased on in vitro studies: (a) Perlecan HS sequesters and traps growthfactors, protecting them from proteolytic degradation; and (b) perle-can HS binds growth factors by acting as a low-affinity receptor andmediates high-affinity growth factor-receptor binding. Data from ourin vitro binding support the notion that although a proportion ofperlecan is secreted into the culture medium (17), a significant amountof perlecan attached to cell surfaces (18, 19). Cell surface perlecanparticipated directly or indirectly in the presentation of growth factorsto its high-affinity receptor. Both the ligand trapping and presentingfunctions of perlecan seemed to be carried out through its HS sidechains.

Based on our and others results, it is likely that the major function

of perlecan HS in the extracellular matrix is to bind FGF-2 and othergrowth factors, protect them from proteolytic degradation, and storethe ligands in a dormant state. Perlecan on the cell surface maydirectly or indirectly present the growth factors to their receptor andfacilitate the growth factor signaling. Perlecan without its HS chainsfailed to trap growth factors at the extracellular matrix of endothelialbasement membranes and was unable to protect or present growthfactors efficiently to the cells surface or higher affinity receptors.Therefore, impaired angiogenesis in the mutant mice was expected.Our results also indicate that the other membrane-associated HSPGsmay also be important in the presentation of growth factors to theirreceptors. Lack of perlecan HS side chains did not result in severedevelopmental defects in neuronal and vascular tissues or total loss ofmitotic response to FGF-2, indicating the functional redundancy in theregulation of growth factors receptors binding by HSPGs.

Acknowledgments

We thank Maria Laisi and Dadi Niu for excellent technical support. We alsothank Neil Portwood for proofreading of the manuscript.

References1. Nugent MA, Iozzo RV. Fibroblast growth factor-2. Int J Biochem Cell Biol 2000;

32:11520.2. Perrimon N, Bernfield M. Specificities of heparan sulphate proteoglycans in devel-

opmental processes. Nature 2000;404:7258.3. Handler M, Yurchenco PD, Iozzo RV. Developmental expression of perlecan during

murine embryogenesis. Dev Dyn 1997;210:13045.4. Iozzo RV. Matrix proteoglycans: from molecular design to cellular function. Annu

Rev Biochem 1998;67:60952.5. Costell M, Gustafsson E, Aszodi A, et al. Perlecan maintains the integrity of cartilage

and some basement membranes. J Cell Biol 1999;147:110922.6. Rossi M, Morita H, Sormunen R, et al. Heparan sulfate chains of perlecan are

indispensable in the lens capsule but not in the kidney. EMBO J 2003;22:23645.7. Malinda KM, Sidhu GS, Mani H, et al. Thymosin 4 accelerates wound healing.

J Invest Dermatol 1999;113:3648.8. Soutter AD, Nguyen M, Watanabe H, Folkman J. Basic fibroblast growth factor

secreted by an animal tumor is detectable in urine. Cancer Res 1993;53:52979.9. Zhou Z, Apte SS, Soininen R, et al. Impaired endochondral ossification and angio-

genesis in mice deficient in membrane-type matrix metalloproteinase I. Proc NatlAcad Sci USA 2000;97:40527.

10. Yayon A, Klagsbrun M, Esko JD, Leder P, Ornitz DM. Cell surface, heparin-likemolecules are required for binding of basic fibroblast growth factor to its high affinityreceptor. Cell 1991;64:8418.

11. Thyboll J, Kortesmaa J, Cao R, et al. Deletion of the laminin 4 chain leads toimpaired microvessel maturation. Mol Cell Biol 2002;22:1194202.

12. Forsberg E, Kjellen L. Heparan sulfate: lessons from knockout mice. J Clin Investig2001;108:17580.

13. Arikawa-Hirasawa E, Watanabe H, Takami H, Hassell JR, Yamada Y. Perlecan isessential for cartilage and cephalic development. Nat Genet 1999;23:3548.

14. Aviezer D, Hecht D, Safran M, Eisinger M, David G, Yayon A. Perlecan, basallamina proteoglycan, promotes basic fibroblast growth factor-receptor binding, mi-togenesis, and angiogenesis. Cell 1994;79:100513.

15. Nugent MA, Karnovsky MJ, Edelman ER. Vascular cell-derived heparan sulfateshows coupled inhibition of basic fibroblast growth factor binding and mitogenesis invascular smooth muscle cells. Circ Res 1993;73:105160.

16. Koyama N, Kinsella MG, Wight TN, Hedin U, Clowes AW. Heparan sulfate pro-teoglycans mediate a potent inhibitory signal for migration of vascular smooth musclecells. Circ Res 1998;83:30513.

17. Forsten KE, Courant NA, Nugent MA. Endothelial proteoglycans inhibit bFGFbinding and mitogenesis. J Cell Physiol 1997;172:20920.

18. Hayashi K, Madri JA, Yurchenco PD. Endothelial cells interact with the core proteinof basement membrane perlecan through 1 and 3 integrins: an adhesion modu-lated by glycosaminoglycan. J Cell Biol 1992;119:94559.

19. Tran PK, Tran-Lundmark K, Soininen R, et al. Increased intimal hyperplasia andsmooth muscle cell proliferation in transgenic mice with heparan sulfate-deficientperlecan. Circ Res 2004;94:5508.

Fig. 4. Effects of perlecan HS on FGF-2 binding. MEF cell surface-bound 125I-FGF-2increased along with the increase in the concentration of 125I-FGF-2. A significantlyhigher mount of 125I-FGF-2 bound to wild-type MEFs than to Hspg23/3 MEFs (,P 0.05; , P 0.01).

4702

PERLECAN AND ANGIOGENESIS

Research. on February 24, 2015. 2004 American Association for Cancercancerres.aacrjournals.org Downloaded from

2004;64:4699-4702. Cancer Res

Zhongjun Zhou, Jianming Wang, Renhai Cao, et al.

Sulfate-Deficient MiceRetarded Tumor Growth in Perlecan Heparan Impaired Angiogenesis, Delayed Wound Healing and

Updated version

http://cancerres.aacrjournals.org/content/64/14/4699Access the most recent version of this article at:

Cited Articles

http://cancerres.aacrjournals.org/content/64/14/4699.full.html#ref-list-1This article cites by 19 articles, 9 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/64/14/4699.full.html#related-urlsThis article has been cited by 18 HighWire-hosted articles. Access the articles at:

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

SubscriptionsReprints and

.pubs@aacr.orgDepartment atTo order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

.permissions@aacr.orgDepartment atTo request permission to re-use all or part of this article, contact the AACR Publications

Research. on February 24, 2015. 2004 American Association for Cancercancerres.aacrjournals.org Downloaded from