Efficient Infection of Tumor Endothelial Cells by a Capsid-Modified Adenovirus

Gene Therapy (2006) 13, 52–59 & 2006 Nature Publishing Group All rights reserved 0969-7128/06 $30.00 www.nature.com/gt ORIGINAL ARTICLE Efﬁcient infection of tumor endothelial cells by a capsid-modiﬁed adenovirus

K Shinozaki1, E Suominen2,3, F Carrick1, B Sauter1, V-M Ka¨ha¨ri2,3,4, A Lieber5, SLC Woo1 and M Savontaus2,3,6 1Carl C. Icahn Center for Gene Therapy and Molecular Medicine, Mount Sinai School of Medicine, New York, NY, USA; 2Turku Centre for Biotechnology, University of Turku and A˚ bo Akademi University, Turku, Finland; 3Department of Medical Biochemistry and Molecular Biology, University of Turku, Turku, Finland; 4Department of Dermatology, University of Turku, Turku, Finland; 5Division of Medical Genetics, School of Medicine, University of Washington, Seattle, WA, USA; and 6Department of Medicine, University of Turku, Turku, Finland

Targeted antiangiogenic gene therapy is an attractive demonstrated that infection of HUVECs and HAECs with approach to treat metastatic cancer. However, the relative Ad5/35 resulted in between 1 and 3 orders of magnitude paucity of the receptors of the commonly used adenovirus higher gene expression than infection with Ad5. Furthermore, serotype 5 in endothelial cells as compared with liver cells various liver-derived cells were less infectable with Ad5/35 undermines the use of this vector for targeting the endothelial than Ad5, indicating a favorable toxicity profile for this virus. cells in tumors. To overcome this problem, we analyzed the In a rat colon carcinoma tumor model, Ad5 was located ability of a hybrid Ad5/35 virus, where the serotype 5 fiber mainly in the liver parenchyma after hepatic artery adminis- has been replaced with the fiber from serotype 35, to target tration. In contrast, Ad5/35 was found only in the angiogen- tumor vasculature. Infection of human umbilical vein esis-rich border region of the tumor. Double immunostaining endothelial cells (HUVECs) with Ad5/35 at MOI 120 infected revealed that Ad5/35 colocalized with CD31 and Flk-1 100% of cells. In contrast, infection with Ad5 at the same positive endothelial cells. These results indicate that Ad5/ MOI infected only 10% HUVECs. Ad5/35 was even more 35 may be useful in anticancer strategies targeting tumor effective in transducing human aortic endothelial cells endothelial cells. (HAECs), as infection with Ad5/35 at MOI 3.6 was sufficient Gene Therapy (2006) 13, 52–59. doi:10.1038/sj.gt.3302598; to transduce 95% of cells. Gene expression analyses published online 18 August 2005

Keywords: hybrid adenovirus; endothelial targeting; tumor vasculature; cancer treatment

Introduction has been shown in a clinical trial for head and neck cancer.6 Easy production in high titers, the broad It is now well recognized that the ability of a tumor spectrum of infectivity of both resting and proliferating to grow and metastasize is dependent on its ability cells along with its highly efficient gene delivery capacity to form new blood vessels, hence providing a rationale are considered to be advantages of this vector. However, for the development of antiangiogenic therapy for the main limitation undermining the practical use of this cancer.1 Targeted destruction of the endothelial cell vector stems from its mechanism of cell entry. The population in tumor vessels is an attractive approach primary cellular receptor for adenovirus serotype 5, to halt tumor growth, as tumor endothelial cells are which is currently used in most applications, is the genetically stable and easily accessible through the coxsackie-adenovirus receptor (CAR). Adenovirus binds bloodstream. The feasibility of this approach has recently to this receptor via its fiber molecule, thereby allowing it been demonstrated by delivering cytotoxic and anti- to attach itself to the cell surface. Attached viruses are angiogenic genes specifically to the tumor endothelium internalized by avb3 and avb5 integrins, which function using cationic nanoparticles and endothelial progenitor- as secondary adenovirus receptors. Although some varia- like cells.2,3 tion exists between species, CAR is expressed in many Adenovirus is currently the most widely used vector normal cell types, including liver, heart, brain, kidney for cancer gene therapy and proof-of-concept using this and lung.7,8 Limited data exist on the expression of the vector has been demonstrated in numerous preclinical CAR receptor in endothelial cells in vivo; nevertheless, it applications.4,5 In addition, a level of therapeutic efficacy has been shown that gene transfer by the standard serotype 5 adenovirus to endothelial cells is inefficient, indicating that this virus has an unfavorable tropism Correspondence: Dr M Savontaus, Turku Centre for Biotechnology, profile for antiangiogenic cancer gene therapy.9,10 In University of Turku, Tykisto¨katu 6B, FIN-20520 Turku, Finland. E-mail: [email protected].fi addition, Chillon and co-workers have demonstrated Received 25 February 2004; revised 27 February 2005; accepted 8 that several other adenovirus serotypes also transfer July 2005; published online 18 August 2005 primary endothelial cells poorly in vitro, with the Endothelial targeting by a hybrid virus K Shinozaki et al 53 exception of serotypes 3, 17 and 30. However, serotype 35 was used to determine the expression levels of CD46, 11 was not included in this analysis. CAR and av integrins for the human cell lines used in the Several approaches have been used in an attempt to study (Figure 1). FACS analyses demonstrated that redirect adenoviral tropism away from CAR. The CAR Hep3B and HepG2 liver cells express high levels of all binding site in the fiber protein has been masked by three adenovirus receptors. In contrast, HUVECs and bispecific antibodies directed against the fiber and a HAECs expressed high levels of CD46 and av integrins, receptor expressed in the target cell population12,13 or by but had very low CAR expression on their cell surface, polyethylene glycol treatment of virus particles.14 These suggesting that viral transduction through CD46, but two-component approaches suffer from potential in- not CAR, would result in efficient transduction to stability as well as having a limited use in retargeting endothelial cells. replication-competent adenoviruses. Retargeting strategies by modification of the viral genome include deleting In vitro transduction of Ad5 and Ad5/35 the binding sites for CAR and avb3/avb5 integrins and In order to analyze the transduction efficiency of Ad5/35 inserting short peptide ligands with specificity for target to endothelial cells, a series of in vitro transduction assays cells.15–19 These approaches have produced somewhat was performed (Figure 2). We chose several liver cell conflicting results with regard to liver transduction. lines as controls in these experiments, since it has Some papers have reported significant reduction in liver previously been demonstrated that after systemic admin- transduction using vectors deleted for both CAR and istration, the majority of a nontargeted adenovirus will integrin binding sites, while others demonstrate that be taken up by liver, potentially causing significant similar deletions have no effect in detargeting the liver. toxicity.29 HUVECs and HAECs as well as BRL, Hep3B These discrepancies have been suggested to be caused and HepG2 liver cells were infected with Ad5/35.LacZ by the differences in the type of the deletion mutation as and Ad5.LacZ with MOI 120 (HUVEC, Hep3B, HepG2 well as in the route of administration of the virus. and BRL) or MOI 12 (HAEC) and infected cells were Another approach is to use adenoviruses with hybrid detected by b-galactosidase staining (Figure 2a). The serotypes by swapping the fiber gene from one serotype lower MOI for HAECs was chosen because Ad5/35.LacZ to another. There are at least 51 serotypes of human infects HAECs so efficiently that infection with MOI 120 adenoviruses, which are divided into six subgroups resulted in the death of all Ad5/35.LacZ-infected cells (A–F). Native viruses in different subgroups have (data not shown). Infection of HUVECs with MOI 120 or different tissue tropisms and it has been demonstrated HAECs with MOI 12 of Ad5/35.LacZ resulted in reporter that at least viruses in subgroup B, which includes gene expression in 100% of cells. In contrast, infection serotype 35, do not use CAR as a physiologically relevant of HUVECs and HAECs with Ad5.LacZ with the same receptor and the receptor for most subgroup B adeno- MOIs resulted in transduction of very few cells. All liver viruses was recently discovered to be CD46, a comple- cells were clearly less transducable with Ad5/35.LacZ ment regulatory protein with multiple isoforms.20–22 This than Ad5.LacZ, the difference being most obvious in BRL suggests that fiber-swapping can be utilized to redirect rat liver cells. Quantification of infected cells with MOIs adenovirus tropism and the feasibility of this approach ranging from 36 to 360 essentially corroborated the above has been successfully demonstrated for several hybrid findings (Figure 2b). Again, for HAECs, MOIs ranging serotypes including Ad5/3, Ad5/7, Ad5/16, Ad5/35 from 36 to 3.6 are shown due to the high infectivity of and Ad2/17. Using this strategy, hybrid adenoviruses Ad5/35.LacZ for these cells. Ad5/35.LacZ infected 100% have shown favorable tropism profiles for several cell of HUVECs with MOI 120 whereas Ad5.LacZ infected types, including ovarian and breast cancer cells, hema- less than 10% of the cells with the same MOI. For topoetic stem cells, dendritic cells, HUVECs, smooth HAECs, infection with Ad5/35.LacZ even with MOI 3.6 muscle cells and neurons.11,23–28 transduced 95% of cells whereas Ad5.LacZ did not In an attempt to develop a specific gene delivery transduce any HAECs at this MOI. BRL cells were system for the tumor endothelium, we have analyzed the untransducable by Ad5/35.LacZ even with MOI 360, ability of the recently developed Ad5/35 hybrid virus to whereas Ad5.LacZ efficiently transduced these cells at transduce endothelial cells in vitro as well as in a rat MOI 36. Hep3B and HepG2 cells were also transduced hepatocellular carcinoma (HCC) model. more efficiently with Ad5.LacZ than Ad5/35.LacZ, but the difference between the two viruses in these cell types was less clear, especially at the highest MOI, where no Results difference in viral transduction efficiency was observed. b-Galactosidase enzyme activity detection was used Characterization of target cells for quantitative measurement of viral transgene expres- The goal of this paper was to analyze the transduction sion (Figure 2c). HUVECs infected with Ad5/35.LacZ at efficiency of the recently described Ad5/35 vector to MOI 120 gave rise to more than 11-fold increase in endothelial cells. In the Ad5/35 system, the fiber shaft transgene expression when compared to the standard and knob domains of the standard adenovirus serotype Ad5.LacZ. In HAECs the difference between the two 5 have been replaced by the shaft and knob domains viruses was even higher, infection with Ad5/35.LacZ of serotype 35. This hybrid virus binds to a cellular and Ad5.LAcZ resulted in 89.5 and 0.01 mU/mg trans- receptor, which was recently identified as CD46, result- gene expression at MOI 36, respectively. In contrast, ing in a different tropism as compared with Ad5.20 In infection of BRL cells by Ad5/35.LacZ resulted in gene order to compare endothelial cell transduction of Ad5/35 expression barely above background levels even at MOI with Ad5 in vitro, human HUVECs and human aortic 360, whereas Ad5.LacZ caused 52-fold more b-galactosi- endothelial cells (HAECs) as well as three liver cell lines dase expression with the same MOI. Ad5.LacZ was also (Hep3B, HepG2 and BRL) were chosen. Flow cytometry more efficient in infecting Hep3B and HepG2 cells than

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 54

Figure 1 Expression of adenovirus receptors on endothelial and liver cells. HAEC, HUVEC, Hep3B and HepG2 cells were incubated with

monoclonal antibody against CAR (RmcB; 1:100 dilution for HUVECs; 1:50 dilution for Hep3Bs and HepG2s), av integrin (L230; 1:50 dilution) or CD46 (E4.3; 1:200 dilution). Control samples were incubated with an isotype speciﬁc control antibody. The binding of primary antibodies was developed with FITC-conjugated secondary antibody and analyzed by ﬂow cytometry. Cells stained with the isotype control antibody were used to set a background level of 1%. Values under the markers represent the percentage of positive cells.

Ad5/35, with up to 2.2-fold higher transgene expression administration of increasing doses of both viruses and after infection with Ad5.LacZ. These results clearly MTD was defined as the highest dose which did not kill demonstrate that the hybrid Ad5/35 virus is more any animals in the group. Based on this experiment, the efficient in infecting endothelial cells and results in MTD doses were 1.2 Â 1012 vp for Ad5/35.LacZ and higher transgene expression than the standard Ad5. 1.2 Â 1011 vp for Ad5.LacZ (data not shown). However, for direct comparison of the differences in viral tropism, Transduction of Ad5/35 in an in vivo rat tumor model the same dose was used for both viruses. In order to analyze the transduction efficiency of Ad5/35 In order to assess the differences in viral tropism in the in vivo, we used an orthotopic solitary tumor model of HCC model, 1.2 Â 1011 vp of Ad5.LacZ and Ad5/35.LacZ HCC in immune-competent Buffalo rats. This model were administered to three tumor-bearing rats for both represents a clinically relevant model of human HCC. groups via the hepatic artery. After 72 h, the animals Moreover, the larger size of the rat as compared with the were killed and analyzed for viral transgene expression. mouse permitted virus administration via the hepatic In the rats infused with Ad5.LacZ, b-galactosidase artery, which is routinely used to administer chemother- expression was detected primarily in the liver parench- apeutic agents to HCC patients in the clinic. Syngeneic yma, whereas only few tumor cells exhibited b-galacto- McA-RH777 HCC cells were implanted directly into the sidase-positive staining (Figure 3). The positive cells in left lateral lobe of the liver. After 14 days, when tumors the liver parenchyma were diffusely scattered in the reached 13–18 mm in diameter, Ad5.LacZ or Ad5/ tissue, with no clear preference for the endothelial cells 35.LacZ was injected via the hepatic artery. Initially, the in the liver. In Ad5/35.LacZ-infused rats, the staining maximal tolerable dose (MTD) was determined for both pattern was clearly different. The majority of positive viruses in this model. Animal survival was assessed after cells were detected in the tumor rim, where most of the

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 55

Figure 2 Infectivity of Ad5.LacZ and Ad5/35.LacZ in endothelial and liver cells. (a) Cells were infected with Ad5.LacZ and Ad5/35.LacZ with MOI 120 (HUVEC, Hep3B, HepG2 and BRL) or MOI 12 (HAEC) in serum-free medium. The lower MOI for HAECs was chosen because Ad5/35.LacZ infects HAECs so efficiently that infection with MOI 120 results in the death of all Ad5/35.LacZ-infected cells. At 48 h postinfection, the cells were stained with X-gal and photographed. (b) Quantification of the infectivity of Ad5.LacZ and Ad5/35.LacZ. Cells were infected with Ad5.LacZ and Ad5/35.LacZ with indicated MOIs in serum-free medium. At 48 h postinfection, the cells were stained with X-gal and b-galactosidase-positive cells were independently counted by two investigators. White and black bars represent Ad5.LacZ and Ad5/35.LacZ, respectively. (c) Quantification of b-galactosidase enzyme activity after infection by Ad5.LacZ and Ad5/35.LacZ. Cells were infected with Ad5.LacZ and Ad5/35.LacZ with indicated MOIs in serum-free medium. At 48 h postinfection, the cells were lysed and b-galactosidase enzyme activity was detected. Protein content in each sample was measured and the enzyme activity was correlated for the total protein content in each sample.

neovasculature in the tumor nodule was localized. Very cells, suggesting that CD46 acts as a receptor for Ad5/35 little b-galactosidase expression was detected in the in our in vivo model. normal liver tissue of Ad5/35.LacZ-infused rats (Figure 3). These results demonstrate that Ad5/35 has a clearly different tropism from Ad5 in vivo and suggest that this Discussion virus could be used to deliver genes to sites of neoangiogenesis in liver tumors. Angiogenesis has been recognized as a critical step In order to further characterize the cellular localization in tumor progression and metastasis and a number of Ad5/35.LacZ in vivo, double staining using mono- of approaches have been undertaken to interfere with the clonal antibodies for CD46 as well as known endothelial angiogenic process to prevent tumor growth. Direct cell markers CD31 and Flk-1 was employed (Figure 4). targeting of the dividing endothelial cells in growing Examination of the double-stained sections at high tumors offers the advantage of aiming at a geneti- magniﬁcation ( Â 100) revealed colocalization of viral cally stable cell population that is easily accessible b-galactosidase with either CD31 or Flk-1 stained cells, through the bloodstream. The feasibility of this approach indicating that Ad5/35 is able to transduce endothelial was initially established by using a bispeciﬁc antibody cells in vivo. In addition, double stainings for viral to target human tissue factor to tumor endothelial b-galactosidase and CD46 revealed that Ad5/35.LacZ is cells. The tissue factor–antibody complex caused throm- either colocalized or in close proximity of CD46-positive bosis in the tumor vasculature resulting in tumor

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 56

Figure 4 Colocalization of Ad5/35.LacZ with cell surface markers. Tumors were produced in male Buffalo rats by implanting 1 Â107 syngeneic McA-RH7777 rat HCC cells. After 14 days, 1.2 Â 1012 vp of Ad5/35.LacZ was injected via the hepatic artery to three rats. At 72 h postinjection, the animals were euthanized and double stained using X-gal (to detect viral transgene expression) with Flk-1 and CD31 (to detect endothelial cells) or CD46. Representative results are shown. Numbers above the panels indicate the magniﬁcation used. Arrowheads indicate colocalization of Flk-1, CD31 and CD46 with viral b-galactosidase.

Figure 3 Tropism of Ad5.LacZ and Ad5/35.LacZ in vivo. Tumors were produced in male Buffalo rats by implanting 1 Â107 syngeneic In this report, we explore the feasibility of targeting McA-RH7777 rat HCC cells. After 14 days, 1.2 Â 1011 vp of adenoviral vectors to tumor endothelial cells using fiber Ad5.LacZ or Ad5/35.LacZ were injected via the hepatic artery to serotype replacement. The fiber replacement strategy has three rats in each group. At 72 h postinjection, the animals were previously been used with several other serotypes euthanized and stained using X-gal to detect viral transgene 22,24,26,28 expression. Representative results are shown. L ¼ normal liver including 3, 7 and 17. These fiber replacements tissue, T ¼ tumor rim. have enabled virus retargeting to, for example, ovarian cancer, central nervous system and airway epithelium. The Ad5/35 hybrid virus has previously been used to regression.30 Subsequently, similar approaches have target hematopoetic stem cells, dendritic cells and certain been employed to target cytotoxic or antiangio- breast cancer cells.11,23,25 A detailed biodistribution study genic proteins to tumor vasculature, including targeting demonstrated an effective uptake by liver nonparench- tumor necrosis factor alpha and interleukin-12 to ymal cells after intravenous administration of Ad5/35.38 ED-B domain of fibronectin.31,32 In addition, chemo- Here we describe a potential new application for this therapeutic agents and proapoptotic peptides have been vector. directed to tumor endothelium using short targeting In vitro infection assays demonstrated the superiority peptides.33,34 of Ad5/35 hybrid vector in infecting endothelial cells as The specificity of direct targeting can be further compared to the standard Ad5 vector. At MOI 120 Ad5/ enhanced by using vectors capable of gene delivery, 35 transduced 100% of HUVECs whereas infection with where therapeutic genes can be placed also under Ad5 resulted in transduction of 10% of the cells (Figure transcriptional control. In addition, this approach en- 2b). The difference between the two viruses was even ables the combination of multiple therapeutic genes in more striking in HAECs, where Ad5/35 infection with a single vector and the use of replication-competent only MOI 3.6 was sufficient to transduce 95% of cells. viruses, a promising new approach in cancer gene In order to achieve comparable transduction with Ad5, therapy.35 Previously, cationic nanoparticles as well as HAECs had to be infected with MOI 1200. Quantification bone marrow-derived, endothelial progenitor-like cells of the lacZ reporter gene expression revealed an over have been successfully used to deliver therapeutic genes 10-fold increase in gene expression from Ad5/35 in to tumor endothelium.2,3 In addition, a retrovirus HUVECs as compared with Ad5 (Figure 2c). For HAECs, harboring a matrix-targeting motif as well as a tran- this difference was even clearer. In contrast, liver cells scriptionally targeted lentivirus have been shown to were far more transducable by Ad5 than Ad5/35. The selectively target angiogenic regions within tumor situation was most pronounced in BRL rat liver cells, nodules.36,37 Adenoviruses are currently the most widely where reporter gene expression was increased over used vectors for cancer gene therapy, and despite several 50-fold in Ad5-infected cells. In human liver cells the successful approaches of targeting adenoviruses to difference was less clear, but infection with Ad5 still endothelial cells in vitro, the in vivo targeting of resulted in more than two-fold reporter gene expression adenoviruses to tumor vessels has been difficult to than Ad5/35. These results are comparable in terms of achieve.12,13,15 the increase in endothelial infectivity to earlier reports

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 57 were Ad5 infectivity was enhanced by adding targeting with the serotype 35 fiber results in a marked improve- motifs to the HI loop of the fiber knob without abolishing ment on transduction efficiency to endothelial cells CAR binding.39 The marked difference between liver in vitro as well as in vivo. This vector may be used as a cells from different species is probably a reflection of the powerful tool to deliver therapeutic genes to tumor different expression profile of CD46 in rat and human.40 vasculature. Nevertheless, when comparing endothelial and liver cells of human origin, the Ad5/35 vector achieves a therapeutic ratio of at least over 20 and it is notable that Materials and methods therapeutic ratios of many conventional cytotoxic drugs range from 1.5:1 to 6:1.41 Cells and viruses To determine the transduction efficiency of Ad5/35 Hep3B, HepG2, BRL and McA-RH7777 cell lines were in vivo, we used a rat hepatoma model. The larger size of obtained from American Type Culture Collection. rat vasculature as compared to mouse enabled us to HAECs were obtained from Clonetics Corporation administer the vectors via the hepatic artery, a route (Wakersville, MD, USA). HUVECs were isolated from that is routinely used in the clinic to deliver cytotoxic umbilical cords (Institutional Review Board-approved compounds. Injection of Ad5 to the hepatic artery cord blood program) by collagenase type IV (Sigma- resulted in b-galactosidase expression mainly in the liver Aldrich Inc., St Louis, MO, USA) perfusion (0.2% in parenchyma, whereas only few tumor cells were Hanks’ balanced salt solution) for 20 min at room b-galactosidase-positive (Figure 3). In contrast, Ad5/35- temperature. Hep3B, HepG2, BRL and McA-RH7777 infused rats demonstrated b-galactosidase expression cells were cultured in DMEM medium supplemented mainly in the periphery of the tumor with little reporter with 10% FBS, 0.2 mM glutamine, and penicillin/strep- gene expression in the normal liver parenchyma. As the tomycin. In addition, Hep3B medium was supplemented tumor periphery has been previously shown to contain with 1 mM sodium pyruvate. HAECs and HUVECs were the so-called angiogenic rim where most active neo- cultured in EGM-2 medium (Clonetics). vasculature formation is taking place, this expression Recombinant adenoviruses Ad5.LacZ and Ad5/ pattern suggested that Ad5/35 was able to transduce 35.LacZ were prepared by cotransfecting 293 cells with endothelial cells in the tumors.42 Double stainings using E1-deleted shuttle plasmid pAd.RSV.LacZ with back- known endothelial cell markers confirmed the localiza- bone plasmids pBHG10 (for Ad5.LacZ), or pAdDCF35 as tion of viral b-galactosidase to endothelial cells (Figure described.44 pAdDCF35 is based on pBHG10 (Microbix 4). In a previous in vivo study using a mouse breast Biosystems Inc., Toronto, Canada), where the Ad5 fiber carcinoma metastasis model, Ad5/35 failed to efficiently gene was replaced with the Ad35 fiber gene, as transduce breast carcinoma metastases in the liver after previously described.25 Viruses were isolated from a tail vein administration of the virus.11 This dissimilarity single plaque, expanded in 293 cells and purified by may be due to the different rodent model as well as the double cesium gradient ultracentrifugation. The viral tumor type. In addition, the route of virus administration particles were measured by optical absorbance at (intrahepatic artery vs tail vein) is likely to play a 260 nm, and the plaque-forming units (PFU) were significant role in viral tropism. determined by standard agarose-overlay plaque assay CD46 has been recently discovered to be a receptor for on 293 cells. The E1-deleted b-gal virus was produced by most group B adenoviruses, including Ad35.20 As CD46 cotransfecting pCMV-lacZ shuttle plasmid with pBHG10 is expressed in most human tissues, the Ad5/35 vector backbone plasmid as described above. can infect other cell types besides endothelial cells. However, as CD46 has many isoforms that have tissue- Virus infections and reporter gene analysis specific expression, it is not currently known whether all 5 Â 104 cells/well were plated on 24-well dishes 24 h isoforms can mediate Ad35 transdcuction with equal before infection. Cells were infected with indicated MOIs efficiency and our results suggest that liver cells are less of virus in 300 ml serum-free medium. After a 2-h susceptible to Ad5/35 infection than endothelial cells incubation at 371C, cells were washed once with despite similar CD46 expression profiles. In addition, prewarmed PBS and 500 ml complete medium was Ad5/35 transduced HAECs far more efficiently than added to each well. For in situ staining of cells for HUVECs even though both cell types express high levels b-galactosidase activity, after an additional 48 h the cells of CD46. This suggests that additional viral receptors or were washed twice with PBS, fixed with 0.25% glutar- cell type-specific factors may influence Ad5/35 transduc- aldehyde in PBS and incubated in X-Gal solution tion. Many cell types, including endothelial cells, express (Promega, Madison, WI, USA) for 3 h at 371C. The cells heparan sulfates, which have been shown to be able to were then washed once with PBS, photographed and the act as secondary adenoviral attachment receptors and percentage of b-galactosidase positive cells were counted can thus influence adenoviral tropism.43 Furthermore, independently by two investigators. the avb3/avb5 integrins that are required for internaliza- For quantitative measurement of b-galactosidase tion of the hybrid virus are highly overexpressed in activity, the b-galactosidase enzyme assay system (Pro- tumor endothelial cells and a cationic nanoparticle mega) was used. Briefly, after 48-h incubation the cells coupled to an integrin avb3-targeting ligand was recently were lysed using reporter lysis buffer and the enzyme used to deliver genes selectively to angiogenic blood activity was measured using assay buffer according to vessels.2 The selectivity of Ad5/35 targeting can also be the instructions provided by the supplier. Protein content markedly improved by locoregional administration as in each sample was measured using Bio-Rad protein demonstrated here by using hepatic artery delivery. assay (Bio-Rad Laboratories, Hercules, CA, USA) and the In summary, we demonstrate here that replacing the enzyme activity was correlated for the total protein fiber molecule on the standard serotype 5 adenovirus content in each sample.

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 58 Flow cytometry Acknowledgements Flow cytometry was used to analyze CD46, CAR and av integrin expression of the human cell lines used in this This work was supported by grants from the Academy of study. 5 Â 105 cells/sample were washed with PBS and Finland, Turku University Foundation, Sigrid Juselius incubated for 30 min on ice with monoclonal antibodies Foundation, Turku University Central Hospital and for CD46 (BD Biosciences, San Diego, CA, 1:200 dilution), Department of Defense (DAMD17-01-1-0336). CAR (Upstate Biotechnology, Lake Placid, NY, USA; 1:100 dilution for HUVECs and HAECs, 1:50 dilution for

HepG2 and Hep3B) or av integrin (ATCC, 1:50 dilution) in PBS supplemented with 1% FBS. Control samples References were incubated with an isotype specific control antibody (Mouse IgG1, Beckham Coulter, Fullerton, CA, USA for 1 Hanahan D, Folkman J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 1996; 86: CAR and a integrin antibodies and Mouse IgG2a, BD v 353–364. Biosciences, for CD46 antibody). Cells were then washed 2 Hood JD, Bednarski M, Frausto R, Guccione S, Reisfeld RA, twice with ice-cold PBS supplemented with 1% FBS and Xiang R et al. Tumor regression by targeted gene delivery to the incubated with FITC-conjugated goat anti-mouse IgG neovasculature. Science 2002; 296: 2404–2407. (Immunotech, Westbrook, ME, USA) for 30 min on ice. 3 Ferrari N, Glod J, Lee J, Kobiler D, Fine HA. Bone marrow- Cells were washed twice and analyzed by flow cyto- derived, endothelial progenitor-like cells as angiogenesis-selec- metry (FACScan, Becton Dickinson, Mountain View, CA, tive gene-targeting vectors. Gene Therapy 2003; 10: 647–656. USA). Data acquisition and analysis was performed 4 Post DE, Khuri FR, Simons JW, Van Meir EG. Replicative using CellQuest software. oncolytic adenoviruses in multimodal cancer regimens. Hum Gene Ther 2003; 14: 933–946. 5 Bauerschmitz GJ, Barker SD, Hemminki A. Adenoviral gene In vivo experiments therapy for cancer: from vectors to targeted and replication All procedures involving animals were approved by and competent agents (review). Int J Oncol 2002; 21: 1161–1174. performed according to the guidelines of the Institutional 6 Khuri FR, Nemunaitis J, Ganly I, Arseneau J, Tannock IF, Romel Animal Care and Use Committee of the Mount Sinai L et al. A controlled trial of intratumoral ONYX-015, a School of Medicine. Male Buffalo rats (6–7 weeks of age; selectively-replicating adenovirus, in combination with cisplatin Harlan, Indianapolis, IN, USA) were implanted orthoto- and 5-fluorouracil in patients with recurrent head and neck pically with 1 Â107 syngeneic McA-RH7777 rat HCC cancer. Nat Med 2000; 6: 879–885. cells in 20 ml Hanks’ solution. In order to determine the 7 Fechner H, Haack A, Wang H, Wang X, Eizema K, Pauschinger MTD for both Ad5.LacZ and Ad5/35.LacZ in this model, M et al. Expression of coxsackie adenovirus receptor and alphav- increasing doses of Ad5.LacZ and Ad5/35.LacZ (4 Â 1010 integrin does not correlate with adenovector targeting in vivo vp, 1.2 Â 1011 vp, 4 Â 1011 vp and 1.2 Â 1012 vp) in 1 ml indicating anatomical vector barriers. Gene Therapy 1999; 6: PBS were injected to the tumor-bearing rats (three rats/ 1520–1535. 8 Tomko RP, Xu R, Philipson L. HCAR and MCAR: the human and vector dose) via the hepatic artery. Based on this mouse cellular receptors for subgroup C adenoviruses and experiment, the MTD doses were 1.2 Â 1012 vp for 11 group B coxsackieviruses. Proc Natl Acad Sci USA 1997; 94: Ad5/35.LacZ and 1.2 Â 10 vp for Ad5.LacZ. However, 3352–3356. for direct comparison of the differences in viral tropism, 9 Wickham TJ, Tzeng E, Shears II LL, Roelvink PW, Li Y, Lee GM 11 the same dose (1.2 Â 10 vp) was used for both viruses. et al. Increased in vitro and in vivo gene transfer by adenovirus Viruses were injected to the rats 14 days after tumor cell vectors containing chimeric fiber proteins. J Virol 1997; 71: implantation. To assess virus spread in the tumor and in 8221–8229. the liver, animals were euthanized at 72 h after vector 10 Merrick AF, Shewring LD, Sawyer GJ, Gustafsson KT, Fabre JW. injection and livers were removed for analysis. Repre- Comparison of adenovirus gene transfer to vascular endothelial sentative results are shown in the figures. cells in cell culture, organ culture, and in vivo. Transplantation For histological examination, immunohistochemistry 1996; 62: 1085–1089. and b-galactosidase expression analysis, the harvested 11 Shayakhmetov DM, Li ZY, Ni S, Lieber A. Targeting of livers were fixed in 1% paraformaldehyde for 4 h, and adenovirus vectors to tumor cells does not enable efficient then in 18% sucrose solution overnight. After fixation, transduction of breast cancer metastases. Cancer Res 2002; 62: the blocks were embedded in Tissue-Tek (Sakura Finetek 1063–1068. USA, Torrance, CA, USA). Cryostat sections of 10 mm 12 Wickham TJ, Segal DM, Roelvink PW, Carrion ME, Lizonova A, thickness were stained by immersion in X-Gal (5-bromo- Lee GM et al. Targeted adenovirus gene transfer to endothelial and smooth muscle cells by using bispecific antibodies. J Virol 4-chloro-3-indolyl-b-D-galactopyranoside) staining solution for 16 h at 371C. Sections were counterstained with 1996; 70: 6831–6838. nuclear fast red (Vector Laboratories, Burlingame, CA, 13 Nettelbeck DM, Miller DW, Jerome V, Zuzarte M, Watkins SJ, Hawkins RE et al. Targeting of adenovirus to endothelial cells by USA). Immunohistochemistry was performed using a bispecific single-chain diabody directed against the adenovirus monoclonal antibodies for known endothelial markers fiber knob domain and human endoglin (CD105). Mol Ther 2001; CD31 (BD Biosciences) and Flk-1 (Santa Cruz Biotech- 3: 882–891. nology Inc., Santa Cruz, CA, USA) as well as for CD46 14 Fisher KD, Stallwood Y, Green NK, Ulbrich K, Mautner V, (Santa Cruz Biotechnology Inc.). For double stainings, Seymour LW et al. Polymer-coated adenovirus permits efficient sections were first incubated with a monoclonal anti- retargeting and evades neutralising antibodies. Gene Therapy body, rinsed twice in PBS, and stained for b-gal. In situ 2001; 8: 341–348. detection of b-galactosidase activity in the sections was 15 Nicklin SA, Von Seggern DJ, Work LM, Pek DC, Dominiczak AF, performed using X-gal as suggested by the supplier Nemerow GR et al. Ablating adenovirus type 5 fiber-CAR (Promega). binding and HI loop insertion of the SIGYPLP peptide generate

Gene Therapy Endothelial targeting by a hybrid virus K Shinozaki et al 59 an endothelial cell-selective adenovirus. Mol Ther 2001; 4: 29 Herz J, Gerard RD. Adenovirus-mediated transfer of low density 534–542. lipoprotein receptor gene acutely accelerates cholesterol clear- 16 Einfeld DA, Schroeder R, Roelvink PW, Lizonova A, King CR, ance in normal mice. Proc Natl Acad Sci USA 1993; 90: 2812–2816. Kovesdi I et al. Reducing the native tropism of adenovirus 30 Huang X, Molema G, King S, Watkins L, Edgington TS, Thorpe vectors requires removal of both CAR and integrin interactions. PE et al. Tumor infarction in mice by antibody-directed targeting J Virol 2001; 75: 11284–11291. of tissue factor to tumor vasculature. Science 1997; 275: 547–550. 17 Martin K, Brie A, Saulnier P, Perricaudet M, Yeh P, Vigne E et al. 31 Borsi L, Balza E, Carnemolla B, Sassi F, Castellani P, Berndt A Simultaneous CAR- and alpha V integrin-binding ablation fails et al. Selective targeted delivery of TNFa to tumor blood vessels. to reduce Ad5 liver tropism. Mol Ther 2003; 8: 485–494. Blood 2003; 102: 4384–4392. 18 Koizumi N, Mizuguchi H, Sakurai F, Yamaguchi T, Watanabe Y, 32 Halin C, Rondini S, Nilsson F, Berndt A, Kosmehl H, Zardi L et al. Hayakawa T et al. Reduction of natural adenovirus tropism to Enhancement of the antitumor activity of interleukin-12 by targeted mouse liver by fiber-shaft exchange in combination with both delivery to neovasculature. Nat Biotechnol 2002; 20: 264–269. CAR- and alphav integrin-binding ablation. J Virol 2003; 77: 33 Arap W, Pasqualini R, Ruoslahti E. Cancer treatment by targeted 13062–13072. drug delivery to tumor vasculature in a mouse model. Science 19 Akiyama M, Thorne S, Kirn D, Roelvink PW, Einfeld DA, King 1998; 279: 377–380. CR et al. Ablating CAR and integrin binding in adenovirus 34 Ellerby HM, Arap W, Ellerby LM, Kain R, Andrusiak R, Rio GD vectors reduces nontarget organ transduction and permits et al. Anti-cancer activity of targeted pro-apoptotic peptides. Nat sustained bloodstream persistence following intraperitoneal Med 1999; 5: 1032–1038. administration. Mol Ther 2004; 9: 218–230. 35 Savontaus MJ, Sauter BV, Huang TG, Woo SL. Transcriptional 20 Gaggar A, Shayakhmetov DM, Lieber A. CD46 is a cellular targeting of conditionally replicating adenovirus to dividing receptor for group B adenoviruses. Nat Med 2003; 9: 1408–1412. endothelial cells. Gene Therapy 2002; 9: 972–979. 21 Roelvink PW et al. Identification of a conserved receptor-binding 36 Gordon EM, Chen ZH, Liu L, Whitley M, Liu L, Wei D et al. site on the fiber proteins of CAR-recognizing adenoviridae. Systemic administration of a matrix-targeted retroviral vector is Science 1999; 286: 1568–1571. efficacious for cancer gene therapy in mice. Hum Gene Ther 2001; 22 Stevenson SC, Rollence M, White B, Weaver L, McClelland A. 12: 193–204. Human adenovirus serotypes 3 and 5 bind to two different 37 De Palma M, Venneri MA, Naldini L. In vivo targeting of tumor cellular receptors via the fiber head domain. J Virol 1995; 69: endothelial cells by systemic delivery of lentiviral vectors. Hum 2850–2857. Gene Ther 2003; 14: 1193–1206. 23 Rea D, Havenga MJ, van Den Assem M, Sutmuller RP, Lemckert 38 Sakurai F, Mizoguchi H, Yamaguchi T, Hayakawa T. Character- A, Hoeben RC et al. Highly efficient transduction of ization of in vitro and in vivo gene transfer properties of human monocyte-derived dendritic cells with subgroup B adenovirus serotype 35 vector. Mol Ther 2003; 8: 813–821. fiber-modified adenovirus vectors enhances transgene-encoded 39 Biermann V, Volpers C, Hussmann S, Stock A, Kewes H, antigen presentation to cytotoxic T cells. J Immunol 2001; 166: Schiedner G et al. Targeting of high-capacity adenoviral vectors. 5236–5244. Hum Gene Ther 2001; 12: 1757–1769. 24 Kanerva A, Zinn KR, Chaudhuri TR, Lam JT, Suzuki K, Uil TG 40 Mead R, Hinchliffe SJ, Morgan BP. Molecular cloning, expression et al. Enhanced therapeutic efficacy for ovarian cancer with a and characterization of the rat analogue of human membrane serotype 3 receptor-targeted oncolytic adenovirus. Mol Ther cofactor protein (MCP/CD46). Immunology 1999; 98: 137–143. 2003; 8: 449–458. 41 Elliott WL, Roberts BJ, Howard CT, Leopold III WR. Chemo- 25 Shayakhmetov DM, Papayannopoulou T, Stamatoyannopoulos therapy with [SP-4-3-(R)]-[1,1-cyclobutanedicarboxylato(2-)](2- G, Lieber A. Efficient gene transfer into human CD34(+) cells by methyl-1,4-butanediamine-N,N’)platinum (CI-973, NK121) in a retargeted adenovirus vector. J Virol 2000; 74: 2567–2583. combination with standard agents against murine tumors 26 Gall J, Kass-Eisler A, Leinwand L, Falck-Pedersen E. Adenovirus in vivo. Cancer Res 1994; 54: 4412–4418. type 5 and 7 capsid chimera: fiber replacement alters receptor 42 Engels K, Fox SB, Whitehouse RM, Gatter KC, Harris AL. tropism without affecting primary immune neutralization Distinct angiogenic patterns are associated with high-grade epitopes. J Virol 1996; 70: 2116–2123. in situ ductal carcinomas of the breast. J Pathol 1997; 181: 207–212. 27 Havenga MJ, Lemckert AA, Grimbergen JM, Vogels R, Huisman 43 Dechecchi MC, Melotti P, Bonizzato A, Santacatterina M, Chilosi LG, Valerio D et al. Improved adenovirus vectors for infection of M, Cabrini G et al. Heparan sulfate glycosaminoglycans are cardiovascular tissues. J Virol 2001; 75: 3335–3342. receptors sufficient to mediate the initial binding of adenovirus 28 Chillon M, Bosch A, Zabner J, Law L, Armentano D, Welsh MJ types 2 and 5. J Virol 2001; 75: 8772–8780. et al. Group D adenoviruses infect primary central nervous 44 Bautista DS, Hitt M, McGrory J, Graham FL. Isolation and system cells more efficiently than those from group C. J Virol characterization of insertion mutants in E1A of adenovirus type 1999; 73: 2537–2540. 5. Virology 1991; 182: 578–596.

Gene Therapy