Adenoviral vectors expressing fusogenic membrane glycoproteins activated via matrix...

12
THE JOURNAL OF GENE MEDICINE RESEARCH ARTICLE J Gene Med 2004; 6: 1216–1227. Published online 30 September 2004 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jgm.616 Adenoviral vectors expressing fusogenic membrane glycoproteins activated via matrix metalloproteinase cleavable linkers have significant antitumor potential in the gene therapy of gliomas Cory Allen Cari McDonald Caterina Giannini Kah Whye Peng Gabriela Rosales Stephen J. Russell Evanthia Galanis* Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA *Correspondence to: Evanthia Galanis, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA. E-mail: [email protected] Received: 22 December 2003 Revised: 29 March 2004 Accepted: 1 April 2004 Abstract Background Fusogenic membrane glycoproteins (FMG) such as the gibbon ape leukemia virus envelope (GALV) glycoprotein are potent therapeutic transgenes with potential utility in the gene therapy of gliomas. Transfection of glioma cell lines with FMG expression constructs results in fusion with massive syncytia formation followed by cytotoxic cell death. Nevertheless, ubiquitous expression of the GALV receptor, Pit-1, makes targeting desirable in order to increase the specificity of the observed cytopathic effect. Here we report on use of matrix metalloproteinase (MMP)-cleavable linkers to target the cytotoxicity of FMG-expressing adenoviral vectors against gliomas. Methods Replication-defective adenoviruses (Ad) were constructed express- ing the hyperfusogenic version of the GALV glycoprotein linked to a blocking ligand (C-terminal extracellular domain of CD40 ligand) through either an MMP-cleavable linker (AdM40) or a non-cleavable linker (AdN40). Both viruses also co-expressed the green fluorescent protein (GFP) via an internal ribosomal entry site. Results The glioma cell lines U87, U118, and U251 characterized by zymog- raphy and MMP-2 activity assay as high, medium and low MMP expressors, respectively, the MMP-poor cell lines TE671 and normal human astrocytes were infected with AdM40 and AdN40 at different multiplicities of infection (MOIs) from 1-30. Fusion was quantitated by counting both number and size of syncytia. Infection of these cell lines with AdN40 did not result in fusion or cytotoxic cell death, despite the presence of infection, as demonstrated by GFP positivity, therefore indicating that the displayed CD40 ligand blocked GALV-induced fusion. Fusion was restored after infection of glioma cells with AdM40 at an MOI as low as 1 to an extent dependent on MMP expression and coxsackie adenovirus receptor (CAR) expression in the specific cell line. Western immunoblotting demonstrated the presence of the cleaved CD40 ligand in the supernatant of fused glioma cells. Use of the MMP inhibitors 1,10 phenanthroline and N-hydroxy-5,5-dimethylpiperazine-2-carboxamide completely abolished AdM40-induced fusion, while the non-specific ser- ine protease inhibitor soybean trypsin inhibitor did not affect it, thus demonstrating specificity of the observed effect. Intratumoral treatment of BalbC/nude mice bearing subcutaneous U87 glioma xenografts with AdM40 at a total dose of 1.2 × 10 10 plaque-forming units (pfu) resulted in statistically Copyright 2004 John Wiley & Sons, Ltd.

Transcript of Adenoviral vectors expressing fusogenic membrane glycoproteins activated via matrix...

THE JOURNAL OF GENE MEDICINE R E S E A R C H A R T I C L EJ Gene Med 2004; 6: 1216–1227.Published online 30 September 2004 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jgm.616

Adenoviral vectors expressing fusogenicmembrane glycoproteins activated via matrixmetalloproteinase cleavable linkers havesignificant antitumor potential in the genetherapy of gliomas

Cory AllenCari McDonaldCaterina GianniniKah Whye PengGabriela RosalesStephen J. RussellEvanthia Galanis*

Mayo Clinic, 200 First Street SW,Rochester, MN 55905, USA

*Correspondence to:Evanthia Galanis, Mayo Clinic, 200First Street SW, Rochester, MN55905, USA. E-mail:[email protected]

Received: 22 December 2003Revised: 29 March 2004Accepted: 1 April 2004

Abstract

Background Fusogenic membrane glycoproteins (FMG) such as the gibbonape leukemia virus envelope (GALV) glycoprotein are potent therapeutictransgenes with potential utility in the gene therapy of gliomas. Transfectionof glioma cell lines with FMG expression constructs results in fusion withmassive syncytia formation followed by cytotoxic cell death. Nevertheless,ubiquitous expression of the GALV receptor, Pit-1, makes targeting desirablein order to increase the specificity of the observed cytopathic effect. Here wereport on use of matrix metalloproteinase (MMP)-cleavable linkers to targetthe cytotoxicity of FMG-expressing adenoviral vectors against gliomas.

Methods Replication-defective adenoviruses (Ad) were constructed express-ing the hyperfusogenic version of the GALV glycoprotein linked to a blockingligand (C-terminal extracellular domain of CD40 ligand) through either anMMP-cleavable linker (AdM40) or a non-cleavable linker (AdN40). Bothviruses also co-expressed the green fluorescent protein (GFP) via an internalribosomal entry site.

Results The glioma cell lines U87, U118, and U251 characterized by zymog-raphy and MMP-2 activity assay as high, medium and low MMP expressors,respectively, the MMP-poor cell lines TE671 and normal human astrocyteswere infected with AdM40 and AdN40 at different multiplicities of infection(MOIs) from 1-30. Fusion was quantitated by counting both number and sizeof syncytia. Infection of these cell lines with AdN40 did not result in fusionor cytotoxic cell death, despite the presence of infection, as demonstrated byGFP positivity, therefore indicating that the displayed CD40 ligand blockedGALV-induced fusion. Fusion was restored after infection of glioma cells withAdM40 at an MOI as low as 1 to an extent dependent on MMP expressionand coxsackie adenovirus receptor (CAR) expression in the specific cell line.Western immunoblotting demonstrated the presence of the cleaved CD40ligand in the supernatant of fused glioma cells. Use of the MMP inhibitors1,10 phenanthroline and N-hydroxy-5,5-dimethylpiperazine-2-carboxamidecompletely abolished AdM40-induced fusion, while the non-specific ser-ine protease inhibitor soybean trypsin inhibitor did not affect it, thusdemonstrating specificity of the observed effect. Intratumoral treatment ofBalbC/nude mice bearing subcutaneous U87 glioma xenografts with AdM40at a total dose of 1.2 × 1010 plaque-forming units (pfu) resulted in statistically

Copyright 2004 John Wiley & Sons, Ltd.

Ad Vectors Activated via MMP-Cleavable Linkers 1217

significant tumor regression as compared with control animals either treated with AdN40 (p = 0.01) or untreatedanimals (p = 0.01). Treatment with AdM40 also resulted in survival improvement as compared with AdN40-treatedanimals (p = 0.006) or untreated animals (p = 0.001). Histopathologic examination of treated tumors demonstratedextensive syncytia formation.

Conclusions Our data indicate that AdM40, a replication-defective adenovirus expressing the GALV fusogenicglycoprotein, attached to a blocking ligand via an MMP-cleavable linker, can target the cytotoxicity of GALV inMMP-overexpressing glioma lines and xenografts, and maintain significant antitumor activity both in vitro and in vivo.Given the high frequency of MMP overexpression in gliomas, AdM40 represents a potentially promising agent in thegene therapy of these tumors. Copyright 2004 John Wiley & Sons, Ltd.

Keywords GALV; targeting; adenovirus; gliomas; MMP; linkers

Introduction

We have recently demonstrated that a hyperfusogenicform of the gibbon ape leukemia virus (GALV) glycopro-tein, characterized by a truncation of the C-terminal Rpeptide, represents a potent therapeutic transgene withpotential utility in the gene therapy of different malig-nancies including gliomas [1–3]. Transfection with GALVleads to cell–cell fusion and massive syncytia formationfollowed by apoptotic cell death in glioma lines [1].The GALV receptor Pit-1 [4], however, is ubiquitouslyexpressed in primate cells. As a result, GALV-inducedfusion of tumor cells can spread in heterologous celllines including normal cells. As it pertains to gliomas, ithas been shown that GALV-transfected U87 glioma cellscould fuse and cause cytotoxic death to untransfectednormal human astrocytes and fibroblasts [1]. Therefore,it becomes evident that targeting the cytotoxicity of GALVcould further improve its therapeutic index and increaseits clinical applicability in the treatment of gliomas.

The hyperfusogenic form of GALV is expressed as atrimer on the cell surface [5]. Thus, as a blocking ligandwe chose to display at the N-terminus of GALV a trimericpolypeptide, i.e., the C-terminal domain of the CD40ligand. Display of trimeric polypeptides at the N-terminusof retroviral glycoproteins has been shown to createsteric impedance and ablate retroviral attachment andentry [6,7]. We hypothesized that a protease cleavagesignal between GALV and CD40 ligand could allow us toexploit the protease-rich tumor environment in order toaccomplish selective cleavage of the CD40 ligand blockingdomain and activation of GALV in this setting. Giventhe overexpression of matrix metalloproteinase (MMP) ingliomas, we chose to insert the MMP-cleavable linkerGGPLGLWAGG between GALV and the CD40 ligand[8]. Matrix metalloproteases are a particularly appealingtarget for the development of targeted therapeutics ingliomas: they are frequently overexpressed in high-gradetumors such as glioblastoma multiforme and anaplasticastrocytoma, while there is no MMP expression in normalbrain tissue [9–12]. Using GALV constructs containingthe MMP-cleavable linker GGPLGLWAGG, we showedthat transfection of glioma lines expressing variable MMPlevels led to MMP-specific glioma cell line fusion andcytotoxic cell death at an extent dependent on MMP

overexpression [8]. Despite the fact, however, that thetargeted plasmid constructs decreased glioma cell linetumorigenicity [8], they lacked therapeutic efficacy inestablished xenografts.

In order to examine the therapeutic potential of thistargeting approach, we proceeded with construction ofadenoviral vectors containing MMP-cleavable GALV con-structs. Construction of GALV-containing adenoviruses,however, has been technically very challenging because ofcytotoxic death of the 293A producer cells due to expres-sion of the GALV Pit-1 receptor. We demonstrated thatproduction of adenoviral vectors coding for the hyper-fusogenic version of GALV and displaying the blockingC-terminus of CD40 ligand via an MMP-cleavable linkerGGPLGLWAGG (AdM40) in high titers is feasible. TheAdM40 adenovirus can target the cytotoxicity of GALVagainst MMP-overexpressing gliomas while sparing nor-mal human astrocytes. In addition, AdM40 has significanttherapeutic activity in vivo leading in tumor regressionand prolongation of survival in a U87 glioma model, and,therefore, it represents a potentially promising approachin the gene therapy of gliomas.

Methods

Cell lines

The glioma cell lines U87, U251, U118, and the rhab-domyosarcoma cell line TE671 were obtained from ATCC(Manassa, VA, USA) and maintained in Dulbecco’s mod-ified Eagle’s medium (DMEM) supplemented with 10%fetal bovine serum (FBS) and 1% penicillin/streptomycin(P/S). Normal human astrocytes were obtained fromClonexpress (Gaithersburg, MD, USA) and maintainedin DMEM/F12, supplemented with 5% FBS and astrocytegrowth factor supplement (Clonexpress).

Adenovirus construction

The template DNA was PCR-amplified from the GALVM40and GALVN40 expression constructs [8] using AmpliTaqGold (Applied Biosystems, Foster City, CA, USA) asper the manufacturer’s instructions with the addition

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

1218 C. Allen et al.

of an equal volume of Vent DNA polymerase (NewEngland Biolabs, Beverly, MA, USA), and the followingprimers: Mo.F.Bgl2 5′-GGC AGA TCT ATG GCG CGT TCAACG CTC TCA AAA-3′ and Mo.R.Bgl2 5′-GGC AGA TCTTTA ACA TGC ACT TAT CCT ATC A-3′. The resultingPCR products were excised from a 1% agarose gel,purified using the High Pure PCR product purificationkit (Roche, Indianapolis, IN, USA) and cloned into thepCR2.1 vector (Invitrogen, Carlsbad, CA, USA) using aTA cloning system. The Pac I site was eliminated fromboth target sequences using the QuickChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA)and the following primers: GALVP.R 5′-GGA GGT CTATAG GTC CTT TAA TCA AGG CAG TTG AAC CAG TACC-3′ and GALVPac.F 5′-GGT ACT GGT TCA ACT GCCTTG ATT AAA GGA CCT ATA GAC CTC C-3′, resultingin a silent mutation in the sequence of interest. This wasverified by diagnostic digestion using Pac I followed bysequencing of each final target region in its entirety. ThepCR2.L GALVM40 and pCR2.1 GALVN40 plasmids weredigested with Bgl II, gel-purified, and ligated into theBgl II site of pAdenoVator-CMV5-IRES-GFP (Qbiogene,Carlsbad, CA, USA). Insertion in the correct orientationwas verified by Hpa I digestion. The resulting plasmidswere linearized with Pme I and co-electroporated withpAdenoVator � E1/E3 into electrocompetent BJ5183 E.coli. Positive recombinants were selected following BstX Idigestion and transformed into One Shot ElectrocompGeneHogs E. coli (Invitrogen). Positive transformantswere selected following Bst I and Pac I digestion.The resulting recombinant pAd containing either theGALV M40 or the GALV N40 sequences (Figure 1) werelinearized with Pac I and transfected into 293A cellsusing Lipofectamine Plus (Life Technologies, Carlsbad,CA, USA). Twenty-four hours post-transfection, the cellswere overlaid with agarose; visible plaques were selected12 days following transfection and expanded. Virus was

purified from 293A cells by three freeze/thaw cycles priorto concentration using cesium chloride centrifugation.Titration was performed on 293A cells [13].

Characterization of MMP activity

The glioma cell lines used in our experiments and normalhuman astrocytes have been previously characterized byzymography regarding MMP expression [14], and bya colorimetric assay for MMP-2 expression [8]. Theyhave also been characterized regarding their ability tocleave the MMP linker GGPLGLWAGG incorporated inour constructs [8].

Flow cytometry for determination ofCAR expression

Cells (1 × 106) were incubated for 45 min on ice with10 µg monoclonal antibody RmcB (ATCC) in 500 µl0.5% PBS-BSA. The cells were washed in 0.5% PBS-BSA and incubated with 5 µg goat anti-mouse IgG FITC(Pierce, Rockford, IL, USA) in phosphate-buffered saline(PBS) containing 0.5% bovine serum albumin (BSA).Cells were washed twice 30 min later with 0.5% PBS-BSA, and fixed with 0.5% paraformaldehyde. Sampleswere assayed on a Bekton-Dickinson FACScan cytometer.Analysis was performed using the CellQuest software(Bekton-Dickinson, San Jose, CA, USA).

Assessment of the CPE in vitro

Twenty-four hours prior to infection, cells were plated in6-well plates at a density of 3 × 105 cells per well. Cellswere infected at 37 ◦C with 1 ml/well of the appropriatemedium containing AdM40 or AdN40 at MOIs of 0, 1, 10

LITR

CMVpromoter

GALV CD40L

IRES

GFP

Poly A+

RITR

Pac I

AD5 DNA

Bgl IIBgl IIPac I

GGGGS

GGPLGLWAGG

N40

M40

Linker

Figure 1. Schematic representation of the AdN40 and AdM40 adenoviruses. Both viruses encode the gibbon ape leukemia virusenvelope glycoprotein displaying the CD40 ligand via either the non-cleavable linker GGGGS (AdN40) or the MMP-cleavable linkerGGPLGLWAGG (AdM40). Both viruses also express the green fluorescent protein (GFP) gene via an internal ribosomal entry(IRES) site

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

Ad Vectors Activated via MMP-Cleavable Linkers 1219

and 30. After 4 h the medium was replaced with 5 ml offresh growth medium. Cells were collected at days 2, 3and 5 post-infection, stained with trypan blue and viablecells were counted. Results were presented as percentageof infected/uninfected cells surviving at each time pointand MOI. In addition two wells per cell line and timepoints were fixed with 0.5% glutaraldehyde and stainedwith crystal violet to better illustrate the observed CPE.

Quantitation of fusion

Cells were plated in 6-well dishes at a density of 3 × 105

cells/well and infected as described above with AdM40or AdN40 at MOIs of 0, 1, 10, and 30. On days 2, 3 and5 cells were fixed with 0.5% glutaraldehyde. The averagearea (A) occupied by the nuclear cluster per syncytiumwas estimated using a 4 × 4 grid under 20× phase in sixdifferent fields per well, in duplicate wells. Cells were thenstained with 0.1% crystal violet in 2% ethanol, rinsed andcovered with 5 ml PBS; the number (N) of syncytia per2 mm square of a grid was counted in two different fieldsper well in duplicate wells under 4× magnification, andthe average number was calculated. The average nucleararea per syncytium multiplied by the average number ofsyncytia per 2 mm square was calculated (A × N, syncytialindex). All experiments were repeated at least twice.

Western immunoblotting

Cells (2 × 107) were plated in T-175 flasks and infected24 h later with either AdM40 or AdN40. At 24 h, 7 ml ofmedium were removed and replaced with fresh growthmedium; another 7 ml of medium were removed 24 hlater, and all 14 ml were mixed with an equal volume of2× lysis buffer containing 50 mM sodium pyrophosphate,50 mM sodium fluoride, 50 mM sodium chloride, 5 mMEDTA, 5 mM EGTA, 100 uM sodium orthovanadate,10 mM HEPES, pH 7.4, 0.1% Triton X-100, and 0.5 mMPMSF. Cell debris was removed by centrifugation and theresulting sample was concentrated using an Amicon Ultra10KD MWCO spin filter (Millipore Corporation, Bedford,MA, USA). The concentrated samples were removed andthe protein concentration was assayed using the PierceBCA protein assay kit. The samples were preheated inLaemmli’s sample buffer (0.3 M Tris, 30% glycerol, 1%SDS, 5% 2-mercaptoethanol, 0.01% bromophenol blue)for 10 min at 99 ◦C and loaded onto a 15% Tris-HClgel (100 µg of protein per well). The proteins weretransferred overnight onto nitrocellulose, and blocked for3 h in 5% BSA/dry milk in TTBS. The blot was incubatedwith 50 ng/mL of biotinylated anti-human CD40 ligandantibody (R&D Systems) in TTBS containing 1% BSA/drymilk for 1 h, were washed in TTBS (3 × 10 min),incubated with 1 µl/ml streptavidin-HRP (AmershamBiosciences) in TTBS containing 1% BSA/dry milk for1 h and then washed in TTBS (3 × 10 min, 2 × 30 min)

prior to being exposed to ECL+ (Amersham Biosciences)for protein detection.

Inhibitor assays

Cells were plated in 6-well plates at a density of3 × 105 cells/well and infected with AdM40 or AdN40at MOIs of 1, 10, and 30 as described above. Four hoursafter infection, the medium was changed to inhibitor-containing medium. We used two broad-spectrum MMPinhibitors: (a) 1,10-phenanthroline (Aldrich, Milwaukee,WI, USA), concentrations 0, 3.125, 6.25, 8, 10,12.5, 25, 50, and 100 µM; and (b) N-hydroxy-5,5-dimethylpiperazine-2-carboxamide (Calbiochem, La Jolla,CA, USA), concentrations 0, 10 and 100 µM. In orderto exclude non-specific protease inhibition, we used, ascontrol, the serine protease inhibitor soybean trypsininhibitor (Roche), concentrations 0, 0.1 and 1 mg/ml.Three days post-infection, cells were fixed with 0.5%glutaraldehyde and the syncytial index was calculated asabove.

In vivo studies

All animal protocols were approved by the Mayo ClinicInstitutional Animal Care and Use Committee (IACUC).

Treatment of subcutaneous U87 tumorxenografts

U87 cells (5 × 106) were injected subcutaneously into theflanks of BalbC/nude mice (8–9 animals per group). Fourdays later, when the tumors had reached an average sizeof 0.3–0.4 cm, the animals in the treatment group weretreated with 0.4 × 109 pfu of AdM40 for 3 consecutivedays (total dose of 1.2 × 1010 pfu), while one group ofthe control animals was treated with the same total doseof AdN40 and the third group was left untreated. Tumorvolume was measured three times per week. As per theIACUC guidelines, animals were euthanized if weight loss>20% of body weight was observed or the tumor diameterexceeded 1 cm. In addition, 3 and 8 days after completionof treatment, one animal in the AdM40-treated group andone in the AdN40 control group were euthanized andtheir tumors were stained with hematoxylin and eosin(H&E).

Statistical analysis

The Kruskall-Wallis test was used to compare tumorgrowth among the different animal groups. The Wilcoxonrank sum test was used to test two group tumor volumegrowth rate comparison. Survival curves were generatedusing the Kaplan-Meier method. The log-rank test wasused to test for differences in survival times among thethree groups and between two group pairs. A p value≤0.05 was considered statistically significant.

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

1220 C. Allen et al.

Results

Characterization of MMP activity

The glioma cell lines used in our experiments and normalhuman astrocytes have been previously characterizedby zymography regarding MMP expression [14]. U87has the highest MMP expression followed by U118 andU251, while normal human astrocytes do not expressMMP. Furthermore, we measured total and active MMP-2levels, using the MMP-2 activity assay system (AmershamPharmacia Biotech, Piscataway, NJ, USA); these resultshave been previously reported [8] and are consistent withthe zymography results: U87 cells produce higher MMP-2levels followed by U118 and U251 cells while TE671 cellshave undetectable MMP-2 levels. In addition, the abilityof these cell lines to cleave the MMP linker GGPLGLWAGGwas assessed using a fluorogenic substrate and was foundto be proportional to their MMP expression [8]. Ofimportance for the adenoviral construction, the 293Acells used for the production of our adenoviral vectorshave no detectable MMP-2 activity (data not shown).

Characterization of CAR receptor level

Given the impact of adenovirus receptor expression on theefficacy of adenoviral gene therapy [15], we determinedthe expression levels of the coxsackie adenovirus receptor(CAR) in our tumor cell lines. All glioma cell linesand the TE671 cells were characterized by fluorescence-activated cell sorting (FACS) for CAR expression levels:U251 and TE671 cells had the highest CAR expression,followed by U87 cells, while U118 cells had very low

CAR expression. Normal human astrocytes have beenpreviously characterized for CAR expression [16] andthey were found to be high CAR expressors with morethan 50% of the cells expressing the receptor.

Targeting of GALV-induced fusion withadenoviral vectors encodingMMP-cleavable linkers

In order to target GALV cytotoxicity against gliomas, weconstructed replication-defective adenoviruses expressingthe hyperfusogenic GALV protein displaying the blockingC-terminus of CD40 ligand via either an MMP-cleavablelinker or a non-cleavable linker using the AdenoVatorsystem (Qbiogene). The low MMP levels of the producer293A cells prevented activation of GALV and 293Acytotoxicity during production of AdM40 resulting inadequate viral titers: 2.5–5.5 × 1010 pfu/mL for AdM40and 2–7 × 1010 pfu/mL for AdN40. Both viruses containthe hyperfusogenic version of the GALV glycoproteindisplaying, on its N-terminus, the C-terminus of CD40ligand via either the MMP-cleavable linker GGPLGLWAGG(AdM40) or the non-cleavable linker GGGGS (AdN40)(Figure 1). The GGPLGLWAGG linker is cleaved by abroad spectrum of MMPs including gelatinases (MMP-2and MMP-1) and MT1-MMP [17,18]. Both viruses co-expressed GFP via an internal ribosomal entry (IRES)site. Infections of the glioma cell lines U87, U118, U251,normal human astrocytes and the MMP-poor non-gliomacell line TE671 were performed (Figures 2–5). Infectionwith AdM40 at an MOI as low as 1 led to tumor cellfusion and cytotoxic cell death to an extent dependent onMMP and CAR expression. No fusion was observed in any

Syn

cytia

l ind

ex

Days

U87

U87 AdN40

U118

U118

U251

U251

TE671

TE671

NHA

NHA

0.1

1

10

2 53

AdM40

AdM40

AdN40

AdM40

AdN40

AdM40

AdN40

AdM40

AdN40

Figure 2. Maximum fusion observed in the glioma cell lines U87, U118, and U251, the MMP-poor non-glioma cell line TE671 andnormal human astrocytes (NHA) at 2, 3, and 5 days after infection with the AdM40 and AdN40 adenoviruses at an MOI of 1. Resultsare expressed as syncytial index (see Methods). The observed AdM40 induced fusion increased over time in MMP-expressing celllines. AdM40 failed to induce fusion in normal human astrocytes, indicating that the proposed MMP-based targeting strategy canprotect normal brain

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

Ad Vectors Activated via MMP-Cleavable Linkers 1221

Uninfected AdM40 MOI 10 AdN40 MOI 10

U87

TE671

NHA

A CB

IHG

E FD

Figure 3. MMP-rich U87 cells 48 h after infection with AdM40 (B) and AdN40 (C) at an MOI of 10. MMP-poor TE671 cells afterinfection with AdM40 (E) and AdN40 (F) at an MOI of 10. Normal human astrocytes after infection with AdM40 (H) and AdN40(I) at an MOI of 10. Extensive syncytial formation is observed after infection with AdM40 in U87 cells. In contrast, AdM40 failedto cause fusion in the MMP-poor cell line TE671 and the MMP-negative normal human astrocytes. Uninfected U87 (A), TE671 (D),and normal human astrocytes (G) are included as controls

cell line after infection with AdN40. In MMP-producinglines there was incremental increase in observed fusionafter infection with AdM40 over a period of 5 days atwhich point maximum syncytial formation was observedpreceding cell death (Figure 2). In contrast no fusionor cytotoxicity was observed in MMP-poor TE671 cellsand normal human astrocytes (Figures 2 and 3). Lack offusion after infection with AdM40 in MMP-poor cell lines,such as TE671 cells, was not due to lack of infection, asdemonstrated by the TE671 GFP positivity (Figure 4).Detection of the 20 kDa CD40 ligand fragment afterinfection with AdM40 in the supernatant of fusing MMP-rich U87 cells, but not in fusion free, MMP-poor, TE671cells, confirmed the association between MMP cleavageof CD40 ligand and activation of GALV-induced fusion.In MMP-expressing glioma lines the AdM40 cytotoxicitywas dose-dependent with MOIs of 30 being required foreradication of monolayer cell cultures (Figure 5).

Inhibitor studies

In order to prove the specificity of our target-ing approach, we used two different broad-spectrum

MMP inhibitors, 1,10-phenanthroline and N-hydroxy-5,5-dimethylpiperazine-2-carboxamide. They both completelyabolished fusion induced by AdM40 in the three gliomalines, in a dose-dependent manner (Figures 6A and 6B).In contrast, a non-specific serine protease inhibitor, thesoybean trypsin inhibitor, did not affect AdM40-inducedfusion (Figure 6C). No fusion was observed after infectionof the three glioma lines with AdN40, and treatment withthe two MMP inhibitors or the soybean trypsin inhibitorhad no effect on the cells. No fusion was observed afterinfection of the MMP-poor TE671 cells with either AdM40or AdN40, and treatment with the two MMP inhibitors orthe soybean trypsin inhibitor had no effect on the cells(data not shown).

Treatment of U87 glioma xenografts

In order to assess if the antitumor activity of GALV M40and GALV N40 persisted in vivo, we established U87xenografts and treated them with either AdM40 at atotal dose of 1.2 × 1010 pfu or the same dose of AdN40,while a third group of animals served as untreated control(8–9 animals per group). Treatment with AdM40 resultedin statistically significant tumor regression (Table 1):

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

1222 C. Allen et al.

(a)

(b)

Figure 4. (a) Selective activation of GALV and fusion after infection with AdM40 (MOI 10) in overexpressing U87 cells (A),but not in MMP-poor TE671 cells (C), despite similar levels of infection as indicated by comparable GFP positivity. (b) Westernimmunoblotting of cell supernatant with an anti-CD40 ligand antibody demonstrated a 20 KDa band, corresponding to the cleavedCD40 ligand in the supernatant of the AdM40 infected cells and fused MMP-rich U87 cells, but not in AdN40 infected U87 cells orin TE671 cells

Table 1. Volume (cm3) on day 26

Untreated AdN40 AdM40 p value

Volume 26:Mean ± SD 0.1234 ± 0.1033 0.0510 ± 0.0289 0.0258 ± 0.0365 0.003Median 0.0704 0.0400 0.0158IQR 0.0656–0.01696 0.0320–0.0625 0.000–0.0214

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

Ad Vectors Activated via MMP-Cleavable Linkers 1223

0

40

80

120

160

0 1 4 6

U87, AdM40, MOI 1

U87, AdM40, MOI 10

U87, AdM40, MOI 30

U87, AdN40, MOI 1

U87, AdN40, MOI 10

U87, AdN40, MOI 30

Sur

vivi

ng c

ells

(%

)

Days

532

(A)

Sur

vivi

ng c

ells

(%

)

Days

20

40

60

80

100

120

140

0 1 4 6

TE671, AdM40, MOI 1

TE671, AdM40, MOI 10

TE671, AdM40, MOI 30

TE671, AdN40, MOI 1

TE671, AdN40, MOI 10

TE671, AdN40, MOI 30

532

(B)

Figure 5. CPE of AdM40 (MOIs 1, 10, 30) and AdN40 (MOIs 1, 10, 30) in MMP-rich U87 cells (A) and MMP-poor TE671 cells (B).Cell viability is determined by trypan blue exclusion. There is dose-dependent cytotoxicity observed in the MMP-rich U87 cells withless than 5% being viable by day 5 at an MOI of 30. In contrast, AdM40 causes no cytopathic effect in the MMP-poor cell line TE671.Similarly, the AdN40 virus containing the non-cleavable linker causes no cytotoxicity in either cell line

p = 0.003 for the three group comparison, p = 0.01 forcomparison between AdM40- and AdN40-treated animals,p = 0.01 for comparison between AdM40-treated anduntreated animals. Tumor volume comparison amongthe groups was performed on day 26, prior to any of theanimals being euthanized as a result of tumor growth.

In addition, treatment with AdM40 resulted in sta-tistically significant improvement in survival (Figure 7):log-rank p value = 0.002 for the three group compari-son, p = 0.006 for comparison between the AdM40- andAdN40-treated groups, p = 0.001 for comparison betweenthe AdM40-treated and untreated groups. Approximately30% of the treated animals were long-term survivorswhile all animals in the two control groups died orhad to be euthanized around day 40 (Figure 7). H&Estains of the treated tumors at 8 days after completion

of treatment demonstrated extensive cytopathic effectswith syncytial formation (Figure 8), therefore, indicat-ing cleavage of the MMP linker and activation of GALVin vivo by tumor MMPs. No syncytia formation or cyto-pathic effect were observed in the AdN40-treated (viruscontrol) tumor (Figure 8) or in the untreated controltumor. Furthermore, there was no evidence of inflamma-tion or cytopathic effects in the normal tissue surroundingAdM40-treated tumors.

Discussion

The truncated hyperfusogenic version of the gibbon apeleukemia virus (GALV) glycoprotein has emerged asa novel therapeutic transgene, with antitumor activity

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

1224 C. Allen et al.

Syn

cytia

l ind

ex

1,10 phenanthroline (µM)

0

40

80

120

160

0 3.125 6.25 8 10 12.5 25

AdM40

AdN40S

yncy

tial i

ndex

N-hydroxy-5, 5-dimethyl-piperazine-2-carboxamide (µM)

0

20

40

60

0 10

AdM40

AdN40

(A)

(B)

(C)

Syn

cytia

l ind

ex

Soybean trypsin inhibitor (mg/mL)

0

20

40

60

80

0 0.1 1

AdM40

AdN40

Figure 6. Effect of MMP inhibitors on fusion induced by AdM40 in the MMP-rich cell line U87. The MMP inhibitors1,10-phenanthroline (A) and N-hydroxy-5,5-piperazine-2-carboxamide (B) and the serine protease inhibitor soybean trypsininhibitor (C) were added 4 h after infection of the MMP-rich U87 cells with AdM40 (MOI 30). There was a dose-dependentinhibitory effect on the AdM40-induced fusion, with inhibitory concentrations of 1,10-phenanthroline 12.5 µM or higher andinhibitory concentrations of N-hydroxy-5,5-piperazine-2-carboxamide 10 µM or higher causing complete inhibition of fusion. Incontrast, the non-specific serine protease inhibitor soybean trypsin inhibitor had no effect on fusion, thus indicating specificity ofthe observed effect. In contrast, no fusion was observed after infection with AdN40 at the same MOI, and the two MMP inhibitorsand the soybean trypsin inhibitor had no effect on the cells (A, B, C)

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

Ad Vectors Activated via MMP-Cleavable Linkers 1225

Ani

mal

s su

rviv

ing

(%)

Follow-up (days)

P = 0.002

0

20

40

60

80

100

20 30 40 50

AD M40AD N40

Untreated

Figure 7. U87 xenografts were established in Balb/c nude mice(8–9 animals per group). When tumors reached a diameterof 0.3–0.4 cm, the animals were treated with either AdM40(total dose 1.2 × 1010 pfu) or the same dose of AdN40, whilea third group remained untreated. Treatment with AdM40caused significant tumor regression (p = 0.003, see Table 1)and improvement of survival (log-rank p = 0.002) as comparedwith untreated mice or mice treated with the control adenovirusAdN40 containing a non-cleavable linker

having been demonstrated in a variety of animal modelsincluding gliomas [1], HT-1080 fibrosarcoma [3], Mel-625 melanoma [19], and Hep 3B hepatoma [20].Nevertheless, the receptor for GALV, Pit-1, is ubiquitouslyexpressed in primate cells [21–23], therefore leading toheterologous fusion of normal cells. The latter can limitthe applicability of GALV-based gene therapy approaches.

Construction of adenoviruses expressing unmodifiedGALV is technically complicated by death of the producerembryonic kidney 293A cells, as a result of expression

of the Pit-1 receptor. The only prior attempt that wasable to overcome this problem included the use ofthe modified human HSP70b promoter, a hyperthermia-responsive promoter [24]. Demonstration of antitumoractivity in vivo in this system required heating at 44 ◦Cfor 30 min. An alternative approach could involve the useof a tissue-specific promoters [19]. We chose to exploittransgene targeting, using an MMP-cleavable linker ratherthan transcriptional targeting as our GALV-targetingstrategy, when generating the adenoviral vectors. Inaddition to avoiding potential difficulties associatedwith transcriptional targeting such as variable promoterexpression and leakage, this strategy can increase theapplicability of the approach, given the frequent MMPoverexpression in a variety of common malignancies(such as colon, breast, pancreatic, ovarian cancer andCNS tumors) [25–29] and the low MMP levels in normaltissues. 293A cytotoxicity did not represent a problemwhen constructing our retargeted viruses since 293Ahuman embryonic kidney cells do not produce MMPs.The lack of activation of the cytotoxic transgene duringthe production process allowed viral production in hightiters ranging from 2.5–5.5 × 1010 pfu/ml for AdM40 and2–7 × 1010 pfu/ml for AdN40.

Using the targeted adenoviral vectors AdM40 andAdN40, we demonstrated that the cytopathic effect of theGALV fusogenic glycoprotein when used as a therapeutictransgene can be targeted against MMP-overexpressingtumors and spare normal cells that have minimal or noMMP expression. That was achieved by sterically blockingGALV recognition of the receptor with a trimeric ligand,the N-terminus of CD40 ligand, attached to the C-terminusof GALV via the MMP-cleavable linker GGPLGLWAGG.The non-cleavable linker GGGS was employed instead ofan MMP linker in the control constructs. We chose gliomas

AdM40 Treated Tumor

AdN40 Treated Tumor

Figure 8. Examination of an AdM40-treated tumor, 8 days post-completion of treatment, revealed replacement of the AdM40-treatedtumor by massive syncytia. Examination of an AdN40-treated (virus control) tumor showed no cytopathic changes (H + E, 400×)

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

1226 C. Allen et al.

as our therapeutic target because of the high-frequencyMMP overexpression in gliomas. Of equal importance,high-grade gliomas for which limited therapeutic optionsexist are more frequent MMP overexpressors, as comparedto low grade gliomas [9–12]. We showed that theAdM40 adenovirus expressing the MMP-cleavable GALVconstruct can result in selective cytotoxicity against MMP-overexpressing glioma cells, but not in normal astrocytesor MMP-poor lines such as the rhabdomyosarcoma cellline TE671. The cleaved CD40 ligand was detectedby Western immunoblotting after AdM40 infection inthe supernatant of fused MMP-rich U87 cells, but notinfected, non-fusing MMP-poor, TE671 cells, confirmingthat GALV-induced fusion was dependent on the cleavageof the MMP-cleavable linker. Use of two differentMMP inhibitors, 1,10-phenanthroline and N-hydroxy-5,5-dimethylpiperazine-2-carboxamide, resulted in inhibitionof AdM40-induced fusion in a dose-dependent manner.In contrast, use of the soybean trypsin inhibitor (a non-specific serine/protease inhibitor) had no effect on fusion,thus proving the specificity of the observed effect.

The GGPLGLWA linker employed in the constructionof our adenoviral vectors can be cleaved by a varietyof MMPs, both soluble and membrane-based [17,18]. Incell mixing experiments, our group has recently shownthat mixing of MMP-poor TE671 cells transfected withGALV M40 plasmid constructs with MMP-overexpressinguntransfected U87 glioma lines led to partial restorationof fusion. Use of U87 supernatant did result in asimilar effect, thus indicating that soluble MMPs in cellsupernatant can activate the GALV M40 construct bycleaving the blocking ligand, after expression of thechimeric protein in the cell surface [8]. Furthermore,establishment of stable transfectants expressing themembrane type MT1-MMP and MT2-MMP did restorefusion in the MMP-poor cell line TE671 after transfectionwith GALV M40, pointing to the fact that membrane MMPscan also activate the MMP-cleavable GALV containingconstructs.

In order to investigate the therapeutic potential ofour approach, we tested our targeted adenovirusesin vivo, a definitely much more complex environment,given the presence of not only proteases, but also avariety of protease inhibitors. Significant therapeuticbenefit was observed in a U87 glioma xenograft modelusing the targeted AdM40 virus. Statistically significanttumor regression and improvement of survival wasobserved in the group of animals treated with AdM40,as compared with AdN40-treated animals (p = 0.01 and0.006, respectively) or untreated controls (p = 0.01 and0.001, respectively). Histological examination of AdM40-treated tumors showed significant cytopathic effect withdiffuse tumor replacement by syncytia, thus indicatingthat cleavage of the blocking ligand with activation ofGALV and induction of fusion can effectively occur inan MMP-rich in vivo environment. Nevertheless, despitethe improvement of survival in the AdM40-treated group,cure occurred in a smaller percentage (approximately30%) of the treated animals, thus pointing to the

importance of further optimization of this strategy. Oneof the most common problems associated with use ofnon-replicating adenoviral vectors pertains to transgeneexpression only in a narrow radius of a few mmfrom the injection site, which compromises antitumorefficacy [30]. Development of conditionally replicatingadenoviruses expressing MMP-cleavable constructs couldpossibly further augment the antitumor efficacy, andcontribute an additional element of tumor specificity.Another possibility would be the use of a differentMMP-cleavable linker. Schneider et al. recently publisheda novel approach that allows the selection of MMP-activatable retroviruses from viral libraries displayingcombinatorially diversified protease substrates [31].Using this approach they have demonstrated that theselected PQGLYQ peptide was cleaved by MMP-2 witha significantly higher turnover rate that the PLGLWApeptide (also used in our construct). It remains to beseen, however, if increased cleavage in this setting willtranslate into increased therapeutic efficacy.

In summary, our results indicate that GALV targetingusing matrix metalloprotease cleavable linkers canbe accomplished with high specificity both in vivoand in vitro, by employing adenoviruses expressingthe targeted transgene. Given the frequent MMPoverexpression in gliomas and other solid tumors, anMMP-based GALV-targeting strategy can further improvethe therapeutic index potential of GALV-based genetransfer approaches.

Acknowledgements

This research was supported by CA 84388-01 (EG).

References

1. Galanis E, Bateman A, Johnson K, et al. Use of viral fusogenicmembrane glycoproteins as novel therapeutic transgenes ingliomas. Hum Gene Ther 2001; 12: 811–821.

2. Bateman A, Bullough F, Murphy S, et al. Fusogenic membraneglycoproteins as a novel class of genes for the local andimmune-mediated control of tumor growth. Cancer Res 2000;60: 1492–1497.

3. Diaz RM, Bateman A, Emiliusen L, et al. A lentiviral vectorexpressing a fusogenic glycoprotein for cancer gene therapy.Gene Ther 2000; 7: 1656–1663.

4. Olah Z, Lehel C, Anderson WB, et al. The cellular receptor forgibbon ape leukemia virus is a novel high affinity phosphatetransporter. J Biol Chem 1994; 269: 25 426–25 431.

5. Forestell SP, et al. Novel retroviral packaging cell lines:complementary tropisms and improved vector production forefficient gene transfer. Gene Ther 1994; 4: 600–610.

6. Morling FJ, Peng KW, Cosset F-L, et al. Masking of retroviralenvelope functions by oligomerizing polypeptide adaptors.Virology 1997; 234: 51–61.

7. Peng K-W, Vile RG, Cosset F-L, et al. Selective transductionof protease-rich tumors by matrix-metalloproteinase-targetedretroviral vectors. Gene Ther 1999; 6: 1552–1557.

8. Johnson KJ, Peng KW, Allen C, et al. Targeting the cytotoxicityof fusogenic membrane glycoproteins in gliomas thoughprotease-substrate interaction. Gene Ther 2003; 10: 725–732.

9. Nakada M, Nakamura H, Ikeda E, et al. Expression andtissue localization of membrane-type 1, 2 and 3 matrixmetalloproteinases in human astrocytic tumors. Am J Pathol1999; 154: 417–428.

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.

Ad Vectors Activated via MMP-Cleavable Linkers 1227

10. Lampert K, Machein U, Machein MR, et al. Expression of matrixmetalloproteinase and their tissue inhibitors in human braintumors. Am J Pathol 1998; 153: 429–437.

11. Apodaca G, Rutka JT, Bouhara K, et al. Expression ofmetalloproteinases and metalloproteinase inhibitors by fetalastrocytes and glioma cells. Cancer Res 1990; 50:2322–2329.

12. Yamamoto M, Mohanam S, Sawaya R, et al. Differentialexpression of membrane-type matrix metalloproteinase and itscorrelation with gelatinase A activation in human malignantbrain tumors in vivo and in vitro. Cancer Res 1996; 56: 384–392.

13. Graham FL, Prevec L. Adenovirus-based expression vectors andrecombinant vaccine. Biotechnology 1992; 20: 363–390.

14. Uhm JH, Dooley NP, Villemure JG, et al. Glioma invasionin vitro: regulation by matrix metalloprotease-2 and proteinkinase C. Clin Exp Metastasis 1996; 14: 421–433.

15. Miller CR, Buschsbaum DJ, Reynolds PN, et al. Differentialsusceptibility of primary and established human glioma cells toadenovirus infection: targeting via the epidermal growth factorreceptor achieves fiber receptor-independent gene transfer.Cancer Res 1998; 58: 5738–5748.

16. Fueyo J, Alemany R, Gomez-Manzano C, et al. Preclinicalcharacterization of the antiglioma activity of a tropism-enhancedadenovirus targeted to the retinoblastoma pathway. J NatlCancer Inst 2003; 95: 652–660.

17. Ye Q-Z, Johnson LL, Yu AE, et al. Reconstructed 19kDa catalyticdomain of gelatinase A is an active proteinase. Biochemistry1995; 34: 4702–4708.

18. Will H, Atkinson SJ, Butler GS, et al. The soluble catalyticdomain of membrane type 1 matrix metalloproteinase cleavesthe propeptide of progelatinase A and initiates autocatalyticactivation. J Biol Chem 1996; 271: 17 119–17 123.

19. Emiliusen L, Gough M, Bateman A, et al. A transcriptionalfeedback loop for tissue-specific expression of highly cytotoxicgenes which incorporates an immunostimulatory component.Gene Ther 2001; 8: 987–998.

20. Higuchi H, Bronk SF, Bateman A, et al. Viral fusogenicmembrane glycoprotein expression causes syncytia formation

with bioenergetic cell death: Implications for gene therapy.Cancer Res 2000; 60: 6396–6402.

21. Johann SV, Gibbons JJ, O’Hara B. GLVR1, a receptor for gibbonape leukemia virus, is homologous to a phosphate permease ofNeurospora crassa and is expressed at high levels in the brainand thymus. J Virol 1992; 66: 1635–1640.

22. Kavanaugh MP, Miller DG, Zhang W, et al. Cell-surfacereceptors for gibbon ape leukemia virus and amphotropicmurine retrovirus are inducible sodium-dependent phosphatesymporters. Proc Natl Acad Sci U S A 1994; 91:7071–7075.

23. Palmer G, Manen D, Bonjour JP, et al. Characterization of thehuman Glvr-1 phosphate transport/retrovirus receptor gene andpromoter region. Gene 1999; 226: 25–33.

24. Brade AM, Szmitko P, Ngo D, et al. Heat-directed tumor cellfusion. Hum Gene Ther 2003; 14: 447–4651.

25. Stetler-Stevenson WG, Aznavoorian S, Liotta LA, et al. Tumorcell interactions with the extracellular matrix during invasionand metastasis. Annu Rev Cell Biol 2003; 9: 541–573.

26. Jones L, Ghaneh P, Humphreys M, et al. The matrixmetalloproteinases and their inhibitors in the treatment ofpancreatic cancer. Ann NY Acad Sci 1999; 880: 288–307.

27. Moser TL, Young TN, Rodriguez GC, et al. Secretion ofextracellular matrix-degrading proteinases is increased inepithelial ovarian carcinoma. Int J Cancer 1994; 56:552–559.

28. Hewitt RE, Leach IH, Powe DG, et al. Distribution of collagenaseand tissue inhibitor of metalloproteases (TIMP) in colorectaltumors. Int J Cancer 1991; 49: 666–672.

29. Ueno H, Nakamura H, Inoul M, et al. Expression and tissuelocalization of membrane-types 1, 2, and 3 matrixmetalloproteases in human invasive breast carcinomas. CancerRes 1997; 57: 2055–2060.

30. Lang FF, Bruner JM, Fuller GN, et al. Phase I trial of adenovirus-mediated p53 gene therapy for recurrent glioma: biological andclinical results. J Clin Oncol 2003; 21: 2508–2518.

31. Schneider RM, Medvedovska Y, Hartl I, et al. Directed evolutionof retroviruses activable by tumor-associated matrixmetalloproteases. Gene Ther 2003; 10: 1370–1380.

Copyright 2004 John Wiley & Sons, Ltd. J Gene Med 2004; 6: 1216–1227.