J Mol Cell Cardiol 33, 575–580 (2001)

doi:10.1006/jmcc.2000.1322, available online at http://www.idealibrary.com on

Rapid Communication

Phosphodiesterase Inhibitor-mediatedPotentiation of Adenovirus Delivery toMyocardiumKoichi Nagata1, Eduardo Marban, John H. Lawrence2 andJ. Kevin DonahueInstitute of Molecular Cardiobiology, Johns Hopkins University School of Medicine, Baltimore,MD 21205, USA

(Received 4 December 2000, accepted in revised form 12 December 2000, published electronically 9 February 2001)

K. N, E. M, J. H. L J. K. D. Phosphodiesterase Inhibitor-mediated Potentiationof Adenovirus Delivery to Myocardium. Journal of Molecular and Cellular Cardiology (2001) 33, 575–580. Currentgene therapy models are limited by inadequate vector delivery. Increases in microvascular permeability havebeen shown to improve adenovirus-mediated gene transfer to ex vivo and in vivo models. We explored theintracellular mechanism underlying the permeabilizing effects of vascular endothelial growth factor (VEGF). Usingan ex vivo model of coronary perfusion in rabbits, we found a dose–response relationship between VEGF and theefficiency of adenoviral gene transfer. Inhibitors of nitric oxide synthase and guanylate cyclase prevented theVEGF effect, and analogues of nitric oxide and cGMP mimicked the effect. Co-administration of phosphodiesterase-5 inhibitors and VEGF caused a synergistic increase in gene delivery. These results can be readily applied toexisting models to further optimize vector delivery for gene therapy. 2001 Academic Press

K W: Gene therapy; VEGF; Adenovirus; Vascular Permeability; Phosphodiesterase inhibitors.

artery cross-clamping and achieved gene transferIntroductionin approximately 50% of ventricular myocytes ina patchy, heterogeneous distribution.4 So far, no inPrior attempts at in vivo myocardial gene transfer

have had limited success, in part due to the poor vivo delivery system has proven capable of affectinga majority of myocytes in a homogeneous fashion.efficiency of vector delivery. The initial delivery

method utilized injection of adenovirus directly into Using ex vivo and in vitro models, we have studiedthe parameters important for adenovirus-mediatedthe myocardium and achieved high density gene

transfer, but only within close proximity to the gene transfer to cardiac myocytes.5,6 We found thatthe percentage of affected cells varied with coronaryneedle track.1 Intracoronary catheterization models

succeeded in broadening the distribution of virus flow rate, virus contact time and virus con-centration. A significant improvement in genedelivery, but the percentage of infected cells within

the target area was low. The best result was 30% transfer efficiency was noted with modulation ofmicrovascular permeability. Since publication ofof ventricular myocytes in a rabbit model.2 A sub-

sequent study, however, delivered a similar amount those reports, the concept of improving gene trans-fer by increasing vascular permeability has beenof virus after visualization and needle puncture

of a coronary artery and <1% of myocytes were validated in two in vivo models.7,8 The originalmethod for increasing permeability used serotonininfected.3 A third delivery system infused virus into

rat left ventricles during aortic and pulmonary exposure, but the permeabilizing effects of serotonin

1 Current address: Tanabe Seiyaku Co., Ltd., 2-2-50, Kawagishi, Toda-shi, Saitama, 335 Japan.2 Current address: Bristol Myers Squibb, Inc., P.O. Box 4000, Mail Stop J44-09, Princeton, NJ 08543.Please address all correspondence to: J. Kevin Donahue. E-mail: kdonahue@jhmi.edu

0022–2828/01/030575+06 $35.00/0 2001 Academic Press

K. Nagata et al.576

Materials and Methods

Adenovirus vectors

Ad�gal was a gift from Dr Frank Graham; the vectorcontained the Escherichia coli lacZ gene driven by thehuman cytomegalovirus immediate early promoter.Virus stock expansion and quality control wereperformed as previously described.6

Langendorff infection

Adult New Zealand white rabbits (2–3 kg) receivedheparin anticoagulation (1000 units i.v.) prior topentobarbital (50 mg/kg i.v.). Each heart was ex-tracted and rinsed twice in ice-cold, modified Krebsbuffer containing 137 mmol/l NaCl, 5.4 mmol/lKCl, 1.2 mmol/l MgSO4, 1.0 mmol/l CaCl2,1.2 mmol/l NaH2PO4, 20 mmol/L HEPES, and15 mmol/l glucose, saturated with O2 at pH 7.4.Next, the aorta was cannulated, and the heart wassuspended in an insulated chamber at 35–37°C.During the subsequent experimental ma-nipulations, heart rate, coronary flow and intra-

Figure 1 Schematic of the intracellular cascade me- aortic pressure were monitored at 10-min intervals.diating VEGF effect on microvascular permeability (com- Langendorff perfusion occurred by retrogradepiled from refs10–14). flow from the cannula in the ascending aorta to

the coronary arteries. Each heart was first perfusedwith 20 ml Krebs buffer at approximately 30 ml/min. After initial perfusion, the heart was pretreatedwith one of the following (Fig. 2): (1) 60 ml Krebsbuffer with 0.3–1.0 nmol/l VEGF165 (R & D Systems,are long-lasting, while the need for increased per-

meability during gene delivery is brief (e.g. a few Minneapolis, MN, USA) over 2 min; (2) 60 ml Krebsbuffer with 90 �mol/l nitroglycerin (TNG, Abbott,minutes).9 Accordingly, we sought a method for

increasing virus access to myocytes that would Chicago, IL, USA) over 2 min; (3) 60 ml Krebs bufferwith 100 �mol/L 8-Br-cGMP (Sigma, St Louis, MO,have efficacy similar to serotonin but that would

also be self-limited. USA) over 2 min; (4) 60 ml of Krebs buffer with300 �mol/l NG-monomethyl--arginine (L-NMMA,In the coronary circulation, VEGF increases

microvascular permeability sufficiently for the es- Sigma) over 2 min followed by 60 ml of Krebs bufferwith 300 �mol/l -NMMA and 1.0 nmol/l VEGFcape of small proteins, such as bovine serum al-

bumin.10 The duration of the permeability effect is over 2 min; (5) 60 ml of Krebs buffer with300 �mol/l 1H-[1,2,4]Oxadiazolo[4,3-a]quinoxalin-approximately 8–10 min, with peak effects oc-

curring 1–5 min after exposure. The signaling cas- 1-one (ODQ, Calbiochem, San Diego, CA, USA)over 2 min followed by 60 ml of Krebs buffer withcade that mediates the VEGF effect on vascular

permeability includes activation of phospholipase 300 �mol/l ODQ and 1.0 nmol/l VEGF over 2 min;(6) 450 ml of Krebs buffer with 10 �mol/l of eitherC�, protein kinase C�, constitutive nitric oxide syn-

thase, guanylate cyclase, and cGMP-dependent pro- sildenafil (Pfizer, New York, NY, USA), or zaprinast(Sigma), or T-1032 (Tanabe, Osaka, Japan) overtein kinases (Fig. 1).10–14 In this report, we show

that VEGF-mediated increases in microvascular per- 15 min; (7) 450 ml of Krebs buffer with 10 �mol/lof either sildenafil, zaprinast, or T-1032 over 15 minmeability suffice to allow virus particles to escape

from the vasculature, and that various elements of followed by 60 ml of Krebs buffer with 0.3 nmol/lVEGF and 10 �mol/l of sildenafil, zaprinast, or T-the intracellular signaling pathway can be exploited

to potentiate gene transfer to cardiac myocytes. 1032 over 2 min.

Phosphodiesterase Inhibitors and Gene Therapy 577

Figure 2 Flow diagram of the experimental protocol.

Following pretreatment, the heart was infected number of viable cells with each isolation, quan-tified by counting morphologically normal myo-with Krebs buffer containing 1 mg/ml bovine serum

albumin and 1×108 pfu/ml Ad�gal over 2 min. cytes, were not affected by any of the experimentalmanipulations. In separate control experiments,Experiments utilizing pretreatment also continued

those agents during the infection step. During in- hearts underwent a sham infection (n=2). Noneof the cells from these experiments stained positivefection, the perfusate was collected and recirculated,

and the flow rate was controlled by a peristaltic for �-galactosidase activity.The animals used in this study were maintainedpump (Masterflex, Cole-Palmer Co.). At the end of

the infection interval, the heart was perfused with in accordance with the guiding principles of theAmerican Physiological Society regarding ex-non-recirculating, virus-free Krebs buffer for a total

Langendorff time of 180 min. Following the in- perimental animals. The experimental protocol wasapproved by the Institutional Animal Care and Usefection and washout phases of the experiments,

myocytes were isolated and cultured by perfusing Committee at the Johns Hopkins University.with nominally calcium-free Krebs for 5 min at10–15 ml/min, followed by an enzyme solution

Reporter gene assays and data analysisconsisting of Krebs buffer with 25 �mol/l CaCl2,1 mg/ml collagenase B (Boehringer-Mannheim, In-

Forty-eight hours after exposure to Ad�gal, thedianapolis, IN, USA), 0.1 mg/ml protease (fraction

myocytes were fixed in 0.05% glutaraldehyde forXIV, Sigma), 60 mmol/l taurine, 8.0 mmol/l glu-

5 min at room temperature. The cells were thentamic acid, and 2.0 mmol/l ,-carnitine. After di-

washed three times in phosphate-buffered salinegestion, the ventricles were excised, minced, and

(PBS) and stained overnight at 37°C in PBS con-agitated by repeated suction through a Pasteur

taining 1.0 mg/ml 5-bromo, 4-chloro, 3-indolyl-�-pipette. The resulting cell suspension was strained

-galactopyranoside (X-gal), 15 mmol/l potassiumthrough a 200-�m nylon mesh filter. The ventricles

ferricyanide, 15 mmol/l potassium ferrocyanide andwere uniformly digested, and the residue remaining

1 mmol/l MgCl2. Four hundred cells were countedon the filter was from the valvular and tendon

for each experiment. Myocyte survival at 48 h didstructures, with only minimal myocardium. Cal-

not vary with changes in the virus concentrationcium was gradually repleted in four steps before

or experimental protocol.the cells were added to medium 199 (Mediatech,Herndon, VA, USA) supplemented with 5.0 mmol/l creatine, 5.0 mmol/l ,-carnitine, 5.0 mmol/l Statistical analysistaurine, 100 units/ml penicillin and 0.1 mg/mlstreptomycin, and placed in laminin-coated dishes, Unless otherwise stated, all experiments were per-

formed in triplicate and the data are presented asin a 37°C incubator. The myocardial digestion and

K. Nagata et al.578

mean±... Statistical significance was de-termined at the 5% level using the unpaired Stu-dent’s t-test.

Results and Discussion

In the ex vivo perfused rabbit heart, VEGF exposureincreased the efficiency of gene transfer in a dose-dependent fashion [Fig. 3(a)]. Control perfusion for2 min with Kreb’s solution prior to a 2-min per-fusion with 1×108 pfu/ml Ad�gal resulted in �-galactosidase (�gal) gene transfer to 1.8±0.4% ofventricular myocytes. Exposure to 0.1 nmol/l VEGFbefore and during Ad�gal delivery increased thepercentage of myocytes to 4.9±0.4% (P=0.005relative to control). Further increases were seenwith 0.3 and 1.0 nmol/l doses (0.3 nmol/l VEGF8.2±0.4%, 1.0 nmol/l VEGF 18.6±0.9%).

Evaluation of the intracellular cascade responsiveto VEGF-receptor binding revealed increases in genetransfer by administration of downstream par-ticipants in the cascade and attenuation of theFigure 3 Effects of VEGF signaling pathway on gene

transfer. A: VEGF perfusion shows a dose–response re- VEGF effects by blockade of the corresponding en-lationship with the percentage of myocytes expressing �- zymes [Fig. 3(b)]. Exogenous administration ofgalactosidase after perfusion with 1×108 pfu/ml Ad�gal either nitric oxide via nitroglycerin (TNG) perfusionfor 2 min. B: Perfusion with 300 �mol/l -NMMA or ODQ

or cGMP from 8-Br-cGMP adminstration resultedin the presence of 1.0 nmol/l VEGF attenuates the VEGFin improved gene delivery. After 2-min exposure toeffect on gene transfer. Perfusion with 90 �mol/l TNG or

100 �mol/l 8-Br-cGMP duplicates the VEGF effect on gene 90 �mol/l TNG and 2-min infection with 1×108

transfer (n=3 for each group). pfu/ml Ad�gal in Krebs buffer with TNG, 36.2±8%of cells expressed �gal. A similar protocol using100 �mol/l 8-Br-cGMP resulted in gene transfer to31.3±1.0% of myocytes. Blockade of nitric oxidesynthase with 300 �mol/l -NMMA reduced theeffect of 1.0 nmol/l VEGF from 18.6±0.9% to6.2±1.3% (P=0.001). A similar decrement in theVEGF effect occurred with guanylate cyclase block-ade by 300 �mol/l ODQ (5.8±2.9%, P=0.008).

Since cGMP is an identified part of the in-tracellular cascade mediating VEGF effects, andincreasing cellular cGMP levels duplicated the VEGFresponse, we hypothesized that slowing the break-down of cGMP would further enhance the effects ongene transfer. Accordingly, we perfused the rabbithearts with inhibitors of phosphodiesterase 5 (PDE-5) to reduce enzymatic breakdown of cGMP.15 Inthe absence of VEGF, sildenafil, zaprinast and T-Figure 4 Augmentation of VEGF effect by perfusion with1032 caused modest improvement in gene transferPDE-5 inhibitors. In the absence (C) or presence (Ε) of

VEGF, hearts were perfused with 10 �mol/l of the PDE- (Fig. 4). After 15 min perfusion with 10 �mol/l of5 inhibitors zaprinast, sildenafil, or T-1032. Each agent each agent, 2-min infection with Ad�gal resultedcaused a significant increase in gene transfer in the in gene transfer to 5.4±0.8% of myocytes forabsence of VEGF (n=3, P<0.005 for each). Co-ad-

zaprinast (P=0.004 relative to control), 9.4±0.8%ministration of VEGF with the PDE-5 inhibitors increasedfor sildenafil (P<0.001), and 12.4±1.2 for T-1032gene transfer beyond that seen by the additive effects of

perfusing with VEGF or PDE-5 inhibitor alone. (P<0.001).

Phosphodiesterase Inhibitors and Gene Therapy 579

When combined with VEGF, these agents ex- tor access to the target organ, we are able to stackinterventions and increase the efficiency of genehibited a synergistic effect to improve gene transfer.

A 15 min perfusion with 10 �mol/l zaprinast fol- transfer. Further development of gene transfermodels for use in research or clinical practice willlowed by a 2-min exposure to 0.3 nmol/l VEGF

with continued zaprinast and then a 2-min infection require continued innovation to improve theefficiency of vector delivery.with 1×108 pfu/ml Ad�gal in Krebs solution with

zaprinast and VEGF resulted in �gal expression in21±7% of myocytes (P=0.03 relative to 0.3 nmol/l VEGF). Similar protocols with 10 �mol/l sildenafil Acknowledgementsand T-1032 caused �gal expression in 34±9% (P=0.007) and 47±10% (P=0.003), respectively. This work was supported by the Richard S. Ross

Previous investigations into the VEGF-mediated Clinician-Scientist Award, Johns Hopkins Universityincrease in vascular permeability have focused on (JKD), the NIH (P50 HL 52307, JKD and EM) andthe leak of organic dyes and small proteins.10–14

by a grant from the Tanabe Seiyaku Co., Ltd. EMThose studies defined portions of the intracellular holds the Michel Mirowski, M.D. Professorship ofpathway underlying the observed VEGF effects. Our Cardiology at the Johns Hopkins University.results extend those findings by demonstrating thatthe permeability increase is sufficient to allow virusparticles to escape the vasculature. We found a

Referencesdose–reponse relationship between VEGF con-centration and gene transfer using the �-galacto-

1. G RJ, L P, C R, E SE,sidase reporter gene. This resultant increase in geneF T. Efficient gene transfer into myocardium

transfer can be replicated by exogenous NO or cGMP by direct injection of adenovirus vectors. Circ Resadministration and can be blocked by inhibition of 1993; 73: 1202–1207.

2. B E, C J, K AM, T SK,either NOS or GC. Furthermore, combined ad-K K, W JM, L J. Efficient catheter-ministration of a PDE-5 inhibitor and VEGF has amediated gene transfer into the heart using rep-synergistic effect on vascular permeability. The endlication-defective adenovirus. Gene Therapy 1994; 1:

result is improvement in the conditions necessary 51–58.for myocardial gene transfer. 3. M J, J M, Y I, C C,

L P, G TR, B B, S S,Previous work has shown the efficacy of PDE-5C R, C MC. Safety and efficacy ofinhibitors for vasodilation in a variety of vascularin vivo gene transfer into the porcine heart withbeds.15 Our data demonstrate that these agents alsoreplication-deficient, recombinant adenovirus vec-

increase vascular permeability. Since the action of tors. Gene Therapy 1996; 3: 145–153.PDE-5 inhibitors is to inhibit breakdown of cGMP, 4. H R, S U, M T, G L, L K,

G J, D G, S M, R A.we presume that the mechanism of PDE-5 inhibitorModulation of ventricular function through geneaction is to amplify the endogenous cGMP, causingtransfer in vivo. Proc Natl Acad Sci USA 1998; 95:a small increase in vascular permeability when5251–5256.

given alone. This hypothesis is supported by the 5. D J, K K, J D, M E,synergistic actions of VEGF and PDE-5 inhibitors. L J. Ultrarapid, highly efficient viral gene

transfer to the heart. Proc Natl Acad Sci USA 1997;VEGF would act at a more proximal portion of the94: 4664–4668.signaling pathway, and then the PDE-5 inhibitors

6. D J, K K, T AD, M E,would inhibit breakdown of the cGMP formed byL J. Acceleration of widespread adenoviral

VEGF exposure. Further investigation of the par- gene transfer to intact rabbit hearts by coronaryticipants in this signaling pathway will be needed perfusion with low calcium and serotonin. Gene

Therapy 1998; 5: 630–634.to fully define the mechanism of PDE-5 inhibitors.7. W H, L K, M A, K K,Attempts to develop myocardial gene transfer

S C, B S, S M, B T, Smodels in large mammals have had limited success.A, U M. Enhanced cardiac contractility after

In general, the delivery systems reported in the gene transfer of V2 vasopressin receptors in vivo byliterature include widespread delivery in small ultrasound-guided injection or transcoronary de-

livery. Circulation 2000; 101: 1578–1585.mammal models or focal delivery in large mam-8. D J, H A, F H, MD A,mals.4,8 So far, the ability to deliver a transgene

M J, R J, E T, M E. Focalhomogeneously to a majority of cells in a targetmodification of electrical conduction in the heart by

organ has eluded investigators. This report adds viral gene transfer. Nature Med 2000; 6: 1395–1398.another piece to that puzzle. By defining the in- 9. B I, K T, B M, V K. Effects of

bradykinin, histamine and serotonin on pulmonarytracellular pathways responsible for improving vec-

K. Nagata et al.580

vascular resistance and permeability. Acta Physio- cium in VEGF-induced venular hyperpermeability.Am J Physiol 1999; 276: H535–H542.logica Scandinavica 1997; 159: 189–198.

10. W HM, H Q, Y Y, G HJ. VEGF 13. K P, G DN. Nitric oxide modulates micro-vascular permeability. Am J Physiol 1992; 262:induces NO-dependent hyperpermeability in cor-

onary venules. Am J Physiol 1996; 271: H2735– H611–H615.14. H V. Endothelial permeability forH2739.

11. X P, A L, I H, J Z, P D, R macromolecules. Arterioscler Thromb 1997; 17:1018–1023.G, T H, N W, J M, K G.

Characterization of vascular endothelial growth fac- 15. W R, C J, F S, E P. Tissue dis-tribution of phosphodiesterase families and the effectstor’s effect on the activation of protein kinase C, its

isoforms, and endothelial cell growth. J Clin Invest of sildenafil on tissue cyclic nucleotides, plateletfunction, and the contractile responses of trabeculae1996; 98: 2018–2026.

12. W H, Y Y, Z D, T J, G H. carneae and aortic rings in vitro. Am J Cardiol 1999;83: 3C–12C.Role of phospholipase C, protein kinase C, and cal-