Avoidance of Immune Response Prolongs Expression of Genes Delivered to the Adult Rat Myocardium by Replication-Defective Adenovirus
Background Gene delivery is a rapidly expanding field with potential applications to every human organ system. Recently, adenoviruses have been used as efficient vectors for in vivo gene transfer into the myocardium. These methods, however, have shown a sharp decline of gene expression after 1 week. To test the hypothesis that an immune-effector mechanism is involved in this decline, we compared the results after injection of adenovirus-5 carrying the β-galactosidase gene (Adβ-gal) into the left ventricular myocardium of athymic nude rats (NDRs) versus immunocompetent Sprague-Dawley rats (SDRs).
Methods and Results Adβ-gal (5.0×109 PFU/mL) was injected into the left ventricle of NDRs (n=16) and SDRs (n=22). Hearts were harvested, embedded in paraffin, and sectioned and stained for β-gal activity, hematoxylin and eosin and picrosirius red at 4, 21, 35, 85, and 120 days. Representative samples were immunostained with antibodies directed at inflammatory markers. β-gal activity was quantified by digital planimetry and expressed as area of staining (%±SEM). Peak β-gal activity was highest at 4 days, with NDRs displaying significantly greater staining (83±3.0% versus 54±8.0%; P=.03). SDRs sustained a rapid drop in activity, such that at 35 (1±0.19%) and 85 (1±0.4%) days, only occasional cells stained positive and by 120 days (0.3±0.0%), activity had been extinguished. NDRs continued to show transgene expression at all time periods (35 and 85 days, 25±7.1% and 7.4±2.7%, respectively) and was still readily detected at 120 days. An inflammatory response was limited in NDRs compared with SDRs, in which there was intense mononuclear cell infiltration, with collagen deposition and scar formation. Immunostaining identified the majority of these inflammatory cells as not being of lymphocyte lineage, although small numbers of lymphocytes and phagocytic and activated plasma cells were identified.
Conclusions Our data suggest that immune-effector mechanisms can severely affect the expression of genes delivered by adenovirus. The present model provides efficient gene expression for at least 120 days without significant inflammatory reaction.
Gene delivery for the investigation and/or correction of disease is a rapidly expanding field with potential applications to every human organ system.1 2 To date, a number of strategies have been used for the delivery of desired genes into a variety of hosts and organs. Among these are the direct injection of naked plasmid DNA, ex vivo genetically engineered transplanted cells, liposome-DNA complexes, and several recombinant and conjugated viruses. We3 and other investigators (reviewed in Reference 4) have tested various techniques for gene delivery into the myocardium and have found short-term results to be very encouraging.
Recently, replication-defective Ads have been exploited as gene delivery vehicles because of their ability to easily infect a wide variety of hosts and tissues.5 In addition, Ads have copious cDNA capacities, can easily be grown to high titers, are not known to cause human malignancies, and are highly efficient in infecting slowly replicating or nonreplicating cells.6 This last feature is especially attractive when gene delivery is directed at the adult mammalian heart, because cardiomyocytes are postmitotic cells.7 More recently, in vivo studies in the heart have shown Ads to be highly efficient at both infection and transgene expression, whether introduced by direct or systemic injection or cardiac catheterization.8 9 10 11 12 The resultant gene expression is noticeable within hours and is stable for at least several days.10 However, only in studies conducted in the neonatal mouse heart8 has it been possible to avoid a sharp decline of gene expression that begins about 1 week after transfection.4
The reasons for the sharp decline have not been fully explained. Identifying the precise mechanism of this phenomenon has paramount importance for future Ad-mediated strategies of gene therapy. Previous investigators9 10 13 14 have reported infiltration of inflammatory cells into the area of Ad transfection and have speculated that immunological processes may be involved in the quick transgene decline observed. The purpose of our study, therefore, was to investigate the hypothesis that cell-mediated immune responses are involved in the decline of Ad-mediated gene expression in the adult mammalian heart.
Adenoviral Constructs and Purification of Virus
The recombinant E1a-deleted adenovirus-5 carrying the bacterial β-galactosidase reporter gene (Adβ-gal) was provided by Robert D. Gerard, University of Texas, Southwestern Medical Center, Dallas. This vector was constructed and packaged as previously described.15 Recombinant virus was used to infect and grow in 293 cells as a monolayer. Plates were incubated in 5% CO2 at 37°C until all cells detached. Cells and supernatant were collected and subjected to two cycles of freezing and thawing. Samples were centrifuged, and supernatant was collected and precipitated in 20% polyethylene glycol for 1 hour on ice. Samples were centrifuged, and pellets were resuspended in HBS (20 mmol/L HEPES, pH 7.3; 150 mmol/L NaCl). Viral particles were purified by passage through a G50 Sephadex column (Pharmacia). Recovered virus was titered by infection of 293 cell lines in decreasing concentrations and histochemical staining for β-gal activity.
Animals and Adenoviral Infection
The study was performed in accordance with the guidelines of The Animal Care and Use Committee of the Good Samaritan Hospital, which conform to the policies of the American Heart Association and the Guide for the Care and Use of Laboratory Animals (Department of Health and Human Services, publication [NIH] 86-23).
Female adult immunocompetent SDRs (Charles River, Wilmington, Mass) and NDRs (nu/nu; Taconic Laboratories, NJ) were used for these experiments. All rats were between 6 and 8 weeks of age. Rats were anesthetized with ketamine/xylazine intramuscularly, intubated, and ventilated with room air. Under sterile technique, the chests were surgically opened and the beating hearts exposed. The pericardium was stripped, and the left ventricular myocardium was injected twice with 2.5×109 PFU of Ad (25 μL of a solution of 1×1011 PFU/mL). In addition, 2 SDRs were also injected with 50 μL of sterile normal saline. After intracardiac injection, air was expelled from the chest, and the surgical wounds were sutured closed. SDRs injected with Adβ-gal were killed at 4 (n=4), 21 (n=5), 35 (n=7), 85 (n=4), and 120 (n=2) days after cardiac injection. SDRs injected with normal saline were killed at 4 days (n=2). NDRs were killed at 4 (n=4), 35 (n=7), 85 (n=4), and 120 (n=1) days. At the time of death, hearts were harvested and processed for further analysis as described below.
Evaluation of Gene Transfer and Expression Efficiency
When the animals were killed, hearts were harvested, washed in ice-cold PBS, and immediately frozen and sectioned or embedded in paraffin (see below). Frozen 6-μm sections were fixed in 0.05% glutaraldehyde for 10 minutes and incubated with 2.5 mmol/L X-gal, 5 mmol/L potassium ferricyanide, 5 mmol/L potassium ferrocyanide, and 2 mmol/L MgCl2 in 0.1 mol/L PBS overnight at 37°C. Representative samples were also counterstained with H&E. Sections stained for β-gal were examined by light microscopy and photographed with Kodak Super-200 film. Expression of transfected Adβ-gal was considered positive when dark blue cytoplasmic staining was observed. Quantification of transduced cells was carried out by digital planimetry with a computer-assisted morphometric program (SigmaPlot, Jandel Scientific Software) and a graphics digitizing table (Summasketch II, Summagraphics Corp). Low-power (×4) 5×7-in photographs were used, and the area of injection as well as areas delineated by blue staining was measured. An index of β-gal gene expression was calculated by the ratio of cytoplasmic blue-staining areas to the overall area of myocardial tissue in the photograph. For each rat heart evaluated, planimetry was performed on three contiguous zones, and the sum of the three areas was calculated.
Histological and Immunocytochemical Evaluation
Representative samples from harvested hearts were collected from locations that coincided with those areas taken for frozen sections, fixed in B-5, embedded in paraffin, and cut into 5-μm sections. These samples were stained with H&E or PSR. The stained, paraffin-embedded sections were examined and qualitatively evaluated by light photomicroscopy for inflammatory and architectural changes (H&E) and collagen deposition (PSR). Sections from NDRs and SDRs were further evaluated by immunocytochemical techniques to identify inflammatory cell types infiltrating the areas of Adβ-gal infection: After paraffin sectioning, samples were deparaffinized with xylene and washed with decreasing concentrations of ethanol (100%, 95%, 70%, and 35%). To dissolve mercury crystals, samples were incubated in an iodine solution for 5 minutes followed by 5% sodium thiosulfate. Samples were rehydrated in PBS for 10 minutes and incubated at 37°C for 30 minutes with one of the following primary antibodies: anti–leukocyte common antigen, which stains for all leukocytes, and anti-lysozyme, which stains for cells displaying phagocytic activity (Lipshaw); anti–LN-1, anti–LN-2, and anti–Leu-M-1, which stain for T lymphocytes (Biogenex); anti–HLA-DR, which recognizes lymphocytes (Sanbio); and anti–λ+κ light chain, which stains for activated plasma cells (Signet). Samples were rinsed in PBS and incubated with commercial immunostaining kits (Lipshaw and Biogenex) following the manufacturers' instructions. In brief, the sections were incubated with a multispecific, biotinylated secondary antibody for 15 minutes at 37°C, followed by a PBS rinse and subsequent incubation with streptavidin-peroxidase reagent. After a PBS rinse, the antibody complex was visualized by incubation with 3-amino-9-ethylcarbazole and counterstained with Mayer's hematoxylin.
All variables are expressed as mean±SEM. Data on Adβ-gal transfection efficiencies in SDRs and NDRs were not normally distributed and were compared by nonparametric methods. Time course measurements of β-gal activity were analyzed by Kruskal-Wallis nonparametric ANOVA. Differences between groups in the time periods studied were analyzed by the Mann-Whitney U test, and a value of P<.05 was considered statistically significant.
Gene Transfer and Expression Efficiency
Cardiomyocytes of samples from hearts of SDRs injected with Adβ-gal showed strong β-gal activity 4 days after infection (Fig 1⇓, top left). The staining was intense, with large areas of grossly visible confluence extending at least 8 to 10 mm from the injection site. Transgene activity in the NDRs at 4 days was reproducibly superior, with >80% of cardiomyocytes displaying β-gal staining in the injected site (Fig 2⇓, top left). Control SDRs injected with normal saline showed no detectable staining (Fig 1⇓, bottom right). By 21 and 35 days, staining in the SDRs had sharply declined. At these time periods, there was no grossly visible staining, and only occasional cardiomyocytes displayed evidence of gene expression (Fig 1⇓, top right and middle left). By 85 and 120 days, transgene activity was extinguished in these immunocompetent rats except for a rare staining cell (Fig 1⇓, middle right and lower left). Morphometric analysis of tissues taken at these time points is shown in Fig 3⇓. The morphometric analysis supports the visual data in that activity seen at 35, 85, and 120 days was significantly lower than activity measured at 4 days after infection (Fig 3⇓; P<.01).
In contrast to the rapid and progressive decline of gene expression observed in the SDRs, the NDRs had highly efficient expression throughout the 120 days of observation. At 35 days after Ad infection, large homogeneous patches of myocardial cells staining the typical dark blue color persisted (Fig 2⇑, top right). Although the SDRs showed a significant decrease in activity by 21 and 35 days, the NDRs continued to display brisk activity during this time period. The significant disparity seen at 35 days was also evident at the later study periods, such that at 85 and 120 days after infection, whereas transgene expression was almost nonexistent in the SDRs (Fig 1⇑, middle right and bottom left), expression was still abundant in the NDRs, although the staining was somewhat reduced in intensity (Fig 2⇑, bottom left and right). By planimetry, NDRs had a significantly greater percentage activity at 4 and 35 days compared with SDRs (P=.03 and P<.006, respectively; Fig 3⇑).
Histological Evaluations for Inflammation and Collagen Deposition
All acute (4 days postinjection) SDRs demonstrated an intense inflammatory response that extended deep into the myocardium beyond the injection site (Fig 4⇓, top left). Magnification of these areas identified the infiltrating cells as being primarily mononuclear, with only rare polymorphonuclear leukocytes present (Fig 4⇓, top right). Furthermore, in the long-term SDRs (85 and 120 days after injection), the myocardial inflammatory infiltrations were accompanied by an obvious disruption of normal myocardial architecture (Fig 4⇓, middle left), with significant collagen deposition and scar formation, as was evidenced by strong PSR staining (Fig 4⇓, middle right). These pathological changes were consistent in all SDRs studied.
In contrast, the injected myocardial areas in NDRs were relatively free of mononuclear infiltration and retained near-normal architecture (Fig 4⇑, lower left). In addition, with the exception of one rat heart sample that did display some intramural collagen deposition, PSR staining revealed that most NDR hearts have little collagen deposition or scar formation (Fig 4⇑, lower right).
Control SDR animals injected with normal saline showed minimal inflammatory infiltrations around the needle track at 4 days.
Heart samples from immunocompetent animals that were stained with H&E revealed a severe inflammatory cell infiltration in the areas of Ad injection (Fig 4⇑, top right and left), whereas immunocompromised NDRs showed little or no evidence of inflammation (Fig 4⇑, lower left). Immunostaining of heart samples, taken from inflammatory areas of immunocompetent animals, revealed some positive staining for leukocytes (anti–leukocyte common antigen), lymphocytes (anti–HLA-DR), phagocytic cells (anti-lysozyme), and activated plasma cells (anti–lambda/kappa; Fig 5⇓, top), whereas immunocompromised animals showed little or no activity for any of these inflammatory markers (Fig 5⇓, bottom). Furthermore, although we did find some staining for various lymphocytic cells in the immunocompetent rat hearts, most of the mononuclear cells did not stain with anti–HLA-DR or with the various specific antibodies directed at a variety of T- and B-cell antigens. Interestingly, anti–leukocyte common antigen, specific for white blood cells in general, stained only a small percentage of the inflammatory cells found at the Ad injection site. We conclude, then, that the majority of mononuclear cells constituting the inflammatory reaction were made up of nonidentifiable inflammatory cells and not infiltrating lymphocytes.
The main findings of our study are that Ad-mediated transfected genes in the myocardium of immunodeficient animals had strikingly stronger expression and longer duration of activity than in immunocompetent animals. The injection of Ad into immunocompetent myocardium was accompanied by prominent mononuclear cell inflammation, with the minority of these cells being lymphocytes, phagocytic cells, and activated plasma cells. Furthermore, focal myocarditis developed in an extensive area, larger than the site of injection. This inflammation was followed by an extensive scarring process. By comparing these observations with control animals, we conclude that this immune response was not related to needle trauma but was most probably directed against foreign antigens, ie, viral proteins or foreign genes. In light of these severe inflammatory reactions, the results of our study support the hypothesis that immune mechanisms contribute to the limited efficiency and long-term expression of genes delivered into the myocardium. This immune reaction undoubtedly plays a role in the sharp decline of the expression of genes delivered into the myocardium.
Although a graduated progressive decrease in transgene expression was also observed among the NDRs, the decrease was significantly less at each time point compared with the residual expression in SDRs. This decay of activity in NDRs may also be a result of immune responses, since NDRs are known to retain the ability to mount antibody responses and develop some T-cell activity in later life.16 This possibility has been demonstrated in previous reports17 18 ; however, our studies showed little or no inflammatory cellular infiltration or immunostaining for such cells.
Our observation of efficient but short-lived Ad-mediated transgene expression in the hearts of immunocompetent animals is similar to the findings of others.9 10 11 12 Only one previous study,8 which used neonatal mice, reported gene activity to have lasted for 12 months. French et al,12 using the luciferase firefly reporter gene, found peak expression activity at 7 days, with subsequent declines at 14 and 21 days. Guzmun et al9 found peak expression at 7 days, but expression was completely extinguished at 30 days. Similarly, in the report by Kass-Eisler et al,10 activity was seen within 15 hours, with peak expression at 5 days. At 43 and 55 days, activity was found to be five to six orders of magnitude lower. Barr et al11 delivered Ads into the hearts of adult rabbits by cardiac catheterization and found 32% of cells to be transduced at 5 days. By 1 month, only 0.01% of cells showed activity. In addition, the strong mononuclear inflammatory reactions with disruption of normal architecture that we observed were similar to those previously reported in the myocardium and other organs.9 10 13 14
The rapid but transient transgene expression found in the myocardium is similar to that seen in other organs.5 The mechanisms responsible for this limited expression have not been completely delineated and could result from any one or a combination of factors: eg, that (1) the Ad transgene is episomal, does not integrate into the host genome, and thus could be lost at the time of cell replication or through degradation; (2) the transgene does persist but inactivates; or (3) transduced cells express some virally encoded proteins, with subsequent induction of immune responses. Although loss during cell division is a possibility in other tissues, adult myocardial cells are not capable of replication, and so it is unlikely that the transgene is lost during cell division. A recent report in which polymerase chain reaction assay was used found that a significant number of hearts with extinguished transgene activity had no detectable transgene DNA.11 Similarly, Yang et al,18 using Southern blot analysis, demonstrated that decline of transgene activity in mouse liver was accompanied by loss of viral genome. Both of these studies raised the possibility that immunological responses could be major factors in the transient nature of Ad-mediated expression.
In support of our findings, other investigators have recently reported longer-lived Ad-mediated transgene expression and less inflammatory reaction in the livers and lungs of nude mice compared with immunocompetent animals.17 18 Taken together, these studies strongly suggest that the immune system is at least partially responsible for the limited expression of transgenes when delivered by Ads even when these Ads are engineered to be replication-defective.
Prince et al13 and Ginsberg et al17 suggested that this immunological response is biphasic. In an early nonspecific response, monocytes, macrophages, and polymorphonuclear cells migrate to affected areas, with release of cytokines and other attractants. In a later, more specific phase, there is a predominance of lymphocytes, with T-cell cytotoxic activity leading to cell death and inflammation. The observed inflammatory reaction to Ad injection is similar to those previously reported in animal models of myocarditis induced by group B coxsackie virus.19 20 This myocarditis was also characterized by an early stage of myocyte infiltration with foci of myocardial necrosis. This was followed by a later period evidenced by undetectable virus, mononuclear cell infiltration, and scar formation.19 20 Of interest, Barr et al11 delivered Ad into rabbit hearts by cardiac catheterization and reported no inflammation or myocardial necrosis in the hearts studied. Thus, one may speculate that inflammatory reactions, as we and others observe, are related to local needle trauma. However, in our study, hearts taken from immunocompetent rats exhibited larger inflammation and scarring (much beyond the needle track) than control animals that were treated with saline injection alone. Thus, we conclude that the majority of the inflammation was related to a viral response.
Although the currently used first-generation Ads are E1-deleted and assumed incapable of viral protein production, the possibility exists that some viral proteins are synthesized by basal “leaky“ transcription of Ad promoters even in the absence of E1. Some cells contain nuclear E1A-like activities, although the identity of these proteins is yet to be described.21 22 23 HepG2 cells have been shown to be able to support productive infection of E1A-deleted mutant virus that is enhanced by IL-6.24 Recently, Spergel et al25 reported that NF-IL6 (an intracellular mediator of the IL-6 receptor26 ) is able to complement an E1A-deleted mutant in viral infection and to regulate E1A-responsive promoters in the absence of E1A. Furthermore, humoral immunity may also have an effect on transgene expression. A recent study showed that cotton rats were capable of developing human Ad–neutralizing antibody when challenged with an E1/E3 replication-defective Ad into their respiratory tracts.27 A repeated dose of the Ad resulted in significantly lower levels of gene expression, and the decrease was directly proportional to the serum human Ad–neutralizing antibody titer.27 We speculate that, had we used a severe combined immunodeficient animal that, in addition to having no cell-mediated immune response, has no antibody capabilities, transgene expression might have been stronger and of longer duration.
Our study is limited in that we did not determine whether immune responses were directed against the viral antigens or against the foreign transgene product. However, we3 and others28 have achieved prolonged expression of β-galactosidase in myocardium of immunocompetent rats after direct DNA injection. No evidence of an inflammatory response was seen except along injection-induced needle tracks. This suggests that immune responses observed after Ad infection are directed primarily at viral antigens.
The results of our study also demonstrate the limitations and the problems associated with Ad-mediated gene therapy for the long term. The immune response may eliminate continuous therapeutic effects. Future approaches will most likely involve the development of new recombinant viruses incapable of transcriptional activity29 or antigen presentation. Until such vectors become widely available, immunosuppression may be required as a means to extend transgene expression. Alternatively, a therapeutic window may exist for tissues of interest in which the inflammatory response can be avoided while acceptable transgene activity is still retained.
Selected Abbreviations and Acronyms
|H&E||=||hematoxylin and eosin|
|NDR||=||athymic nude rat|
The authors are grateful to Dr Peter Whittaker for guidance in morphometric analysis. Terry Saluna, Seda Dzhandzhapanyan, and Michelle K. MacVeighthe provided expert technical assistance. This work was supported in part by grants from the NIH (SCOR grant P50-HL-44404). Dr Quin˜ones is a recipient of an NIH training grant (P50-HL-44404-02) as a supplement to the SCOR.
The first two authors contributed equally to the preparation of this manuscript.
- Received July 5, 1995.
- Revision received March 13, 1996.
- Accepted March 26, 1996.
- Copyright © 1996 by American Heart Association
Anderson WF. Human gene therapy. Science. 1992;256:808-813.
Prentice H, Kloner RA, Prigozy T, Christensen T, Newman L, Li Y, Kedes L. Tissue restricted gene expression assayed by direct DNA injection into cardiac and skeletal muscle. J MolCell Cardiol. 1994;26:1393-1401.
Nabel EG. Gene therapy for cardiovascular disease. Circulation. 1995;91:541-548.
Zak R. Development and proliferative capacity of cardiac muscle cells. Circ Res. 1974;35(suppl V):V-17-V-28.
Stratford-Perricaudet LD, Makeh I, Perricaudet M, Briand P. Widespread long-term gene transfer to mouse skeletal muscles and heart. J Clin Invest. 1992;90:626-630.
Guzman RJ, Lemarchand P, Crystal RG, Epstein SE, Finkel T. Efficient gene transfer into myocardium by direct injection of adenovirus vectors. Circ Res. 1993;73:1202-1207.
Kass-Eisler A, Falck PE, Alvira M, Rivera J, Buttrick PM, Wittenberg BA, Cipriani L, Leinwand LA. Quantitative determination of adenovirus-mediated gene delivery to rat cardiac myocytes in vitro and in vivo. Proc Natl Acad Sci U S A. 1993;90:11498-11502.
French BA, Mazur W, Geske RS, Bolli R. Direct in vivo gene transfer into porcine myocardium using replication-deficient adenoviral vectors. Circulation. 1994;90:2414-2424.
Prince GA, Porter DD, Jenson AB, Horswood RL, Chanock RM, Ginsberg HS. Pathogenesis of adenovirus type-5 pneumonia in cotton rats (Sigmodon hispidus). J Virol. 1993;67:101-111.
Herz J, Gerard RD. Adenovirus-mediated transfer of low density lipoprotein receptor gene acutely accelerates cholesterol clearance in normal mice. Proc Natl Acad Sci U S A. 1993;90:2812-2816.
Gomez FAM, Coats WS, Baque S, Alam T, Gerard RD, Newgard CB. Adenovirus-mediated transfer of the muscle glycogen phosphorylase gene into hepatocytes confers altered regulation of glycogen metabolism. J Biol Chem. 1992;267:25129-25134.
Juul P, Christensen HB, Hougen HP, Svendsen O, Thygesen P, Rygaard J. Athymic experimental animals in pharmaco-immunological research. Toxicol Lett. 1992;64/65:85-92.
Ginsberg HS, Moldawer LL, Sehgal PB, Redington M, Kilian PL, Chanock RM, Prince GA. A mouse model for investigating the molecular pathogenesis of adenovirus pneumonia. Proc Natl Acad Sci U S A. 1991;88:1651-1655.
Yang Y, Nunes FA, Berencsi K, Furth EE, Gonczeol E, Wilson JM. Cellular immunity to viral antigens limits E1-deleted adenoviruses for gene therapy. Proc Natl Acad Sci U S A. 1994;91:4407-4411.
Imperiale MJ, Kao H-T, Feldman LT, Nevins JR, Strickland S. Common control of the heat shock gene and early adenovirus genes: evidence for a cellular E1A-like activity. Mol Cell Biol. 1984;4:867-874.
Dooley TP, Miranda M, Jones NC, DePamphilis ML. Transactivation of the adenovirus EIIa promoter in the absence of adenovirus E1A protein is restricted to mouse oocytes and preimplantation embryos. Development. 1989;107:945-956.
Spergel JM, Chen-Kiang S. Interleukin 6 enhances a cellular activity that functionally substitutes for E1A protein in transactivation. Proc Natl Acad Sci U S A. 1991;88:6472-6476.
Spergel JM, Hsu W, Akira S, Thimmappaya B. NF-IL6, a member of the C/EBP family, regulates E1A-responsive promoters in the absence of E1A. J Virol. 1992;66:1021-1030.
Lin H, Parmacek MS, Morle G, Bolling S, Leiden J. Expression of recombinant genes in myocardium in vivo after direct injection of DNA. Circulation. 1990;82:2217-2221.
Engelhardt JF, Ye X, Doranz B, Wilson JM. Ablation of E2A in recombinant adenoviruses improves transgene persistence and decreases inflammatory response in mouse liver. Proc Natl Acad Sci U S A. 1994;91:6196-6200.