Stranded Adeno-Associated Virus Vectors in Vitro and in Vivo

Home , Transduction (genetics)

Gene Therapy (2003) 10, 2105–2111 & 2003 Nature Publishing Group All rights reserved 0969-7128/03 $25.00 www.nature.com/gt RESEARCH ARTICLE Rapid and highly efﬁcient transduction by double- stranded adeno-associated virus vectors in vitro and in vivo

Z Wang1, H-I Ma1,3,JLi1, L Sun1, J Zhang1 and X Xiao1,2 1Department of Molecular Genetics and Biochemistry, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA; 2Department of Orthopedic Surgery, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA; and 3Department of Neurological Surgery, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan

Adeno-associated virus (AAV) is a promising gene vector the conventional ssAAV vectors. Similar increases were also based on a single-stranded (ss) DNA virus. Its transgene observed in vivo in a variety of tissues including muscle expression requires the conversion of ssDNA to double- and liver. The dsAAV transduced a vast majority of the stranded (ds) genome, a slow process responsible for the hepatocytes for more than 6 months, while the ssAAV delayed transduction and occasional inefﬁciency. By mutat- transduced only a small fraction. In addition to circumventing ing the inverted terminal repeat, we have made novel AAV the requirement for DNA synthesis, the dsAAV exhibited vectors that predominantly package the self-complementary higher in vivo DNA stability and more effective circularization dsDNA genome. The dsAAV consistently demonstrated than the ssAAV, suggesting potential molecular mechanisms superior and accelerated transduction in vitro and in vivo. for the faster, stronger and prolonged transgene expression. Dramatic increases in transgene expression were observed Gene Therapy (2003) 10, 2105–2111. doi:10.1038/ in most of the cell lines examined, including B16 melanoma sj.gt.3302133 and 3LL lung cancer that are difﬁcult to be transduced by

Keywords: AAV; gene therapy; vector; double-stranded; dsAAV

Introduction is sufficient for many gene therapy purposes,10–12 it is insufficient for others such as therapy for hepatitis (eg, Gene vectors derived from adeno-associated virus (AAV) ribozyme or siRNA strategies) and for certain metabolic can infect both dividing and nondividing cells in vitro diseases that require the majority of the liver cells to be and in vivo, establishing long-term and efficient trans- transduced. In addition, certain cancer gene therapy gene expression with minimal toxicity and cellular strategies such as antiangiogenesis and tumorcidal genes immune responses.1 However, unlike other DNA-based also need timely and high-level expression. A frequently viral and nonviral vectors such as adenovirus and used solution to this shortcoming is the coinfection of plasmid, AAV packages and delivers a single-stranded wild-type adenovirus that greatly accelerates and en- (ss) DNA genome that is transcriptionally inactive until it hances AAV vector-mediated transduction both in vitro is converted into a double-stranded (ds) template. The and in vivo.2,3 However, the improved transduction also ssDNA to dsDNA conversion is a well-documented, comes with a price of unacceptable toxicity due to rate-limiting step involving either the de novo synthesis of adenovirus infection. Alternatively, chemical and physi- the second-strand DNA2,3 or the annealing of the plus cal treatments of cells and animals demonstrated variable and minus strands from two separate viral particles enhancement of AAV transgene expression but accom- coinfected into the same cell.4 This process contributes in panied by considerable toxicity.2,3,13,14 part to the slow onset of transgene expression, especially Recent reports showed that AAV vectors of half size of a in vivo, where a few weeks are often required to achieve wild-type genome could package a ds, hairpin-like significant transgene expression after vector delivery. genome that is self-complementary.15 Such vectors can The delayed transgene expression limits the usefulness circumvent the ss- to dsDNA conversion. However, the of AAV vectors for applications that require immediate vector was inefficient to produce and laborious to purify therapeutic intervention.5–7 Another limitation of AAV because it was a by-product of ssAAV vector produc- vectors (eg, AAV2) is in the liver where only a few tion.15,16 To generate AAV vectors that predominantly percent of the hepatocytes can be transduced despite package the dsDNA genome, we have mutated one of the the use of high vector doses.8,9 Although such efficiency AAVinverted terminal repeats (ITRs). This mutation led to nearly exclusive packaging of hairpin-like, dsAAV DNA genomes. Examination of the dsAAV vector in vitro and in Correspondence: Dr X Xiao, Department of Molecular Genetics and Biochemistry, University of Pittsburgh School of Medicine, Room W1244 vivo and molecular characterization of the vector DNA BST, Pittsburgh, PA 15261, USA support the conclusion that dsAAV vectors render supe- Received 25 March 2003; accepted 14 June 2003 rior gene transfer over the conventional ssAAV vectors. Double-stranded AAV vector transduction Z Wang et al 2106 Results serotype 2 (Figure 1a) and obtained very similar titers as determined by DNA dot blot method (data not shown), Efficient packaging of dsAAV after mutation of one ITR indicating that the deletion of D-sequence on one ITR did A ssAAV DNA genome is flanked by two ITRs of 145 not affect the efficiency of viral replication and packa- bases (Figure 1a and b). The 30 ITR serves as a replication ging. The dsAAV particles were readily purified by origin and a packaging signal.17 AAV replication inter- a single-step heparin affinity column chromatography mediates include head-to-head and tail-to-tail dimer without the need for the tedious density-gradient DNA molecules. The dimers are cleaved by AAV Rep ultracentrifugation.18 By contrast, the centrifugation step proteins at the junction ITR to yield monomers. Each was a requirement for the purification of the early monomer is then packaged into an AAV particle as version dsAAV, because it was a minor product among ssDNA (Figure 1b, left panel). However, if one of the the ssAAV monomers.15,16 By deleting the D-sequence ITRs (for example, the left ITR) is deleted of D-sequence and the trs on the left ITR, we have prevented the (the packaging signal) together with the adjacent monomer formation and made the head-to-head dimers terminal resolution site (trs) (Figure 1a and b, right as the predominant product. Analysis of the purified panel), then the dimers fail to be resolved into mono- vector DNA by neutral agarose gel electrophoresis mers. Nonetheless, each dimer can still be packaged via confirmed that the dsAAV DNA was a ds hairpin of the remaining wild-type ITR into AAV particles in the 2.2 kilo-base pair, while the conventional ssAAV DNA form of a large hairpin DNA, hence a ds genome (Figure migrated faster in the gel as a ss molecule of 4.5 kilo- 1b, right panel). nucleotides (kilo-nt) next to the 2 kb dsDNA marker To generate AAV vectors that predominantly package (Figure 1c, left). As expected, the dsAAV hairpin in the the ds hairpin-like dimer DNA, we have deleted the D- alkaline agarose gel was denatured into a 4.4 kilo-nt sequence and the trs of the left ITR on the AAV vector linear ssDNA, and migrated faster than the 4.5 kilo-nt while keeping the right ITR intact (Figure 1a; see section conventional ssAAV DNA (Figure 1c, right). No detect- Materials and methods). Furthermore, the entire AAV able linear ss monomers of 2.2 kilo-nt were present in the vector genome was made smaller than 2.5 kb so that a alkaline gel, indicating that the dsAAV particles pre- dimer molecule can be packaged without exceeding the dominantly packaged the dimer hairpin DNA. viral packaging limit of 5 kb. Using the enhanced green fluorescent protein (GFP) as the reporter gene, we have Dramatic improvement of transduction by dsAAV produced both ssAAV and dsAAV vectors based on AAV vector in vitro To examine if the dsAAV vectors can transduce more effectively than the ssAAV vectors in vitro, we have compared these vectors in more than 20 cell lines of human, monkey and rodent origins. In the commonly used 293, HeLa and COS-7, the dsAAV-CMV-GFP vector yielded 10–20% green fluorescent cells 1 day after infection at a dose of 500 v.g. (viral genome)/cell without the coinfection of adenovirus. The control ssAAV-CMV- GFP vector yielded less than 1% green fluorescent cells under the same condition. On day 3 after infection, the dsAAV turned 80–100% of the cells green with intense fluorescence, while the ssAAV had approximately 10% of the cells green with dim fluorescence (data not shown). Because the above cell lines are highly permissive for ssAAV vectors, we further examined a number of cell lines that have been shown as very difficult to be Figure 1 Construction and characterization of ssAAV and dsAAV vectors. transduced by ssAAV, for example, the B16 melanoma, (a) AAV vector construction. The conventional ssAAV-CMV-GFP vector 19,20 (4.5 kb) had two wild-type ITRs flanking the GFP and Neor gene 3LL Lewis lung carcinoma and NIH3T3 fibroblast. expression cassettes. The dsAAV-CMV-GFP had deleted the D-sequence Our results showed that the dsAAV could effectively and the terminal resolution site (trs) in the left ITR, while its right ITR transduce the above cells and yield overwhelmingly remained intact in the wild-type form. The Neor gene was also removed to more green cells than the ssAAV did (Figure 2a). decrease the size of the entire vector cassette to 2.2 kb. (b) Diagrams of Furthermore, a dsAAV vector carrying the gene of a ssAAV and dsAAV viral DNA structures in the viral particles and in the secretable protein (angiostatin) also showed much faster host cells after infection. No conversion of ss- to dsDNA is required for the dsAAV after infection. (c) Characterization of ss- and dsAAV viral DNA transduction than the ssAAV counterpart (Figure 2b). In from purified viral particles. For the analysis on neutral agarose gel all the cell lines tested thus far, we have consistently (0.8%), the viral DNA dissolved in TE buffer (10 mM Tris-HCl, 1 mM observed superior transduction by dsAAV over ssAAV. EDTA, pH 7.5) was boiled for 5 min and quickly chilled on ice before loading. The standard DNA marker (M) was loaded without boiling and therefore still in the ds form. Note that the dsAAV DNA hairpin quickly Accelerated and long-term transduction by dsAAV snapped back into the 2.2 kb dsDNA, while the ssAAV DNA (4.5 kilo- vector in vivo nucleotides) remained in ss form and migrated aberrantly next to the 2 kb It has been well documented that the ssAAV vectors can dsDNA marker. For the analysis on alkaline denaturing agarose gel efficiently transduce muscle tissues but at a slow pace. (0.8%), the viral DNA or the DNA marker (M) was dissolved in 100 mM We therefore compared the time course and efficiency of NaOH, 1 mM EDTA for denaturation. Note that the dsAAV DNA was denatured into a 4.4 kilo-nt ssDNA while the ssAAV DNA remained as a dsAAV versus ssAAV in mouse leg tibialis anterior (TA) 11 4.5 kilo-nt ssDNA. Both DNAs migrated with expected molecular weight muscle by intramuscular injection of 1 Â 10 v.g. of either next to the denatured DNA marker (M). the ds or the ss GFP reporter vectors. Our results showed

Gene Therapy Double-stranded AAV vector transduction Z Wang et al 2107

Figure 2 Superior in vitro transduction by dsAAV vectors. (a) Comparison of transduction efficiencies by dsAAV and ssAAV vectors in poorly permissive cell lines. B16 F10, NIH3T3 and 3LL cells were infected with dsAAV-CMV-GFP or ssAAV-CMV-GFP vectors at 10 000 v.g./cell. Photos were taken 3 days after infection. (b) Secretion of angiostatin into Figure 3 Superior in vivo transduction by dsAAV vectors in the muscle the culture media from 293 cells after infection with dsAAV-CMV- and liver. (a) Comparison of transduction efficiencies by dsAAV and angiostatin or ssAAV-CMV-angiostatin vectors at a dose of 500 v.g/cell. ssAAV vectors in mouse tibialis interior (TA) muscle. Tissues were Lane 1 was the culture medium from uninfected 293 cells as a negative collected at 1 week, 2 weeks, 6 weeks and 6 months after injection of control. The relative levels of angiostatin secretion by dsAAV and ssAAV 1 Â 1011 v.g. of dsAAV-CMV-GFP or ssAAV-CMV-GFP. Photographs vector-transduced 293 cells at different time points were quantitated by of muscle cryo thin sections (8 mm) were taken with indicated exposure densitometry, and presented as relative numbers, where the lowest level time (2 or 30 s) with a Nikon TE-300 inverted fluorescent microscope (lane 3) was arbitrarily defined as 1 Â . equipped with a Spot CCD digital camera. Photos of whole muscle tissues were also taken at 6 weeks after vector injection (panels E and J). Note the green color of the TA muscle (panel E) as a result of robust GFP gene that the dsAAV vectors achieved accelerated and more expression from the dsAAV-CMV-GFP vector, whereas the opposite leg robust transgene expression than the ssAAV vectors. By 1 injected with the ssAAV vector had no noticeable green color on the TA muscle. (b) Comparison of transduction efficiencies by dsAAV and ssAAV week after vector delivery, strong GFP expression was vectors in mouse liver. Liver tissues were collected at 3 days, 1 week, 2 already observed in dsAAV-injected muscle, but only a weeks, 2 months and 6 months after i.v. injection of 2 Â 1011 v.g of dsAAV- background level of GFP expression was detectable in CB-GFP or ssAAV-CB-GFP per mouse. Photographs of liver cryo thin ssAAV-injected muscle (Figure 3a, panel A versus f). By 2 sections (8 mm) were taken with an exposure time of 5 s with the Nikon weeks, GFP expression in dsAAV-treated muscle was at TE-300 microscope and the Spot CCD digital camera. least 50 times stronger than the ssAAV-treated muscle as judged by the intensity of the fluorescence. The photographs of dsAAV-treated muscle under a 2-s exposure hepatocytes even with high vector doses.4,21 We admi- time showed much stronger green fluorescence intensity nistered dsAAV or ssAAV GFP vectors intravenously at than a 30-s exposure of the ssAAV-treated muscle (Figure a dose of 2 Â 1011 v.g. per mouse. By 3 days after vector 3a, panel B versus g). The transgene expression of dsAAV injection, dsAAV-treated liver showed 2–5% of the reached a plateau by 6 weeks (Figure 3a, panel C) and hepatocytes expressing GFP. The GFP-positive hepato- maintained the high level for the 6 months duration of cytes increased to more than 10% in the liver by 1 week the experiment (Figure 3a, panel D), whereas the and to more than 50% by 2 weeks (Figure 3b, panels transgene expression of ssAAV increased slowly by 6 A–C). Up to 90% of the hepatocytes became GFP positive weeks and continued to 6 months (Figure 3a, panels and roughly one-third of those cells had very intense H and I). The GFP expression after dsAAV treatment was GFP expression by 2 months (Figure 3b, panel D) after so intense that the muscle turned green and could be dsAAV vector injection and maintained the high levels to observed without the need of a microscope (Figure 3a, the end of the 6 months duration of the experiment panels E and J). Throughout the entire 6-month time (panel E). In contrast, the ssAAV-treated liver had no course, the dsAAV gene expression was at a minimum of detectable GFP expression until 2 weeks after vector 15 times stronger than that of the ssAAV. (The photo- injection (Figure 3b, panels F–H). The GFP expression graphic exposure time of ssAAV-treated muscles in increased slightly by 2 months and continued to increase Figure 3a was 15 times as long as the dsAAV-treated up to 6 months; nevertheless, no more than 2% of the muscles. However, the dsAAV-treated muscles still total hepatocytes became transduced. In comparison, the showed stronger fluorescence than the ssAAV-treated livers treated with dsAAV displayed more green ones.) hepatocytes in 3 days than the livers treated with ssAAV We next compared the time course and efficiency of in 2 weeks (Figure 3b, panel A versus H), and displayed dsAAV- versus ssAAV-mediated transgene expression in similar numbers of green cells as the livers treated with the liver. It has been reported that ssAAV transduction ssAAV for 2 months (Figure 3b, panel A versus I). These is slow and restricted to a small fraction (o10%) of the results strongly demonstrate that dsAAV can achieve

Gene Therapy Double-stranded AAV vector transduction Z Wang et al 2108 more accelerated and robust transduction in the liver corroborate the rapid and efficient transgene expression than ssAAV. of dsAAV in the liver (Figure 3b, panels A–E). Another important observation was the much higher Molecular analysis of AAV genomes in liver in vivo vector DNA stability of dsAAV than the ssAAV (comparing Figure 4a and b). The input dsAAV vector To investigate if the ds- and ssAAV genomes had DNA did not decline significantly from 3 days to 6 different molecular fate in vivo, especially in the liver, months (Figure 4a, lanes 12–15), maintaining an average we performed Southern analysis of the liver DNA of 4–5 copies per cell. In contrast, the input ssAAV vector isolated at various time points after vector injection, DNA declined rapidly from 3 days to 2 weeks, and from 3 days, 2 weeks, 2 months to 6 months. A fragment further declined until 2 months post-vector injection of GFP gene, common for both ds- and ssAAV vectors, (Figure 4b), maintaining an average of 0.2 copies per cell. was used as the probe. At 3 days after injection, the Furthermore, the majority of the ssAAV DNA remained dsAAV vector was primarily in the form of linear in the ss form in the first 2 weeks (Figure 4a, lanes 4 and dsDNA of 2.1 kb (Figure 4a, lane 2, also see Figure 1b 5, and 9 and 10), and was slowly converted to sc circular and c for diagrams), which was confirmed by BamHI dsDNA and high-molecular-weight DNA by 2–6 months digestion that cut once in the genome (lane 7, and (Figure 4b, lanes 11 and 12). The sc circular dsDNA was diagram below). By 2 weeks, approximately half of the confirmed by SacI digestion into a linearized unit-length input linear dsDNA was converted to the supercoiled dsDNA (Figure 4b, lanes 6 and 7; note the disappearance (sc) circular monomer DNA that migrated faster in the of the sc DNA as seen in lanes 11 and 12). The rapid agarose gel than the linear DNA (Figure 4a, lane 3). The decline of input vector, the slow and inefficient conver- sc circular DNA could be linearized by BamHI digestion sion of ssDNA to sc circular dsDNA also corroborates the into a 2.1 kb unit-length dsAAV genome (lane 8, upper delayed and weaker transgene expression of ssAAV band), whereas the remaining input linear viral genome vectors in the liver (Figure 3b, panels F–J). was digested into two fragments of 0.83 and 1.27 kb (lane 8, lower band). By 2 months and up to 6 months, the linear input dsAAV genomes were no longer detectable. They were converted primarily to the sc circular Discussion monomer DNA (Figure 4a, lanes 4 and 5, and 9 and Here we have presented a new strategy for efficient 10). Interestingly, the rapid and efficient conversion of production of dsAAV vectors and demonstrated the the linear dsDNA to the sc circular DNA appeared to usefulness of these vectors both in vitro and in vivo for accelerated and robust transgene expression. The dsAAV not only rendered much improved transduction in cells that are highly permissive for ssAAV, but also overcame barriers in the poorly permissive cells. For example, the melanoma B16F10 cell is a very difficult cell line for ssAAV to transduce. This limited transduction has been attributed to the low abundance of the AAV2 cell surface receptor.19 However, the dsAAV was able to transduce efficiently the B16F10 cells (Figure 2a). Since both the ss- and dsAAV vectors used identical viral capsids, the profound difference in transduction efficiencies could not be accounted for by the lack of cell surface receptors. In addition, another cell line NIH3T3 fibroblast was readily transduced by the dsAAV but only slightly transduced by the ssAAV (Figure 2a), although both NIH3T3 and the permissive 293 cells had similar binding capacity to AAV2,22 again suggesting that factors Figure 4 Southern analyses of AAV vector DNA molecular fate in the other than the receptor play an important role. The mouse livers at various time points after in vivo administration. (a) dsAAV also showed superior transduction in primary Analysis of dsAAV vector fate. Total DNA isolated from the dsAAV-CB- islet and neural stem cell cultures (data not shown). Our GFP vector-treated livers at various time points, or from the untreated control liver, was analyzed by 0.8% agarose gel electrophoresis. A total of in vitro results demonstrated that the dsAAV vector 10 mg of DNA from each sample was loaded in each lane either without genome could effectively circumvent the key rate-limit- restriction digestion (lanes 1–5), with digestion by BamHI (lanes 6–10) or ing step of ssDNA to dsDNA conversion. These findings double-digestion by BamHI and HindIII (lanes 11–15). The plasmid DNA should facilitate new applications of dsAAV in gene was used as a copy number control ranging from 2, 10 to 50 copies per cell therapy for cancer, diabetes and neural degenerative (lanes 16–18). Linear vector DNA (L) is indicated by an open arrowhead diseases. and the sc vector DNA is indicated by a solid arrowhead. The autoradiograph was obtained after exposure against the X-ray film The in vivo results also support the conclusion that the overnight. (b) Analysis of ssAAV vector fate. Total DNA isolated from effectiveness of dsAAV transduction is achieved at the the ssAAV-CB-GFP vector-treated livers at various time points post-vector post-cell-entry levels, because the parallel studies using injection was similarly analyzed as the dsAAV counterparts. A total of ds- and ssAAV vectors were performed on the same 10 mg of DNA from each sample was loaded in each lane either without tissues but revealed profound differences. Of particular restriction digestion (lanes 8–12) or with digestion by SacI (lanes 3–7). interest is the transduction in the liver. The mediocre The ss vector DNA (SS) is indicated by a gray arrowhead. The plasmid DNA was used as a copy number control ranging from two copies (lane 2) efficiency in our studies using the ssAAV vector was 9,21 9 to 10 copies (lane 1) per cell. The autoradiograph was obtained after consistent with the results by others. Song et al exposure against the X-ray film for 2 weeks. reported that after administration of 3 Â 1010 infectious

Gene Therapy Double-stranded AAV vector transduction Z Wang et al 2109 units of ssAAV-GFP vector into the liver, only a small of dsAAV into circular monomers does not favor the portion of hepatocytes (about 5%) expressed the trans- formation of high-molecular-weight DNA. gene. Kay and co-workers showed similar results (o10% A major drawback of the dsAAV is its decrease in the transduction) with ssAAV vector containing the clotting size of foreign genes that can be packaged. An AAV viral factor IX gene or beta-galactosidase gene in the liver at particle can usually package approximately 5000 nucleo- doses up to 3.9 Â 1012 genomes/animal.8,21 They con- tides of ssDNA. The dsAAV vectors in effect contain two cluded that only a small fraction of the hepatocytes was copies of the ssAAV cassette in a head-to-head hairpin permissive for ssAAV transduction despite the fact that form (Figure 1). As a result, each vector cassette most of the hepatocytes had taken up the viral DNA, as including the ITRs could not exceed 2.5 kb in size for being demonstrated by in situ hybridization8 and by efficient packaging. When common promoters such as adenovirus infection.3 Adenovirus effectively converted the CMV, CB and MCK are used, a cDNA of about 1.2 kb the transcriptionally inactive ssAAV vector DNA into can be packaged into the dsAAV. While such packaging functional templates in the majority of the hepatocytes.3 size excludes many candidate genes, it is still useful for The capability of dsAAV to render high-level transgene a large number of genes that encode for cytokines, expression in the majority of the hepatocytes substanti- growth factors, other secretable proteins (eg, endostatin ates the notion that ss- to dsDNA conversion is a major and angiostatin), muscle proteins (eg, sarcoglycans) and barrier for ssAAV transduction in vivo, particularly in the small RNA genes (ribozymes,5 antisense28 and small liver, irrespective of how the ssDNA was converted to interference RNA29) for the purposes of prompt, robust dsDNA, either by second-strand DNA synthesis2,3 or by and prolonged transgene expression. annealing of the plus and minus strands of two ss vector genomes.4 The double-strandedness of the AAV genome appears not to be the sole factor that contributes to the high Materials and methods efficiency of dsAAV. Unlike other dsDNA vectors such as Cell lines adenovirus and plasmid DNA, dsAAV-mediated trans- Human cell lines including 293, cervical cancer HeLa, duction does not rapidly decline with time. Its transgene prostate cancer PC3, DU145, TSU and LNCaP, colon expression increases with time, especially in vivo, cancer HCT116, follicular thyroid carcinoma FTC133, suggesting continuous recruitment and conversion of head and neck cancer 1483, osteosarcoma cell SaOS2, the dsAAV genomes into transcriptionally active tem- glioma U87, breast cancer MCF7 and BT549, hepato- plates. Southern blot analyses of the ds- and ssAAV carcinoma HepG2 and monkey kidney cell COS-7 and vector DNA isolated from the livers at different intervals murine cell lines melanoma B16 and liver cancer Hepa1- (Figure 4) not only revealed higher in vivo DNA stability 6 and Lewis lung carcinoma 3LL and NIH3T3 fibroblast, of the dsAAV, but also showed more efficient conversion and Chinese hamster ovary (CHO) cell were obtained into sc circular DNA (comparing Figure 4a and b). When from American Type Culture Collection (Rockville, MD, identical doses of viral genome particles (each dsAAV USA) or from various laboratories. is considered as two ssAAV) were administered, the dsAAV DNA persisted more effectively than the ssAAV DNA throughout the time course in the liver. The copy AAV vector construction and production numbers of dsAAV were more than 10 times higher than The conventional AAV-CMV-GFP vector plasmid (Figure the ssAAV 2 months after vector administration. It is 1a) was published earlier.30 The AAV vector plasmid plausible to believe that the higher vector genome dsAAV-CMV-GFP (Figure 1a) that generates ds viral stability also contributed to the higher in vivo transduc- DNA was constructed by deleting the D-sequence of the tion efficiency. Another interesting phenomenon is the 50 ITR with MscI digestion. MscI removed the D-sequence more effective conversion of linear vector DNA to and the terminal resolution site (trs) (nucleotides 122–144 the circular sc form.23 A high proportion of the dsAAV of AAV2 genome, GenBank #NC_001401). The ITR on the in the liver persisted as the sc monomer (Figure 4a). Since 30 terminus of the vector remained intact (wild type). To the sc DNA is transcriptionally more active than the make the vector smaller than 2.5 kb for dsAAV packa- linear DNA,24 it is reasonable to speculate that the ging, the Neor cassette was also deleted (Figure 1a). gradual increase of sc dsDNA may be responsible, in a Plasmid dsAAV-CMV-angiostatin was made by repla- major part, for the gradual increase of gene expression in cing the GFP gene and the SV40 polyA site of dsAAV- vivo. A recent study of ssAAV dose escalation in liver CMV-GFP with the angiostatin cDNA coupled with a indicated that the gene expression levels correlated more miniature polyA site. For stable liver gene expression, closely with the formation of sc vector DNA than with the CMV promoter was replaced by the CB promoter the high-molecular-weight concatemers.21 Long-term (CMV enhancer/chicken beta-actin promoter), generat- persistence of the sc vector DNA in vivo has been ing plasmids ssAAV-CB-GFP and dsAAV-CB-GFP. The reported in numerous cases including in tissues such as recombinant viral stocks were produced by the adeno- the lung,14,25 muscle23,26 and liver.9,27 Whether the virus free, triple-plasmid cotransfection method.30 The sc vector DNA is the primary contributor to the AAV particles were subsequently purified by heparin transgene expression in vivo remains to be seen. Inter- affinity column chromatography.18 Titers of viral genome estingly, in dsAAV-transduced livers, the vector DNA (v.g.) particle number were determined by quantitative primarily remained as the circular low-molecular-weight DNA dot blot method.31 Each dsAAV particle was episomal DNA, whereas the ssAAV vector transduced calculated as containing two copies of ss viral genome livers had much more high-molecular-weight DNA, (versus one copy in each conventional ssAAV particle). which could be either multimers of vector DNA and/ All AAV vectors used in this study were from AAV or integrated forms, suggesting that rapid conversion serotype 2.

Gene Therapy Double-stranded AAV vector transduction Z Wang et al 2110 In vitro and in vivo infection 3 Fisher KJ et al. Transduction with recombinant adeno-associated For the in vitro assays, 2 Â 105 cells were seeded in each virus for gene therapy is limited by leading-strand synthesis. well of a six-well plate 1 day before infection. AAV J Virol 1996; 70: 520–532. vectors diluted in DMEM medium were added to the cell 4 Nakai H, Storm TA, Kay MA. Recruitment of single-stranded culture without wild-type adenovirus coinfection. The recombinant adeno-associated virus vector genomes and GFP expression was examined 1 day and 3 days after intermolecular recombination are responsible for stable infection on a Nikon TE-300 inverted fluorescent micro- transduction of liver in vivo. J Virol 2000; 74: 9451–9463. scope. The secretion of angiostatin in culture medium 5 Lewin AS et al. Ribozyme rescue of photoreceptor cells in a transgenic rat model of autosomal dominant retinitis was detected by immunoprecipitation (IP) Western blot 6 pigmentosa [published erratum appears in Nat Med 1998 as described previously. For in vivo assays in the muscle Sep;4(9):1081]. Nat Med 1998; 4: 967–971. and liver, female C57/B10 mice (8–10 weeks of age) were 6MaHIet al. Suppression of intracranial human glioma growth used (Jackson Laboratory, Bar Harbor, ME, USA). Tibialis after intramuscular administration of an adeno-associated viral 11 anterior muscles were injected with 50 ml (1.0 Â 10 total vector expressing angiostatin. Cancer Res 2002; 62: 756–763. viral genome particles) of either the ss (ssAAV-CMV- 7 Shimazaki K et al. Adeno-associated virus vector-mediated bcl-2 GFP) or the ds (dsAAV-CMV-GFP) AAV vectors. For gene transfer into post-ischemic gerbil brain in vivo: prospects liver gene transfer, 200 ml (total 2 Â 1011 viral particles per for gene therapy of ischemia-induced neuronal death. Gene mouse) of either ssAAV-CB-GFP or dsAAV-CB-GFP was Therapy 2000; 7: 1244–1249. injected intravenously. Muscle and liver samples were 8 Miao CH et al. Nonrandom transduction of recombinant adeno- collected at various time points for analysis. For the associated virus vectors in mouse hepatocytes in vivo: cell detection of GFP expression in the muscle and liver, cryo cycling does not influence hepatocyte transduction. J Virol 2000; thin sections (8 mm) were directly mounted with Gel/ 74: 3793–3803. Mount (Biomeda, Foster City, CA, USA) and observed 9 Song S et al. Stable therapeutic serum levels of human alpha-1 under Nikon TE-300 inverted fluorescent microscope. antitrypsin (AAT) after portal vein injection of recombinant adeno-associated virus (rAAV) vectors. Gene Therapy 2001; 8: 1299–1306. Southern analysis of liver tissue DNA 10 Mount JD et al. Sustained phenotypic correction of hemophilia B Liver tissues were harvested, minced and suspended in dogs with a factor IX null mutation by liver-directed gene lysis buffer (10 mM Tris-HCl, 100 mM EDTA, 0.5% SDS therapy. Blood 2002; 99: 2670–2676. and 100 mg/ml proteinase K, pH 8.0). The samples were 11 Chao H, Mao L, Bruce AT, Walsh CE. Sustained expression of digested at 371C overnight and incubated for additional human factor VIII in mice using a parvovirus-based vector. Blood 2 h after adding 20 mg/ml RNase A. After phenol, 2000; 95: 1594–1599. phenol–chloroform and chloroform extraction, the total 12 Xu L et al. CMV-beta-actin promoter directs higher expression DNA (both low- and highmolecular-weight DNA) was from an adeno-associated viral vector in the liver than the cytomegalovirus or elongation factor 1 alpha promoter and precipitated with 0.3 M sodium acetate and 2.5 volumes of ethanol and by centrifugation, washed twice with 70% results in therapeutic levels of human factor X in mice. Hum Gene Ther 2001; 12: 563–573. ethanol and redissolved in TE (10 mM Tris-HCl, 1 mM 13 Qing K et al. Role of tyrosine phosphorylation of a cellular EDTA, pH 8.0). Restriction enzyme digestions (New protein in adeno-associated virus 2-mediated transgene England BioLabs) were carried out overnight. DNA expression. Proc Natl Acad Sci USA 1997; 94: 10879–10884. samples were ethanol-precipitated after digestion to 14 Duan D et al. Endosomal processing limits gene transfer to remove extra salt before they were separated on a 0.8% polarized airway epithelia by adeno-associated virus. J Clin agarose gel. Southern hybridization was performed with Invest 2000; 105: 1573–1587. GeneScreen Plus membrane according to the manufac- 15 McCarty DM, Monahan PE, Samulski RJ. Self-complementary turer’s standard protocol (NEN-DuPont). The EGFP recombinant adeno-associated virus (scAAV) vectors promote DNA probe was a 0.7 kb BamHI and NotI fragment efficient transduction independently of DNA synthesis. Gene labeled with alpha-32P dATP and Random Primer Kit Therapy 2001; 8: 1248–1254. (Boehringer Mannheim). 16 Hirata RK, Russell DW. Design and packaging of adeno-associated virus gene targeting vectors. J Virol 2000; 74: 4612–4620. 17 King JA, Dubielzig R, Grimm D, Kleinschmidt JA. DNA Acknowledgements helicase-mediated packaging of adeno-associated virus type 2 We thank Paul. Monahan and Michael Xiao for critical genomes into preformed capsids. EMBO J 2001; 20: 3282–3291. reading of the manuscript. This work is supported by 18 Clark KR, Liu X, McGrath JP, Johnson PR. Highly purified recombinant adeno-associated virus vectors are biologically NIH Grants AR45967 and AR45925. active and free of detectable helper and wild-type viruses. Hum Gene Ther 1999; 10: 1031–1039. 19 Girod A et al. Genetic capsid modifications allow efficient References re-targeting of adeno-associated virus type 2. Nat Med 1999; 5: 1052–1056. 1 Samulski RJ, Sally M, Muzyczka N. Adeno-associated viral 20 Hansen J, Qing K, Srivastava A. Adeno-associated virus type 2- vectors. In: Friedmann T (ed). Development of Human Gene mediated gene transfer: altered endocytic processing enhances Therapy. Cold Spring Harbor Laboratory Press: Cold Spring transduction efficiency in murine fibroblasts. J Virol 2001; 75: Harbor, 1999, pp 131–172. 4080–4090. 2 Ferrari FK, Samulski T, Shenk T, Samulski RJ. Second-strand 21 Nakai H et al. A limited number of transducible hepato- synthesis is a rate-limiting step for efficient transduction by cytes restricts a wide-range linear vector dose response in recombinant adeno-associated virus vectors. J Virol 1996; 70: recombinant adeno-associated virus-mediated liver trans- 3227–3234. duction. J Virol 2002; 76: 11343–11349.

Gene Therapy Double-stranded AAV vector transduction Z Wang et al 2111 22 Hansen J et al. Impaired intracellular trafficking of adeno- 27 Nakai H et al. Extrachromosomal recombinant adeno-associated associated virus type 2 vectors limits efficient transduction of virus vector genomes are primarily responsible for stable liver murine fibroblasts. J Virol 2000; 74: 992–996. transduction in vivo. J Virol 2001; 75: 6969–6976. 23 Duan D et al. Circular intermediates of recombinant adeno- 28 Xiao X et al. Adeno-associated virus (AAV) vector antisense gene associated virus have defined structural characteristics transfer in vivo decreases GABA(A) alpha1 containing receptors responsible for long-term episomal persistence in muscle tissue and increases inferior collicular seizure sensitivity. Brain Res [published erratum appears in J Virol 1999 Jan;73(1):861]. J Virol 1997; 756: 76–83. 1998; 72: 8568–8577. 29 McCaffrey AP et al. RNA interference in adult mice. Nature 2002; 24 Wang JC, Lynch AS. Transcription and DNA supercoiling. Curr 418: 38–39. Opin Genet Dev 1993; 3: 764–768. 30 Xiao X, Li J, Samulski RJ. Production of high-titer recombinant 25 Flotte TR et al. Gene expression from adeno-associated virus adeno-associated virus vectors in the absence of helper vectors in airway epithelial cells. Am J Respir Cell Mol Biol 1992; adenovirus. J Virol 1998; 72: 2224–2232. 7: 349–356. 31 Snyder R, Xiao X, Samulski RJ. Production of recombinant 26 Song S, Laipis PJ, Berns KI, Flotte TR. Effect of DNA-dependent adeno-asociated viral vectors. In: Smith D (ed). Current Protocols protein kinase on the molecular fate of the rAAV2 genome in in Human Genetics. John Wiley & Sons Ltd: New York, 1996: skeletal muscle. Proc Natl Acad Sci USA 2001; 98: 4084–4088. 12.1.1–12.2.23.

Gene Therapy