SEAL Team Season 2 Season 2 of SEAL Team premiered on October 3, 2018. SEAL Team is an American military drama television series created by Benjamin Cavell. The series is produced by CBS Television Studios. | AU ROYAUME DES SINGES STREAMING | solidwor

Altering AAV tropism with mosaic viral capsids

Altering AAV tropism with mosaic viral capsids

ARTICLE doi:10.1016/j.ymthe.2005.03.005 Altering AAV Tropism with Mosaic Viral Capsids Laure Gigout,1 Patricia Rebollo,1 Nathalie Clement,1 Kenneth ...

531KB Sizes 1 Downloads 9 Views

ARTICLE

doi:10.1016/j.ymthe.2005.03.005

Altering AAV Tropism with Mosaic Viral Capsids Laure Gigout,1 Patricia Rebollo,1 Nathalie Clement,1 Kenneth H. Warrington Jr.,2,3 Nicholas Muzyczka,3,4 R. Michael Linden,1 and Thomas Weber1,5,* 1

Department of Gene and Cell Medicine and 5Department of Molecular, Cell and Developmental Biology, Mount Sinai School of Medicine, 1425 Madison Avenue, New York, NY 10029-6574, USA 2 Pediatrics–Division of Cellular and Molecular Therapy, University of Florida, Alachua, FL 32615-9586, USA 3 Powell Gene Therapy Center and 4Department of Molecular Genetics and Microbiology, College of Medicine, University of Florida, Gainesville, FL 32610-0266, USA *To whom correspondence and reprint requests should be addressed. Fax: +1 212 849 2437. E-mail: [email protected]

Available online 16 April 2005

Over the past decade, AAV-based vectors have emerged as promising candidates for gene therapeutic applications. Despite the broad tropism of the first eight serotypes identified, certain cell types are refractory to transduction with AAV-based vectors. Furthermore, for certain applications the targeting of specific cell types is desirable. To improve on present methods to alter AAV2 tropism, we take advantage of AAV2 mosaics. Here, we show that AAV2 mosaics have improved infectivity compared with all-mutant virions. Using an AAV2 mutant that contains the immunoglobulin-binding Z34C fragment of protein A, we demonstrate the utility of AAV2 mosaics to alter AAV2 tropism. This system allows us to transduce selectively and efficiently MO7e and Jurkat cells. The use of AAV2 mosaics with a protein A fragment inserted into their capsid, together with targeting antibodies, is a versatile method that allows the specific transduction of a wide array of cell types.

INTRODUCTION Gene therapy shows great promise for the treatment of a variety of genetic diseases including classical genetic disorders such as cystic fibrosis, sickle cell anemia, and lysosomal storage disorders. In addition, Parkinson disease, Alzheimer disease, type I diabetes, and even cardiovascular disorders and cancer are among the diseases that are targets for gene therapy. So far, the progress in the clinical application of gene therapy has been limited by the lack of ideal gene delivery vehicles. Among the most promising vectors currently used are adeno-associated virus (AAV)-based vectors. Adeno-associated viruses are small (20–25 nm), singlestranded DNA viruses of the family Parvoviridae. The icosahedral capsid of 60 capsid proteins is formed by three different proteins, VP1, VP2, and VP3, in a ratio of about 1:1:10 [3]. To date, eight distinct serotypes of AAV with different host cell tropisms have been described, although most recently, additional AAV subtypes have been reported [6]. Of the known serotypes, AAV2 is by far the most widely used and best characterized serotype. The primary receptor of AAV2 is heparan sulfate proteoglycan (HSPG) [30]. Additionally, the FGF receptor [21], avh5 integrins [29], and hepatocyte growth factor receptor [12] have been proposed to serve as coreceptors. 856

The broad tropism of AAV2 is a result of the ubiquitous expression pattern of HSPG. Especially for ex vivo transduction of homogeneous cell preparations, the broad array of cells that are infectable by AAV2 can be an advantage. Nevertheless, certain cells cannot be efficiently infected with AAV2 or any of the other AAV serotypes. As a result, the expansion of AAV2 tropism is highly desirable. For other applications, including certain ex vivo and many in vivo applications, it is advantageous to limit the tropism to one specific cell type, requiring both the elimination of the natural tropism of AAV2 and the addition of new targeting information. Over the past few years, several methods have been developed to target AAV2 to additional cell types. The published efforts to alter AAV2 tropism took advantage of an array of diverse methods [2,7,8,16,17,19,20,24, 26,32,35]. Chief among these strategies was the introduction of a peptide or protein ligand coding sequence into the AAV2 capsid open reading frame [2,16,19,20,24]. While these methods showed some success, the modification of the AAV2 capsid often resulted in significant reduction in both virus particle and transducing titers, especially when large ligands were introduced into the AAV2 capsid. In certain instances, if the ligand is displayed exclusively at the N-terminus of VP2, the

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy 1525-0016/$30.00

doi:10.1016/j.ymthe.2005.03.005

reduction in titers can be mitigated by supplying VP1 and VP3 in trans [31]. Our work demonstrates that an additional method to alleviate the detrimental effects of capsid modification is the use of AAV2 mosaics. Recently, chimeras between two different AAV serotypes have been described [10,22]. These chimeras showed the combined tropism of both serotypes, demonstrating the utility of hybrid AAV vectors. In another publication it was reported that AAV2 viral particles composed of wildtype proteins and capsid proteins with a single-chain antibody against CD34 inserted at the N-terminus of VP2 can successfully transduce CD34-positive cells [34]. The transducing titers reported in this study were, however, extraordinarily low (2  102 tu/ml). For this reason, as also pointed out by Rabinowitz and Samulski [23], it cannot be excluded that the apparent transduction was indeed a result of pseudotransduction [1]. Here, we show unequivocally that the use of AAV2 mosaics with ligand insertions in only a subset of the VP1, VP2, and VP3 capsid proteins results in increased virus particle and transducing titers compared with viruses that carry the insertion in all 60 capsid proteins. Furthermore, we demonstrate the utility of viral mosaics in altering AAV2 tropism using a versatile system that will allow the specific transduction of a variety of cells.

RESULTS To test if AAV2 mosaics composed of both wild-type and capsid proteins with peptide insertions have higher infectivities than their all-mutant counterparts we decided to take advantage of two mutants described earlier [32]. These mutants, called L4 and L5, contain a HA epitope (YPVDVPDYA) insertion after amino acid 522 and 553 (of VP1), respectively. Both these viral mutants can produce viral particles [32] at satisfactory titers, but the virions generated are noninfectious. We reasoned that mosaic viruses that contain either L4 or L5 capsid proteins as well as wild-type proteins should be infectious because a limited number of wild-type proteins could provide essential functions that are lost in L4 or L5 mutant virions. Production of L4 and L5 Mosaics To produce L4/wild-type and L5/wild-type mosaics we subcloned the Cap coding regions of L4-Cap and L5-Cap into the plasmid pDG [9]. pDG encodes both the nonstructural AAV2 Rep proteins and the capsid proteins VP1, VP2, and VP3. In addition, the plasmid provides all necessary adenovirus helper functions for productive replication of AAV2 in 293 cells. Cotransfection of pDG with a plasmid that contains a transgene between two AAV2 inverted terminal repeats (ITRs) results in the production of recombinant AAV2 [9]. To produce AAV2 mosaics we used a modification of this two-plasmid system [9]. In this approach, three

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

ARTICLE

plasmids are co-transfected into 293 cells by calcium phosphate transfection. The plasmids pDG and pDG-L4 or pDG-L5 encode the wild-type or mutant capsid proteins, respectively. pTRUF11 encodes a green fluorescent protein (GFP) expression cassette placed between two AAV2 ITRs [14]. We used between 0 and 100% pDG-L4 or pDG-L5, and the balance was made up by wild-type pDG. Three days after transfection we harvested the cells and produced cell lysates. We then isolated the virus via iodixanol density step gradient [36] and checked the purity by SDS–PAGE and silver staining (data not shown). Analysis of L4 and L5 Mosaics To determine the ratio of mutant and wild-type proteins in the purified virions, we analyzed the capsid composition of L4 mosaics by Western blot analysis using both an antibody that recognizes all three capsid proteins and an antibody against the HA epitope. Two bands could be detected with the anti-capsid-protein antibody, but only the higher band reacted with an anti-HA antibody, demonstrating that mosaic virions were formed. As expected, the ratio of wild-type to mutant capsid proteins, estimated by the relative intensities of the two VP3 bands, was determined by the ratio of wild-type and mutant pDG plasmid during the transfection (Fig. 1). Similar results were obtained for L5 mosaics (data not shown). We then

FIG. 1. Western blot and immunoprecipitation analysis of L4- and L5-AAV2 mosaics. (A) Iodixanol-purified virus preparations produced with the indicated amount of mutant pDG-L4 were analyzed by SDS–PAGE and Western blot as described under Material and Methods. The left side shows capsid proteins detected with the antibody B1 against VP3. The right side shows the same membrane after being reprobed with anti-HA antibody 16B12. The higher molecular weight band detected with the anti VP3 antibody (left) represents a capsid protein with a HA-epitope insertion, as confirmed by its reactivity with an anti HA-epitope antibody (right). The slightly higher molecular weight of L4-VP3 in the 25% mosaic preparation is due to a gel artifact. To be able to detect both the wild-type and the L4-VP3 band short exposure times are necessary. On longer exposures two bands, the higher band being detectable with anti-HA antibody 16B12, can be detected for all capsid proteins. (B) 3.5  109 genome-containing, iodixanol-purified b wildtypeQ or 75% L4-AAV2 mosaic virus particles were immunoprecipitated with anti-HA (16B12) or anti-AU1 beads and analyzed by SDS–PAGE and Western blot as described under Material and Methods. On longer exposures VP1 and VP2 are visible in all positive samples.

857

ARTICLE

determined the number of genome-containing particles by real-time PCR [25] and the transducing titers by infecting C12 cells (expressing AAV2 Rep and Cap) with recombinant AAV2 and co-infection with adenovirus followed by FACS analysis [4,5,11]. As previously published [32], 100% mutant AAV2 (L4 and L5 insertion mutants) were able to form virus particles, but these viruses were not infectious (Figs. 2A and 2B). Interestingly, even when we added only small amounts of wild-type plasmid (25% of total) to the transfections—resulting in approximately 25% wild-type protein within the capsid (Fig. 1A)—the mosaic AAV2 particles regained infectivity (Figs. 2A and 2B). For L5, the wild-type virus and the 25% L5 virus showed comparable transducing titers (Fig. 2B). Transducing titers of 25% L4 mosaics, on the other hand, were approximately 10-fold lower than titers of virus with wild-type capsids (Fig. 2A). Further increasing the proportion of mutant protein in the viral capsids resulted in a progressive loss of infectivity (Fig. 2). These data suggested that the inclusion of wild-type capsid proteins, generating AAV2 mosaics, can rescue

doi:10.1016/j.ymthe.2005.03.005

viral infectivity of AAV2 mutants that, on their own, form noninfectious virions. These results also eliminate the possibility of pseudotransduction [1] because we could observe no transduction with either 100% L4 or 100% L5 virus preparations. We could not rigorously exclude, however, that the observed transduction was not due to bcontaminatingQ virions with wild-type capsids. To address this potential concern, we analyzed mosaic virions by immunoprecipitation with an anti-HA antibody coupled to beads. When we precipitated 75% L4 virions with an antibody against HA coupled to beads, we were able to precipitate quantitatively the viral particles; viral particles with wild-type capsids, on the other hand, remained in the supernatant (Fig. 1B). A nonspecific control antibody against the unrelated epitope AU1 coupled to beads was unable to precipitate HA-containing or wild-type virus (Fig. 1B). We obtained similar results for mosaic virus preparations with different compositions (data not shown). Because no virions were detectable in the supernatant of the anti-HA immunoprecipitation of 75% L4, these data strongly indicated

FIG. 2. Viral particle and transducing titers of L4 and L5 mosaics. (A) L4 mosaics (encoding the transgene GFP) were produced as described under Material and Methods using between 0 and 100% pDG-L4 as indicated (the balance being wild-type pDG) and constant amounts of pTRUF11. Viral particle titers (gcp/ml) were then determined with real-time PCR [25]. Transducing titers (tu/ml) were obtained by infection of C12 cells with increasing amounts of L4 or L5 AAV mosaics and co-infection with adenovirus as described under Material and Methods. The ratio of gcp/tu—a measure inversely proportional to the infectivity of a particular virus preparation—was calculated from the particle and transducing titers and is shown on the right. (B) L5 mosaics were analyzed as described under (A) for L4 mosaics.

858

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

doi:10.1016/j.ymthe.2005.03.005

that the transduction observed was indeed due to mosaic AAV2 particles and not due to contaminating virions with wild-type capsids. Nevertheless, formally, we could not exclude the possibility that the amount of contaminating viral particles with wild-type capsids was below the detection limit. To rule out this possibility, we immunoprecipitated 25% L4 and 50% L4 virions with anti-HA beads and eluted the viral particles with excess HA peptide. As shown in Fig. 3 for 25 and 50% L4 mosaics, the eluted virus was clearly infectious. This demonstrates unambiguously that virions with mosaic capsids are indeed infectious. We obtained similar results for all mosaic virus preparations tested (data not shown). Mosaics Composed of Two Mutant Sets of Capsid Proteins To prove independently that AAV2 mosaics can have increased infectivities compared with their all-mutant counterparts, we decided to produce mosaics composed of 50% L4 and 50% L5 capsid proteins. Because both 100% L4 and 100% L5 viruses are completely noninfectious, infectivity of 50% L4/50% L5 virions demonstrates that mosaics assembled from two defective sets of capsid proteins can be infectious. As expected, mosaics composed of 50% L4 and 50% L5 are able to transduce C12 cells, albeit at low efficiencies (titers 4.6  103 tu/ml). One hundred percent viruses were inactive; FACS profiles from transduction experiments using these viruses were indistinguishable from profiles of samples without virus, excluding the possibility of pseudotransduction. The low titers notwithstanding, these experiments provide independent proof that AAV2

ARTICLE

mosaics derived from two mutant, noninfectious viruses can be infectious. Altering AAV2 Tropism with Viral Mosaics Hallek and colleagues [24] previously described a flexible system that allows the specific transduction of cells that are otherwise poorly transduceable with AAV2. The insertion of an immunoglobulin-binding fragment of protein A (Z34C) [28], combined with a targeting antibody, allowed the specific transduction of a variety of cell types including the erythroleukemia cell line MO7e and T-lymphoma-derived Jurkat cells [24]. Unfortunately, however, the insertion of this relatively large ligand (34 amino acids) resulted in an at least 10-fold reduction in particle titers. Even more dramatically, the infectious titers, when tested on HeLa cells, were roughly 4 orders of magnitude lower than wild-type titers, suggesting a serious disturbance of viral capsid functions [24]. We reasoned that for the expansion of AAV2 tropism with this system, Z34C would not have to be present in all capsid proteins. To test this hypothesis, we produced mosaic viruses that contained between 25 and 75% Z34C capsid proteins, the rest consisting of wild-type proteins, and confirmed the capsid composition of the virions by Western blot (data not shown). Using different ratios of wild-type and Z34C-pDG, we were able to produce 25, 50, and 75% but not 100% Z34C-AAV2 at satisfactory titers (Table 1). Consistent with the deleterious effect of the Z34C insertion, transducing titers of wild-type capsid virus were higher than titers of 25 (2.5-fold), 50 (6- to 7-fold), and 75% (200- to 300-fold) mosaic viruses when tested on HeLa cells (Table 1).

FIG. 3. Immunopurified AAV mosaic virions retain their infectivity. 25 or 50% L4 mosaics were immunoprecipitated with a monoclonal antibody against the HA epitope coupled to beads (16B12; Covance) as described under Material and Methods. The bound virus was then eluted by incubation of the beads with excess HA-peptide. The infectivity of the mosaic virions before and after immunopurification was analyzed on C12 cells as described under Material and Methods. Immunopurification of the virus results in a partial loss of infectivity. To correct for this, the particle-to-cell ratio was increased fivefold compared to nonimmunopurified virus. Because of the seriously impaired infectivity of 75% virus, a similar experiment resulted in a significant but low percentage of GFPpositive cells.

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

859

ARTICLE

doi:10.1016/j.ymthe.2005.03.005

In a next step, we examined the usefulness of these mosaics to alter AAV2 tropism. To this end, we analyzed the infectivity of these viruses on MO7e and Jurkat cells in the presence and absence of targeting antibodies and various inhibitors. As previously described [24], wildtype-capsid AAV2 was able to transduce both MO7e and Jurkat cells to a certain extent. Similarly, all AAV2/Z34C mosaics were able to transduce these cells even in the absence of targeting antibody; however, in the presence of targeting antibody against the stem cell factor receptor c-kit (CD117), the transduction of MO7e using 25% Z34C mosaic virions increased 2- to 4-fold (Fig. 4). Likewise, antibodies against h1 integrins (CD29) together with 25% Z34C mosaic virions resulted in a 1.5- to 2-fold improvement of transduction of Jurkat cells compared to wild-type or 25% Z34C mosaic virus in the absence of targeting antibody (Fig. 5). Presumably, because of the deleterious effect of the large ligand

insertion, 50% mosaics and particularly 75% mosaics were considerably less infectious than 25% mosaics, both in the absence and in the presence of targeting antibodies (Figs. 4 and 5). These results with mosaic AAV2 are in sharp contrast to the transducing titers for all-mutant virus reported by Ried et al. [24]. The specific titers reported for Jurkat and MO7e cells reported in this publication [24] were indeed 20- and 100-fold lower, respectively, than the transducing titers obtained with wild-type capsid virus. Significantly, in the presence of heparin, the transduction with wild-type AAV2 (Figs. 4 and 5) was eliminated. The transduction efficiencies of the viral mosaics in the presence of targeting antibodies, on the other hand, were only modestly affected by the addition of heparin, demonstrating the specificity of the antibody-mediated transduction. Under these conditions, the transduction efficiency with 25% Z34C mosaic

TABLE 1: Particle and transducing titers of wild-type and Z34C-mosaic AAV2 in the presence and absence of targeting antibodies and inhibitors Virus and conditions a

0% 0% + Hep 0% + CD29/CD117 AB 0% + CD29/CD117 AB + Hep 0% + CD29/CD117 AB + Hep + IgG 0% + CD29/CD117 AB + Hep + Prot A 0%-R585/588A + CD29 AB 0%-R585/588A + CD29 AB + Hep 25% 25% + Hep 25% + CD29/CD117 AB 25% + CD29/CD117 AB + Hep 25% + CD29/CD117 AB + Hep + IgG 25% + CD29/CD117 AB + Hep + Prot A 25%-R585/588A + CD29 AB 25%-R585/588A + CD29 AB + Hep 50% 50% + Hep 50% + CD29/CD117 AB 50% + CD29/CD117 AB + Hep 50% + CD29/CD117 AB + Hep + IgG 50% + CD29/CD117 AB + Hep + Prot A 50%-R585/588A + CD29 AB 50%-R585/588A + CD29 AB + Hep 75% 75% + Hep 75% + CD29/CD117 AB 75% + CD29/CD117 AB + Hep 75% + CD29/CD117 AB + Hep + IgG 75% + CD29/CD117 AB + Hep + Prot A 75%-R585/588A + CD29 AB 75%-R585/588A + CD29 AB + Hep

gcp/ml 12

6.8  10

HeLa (tu/ml) 2.2  10

8

3.7  1012 7.5  1012

8.9  107

7.1  1012 7.3  1012

3.3  107

7.2  1012 1.4  1012

1.8  1012

8.0  105

Jurkat (tu/ml) 7

9.5  10 5.0  106 8.3  107 8.0  106 6.4  106 7.1  106 b104 b104 6.4  107 7.7  106 1.4  108 8.5  107 8.6  106 4.7  106 2.3  107 2.4  107 1.9  107 3.4  106 7.7  107 3.2  107 5.4  106 1.4  107 1.6  107 1.9  107 2.1  105 1.9  105 2.3  106 1.1  106 1.8  105 4.1  105 1.3  106 1.5  106

MO7e (tu/ml) 5.1 2.8 5.3 5.5 6.6 5.9

     

106 105 106 105 105 105

5.0 5.1 1.9 9.8 4.2 2.9

     

106 105 107 106 105 105

4.3 9.3 1.1 4.6 4.2 5.9

     

106 105 107 106 105 105

4.7 1.6 1.1 4.2 1.0 1.6

     

105 105 106 105 105 105

Hep, heparin; CD117 AB, antibody against CD117 (c-kit); CD29 AB, antibody against CD29 (h1 integrin); IgG, rabbit IgG; Prot A soluble protein A. In experiments with MO7e cells, antiCD117 antibodies were used; in experiments with Jurkat cells, anti-CD29 antibodies were used. For experimental details see Material and Methods. a %, percentage Z34C capsid protein.

860

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

doi:10.1016/j.ymthe.2005.03.005

ARTICLE

FIG. 4. Specific transduction of MO7e cells with AAV2 mosaics. MO7e cells were transduced with either wildtype or mosaic-capsid virus in the presence or absence of targeting antibody against c-kit (CD117) and inhibitors as indicated. Transduction efficiency was determined by FACS analysis as described under Material and Methods and is shown as percentage GFP-positive cells F standard deviation. The assays were performed in triplicate in the presence or absence of inhibitors as indicated. The percentage of mutant capsid protein in the viral particles (percentage of pDG-Z34C used in virus production) is as indicated. Hep, heparin; CD117, antibody against CD117 (c-kit); IgG, rabbit IgG; Prot A, soluble protein A.

virions was about 11- to 18-fold higher than with wildtype AAV2. As expected, the transduction efficiency with the 25% mosaic was the highest among the mosaic

virions. The transduction efficiency of 50% mosaic virus was reduced approximately 2-fold and the transducing titers of 75% mosaic virus were indeed near background

FIG. 5. Specific transduction of Jurkat cells with AAV2 mosaics. Jurkat cells were transduced with either wildtype or mosaic-capsid virus in the presence or absence of targeting antibody against h1 integrin (CD29) and inhibitors as indicated. Transduction efficiency was determined by FACS analysis as described under Material and Methods and is shown as percentage GFP-positive cells F standard deviation. The assays were performed in triplicate in the presence or absence of inhibitors as indicated. The percentage of mutant capsid protein in the viral particles (percentage of pDG-Z34C used in virus production) is as indicated. Hep, heparin; CD29, antibody against CD29 (h1 integrin); IgG, rabbit IgG; Prot A, soluble protein A.

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

861

ARTICLE

levels. To demonstrate specificity further we also analyzed the transduction efficiencies of the mosaic virions in the presence of targeting antibodies, heparin, and the competitive inhibitors IgG and protein A. Consistent with specific transduction, the addition of either IgG or protein A together with heparin reduced the transduction to background levels (Figs. 4 and 5). Using secreted alkaline phosphatase (SEAP) as a transgene allowed a reduction in virus per cell ratios. Under these conditions the differences between wild-type capsid AAV2 (or 25% mosaics in the absence of antibody) and 25% mosaics targeted to Jurkat cells via an anti-CD29 antibody were even more pronounced. The mosaic AAV produced about 10-fold more SEAP in the presence of targeting antibody than in its absence. If in addition heparin was present, at least 1000-fold more secreted alkaline phosphatase was synthesized (Supplementary Fig. 1). These experiments clearly demonstrate that it is possible to alter AAV tropism with AAV mosaics and to target these mosaics to a specific receptor. Because the wild-type tropism of AAV2 has not been eliminated in these experiments, they do not strictly constitute retargeting. Hence, in a next step, we eliminated AAV2 wildtype tropism by mutating the arginines at positions 585 and 588 to alanines [13,18]. We then tested wild-typeR585/588A and 25, 50, and 75% mosaics carrying the same mutations for their ability to transduce specifically Jurkat cells in the presence of an antibody against CD29.

doi:10.1016/j.ymthe.2005.03.005

As can be seen from Fig. 6, wild-type-R585/588A virus was completely unable to transduce Jurkat cells. On the other hand, 25, 50, and 75% Z34C mosaics efficiently transduced Jurkat cells in the presence of antibodies against CD29. Not surprisingly, this transduction was unaffected by the addition of heparin, an inhibitor of AAV2 wild-type tropism. Addition of protein A or rabbit IgG, however, completely abolished transduction, demonstrating the exquisite specificity achievable. As expected, the transduction efficiencies were highest with 25% mosaics, intermediate with 50% mosaics, and lowest with 75% mosaics. These results show unambiguously that AAV mosaics are useful to retarget AAV to specific cell types.

DISCUSSION The targeting of specific cell types with gene delivery vehicles remains a focus of research in the gene therapy field. Here, we show that viral mosaics allow the specific and efficient transduction of cells that are poorly transduced with AAV2 vectors. In the vast majority of studies in which a targeting ligand has been inserted into the viral capsid the mutant virions demonstrate reduced transducing titers on cell lines permissive to wild-type AAV2 infection [8,16,17, 19,24,26,27,32,34]. In some cases these reductions are drastic or eliminate infectivity altogether, as is the case with the two HA-insertion mutants L4 and L5 [32]. Using

FIG. 6. Retargeting AAV2 mosaics to Jurkat cells. All viruses contained the arginine-to-alanine mutations at positions 585 and 588, eliminating wild-type tropism. Jurkat cells were transduced with either wild-type R585/ 588A or mosaic-capsid virus in the presence or absence of targeting antibody against h1 integrin (CD29) and inhibitors as indicated. Transduction efficiency was determined by FACS analysis as described under Material and Methods and is shown as percentage GFP-positive cells F standard deviation. The assays were performed in triplicate in the presence or absence of inhibitors as indicated. The percentage of mutant capsid protein in the viral particles (percentage of pDGZ34C used in virus production) is as indicated. Hep, heparin; CD29, antibody against CD29 (h1 integrin); IgG, rabbit IgG; Prot A, soluble protein A.

862

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

ARTICLE

doi:10.1016/j.ymthe.2005.03.005

these mutants, we were able to demonstrate that AAV2 mosaics containing both wild-type and mutant capsid proteins are infectious, whereas the all-mutant virions showed no transduction ability. With increasing amounts of wild-type capsid, progressively higher particle titers were obtained with either the L4 or the L5 mosaic. Maybe more importantly, the inclusion of small amounts of wild-type capsid protein rescued the ability of the viruses to transduce cells. Inclusion of higher amounts of wild-type capsid resulted in titers approaching wild-type levels. Similar to the results observed by Rabinowitz et al. [22] using serotype chimeras, the reduction in viral infectivity that we observed when we added increasing amounts of heparin-binding impaired L4-capsid protein was not directly proportional to the ratio of wild-type to mutant capsid protein. Indeed, while the inclusion of 25% L4 protein resulted in virions with gcp/tu ratios that were nearly identical to wild-type-capsid virus, the 50 and 75% mosaics had approximately 8- and 40-fold higher gcp/tu ratios. Similar results could be observed with L5 mosaics. This nonlinear reduction in infectivity also explains why the 50% L4/L5 mosaic displayed very low infectivity. To demonstrate the utility of the concept of viral mosaicism to alter the tropism of AAV2 we took advantage of a system developed previously by Ried et al. [24] that allows the use of receptor-specific antibodies to target AAV to specific cell types. Unfortunately, the introduction of the protein A fragment results in a significant reduction of viral particle and especially transducing titers. Using Z34C/wild-type capsid mosaics and targeting antibodies against the growth factor receptor c-kit (CD117) and h1 integrin (CD29), we were able to transduce specifically and efficiently the human erythroleukemia cell line MO7e and the T cell leukemia-derived Jurkat cells, respectively. The use of Z34C mosaics resulted in up to 10-fold increased titers compared to wild-type virus (Table 1). Even more strikingly, compared with the titers of 100% Z34C virus that have been reported by Hallek and colleagues [24], 25% Z34C titers were about 4 to 5 orders of magnitude more infectious than all-mutant virus, demonstrating the dramatic benefit of using AAV2 mosaic. In addition to showing the efficient and specific transduction of MO7e and Jurkat cells with AAV2 mosaics, the results presented also suggest that ligand density under certain conditions is not a dominant parameter for viral infectivity. While we did not determine a lower limit for the number of ligands necessary for efficient transduction, the high transduction efficiencies obtained with 25% Z34C virus demonstrate that, in this system, an average of 15 ligands per capsid is sufficient. Interestingly, in the case of L4 mosaics the reduction in transducing titers was very significant in the presence of 25% wild-type—HSPG-binding-competent—capsid protein compared to all wild-type capsid virus (Fig. 2A). This is consistent with a model in which the nature of the

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

receptor influences the requirement for ligand density. The residues in the AAV2 capsid that are responsible for HSPG binding have been identified by mutation analysis [13,18,32] and are located on a spike protrusion and a positively charged valley on the capsid surface [13,18,33] located at the threefold axis. It is tempting to speculate that for efficient binding to HSPG, or the triggering of endocytosis, all three monomers must be binding competent. This model could explain the hysteretic effect in transduction efficiencies observed by us and others [22]. For retargeting experiments, the AAV2 wild-type tropism must be eliminated. To achieve this we mutated the arginines at positions 585 and 588 to alanine [13,18]. Wild-type-R585/588A and mosaic-R585/588A virions with these mutations lost the ability to transduce Jurkat cells in the absence of targeting antibody. In the presence of targeting antibody, on the other hand, transduction is very efficient. To our knowledge, this is the first time that the addition of new targeting information and the elimination of wild-type tropism have been combined. These results also set our work apart from studies with serotype chimeras [10,22]. In these cases, the tropism was expanded and could not be restricted to a specific receptor. Clinical applications of AAV-vector-based gene therapy will greatly benefit from methods allowing the transduction of specific cell types with AAV. Here we demonstrate that AAV2 mosaicism is an efficient approach to alter AAV2 tropism and to retarget AAV2 to specific cell types. The use of AAV2 mosaics will undoubtedly facilitate the targeting of cell types that are difficult to achieve with present methods. In particular, the combination of viruses containing a protein A fragment in their capsid proteins with viral mosaicism provides a powerful tool to transduce cell types specifically and efficiently. At least in principle, the versatility of this system is limited only by the availability of specific antibodies of sufficient affinity against suitable receptors.

MATERIAL

AND

METHODS

Cell culture. Low-passage HEK 293 cells (20–40 passages) and HeLa C12 [4,5] cells were cultured in DulbeccoTs modified Eagle medium (DMEM; Cellgro, Hendon, VA, USA). MO7e (a kind gift from Dr. Jonathan Licht, Mount Sinai School of Medicine, New York, NY, USA) and Jurkat clone E6-1 cells (ATCC TIB-152) were maintained in RPMI 1640 (Cellgro). For MO7e cells 100 ng/ml human stem cell factor (R&D Systems, Minneapolis, MN, USA) and 30 ng/ml human IL-3 (R&D Systems) were added to the medium. All media were supplemented with 10% heatinactivated fetal bovine serum (Cellgro), and the cells were grown at 378C and 5% CO2. Construction of plasmids expressing AAV capsid mutants. The constructions of all new plasmids used in this article are described in detail in the supplementary material. Production of mosaic AAV particles. Wild-type capsid AAV-GFP was produced with the two-plasmid transfection protocol [9]. Mosaic-capsid

863

ARTICLE

AAV2 was produced with a modified version of this method that uses three plasmids: wild-type pDG, mutant pDG, and pTR-UF11 (encoding the humanized GFP gene driven by the CMV/chicken h-actin promoter and flanked by AAV ITRs). Plasmid DNA for transfection was produced using the Qiagen MegaPrep Kit as recommended by the supplier (Qiagen, Valencia, CA, USA). To produce AAV-GFP, 293 cells were transfected with the helper plasmid pDG (wild type and/or mutant) and the reporter plasmid pTRUF11 at a molar ratio of 1:1. The percentage of pDG-L4, pDG-L5, or pDGZ34C DNA used for transfections was 0, 25, 50, 75, or 100% of the total amount of helper plasmid used for transfection. To keep the amount of helper plasmid constant, the transfections were supplemented with an appropriate amount of pDG DNA. The same protocol was used to produce viruses that lost their wild-type tropism due to the mutation of both arginines 585 and 588 to alanine. The 293 cells (2.5  107) cells were seeded in a triple-layer flask (Nunc, Naperville, IL, USA). Twenty-four hours later the cells were transfected using the calcium phosphate method and then incubated in DMEM low glucose (Gibco, Grand Island, NY, USA) supplemented with 2% FBS at 378C and 5% CO2. Three days later, the cells were harvested by centrifugation at 1000 rpm (Sorvall RC 3C) for 10 min and washed in PBS, and the cell pellet was resuspended in 5 ml of lysis buffer (0.15 M NaCl, 50 mM Tris–HCl (pH 8.5)). The viral particles were released by freezing and thawing three times. The AAV lysates were treated with 50 U/ml benzonase (Sigma, St. Louis, MO, USA) for 30 min at 378C and then clarified by centrifugation at 3700 g for 30 min. The virions were then purified by iodixanol step gradient [36]. The gradient was formed in Ultra-Clear centrifuge tubes (Beckman, Palo Alto, CA, USA) by first adding 1.55 ml of 60% iodixanol (Sigma) containing 0.01 Ag/ml phenol red (Gibco) and then overlaying, in succession, 1.55 ml of 40% iodixanol in PBS-MK buffer (phosphate-buffered saline, 1 mM MgCl2, and 25 mM KCl), 1.88 ml 25% iodixanol in PBS-MK buffer containing 0.01 mg/ml phenol red, and finally 2.8 ml of 15% iodixanol in PBS-MK buffer containing 1 M NaCl. The lysates were then carefully applied on top of the gradient and the tubes centrifuged for 3 h at 40,000 rpm at 188C in an SW40Ti rotor. The 60%/40% interface phase containing the virus was collected using an 18-gauge needle. Immunoblotting and immunoprecipitation. SDS–PAGE was performed according to Laemmli [15]. Iodixanol-purified AAV samples were mixed with loading buffer, incubated at 958C for 5 min, and loaded on a 7.5% SDS–polyacrylamide gel. For immunoblotting, the proteins were transferred onto a Hybond-P (Amersham, Piscataway, NJ, USA) membrane. After being blocked in TBS (50 mM Tris–HCl, pH 7.5, 150 mM NaCl), 5% (w/v) milk, the membrane was decorated with the monoclonal antibody B1 (1:200 dilution; Research Diagnostics, Flanders, NJ, USA) or one of the monoclonal anti-HA antibodies, HA1.1 (1:500 dilution) or clone 16B12 (1:200 dilution), both from Covance (Berkeley, CA, USA). The bands were visualized by peroxidase-coupled secondary anti-mouse antibody (BioRad, Hercules, CA, USA) at a dilution of 1:10,000 and the ECL+ detection system (Amersham) as suggested by the supplier. In some experiments, after immunodetection with B1 antibody, the membranes were stripped in 62.5 mM Tris–HCl, pH 6.7, 2% SDS, 100 mM h-mercaptoethanol for 30 min at 658C. After several washes in TBS, the membranes were blocked with 5% milk in TBS and reprobed with the monoclonal anti-HA antibody HA1.1 (1:500 dilution). For the immunoprecipitation experiment shown in Fig. 1B, 3.5  109 genome-containing, iodixanol-purified AAV particles were incubated with beads with covalently coupled monoclonal anti-HA antibody (clone 16B12; Covance) in NETN buffer (0.1 M NaCl, 1 mM EDTA, 20 mM Tris– HCl (pH 7.5), 0.5% Nonidet P-40 [32]). The monoclonal antibody against the AU-1 epitope (DTYRYL), conjugated to beads (Covance), served as a negative control. After overnight incubation at 48C, the beads were collected by centrifugation for 1 min at 6800 g at 48C. The beads were washed three times with 1 ml NETN buffer. Sample buffer was then added to the beads and the supernatant and the samples were analyzed by Western blot as described above. For the experiment shown in Fig. 3, 1010 iodixanol-purified, genome-containing particles were immunoprecipitated in an identical fashion. The virions were then eluted by incubating

864

doi:10.1016/j.ymthe.2005.03.005

with 1 mg/ml HA peptide (Anaspec, San Jose, CA, USA) in NETN buffer for 1 h at 378C. The beads were sedimented by centrifugation and the supernatant containing the eluted virus was collected and subjected to real-time PCR quantification and the transduction experiment shown in Fig. 3. Determination of virus particle titers. Titers of DNA-containing particles were determined by real-time PCR [25] using 2 QuantiTect SYBR Green PCR Master Mix (Qiagen) on a LightCycler (Roche, Indianapolis, IN, USA). A DNA plasmid standard curve was set up using 104–107 genome equivalents. For viruses generated with the plasmid pTR-UF11 primers within the GFP open reading frame were used (forward, 5V-ATGAGCAAGG GCGAGGAAC-3V, and reverse, 5V-AGCGCGGGTC TTGTAGTTC-3V). Transducing titers. Transducing titers were determined using the Repand Cap-expressing HeLa derivative C12 [4] using previously published protocols [5,11]. Briefly, C12 cells were seeded in 48-well plates (Falcon, Franklin Lakes, NJ, USA) the day before infection to be 80% confluent the next day. Serial dilutions of purified AAV-GFP viruses or crude lysates were added to the cells in the presence of Ad5 (m.o.i. of 10). After 40 h of incubation at 378C, the cells were harvested, washed in PBS, counted, and fixed in 1% paraformaldehyde. The percentage of GFP-expressing cells was analyzed by flow cytometry (FACScalibur; Becton–Dickinson, San Diego, CA, USA) using CellQuest software (Becton–Dickinson). A minimum of 10,000 cells were analyzed for each sample. The infectious titers were determined using the dilution of AAV that yielded approximately 20% of GFP-positive cells. If the percentage of transduced cells is significantly lower or higher, the titers are no longer in the linear range of the titration curve. Performing all titer determinations with samples with approximately 20% transduced cells eliminates these potential complications and yields the most reproducible titers. Cell-specific transduction. Two models of cell-specific transduction were studied. In a first model, AAV2 was targeted to MO7e cells using a monoclonal mouse anti-human antibody directed against c-kit (CD117). Two different antibodies were used: the clone K45 (Biosource, Camarillo, CA, USA) or the antibodies from hybridoma BA7.3C.9 (ATCC HB-10716) purified via protein A affinity chromatography. In the second system, AAV2 was targeted to Jurkat E6-1 cells with a mouse anti-human h1 integrin (CD29) antibody (clone K20; Dako Cytomation, Carpinteria, CA, USA). Transduction of MO7e cells. MO7e cells (3  105) suspended in 100 Al RPMI were incubated on ice with 1 Ag of the targeting antibody CD117. After 1 h, the cells were washed three times in 1 ml cold RPMI, resuspended in 100 Al cold RPMI, and incubated with AAV2 encoding GFP at a particle-to-cell ratio of 105 gcp/cell. In certain experiments, the samples were supplemented with 100 Ag/ml heparin (Sigma), 10 Ag/ml soluble protein A (Sigma), or 2 Ag/ml rabbit immunoglobulin (Jackson Laboratories, West Grove, PA, USA). The cells were then incubated between 1 and 2 h at 378C, 5% CO2, and reisolated by centrifugation for 5 min at 240 g. The supernatant was discarded and the cells were resuspended in 200 Al RPMI supplemented with 5% serum, 30 ng/ml IL-3, and 100 ng/ml SCF and incubated at 378C, 5% CO2. After 48 h, the cells were counted, the percentage of GFP-expressing cells was determined by flow cytometry (FACScalibur; Becton–Dickinson), and the titers were calculated. Transduction of Jurkat cells. A nearly identical protocol was followed for the specific transduction of Jurkat cells with GFP-expressing AAV, with the following modifications: 4  105 cells were infected with either 10,000 genome-containing particles per cell of mosaic AAV or 40,000 genomecontaining particles per cell of heparan-binding negative mosaic AAV. The three washes were done in RPMI supplemented with 0.1% BSA (Roche) and the final incubation at 378C for 48 h was in RPMI supplemented with 2% serum.

ACKNOWLEDGMENTS This article is dedicated to Josef Brunner (Swiss Federal Institute of Technology, Zqrich, Switzerland), on the occasion of his retirement, for

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

doi:10.1016/j.ymthe.2005.03.005

introducing me (T.W.) to science. We thank Jqrgen Kleinschmidt (Deutsches Krebsforschungszentrum, Heidelberg, Germany) for providing us with the plasmid pDG and Jonathan Licht (Mount Sinai School of Medicine, New York, NY, USA) for giving us MO7e cells. This work was supported by NIH Grants GM062234 and DK062345 (R.M.L) and PO1 HL51811 and PO1 HL59412 and the Edward R. Koger Chair (N.M.). N.M. and K.H.W are investors on patents related to recombinant AAV technology and N.M owns equity in a gene therapy company that is commercializing AAV for gene therapy applications. RECEIVED FOR PUBLICATION JANUARY 18, 2005; ACCEPTED MARCH 7, 2005.

APPENDIX A. SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ymthe. 2005.03.005. REFERENCES 1. Alexander, I. E., Russell, D. W., and Miller, A. D. (1997). Transfer of contaminants in adeno-associated virus vector stocks can mimic transduction and lead to artifactual results. Hum. Gene Ther. 8: 1911 – 1920. 2. Bartlett, J. S., Kleinschmidt, J., Boucher, R. C., and Samulski, R. J. (1999). Targeted adeno-associated virus vector transduction of nonpermissive cells mediated by a bispecific F(abVgamma)2 antibody. Nat. Biotechnol. 17: 181 – 186. 3. Buller, R. M., and Rose, J. A. (1978). Characterization of adenovirus-associated virusinduced polypeptides in KB cells. J. Virol. 25: 331 – 338. 4. Clark, K. R., Voulgaropoulou, F., Fraley, D. M., and Johnson, P. R. (1995). Cell lines for the production of recombinant adeno-associated virus. Hum. Gene Ther. 6: 1329 – 1341. 5. Clark, K. R., Voulgaropoulou, F., and Johnson, P. R. (1996). A stable cell line carrying adenovirus-inducible rep and cap genes allows for infectivity titration of adenoassociated virus vectors. Gene Ther. 3: 1124 – 1132. 6. Gao, G., et al. (2004). Clades of adeno-associated viruses are widely disseminated in human tissues. J. Virol. 78: 6381 – 6388. 7. Girod, A., et al. (1999). Genetic capsid modifications allow efficient re-targeting of adeno-associated virus type 2. Nat. Med. 5: 1052 – 1056. 8. Grifman, M., et al. (2001). Incorporation of tumor-targeting peptides into recombinant adeno-associated virus capsids. Mol. Ther. 3: 964 – 975. 9. Grimm, D., Kern, A., Rittner, K., and Kleinschmidt, J. A. (1998). Novel tools for production and purification of recombinant adenoassociated virus vectors. Hum. Gene Ther. 9: 2745 – 2760. 10. Hauck, B., Chen, L., and Xiao, W. (2003). Generation and characterization of chimeric recombinant AAV vectors. Mol. Ther. 7: 419 – 425. 11. Horster, A., Teichmann, B., Hormes, R., Grimm, D., Kleinschmidt, J., and Sczakiel, G. (1999). Recombinant AAV-2 harboring gfp-antisense/ribozyme fusion sequences monitor transduction, gene expression, and show anti-HIV-1 efficacy. Gene Ther. 6: 1231 – 1238. 12. Kashiwakura, Y., et al. (2005). Hepatocyte growth factor receptor is a coreceptor for adeno-associated virus type 2 infection. J. Virol. 79: 609 – 614. 13. Kern, A., et al. (2003). Identification of a heparin-binding motif on adeno-associated virus type 2 capsids. J. Virol. 77: 11072 – 11081. 14. Klein, R. L., et al. (1998). Neuron-specific transduction in the rat septohippocampal or nigrostriatal pathway by recombinant adeno-associated virus vectors. Exp. Neurol. 150: 183 – 194. 15. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227: 680 – 685.

MOLECULAR THERAPY Vol. 11, No. 6, June 2005 Copyright C The American Society of Gene Therapy

ARTICLE

16. Muller, O. J., et al. (2003). Random peptide libraries displayed on adenoassociated virus to select for targeted gene therapy vectors. Nat. Biotechnol. 21: 1040 – 1046. 17. Nicklin, S. A., et al. (2001). Efficient and selective AAV2-mediated gene transfer directed to human vascular endothelial cells. Mol. Ther. 4: 174 – 181. 18. Opie, S. R., Warrington, K. H., Jr., Agbandje-McKenna, M., , Zolotukhin, S., and Muzyczka, N. (2003). Identification of amino acid residues in the capsid proteins of adeno-associated virus type 2 that contribute to heparan sulfate proteoglycan binding. J. Virol. 77: 6995 – 7006. 19. Perabo, L., et al. (2003). In vitro selection of viral vectors with modified tropism: the adeno-associated virus display. Mol. Ther. 8: 151 – 157. 20. Ponnazhagan, S., Mahendra, G., Kumar, S., Thompson, J. A., and Castillas, M., Jr. (2002). Conjugate-based targeting of recombinant adeno-associated virus type 2 vectors by using avidin-linked ligands. J. Virol. 76: 12900 – 12907. 21. Qing, K., Mah, C., Hansen, J., Zhou, S., Dwarki, V., and Srivastava, A. (1999). Human fibroblast growth factor receptor 1 is a co-receptor for infection by adeno-associated virus 2. Nat. Med. 5: 71 – 77. 22. Rabinowitz, J. E., Bowles, D. E., Faust, S. M., Ledford, J. G., Cunningham, S. E., and Samulski, R. J. (2004). Cross-dressing the virion: the transcapsidation of adeno-associated virus serotypes functionally defines subgroups. J. Virol. 78: 4421 – 4432. 23. Rabinowitz, J. E., and Samulski, R. J. (2000). Building a better vector: the manipulation of AAV virions. Virology 278: 301 – 308. 24. Ried, M. U., Girod, A., Leike, K., Buning, H., and Hallek, M. (2002). Adenoassociated virus capsids displaying immunoglobulin-binding domains permit antibody-mediated vector retargeting to specific cell surface receptors. J. Virol. 76: 4559 – 4566. 25. Rohr, U. P., Wulf, M. A., Stahn, S., Steidl, U., Haas, R., and Kronenwett, R. (2002). Fast and reliable titration of recombinant adeno-associated virus type-2 using quantitative real-time PCR. J. Virol. Methods 106: 81 – 88. 26. Shi, W., Arnold, G. S., and Bartlett, J. S. (2001). Insertional mutagenesis of the adenoassociated virus type 2 (AAV2) capsid gene and generation of AAV2 vectors targeted to alternative cell-surface receptors. Hum. Gene Ther. 12: 1697 – 1711. 27. Shi, W., and Bartlett, J. S. (2003). RGD inclusion in VP3 provides adeno-associated virus type 2 (AAV2)-based vectors with a heparan sulfate-independent cell entry mechanism. Mol. Ther. 7: 515 – 525. 28. Starovasnik, M. A., Braisted, A. C., and Wells, J. A. (1997). Structural mimicry of a native protein by a minimized binding domain. Proc. Natl. Acad. Sci. USA 94: 10080 – 10085. 29. Summerford, C., Bartlett, J. S., and Samulski, R. J. (1999). AlphaVbeta5 integrin: a coreceptor for adeno-associated virus type 2 infection. Nat. Med. 5: 78 – 82. 30. Summerford, C., and Samulski, R. J. (1998). Membrane-associated heparan sulfate proteoglycan is a receptor for adeno-associated virus type 2 virions. J. Virol. 72: 1438 – 1445. 31. Warrington, K. H., Gorbatyuk, O. S., Harrison, J. K., Opie, S. R., Zolotukhin, S., and Muzyczka, N. (2004). Adeno-associated virus type 2 VP2 capsid protein is nonessential and can tolerate large peptide insertions at its N terminus. J. Virol. 78: 6595 – 6609. 32. Wu, P., et al. (2000). Mutational analysis of the adeno-associated virus type 2 (AAV2) capsid gene and construction of AAV2 vectors with altered tropism. J. Virol. 74: 8635 – 8647. 33. Xie, Q., et al. (2002). The atomic structure of adeno-associated virus (AAV-2), a vector for human gene therapy. Proc. Natl. Acad. Sci. USA 99: 10405 – 10410. 34. Yang, Q., et al. (1998). Development of novel cell surface CD34-targeted recombinant adenoassociated virus vectors for gene therapy. Hum. Gene Ther. 9: 1929 – 1937. 35. Zhong, Q., et al. (2001). Efficient c-kit receptor-targeted gene transfer to primary human CD34-selected hematopoietic stem cells. J. Virol. 75: 10393 – 10400. 36. Zolotukhin, S., et al. (1999). Recombinant adeno-associated virus purification using novel methods improves infectious titer and yield. Gene Ther. 6: 973 – 985.

865