Effect of Genome Size on AAV Vector Packaging (2024)

Abstract

Adeno-associated virus (AAV) vector genomes have been limited to 5 kilobases (kb) in length because their packaging limit was thought to be similar to the size of the parent AAV genome. Recent reports claim that significantly larger vector genomes can be packaged intact. We examined the packaged vector genomes from plasmid-encoded AAV vectors that ranged from 4.7 to 8.7kb in length, using AAV types 2, 5, and 8 capsids. Southern blot analysis indicated that packaged AAV vector genomes never exceeded 5.2kb in length irrespective of the size of the plasmid-encoded vector or the capsid type. This result was confirmed by vector genome probing with strand-specific oligonucleotides. The packaged vector genomes derived from plasmid-encoded vectors exceeding 5kb were heterogeneous in length and truncated on the 5′ end. Despite their truncated genomes, vector preparations produced from plasmid-encoded vectors exceeding 5.2kb mediated reporter gene expression in vitro at high multiplicity of infection (MOI). The efficiency of expression was substantially lower than that of reporter vectors with genomes <5kb in length. We propose that transcriptionally functional, intact vector genomes are generated in cells transduced at high MOI from the fragmentary genomes of these larger vectors, probably by recombination.

Introduction

The adeno-associated viruses (AAVs) are single-stranded DNA viruses with genomes that are ~4.7 kilobases (kb) in length. They comprise the Dependovirus branch of the parvovirus family and require co-infection with a helper virus, typically adenovirus or herpes virus, to replicate. Recombinant vectors derived from AAVs are attractive for the development of human therapeutics because they mediate long-term gene expression and the parent AAVs produce no known pathology in humans. AAV vectors were recently used to successfully treat patients with Leber congenital amaurosis, a severe form of genetically acquired childhood blindness.^1,2,3 A substantial limitation of AAV vectors is their small packaging capacity that is generally considered to be <5kb. This limitation makes vector design for diseases involving larger genes challenging or impossible. Larger genes are relatively common and are involved in diseases such as cystic fibrosis, duch*enne muscular dystrophy, hemophilia A, and some forms of Leber congenital amaurosis and retinitis pigmentosa. To address this problem, strategies based on trans-splicing^4,5,6 and hom*ologous recombination⁷ were developed. All of these methods require multiple vectors, and the efficiency of some of these methods is inherently limited by the efficiencies of intron processing or hom*ologous recombination. Two recent reports suggest that AAV capsids are capable of packaging substantially larger vector genomes in an intact form. Grieger and Samulski⁸ reported that AAV types 1–5 can package up to 6kb DNA. Most recently, Allocca et al.⁹ reported that AAV type 5 can package up to 8.9kb DNA and mediate efficient transgene expression both in vitro and in vivo.

In this study, we evaluated the packaged genome size and in vitro expression characteristics of AAV vectors ranging from 4.7 to 8.7kb in length. The vectors examined were a mouse CEP290 vector and a set of lacZ reporter vectors of increasing length. These were packaged into AAV types 2, 5, and 8 capsids using the three-plasmid transfection method. We found that the size of packaged vector genome DNAs never exceeded 5.2kb no matter what size vector was encoded by the production plasmid. This was true for all capsid types and vectors tested. The packaged vector genome DNA from plasmid-encoded vectors exceeding 5.2kb consisted of a heterogeneous population of genomes with a broad size distribution and a maximum length of ~5kb. The heterogeneity was generated by deletion on the 5′ end. Vector preparations consisting of heterogeneous fragmentary vector genomes were capable of gene expression in vitro when used at high multiplicity of infection (MOI). The expression efficiency of these vector preparations was substantially lower than vector preparations with genomes that were <5kb in length.

Results

Discussion

Several groups have examined AAV packaging and have tried to assess the maximum packaging capacity of AAV vectors. Dong et al.¹⁴ were the first to systematically address the subject and reported that AAV type 2 vectors made from plasmid-encoded vectors exceeding 4.9kb showed substantially reduced expression when used to transduce cells in vitro. Subsequently, Grieger and Samulski⁸ reported that genomes up to 6.0kb could be packaged in AAV types 1–5 capsids, but that vectors exceeding 5.2kb were largely degraded in infected cells by the proteasome. Two years later, Wu et al.¹³ reported that a self-complementary vector with an intact 6.6kb genome could be produced using the AAV type 2 capsid. Most recently, Allocca et al.⁹ reported that AAV type 5 was able to package intact AAV genomes up to 8.9kb in length and that these vector preparations mediated gene expression in vitro and in vivo.

In this study, we investigated ability of AAV types 2, 5, and 8 capsids to package nine different plasmid-encoded AAV vectors. The vectors examined were an 8.6kb mouse CEP290 vector and a set of lacZ reporter vectors that ranged from 4.7 to 8.7kb in length. In contrast to some of the previous studies, we found that maximum size of the packaged vector DNA never exceeded ~5.2kb, no matter what size plasmid-encoded vector or AAV serotype was used. Given the limit of sensitivity of our Southern analysis (0.33% of the applied vector genome DNA), if intact 8.6kb CEP290 or 8.7kb lacZ genomes were packaged, they comprised <1 part in 300 (by mass) of the total packaged DNA. Additionally, we found that vector genomes isolated from vector preparations produced with plasmid-encoded vectors exceeding 5.2kb were heterogeneous with a broad size distribution and had a maximum length of ~5kb. Strand-specific probing of packaged CEP290 vector genomes indicated they were exclusively derived from the 5kb region located at the 3′ end of the vector genomes (of both polarities). Apparently, the genomes were truncated at the 5′ end. This may be due to the fact that AAV vectors package genomes unidirectionally into preformed capsids starting with the 3′ end.¹⁵ With larger vectors, the packaging process may stall after the capsid is full. The remaining unpackaged 5′ end of the genome may then be removed by cellular nucleases.

We then asked whether these fragmentary vector genomes, packaged in AAV type 2 capsids, were capable of mediating gene expression in cells. Somewhat surprisingly, all of the AAV2 lacZ vector preparations mediated some degree of reporter gene expression when used to transduce 293 cells at MOIs exceeding 1. The reason that vector preparations with fragmentary genomes mediate gene expression is not clear. A likely possibility is that overlapping vector genomes of opposite polarities anneal and regenerate full-length vector genomes in cells that are infected by multiple vector particles (Figure 5). Generation of full-length vector genomes may also occur by hom*ologous recombination. Multiple infections are probably common at the high MOIs used in our experiment. Intracellular vector recombination of AAV vectors has been previously described.⁷

The location of the 4.4kb CMV lacZ cassette in the vector genome had a large effect on transduction activity of the vectors. Vector preparations produced from 5.2 to 8.7kb plasmid-encoded vectors with intact lacZ expression cassettes and one ITR located entirely within 5,050bp of continuous DNA-mediated expression are about tenfold more efficiently than vector preparations derived from plasmid-encoded vectors that did not have intact lacZ expression cassettes and one ITR located entirely within 5,050bp of continuous DNA. The higher efficiency of vectors with ITR-proximal lacZ expression cassettes may indicate that these vectors simply package the first 5,050bp from the 3′ end, which encodes one ITR and an intact lacZ expression, and that this partial genome is capable of mediating gene expression without having to anneal to a second vector. This idea is illustrated in Figure 6.

The transduction efficiency of the AAV2-lacZ-C8.7k vector preparation alone demonstrated a very high sensitivity to MOI, jumping 18-fold, versus two- to fourfold for the other vectors, in response to a fivefold increase in MOI. This vector and the AAV2-lacZ-C6.7k vector do not have intact lacZ expression cassettes within 5.2kb of either end of their genomes and would likely require intracellular recombination to regenerate an intact lacZ expression cassette. Assuming a maximum packaged genome size of 5.2kb, the AAV2-lacZ-C8.7k vector would have the smallest overlap between genomes of opposite polarities (1.7kb versus 3.7kb for the AAV2-lacZ-C6.7k vector). This smaller overlap may have made the expression efficiency of this vector more MOI sensitive. The genome populations in these vector preparations are heterogeneous with respect to size. Certainly, the smaller overlap would reduce the fraction of genomes capable of annealing within the cell.

It is also unclear why our data do not agree with previous reports that vector genomes as large as 8.9kb can be packaged and are functional for transduction. To control for differences in packaging plasmids and vector purification techniques, we repeated our experiments with two other sets of packaging plasmids^9,16 and one other purification protocol¹⁷ with no change in outcome (data not shown).

In conclusion, we found no evidence that AAV capsids efficiently package vector genomes that are substantially larger than the wild-type viral genomes. The limitations imposed by the packaging limits of AAV capsids are major hindrances to the development of many otherwise promising and clinically relevant applications. If AAV vector-based gene delivery is to be used for these applications, strategies need to be developed to accommodate the genome size limitation. The recent reports of large genome packaging usually come with steep reductions in packaging or transduction efficiency. Considering our data, or the data of Dong et al.,¹⁴ this is not surprising. More work certainly needs to be done to see whether capsid or genome modifications can be found that will allow the efficient packaging of larger genomes. But we believe that the current trend of simply trying to package larger genomes into the commonly used AAV capsids is akin to fitting a small shoe onto a large foot.

Materials and Methods

AAV vector constructs and production. AAV2 ITRs were used in all AAV vector constructs. The mouse CEP290 complementary DNA was provided by Anand Swaroop (National Eye Institute/National Institutes of Health). To clone the mCEP290 complementary DNA into an AAV vector plasmid, PCR was carried out using the following primers: 5′-AGT CAT CGA TGC CAC CAT GCC ACC TAA TAT AAA GTG GAA AGA ATT AA-3′ and 5′-AGT CAG ATC TCT AAT AAA TAG GGA AAC TAT-3′. The PCR product was digested with Cla I and Bgl II, and then inserted into an AAV vector plasmid with an expression cassette consisting of a CMV promoter and a human β-globin polyadenylation site. The parent plasmid was described previously and was used for enhanced green fluorescent protein expression.¹⁸ The resulting plasmid was designated pV4.6c mCEP290. The plasmid encoding the 4.7kb, CMV lacZ vector was previously described.¹⁹ Phage lambda DNA of different lengths was used to create the larger lacZ vectors. PCR was used to create lambda DNA fragments of the appropriate sizes. The PCR products were inserted with either 3′ or 5′ of the CMV lacZ expression cassette. The designations of the resulting vectors are shown in Figure 3.

AAV production and purification was carried out as previously described.¹² Real-time PCR was employed to titer the vectors. The following fluorescence-labeled probe (5′-TCC AAA ATG TCG TAA CAA CT-3′) and primers (5′-TGG GAG TTT GTT TTG CAC CAA-3′ and 5′-CGC CTA CCG CCC ATT TG-3′) were used for all vectors.

Alkaline gel electrophoresis and Southern blot analysis of AAV vector DNA. DNA was extracted from 1 × 10⁹ or 1 × 10¹⁰ full vector particles as previously described.⁹ Before extraction, the vector particles were treated with DNase I to remove unpackaged DNA. Alkaline agarose gel electrophoreses was performed as previously described.²⁰ DIG-labeled (Roche, Indianapolis, IN) or ³²P-labeled probes were employed for DNA hybridization. To analyze AAV mCEP290 vectors, a 408bp Eco RI fragment on pV4.6c mCEP290 was used as a probe. The 5.1 and 8.3kb markers were produced by digestion of the pV4.6c mCEP290 with Sac II and Sph I, or Not I alone. To analyze AAV LacZ vectors, a 657bp Spe I-Age I fragment from the vector plasmid was used as a probe. The 5.1 and 8.6kb markers were generated by Sma I digestion of pAAV-lacZ-I5.2k and pAAV-lacZ-C8.7k plasmids, respectively.

Strand-specific dot-blot hybridization of AAV5 mCEP290 vector. AAV5 mCEP290 vector DNA was extracted as previously described.⁹ Extracted vector DNA from 2.3 × 10⁹ vector genomes was applied to the first well of the manifold. Twofold serial dilutions of the extracted vector DNA were applied to the remaining seven wells of the manifold. The corresponding amounts of vector DNA applied to the eight wells of the manifold were 7, 3.5, 1.7, 0.86, 0.43, 0.21, 0.11, and 0.05ng. Standard DNA generated by Sma I digestion of pV4.6c mCEP290 was diluted in the same fashion. The amounts of plasmid DNA dotted onto the eight wells of the manifold were 50, 25, 12.5, 6.25, 3.1, 1.6, 0.78, and 0.39ng. Both vector and plasmid DNA were denatured by 0.25mol/l NaOH before loading to nylon membrane. Single-strand oligonucleotides corresponding to several positions in the vector genome (Table 1) were designed, synthesized, and labeled using DIG Oligonucleotide 3′-End Labeling Kit (Roche). Hybridization was carried out overnight at 10°C below the melting temperature of each probe. After probing, the membrane was stripped of DIG-labeled probe by washing twice with 0.2mol/l NaOH/0.1% sodium dodecyl sulfate for 15 minutes at 37°C and rinsing in 2× saline-sodium citrate for 5 minutes.

Infection and quantification of transduced cells. HEK 293 cells were seeded on 24-well plates at a concentration of 1 × 10⁵ cells per well 40 hours before infection. The cells were transfected with pAdeno5, plasmid encoding the E2A, VA RNA, and E4 regions of adenovirus type 2 (ref. 19). The cells were then infected with the AAV2 LacZ vectors at MOIs of 2 × 10³ or 1 × 10⁴. Etoposide was added to each well to produce a final concentration of 20nmol/l. After 24 hours, the cells were fixed with phosphate-buffered saline containing 2% formaldehyde and 0.2% glutaraldehyde for 5 minutes at room temperature. After fixation, lacZ expression was evaluated by histochemical staining with a solution of 1mg/ml X-gal in phosphate-buffered saline containing 5mmol/l K₃Fe(CN)₆, 5mmol/l K₄Fe(CN)₆3H₂O, and 2mmol/l MgCl_2. The cells were stained for 4 hours at 37°C. Eight fields per well were counted, and the average number of blue cells per well was determined.

Acknowledgments

We thank Chun Y Gao of the National Eye Institute for assistance with cell imaging.

REFERENCES