Generation and Characterization of Chimeric Recombinant AAV Vectors

Abstract

Although most animal experiments with recombinant adeno-associated virus (AAV) vectors have been based on AAV serotype 2, recent studies showed that AAV vectors based on AAV serotype 1 performed more efficiently in muscle and other tissues. On the other hand, AAV2-based vectors can be readily purified by heparin column. To combine the advantages of both types of vectors, we developed a strategy to generate chimeric vectors by using a mixture of AAV helper plasmids encoding both serotypes in the transfection process. Because the AAV packaging machinery cannot distinguish between closely related AAV1 and AAV2 capsid proteins, each packaged virion contains capsid proteins from both serotypes. As expected, the resulting chimeric vectors could be purified by heparin column. Neutralizing antibody assays showed that the chimeric vectors can be inhibited by either AAV1 or AAV2 antiserum. In vivo, the chimeric vectors direct levels of expression similar to those of AAV1 in muscle or AAV2 in liver; that is, they combine the best transduction characteristics of both parent vectors. In summary, this study provides a straightforward method for combining various properties of different AAV serotypes into one vector. Potential limitations of the chimeric vectors are also discussed.

Introduction

Adeno-associated virus (AAV) is a member of the Parvovirus family. Unlike autonomous parvoviruses, AAV is capable of establishing a latent infection in which its genome is incorporated into host chromosome [1]. For the most studied serotype, AAV type 2, such integration events frequently occur in a site-specific manner at human chromosome 19q13.3 [2]. Because of its defective nature, AAV generally requires a helper virus such as adenovirus to complete its lytic infection. Due to the limited size of its virion, AAV is capable of packaging approximately 4700 nucleotides. In the AAV genome, the major cis element is the two copies of AAV terminal repeats. They are critical for AAV packaging, replication, and integration and they are the only AAV sequences retained in recombinant AAV vectors.

Up to now, all five original AAV serotypes have been characterized and sequenced. AAV1 and AAV4 are of nonhuman primate origin based on seroepidemiology studies. Seropositivities to AAV2, AAV3, and AAV5 are found widely in the human population [3]. All these AAV serotypes share considerable homology in both rep and cap genes except AAV5, which is a more distal member of the dependovirus family [4]. Despite their homologies, different AAV serotypes appear to have distinctive tissue tropisms and use their own receptors for entry into host cells [5]. In addition, a sixth AAV serotype has been isolated and shown to be a recombination product of AAV1 and AAV2 [6,7]. However, the exact nature of this hybrid virus remains unknown.

Recombinant AAV vectors are becoming popular since they are able to direct long-term gene expression without destructive T cell responses [8,9]. A variety of transgenes using AAV vectors have been tested in brain, liver, muscle, and other tissues and positive results have been published [10-14]. In phase I clinical trials with AAV vectors for hemophilia B, AAV delivered to muscle exhibited no cytotoxic effect and showed evidence of transgene expression in the initial low doses [15].

AAV2 was the first AAV serotype fully cloned and sequenced [16,17]. Virtually all AAV vectors were initially based on AAV serotype 2 since it was the only available AAV vector. The receptors and coreceptors identified for AAV2 include HSPG, FGFR, and integrin α_vβ₅ [18-20]. However, with the successful characterization of additional AAV vectors, the use of alternative AAV serotypes for gene transfer is becoming popular. In many reports, alternative AAV serotypes exhibit improved performance over AAV serotype 2. AAV5 appears to be a better vector for gene delivery to the brain, liver, and airway epithelial cells [4,21-23]. It is perceived that the AAV receptors largely determine the vector’s performance in vivo; presumably these tissues have more receptors for AAV5. Our previous studies demonstrated that AAV1 was better suited to muscle-based gene transfer [6]. The improvement of AAV1 over AAV2 in muscle is in the range of 10- to 100-fold depending on the half-life of the transgenes [5,24]. On the other hand, AAV2 is generally a better vector than AAV1 for liver-directed gene transfer [6].

In the current study, we tried to design a strategy to combine the advantages of AAV1 and AAV2 vectors by generating new vectors that are chimeric products of AAV1 and AAV2. The extraordinarily high level of similarity of AAV1 and AAV2 allows two kinds of cap proteins to be packaged indiscriminately into the same mature virions. Using human factor IX and human α1-antitrypsin as reporter genes we show that the new vectors possess properties that allow transduction of both liver and muscle. In addition, the new vectors can be purified on a heparin column. Our results suggest that such chimeric vectors could be used for delivering the factor IX gene and other genes for human gene therapy.

Results

Strategy to Generate Chimeric Vector Based on AAV1 and AAV2

Generally, AAV vectors are made by using a three-component system that includes a vector construct, a helper virus construct, and an AAV helper construct [29,30]. The vector construct is a plasmid that has the cis signal from AAV for replication and packaging. The helper virus constructs provide necessary helper function from adenovirus, including E1a, E1b, E2a, E4, and VA RNA. The AAV helper construct will supply the rep gene and cap gene products to complement the replication and packaging process. In ordinary AAV production systems, only one AAV helper is used. In our novel approach, we tried to make rAAV using a mixture of AAV1 helper and AAV2 helper. This strategy is illustrated in Fig. 1. The AAV1 helper includes the AAV2 rep gene and AAV1 cap gene while the AAV2 helper carries AAV2 rep gene and AAV2 cap gene. Since AAV1 and AAV2 cap genes share more than 80% identity at the amino acid level, we hypothesized that the AAV packaging machinery would not be able to distinguish the two types of capsid proteins in the cell, resulting in assembly of a chimeric virion made of both AAV1 and AAV2 capsid proteins. Because the AAV capsids are expressed very abundantly in the late stage of AAV replication and packaging, statistically, the vast pool of available capsid proteins will allow a near uniform composition of AAV1 and AAV2 capsid proteins in most packaged virions.

FIG. 1.

(A) An illustration of the strategy to produce chimeric recombinant AAV vectors based on AAV1 and AAV2. AAV1 and AAV2 genomes and expressed proteins are shown in different colors. The mRNAs for Rep78, 68, 52, and 40 and VP1, 2, and 3 are indicated. Promoters P5, P19, and P40 are marked by arrows. (B) The helper effects on the composition of chimeric AAV vectors. The hypothetical composition of vector capsid is calculated based on the different ratios of helper plasmids used for transfection. A total of 50 μg DNA was used to transfect a 15-cm dish of 293 cells. For example, AAV12_1:9 can be made by transfecting 5 μg AAV1 helper and 45 μg AAV2 helper along with adenovirus miniplasmid and vector plasmid into 293 cells.

Production and Purification of Chimeric AAV Vectors Based on AAV1 and AAV2

To test the above hypothesis, we designed the experiments outlined in Fig. 1B. In addition to the vector plasmids and adenovirus helper plasmids, we used different combinations of AAV helpers as shown in the table. The vector plasmid, helper plasmids mix, and adenovirus helper plasmids were cotransfected into human 293 cells. At 96 h posttransfection, the cells were harvested and the cell pellets were collected. To determine vector yield, we purified vectors by standard CsCl gradient. All vectors in a density of 1.38 -1.42 g/ml were collected. In accordance with previous results [6], yield of AAV1 vector is three- to fourfold lower than that of AAV2. The chimeric vector yields are in the range of AAV1 and AAV2 yields depending on the composition of the helpers. Although it is anticipated that the mixing of helpers would not interfere with AAV packaging, the 1:1 ratio of AAV1:AAV2 consistently gave lower yield than either AAV1 and AAV2 (Fig. 2).

FIG. 2.

Production of chimeric vectors. AAV1 and AAV2 chimeric vectors were generated by cotransfection of pAAV.F.IX, pAd, and AAV helper into 3 × 10⁸ 293 cells. The ratios of AAV1 and AAV2 helper are 10:0, 9:1, 1:1, 1:9, and 0:10. The vectors were purified by CsCl gradient. Vectors at the density of 1.38 -1.42 g/ml were collected. The titer was determined by quantitative PCR. The experiment was repeated three times and similar results were obtained. The y axis shows the total amount of vectors isolated in genomes. The standard deviation is identified.

Heparin sulfate proteoglycan has been identified as a receptor for AAV2 [20]. The strong affinity of AAV2 for heparin allows AAV2 to be purified by heparin column. AAV vectors purified by this method can reach much higher purity than those processed by CsCl gradient. AAV1, in contrast, has no significant affinity to heparin. Therefore, heparin-affinity chromatography cannot be used to purify vectors based on this serotype. Since AAV2 capsid proteins would be incorporated into the chimeric virion, we hypothesized that the presence of these AAV2 peptide epitopes would allow the purification of these novel vectors by heparin-affinity chromatography. To test this hypothesis, we attempted AAV2 heparin column purification procedures to isolate chimeric AAV vectors (Fig. 3). AAV1 cannot be purified by heparin column. A chimeric vector with 90% AAV1 capsid proteins (AAV12_9:1) also has no significant affinity to heparin column. These two vectors can be purified only by CsCl gradient based on their buoyant density (Fig. 2). AAV12_1:1, AAV12_1:9, and AAV2 vectors all showed affinity to heparin column. The elution peak for AAV12_1:1 shifted to a lower salt concentration, indicating the existence of chimeric particles with fewer AAV2 capsid protein molecules in each individual particle compared to pure AAV2. This results in a lower affinity to heparin. If individual AAV1 and AAV2 populations were present in the preparation, the recovered vector would have appeared in the same fraction as the native AAV2 vector. The different binding properties to heparin suggest that the virion particles indeed contain different amounts of AAV1 and AAV2 capsid proteins, corresponding to the ratio used in virus production.

FIG. 3.

Purification of chimeric vectors by heparin column. The AAV1 and AAV2 chimeric vectors were generated by cotransfection of pAAV.F.IX, pAd, and AAV helper into 3 × 10⁸ 293 cells. The ratios of AAV1 and AAV2 helper are 10:0, 9:1, 1:1, 1:9, and 0:10. The vectors were purified by using 1.5 g heparin resin per sample. Vectors were eluted by a NaCl gradient ranging from 200 mM to 1 M. The elution volume for each fraction was 10 ml. The vector titer was determined by quantitative PCR. The experiment was repeated three times. Shown is the vector distribution in the elution gradient. The y axis shows vector titer (genomes/ml) in each fraction. To obtain the total vector yield, it must be multiplied by a factor of 10. The x axis is the salt concentration used for elution. The mean yield and standard deviation in each fraction are calculated from three independent experiments.

Characterization of Antigenic Properties of Chimeric AAV Vectors in Vitro

To define the antigenic properties of chimeric vectors, we tested their reactions to anti-AAV sera. COS cells were plated 24 h before infection with rAAV vectors at an m.o.i. of 5000 particles per cell. AAV1, AAV2, and chimeric vectors made at three different ratios of helper as outlined in Fig. 1B were incubated with AAV1 and AAV2 antisera from mouse at a dilution of 1:100 before infection. The control groups were incubated with normal mouse sera at the same dilution. Secreted factor IX was determined 48 h postinfection by ELISA. The results are presented in Table 1. AAV1 neutralizing antibody (NAB) had little effect on the transduction capability of AAV2 vectors as evidenced by the fact that it reduced expression by only 5%. However, it is quite effective at inhibiting the chimeric AAV12_1:1 since expression was reduced by approximately 90%. Similar results were observed with AAV2 NAB since expression from AAV12_1:1 was also reduced by more than 90%. The data demonstrate that the chimeric vectors can be neutralized by both AAV1 and AAV2 NAB, which is in contrast to either pure AAV1 or pure AAV2 particles. Since both AAV1 and AAV2 NAB can affect the chimeric vectors, the data confirm that our production method produced a vector with both AAV1 and AAV2 capsid proteins on the surface.

TABLE 1.

Characterization of chimeric AAV1/2 vectors in vitro

	AAV1	AAV12_9:1	AAV12_1:1	AAV12_1:9	AAV2
AAV1 antiserum	30.5 ± 17.2%	26.1 ± 8.4%	11.3 ± 1.9%	85.2 ± 5.4%	96.2 ± 6.2%
AAV2 antiserum	85.2 ± 17.6%	78.5 ± 11.8%	3.4 ± 2.4%	1.9 ± 1.8%	3.7 ± 3.8%

COS cells were plated 24 h before infection with rAAV vector. The vectors AAV1, AAV2, and chimeric AAV1 and AAV2, at ratios of 9:1, 1:1, and 1:9, were incubated with AAV1 and AAV2 antisera at dilution of 1:100. The control groups were incubated with normal mouse serum at the same dilution. Secreted factor IX was measured 48 h postinfection by ELISA. The data show the remaining expression following the incubation with AAV antiserum compared to the normal mouse serum control.

Characterization of the Performance of Chimeric AAV12 Vectors in Vivo

To investigate how these chimeric vectors behave in vivo, we injected approximately 5 × 10¹⁰ vector particles of each of the vectors AAV1, AAV2, and chimeric AAV1/AAV2, purified by CsCl gradient, at ratios of 9:1, 1:1, and 1:9 into the muscles of CD4 KO mice. We selected human factor IX or human α1-antitrypsin as the secretion marker for the study. CD4 KO mice were chosen for muscle to avoid the confounding influence of neutralizing antibodies generated against human factor IX or human α1-antitrypsin in immunocompetent mice via intramuscular injections. In other experiments, we used C57BL6 mice to test the performance of vectors in liver. The experiments are summarized in Table 2. Since these experiments were repeated using different batches of vectors and the absolute expression varies, we define the performance of AAV1 in liver and of AAV2 in muscle as 1. Since the expression from AAV1 in liver and from AAV2 in muscle is the lowest, the expression from other vectors is shown as “fold” compared to the standard. The chimeric vectors exhibited improved performance over AAV2 in muscle. Although the AAV12_1:1 vector consists of only 50% AAV1 capsid proteins, expression levels were nearly the same as those of AAV1, which is approximately 4- to 7-fold higher than AAV2 in this study. As a comparison, the same amount of each vector was administered to the liver of C57BL6 mice via tail vein injection. As shown in Table 2, transduction with chimeric vectors containing at least 50% AAV2 capsid protein results in higher levels of transgene expression. Surprisingly, the expression of human factor IX or human α1-antitrypsin from chimeric vector improved to match the best performance of their parental vectors in the liver, which exhibited approximately 10-fold increases over that of AAV1. The data suggest that the chimeric vectors combine useful features of both AAV1 and AAV2.

TABLE 2.

Summary of the performance of chimeric AAV vectors in muscle and liver

	Liver		Muscle
	a1at	hFIX	a1at	hFIX
AAV1	1	1	4.6-6.3	5.4-8.9
AAV12_9:1	1.6-10.3	1.4-5.4	2-3.4	1.8-3.1
AAV12_1:1	3.8-17.4	6.6-23.7	8.5-11.7	2.3-7.1
AAV12_1:9	2.4-13.5	5.3-22.7	1.4-2.9	1.6-3.0
AAV2	2.1-5.7	2.7-12.2	1	1

Data from two parallel experiments using either human factor IX (hFIX) or human α1-antitrypsin (a1at) as reporter genes, levels of which were assayed by ELISA, are presented. Approximately 5 × 10¹⁰ particles were injected per mouse. For studies targeting liver, vectors were injected intravenously into C57BL6 mice. For muscle studies, factor IX vectors were administered to CD4 KO mice while a1at vectors were injected intramuscularly into C57BL6 mice. The range of expression in fold compared to that of AAV1 in liver and AAV2 in muscle at week 6 postinjection is shown. The basal expression levels of transgene expression for human factor IX and human α1-antitrypsin were designated “1”. The following vectors were used: CsCl-purified AAV1, AAV2, and chimeric AAV1 and AAV2 at ratios of 9:1, 1:1, and 1:9.

Discussion

AAV vectors are very promising in gene therapy of genetic diseases. As shown in the initial result from a phase I clinical trial of hemophilia B, the doses used in initial human clinical trial did not allow the expression of sufficient amounts of factor IX (FIX) in the subjects [15]. Therefore, any improvements in vectors that enhance transgene expression are highly desirable. In this initial study, we demonstrated that chimeric vectors can be used to improve AAV performances in both liver and muscle.

Our new vectors are classified as chimeric vectors, which are distinct from hybrid vectors such as AAV6. AAV6 is composed of both AAV1 and AAV2 sequence. The overall virions are still homogeneous in terms of composition. For chimeric vectors of AAV1 and AAV2, each virion has six kinds of capsid proteins, which include AAV1 VP1, VP2, and VP3 and AAV2 VP1, VP2, and VP3. Since all cells are generally transfected with a high copy number of AAV helper and it is estimated that more than a million copies of capsid proteins can be expressed in a cell, statistically, the packaged virions would be in a nearly uniform composition. The ability of the chimeric vectors to bind to heparin and to react with both AAV1 and AAV2 neutralizing antibodies confirms that the vectors produced are not a mixture of AAV1 and AAV2 vectors.

One major advantage of this chimeric vector is that it can be purified efficiently using a heparin column. At this stage, AAV1 cannot be purified by heparin column. Since AAV2 vectors bind to a heparin column and AAV1 does not have this property, the incorporation of AAV2 capsid proteins into AAV1 vector allows the chimeric vectors to bind to heparin. As heparin column-purified vectors are considerably cleaner that those purified by CsCl gradient, the chimeric AAV vectors may have a better safety profile after this modification. As shown in Fig. 3, approximately 50% of the capsid proteins in virions must be of AAV2 origin to allow the chimeric vectors to gain such properties. So for better binding activity, it is desirable to have more AAV2 capsid protein in the virion. On the other hand, to gain the properties of AAV1, more AAV1 capsid proteins would be beneficial. Therefore, there will be an optimized ratio for performance. As shown in Table 2, in our testing so far, the optimal ratio for AAV12 chimeric vectors is 1:1. At this ratio, the chimeric vectors can be purified by heparin column efficiently and have improved performance in both liver and muscle compared to the less effective serotype.

One concern is that the ratios of AAV1 and AAV2 cap proteins will vary from one experiment to another. In practical use, a packaging cell line expressing a fixed ratio of AAV1 and AAV2 cap proteins could be established to avoid this problem. Another potential disadvantage for chimeric vectors is that their transduction may be blocked by antibodies to either AAV1 or AAV2 as chimeric vectors possess properties of both AAV1 and AAV2. However, the successful use of AAV2 vector in patients with NAB against AAV2 suggests that administration of vectors in the presence of NAB is a rather complicated issue [15]. Further investigation will certainly be necessary to address this question.

Since each AAV serotype appears to have different tissue tropisms and all AAV serotypes share a high degree of homology, it seems likely that additional chimeric vectors could be made using the same strategy. One potential use is to apply this strategy to improve on AAV vectors made by epitope insertions in capsid region [31-34]. These vectors often have a low yield because of the additional sequence insertion. The chimeric strategy could be a way to improve rAAV yield by decreasing unnatural capsid components. However, an effective ratio will have to be tested individually based on the vectors to avoid possible interference. It is also anticipated that these new chimeric vectors would have the combined properties of both or all their parents and different tissues could be targeted with the same vector.

Materials and Methods

Vector production and titering

Human 293 cells were maintained in DMEM supplemented with 10% fetal bovine serum and 1× penicillin and streptomycin and incubated in humidified environment with 5% CO₂ at 37°C. Recombinant AAV vectors were produced as described previously [25]. The vector plasmids pAAV-CMV-F.IX, pAAV1-EF-F.IX, and pAAV-MFG-a1at were used in a previous study [26,27]. Briefly, helper plasmid, vector plasmid, and mini-adenovirus helper plasmid were transfected into 293 cells at a ratio of 1:1:2 by calcium phosphate precipitation. For each 15-cm tissue culture dish, a total of 50 μg DNA was used. The transfected cells were then maintained in DMEM containing 2% fetal bovine serum. At 4 days posttransfection, the cells were harvested and vectors were purified by CsCl gradient or heparin column.

Vectors were titered by quantitative PCR using ABI Biosystems 7700 according to procedures described before [25]. The reactions were performed following the instructions of the manufacturer using PCR core reagents kit (PE Biosystems). A DNA plasmid standard curve was set up using 10³-10⁷ AAV genome equivalents. The virus DNA samples were prepared by digesting DNase-resistant virus with proteinase K in 1× PCR buffer at 37°C overnight and boiled for 20 min after digestion. The titer was determined by comparing the amplification curve to the standard curve and also confirmed by slot blot.

Vector purification by CsCl gradient

The harvested cell pellet after transfection and incubation was resuspended in 10 mM Tris, pH 8.0, and sonicated for 2 min to break down cell membranes. After incubating with DNase and RNase, 1/20 volume of 10% sodium deoxycholate solution was added to the mix. The obtained lysates were then loaded onto a CsCl step gradient and subjected to two rounds of ultracentrifugation. The fractions with refractory index in the range of 1.370 -1.377 in the first round were collected and used for the second round of centrifugation. The resulting peak fractions were then collected, pooled together, and dialyzed against PBS containing 10% glycerol. The obtained vector stocks were stored at -80°C before use.

Vector purification by heparin column

The harvested cell pellet was resuspended in 10 mM Tris, pH 8.0, + 150 mM sodium chloride and sonicated for 2 min to release the virus. The resulting mixture was then centrifuged at 3000g and the debris was discarded. The supernatant was incubated with Benzonase (Sigma) in the presence of 0.5% sodium deoxycholate for 1 h at 37°C. Heparin-Sepharose resin (1.5 g; Amersham Pharmacia) was pretreated according to the manufacturer’s protocol, added to the mixture, and incubated overnight at room temperature with constant agitation. The resin was then loaded onto a Bio-Rad minicolumn. The vectors were eluted using gradient salt solution from 200 to 1000 mM NaCl in 10 mM Tris, pH 8.0, in a volume of 10 ml/fraction.

Animal handling and vector administration

C57BL6 (Charles River Breeding Laboratories) mice were selected for intravenous injection with rAAV. CD4 KO mice were purchased from The Jackson Laboratory for intramuscular injection. The animals were age 4 - 6 weeks when vectors were administered. Standard procedures for animal handling were followed. AAV-F.IX was injected into the tibialis anterior (25 μl) and the quadriceps muscle (50 μl) of the left leg using a Hamilton syringe to target muscle. To target liver, vectors were injected through the tail vein. Blood samples were collected every 2 weeks from the retro-orbital plexus in microhematocrit capillary tubes and plasma was assayed for human FIX by ELISA (vide infra). A typical group usually consists of three mice. Each experiment was repeated to confirm the results.

In vitro neutralization assay

For neutralization assays, COS cells were infected with vectors at an m.o.i. of 5000. To test whether the vector can be neutralized by AAV1 and AAV2 antiserum, the vectors were incubated with AAV antiserum at 1:100 dilution for 30 min at 37°C. The vectors along with appropriate controls were then used to infect COS cells. At 48 h postinfection, the medium was collected and the concentration of human factor IX secreted into the medium determined by ELISA.

Factor IX and human α1-antitrypsin ELISA

Human FIX antigen in mouse plasma was determined by ELISA as described by Herzog et al. [28]. This ELISA did not cross-react with mouse FIX. All samples were measured in duplicate. The dilution of mouse plasma was usually in the range of 5- to 50-fold depending on the level of human factor IX in serum or medium.

The concentration of human α1-antitrypsin in the mouse serum was measured by ELISA. Briefly, 10 μg/ml rabbit anti-human α1-antitrypsin (Sigma) was coated to ELISA plates in 0.1 M/sodium bicarbonate, pH 9.6, for 2 h. After being blocked with 3% BSA/PBS, different dilutions (50 to 200 times dilution) of mouse serum samples were applied. The captured human α1-antitrypsin was incubated with goat anti-human α-1-antitrypsin (2 μg/ml; Sigma) followed by mouse anti-goat IgG peroxidase conjugate (Sigma). Signal detection was achieved by addition of ABTS (Boeh-ringer Mannheim). The linear range of this assay is between 0.3 and 30 ng/ml. The human α1-antitrypsin standard was purchased from The Scripps Institute (La Jolla, CA).