AAVone: A cost-effective, single-plasmid solution for efficient AAV production with reduced DNA impurities

Abstract

Currently, the most common approach for manufacturing good manufacturing practice (GMP)-grade adeno-associated virus (AAV) vectors involves transiently transfecting mammalian cells with three plasmids that carry the essential components for production. Here, we developed an all-in-one, single-plasmid AAV production system, called AAVone, in which the adenovirus helper genes (E2A, E4orf6, and VA RNA), AAV packaging genes (rep and cap), and the vector transgene cassette are consolidated into a single compact plasmid with a 13-kb backbone. The AAVone system achieves a 2- to 4-fold increase in yields, exhibits low batch-to-batch variation, eliminates the need for fine-tuning the ratios of the three plasmids, and simplifies the production process, compared with the traditional triple-plasmid system. AAVs generated by the AAVone system show similar in vitro and in vivo transduction efficiency, but a substantial reduction in DNA impurities from plasmid bacterial backbones and a marked reduction in non-functional snap-back genomes. The AAVone system does not pose a risk for generating replication-competent AAV contaminants. Furthermore, the AAVone system requires significantly less DNAs for AAV production, while achieving favorable full:empty particle ratios and further reducing impurities. In summary, the AAVone platform is a highly efficient, straightforward, cost-effective, and highly consistent AAV production system, making it particularly suitable for manufacturing of GMP-grade AAV vectors.

Graphical abstract

The AAVone system integrates all necessary AAV production components into a single plasmid, enabling the generation of AAV vectors by transfecting packaging cells with minimal amounts of this plasmid. AAVone is simple, economical, and highly dependable system for AAV manufacturing, making it especially suitable for generating GMP-grade AAV vectors.

Introduction

Adeno-associated virus (AAV) vectors have been successfully used in numerous clinical trials and have received regulatory approvals as drugs for human use.^1,2,3 Although alternative production systems exist,^{4,5,6,7,8,9,10,11} the primary technology for AAV vector production remains co-transfection of three plasmids into HEK-293 cells (tri-plasmid system).¹² The AAV vector plasmid, also called the cis plasmid, carries two inverted terminal repeats (ITRs) and the gene of interest (GOI) expression cassette. The AAV helper plasmid, also called the trans plasmid, provides the AAV genes (rep and cap) necessary for production.^13,14,15 To facilitate AAV vector production, a third plasmid is used to provide the essential adenoviral (AdV) genes, including E2A, E4orf6, and VA RNA.¹⁶ Recent studies have also shown that the AdV 22K protein is essential for AAV vector production and the 33K protein synergistically increases vector yield.¹⁷

AAV vectors produced using the tri-plasmid method have demonstrated safety, convenience, and effectiveness. However, this method has limited productivity and scalability^18,19 and requires co-transfection of all three plasmids into a single cell.²⁰ Moreover, the requirement for high quantities (∼1 μg/1e6 cells) of plasmid DNA (pDNA) and the need for ratio optimization among the three plasmids not only increases time and labor costs, but also introduces variability among production batches.^21,22,23 The tri-plasmid systems yield relatively high amounts of plasmid DNA impurities,^{10,18,24,25,26} and the extent of those impurities in AAV products varies significantly depending on the vector design and production process.^{24,27,28,29,30} This concern has become more important, since recent trials have revealed significant safety concerns at doses exceeding 1E14 vg/kg.^31,32

Many approaches have been proposed to improve current AAV production systems. One approach is to develop more efficient upstream processes, including new cell lines, culture media, transfection reagents, and enhancers/boosters.^18,33,34,35 Another approach is to optimize the design and function of the AdV helper, trans, and cis plasmids used in tri-plasmid systems.^12,36,37 A third approach is to streamline the transient-transfection step by reducing the number of plasmids. In a dual-plasmid system, one plasmid (pDG) carries the necessary AdV genes, as well as rep and cap, while the other contains the GOI cassette.^38,39,40 However, the large plasmid size of pDG (∼22 kb) poses challenges during manufacturing, particularly in terms of scalability, purification, and stability. Additionally, large plasmids exhibit lower transfection rates and higher toxicities compared with smaller plasmids, limiting their use in AAV vector production.^41,42,43 Recently, another dual-plasmid system (pOXB) has been described that combines the cis and trans elements together on a single plasmid, and provides AdV genes on a separate plasmid.⁴⁴ A third possible dual-plasmid configuration (pLV) is to combine the cis elements with the AdV helper, while expressing rep and cap on a separate plasmid. Interestingly, the pOXB design outperforms the pDG and pLV configuration in terms of productivity.⁴⁴

Here, we have developed an all-in-one single-plasmid AAV production system called AAVone. In the AAVone system, the AdV helper genes (E2A, E4orf6, and VA RNA), AAV helper genes (rep and cap), and the AAV vector genome are assembled into a compact pAAVone plasmid with a 13-kb backbone. The AAVone system demonstrates 2- to 4-fold higher vector yields compared with those achieved by a traditional tri-plasmid approach, while significantly reducing batch-to-batch variation and eliminating the need for plasmid ratio optimization. Vectors produced by AAVone exhibit in vitro and in vivo transduction efficiencies that are comparable with the efficiencies of vectors produced by conventional methods, but with substantially fewer contaminating plasmid-derived backbone sequences and a marked reduction in non-functional snap-back genomes. Importantly, AAVone does not pose a risk for generating replication-competent AAV (rcAAV) contaminants. Additionally, AAVone requires only 0.25–0.50 μg of pDNA per million cells, further reducing impurities, while maintaining high full-to-empty capsid ratios. In summary, AAVone represents a highly efficient, robust, and cost-effective platform for good manufacturing process (GMP)-grade AAV production, offering simplified scalability, improved consistency, and superior vector purity—making it an ideal solution for both research and clinical applications.

Results

Creation of the AAVone single-plasmid system for AAV vector production

To create an efficient single-plasmid system for AAV vector production, it is critical to control the total plasmid size, as it can modulate gene transfer efficiency.^41,42,43 Among the three plasmids for AAV vector production, the AdV helper plasmid is the largest. Researchers have continually made efforts to reduce the size of the AdV helper, such as pBHG10, pXX5, pXX6, and pAdDeltaF6.^12,45 The AdV helper plasmid used in this study, pHelper, is 11.6 kb in size and contains 9.3 kb of the AdV2 genome. We minimized the pHelper by removing the introns present in the E2A and E4orf6 genes (Figure S1). This modification led to a mini-pHelper-1.0 that retains only 6.1 kb of the AdV2 genome and has a total plasmid size of 8.4 kb (Figure 1A). The 22K and 33K proteins were recently reported to be involved in AAV production and are preserved in the mini-pHelper-1.0 plasmid.¹⁷ When co-transfected with the cis (pAAV-GOI) and trans (pRCap) plasmids, the mini-pHelper-1.0 (mTri-plasmid system) has higher productivity than the parental pHelper (pTri-plasmid system) (Figure 1B). We then created pDG-like and pLV-like (AAVdual) dual-plasmid systems based on the mini-pHelper-1.0. The AAVdual system matched or exceeded the productivity of pTri-plasmid system, whereas the pDG-like system displayed comparable or reduced packaging efficiencies (Figure 1B).

Figure 1.

Design and testing of the AAVone production system

(A) Schematic illustrations of the AAVone, triple-plasmid (pHelper, pTri-plasmid; and mini-pHelper-1.0, mTri-plasmid) systems, dual-plasmid (pDG-like and AAVdual) systems, and the AAVone system. (B) A comparative analysis of AAV productivity using AAVone and alternative systems in adherent HEK-293T cells. For every system compared, the total pDNA amount was maintained at 0.5 μg/well (24-well plate), and the PEImax to pDNA ratio was consistently kept at 3:1. The plasmid molecular ratios for each system is displayed under each dataset. (C) A comparison of vector yields with different serotypes between the AAVone and the pTri-plasmid system in suspension HEK-293T cells. The pTri-plasmid set was transfected at equal plasmid molecular ratios at a total DNA amount of 1 μg/1E6 cells, at a cell density of 3E6 cells/mL. Crude lysate samples were quantified using qPCR with primers targeting Egfp, and data were normalized to the average values of each pTri-plasmid group. (D) Productivity optimization of the AAVone system in suspension cultured HEK-293T cells as assessed by crude lysate titration. Different cell densities, total pDNA, and transfection reagent (PEImax) were examined.

Encouraged by the positive outcomes from the dual-plasmid systems, we consolidated all of the cis and trans essential elements for AAV production, and the AdV genes into a single-plasmid with a 13-kb compact backbone (pAAVone) (Figure 1A). In the AAVone system, the GOI cassette is situated between the bacterial backbone and VA RNA, while the rep-cap expression cassette is positioned between VA RNA and E4orf6. In this configuration, the VA RNA is close to the right (R)-ITR and bacterial elements are located before the left (L)-ITR. The compact pAAVone plasmids are easy to clone and amplify without reducing yields (Figure S2). Additionally, creating pAAVone plasmids with different vector designs, serotypes, and transgenes was straightforward. With the AAVone system, AAV vectors can be produced by transfecting a single-plasmid into producer cells. For example, AAV2-CMV-Egfp vectors were generated by transfecting pAAVone-AAV2-CMV-Egfp plasmid into adherent HEK-293T cells. As shown in Figure 1B, the output from the tri- and dual-plasmid systems varied greatly with changes in the proportions of the three plasmids, while the AAVone system produces reproducible and higher yields in multiple tests.

When the plasmids were tested by transient transfection (Figure S3), the AAVone system demonstrated comparable levels of EGFP expression and an equivalent percentage of positively transfected cells as those achieved by the pTri-plasmid system. It is important to highlight that, in these experiments, the 15.8-kb pAAVone-AAV2-CMV-Egfp construct that is used in the AAVone system and the 5.4-kb pAAV-CMV-Egfp construct that is used in pTri-plasmid system carried the same AAV-CMV-Egfp genome design. In addition, the AAVone and the pTri-plasmid system (with a molecular ratio of pHelper:pRCap:pAAV-CMV-Egfp = 1:1:1) used the same amount of total pDNA by mass. Under this experimental parameter, the molecular ratio of pAAVone-AAV2-CMV-Egfp to pAAV-CMV-Egfp plasmid is 1:0.64. When pAAVone-AAV2-CMV-Egfp and pAAV-CMV-Egfp were transfected with the same mass (molecular ratio of 1:2.93), EGFP expression from the pAAV-CMV-Egfp plasmid was much stronger than what was produced by transfection with pAAVone (Figure S3). These data suggest that the enhanced AAV productivity of the AAVone system is not due to increased transfection efficiency, but rather the co-localization of all packaging components within the same cells.²⁰

The AAVone system achieves high AAV vector yields in suspension-cultured cells

To assess the compatibility of the AAVone platform for large-scale manufacturing, we initially evaluated its packaging yield in suspension-cultured HEK-293T cells. AAVone consistently resulted in a 2- to 4-fold increase in packaging yield compared with the pTri-plasmid system across various serotypes (Figure 1C; Table S1). We next tested various parameters, including different cell densities (2.5E6 to 3.5E6 cells/mL), PEI/DNA ratios (2:1 to 3:1), and pDNA amounts (0.5–1.0 μg per 1E6 cells). Within those conditons, the AAVone system demonstrated optimal performance when 0.5 μg of pDNA was transfected for every 1E6 cells, with a PEI/DNA ratio of 2:1, and a cell density of 3E6 cells/mL (Figure 1D). The most important parameter for vector production was the pDNA amount. The highest yields were obtained for transfections using 0.5 μg pDNA per 1E6 cells condition. For high-yield AAV serotypes such as AAV1, AAV8, AAV9, and AAVrh.10, the AAVone system easily achieved Egfp specific crude lysate titers exceeding 1E15 vg/L in different production scales without any process optimization (Table S1). Among these serotypes, AAV9 demonstrated the highest yields, with titers obtained from crude cell lysates collected after AAV production ranging from 1.96E15 to 3.7E15 vg/L, and purified titers after two rounds of CsCl ultracentrifuge ranging from 4.22E14 to 1.03E15 vg/L. AAV9 also exhibited the highest fold change with AAVone, achieving up to a 4-fold increase in yield compared with the pTri-plasmid system. Even with low-yield AAV serotypes, like AAV2, the AAVone system reliably produced between 4.9E14 and 8.9E14 vg/L of crude titers across various production scales, showcasing its consistent performance. Remarkably, the AAVone system demonstrated the ability to generate AAV vectors with transgene genome sizes ranging from 2.2 kb to 4.7 kb, while the sizes of the corresponding pAAVone plasmids varied between 15.2 kb and 17.7 kb (Table S1). Despite a slight decrease in AAV vector yields with increasing payload sizes, the AAVone system achieved a crude titer of 1.94E15 vg/L for a 4.7-kb vector packaged with AAV9.

We further evaluated the productivity of the AAVone system in suspension-cultured HEK-293 cells, which are commonly used in GMP-level AAV production. In general, the yields achieved by the AAVone system in the suspension HEK-293 cells were about half of those obtained with HEK-293T cells (Table S1). The AAVone system showed adaptability with various transfection reagents, yielding high outputs with both PEImax and FectoVIR. Specifically, for AAV9 packaged with a 2.2-kb cassette, AAVone reached up to 1.53E15 vg/L of crude lysate titers with PEImax. With AAV2, the system produced over 4E14 vg/L with PEImax, and more than 6E14 vg/L with FectoVIR. These results collectively indicate that the AAVone system is highly effective across various transgenes, AAV serotypes, and cell types. It is also compatible with different transfection reagents. Moreover, the AAVone system can be easily scaled up for production using small (Ambr 250) and large (XDR200) bioreactors (R.Y. data not shown).

AAV vectors generated by AAVone are similar to those produced by tri-plasmid systems

Assessing the quality of AAV vectors is of paramount importance when appraising a novel production system. To directly compare the performance of the AAVone system with tri-plasmid systems, AAV2-CMV-Egfp vectors were produced across all platforms using suspension-cultured HEK-293T cells. A total of 0.75 μg/1E6 cells of pDNA was used for this examination. AAV vectors were isolated at specific refractive index values, ranging from 1.369 to 1.371, following two rounds of CsCl gradient ultracentrifugation. The composition of the capsid proteins, VP1, VP2, and VP3, showed an approximate 1:1:10 ratio across the three systems, with no observable differences between the vectors (Figure 2A). Purified vector DNA also showed that the predominant species were the expected 2.8-kb full-length genomes (Figure 2B). All three preparations indicated the presence of truncated or partial genomes with no visually discernable differences in the sizes and abundances of the partial genomes. It is notable that the Tapestation software identified additional low-molecular-weight species under 500 bp in size for the pTri- and mTri-plasmid systems, but these species were not identified in the AAVone system (Figure 2B). Mass photometry analyses also did not reveal major differences in particle masses between vectors produced by the three platforms (Figure 2C). We next evaluated whether there were differences in the transduction profiles between vectors produced by the three systems. Transduction of HeLa cells indicated no clear differences in EGFP expression at 48-h post-transduction (Figure 2D).

Figure 2.

Characterization of AAV vectors produced by the AAVone and tri-plasmid systems

Purified AAV vectors produced in suspension HEK-293T under different production platforms. (A) Purified AAV vectors (∼5E11 particles) from pTri-plasmid, mTri-plasmid, and AAVone were electrophoresed and visualized by Coomassie G-250 staining. (B) Automated electrophoresis analysis (TapeStation) of vector DNAs illustrating full-length genomes (arrow) and partial genomes (bracket). Arrow heads designate specific bands identified by software. (C) Mass photometry analysis of AAV particles. y axis, particle counts; x axis, mass of particles. (D) Transduction activity of vectors packaging the AAV2-CMV-Egfp reporter produced by AAVone or triple-plasmid systems in HeLa cells at a multiplicity of infection (MOI) of 10,000 vg/cell. and EGFP expression was observed 48 h post-transduction (scape bar, 100 μm).

Vectors packaged by the AAVone system and tri-plasmid systems exhibited similar retinal transduction profiles

Retinal tissues are among the most prominent targets for AAV vector-based gene therapies being tested in preclinical and clinical trials, especially for platforms that are based on AAV2. Vectors packaged with AAV2-CMV-Egfp using the AAVone, mTri-, and pTri-plasmid systems were administered to mouse retinas by intravitreal (IVT) injection at 1E9 vg/eye. At 4 weeks post-injection, we conducted fundoscopy imaging to assess EGFP expression and distribution across the retina. We found that administration of the three vectors led to similar levels of EGFP expression and distribution (Figure 3A). At 6 weeks post-injection, retinas were collected, cryo-sectioned, and stained with anti-EGFP antibody (green) and DAPI (DNA, blue). We observed comparable levels of EGFP signal overlap across all three vector groups (Figure 3B) and similar percentages of EGFP-positive photoreceptor outer segments per section (56%–59%) (Figure 3C). This finding suggested that vectors produced by the AAVone system performed similarly to those produced by the tri-plasmid systems. To quantify the transduction efficiencies of the vectors produced by the three systems, we extracted bulk DNA and RNA from harvested retinas to measure the presence of vector genomes and transgene transcripts. Using digital droplet PCR (ddPCR) we found that the vector packaging systems did not influence transduction (Figures 3D and 3E).

Figure 3.

AAVone-packaged AAV2 vectors exhibit similar in vivo efficacies as vectors produce d by triple-plasmid systems

(A) Representative fundoscopy images of murine retinas at 4 weeks post-injection. (B) Representative retina cross-sections stained to visualize EGFP (anti-EGFP, green), nuclei (DAPI, blue, and microglia (IBA1, red) at 6 weeks post-injection. Cross-sections of representative whole eye cups (scale bars, 0.5 mM). (C) Quantification of EGFP⁺ cells in the photoreceptor (PR) layer. (D and E) ddPCR quantification of vector genomes (D) and vector transcripts (E) in retina samples at 6 weeks post-injection (n = 4). (F) Counts of total microglia in retina cross-sections at 6 weeks post-injection. (G) Representative cross-section of a vector-treated retina illustrating microglial (IBA1⁺) infiltration in retinal layers: photoreceptor segment layer (PS); outer nuclear layer (ONL); outer plexiform layer (OPL); inner nuclear layer (INL); inner plexiform layer (IPL); and ganglion cell layer (GCL). IBA1⁺ cells (red) and DAPI (blue). (H) Percentage of IBA1⁺ cells in each retinal layer. Values represent means ± SD, ∗∗p < 0.01; ∗∗∗∗p < 0.0001, ns = not significant; by one-way ANOVA and Tukey’s multiple comparison.

We next aimed to assess whether vectors produced by the three AAV packaging systems elicited differences in immune response. It has been well documented that following IVT injection of AAV vectors, microglia are activated and migrate to regions of pathogen-associated molecular pattern activation.^46,47,48,49 We therefore stained cryosections of treated retinas to mark microglia (anti-IBA1). Cells positive for IBA1 were found to increase in the retinas treated by AAV vectors (Figure 3F). Notably, we found similar counts of IBA1-positive cells per section across all three treatment groups. We also examined the distribution of IBA1-positive cells infiltrating into the deeper layers of the transduced retinas (Figures 3G and 3H). All three vectors exhibited equivalent stimulations of IBA1-positive cell infiltration across the outer plexiform layer, the inner nuclear layer, the inner plexiform layer, and the ganglion cell layer.

Vectors produced by AAVone and tri-plasmid systems exhibited similar muscle and brain transduction profiles

We next aimed to determine whether the AAVone system was also compatible with other serotypes. We packaged the same ssAAV-CMV-Egfp transgene as above with AAV1, AAV8, AAVrh.10, and AAV-PHP.eB capsids by the AAVone, pTri-, and mTri-plasmid systems. Since AAV1 and AAV8 are known to be tropic to muscle, we injected the AAV1- and AAV8-packaged vectors into the tibialis anterior (TA) muscles of adult female mice via intramuscular (IM) injection. Animals injected with AAV1 vectors were sacrificed after 2 weeks. Animals injected with AAV8 vectors were sacrificed at week 8. TA muscles were harvested and cryosectioned to assess differences in transduction efficiency between the three vectors systems. Fluorescence microscopy of TA cross-sections revealed similar intensities and distributions of EGFP throughout the TA across the three treatment groups for AAV1 and AAV8 (Figure S4A) vectors. Additionally, we assessed the immunogenic profiles of vectors produced by the AAVone system compared with vectors produced by pTri- or mTri-plasmid systems. Two weeks post-injection, there were no observable differences in cellular infiltrates of the TA muscle across the treatment groups (Figure S4B).

Both AAV-PHP.eB and AAVrh.10 are well known for their neurotropic properties. We, therefore, injected adult male C57BL/6J mice at a dose of 1E14 vg/kg by intravenous administration. Two weeks post-injection, animals were euthanized and brains were harvested. Immunohistochemical staining of brains cryosections showed that AAV-PHP.eB and AAVrh.10 exhibited different patterns of EGFP-positive cells, as expected (Figure S5); AAV-PHP.eB primarily transduced neurons, while AAVrh.10 transduced both neurons and glial cells. Importantly, there were no significant differences in EGFP transduction between vectors produced by the three packaging systems among the AAVrh.10 and AAV-PHP.eB groups.

The AAVone system produces vectors with slightly reduced genome heterogeneity

We next assessed the heterogeneity of the vector genomes using single molecule, real-time (SMRT) sequencing and AAV genome population (AAV-GPseq) analyses.^11,29,50 Using this approach, we were able to analyze individual AAV vector genomes at a single molecule level. The majority of reads obtained by SMRT sequencing spanned the entirety of the ssAAV-CMV-Egfp reference genome from ITR to ITR (Figure S6). Truncated genomes with snap-back configurations^51,52 were also observed as reads that only partially spanned the reference. These species seem to be similar across the three production platforms, suggesting that there were no substantial differences in truncation events between the vectors analyzed. Instead, the major truncations seem to be centered at the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) (Figure S5). We next aimed to assess whether the AAVone system produced vectors with different amounts of non-unit length genomes. The lengths of the reads mapping to the references were tabulated and graphed (Figure 4A). The vector genomes from the two tri-plasmid systems had nearly identical read-length distributions, with a major peak at 2,770 nt and multiple smaller peaks at 347 nt, 591 nt, and 1,878 nt. The 2,770-nt population represents the intact AAV genome size, plus two intact 145-nt ITRs, suggesting at ITR repair.⁵³ Interestingly, the vector produced by the AAVone system displayed a predominant read-length peak at 2,770 nt, with some minor peaks at other lengths. The AAVone system exhibited a slightly higher percentage of intact genomes (58.77%), compared with the pTri-plasmid (51.34%) and the mTri-plasmid (51.41%) systems (Figure 4A). Analyses of read alignment start and end positions also indicated a clear reduction in truncation hotspots with the AAVone system when compared with the two tri-plasmid systems (Figure 4B). As noted before, the major truncation hotspot seems to be centered at the WPRE. Additional truncations are revealed to be centered at the hGH polyA sequences (Figure 4B).

Figure 4.

Characterization of AAV vectors produced by the AAVone and tri-plasmid systems

(A) Distribution of SMRT sequencing read lengths from vectors produced by pTri-plasmid system (left), mTri-plasmid (center), and AAVone (right). Read lengths are binned into 10-nt increments. The percentage of full-length genomes (2,770 ± 10 nt) are shown for each graph. The y axes shown are separated into two segments to highlight the low prevalence of truncated genomes. (B) Relative counts of mapped read start (blue) and end (red) positions, indicating truncation hotspots. The y axes shown are separated into two segments to highlight the low prevalence of start and end positions beyond the ITR ends. (C) Sequences of the 119-nt L-ITR and the 130-nt R-ITR used in the plasmid vector designs. (D) Length distribution of the L-ITRs and R-ITRs from the three production systems examined. Read counts are on the x axes, and the read lengths are on the y axes. (E) A summary table of ITR integrity (especially the number of 11-nt C arm deleted ITRs at the L-ITRs and R-ITRs identified among the SMRT reads) and major impurities (snap-back gnomes, total plasmid related impurities, and hg38-related impurities).

Intact ITRs in AAV vectors are fundamental for transgene expression and for the overall success of the gene therapy. wtAAV2 ITRs are 145-nt in length and can take on two configurations, flip and flop.⁵⁴ Packaged vectors that undergo proper rolling-hairpin replication show equal distribution of flip and flop at both ITRs. As mentioned before, the AAV GOI cassette used by the different production systems are the same. Within the plasmid construct, the R-ITR is 130 nt long, which is 15 nt shorter than the full-length 145-nt ITR. The L-ITR is 119 nt in length and differs from the L-ITR by an 11-nt deletion in the C arm (Figure 4C). Both ITRs were designed in the flop orientation. We found that the ITRs in all AAV vectors were 145 bp in length at both the L-ITRs and R-ITRs (Figure 4D). The majority of reads obtained from vectors produced by the three systems were fully aligned to the reference ITRs. Flip and flop are the two major configurations observed in both L- and R-ITRs in AAVone and tri-plasmid systems. They occupy over 90% of the population and display similar flip:flop ratios (Figure S7). Other mutated and truncated forms were also observed at both L- and R-ITRs, but they were in the minority. The presence of ITRs missing 11 nt in the C arm constituted only approximately 1.5% of the total genomes, implying that the deletion was largely repaired during replication (Figure 4E).⁵⁵ Taken together, AAV vectors produced by the AAVone system have nearly identical genomic compositions as compared with those produced by the tri-plasmid systems.

The AAVone system exhibits alternative plasmid-related impurities in the AAV vectors

Residual input pDNAs are one major concern for AAV production. In the tri-plasmid systems, contaminants mainly originate from the backbone of the cis plasmid, while trans plasmid sequences are packaged at lower levels.²⁹ One major difference between the AAVone and the tri-plasmid systems is the oversized 13-kb backbone, which not only includes the bacterial ori and the antibiotic selection gene (kanR) but also contains rep, cap, E2A, E4orf6, and VA RNA. To examine the impurities originating from plasmids found in vectors produced by AAVone and tri-plasmid systems, we aligned SMRT reads to the AAV-GOI plasmid references. We found that the two ITRs were where the majority of reads mapped (Figures 5A–5C). We also showed that the reads mapping to the plasmid backbone tended aggregate proximal to the ITRs, with decreasing coverage when sequences are further away from the ITRs. Within the AAVone system, VA RNA constituted 2.35% of the total sequences (Table S2), making it the most prevalent contaminant. The next most abundant type of contaminant was the plasmid bacterial backbone (0.87%). Other elements, such as E2, E4, rep, and cap, which are more distant from the ITRs, exhibited a considerably lower percentage of mapped reads (Table S2). The dominant contaminant in vectors produced by the tri-plasmid systems originated from the cis plasmid backbone (3.95% in the pTri-plasmid and 2.69% in the mTri-plasmid system). These observations were validated through qPCR analysis focusing on various regions of the pAAVone plasmid (Figure 5D). Thus, the dominant contaminants in AAVone-produced vectors were different from the contaminants related to vectors produced by the tri-plasmid systems.

Figure 5.

Abundance of producer plasmid contaminants in AAV vectors generated by tri-plasmid systems and AAVone

(A and B) Coverage traces of aligned SMRT reads to the pAAVone (A), pAAV-CMV-EGFP, (B) and AAVone (C) plasmid references. Dashed lines indicate the boundaries of the ITR references. (D) ddPCR quantification of trans plasmid sequences (rep and cap), AdV helper plasmid (VA RNA), backbone sequences (pMB1 ori and F1 ori), and rcAAV in vector preparations. Values are means ± SD (n = 3). ∗p ≤ 0.05; ∗∗∗∗p ≤ 0.0001 by two-way ANOVA. Color of asterisks correspond to p values when compared with pTri-plasmid (red), mTri-plasmid (blue), and AAVone (purple) data.

Host cell DNA (hcDNA) contaminants originating from the packaging cell line used in vector production represent another form of genomic contaminant that may impact the safety of gene therapy vectors. Overall, there was no notable difference in hcDNAs among the AAV vectors produced by the three production systems, with each showing approximately 3% of the total reads (Table S2; Figure S8). In all of the reads that mapped to the hg38 genome, the average (∼1,300 nt) and mean (∼1,140 nt) read lengths were very similar among the three systems (Figure S9). Notably, these read lengths are shorter than the smallest oncogene size (1,664 bp).⁵⁶ The number of unique reads spread across each chromosome, suggested at an overall random distribution among all three AAV samples (Figure S8).^28,29 Interestingly, reads mapping to mitochondrial DNA (chrM) seemed to be enriched at the D loop²⁸ (Figure S10). However, chrM contaminants that were chimeric with the GOI were also rare (data not shown). Among all of the hcDNA-mapped reads, only 15% seemed to be chimeric genomes (i.e., vector genomic sequences that are recombined with hcDNAs), irrespective of the production system (Table S2; Figure S8). However, when these chimeric reads were mapped back to the GOI reference and visualized, the majority of reads mapped imperfectly with the hGH polyA (Figure S11). These results indicated that the chimeric reads were predominantly false positives.

In summary, our SMRT sequencing results demonstrate that the AAVone system exhibits significantly lower levels of bacterial backbone sequences, comparable or slightly higher rep, cap, E2, E4, and hcDNA contaminations, and an increase in VA RNA sequences compared with the tri-plasmid systems. Despite this shift in the nature of contaminants, the overall total number of genomic contaminations for the AAVone system were decreased (Table S2).

The AAVone system does not pose a risk for generating rcAAV contaminants

Another concern for the AAVone system is the potential for generating rcAAVs. In tri-plasmid systems, rcAAVs arise through the recombination of DNAs carrying ITRs with rep and cap sequences originating from the rep-cap plasmid.⁵⁷ In the AAVone system, rep, cap, and the ITRs co-exist within the same plasmid. A cell-based rcAAV assay performed by an external contract research organization indicated that rcAAV levels were less than 1 per 1E9 vector genomes AAV vectors generated by the AAVone system. After three rounds of amplification with AdV, all three flasks in the 1E9 vector genome group showed no presence of rcAAV, and only one out of three samples tested positive for rcAAV in 1E10 vector genomes. This shows that the AAVone system can effectively produce safe AAV vectors with low rcAAV. Using qPCR primers targeting both ITR and rep to detect ITR-rep hybrids,⁵⁸ which probes for the presence of rcAAV contamination, comparable levels were observed among the AAVone and tri-plasmid systems (Figure 5D). Analyses of the SMRT reads representing genomes obtained from vectors produced by the AAVone system showed that the majority of reads mapping to the rep or cap genes also co-mapped with the GOI sequence (89 out of 99 [89.9%] and 201 out of 230 [87.39%], respectively). It is noteworthy that, in the tri-plasmid systems, reads that map to either rep or cap also displayed a high degree of mapping to ITRs, despite the fact that the trans plasmid lacks ITRs (Table S2). For a read to represent an rcAAV genome, it has to contain the ITRs, rep, and cap in one sequence.⁵⁷ Our analysis revealed that among the vector genomes from the AAVone system, such chimeric reads were present in 90 out of 196,753 reads (0.045%) (Figure 6), which was slightly higher than those of the pTri-plasmid system (27 out of 131,989 [0.021%]) and mTri-plasmid system (22 out of 119,029 [0.018%]). Nevertheless, among all of these tri-chimeric sequences, none contained an intact ITR-rep-cap-ITR configuration (Figure 6).

Figure 6.

Examination of SMRT reads for the presence of rcAAV

IGV display of SMRT reads representing genomes obtained from vectors produced by pTri-plasmid (top), mTri-plasmid (middle), and AAVone (bottom) systems. Alignments are shown with soft-clipped bases visible. Read matches (gray), mismatches (colored), and insertions/deletions (speckles) are shown.

The AAVone system enhances production and full:empty particle ratios and reduces impurities by using less pDNA

In the tri-plasmid systems, a relatively high amount of input total plasmids (∼1 μg/1E6 cells) is typically needed to achieve high AAV vector yields.^34,59,60,61 The AAVone system allows for a substantial reduction of input plasmid, while still increasing AAV yield in both HEK-293T and HEK-293 cells (Figures 7A and 7B). In HEK-293T cells, using 15.2-kb or 15.8-kb pAAVone plasmids that carrying CMV-Egfp genomes, the AAVone system reached its highest yield at 0.25 μg/1E6 cells, a result that was independent of the serotype (Figure S12). In HEK-293 cells, the optimal plasmid amounts were between 0.25 and 0.50 μg/1E6 cells (Figure 7A). The reduction in input plasmid also increased transfection efficiency for AAV2 vectors (Figure S12A). The use of reduced plasmids slightly increased cell viability during production of both AAV2 and AAV9 vectors (Figure S12B), while cell densities reached up to 6E6 cells/mL at the time of harvest (Figure S12C). Overall, the optimal plasmid dose required for the AAVone system is approximately one-half to one-fourth of the amount typically used in tri-plasmid systems.

Figure 7.

Reduction of AAVone plasmid increases the quality of vectors

(A and B) Effect of pDNA amount on AAV productivity of suspension HEK 293T(A) and HEK 293 (B) cells. (C–E) Analytical ultracentrifugation analyses of AAV2 vectors produced by triple-plasmid transfections (pTri-plasmid and mTri-plasmid) and AAVone produced under different plasmid amounts (C). AUC analyses of AAV8 (D) and AAV9 (E) vectors under different plasmid amounts x axis = sedimentation coefficient (S). The y axis is the normalized distribution c(s). (F) qPCR analyses of DNase-resistant target regions packaged with AAVone-produced vectors. The x axes represent positions moving away from the ITRs.

The use of reduced input plasmid with the AAVone system has additional benefits. First, it led to increased full:empty capsid ratios during AAV production (Figures 7C–7E), which are desirable for obtaining higher-quality AAV vectors with a larger proportion of functional, transgene-carrying capsids. To analysis the full:empty particle ratios by AUC, AAV vectors from CsCl index 1.360–1.380 were collected after one round of ultracentrifugation, which included both empty and full AAV particles. When using the same quantity of input pDNAs to produce AAV2 vectors (0.75 μg/1E6 cells), the AAVone system exhibited a similar full:empty ratio as compared with what was achieved by the tri-plasmid systems (Figure 7C). All systems demonstrate two prominent peaks at 66 S and 92 S, which corresponds with empty AAV2 particles and full capsids with 2.8-kb genomes, respectively. Approximately 30% of AAV particles found in crude lysates for each production system were full capsids. When the plasmid was reduced to 0.25 μg/1E6 cells with the AAVone system, the full:empty particle ratio increased from 30.02% to 38.48%. Similarly, titers for AAV8 and AAV9 exhibited an approximately 10% increase in full AAV particles when using a reduced amount of input pDNA (Figures 7D and 7E). Second, the use of reduced pDNA with the AAVone system can also help to minimize plasmid-related impurities. Reducing the pDNA amount led to a decrease in plasmid impurities on both sides of the ITRs (Figure 7F).

Discussion

Advantages of the AAVone system

The AAVone system described here is designed to bring all the essential components for AAV production into a single-plasmid (Figure 1), while retaining all of the essential vector performance features and qualities expected of those produced by standard tri-plasmid transfection platforms (Figures 2 and 4). These features include the preservation of important characteristics, such as capsid content, genome and ITR integrity, transduction efficiency, and full:empty ratio. With this system, users can integrate their existing GOIs into the AAVone manufacturing process without the need to develop additional upstream and downstream processes, analytical release, or characterization assays.

A notable benefit of using the AAVone system is the considerable reduction in costs associated with raw materials, as it only uses a small quantity of a single-plasmid, instead of three plasmids. Plasmids represent one of the most expensive components in AAV vector production. The large-scale production of three GMP grade plasmids with greater than 95% purity and absent of process-related impurities, continues to be a significant challenge. Although numerous researchers have dedicated efforts to developing dual-plasmid transfection systems for AAV vector production,^38,39,40,44 this study stands out as the first to investigate the production of AAV vectors using a single plasmid. The system enables the production of AAV vectors using only approximately one-quarter to one-half of the total pDNA normally required with conventional approaches (Figure 7). Combined with a 2- to 4-fold increase in AAV vector productivity, the AAVone system uses only one-tenth of the total pDNA to achieve the same titers produced from the pTri-plasmid system.

The AAVone system’s single-plasmid approach also addresses the challenges faced with traditional tri-plasmid systems, such as obtaining a consistent vector product (Figure 1B; Table S1). Finding the optimal ratio of the three plasmids, maintaining plasmid stability, and ensuring consistent copy numbers across different production batches is complex and is often times challenging when considering the different cell lines, AAV serotype, and the size and composition of the AAV cassette used.^34,59,60,61 The single-plasmid system minimizes the potential for introducing plasmid-related variation. Moreover, in the AAVone system, the Cap proteins and replicated AAV genomes, which are the two essential components for AAV assembly, are always present in the same cells. This ensures that they maintain the same ratio, thereby enhancing the consistency of the produced AAV vectors, including the full:empty ratios.

The AAVone system potentially offers a higher effective AAV vector production efficiency compared with the pTri-plasmid system. The increased productivity of AAVone system is not due to increased transfection efficiency (Figure S3), but increased production efficiency (Figure S13). Theoretically, within the AAVone system, there is a linear relationship between AAV vector production efficiency and transfection efficiency. However, in the tri-plasmid systems, transfection efficiency cannot linearly convert to AAV vector productivity, as only cells that are co-transfected with all three plasmids are able to produce AAV vectors (Figure S13). This effect becomes significantly pronounced at lower transfection efficiencies. It has been reported that, at a transfection efficiency of approximately 60%, only approximately 7% of cells produce measurable amounts of assembled AAV capsid with the tri-plasmid transfection method, while transfection with the dual-plasmid system increased cells positive for assembled AAV capsid by 4- to 5-fold.²⁰ Moreover, it has been reported that with three different dual-plasmid configurations, pOXB showed significantly superior results compared with the pDG-like and pLV systems.⁴⁴ We also found that the AAVdual configuration exhibited comparable productivity, and the pDG-like configuration showed a lower production efficiency than achieved with the tri-plasmid method (Figure 1B). These findings suggest that co-existence of the AAV vector genome and AAV helper genes may play an important role in the increased vector productivity by AAVone.

In tri-plasmid systems, both productivity and full:empty particle ratios vary with changes in plasmid ratios. Among the three plasmids, the cis plasmid is typically used in excess, while the trans plasmids require higher absolute amounts to sustain viral replication and packaging.⁶² Notably, reducing the cis plasmid to 10% of the standard dose does not significantly diminish AAV yield, but reducing the trans plasmids results in a linear decrease in AAV productivity.⁶² ITR-flanked vector DNA in the cis plasmid undergoes replication after transfection. Replication efficiency, rather than the initial plasmid amount, determines packaging efficiency when capsid protein abundance is sufficient.^63,64 This notion is supported by the lower pDNA requirement in HEK-293T cells compared with HEK-293 cells, likely due to the presence of the SV40 large T antigen in 293T cells, which stimulates Rep expression and enhances the replication of AAV genomes.^56,65,66 Moreover, the cells are healthier in low pDNA transfection condition, as Rep and PEI-induced toxicity is reduced (Figure S12). The increased full:empty ratio at the lower 0.25 μg/1E6 transfection condition may be due to reduced Cap expression and more replicated AAV genomes, achieving a good balance for AAV encapsidation, as saturation of capsid production and delayed, or extremely reduce DNA replication, result in empty capsids.^62,67

Although AAV production systems based on herpes simplex virus, vaccinia virus, and AdV have been developed, they have not been widely adopted due to concerns related to potency, quality, safety, and processing complexity.^{4,5,6,7,8,9,10,11} Previously, we developed the VV-AdV/HeLa system, which uses a vaccinia virus and an AdV to co-infect the suspension E1-expressing HeLaS1 cell line, QW158-7.⁶ The recently developed TESSA/HEK-293 system also uses AdV (two AdVs or one AdV plus one seed AAV) to produce AAV vectors.^7,68 However, producing AdVs requires extra time and effort, and the AAV products have the potential for contamination with AdVs, posing safety and quality concerns. Baculovirus/Sf9 systems are another way to produce AAVs approved by the U.S. Food and Drug Administration (FDA). They have unique advantages, such as high productivity and ease of scaling up.^5,69 However, the AAVs produced by these systems have lower potencies, often requiring higher dosages to achieve the same efficacies as those produced in human cell lines.¹¹ Additionally, producing baculovirus vectors takes extra time and effort. These vectors are also not very stable, as they can undergo mutations and recombination.³³ Producer cell lines, which integrate parts or all of the elements required for AAV production, are another appealing system for large-scale production, as they can significantly reduce costs.⁷⁰ However, generating these cell lines takes a long time, and achieving a productive and stable cell line is challenging. The mainstay technology for AAV vector production in gene therapy remains the tri-plasmid transfection method, which uses human cell lines, mainly HEK-293, to produce AAV vectors. Tri-plasmid systems are flexible and easy to use, but have relatively low productivity and are difficult to scale up. With the AAVone system, we can improve AAV productivity, while maintaining flexibility and high vector potencies, but without additional safety concerns by using viruses as helper. We have optimized the AAVone system in different sources of HEK-293, with various culture media and transfection reagents and achieved crude titers over 5E12 vg/mL for AAV9 (data not shown), which is approximately 1E6 vg/cell. All these advantages make the AAVone packaging system a simple, cost-effective, highly scalable, and reproducible AAV production system, particularly suitable for GMP-grade AAV production.

Genomic impurities by the AAVone system

Besides intact GOI genomes, AAV vector production also yields contaminants in the final product, including incomplete AAV vector genomes, and DNAs from the producer plasmids and host genome.²⁴ The presence of contaminants in clinical vectors is considered undesirable because it can lead to increased immune responses against the vector, potential integration into the host genome, and other safety concerns. Different types of packaged incomplete vector genome species are present in AAV vectors,⁵² but most are snap-back forms (Figure S4), which are double-stranded DNAs containing either 5′ or 3′ ITRs.²⁴ The proportion of incomplete genomes are sequence dependent, and regions with a high secondary structure, such as short hairpin RNA cassettes, result in a greater proportion of snap-back genomes.^29,51 In our AAV-CMV-Egfp constructs, the major truncation hotspot appears to be centered at the WPRE, hGH polyA, and CMV sequences, which are close to the ITRs (Figures 4A and 4B; Figure S6). In our previous studies, byproduct empty AAV particles were not truly empty, but contained different types of short vector genomes that were enriched for sequences corresponding with the ITRs.^11,71 We were unable to completely remove these aberrant genomes from AAV preparations, even after extracting thin bands related to full capsid fractions after CsCl ultracentrifugation (Figure S2; Figures 4A and 4B). However, the larger size of the pAAVone plasmid may contribute to the reduction of aberrant genomes, as the AAVone system yielded slightly more intact genomes and significantly less snap-back genomes in the final purification when compared with what was produced by the tri-plasmid systems (Figure 4; Table S2). Furthermore, decreasing the amount of the pAAVone plasmid during production significantly reduced impurities and empty particles (Figures 5 and 8). Both the template-switching during replication and non-homologous end-joining may result in the formation of snap-back genomes.⁵¹ The use of one large plasmid may reduce recombination events between the ITRs and other plasmid elements, which in turn may lead to decreases in genomic impurities. Previous studies have concluded that increasing the size of the plasmid backbone with stuffer DNA can reduce the level of plasmid backbone contaminants, as the two ITRs are not in range of each other to facilitate reverse packaging.^24,72,73 The pAAVone plasmid has 13 kb of non-GOI cassette sequences existing beyond the two ITRs. The oversized design did decrease the abundance of sequencing spanning the bacterial ori and the antibiotic resistance gene. All of the high abundance elements were positioned close to ITRs on the plasmid, which supports the idea that the majority of vector product-related impurities are ITR-containing fragments. In addition, the abundance of contaminants is mainly determined by the distance from the ITRs (Figures 5 and 8D) and sequences positioned more than 2.5 kb away from the ITRs, such as rep, E2, and E4 genes, are rarely packaged into AAV vectors (Figure 5). Thus, incorporating genes into a single plasmid did not significantly increase their related impurities.

The FDA requires that AAV products be screened for the presence of possible rcAAVs. Current vector production systems in use have been optimized to reduce rcAAV to levels of 1 or fewer rcAAV per 1E8 vector genomes.²⁵ Tri-plasmid systems using native rep-cap genes present the possibility of generating rcAAV through homology between the ITR and p5 promoter and through non-homologous recombination.^57,74 Spatially separating the rep-cap ORFs from the AAV GOI,^6,75 isolating them on distinct expression plasmids,⁷⁴ or increasing the rep-cap size by insertion of an additional intron,⁷⁶ can in theory eliminate or decrease the formation of rcAAVs. In the pOXB dual plasmid system, where the rep-cap and AAV-GOI sequences are located on the same plasmid, there was no significant increase in rep-cap packaging compared with tri-plasmid systems.⁴⁴ Additionally, in the pAAVone design, the rep gene is situated 5.6 kb away from the R-ITR, exceeding the AAV packaging capacity of approximately 5.0 kb. This could potentially reduce the efficiency of packaging rcAAVs. As summarized in Table S2, the pAAVone design did not increase reads containing rep sequences recombined with the GOI (rep chimeric with GOI). This finding suggests that the risk of recombination to generate rcAAV is minimal, despite the proximity of the sequences in the plasmid design.

Materials and methods

AAV plasmids

All vector constructs were generated using HiFi assembly (NEB), Gibson assembly, or T4 ligation and were scaled up with midi/maxi preparation kits (Takara Bio). The mini-pHelper-1.0 (Addgene #230933) was generated by deleting introns of E2 and E4 genes from the pHelper (AAVnerGene, SM000002) parent construct. pAAVdual-CMV-Egfp plasmid (Addgene #230931) was cloned by inserting AAV-CMV-Egfp genome (including ITRs) into mini-pHelper-1.0 at the PmeI site, located between pMB1 ori and VA RNA. mini-pHelper-AAV plasmids were cloned by inserting the rep-cap cassette into the ClaI site of mini-pHelper-1.0, which is located between E4 and VA RNA. pAAVone-AAV2-CMV-Egfp plasmid (Addgene #230930) was achieved by insertion AAV2-CMV-Egfp genome into the mini-pHelper-AAV2 at the PmeI site. pAAVone plasmids with different capsids and transgenes are based on pAAVone-AAV2-CMV-Egfp and generated by replacing the Cap ORF(such as pAAVone-AAV9-CMV-Egfp (Addgene #230928) and pAAVone-AAV-PHP.eB-CMV-Egfp (Addgene #230928) or the GOI (such as pAAVone-AAV9-CMV-Egfp-2.2kb, and -4.7kb). All the ITR-bearing plasmids were subjected to diagnostic restriction enzyme digests using SmaI and/or AheI, ITR sequencing, and/or whole plasmid sequencing to confirm the ITR structures. All the plasmids mentioned in this paper can be found at AAVnerGene’s website (www.aavnergene.com).

AAV production, purification, and titration

Adherent HEK-293T cells were cultured in DMEM medium with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S) in various culture vessels, including 24-well plates, 15-cm dishes, or 2-L roller bottles. Suspension HEK-293T and HEK-293 cells were expanded for at least three passages and inoculated into a shake flask with FreeStyle F17 expression medium (Thermo Fisher) with 4 mM GlutaMax before transfection. The transfections were carried out using either PEImax (Polyscience) or FectoVIR (Polyplus), adhering strictly to the provided guidelines. Experimental variations included different cell densities, amounts of pDNA, and ratios of transfection reagent to pDNA. Cells were collected for analyses 48 or 72 h after transfection. For crude titer estimations, cell samples underwent a single freeze-thaw cycle and sonication for lysis. Subsequent centrifugation allowed for the collection of supernatants, from which crude titers were assessed. The AAV vectors extracted from these samples were then purified through one or two CsCl-gradient ultracentrifugation steps before undergoing quality evaluation. Before vector titrations, both crude lysate and purified samples were treated with DNase I to remove un-encapsidated residual DNA, followed by proteinase K treatment. The quantification of AAV vector productivity was achieved using either ddPCR or qPCR, using primers and/or probes targeting the vector’s transgene. To establish a standard curve, a plasmid carrying the relevant sequence was linearized. Each experimental batch included a well-characterized AAV reference (AAVnerGene, DJ001001-AAV9) as an internal standard, ensuring data normalization against this control.

Mice

Six- to eight-week-old female C57BL/6J mice were administered AAV2-CMV-Egfp or saline IVT injection to the left and right eyes at 1E9 GC/dose (n = 10 per group). All AAV2 vectors were diluted with saline and fast green to 1.0E9 vg/μL. The IVT injections were performed with glass needles (Clunbury Scientific LLC; Cat no. B100-58-50) to deliver approximately 1 μL of fluid into the vitreous using the FemtoJet from Eppendorf with a constant pressure of 300 psi and injection time of 1.5 s. At 4 weeks post-injection, fundoscopy imaging was performed to assess the EGFP expression in the mouse retina. Mice were euthanized at 4 and 6 weeks postinjection. Eye cups were collected for immune-staining and molecular quantification.

Six- to eight-week-old female C57BL/6 mice were administered AAV1-CMV-Egfp, AAV8-CMV-Egfp, or saline via IM injection to the left and right TA muscle. Vectors were diluted with saline to 1E10 vg/μL and administered at 1E11 GC/dose (n = 10 per group). At 2 and 6 weeks postinjection, TA muscles were harvested and either placed in 4% paraformaldehyde (PFA, for histology) or snap frozen (for RNA and DNA extraction). Following 4% PFA overnight fixation, tissues were sent to UMass Chan Medical School Morphology Core for hematoxylin and eosin staining or placed in 30% sucrose for 72 h before embedding in Tissue-Tek Optimal Cutting Temperature Compound (Sakura Finetek USA Inc #4583). The embedded tissues were cryosectioned (10-μM slices) and mounted on slides with VECTASHIELD HardSet Antifade Mounting Medium, with DAPI (Vector Laboratories #H-1500-10) for microscopy. All images were visualized with a Leica DM6 Thunder microscope with a 16-bit monochrome camera. Images were processed by LAS X Life Science microscope software.

AAV vectors were intravenously administered to adult male C57BL/6J mice (8–10 weeks old, n = 3) at a dose of 1E14 vg/kg per mouse. Two weeks postinjection, animals were euthanized and intracardially perfused with cold 0.1 M PBS followed by 4% PFA in 0.1 M PBS. The brains were harvested, fixed overnight in 4% PFA, and cryoprotected in 30% sucrose at 4°C for 2–3 days. Eight series of 40-μm-thick sections were cut using a frozen sliding microtome, collected in sterile Eppendorf tubes containing 30% sucrose in 0.1M PBS, and stored at −20°C until use.

Immunohistochemistry

Treated retinas were cross-sectioned as previously described.⁴⁸ In brief, eye cups were dissected in cold 1× PBS and fixed in 4% PFA overnight at 4°C. Cryosections were cut to 12-micron thickness. The following primary antibodies and dilutions were used: chicken anti-EGFP antibody (1:1,000; Abcam; Cat no. ab13970), and rabbit anti-IBA1 (1:300; Wako; Cat no. 019–19741). All antibodies were diluted in PBS with 0.3% Triton X-100 and 5% BSA (CST). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (1:2,000, Sigma-Aldrich; Cat no. 9542). The following secondary antibodies were used: Donkey anti-Chicken IgY (H + L) Highly Cross Adsorbed Secondary Antibody, Alexa Fluor 488, A78948 and Goat anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 594, A-11037. All secondary antibodies were diluted 1:1000 in PBST (PBS with 0.25% tween 20). All images were visualized with a Leica DM6 Thunder microscope with a 16-bit monochrome camera. Images were processed by LAS X Life Science Microscope Software. Floating brain sections were washed three times with 0.5% Triton X-100/Tris-buffered saline (TBS) and blocked for 30 min in a blocking buffer containing 2% BSA in TBS with 0.5% Triton X-100. Primary anti-GFP antibodies were then added, and the sections were incubated overnight at 4°C. After three washes, the sections were treated with Alexa fluorescent-labeled secondary antibodies in blocking buffer for 1 h at room temperature. After washing and Hoechst counterstaining, brain sections were mounted onto glass microscope slides and cover slipped with Fluoroshield (Sigma Aldrich, Cat. #F6182). Tiled confocal images of entire brain sections were acquired and analyzed using the Leica Stellaris-5 confocal microscope system.

RNA and DNA extractions

Snap-frozen TA muscles were homogenized using Bullet Blender Green Bead Lysis Kit (NextAdvance #GREENE5) in conjunction with a Qiagen TissueLyzer II. One-half of the homogenized tissues were processed for RNA extraction using mirVana miRNA Isolation Kit (ThermoScientific #AM1561) according to the standard protocol. The remaining homogenized tissues were processed using QIAamp DNA Mini Kit (Qiagen #51306) according to manufacturer’s recommended procedures. Extracted RNAs were reverse transcribed into cDNA using Iscript Reverse Transcription Supermix (BioRad #1708841) and Iscript gDNA Clear cDNA Synthesis Kit (BioRad #1725035).

Snap-frozen retinas were homogenized using Qiagen 5mm TissueLyser bead (Qiagen #69965) in conjunction with Omni Bead Ruptor Elite. Homogenized retinal tissues were processed for RNA and DNA extraction using the Norgen Biotek RNA/DNA/Protein Purification Plus Kit (Norgen Biotek #47700) according to the standard protocol. Extracted RNA was reverse transcribed into cDNA using Iscript Reverse Transcription Supermix (BioRad #1708841).

ddPCR

ddPCR was performed on gDNA and cDNA using ddPCR Supermix for Probes (no dUTP), 2× (BioRad #1863024) with the following probe sets: Egfp-FAM (Thermo Fisher Scientific #4400291, Assay ID Mr00660654_cn), mouse Gapdh-VIC (Thermo Fisher Scientific #4448484, Assay ID Mm99999915_g1), and TaqMan Copy Number Reference Assay, mouse, Tfrc-VIC (Thermo Fisher Scientific #4458366). Input for AL 479, AL 480, and AL 481 ddPCR reactions were as follows: 40 ng gDNA per 20 μL reaction and 0.5 μL cDNA (diluted 1:15) per 20 μL reaction. Droplet generation was performed using a BioRad Automated Droplet Generator using the corresponding manufacturer recommended consumables into a 96-well plate. PCR reactions were conducted on a thermocycler using the following conditions: 95°C for 10 min, 32 cycles of 95°C for 30 s and 61°C for 1 min, and a final step of 95°C for 10 min. Droplets were read using a BioRad QX200 Droplet Reader and analyzed using QuantSoft software. Raw reads were exported and Egfp values were normalized either to diploid genomes or to housekeeping gene transcripts.

AUC analysis

AUC was used to quantify empty and full particles. AAV from CsCl index 1.360–1.380 were collected after one round of ultracentrifugation. AUC measurements were performed at 20,000 rpm and 20°C on a Beckman AUC Optima instrument (Beckman Coulter) equipped with an An-50 Ti analytical rotor (Beckman Coulter, Brea) using a total number of 150 scans per sample. Analysis of AAV samples by the SEDFIT c(S) model yields a distribution of sedimentation coefficients with each peak in the distribution representing AAV particles with full genomes or empty capsids. Integration of individual peaks yields the sedimentation coefficient (S) and the relative concentration of each species in the distribution.

Mass photometry

Mass photometry measurements were carried out on a SamuxMP instrument (Refeyn). Before each measurement, a calibration was conducted using truly-empty AAV2 particles (AAVnerGene, DE000000-AAV2), which was produced by transfection of one mini-pHelper-AAV2 plasmid (AAVnerGene, SM000002-AAV2) in HEK-293T cells.

qPCR titration plasmid related impurities contaminant amplicons

Purified AAV vectors were treated by DNase I and proteinase K. Vector genome titers and contaminant amplicon titers were assessed by qPCR using SYBR Green. Viral titers were quantified on a StepOnePlus Real-Time PCR System (Applied Biosystems) using the default settings. The procedure involved denaturation at 95°C for 5 min, followed by 40 cycles of amplification (95°C for 10 s, 60°C for 30 s), and concluding with a melting curve. qPCR primers targeting different regions of the pAAVone plasmid are used to qualify the impurities in final AAV products. A plasmid containing the corresponding target sequences was used to make the standard curve. All data were normalized to Egfp transgene ct values.

Characterization of the AAV capsid

The capsid compositions of AAV vectors were determined by Bio-Safe Coomassie G-250 stain. AAV vectors (∼5E11 particles) were resolved by electrophoresis on a 10% SDS-PAGE gel and standard Coomassie G-250 stain was performed according to the manufacturer’s procedures (Bio-Rad).

AAV vector transduction in vitro

The HeLa cell line used in this study were grown in DMEM with 10% FBS and 1% P/S at 37°C in a humidified environment supplied with 5% CO₂. For each transduction experiment, approximately 50,000 viable cells were seeded onto a 24-well plate 24 h before transduction. AAV vectors were added directly to each well at a multiplicity of infection of 10,000. EGFP expression was determined 48 h post-transduction by epifluorescence microscopy.

Flow cytometry

Cells transfected with the tri-plasmid systems or the AAVone system was collected at approximately 48 h and 72 h after transfection and processed using a BD Accuri C6 Plus flow cytometer equipped with an autosampler (BD Biosciences). At least 50,000 events were collected per sample with the fluidics rate set to slow (14 μL/min) with one agitation/SIP clean cycle between samples. The acquired data were next analyzed with FlowJo software (v10.8; FlowJo) using standard gating strategies to select for single cells (based on SSC-A vs. FSC-A and FSC-H vs. FSC-A scatterplots) and to calculate the percentage of GFP-expressing cells (based on FITC-A histogram).

Next-generation sequencing analysis of AAV vector genomes

Next-generation sequencing (NGS) of AAV vector genomes was performed by Azenta. The ssAAV viral DNA was extracted using the Invitrogen PureLink Viral RNA/DNA kit following manufacturer’s instructions. DNA samples were quantified using a Qubit 2.0 Fluorometer (Life Technologies) and sample integrity was checked using Agilent TapeStation 4200 (Agilent Technologies). Amplicon libraries were prepared using SMRTbell Express Template Prep Kit 3.0. The sequencing library was validated on the Agilent TapeStation and quantified by using Qubit 2.0 Fluorometer. All libraries were combined into a single sample and sequenced across two PacBio Sequel IIe SMRT cells using Sequel II binding kit 3.0 with heteroduplex detection mode on.

SMRT sequencing and AAV-GPseq

The bioinformatic workflows were conducted using the Galaxy platform.^77,78 Resulting consensus fastq files were mapped to the references as reported in the study using the Burrows–Wheeler aligner-maximal exact match (BWA-MEM)^79,80 in PacBio mode (-x pacbio). Aligned reads were visualized with the Integrated Genomics Viewer (IGV) tool version 2.14.0 with soft clipping on.⁸¹

Statistical analyses

The results were analyzed by one-way or two-way ANOVA statistical tests were performed in GraphPad Prism 10 along with Tukey’s multiple comparison statistical tests where applicable as described.

Data availability

All data are provided in the main text, figures, or the supplemental information, and raw data files are available upon request. Requests for reagents or cell lines used in this study can directed to the corresponding author, Qizhao Wang.

Acknowledgments

G.G. is supported by grants from the UMass Chan Medical School, United States (an internal grant) and by the National Institutes of Health, United States (R01NS076991-01, P01HL131471–05, R01AI121135, UG3HL147367–01, R01HL097088, R01HL152723–02, U19AI149646-01, and UH3HL147367-04). P.W.L.T. is supported by an award from The Bassick Family Foundation, a BRIDGE Fund Award (a UMass Chan Medical School internal grant), and by the National Institutes of Health, United States (1R21AI183080-01A1). W.H. is supported by the National Institutes of Health, United States (R01DA056876).

Author contributions

R.Y., J.Z., and Y.L. designed and performed the AAV productivity experiments; Y.W. and Y.L. conducted all AAV-related characterizations; N.T.T. analyzed the NGS data and prepared figures. M.Cu., T.C., and T.S. designed and carried out animal studies involving AAV1, AAV2, and AAV8, and drafted the parts of the manuscript that encompass this work; X.Y., D.Z., and D.J. conducted animal studies for AAV-PHP.eB and AAVrh.10; Y.L., Z.S., Y.D., L.F., H.H., B.W., and C.X. constructed and prepared all plasmids; L.W., H.Z., Y.L., M.Ch, J.L., and C.C. performed AAV production, purification, and quality checks; G.G. revised and organized the manuscript, and supported the AAV1, AAV2, and AAV8 animal studies; W.H. provided support for the AAV-PHP.eB and AAVrh.10 animal studies; D.Y. revised the manuscript, contributed to the design of the experimental strategy, and provided overall support for the study; P.W.L.T. analyzed the data, drafted the manuscript, and prepared the figures; Q.W. designed constructs, drafted the manuscript and figures, and provided overarching support for the study.

Declaration of interests

D.Y. and Q.W. are cofounders of AAVnerGene. D.Y., Q.W., Y.L., R.Y., and C.C. are co-investors on the AAVone patent (US20240132911A1). G.G. is a scientific co-founder of Voyager Therapeutics and Aspa Therapeutics and holds equity in these companies. G.G. and P.W.L.T. are inventors on patents with royalties licensed to biopharmaceutical companies.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) used ChatGPT to assist with revising the format, grammar, and spelling. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.