Adeno-associated virus (AAV) vector-based gene therapy is potentially curative for various genetic diseases; however, the development of a scalable purification method for full-genome AAV

vectors remains crucial to increase productivity and reduce cost of GMP production. In this study, we developed a large-scale short-term purification method for functional full-genome AAV

particles by using 2-step cesium chloride (CsCl) density-gradient ultracentrifugation with a zonal rotor. The 2-step CsCl method with a zonal rotor improves separation between empty and

full-genome AAV particles, reducing the ultracentrifugation time (4–5 h) and increasing the AAV volume for purification. The highly purified full-genome AAV particles were confirmed by

column. Interestingly, ddPCR analysis revealed that “empty” AAV particles contain small fragments of the inverted terminal repeat (ITR), probably due to unexpected packaging of Rep-mediated

ITR fragments. This large-scale functional AAV vector purification with ultracentrifugation would be effective for gene therapy.

Adeno-associated virus (AAV) vectors that express therapeutic gene products have shown great promise for gene therapy. Recently, AAV vector-based gene therapy trials have been reported in

various hereditary diseases, including Duchenne muscular dystrophy (DMD), X-linked myotubular myopathy (XLMTM), hemophilia A, and hemophilia B [1, 2]. However, in DMD and XLMTM gene therapy

trials, systemic injection of high-dose AAV9 vectors resulted in lethal liver failure at an early phase and death, most likely due to innate immunoreaction against the AAV genome and

complement activation with AAV particles [1, 3,4,5]. In gene therapy in hemophilia A and hemophilia B using AAV2, AAV8, and AAV10 vectors, liver enzyme elevation and AAV capsid-specific

T-cell activation were detected with subsequent declines in factor VIII and factor IX activity, respectively [2, 6].

To date, manufacturing purification methods for AAV vectors are generally based on ion-exchange and affinity chromatography [7] and this process can remove host cell proteins (HCPs).

Recently, AAV8 and AAV9 serotypes have been more commonly used because of their higher efficiency gene delivery compared to AAV2, enabling the harvesting of AAV vectors from cell culture

supernatant instead of cell lysate [8]. Moreover, these purified AAV vectors still vary according to the packaged genome sizes, including full-genome, intermediate, and empty particles,

which are produced during the AAV biomanufacturing process [4]. Empty capsids are thought to reduce transduction efficiency and induce unnecessary immune responses. In addition,

double-stranded RNA (dsRNA) can be generated by bidirectional promoter activity from the inverted terminal repeat (ITR) of AAV, enhancing innate immunity [9]. Ultracentrifugation with a

density gradient of cesium chloride (CsCl) or iodixanol allows AAV vectors to more efficiently separate the full-genome and empty particles compared to chromatography [10,11,12,13]. However,

this system is limited by its small scale, and long exposure to CsCl (conventionally for 2 days) reduces the transduction efficiency of AAV vectors [5, 14]. In addition, iodixanol is not

suitable for clinical use because of its cross-reactivity with iodine allergy.

For short-term purification of full-genome AAV vectors, we previously developed a 2-step CsCl density-gradient ultracentrifugation method; however, this is limited to a small scale (180 mL).

Therefore, in this study, we developed a large-scale (1000 mL), short-term purification system for functional full-genome AAV vectors using ultracentrifugation. In this system, a zonal

rotor (1.7 L capacity) was used to increase the AAV vector loading volume during ultracentrifugation, and a 2-step CsCl density gradient in the zonal rotor allowed for faster separation of

full-genome particles, resulting in shorter exposure to CsCl during ultracentrifugation and efficient recovery of full-genome AAV vectors.

AAV vectors were prepared in a large scale and harvested from culture supernatant (conditioned media), as previously described [3]. In brief, a 293EB cell line expressing adenoviral E1a,

adenoviral E1b, and Bcl-xL [15] was expanded in two 500 mL flasks (HYPERFlask, Corning, Corning, NY, USA) or a 1 L bioreactor (iCELLis Nano Bioreactor, Pall, Port Washington, NY, USA) for 5

days or 4 days, respectively, in Dulbecco’s Modified Eagle Medium (DMEM high glucose, FUJIFILM Wako, Chuo-ku, Osaka, Japan) with 10% fetal bovine serum (Thermo Fisher, Waltham, MA, USA).

Transfection was then performed with polyethylenimine max, (Polysciences, Warrington, PA, USA) using pAAV-ZsGreen1 (TaKaRa Bio, Kusatsu, Shiga, Japan), pRC9 (serotype 9), and helper plasmids

in DMEM including 2 mM L-Alanyl-L-glutamine Solution(100x) (Nacalai Tesque, Nakagyo-ku, Kyoto, Japan), 0.12% NaHCO3 (Nacalai Tesque), and 0.13% D-glucose (Nacalai Tesque) without serum.

Five days post-transfection, culture supernatants were harvested and treated with 18.5 U/mL endonuclease (KANEKA CORPORATION, Minato-ku, Tokyo, Japan) with 5 mM MgCl2 (Nacalai Tesque) for 30

 min at 37 °C. All cells were checked for mycoplasma contaminations resulting were reported negative.

Five percent CsCl (FUJIFILM Wako) in HNE buffer (50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES, FUJIFILM Wako), 0.15 M NaCl (Nacalai Tesque), and 25 mM

ethylenediaminetetraacetic acid (EDTA, Nacalai Tesque), pH7.4) or HN buffer (50 mM HEPES and 0.15 M NaCl, pH7.4) were added to the culture supernatant, including AAV vectors (Table 1). A

zonal rotor consists of a large cylindrical chamber subdivided into four sector-shaped compartments by vertical septa that radiate from the axial core to rotor wall. The entire chamber was

used during centrifugation and loaded with a single density gradient, and each sector-shaped compartment served as a large centrifuge tube. The large chamber capacity of these rotors (1.7 L)

eliminates the need for multiple runs and density gradients. A CsCl density gradient was generated in a zonal rotor (P32CT or P35ZT, Eppendorf Himac Technologies, Hitachinaka, Ibaraki,

Japan) at 3000 rpm by loaded 200 mL HNE or HN buffer, AAV vector containing 5% CsCl, 300 mL of 25–27% CsCl in HNE or HN buffer, and 300 mL of 38–40% CsCl in HNE or HN buffer. AAV vectors

were separated by ultracentrifugation (Himac CP 80NX, Eppendorf Himac Technologies) at 30,000–35,000 rpm for 4–10 h. After separation, 2 L of 42–45% CsCl buffer was slowly added to the

inside of the zonal rotor at 3000 rpm, and each fraction within the zonal rotor was pushed out from the outside (Tables 2, 3). RI were measured in each fraction using an refractometer NAR-1T

LIQUID or RX 5000i (Atago, Minato-ku, Tokyo, Japan). Each fraction sample was dialyzed with 20 kDa molecular weight cut-off dialysis cassettes (#66003 Thermo Fisher) in 0.5 mM MgCl2

(Nacalai Tesque) in water for ~2 h at 4 °C, and 0.5 mM MgCl2 in PBS (#27575–31, Nacalai Tesque) overnight at 4 °C.

After ultracentrifugation with a zonal rotor, AAV genome copies of each fraction were evaluated using the AAVpro Titration Kit (for Real Time PCR) Ver.2 (TaKaRa Bio, Kusatsu, Shiga, Japan)

in a QuantStudio 3 Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). The sample size was n = 3 minimally needed for statistically significance.

The AAV capsid proteins in each fraction were evaluated by western blot analysis. The samples were degraded with NuPAGE LDS sample buffer (Thermo Fisher) and NuPAGE Reducing Agent (Thermo

Fisher), electrophoresed on a 4–15% (v/v) gradient polyacrylamide gel (Criterion TG Precast Gels, Bio-Rad, Hercules, CA, USA) with SDS running buffer (Nacalai Tesque), transferred to a PVDF

membrane (Trans-Blot Turbo Midi PVDF Transfer Packs, Bio-Rad), and detected using anti-AAV VP1/VP2/VP3 mouse antibody (clone B1, Progen, Heidelberg, Germany) and Amersham ECL Mouse IgG,

HRP-linked whole Ab (Cytiva, Marlborough, MA, USA)5.

Transduction efficiency was evaluated using ZsGreen1-positive percentages (%ZsGreen1) in the transduced 293EB cells. 293EB cells (1 × 105 cells) were cultured in 24-well plates overnight and

transduced with each sample fraction (300 μL per well) in serum-free DMEM containing 2 mM L-glutamine, 12.1% NaHCO3, and 12.9% D-glucose (300 μL per well). The next day, 600 μL of the same

culture medium was added to each well, and %ZsGreen1 was evaluated by flow cytometry (FACSMelody, Becton Dickinson, Franklin Lakes, NJ, USA) at 3 days post-transduction and analyzed using

FlowJo Version 7.1 (Becton Dickinson).

The purity of AAV vectors was analyzed using a Proteome Lab XL-I ultracentrifuge (Beckman Coulter, Indianapolis, IN, USA). Bulk AAV vector samples (400 μL) were applied to the Centerpiece on

Cell Housing, and three cell houses with samples and one counterbalance were inserted into an AUC rotor. After equilibrating to 20 °C, samples were ultracentrifuged at 12,000 rpm at 20 °C,

and the absorbance (260 nm) and interference were measured at 92 timepoints for 4–5 h. The percentages of full-genome, intermediate, and empty AAV particles were analyzed using SEDFIT

(National Institutes of Health, Bethesda, MD, USA) [16] and visualized using GUSSI (UT Southwestern Medical Center, Dallas, TX, USA).

Whole regions of the AAV vector genome were evaluated in each fraction of the samples from ultracentrifugation with a zonal rotor using ddPCR. 1.1 μL of sample less than 10,000 copies/µL

(total 22 μL) was mixed with target primer/probe mixes (ddPCR Copy Number Assy, BioRad) (Table 4); 900 nM primers and 250 nM probe in droplets containing these materials were generated by an

Automated Droplet Generator (BioRad) followed by PCR reactions in a C1000 Touch Thermal Cycler (BioRad). A QX200 droplet reader (Bio-Rad) using the QuantaSoft software package (Bio-Rad) was

used to detect fluorescent signals in each droplet.

Collodion membranes (Nissin EM, Shinjuku, Tokyo, Japan) were hydrophilized using an ion bombarder (Nisshin EM Co., type PIB-10), and 3 μL of AAV samples were placed to a hydrophilized grid

for 1 min. After three times washing with 3 μL water, samples were stained with phosphotungstic acid (PTA) for 10 s. The samples loaded onto the membrane were analyzed using TEM (HT7800,

Hitachi High-Tech, Minato-ku, Tokyo, Japan).

Chromatography was performed using an ÄKTA avant 25 system (Cytiva, Marlborough, MA, USA) with a SuperloopTM 150 mL at a flow rate of 1.0 mL/min. A column (4.6 × 35 mm, Sugiyama Shoji Co.,

Ltd. Kanagawa, Japan) packed with CHT Ceramic Hydroxyapatite Type I, 40 m (Bio-Rad Laboratories Inc., Hercules, CA, USA) was equilibrated with 10 mM HEPES and 150 mM sodium chloride, pH 7.2.

The samples were loaded onto the column and eluted with 50 mM sodium phosphate buffer and 150 mM sodium chloride at pH 7.2. The resulting eluate was monitored for ultraviolet (UV)

absorbance at 260 and 280 nm and conductivity. The collected fractions were evaluated by qPCR, using primers and probes targeting ZsGreen1.

All values are expressed as means ± SEM. Statistical analysis of the data was conducted using a one-way ANOVA. For all statistical analyses, significance was defined as P