Development of a novel purification method for AAV vectors using tangential flow filtration

Adeno‐associated virus (AAV) vector can efficiently transduce therapeutic genes in various tissue types with less side effects; however, owing to complex multistep processes during manufacture, there have been surges in the pricing of recently approved AAV vector‐based gene therapy products. This study aimed to develop a simple and efficient method for high‐quality purification of AAV vector via tangential flow filtration (TFF), which is commonly used for concentration and diafiltration of solutions during AAV vector purification. We established a novel purification method using TFF and surfactants. Treatment with two classes of surfactants (anionic and zwitterionic) successfully inhibited the aggregation of residual proteins separated from the AAV vector in the crude product by TFF, obtaining a clearance of 99.5% residual proteins. Infectivity of the AAV vector purified using the new method was confirmed both in vitro and in vivo, and no remarkable inflammation or tissue damage was observed in mouse skeletal muscle after local administration. Overall, our proposed method could be used to establish a platform for the purification of AAV vector.

culture supernatants, or both) contains many cell-derived impurities, such as host cell proteins (HCP), host cell DNA, and plasmids used in transfection.Removal of these impurities from AAV vectors is critical for safety, especially for suppressing immunogenicity concerns in gene therapy.During the harvesting steps, AAV vectors are usually obtained from cell lysates (Kotterman & Schaffer, 2014); however, they are efficiently released into culture supernatant under nutrient-limiting conditions, except AAV2 (Vandenberghe et al., 2010).Because the suitable serotype is chosen based on their ability to target the desired cell/tissue type and the lower levels of neutralization antibodies, serotypes other than AAV2 have also been used in clinical settings, and a method to recover AAV vectors not only from cell lysates, but also from culture supernatant has been developed to lower the contamination of HCP and host cell DNA (Lock et al., 2010;Okada et al., 2009;Tomono et al., 2016).
During purification, the combination of polyethylene glycol or ammonium sulfate precipitation and density gradient ultracentrifugation with cesium chloride or iodixanol is the most common method for laboratory-scale purification of AAV vectors (Hermens et al., 1999;Merten et al., 2005;Zolotukhin et al., 1999).Although well-established at the laboratory scale, this method has technical difficulties to scale up to the industrial level due to limitations of working volume and complicated operations.Recently, AAV vector purification methods without ultracentrifugation have been developed and are expected to be applied in industrial applications (Okada et al., 2009;Tomono et al., 2016).Among many kinds of chromatographic techniques, the simple one-step antibody-based affinity chromatography has become attractive as a versatile and efficient method for purifying a broad range of AAV vector serotypes (Nass et al., 2018;Smith et al., 2009).However, affinity chromatography requires low pH elution that decreases viral infectivity (Marichal-Gallardo et al., 2021).Thus, developing a simple and high-quality purification method for AAV vectors is crucial to expand its potential use in gene therapy.
Tangential flow filtration (TFF) is a process of size-dependent separation, which is widely used in the biopharmaceutical and food industries.Fluid flows parallel to the filter in this filtration system; hence, the filter does not clog easily.TFF using ultrafiltration (UF) membranes is a commonly used method at the industrial scale, mainly for concentration and dialysis.Previous studies on viral vector purification with UF membranes claimed that residual host cell DNA and proteins can be removed using TFF (Grzenia et al., 2008), however, only partial reduction of residual impurities was demonstrated, and the sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) results showed that much more contaminated proteins remained in the TFF-purified samples than in the ultracentrifuged samples (Doria et al., 2013).
Here, we aimed to develop a simple method for high-quality purification of AAV vectors using TFF and surfactants.We focused on the characteristic that the size of monomeric proteins is usually less than 500 kDa unless the proteins form a complex.Therefore, we hypothesized that most HCP could be removed using TFF with 500 kDa cut-off UF membranes by increasing protein solubility using surfactants.We speculated that surfactant treatment does not affect AAV vector infectivity, because AAV vector is a non-enveloped virus that is strongly resistant to surfactant and conventionally, AAV vectors are recovered through cell lysis using surfactants (Srivastava et al., 2021).

| Production of AAV vector
HEK293EB cells were grown for 3 days to reach approximately 90% confluence.Plasmids used for AAV vector production were (1) the pAAV packaging plasmid with AAV rep and cap genes, (2) pAAV cytomegalovirus promoter (CMV)-enhanced ZsGreen1 fluorescent protein flanked by two identical inverted terminal repeats, and (3) pAd helper plasmid containing the adenovirus E2A, E4, and VA RNA helper genes (Xiao et al., 1998).pAAV packaging plasmids for AAV1, 2, and 5, pAAV CMV-enhanced ZsGreen1, and pAd helper plasmids were purchased from Takara Bio, and pAAV packaging plasmids for AAV8 and 9 were from addgene.The Rep/Cap, GOI (ZsGreen1), and helper plasmids were transfected at a ratio of 1:1:2 (33.3 µg of Rep/ Cap plasmid, 33.3 µg of GOI plasmid, and 66.5 µg of pHelper per flask) to cells maintained in DMEM without serum and supplemented with 1% penicillin and streptomycin, 1% GlutaMAX-I (Thermo Fisher Scientific), 12.6% NaHCO 3 , 13.4% D-glucose, and polyethyleneimine (PEI Max; Polysciences) at a DNA:PEI weight ratio of 1:3.For AAV1, 5, 8, and 9 vectors, the culture supernatant was collected 120 h after transfection, followed by centrifugation at 10,000×g for 15 min at 4°C and filtration using a 0.45 µm filter (Thermo Fisher Scientific).For AAV2 vector, 10 mL of 500 mM ethylenediaminetetraacetic acid and 50 mL of distilled water were added to the HYPERFlask after removing the culture medium to peel off the cells.The cell solution was centrifuged at 2000×g for 10 min, and the supernatant was separated.The precipitated cells were frozen in liquid nitrogen and thawed in a water bath at 37°C thrice.
To this, 30 mL of HNM buffer (50 mM HEPES, 100 mM NaCl, 1 mM MgCl 2 , pH 7.4) was added, centrifuged at 5000×g for 10 min, and the supernatant was filtered using a 0.45 µm filter, followed by the addition of HNM buffer to a final volume of 560 mL equal to the volume of HYPERFlask.
The flow rate was 200−800 mL/min.The pressure for the membrane was not regulated; during filtration, the transmembrane pressure was approximately 2−4 psi.Furthermore, buffer exchange was performed using 2 L of HNM buffer.After complete buffer exchange, 50 mL of the sample was collected, and the permeate line for filtration was closed; the membrane was washed with 50 mL of HNM twice by circulation of HNM inside the hollow fiber cartridge and tubes, and the wash buffer (100 mL in total) was collected.All TFF purifications were performed using the same initial volume of the supernatant (200 mL), which was finally reduced to 150 mL (50 mL of purified sample plus 100 mL of wash buffer).
For affinity chromatography, 1 L of AAV1 vector supernatant was concentrated to 50 mL using a 500 kDa UF hollow fiber cartridge without surfactant.Then, 10 mL of the concentrated solution was used as the starting material.One milliliter of POROS CaptureSelect AAVX Affinity Resin (Thermo Fisher Scientific) was added to the empty gravity chromatography columns to achieve a 0.58 mL bed volume.The affinity column was equilibrated with 10 column volumes of phosphate-buffered saline (PBS) with 1 mM MgCl 2 before loading the concentrated sample.The flow rate was set to 0.2 mL/ min.Following sample loading, the columns were washed with 20 mL of PBS with 1 mM MgCl 2 and subsequently eluted with 10 mL of 0.1 M citric acid (pH 2.5).The eluate was immediately neutralized with 1/7 vol. of 2 M Tris HCl (pH 11.2), which was already added into the eluate fraction wells.After elution, the columns were washed with 10 mL of 1 M NaCl.

| SDS-PAGE
First, 100 mL of the supernatant and TFF-purified sample were concentrated to 500 μL using a 100 kDa amicon Ultra-15 Centrifugal Filter Unit (Merck) to be visualized in SDS-PAGE.The loading volume for SDS-PAGE was adjusted so that the volume of the starting material was the same in each loaded sample.The samples were separated on a 12% (v/v) polyacrylamide gel (FUJIFILM Wako Pure Chemical Corporation) in a running buffer containing SDS.To visualize and analyze the SDS-PAGE bands, oriole fluorescent gel stain (Bio-Rad Laboratories) was used.
Proteins were detected using enhanced chemiluminescence (ChemiDoc Touch Imaging System; Bio-Rad Laboratories).
2.5 | Quantification of AAV vector, HCP, and double-stranded DNA AAV titer was quantified using quantitative polymerase chain reaction (qPCR) with the AAVpro Titration Kit (Takara).AAV vector recovery was calculated using the volume and AAV vector concentration of each purified solution by comparison with those of the starting material.The concentration of contaminating HCP and DNA in the supernatant and the purified solution were measured using the HEK293 HCP enzyme-linked immunosorbent assay (ELISA) Kit (Cygnus Technologies) and PicoGreen dsDNA Assay Kits (Thermo Fisher Scientific), respectively.HCP and DNA clearance were calculated by subtracting the recovery of HCP and DNA from 100.

| Morphological analysis of AAV vectors by transmission electron microscopy (TEM)
AAV vectors were hydrophilized using an ion bombarder (PIB-10; VACCUM DEVICE), and 3 µL hydrophilized samples were attached to a collodion membrane (Nissin EM) for 1 min.After washing thrice with 3 µL water, samples were stained using phosphotungstic acid for 10 s.The samples loaded on the membrane were analyzed using TEM (HT7800; Hitachi High-Tech).

| In vitro infection of AAV1 vector purified using different methods
Triplicate wells containing 1.4 × 10 5 HEK293EB cells per well in 12well plates (Corning) were cultivated for 3 days.The cells were infected with AAV1 vector that was either purified with TFF or affinity chromatography (2 × 10 9 vg/well).Three days after infection, the images of the cells were captured using a microscope (ECLIPSE Ts2; Nikon), and the proportion of ZsGreen1-positive cells were analyzed using flow cytometry (BD FACSMelod Cell Sorter; BD Biosciences).

| Statistical analyses
Data are presented as mean ± standard deviation.Two-tailed Student's t-test was used to determine statistically significant differences between the AAV1 vectors purified using TFF and those using affinity chromatography.There were three technical replicates, and p < 0.05 was considered significant.

| Animals
Nine-week-old wild-type female B6 mice were used.All mice were purchased from Nihon SLC, and all animal experiments were performed in accordance with the guidelines approved by the Ethics Committee for the Treatment of Laboratory Animals at The Institute of Medical Science, The University of Tokyo (PA20-24).Age-matched littermate mice were used in the experiments.

| In vivo administration of AAV vector
Three mice were anesthetized by inhalation of 2.0% isoflurane and oxygen before injection.One hundred microliter of AAV1-ZsGreen (1.0 × 10 11 vg in total) and HNM buffer were intramuscularly injected into mice in the right and left of tibialis anterior (TA) muscle, respectively at one site.The animals were euthanized 2 weeks after injection.

| Histopathological and immunohistochemical analyses
The individual muscles at the injection site (TA muscle) and back site (uninjected site, gastrocnemius medial head; GL muscle) were immediately frozen in liquid nitrogen-cooled isopentane for histological analysis.Transverse cryosections (8-µm-thick sections) were prepared from the skeletal muscles and stained with hematoxylin and eosin (H&E) using standard procedures (Feldman & Wolfe, 2014).
To confirm the expression of ZsGreen1 at the injection sites, the coverslip slides were mounted in Vectashield (Vector Laboratories Inc.) with 4,6′-diamidino-2-phenylindole.Immunofluorescence images were obtained using an IX81 fluorescence microscope (Olympus).

| RESULTS
3.1 | Concentration and buffer exchange of crude culture supernatant using TFF showed certain but not complete protein clearance The average diameter of AAV vector is 22−25 nm, and the radius of the protein at 500 kDa is approximately 5.21 nm if the protein has the simplest shape (Erickson, 2009); therefore, TFF with a 500 kDa cut-off UF membrane can efficiently concentrate AAV vector from crude culture supernatant, before subsequent purification processes (Tomono et al., 2018).Since most of the endogenous monomeric proteins present in mammalian cell lines are less than 75 kDa in size (Jin et al., 2010), exchanging the culture medium with a diafiltration buffer using a 500 kDa UF membrane can theoretically separate AAV vector from most residual proteins.To test this hypothesis, 200 mL crude culture supernatant of AAV1 vector produced by HEK293EB cell line was concentrated and the buffer was exchanged using a 500 kDa cut-off UF hollow fiber made of polysulfone, with an inner diameter of 0.5 mm.The total volume of purified solution was 150 mL derived from 200 mL of culture supernatant.AAV vector purity in the buffer-exchanged TFF solution was determined using SDS-PAGE with fluorescent gel staining and HCP ELISA (Figure 1a and Table 1).As hypothesized, simple buffer exchange using TFF without any additives F I G U R E 1 TFF purification without surfactants.Adeno-associated virus 1 (AAV1) vector was analyzed using (a) 12% (v/v) sodium dodecyl sulfate-polyacrylamide gel electrophoresis which was stained with oriole fluorescent gel stain, and were visualized using (b) transmission electron microscopy (TEM).cleared nearly 98% of the residual HCP proteins from the crude culture supernatant, and most protein bands derived from HCP observed in lane SM were not observed in lane 1 (Figure 1a and Table 1).To visualize the AAV particle proteins by SDS-PAGE, post-TFF samples were concentrated using a 100 kDa UF membrane.
However, analysis of the concentrated sample revealed the presence of many residual proteins other than AAV capsid proteins VP1, VP2, and VP3 (lane 2 in Figure 1a).The bands of residual proteins were smaller than 500 kDa when samples were denatured by SDS, suggesting that these residual proteins may form aggregates larger than the pore size of the UF hollow fiber membrane or may also have specific or nonspecific interactions with the AAV vector, thus enabling them to stay in the sample after TFF.TEM imaging of the concentrated purified sample also showed large structures, presumably aggregated impurities, apart from AAV vector particles (Figure 1b).

| A novel AAV vector purification method was established using a combination of TFF and surfactant treatment
According to our observations (Figure 1), it was hypothesized that inhibition of protein aggregation and interaction is required to remove residual proteins from the crude AAV vector supernatant.To avoid protein aggregation and interaction, three kinds of surfactants (non-ionic: octyl glycoside, anionic: sodium deoxycholate, and zwitterionic: CHAPS) were investigated.Two hundred milliliters of the supernatant in the same culture lot were incubated with 0.5% of each surfactant at 37°C for 30 min and sequentially concentrated to 50 mL; then, buffer exchange was performed with 2 L of HNM diafiltration buffer using TFF.To generate control samples, the same process was repeated using the supernatant without the surfactants.Then, SDS-PAGE with fluorescent staining was performed for the concentrated samples to visualize the variation in residual protein content (Figure 2a).Addition of octyl glycoside showed little effect on residual protein clearance (Figure 2a, lane 2).In contrast, that of CHAPS and sodium deoxycholate showed remarkable clearance of residual proteins, although not all residual proteins were cleared (Figure 2a, lanes 3, 4).Notably, the remaining residual protein bands after CHAPS and sodium deoxycholate treatment showed different patterns, suggesting that CHAPS and sodium deoxycholate inhibited the aggregation/interaction of different residual protein populations.Based on this result, culture supernatant was treated with both sodium deoxycholate and CHAPS.Surprisingly, incubation with 0.5% sodium deoxycholate and 1% CHAPS removed all residual protein bands, except those corresponding to VP1, VP2, and VP3 (Figure 2a, lane 5).The clearance of residual proteins in the sample washed with 0.5% sodium deoxycholate and 1% CHAPS was further confirmed by TEM imaging (Figure 2b).Thus, AAV vector could be simply purified by TFF combined with two different types of surfactants.The residual HCP and DNA contents were quantified using HCP-ELISA and Picogreen assay, respectively (Table 1 row 2−4).Although the difference in the residual HCP concentration between samples treated with different surfactants was not large, it was confirmed that more than 99.5% of HCP and 95% of DNA were cleared by this method.

| Novel AAV vector purification method has improved response to AAV vector biological activity compared to affinity chromatography
It was showed that TFF and surfactant treatment could be a good candidate for the AAV vector purification method.Although sodium deoxycholate and CHAPS are mild surfactants derived from bile acids predominantly existing endogenously in the bile of vertebrates, surfactant treatment may denature AAV capsid and affect the biological activity of AAV vector.To compare the AAV biological activity of TFF purification with that of a conventional purification method, AAV infection assay was performed.Affinity chromatography was chosen as the conventional method for comparison, as it is a major method for AAV vector purification in industrial processes (Marichal-Gallardo et al., 2021).For affinity purification, POROS AAVX affinity resin, a commonly used affinity resin for AAV vector purification, was used.The same starting material was used for both TFF purification and affinity T A B L E 1 AAV vector concentration, impurity concentration, and recovery of each material.chromatography.The peak fraction of the eluate of affinity chromatography, which showed the highest peak of absorbance at 260 and 280 nm, was collected as a sample for infection (Supporting Information: Figure 2a).HEK293 cells were infected with the same genomic copy number of TFF-and affinity-purified AAV1-ZsGreen1 vectors.Three days after infection, the number of ZsGreen1-positive cells was analyzed using flow cytometry to compare the biological activity of AAV vectors.Notably, 24.9 ± 0.95% of the cells was ZsGreen1 positive with TFFpurified AAV vector infection, whereas 20.8 ± 0.14% of the cells was ZsGreen1 positive with affinity-purified AAV vector infection, showing a statistical difference between the three biological replicates (Figure 3 and Table 2).This result suggested that the novel AAV vector purification method has comparable or improved response in terms of AAV vector biological activity compared to affinity chromatography.
Regarding impurity removal, TFF purification showed more than 99.98% and 95% clearance rates of HCP and DNA, respectively, which are similar to that of affinity chromatography purification (Table 3).
SDS-PAGE showed that the bands derived from impurities other than VP1, VP2, and VP3 were present in the affinity-purified sample, whereas only the bands derived from AAV vectors were detected in the TFF-purified sample (Supporting Information: Figure 2b), suggesting that, compared to affinity chromatography purification, TFF purification has superior or equal activity of impurity removal.

| AAV vector purified with the novel purification method accompanies negligible toxicity in live tissues
Although AAV vector infectivity was less affected by surfactant treatment than low pH treatment in affinity chromatography, the F I G U R E 2 TFF purification with surfactants.AAV1 vector was analyzed using (a) 12% (v/v) sodium dodecyl sulfate-polyacrylamide gel electrophoresis which was stained with oriole fluorescent gel stain before and after TFF purification, and were visualized using (b) transmission electron microscopy (TEM).TFF samples were ultracentrifuged using a 100 kDa amicon membrane after TFF purification.(right panels).Similar to the control injection site (PBS injection), neither inflammation nor muscle fiber necrosis was observed at the purified AAV1-ZsGreen injection site in light of nuclear infiltration.In summary, the novel purification method developed in this study did not affect AAV vector infectivity and safety.
3.5 | Novel AAV vector purification method is applicable to many serotypes from culture supernatant Finally, we aimed to determine whether the newly developed AAV vector purification method is applicable to other serotypes.In addition to the AAV1 vector crude culture supernatant, AAV2, AAV5, AAV8, and AAV9 vectors were produced by plasmid triple transfection in HEK293EB cells.The crude culture supernatant of AAV5, AAV8, and AAV9 vectors was prepared using the same method as AAV1 vector, and AAV2 vector was obtained by freezethawing the transfected cells.Crude AAV vector of all the five serotypes were purified, and protein clearance was determined using SDS-PAGE.AAV5, AAV8, and AAV9 vectors were successfully purified like AAV1 vector; however, for AAV2 vector, many residual HCP protein bands were observed in the same or higher expression levels of the bands derived from AAV2 vector (Supporting Information: Figure S3).Since the amount of HCP compared to AAV particle is very high in crude AAV2 lysate (Bennett et al., 2017;Rayaprolu et al., 2013), further optimization is required to purify AAV2 vector using this method.Thus, these data showed that novel AAV vector purification method is applicable to AAV1, 5, 8, and 9 but currently not AAV2 vector.AAV vector concentration and recovery of each serotype are shown in Supporting Information: Table 1.Low AAV2 vector recovery of post-100-kDa UF membrane might be caused by adsorption to 100 kDa UF membrane during concentration, and that is why AAV2-derived bands at SDS-PAGE were hardly visible.

| DISCUSSION
In this study, we developed a novel and simple method for highquality purification of AAV vectors using TFF and surfactants.After evaluating the addition of surfactants under various conditions, we successfully separated most of the host cell-derived protein and nucleic acid impurities by combining an anionic and a zwitterionic surfactant.AAV vector purified using this method was confirmed to have a higher infectivity than affinity-purified AAV vector in vitro.
Furthermore, no remarkable inflammation was observed on local administration of the AAV vector to mice.Thus, this method could be a very promising purification technique to make AAV products for gene therapy.
Previously, TFF using an UF membrane was generally introduced with the purpose of concentration and buffer exchange of AAV vector crude solution for the following purification steps, such as affinity chromatography, ion-exchange chromatography, and ultracentrifuge.During this process, certain amounts of residual HCPs and nucleic acids can be cleared through the pores of the UF; however, only partial reduction of residual materials would be expected (Doria T A B L E 2 AAV vector concentration, impurity concentration, and recovery of each material.Note: AAV vector titer, host cell protein (HCP) concentration, and DNA concentration were measured using quantitative polymerase chain reaction (qPCR), HCP enzyme-linked immunosorbent assay, and picogreen assay, respectively.SM, starting material; TFF, TFF purification with CHAPS and deoxycholate; AF peak, the peak fraction of the eluate of affinity chromatography, which showed the highest peak of absorbance at 260 and 280 nm.
T A B L E 3 In vitro assessment of AAV1-ZsGreen1 fluorescence.Note: Infection rate in HEK293 cells of purified AAV1 vector by TFF (post-TFF with deoxycholate and CHAPS) and affinity chromatography was analyzed using flow cytometry (n = 3).Statistical analysis was performed using an unpaired t-test.(*p < 0.05).
Abbreviations: AAV, adeno-associated virus; TTF, tangential flow filtration.et al., 2013).TEM images of post-TFF samples in our study clearly visualized structures larger than 100 nm in size, which are presumed aggregates of the residual protein (Figure 1b).Based on this, we hypothesized that weakening protein−protein interactions will prevent residual protein aggregation, thus allowing the efficient removal of residual proteins from the AAV vector crude product using TFF.
When a protein loses its three-dimensional structure, the hydrophobic amino acids buried within this structure are exposed and can easily interact with other molecules to form a large aggregate.To inhibit such disorganized protein aggregations, an anti-polymerizer was added.Among many reported anti-polymerizers (surfactants [Rosenow et al., 2002], chaotropic salts [Hatefi & Hanstein, 1969], arginine [Golovanov et al., 2004], alkylamines, and polyamines [Yasui et al., 2010]), surfactants were investigated in this study since they are commonly used for cell lysis to extract AAV vector, and AAV vector are resistant to surfactants (Blessing et al., 2019;Guo et al., 2012).Surfactants are commonly classified as follows according to the charge on the hydrophilic head: ionic (negatively or positively charged), non-ionic (uncharged), and zwitterionic (contains an equal number of positively and negatively charged functional groups).Among each category of surfactants, the investigated surfactants were selected using a criterion that they should cause minimum protein denaturation and have a low risk of being added to the authorization list (Annex XIV) of Registration, Evaluation, Authorization, and Restriction of Chemicals (REACH) regulation (Agency, 2022).Triton X-100 has been used primarily for cell lysis and provides sufficient yields for viral vector purification (Dias Florencio et al., 2015).However, Triton X-100 exhibits acute oral toxicity, eye damage, skin irritation, and chronic aquatic toxicity, resulting in its classification as a substance of very high concern by the European Chemicals Agency under the REACH regulation in December 2016 (Moleirinho et al., 2018).Thus, it can be mainly used for research purposes only.
Three surfactants were selected according to the previously mentioned classification (ionic: sodium deoxycholate, non-ionic: octyl glycoside, and zwitterionic: CHAPS).Octyl glycoside treatment had limited effect on residual protein clearance (Figure 2a), since nonionic surfactants usually disrupt lipid−lipid and lipid−protein interactions, not protein−protein interactions (le Maire et al., 2000;Sjögren et al., 2005).Addition of sodium deoxycholate and CHAPS showed a remarkable clearance of residual proteins, and residual protein bands obtained with sodium deoxycholate and with CHAPS treatment showed different patterns to each other (Figure 2a, lanes 3, 4), suggesting that sodium deoxycholate and CHAPS inhibited aggregation of different residual protein populations.CHAPS, a zwitterion, is a derivative of bile salt and contains positive and negative charges, which strongly inhibit nonspecific binding of proteins (Kondo et al., 2012;Rajan & Matsumura, 2017).CHAPS is very effective in breaking protein−protein interactions and degrading protein  2).HCP levels are generally measured using multiparameter ELISA; however, the coverage of antibody binding-HCP by ELISA kit is broad, although not all HCPs can be detected.Regarding comparison with a small amount of HCP, it is difficult to detect a slight difference using ELISA.Therefore, total HCP concentrations after performing various purification methods were confirmed using SDS-PAGE fluorescent gel staining.Additionally, the efficiency of DNA removal also increased after treatment with a mixture of sodium deoxycholate and CHAPS (95% of the total DNA was removed).Although nuclease treatment has become a standard element in the AAV vector manufacturing process (Hebben, 2018), TFF purification can reduce the amount of endonuclease in such process, which makes the purification process more cost-effective and reduces the risk of contamination owing to the endonuclease.This study also showed that TFF purification method was applicable to many serotypes such as AAV1, 5, 8, and 9 vectors, but not AAV2 vector (Supporting Information: Figure 3).For AAV2 vector production, the cells need to be lysed to collect AAV particle due to less release to culture suspension compared to other serotypes (Vandenberghe et al., 2010).Because the amount of HCP compared to AAV particle is very high in crude AAV2 product, the current method is not adequate to clear HCP.Thus, this method is not capable to use for AAV2 currently and more optimization is required.
In this study, TFF purification of AAV vector was performed using a lab scale peristaltic pump and a 650 cm 2 UF membrane.
Since filtration of hollow fibers can be scalable by increasing the surface area of the hollow fibers, this method could be applied to large-scale industrial AAV vector purification.The industry needs a platform purification process that accommodates a variety of viral vectors and potentially reduces the number of unit operations.In addition, minimizing the physical and chemical degradation of the capsid protein (by aggregation, proteolysis, oxidation, or deamidation) at all stages of downstream purification is vital to ensure retention of viral infectivity (Rodrigues et al., 2018).The infectivity and safety of AAV vector purified using the TFF purification method were tested both in vitro and in vivo (Figures 3 and 4).TFF purification method has comparable response on AAV vector biological activity compared to affinity chromatography.Moreover, the TFF method takes 1.5 h to purify 200 mL AAV culture media, while affinity purification on the same scale requires 3 h from column equivalation to elution.Thus, the TFF method provides a faster purification process than conventional methods.Although this method is effective in removing residual impurities, it cannot separate empty and full capsids.Therefore, it is expected to be the first step in the AAV purification process.Mice that were administered purified AAV vector via intramuscular injections did not show any significant inflammation at the site of administration (Figure 4).Although the efficacy and safety of the AAV vector purified using the proposed method were confirmed in mice, further safety evaluation, such as confirmation of surfactant removal, is recommended in higher mammals, such as dogs and monkeys, for application to clinical trials.
A limitation of the TFF purification in the present study is that the recovery percentage of AAV vector was not high (28%−42.5%) (Tables 1 and 2).Importantly, during TFF purification, AAV vector was barely detectable in the permeate solution sampled directly from the permeate side of UF membrane using qPCR, suggesting that AAV vector was lost owing to adsorption on the UF membrane when the sample purity was increased.We used a 500 kDa UF membrane made with polyether sulfone in this study.The pore size of 500 kDa was chosen to retain AAV vector and remove impurities after trying different pore sizes of UF membrane; however, to improve AAV vector recovery, optimization of the material of the UF membrane was needed for preventing the adsorption of AAV vector on the membrane.Additionally, to decrease the potential damage to AAV particles by shear, TFF condition (flow rate, pressure, temperature of solution) must be optimized.
We used a combination of surfactant treatment and TFF with UF membranes to purify AAV vectors.This principle could be applied to purify other biomaterials, such as recombinant antibodies, exosomes, and other viruses.Inhibition of protein aggregation and interaction using additives by optimization of the type and concentration of additives and buffer conditions can make TFF a scalable sizeexclusion method.

Rimi
(a) M, protein size marker; SM, starting material (pre-TFF); lane 1, post-TFF; lane 2, post-TFF sample after concentration using a 100 kDa amicon membrane.(b) The post-TFF sample after concentration was analyzed.The black arrowhead indicates the full capsid, and the white arrowhead indicates the empty capsid.The white arrows indicate the protein aggregates.Scale bar = 200 nm.TTF, tangential flow filtration.
(a) M, protein size marker; SM, starting material (pre-TFF); lane 1, TFF without detergent; lane 2, TFF with octyl glycoside; lane 3, TFF with 3-cholamidopropyl dimethylammonio 1-propanesulfonate (CHAPS); lane 4, TFF with deoxycholate; lane 5, TFF with CHAPS and deoxycholate.(b) The post-TFF sample with CHAPS and deoxycholate were analyzed after concentration using a 100 kDa amicon membrane.The black arrowhead indicates the full capsid, and the white arrowhead indicates the empty capsid.Scale bar = 500 nm.AAV, adeno-associated virus; TTF, tangential flow filtration.F I G U R E 3 In vitro assessment of AAV1-ZsGreen1 vector fluorescence.ZsGreen fluorescence images show AAV1-infected cells.Scale bar = 500 μm.AAV, adeno-associated virus.impact on in vivo toxicity was unclear.To test the toxicity of the AAV vector purified using the novel TFF purification method, mice were injected intramuscularly with TFF-purified AAV1-ZsGreen1 or the control PBS.Two weeks after injection, their muscles were cryosectioned and ZsGreen1 expression and tissue inflammation were analyzed.Fluorescence image of the cross-section of the AAVinjected muscle showed high expression of ZsGreen1, suggesting high transduction efficiency of purified AAV1-ZsGreen1 vector (Figure4, left panels).To determine any damage or inflammation in the muscle tissue at the injected site, H&E staining was performed, as seen in the figures adjacent to the fluorescence images in Figure4

F
I G U R E 4 In vivo expression of AAV1-ZsGreen1 vector in the skeletal muscle.Immunofluorescence images (left panels) and hematoxylin and eosin staining (right panels) of the injected site, tibialis anterior muscle derived from AAV1-ZsGreen1 vector injection (upper panel) and phosphate-buffered saline injection (lower panels) 2 weeks after injection.Immunofluorescence images contain nuclear stain 4′,6′-diamidino-2phenylindole (DAPI, blue).Inflammatory reaction was comparable between AAV and PBS injection samples.Neither inflammation nor muscle fiber necrosis was observed at the purified AAV1-ZsGreen injection site in light of nuclear infiltration.Scale bar = 100 μm.AAV, adenoassociated virus; PBS, phosphate-buffered saline.complexeswithout affecting secondary or tertiary structures(Rodi et al., 2014).Sodium deoxycholate is an anionic surfactant derived from bile acids and used for protein solubilization, similar to urea/ CHAPS solution(Lau & Othman, 2019).The detailed molecular mechanism by which sodium deoxycholate and CHAPS inhibit protein aggregation is unknown; however, our data indicated that both surfactants caused a reduction in large aggregates of different protein populations.A mixture of sodium deoxycholate and CHAPS treatment with TFF removed almost all residual proteins from the crude culture supernatant (Figure2a, lane 5).TFF purification removed 99.98% residual HCP, compared with 90% observed in hydrophobic interaction chromatography(McNally et al., 2020) and 99.9% in affinity chromatography (Table Miyaoka conducted the experiments and wrote the paper.Yuji Tsunekawa designed the experiments and wrote the paper.Yae Kurosawa, Takako Sasaki, Azusa Onodera, Kenji Sakamoto, and Yuko Kakiuchi conducted the experiments.Mikako Wada and Yuko Nitahara-Kasahara conducted the experiments and helped write the manuscript.Hiromi Hayashita-Kinoh and Takashi Okada supervised the study.