AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Original Article

AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations Simon Pacouret,1,7 Mohammed Bouzelha,7 Rajani Shelke,1 Eva Andres-Mateos,1 Ru Xiao,1 Anna Maurer,1,6 Mathieu Mevel,7 Heikki Turunen,1 Trisha Barungi,1 Magalie Penaud-Budloo,7 Frédéric Broucque,7 Véronique Blouin,7 Philippe Moullier,7 Eduard Ayuso,7 and Luk H. Vandenberghe1,2,3,4,5 1Grousbeck

Gene Therapy Center, 20 Staniford Street, Boston, MA 02114, USA; 2Department of Ophthalmology, Ocular Genomics Institute, Harvard Medical School, 243

Charles Street, Boston, MA 02114, USA; 3Schepens Eye Research Institute, 20 Staniford Street, Boston MA 02114, USA; 4Massachusetts Eye and Ear Infirmary, 243 Charles Street, Boston, MA 02114, USA; 5Harvard Stem Cell Institute, Harvard University, Cambridge, MA 02138, USA; 6Biological and Biomedical Sciences Program, Division of Medical Sciences, Harvard Medical School, Boston, MA 02115, USA; 7Atlantic Gene Therapies, INSERM UMR 1089, University of Nantes, Nantes University Hospital, 22 Boulevard Benoni Goullin, 44200 Nantes, France

Adeno-associated virus (AAV) vectors are promising clinical candidates for therapeutic gene transfer, and a number of AAV-based drugs may emerge on the market over the coming years. To insure the consistency in efficacy and safety of any drug vial that reaches the patient, regulatory agencies require extensive characterization of the final product. Identity is a key characteristic of a therapeutic product, as it ensures its proper labeling and batch-to-batch consistency. Currently, there is no facile, fast, and robust characterization assay enabling to probe the identity of AAV products at the protein level. Here, we investigated whether the thermostability of AAV particles could inform us on the composition of vector preparations. AAV-ID, an assay based on differential scanning fluorimetry (DSF), was evaluated in two AAV research laboratories for specificity, sensitivity, and reproducibility, for six different serotypes (AAV1, 2, 5, 6.2, 8, and 9), using 67 randomly selected AAV preparations. In addition to enabling discrimination of AAV serotypes based on their melting temperatures, the obtained fluorescent fingerprints also provided information on sample homogeneity, particle concentration, and buffer composition. Our data support the use of AAV-ID as a reproducible, fast, and low-cost method to ensure batch-tobatch consistency in manufacturing facilities and academic laboratories.

INTRODUCTION Adeno-associated viruses (AAVs) are 20- to 25-nm single-stranded DNA (ssDNA) viruses that belong to the family Parvoviridae. Their genome consists of two genes, rep and cap, flanked by two 145-bp inverted terminal repeats (ITRs) and packaged into a T = 1 icosahedral capsid composed of 60 protein subunits following a precise stoichiometry.1 The cap gene encodes the VP1, 2, and 3 C-terminally overlapping proteins that form the viral capsid,1 whereas rep codes for four multifunctional proteins involved in essential tasks of viral replication, such as DNA binding, site-specific endonuclease activity, and

helicase activity.2 Given its favorable toxicity and immunogenicity profile,3 its ability to mediate stable and long-term expression of therapeutic transgenes3 and to efficiently transduce both dividing and non-dividing cells,4 AAV has emerged as a potent gene transfer vector system for gene therapy of inherited and acquired diseases.5 AAV-based gene therapies are progressing to the therapeutic drug market, as illustrated by the EMA marketing authorization of Glybera, for the treatment of lipoprotein lipase deficiency.6,7 Therefore, increased emphasis is now placed on manufacturing and assay development to support the prospect of broad applications of AAV-based drugs.8–10 According to the FDA’s chemistry, manufacturing, and control (CMC) guidelines for human gene therapy investigational new drug applications (INDs),11 testing should provide information on the product sterility, stability, purity, potency, and identity. Adequate assays are available for many of these properties, but protein identity assessment of an AAV preparation at a particle level remains complex. Identity testing enables the verification of proper labeling of a given product and batch-to-batch consistency. Molecular methods, such as electrophoresis, PCR, and DNA sequencing, can establish the identity of the vector genome content. However, establishing capsid identity has proved to be more challenging. Protein sequencing,12 mass spectrometry (MS),13 and immunological identification techniques using capsid-specific antibodies are available but remain limited in applicability due to the need of high Received 1 December 2016; accepted 1 April 2017; http://dx.doi.org/10.1016/j.ymthe.2017.04.001. Correspondence: Luk H. Vandenberghe, 20 Staniford Street, Boston, MA 02114. E-mail: [email protected]

Molecular Therapy Vol. 25 No 6 June 2017 ª 2017 The American Society of Gene and Cell Therapy.

1

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

Figure 1. Comparative Analysis of Thermostability Assays (A) VP1-VP3 sequence homology matrix of AAVrh32.33, AAV9, and AAV5. (B) Normalized SYPRO Orange fluorescence signals obtained for AAVrh32.33 (plain line; Tm = 64.6 C), AAV9 (dashed line, Tm = 76.2 C), and AAV5 (dotted line, Tm = 87.4 C). (C) Half-maximum emission wavelengths (excitation: 293 nm; emission: 323–400 nm) plotted as a function of the incubation temperature for AAVrh32.33, AAV9, and AAV5. Error bars represent the SD of the mean obtained from three independent experiments. (D) Migration of AAVrh32.33, AAV9, and AAV5 VPs through an SDS-PAGE gel following incubation at 65 C, 78 C, or 95 C.

amount of substrate, detection bias, their time-consuming nature, availability of instrumentation, operator dependency, and/or lack of validation. An ideal AAV protein identity assay would convey information about primary, secondary, tertiary, and quaternary structure. Verification of primary protein sequence identity of a manufacturing run can monitor for operator-error (e.g., transfection of incorrect DNA construct in a triple transfection production), mutations within production reagents (e.g., clonal expansion of mutated AAV producer cell line), or the chemical modification of residues in a manufacturing process. MS and protein sequencing can both do so; however, neither provide information on structural conformation beyond primary sequence and are therefore incapable of distinguishing a partially denatured AAV sample from a purely intact sample. In addition, these techniques require specific buffer conditions and thus cannot be used for analysis of serotype identity in the final vector formulation. Higher-level structural information can be gleaned from other methods, but those methods are often limited in resolution (e.g., electron microscopy), suffer from detection bias (e.g., affinity reagents), or require substantial amount of substrate (e.g., analytical chromatography). Importantly, none of these methods are likely to provide primary sequence identity information in direct conjunction with the structural assessment (except, for example, if a mutation disrupts capsid structure as a whole or the particular epitope). One low-cost and easy-to-implement experimental approach that has become popular to study protein thermostability is differential scanning fluorimetry (DSF).14–17 This method is used to monitor the unfolding of proteins in response to a temperature gradient in the presence of a fluorescent dye such as SYPRO Orange. The fluorescence of this dye is quenched by solvent molecules yet increases upon binding to the hydrophobic sites that are externalized during thermally induced protein unfolding.15 This technique was recently introduced to the AAV research field to study capsid thermostability.18

2

Molecular Therapy Vol. 25 No 6 June 2017

Major differences between the melting temperatures of naturally occurring AAV serotypes were observed (AAV2: melting temperature [Tm] = 69.6 C ± 0.5 C; AAV5: Tm = 89.7 C ± 0.7 C),18 whereas reconstructed ancestral AAV particles exhibited enhanced thermostability in comparison to most of their contemporary, naturally occurring homologs.19 In this study, we explore the possibility of using the biophysical measure of thermostability for AAV serotype identification at the protein level. We show, through the analysis of 67 AAV samples by DSF, that capsid thermostability can be used to discriminate AAV1, AAV2, AAV5, AAV6.2, AAV8, and AAV9 preparations, regardless of the packaged transgene and vector concentration. This assay, referred to as AAV-ID, also provides information on vector concentration, homogeneity, and formulation. The sensitivity, linearity, and reproducibility of AAV-ID is evaluated and its potential uses and limitations discussed.

RESULTS Comparative Analysis of AAV Capsid Thermostability Assays

First, to assess whether thermostability reflects an intrinsic AAV property and is independent of the measurement methodology, the highly divergent AAV serotypes 5, 9, and rh32.33 (Figure 1A) were analyzed using DSF (Figure 1B), intrinsic fluorescence spectroscopy (IFS) (Figure 1C), and SDS-PAGE (Figure 1D). The IFS approach is based on the intrinsic fluorescence of tryptophan residues, which are highly conserved across AAV serotypes and whose spectral properties are closely dependent on their local environment. Changes from a hydrophobic to a hydrophilic environment lead to a red shift of the emission spectrum of tryptophan upon excitation in the UV range.20 The SDS-PAGE method relies on the differential ability of intact particles and monomeric proteins to enter and migrate through a 4%–12% polyacrylamide gel under electrophoretic conditions. In the IFS experiment presented in Figure 1C, the half-maximum emission wavelength (l50%) between 323 and 400 nm, in response to an excitation at 293 nm, was measured for each AAV sample,

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

www.moleculartherapy.org

peratures greater than or equal to 65 C, 78 C, and 95 C, suggesting that capsid denaturation had taken place at those critical temperatures (Figure 1D). Based on these results, and the fact that the resolution, assay time, and convenience of the DSF method were superior to the other methods, further studies were performed exclusively using the DSF assay, referred to as AAV-ID. A detailed stepwise protocol for AAV-ID is provided in Supplemental Materials and Methods. Subsequent studies were designed to validate AAV-ID for specificity, reproducibility, and sensitivity, as well as outlining its other potential uses and limitations. Specificity of AAV-ID

Figure 2. Statistical Analysis of AAV Sample Thermostability Average melting temperatures obtained for a set of 67 AAV preparations of six different serotypes. Error bars represent the SD of the mean. n represents the number of preparations per group. Statistically significant differences were found between the melting temperatures of all the serotypes considered in this experiment (one-way ANOVA and multiple comparison test with Bonferroni correction: p < 0.0001). The assay was run at constant volume per well (45 mL).

upon incubation at room temperature (RT), 50 C, 60 C, 70 C, 80 C, or 90 C. An increase of l50% indicated tryptophan exposure to solvent due to capsid denaturation.20 In the SDS-PAGE experiment, AAV samples were incubated at RT, 65 C, 78 C, or 95 C, prior to being run on a polyacrylamide gel. Intact capsids were retained in the wells of the gel, whereas VP monomers resulting from capsid denaturation migrated readily through the gel. In line with previous findings,18,19,21 DSF experiments showed that AAV5 was stable up to temperatures approaching 90 C (Tm = 87.4 C) and AAV9 denatured at 76.2 C. AAVrh32.33, a serotype that has not been previously studied, denatured at 64.6 C (Figure 1B). These measurements were validated by IFS (Figure 1C) and SDSPAGE (Figure 1D) experiments. An irreversible red shift in the emission spectrum of AAVrh32.33, AAV9, and AAV5 samples was observed at 70 C, 80 C, and 90 C, respectively (Figures S1 and 1C). Furthermore, all serotype capsids were retained in the wells of the gel following electrophoresis after sample pre incubation at RT (Figure 1D). Additionally, AAVrh32.33, AAV9 and AAV5 VP components migrated readily through the gel after preincubation at tem-

First, our goal was to extend the initial observations made in Figure 1 that AAV serotypes have a unique thermostability profile that could be indicative of its protein identity. To test this hypothesis, and determine whether Tm measures using DSF can provide a “fingerprint” of the capsid protein composition of an AAV preparation, we analyzed a set of 67 randomly chosen AAV preparations using AAV-ID (Table S1). The selection was limited to vector preparations of six distinct serotypes (AAV1, AAV2, AAV5, AAV6.2, AAV8, and AAV9) with various titers (4  1010 to 4.33  1013 vg/mL) and ITR-to-ITR genome sizes (2.1 to 4.7 kilobase pairs [kb]). All of the preparations were produced by triple transfection in HEK293 cells, purified by iodixanol ultracentrifugation, and resuspended in PBS without calcium and magnesium, supplemented with 5% glycerol. As shown in Figure 2, a specific average melting temperature could be derived for every serotype tested in this experiment. Melting temperatures were closely distributed around their mean value, with a maximum SD of 0.77 C, obtained for AAV6.2. Statistically significant differences were found between the melting temperatures of every serotype pair considered in this experiment (one-way ANOVA and multiple comparison test with Bonferroni correction: p < 0.0001) (Figure 2). Next, we analyzed the data to determine whether the viral titer of the input sample and transgene size influenced the AAV-ID readout. Results showed no significant impact of these parameters on the melting temperatures of AAV particles (Table S1). In addition, when comparing thermostability profiles of preparations that contained empty (E) only and enriched full (F) (or genome-containing) particles, we determined that for AAV5 (Tm E = 87.8 C versus Tm F = 87.8 C) and AAV8 (Tm E = 70.2 C versus Tm F = 71.4 C), the presence or absence of genomes had no major impact on the melting temperature. The difference of 1.2 C observed for AAV8 could result from interactions between the viral genome and capsid proteins and therefore requires further characterization. Test-Retest and Lab-to-Lab Reproducibility of AAV-ID

In order to evaluate the robustness of AAV-ID, six AAV preparations of four different serotypes (Table S2) were analyzed in triplicates, in

Molecular Therapy Vol. 25 No 6 June 2017

3

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

Figure 3. Reproducibility of AAV-ID (A) Comparison of the average fluorescence fingerprints obtained in lab A (blue) and lab B (red) for six different AAV preparations. The shaded regions represent the standard deviation of the mean obtained from three independent experiments. (B) Comparison of the melting temperatures obtained in lab A (blue) and lab B (red). Error bars represent the SDs of the mean obtained from three independent experiments. (C) SDS-PAGE run at constant volume per well showing the protein content of each of the six AAV preparations analyzed by DSF.

additional AAV preparations with various levels of protein contaminants were analyzed by DSF (Figure 4).

two gene therapy laboratories: the Grousbeck Gene Therapy Center (MEEI/Harvard, Boston, MA, USA) and INSERM UMR 1089 (University of Nantes, Nantes, France). The experimental protocol used for this comparative study was carefully harmonized between both of the laboratories, equipped with two different qPCR instruments (7500 versus StepOnePlus Real-Time PCR System, Thermo Fisher Scientific). In each laboratory, fluorescence transitions occurred within a relatively narrow range of temperatures (Figure 3A). The SDs of the mean Tm values were below 0.23 C (Figure 3B), indicating the high internal reproducibility of this assay for determination of AAV capsid melting temperatures. The normalized averages obtained from triplicates in both laboratories approximated each other closely, particularly in the region of the temperature transition, where curves were found to overlap with minimal SDs (Figure 3A). The differences between the mean Tm values obtained in both laboratories for each preparation ranged from 0.13 C (AAV8_#2) to 1.07 C (AAV5) (Figure 3B), suggesting that AAV-ID is a highly reproducible assay.

Influence of Protein Contaminants on AAV-ID

In an effort to identify the causes of the high fluorescence values preceding AAV2 fluorescence transition (Figure 3A), the six AAV preparations were run at constant volume on a polyacrylamide gel, under denaturing conditions (Figure 3C). In this particular case, the AAV2 preparation showed the highest levels of contaminants in comparison to the other preparations investigated in that experiment. In order to confirm this correlation between the pre-transition SYPRO Orange fluorescence levels and the amount of impurities,

4

Molecular Therapy Vol. 25 No 6 June 2017

Highly pure AAV preparations (AAV8 #1 and AAV9 #1), as indicated by SDS-PAGE (Figures 4C and 4D), resulted in DSF fingerprints with low initial fluorescence levels followed by a rapid increase and then a progressive decrease of the fluorescence (Figures 4A and 4B, plain lines), characteristic of protein unfolding with subsequent aggregation.22 On the contrary, AAV preparations with higher levels of protein contaminants (AAV8 #2 and AAV9 #2) (Figures 4C and 4D) presented high SYPRO Orange fluorescence values at temperatures below the expected AAV particle Tm (Figures 4A and 4B, dotted lines). For both low-purity preparations, the intrinsic AAV melting temperature remained detectable. However, the fluorescence transition was partially obscured, indicating a decrease of the assay’s sensitivity at high contaminant levels. With the purpose of gaining further insight on the type of contaminants responsible for the distorted fluorescence signals obtained for AAV8 #2 and AAV9 #2, we analyzed these four preparations by western blot using the anti-VP1/2/3 B1 antibody (Figures 4E and 4F). In the case of AAV8, higher levels of capsid-derived contaminants (Figure 4E) could be observed for the preparation giving high fluorescence at low temperatures (Figure 4A). However, these contaminants could only be observed upon over exposure of the polyvinylidene fluoride (PVDF) membranes (Figure 4E), whereas the two AAV9 preparations exhibited quasi-identical capsid-derived protein profiles (Figure 4F) in spite of their very different fluorescence fingerprints (Figure 4B). Although capsid-derived contaminants resulting from capsid proteolysis are likely to interfere with the assay, this piece of evidence indicates that other contaminants, such as cellular proteins, are likely to be responsible for the high fluorescence values preceding the denaturation transition of the AAV9 #2 preparation (Figure 4B). Complementary assays would be required to identify the type and levels of contaminants responsible for these distorted fluorescence

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

www.moleculartherapy.org

Figure 4. Influence of Protein Contaminants on AAV Sample Fluorescence Fingerprints (A and B) Normalized fluorescence fingerprints obtained for two AAV8 preparations (A) and two AAV9 preparations (B). (C and D) SYPRO-Ruby-stained SDS-PAGE run at constant volume per well showing the protein electrophoretic profiles of each AAV sample. (E and F) Analysis of capsid-derived protein contaminants by western blot using the anti-VP1/2/3 B1 antibody.

results, the sample concentration did not affect the measured melting temperature (Tm = 70.67 C ± 0.53 C). Impact of the Vector Formulation

fingerprints. Nevertheless, these observations suggest that AAVID can provide information on the presence or absence of unwanted protein contaminants in AAV preparations and that AAV-ID is limited to AAV preparations above a certain level of purity. AAV-ID Signal Amplitude Is a Function of Virion Concentration

Previous analyses illustrated a large spectrum of absolute fluorescence amplitudes at the apex of the Tm peaks. We next asked whether these amplitudes were related to amount of AAV capsids in a given preparation. In order to determine the relationship between AAV particle levels and AAV-ID signal amplitudes at the apex of the Tm peak, equal volume from 30 AAV8 preparations was subjected to AAV-ID and SDS-PAGE (Figures 5A and 5B). Results demonstrate a clear linear relationship between the capsid protein values (measured as total pixel density of VP1+ VP2+VP3) and the corresponding SYPRO Orange fluorescence amplitudes measured at the apex of the Tm peaks (R2 = 0.9222) (Figure 5C). We additionally tested the linearity of the assay, as well as its sensitivity, by diluting a single AAV8 preparation serially. Sample with titers between 3  1011 and 3  1012 particles/mL (pt/mL), as titrated by ELISA and calibrated using the rAAV8 Reference Standard Materials (RSM8),23 were subjected to AAVID (Figure 5D). Results show linearity of particle dose and DSF amplitude for one order of magnitude, yet there was not an indication of a loss of linearity at the highest concentration tested (Figure 5E). The limit of detection was established to be at 3.95  1011 pt/mL (i.e., 1.78  1010 pt/well). Furthermore, in line with previous

The formulation buffer of an AAV preparation has the potential to impact aspects of safety, potency, and therapeutic outcome. The ideal formulation would ensure vector stability while also minimizing particle aggregation, autoproteolysis, and oxidation, as well as adsorption to surfaces during shipping, storage, and administration.24 To determine the impact of the vector formulation on AAV-ID, an AAV9 preparation was supplemented with 0.001% PLURONICF68, 5% D-sorbitol, or 0.25% BSA, a protein homolog to human serum albumin (HSA). These three excipients have been used to inhibit adsorption of AAV capsids to container and instrument surfaces in hemophilia B25,26 and Leber congenital amaurosis (LCA)27 gene therapy clinical trials. Results show that PLURONIC F-68 did not interfere with the fluorescent dye and that it had no effect on the melting temperature of AAV9 particles (Figure 6A). D-sorbitol also did not interfere with SYPRO Orange, yet it resulted in a consistent increase of the AAV9 Tm of 1.2 C (Figure 6A). Finally, supplementation of the buffer formulation with BSA yielded maximum fluorescence levels at low temperatures, and significantly decreased the amplitude of the AAV9 capsid-specific transition (Figure 6B). Nevertheless, such transition could be recovered through low-pass filtering of the sample (AAV9+BSA) and blank (BSA) fluorescence signals, followed by subtraction of the blank to the sample signal. The melting temperature measured from the resulting curve (Tm = 75.4 C) (Figure 6C) suggested that BSA had little to no impact on the thermostability of AAV9 particles. We further investigated whether the ionic strength could impact the readout of AAV-ID by analyzing an AAV9 preparation formulated in PBS, supplemented with increasing amounts of NaCl. Results show that the melting temperature of AAV9 remained unchanged in the salt concentration range (0–800 mM) tested in this study (Figure S2). Finally, to evaluate the impact of the pH on the fluorescence signature of AAV preparations, AAV-ID was run with AAV1, AAV2, AAV5, AAV6.2, AAV8, and AAV9 particles formulated in citrate

Molecular Therapy Vol. 25 No 6 June 2017

5

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

Figure 5. Linearity and Sensitivity of AAV-ID (A) SYPRO Orange fluorescence fingerprints obtained for 30 AAV8 preparations analyzed by DSF. (B) SDSPAGE at constant volume per well obtained for these 30 AAV8 preparations. (C) SYPRO Orange fluorescence amplitude plotted as a function of the VP1-3 pixel density values. The dot plot obtained was analyzed by linear regression (R2 = 0.9222). (D) Background subtracted fluorescence fingerprints obtained for a set of AAV8 serial dilutions, using an AAV8 internal control of 3  1012 particles/mL (pt/mL). A clear fluorescence transition was obtained at concentrations greater or equal to 3.95  1011 pt/mL, equivalent to 1.78E10 pt/ well (E) SYPRO Orange fluorescence amplitude plotted as a function of the AAV particle concentration (pt/mL). The dot plot obtained was analyzed by linear regression (R2 = 0.9852).

an affinity not observed on AAV1.28 The crystal structure of AAV6 reveals the presence of a positively charged patch located at the base of the icosahedral threefold axis protrusions (Figure 7B), formed by residues R576, K493, K459, and K531, and responsible for the affinity toward HSPG.30 One interesting observation is that these major phenotypic changes can be attributed to a single amino acid difference between AAV1 and AAV6 VP1 protein located at position 531 in both serotypes (Figures

buffer at various pH, ranging from 7 to 3. Remarkably, the melting temperature of AAV vectors was strongly affected by changes in pH, in a serotype-dependent manner (Figures 6D and 6E). In particular, the Tm of AAV1, AAV6.2, and AAV9 was little affected between pH 7 and 5 and significantly decreased between pH 5 and 3 (Figures 6E and S3). The Tm of AAV2 and AAV8 reached a maximum at pH 5 (Figure 6E). Finally, the melting temperature of AAV5 exhibited a progressive decrease between pH 7 and pH 3 (Figures 6D and 6E). These data indicate that the pH of the vector formulation has a major impact on the readout of AAV-ID.

sequences,28,29 7A and 7B).

Applications and Limitations of AAV-ID

In accordance with previous studies,28,29 residue swapping between AAV1 and AAV6.2 at position 531 was detrimental to AAV1 and beneficial to AAV6.2 production yields (Figure 7C). Remarkably, the fluorescence fingerprints obtained for AAV1-E531K and AAV6.2K531E were nearly overlapping with those obtained for AAV6.2 and AAV1, respectively, indicating that residue 531 is a key determinant not only of virus assembly/packaging efficiency and tropism but also of its capsid thermostability (Figure 7D).

Finally, we aimed at testing the applicability of AAV-ID in addressing practical, translational, and/or biological questions. First, we wanted to determine whether minimal structural alteration to capsid composition impacted AAV-ID readout and whether these could inform us on underlying biology. As a case study, AAV1 and AAV6 were investigated, as their VP1 amino acid sequences differ by only six residues (Figure 7), yet these two serotypes exhibit very different biological phenotypes.28,29 Previous studies indicated that AAV1 production led to higher yields than AAV6 by 2- to 4-fold, whereas AAV6 outperformed AAV1 in vivo following intravenous injection.28,29 Furthermore, AAV6 binds heparin sulfate proteoglycan (HSPG),

6

Molecular Therapy Vol. 25 No 6 June 2017

A difference of 4.7 C was found between the capsid thermostability of AAV1 (82 C ± 0.66 C) and AAV6.2 (77.31 C ± 0.77 C), an AAV6 variant with enhanced in vivo gene transfer obtained by a single amino acid change in its VP1/2 domain.29 In order to assess whether residue 531 played a role in capsid thermostability, we generated the AAV1-E531K and AAV6.2K531E mutants by site-directed mutagenesis and subjected them to AAV-ID.

Next, we compared the AAV-ID fluorescence profile of AAV2 and AAV2-R585A-R588A (AAV_HSPG-). The RXXR motif, disrupted in AAV_HSPG-, forms a cationic patch on the spikes of the threefold axis (Figure 7E), which was shown to be critical for AAV2 heparin binding. Furthermore, RXXR was also shown to be determinant of

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

www.moleculartherapy.org

Figure 6. Impact of the Vector Formulation on the SYPRO Orange Fluorescence Fingerprint (A) Average melting temperatures measured from an AAV9 preparation formulated in PBS, PBS/PLURONIC F-68, and PBS/D-sorbitol. Results are given as mean ± SD of the mean, obtained from three independent experiments. (B) Fluorescence signals obtained for the same AAV9 preparation, formulated in PBS supplemented with 0.25% BSA and PBS 0.25% BSA only. (C) Background subtracted fluorescence signal represented between 65 C and 85 C, region delimited by the gray rectangle in (B), where denaturation of AAV9 particles formulated in PBS typically occurs. The melting temperature measured from this signal was 75.4 C. (D) Normalized fluorescence signals obtained for AAV5 particles re suspended in citrate buffer at pH 7-3. (E) Melting temperatures plotted as a function of the pH for AAV1, AAV2, AAV5, AAV6.2, AAV8, and AAV9.

virus uptake into human dendritic cells (DCs) and capsid-specific T cell activation.31 In line with the observations made for AAV1 and AAV6.2, disruption of the RXXR positively charged patch in AAV2 triggered an increase of 5 C in capsid thermostability (Figure 7F). Based on these observations, we conclude that a single residue change affecting the virus capsid biological properties, such as assembly/ packaging or tropism, can also play a key role in its thermostability.

Additionally, the two AAV9 preparations, produced by triple transfection and purified by iodixanol ultracentrifugation or affinity chromatography using the POROS CaptureSelect AAV9 affinity resin, also exhibited the same melting temperature. In the case affinity chromatography, vector elution at very acidic pH (0.1 M sodium acetate, 0.5 M NaCl [pH 2.5]) followed by neutralization in basic buffer (1 M Tris-Base/HCl [pH 10]) did not impact the structural stability of AAV9 particles, indicating the reversibility of the loss of stability observed at very acidic pH (Figure 6E). To summarize, these data suggest that the system used for vector production and purification has little to no impact on AAV-ID readout.

DISCUSSION To further investigate whether subtle capsid composition differences could impact the readout of AAV-ID, we tested delta-VP1 AAV8 capsids, obtained by introduction of a stop codon in cap to prevent VP1 expression. In accordance with previous studies,18 AAV8 and deltaVP1 AAV8 exhibited very similar fluorescence fingerprints, with a Tm of 70.5 C (Figure 8A). Lastly, to investigate the impact of the different production methods on AAV-ID readout, we analyzed two AAV8 (Figure 8B) and AAV9 (Figure 8C) preparations produced and purified using different production systems. AAV8 Baculo/sf9-CsCl particles were produced in lab B using the baculo/Sf9 system and purified by double CsCl ultracentrifugation, whereas AAV8 HEK293-IDX particles were produced in lab A using the HEK293 triple transfection system and purified by iodixanol ultracentrifugation. No major difference was observed between the fluorescence fingerprints obtained for these two AAV8 preparations.

Current US Food and Drug Administration guidance states that “the content of a final container of each filling of each lot shall be tested for identity after all labeling operations shall have been completed.”32 To date, the available methods to evaluate batch-to-batch consistency and identity of AAV vector preparations at the protein level are limited. One dimensional SDS-PAGE analysis does not allow for differentiation of assembled particles versus their monomeric components and is not suitable for discrimination of AAV capsid serotypes with limited VP sequence divergence (e.g., AAV1 and AAV6). Antibody-based identity assays, such as ELISA, are limited in terms of specificity due to cross reactivity between the antibodies directed against AAV capsids. One method that was successfully implemented to distinguish AAV capsids with minimal sequence divergence is MS. Nevertheless, current MS-based approaches for AAV capsid serotype identification are sample and time-consuming and only provide information on the sequence of capsid proteins. In addition, MS is not suitable for analysis of vectors in their final formulation and is thus of limited use for evaluation of batch-to-batch consistency of AAV preparations.

Molecular Therapy Vol. 25 No 6 June 2017

7

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

Figure 7. Effect of Amino Acid Mutation on AAV1, AAV6.2, and AAV2 Thermostability (A and B) Solvent accessible surface and calculated electrostatic potential projected on the solvent-accessible surface of AAV1 (PDB: 3NG9; http://www.rcsb.org/pdb/ explore.do?structureId=3ng9) (A) and AAV6 (PDB: 3SHM30) (B) viewed parallel to the threefold axis. Black residues are divergent between AAV1 and AAV6.2. Residues E/K531 are represented in red. The positive electrostatic potentials are represented in blue and the negative in red. (C) vg viral vector yields measured for AAV1, AAV6.2, AAV1-E531K, and AAV6.2-K531E by qPCR following triple transfection in HEK293 cells (n = 3). Results are given as mean ± SD of the mean, obtained from three independent experiments. (D) DSF analysis of AAV1 (blue line), AAV6.2 (red line), AAV1-E531K (blue dashed line), and AAV6.2-K531E (red dashed line). (E) Solvent-accessible surface and calculated electrostatic potential projected on the solvent accessible surface of AAV2 (PDB: 1LP346) viewed in the vicinity of the threefold axis. (F) DSF analysis of AAV2 (blue line) and AAV2_HSPG- (red line).

Our data show that AAV-ID overcomes some of the limitations of other assays, and may have utility in laboratory, manufacturing, or pharmacy routine testing of AAV lots aimed to demonstrate identity relatively to a reference sample. First, AAV-ID allows for simple, rapid (6 hr), and high-throughput (96-well plates) testing of AAV serotype identity, at low cost, and with limited input material. This assay can be implemented in every research or manufacturing facility equipped with a spectrophotometer with a temperature gradient control system and a filter cube compatible with the SYPRO Orange dye (470/570 nm). Second, as shown here, AAV-ID is a robust, reproducible, and hence easy-to-validate and broadly applicable assay. Following calibration of the assay with AAV products of interest, this method could be used to confirm the identity of vector preparations prior to their

8

Molecular Therapy Vol. 25 No 6 June 2017

release to the clinic. Any significant deviation of the preparation fluorescence fingerprint from a reference signal could be indicative of operator errors, biological divergence of a production process, issues at a fill-finish and/or pharmacy level, the presence of unwanted protein contaminants, or suboptimal vector concentration. Considering the linearity of AAV-ID, this assay could also be calibrated using reference AAV preparations such as RSM233 or RSM823 and further used for quantification of VP proteins levels in therapeutic products. One limitation of AAV-ID is its moderate resolution. For instance, AAV9 and AAV6.2 exhibited similar melting temperatures (Tm_AAV6.2 = 77.31 C ± 0.77 C and Tm_AAV9 = 76.17 C ± 0.2 C) despite a limited VP3 sequence homology (79.6%). Furthermore, not all mutations are equally important to the thermostability of virus capsids.34 While decreasing the temperature step in the gradient may allow further discrimination of AAV serotypes based

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

www.moleculartherapy.org

Figure 8. Effect of Capsid VP1-2-3 Composition and Production Method on the Fluorescence Fingerprint (A) Normalized SYPRO Orange fluorescence signal obtained for AAV8 and delta-VP1 AAV8, both produced by triple transfection in HEK293 cells, purified by double cesium-chloride gradient centrifugation and resuspended in PBS2+ (21-030-CV, Corning). (B) Normalized SYPRO Orange fluorescence signals obtained for AAV8 Baculo/sf9-CsCl and AAV8 HEK293-IDX preparations. (C) Normalized SYPRO Orange fluorescence signals obtained for AAV9 HEK293-POROS and AAV9 HEK293-IDX preparations.

on their melting temperature, it remains possible that two distinct AAV capsids have close to identical Tm that cannot be distinguished using this assay. In these situations, other assays, such as MS, will be required to complement AAV-ID for capsid serotype identification. MS and AAV-ID emerge as two complementary techniques for assessment of AAV product identity prior to their release to the clinic. While the high resolution offered by MS enables us to distinguish AAV capsid variants with single amino acid mutations, AAV-ID enables us to probe the identity of intact AAV particles in their final formulation. In addition to the characterization of AAV-ID in the context of quality control of AAV products, we also used this assay to investigate the relationship between the biological and biophysical properties of AAV capsids. In particular, the study of AAV2, AAV2_HSPG-, AAV1 and AAV6.2 thermostability (Figure 6) enabled us to show that limited sets of residues critical to the tropism, immunogenicity, and assembly/packaging of AAV particles could also be a key determinant of their capsid thermostability. The structural stability of AAV capsids was also shown to vary with the pH in a serotype-dependent manner. With the exception of AAV5, the thermostability of every serotype increased between pH 7 and pH 5–6, a pH transition observed between physiological conditions and early/late endosome. This increase was small for AAV1, AAV6.2, and AAV9, moderate for AAV8, and more significant for AAV2. In addition, a decrease in thermostability, ranging from 4 C (AAV8) to 8 C (AAV2), was observed between pH 5 to pH 4. In a recent study,35 AAV2-R432, a packaging-deficient AAV2 mutant, was shown to be 10 C less thermostable than wt-AAV2, suggesting that capsid structural stability was a key determinant of genome packaging. Then, in vitro genome release was shown to be gradually facilitated for AAV2 between pH 5 and 7.36 Finally, a structural rearrangement of AAV8 capsid was observed at pH 4, resulting in a decrease in association energy and buried surface area between VP monomers at the twofold interface.37 Together with our data, these observations suggest a link between capsid stability and genome release during the virus life cycle. Acidification of the local environment of the AAV particle in lysosomal compartments could disrupt the tight balance of forces between the packaged genome and the capsid walls, hence facilitating genome release. In summary, we conclude that thermostability is a robust and absolute measure that can be used to probe the identity of AAV capsid

particles in their final formulation. This measure was little affected by the size of their packaged genome, method of production, or level of protein contaminants in the vector suspension but strongly affected by the pH of the vector formulation. AAV-ID emerges as a valuable tool to probe the identity of AAV lots for experimental and clinical use and will help to ensure patient’s safety.

MATERIALS AND METHODS Site-Directed Mutagenesis

pAAV2/1-E531K and pAAV2/6.2-K531E plasmids were generated from pAAV2/1 and pAAV2/6.2 ITR-free RepCap production plasmids using the Phusion Site-Directed Mutagenesis Kit (New England Biolabs) and two pairs of complementary mutagenic primers designed using the QuickChange Primer Design Program (Agilent Technologies) (Table S3). Production, Purification, and Titration of rAAV Vectors

Unless mentioned in the text, AAV vectors were produced using the triple transfection system in hyperflasks with 70%–80% confluent HEK293 cell monolayers according to methods previously described.38 Cells were transfected with the AAV cis, AAV trans, and adenovirus helper plasmids, at a ratio 1:1:2, using polyethylenimine (PEI) (Polysciences). Three days after, cells and supernatant were collected, benzonase-treated (250 U/mL) (EMD Millipore), and precipitated overnight at 4 C in high-salt solution. Samples were filtrated and concentrated by tangential flow filtration (TFF) (Vivaflow 200, Sartorius) prior to undergoing iodixanol density gradient centrifugation. The fraction enriched in full capsids was buffer exchanged and concentrated using an Amicon filter with a molecular weight cutoff of 100 kDa (EMD Millipore). DNase I-resistant vector genomes (vg) were quantified by quantitative polymerase chain reaction (qPCR) using the TaqMan (Life Technologies) system with primers and probes targeting promoter, transgene, or poly(A) sequences of the transgene cassette. PAGE

All materials and reagents were purchased from Life Technologies. For each AAV preparation, 5 mL sample was mixed with 5 mL 4X NuPAGE lithium dodecyl sulfate (LDS) sample buffer and 10 mL PBS (21-031-CM; Corning) and incubated at 99 C for 10 min. Polyacrylamide gels were loaded with 10–15 mL of this sample and run at 100 V for 2.30–3 hr. Gels were further stained using SYPRO Ruby per

Molecular Therapy Vol. 25 No 6 June 2017

9

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

the manufacturer instructions. To assess the thermostability profiles of AAVrh32.33, AAV9, and AAV5, 20 mL protein preparation were incubated at RT, 65 C, 78 C, or 95 C for 10 min prior to be loaded in the gel. Western Blot

AAV preparations were run on polyacrylamide gels at 1 mL/well, 100 V for 2 hr 30 min. Proteins were transferred onto PVDF membranes (88518, Thermo Fisher Scientific) for 1 hr at 25 V using a Trans-Blot SD semi-dry transfer cell (Bio-Rad). Membranes were incubated for 5 hr at 4 C in blocking buffer (PBS 0.2%, Tween 20, 5% milk powder) followed by an overnight (ON) incubation at 4 C in primary antibody solution (PBS 0.2%, Tween 20, 0.5% milk powder, 1:200 B1 antibody) (ARP). Membranes were washed three times for 5 min in PBS 0.2%, Tween 20, incubated for 1 hr at RT in secondary antibody solution (PBS 0.2%, Tween 20, 1:5,000 HRP-linked sheep anti-mouse) (NXA931, GE Healthcare Life Sciences) and washed an additional three times for 5 min prior to being imaged.

maximum.18 A detailed stepwise protocol for AAV-ID is provided in Supplemental Materials and Methods. To derive the melting temperature of AAV9 vectors formulated in PBS supplemented with 0.25% BSA (Figure 8C), the sample and blank (PBS-0.25% BSA) signals (Figure 8B) were smoothed using a moving average filter (span: 10) prior to blank subtraction. To investigate the effect of the pH on the melting temperature of AAV vectors, 5 mL pure vectors, formulated in PBS, was diluted in 40 mL 0.6M acetate buffer at pH 7-3. The mix was supplemented with 5 mL 50X SYPRO Orange (final volume: 50 mL) prior to running AAV-ID. Addition of one volume of PBS to 8 volumes of acetate buffer did not significantly affect the pH of the solution. Variations in pH in response to changes in temperature were measured experimentally for each solution (pH 3: dpH/dT = 0.0119 C 1; pH 4: dpH/dT = 0.0093 C 1; pH 5: dpH/dT = 0.0029 C 1; pH 6: dpH/dT = 0.0012 C 1; pH 7: dpH/dT = 0.0047 C 1). Densitometry Analysis

Intrinsic Fluorescence Spectroscopy

For each AAV preparation, 2 mL aliquots were prepared and incubated at RT, 50 C, 60 C, 70 C, 80 C, or 90 C for 4 min. 2 mL of each sample were loaded onto a Take 3 Micro Volume Plate (Bio-Tek). Samples were excited at 293 nm, and emission spectra were acquired from 323 to 400 nm with a resolution of 1 nm, using a Synergy HI Hybrid Plate Reader (Bio-Tek) with the following acquisition parameters: gain: 100; detector position: 7 mm (top). Sample and blank emission spectra were further smoothed with MATLAB (MathWorks) using a moving average filter (span: 15). After background subtraction, spectra were normalized between 0% and 100%, and the wavelength corresponding to a normalized fluorescence of 50% was determined for each serotype and temperature condition. DSF Experiments

The protocols, adapted from Rayaprolu et al.,18 were carefully harmonized between both AAV research laboratories. SYPRO Orange 5000X (S6651, Thermo Fisher Scientific) was aliquoted (5–10 mL) and stored at 30 C for up to 18 months. For each new experiment, a 500 mL sample of SYPRO Orange 50X was prepared using PBS2+ (21-030-CV, Corning) as a solvent. 96-well plates were loaded with 45 mL samples, supplemented with 5 mL Sypro Orange 50X. PBS2+ and 0.25 mg/mL Lyzozyme (L6876, Sigma-Aldrich) solutions were used as negative and positive controls, respectively. Plates were sealed and centrifuged at 3,000 rpm for 2 min and subsequently loaded into the real-time PCR instrument (Boston: 7500 Real-Time PCR System; Nantes: StepOnePlus Real-Time PCR System, Thermo Fisher Scientific). Samples were incubated at 25 C for 2 min prior to undergo a temperature gradient (25 C to 99 C, 2 C/10 min, step and hold mode with 0.4 C temperature increments), while monitoring the fluorescence of the SYPRO Orange dye using the ROX filter cube available on both qPCR systems. Fluorescence signals F were normalized between 0% and 100% and melting temperatures were defined as the temperature for which the numerical derivative dF/dT reached its

10

Molecular Therapy Vol. 25 No 6 June 2017

Polyacrylamide gels were imaged using the ChemiDoc MP imaging system and the Image Lab software (Bio-Rad), optimizing the exposure time for intense bands to prevent pixel saturation. For pixel density measurements of VP1-3 protein bands, the gel analysis plugin of the Fiji software.39 Electrostatic Calculations

AAV1 (PDB: 3NG9; http://www.rcsb.org/pdb/explore.do?structureId= 3ng9) and AAV6 (PDB: 3SHM30) crystal structures were obtained from the Protein Data Bank.40 60-mer PDBs were further reconstructed using the UCSF Chimera package41 and prepared for electrostatic calculations using the PDB2PQR tool42 with the PARSE force field. Protonation states were assigned at pH 7 using PROPKA.43 Electrostatic calculations were adapted from the literature44 and run using the adaptive Poisson-Boltzmann solver (APBS).45 Briefly, the linear PoissonBoltzmann equation was solved on a 400  400  400 Å3 grid to a resolution of 1 Å using the single Debye-Hückel boundary condition, with the following parameters: dielectric constant of solute, 2; dielectric constant of solvent, 78.5; 0.150 M monocations/anions with an exclusion radius of 2 Å. Calculated electrostatic potentials were further projected on the solvent-accessible surface of viral capsids and visualized between 5 kT/e and +5 kT/e using the PyMOL Molecular Graphics System (version 1.6.0.0, Schrödinger).

SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Materials and Methods, three figures, and three tables and can be found with this article online at http://dx.doi.org/10.1016/j.ymthe.2017.04.001.

AUTHOR CONTRIBUTIONS Conceptualization, L.H.V., E.A., P.M., V.B., and S.P.; Methodology, Validation, and Formal Analysis, R.S., M.B., and S.P.; Resources, R.X., T.B., A.M., H.T., M.P.B., F.B., M.M., and E.A.M.; Writing – Original Draft, S.P.; Writing – Review & Editing, L.H.V., E.A.,

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

www.moleculartherapy.org

P.M., and E.A.M.; Supervision, L.H.V., E.A., P.M., M.P.B., and E.A.M.; Funding Acquisition, L.H.V., E.A., and P.M.

14. Layton, C.J., and Hellinga, H.W. (2010). Thermodynamic analysis of ligand-induced changes in protein thermal unfolding applied to high-throughput determination of ligand affinities with extrinsic fluorescent dyes. Biochemistry 49, 10831–10841.

CONFLICTS OF INTEREST

15. Niesen, F.H., Berglund, H., and Vedadi, M. (2007). The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability. Nat. Protoc. 2, 2212–2221.

L.H.V. is an inventor on gene therapy technologies licensed to several biotechnology and pharmaceutical companies, a consultant to several gene and genome editing gene therapy companies, and co-founder of and holds equity in GenSight Biologics, a retinal gene therapy biotechnology company.

ACKNOWLEDGMENTS We thank Daniel Navarro (MEEI Bioinformatic Center), Kevin Eappen, Sean Gai, Sandy Michaud and Kelsey Stappen (Schepens Eye Research Institute), and Nicole Brument (Nantes University Hospital) for technical support and Eric Zinn, Sarah Wassmer, Livia Carvalho, Amanda Dudek, Christophe Chevé, and Achille François for discussions. Support for this work comes from Mass Eye and Ear, Giving/Grousbeck, INSERM, and Nantes University.

REFERENCES 1. Rose, J.A., Maizel, J.V., Jr., Inman, J.K., and Shatkin, A.J. (1971). Structural proteins of adenovirus-associated viruses. J. Virol. 8, 766–770. 2. Fields, B.N., Knipe, D.M., and Howley, P.M. (2013). Fields Virology (Wolters Kluwer Health/Lippincott Williams & Wilkins). 3. Brockstedt, D.G., Podsakoff, G.M., Fong, L., Kurtzman, G., Mueller-Ruchholtz, W., and Engleman, E.G. (1999). Induction of immunity to antigens expressed by recombinant adeno-associated virus depends on the route of administration. Clin. Immunol. 92, 67–75. 4. Podsakoff, G., Wong, K.K., Jr., and Chatterjee, S. (1994). Efficient gene transfer into nondividing cells by adeno-associated virus-based vectors. J. Virol. 68, 5656–5666. 5. Mingozzi, F., and High, K.A. (2011). Therapeutic in vivo gene transfer for genetic disease using AAV: progress and challenges. Nat. Rev. Genet. 12, 341–355. 6. Carpentier, A.C., Frisch, F., Labbé, S.M., Gagnon, R., de Wal, J., Greentree, S., Petry, H., Twisk, J., Brisson, D., and Gaudet, D. (2012). Effect of alipogene tiparvovec (AAV1-LPL(S447X)) on postprandial chylomicron metabolism in lipoprotein lipase-deficient patients. J. Clin. Endocrinol. Metab. 97, 1635–1644. 7. Stroes, E.S., Nierman, M.C., Meulenberg, J.J., Franssen, R., Twisk, J., Henny, C.P., Maas, M.M., Zwinderman, A.H., Ross, C., Aronica, E., et al. (2008). Intramuscular administration of AAV1-lipoprotein lipase S447X lowers triglycerides in lipoprotein lipase-deficient patients. Arterioscler. Thromb. Vasc. Biol. 28, 2303–2304. 8. Wilson, J.M. (2015). A call to arms for improved vector analytics! Hum. Gene Ther. Methods 26, 1–2. 9. Gavin, D.K. (2015). FDA statement regarding the use of adeno-associated virus reference standard materials. Hum. Gene Ther. Methods 26, 3. 10. Ayuso, E. (2016). Manufacturing of recombinant adeno-associated viral vectors: new technologies are welcome. Mol. Ther. Methods Clin. Dev. 3, 15049. 11. U.S. Food and Drug Administration. (2008). Guidance for FDA reviewers and sponsors: content and review of chemistry, manufacturing, and control (CMC) information for human gene therapy investigational new drug applications (INDs). https:// www.fda.gov/BiologicsBloodVaccines/GuidanceComplianceRegulatoryInformation/ Guidances/CellularandGeneTherapy/ucm072587.htm.

16. Reinhard, L., Mayerhofer, H., Geerlof, A., Mueller-Dieckmann, J., and Weiss, M.S. (2013). Optimization of protein buffer cocktails using Thermofluor. Acta Crystallogr. Sect. F Struct. Biol. Cryst. Commun. 69, 209–214. 17. Seo, D.-H., Jung, J.-H., Kim, H.-Y., and Park, C.-S. (2014). Direct and simple detection of recombinant proteins from cell lysates using differential scanning fluorimetry. Anal. Biochem. 444, 75–80. 18. Rayaprolu, V., Kruse, S., Kant, R., Venkatakrishnan, B., Movahed, N., Brooke, D., Lins, B., Bennett, A., Potter, T., McKenna, R., et al. (2013). Comparative analysis of adeno-associated virus capsid stability and dynamics. J. Virol. 87, 13150–13160. 19. Zinn, E., Pacouret, S., Khaychuk, V., Turunen, H.T., Carvalho, L.S., Andres-Mateos, E., Shah, S., Shelke, R., Maurer, A.C., Plovie, E., et al. (2015). In silico reconstruction of the viral evolutionary lineage Yields a potent gene therapy vector. Cell Rep. 12, 1056–1068. 20. Carreira, A., Menéndez, M., Reguera, J., Almendral, J.M., and Mateu, M.G. (2004). In vitro disassembly of a parvovirus capsid and effect on capsid stability of heterologous peptide insertions in surface loops. J. Biol. Chem. 279, 6517–6525. 21. Boye, S.L., Bennett, A., Scalabrino, M.L., McCullough, K.T., Van Vliet, K., Choudhury, S., Ruan, Q., Peterson, J., Agbandje-McKenna, M., and Boye, S.E. (2016). The impact of heparan sulfate binding on transduction of retina by rAAV vectors. J. Virol. 90, 4215–4231. 22. Vedadi, M., Arrowsmith, C.H., Allali-Hassani, A., Senisterra, G., and Wasney, G.A. (2010). Biophysical characterization of recombinant proteins: a key to higher structural genomics success. J. Struct. Biol. 172, 107–119. 23. Ayuso, E., Blouin, V., Lock, M., McGorray, S., Leon, X., Alvira, M.R., Auricchio, A., Bucher, S., Chtarto, A., Clark, K.R., et al. (2014). Manufacturing and characterization of a recombinant adeno-associated virus type 8 reference standard material. Hum. Gene Ther. 25, 977–987. 24. Wright, J.F., Qu, G., Tang, C., and Sommer, J.M. (2003). Recombinant adeno-associated virus: formulation challenges and strategies for a gene therapy vector. Curr. Opin. Drug Discov. Devel. 6, 174–178. 25. Allay, J.A., Sleep, S., Long, S., Tillman, D.M., Clark, R., Carney, G., Fagone, P., McIntosh, J.H., Nienhuis, A.W., Davidoff, A.M., et al. (2011). Good manufacturing practice production of self-complementary serotype 8 adeno-associated viral vector for a hemophilia B clinical trial. Hum. Gene Ther. 22, 595–604. 26. Manno, C.S., Pierce, G.F., Arruda, V.R., Glader, B., Ragni, M., Rasko, J.J., Ozelo, M.C., Hoots, K., Blatt, P., Konkle, B., et al. (2006). Successful transduction of liver in hemophilia by AAV-Factor IX and limitations imposed by the host immune response. Nat. Med. 12, 342–347. 27. Maguire, A.M., Simonelli, F., Pierce, E.A., Pugh, E.N., Jr., Mingozzi, F., Bennicelli, J., Banfi, S., Marshall, K.A., Testa, F., Surace, E.M., et al. (2008). Safety and efficacy of gene transfer for Leber’s congenital amaurosis. N. Engl. J. Med. 358, 2240–2248. 28. Wu, Z., Asokan, A., Grieger, J.C., Govindasamy, L., Agbandje-McKenna, M., and Samulski, R.J. (2006). Single amino acid changes can influence titer, heparin binding, and tissue tropism in different adeno-associated virus serotypes. J. Virol. 80, 11393– 11397. 29. Vandenberghe, L.H., Breous, E., Nam, H.J., Gao, G., Xiao, R., Sandhu, A., Johnston, J., Debyser, Z., Agbandje-McKenna, M., and Wilson, J.M. (2009). Naturally occurring singleton residues in AAV capsid impact vector performance and illustrate structural constraints. Gene Ther. 16, 1416–1428.

12. Salganik, M., Venkatakrishnan, B., Bennett, A., Lins, B., Yarbrough, J., Muzyczka, N., Agbandje-McKenna, M., and McKenna, R. (2012). Evidence for pH-dependent protease activity in the adeno-associated virus capsid. J. Virol. 86, 11877–11885.

30. Xie, Q., Lerch, T.F., Meyer, N.L., and Chapman, M.S. (2011). Structure-function analysis of receptor-binding in adeno-associated virus serotype 6 (AAV-6). Virology 420, 10–19.

13. Van Vliet, K., Mohiuddin, Y., McClung, S., Blouin, V., Rolling, F., Moullier, P., Agbandje-McKenna, M., and Snyder, R.O. (2009). Adeno-associated virus capsid serotype identification: Analytical methods development and application. J. Virol. Methods 159, 167–177.

31. Vandenberghe, L.H., Wang, L., Somanathan, S., Zhi, Y., Figueredo, J., Calcedo, R., Sanmiguel, J., Desai, R.A., Chen, C.S., Johnston, J., et al. (2006). Heparin binding directs activation of T cells against adeno-associated virus serotype 2 capsid. Nat. Med. 12, 967–971.

Molecular Therapy Vol. 25 No 6 June 2017

11

Please cite this article in press as: Pacouret et al., AAV-ID: A Rapid and Robust Assay for Batch-to-Batch Consistency Evaluation of AAV Preparations, Molecular Therapy (2017), http://dx.doi.org/10.1016/j.ymthe.2017.04.001

Molecular Therapy

32. U.S. Food and Drug Administration. (2016). Code of Federal Regulations. Title 21 Sec. 610.14 Identity. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/ cfcfr/CFRSearch.cfm?fr=610.14.

39. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682.

33. Lock, M., McGorray, S., Auricchio, A., Ayuso, E., Beecham, E.J., Blouin-Tavel, V., Bosch, F., Bose, M., Byrne, B.J., Caton, T., et al. (2010). Characterization of a recombinant adeno-associated virus type 2 Reference Standard Material. Hum. Gene Ther. 21, 1273–1285.

40. Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov, I.N., and Bourne, P.E. (2000). The Protein Data Bank. Nucleic Acids Res. 28, 235–242.

34. Reguera, J., Carreira, A., Riolobos, L., Almendral, J.M., and Mateu, M.G. (2004). Role of interfacial amino acid residues in assembly, stability, and conformation of a spherical virus capsid. Proc. Natl. Acad. Sci. USA 101, 2724–2729. 35. Drouin, L.M., Lins, B., Janssen, M., Bennett, A., Chipman, P., McKenna, R., Chen, W., Muzyczka, N., Cardone, G., Baker, T.S., and Agbandje-McKenna, M. (2016). Cryoelectron microscopy reconstruction and stability studies of the wild type and the R432A variant of adeno-associated virus type 2 reveal that capsid structural stability is a major factor in genome packaging. J. Virol. 90, 8542–8551. 36. Horowitz, E.D., Rahman, K.S., Bower, B.D., Dismuke, D.J., Falvo, M.R., Griffith, J.D., Harvey, S.C., and Asokan, A. (2013). Biophysical and ultrastructural characterization of adeno-associated virus capsid uncoating and genome release. J. Virol. 87, 2994– 3002. 37. Nam, H.J., Gurda, B.L., McKenna, R., Potter, M., Byrne, B., Salganik, M., Muzyczka, N., and Agbandje-McKenna, M. (2011). Structural studies of adeno-associated virus serotype 8 capsid transitions associated with endosomal trafficking. J. Virol. 85, 11791–11799. 38. Lock, M., Alvira, M., Vandenberghe, L.H., Samanta, A., Toelen, J., Debyser, Z., and Wilson, J.M. (2010). Rapid, simple, and versatile manufacturing of recombinant adeno-associated viral vectors at scale. Hum. Gene Ther. 21, 1259–1271.

12

Molecular Therapy Vol. 25 No 6 June 2017

41. Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M., Meng, E.C., and Ferrin, T.E. (2004). UCSF Chimera–a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612. 42. Dolinsky, T.J., Nielsen, J.E., McCammon, J.A., and Baker, N.A. (2004). PDB2PQR: an automated pipeline for the setup of Poisson-Boltzmann electrostatics calculations. Nucleic Acids Res. 32, W665–W667. 43. Rostkowski, M., Olsson, M.H., Søndergaard, C.R., and Jensen, J.H. (2011). Graphical analysis of pH-dependent properties of proteins predicted using PROPKA. BMC Struct. Biol. 11, 6. 44. Konecny, R., Trylska, J., Tama, F., Zhang, D., Baker, N.A., Brooks, C.L., 3rd, and McCammon, J.A. (2006). Electrostatic properties of cowpea chlorotic mottle virus and cucumber mosaic virus capsids. Biopolymers 82, 106–120. 45. Baker, N.A., Sept, D., Joseph, S., Holst, M.J., and McCammon, J.A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl. Acad. Sci. USA 98, 10037–10041. 46. Xie, Q., Bu, W., Bhatia, S., Hare, J., Somasundaram, T., Azzi, A., and Chapman, M.S. (2002). The atomic structure of adeno-associated virus (AAV-2), a vector for human gene therapy. Proc. Natl. Acad. Sci. USA 99, 10405–10410.