Fast and efficient size exclusion chromatography of adeno associated viral vectors with 2.5 micrometer particle low adsorption columns

Highlights

• Prototype 2.5 µm particle column enabled high throughput AAV SEC separations.
• Fine particle packing materials yielded efficient separations without sample stress.
• Use of low adsorption hardware allowed the development of robust platform methods.
• Different format columns provided reproducible results with multi attribute MALS.

Abstract

More and more transformative gene therapies (GTx) are reaching commercialization stage and many of them use Adeno Associated Viruses (AAVs) as their vector. Being larger than therapeutic antibodies, their size variant analysis poses an analytical challenge that must be addressed to speed up the development processes. Size exclusion chromatography (SEC) can provide critical information on the quality and purity of the product, but its full potential is not yet utilized by currently applied columns that are (i) packed with relatively large particles, (ii) prepared exclusively in large formats and (iii) built using metal hardware that is prone to secondary interactions. In this paper, we investigate the use of state-of-the-art sub-3 µm particles to address existing limitations. A prototype 2.5 µm column was found to deliver superior kinetic efficiency, significant reduction in run times and increased resolution of separations. No evidence for shear or sample sieving effects were found during comparisons with conventional 5 µm columns. Moreover, use of low adsorption hardware enabled the application of a wide range of mobile phase conditions and a chance to apply a more robust platform method for several AAV serotypes. The resulting method was tested for its reproducibility as well as utility for critical quality attribute assays, including multiangle light scattering based (MALS) measurements of size and molar mass. Thus, a new tool for higher resolution, higher throughput size variant analysis of AAVs has been described.

Keywords: Adeno associated virus, AAVs, Size exclusion chromatography, Size variants analysis, Gene therapy, SEC

1. Introduction

Adeno-associated viruses (AAVs) are one of the leading vehicles for an in vivo gene delivery and are commonly used in many established and emerging therapies [1,2]. They owe their popularity as genetic information vehicles to their increasingly verified clinical potential, relatively low toxicity, tissue tropism and prolonged gene expression without permanent DNA integration [3]. Nevertheless, despite these favorable characteristics, only 5 AAV based medicines have reached the US market so far [4], with factors such as i) cost, ii) dose dependent safety and iii) manufacturing challenges cited as the major hurdles [5]. Some of these aspects are being addressed by progress in viral engineering and an improved understanding of the underlying biology [6,7]. Ultimately though, it has been realized that purification, characterization, and quality control techniques will also need to be improved [8].

Critical quality attributes (CQAs) define a drug in terms of potency, purity, and safety and thus their proper identification and rigorous measurement is essential for successful commercialization. Several CQAs have been proposed for AAV products [9]. Among them, it is (i) virus titer, (ii) aggregation of capsids and (iii) content ratio that are understood to be most important, because the number of viral particles directly affects potency, while oligomeric aggregates are a known risk factor for biopharmaceuticals [10]. Moreover, the ratio of full to empty viral particles is coming to be seen as an important parameter that influences dosing and safety/efficacy profiles [11]. It has been recognized that currently employed analytical techniques for measuring these AAV specific CQAs do not provide sufficient throughput [12] which can impede both process development, formulation development and clinical trials.

Further complications arise from there being only a handful of methods that can report on more than one attribute during a single measurement. For example, analytical ultracentrifugation (AUC) can reveal content ratio and aggregation profiles while optical density measurements can yield the empty/full ratio and virus titer. However, only a multidetector equipped sizing separation is able to inform on all 3 discussed CQAs [13]. Field flow fractionation can provide the sizing separation including an analysis of submicron particles, but it suffers from deployability challenges and low resolution on oligomeric aggregates and low molecular weight impurities [14]. Size exclusion chromatography can instead be applied to address these latter two points. From this point of view, a potential solution to the analytical bottleneck would be the wider adoption of SEC methods, which are currently stymied by a lack of validated methods and suitable materials [15].

An AAV capsid is a stochastic ensemble of 3 viral proteins folded together to form an icosahedral structure of around 20–25 nm diameter and 3.5–6 MDa in size [16]. Species of similar characteristics, notably other viruses and virus-like particles [17,18,19], have been successfully analyzed before with SEC, but reports on AAV separations are more scarce and have been predominantly focused on its application to single CQA determination such as virus titer measurements [20], and content ratio determination [21]. Importantly, in those reports method development considerations are largely omitted. Testing has primarily been limited to the selection of particles with suitable average pore diameters and pore size distributions. Other literature often simply lists a set of final conditions without revealing the experimental details on the conditions explored during screening experiments [22], [23], [24].

In this work, we investigate key SEC method development parameters using new low-adsorption column hardware packed with 2.5 µm diol bonded ethylene bridged hybrid (BEH) particles with an aim to achieve a reliable and fast analysis of aggregation profiles. The use of hydrophilically modified hybrid organic−inorganic column hardware (based on BEH-like siloxane chemistry - hydrophilic hybrid surface technology (h-HST)) has already been demonstrated to significantly reduce unwanted non-specific interactions involving antibody therapeutics [25]. In our current work, the benefit is clearly demonstrated to extend to AAVs.

This paper also constitutes the first demonstration of AAVs being analyzed with kinetically advantaged 2.5 µm particles. Our measurement of aggregates is consistent with results obtained from columns packed with larger, 5 µm silica particles. We additionally explored possible shear and sample sieving effects that may hypothetically arise due to the use of a smaller particles and higher pressures. No such phenomena were observed for oligomers that are measurable via SEC [26]. On the whole, our main focus was to reduce run times while maintaining high separation efficiency. Finally, we demonstrated the utility of the new SEC methods for MALS measurements of molar mass and the analysis of highly aggregated samples produced by forced degradation experiments. To the best of our knowledge, many of the insights uncovered in this investigation have yet not been made publicly available and may therefore be of high interest to the gene therapy community.

2. Material and methods

Aliquots of AAV samples (serotype 2, 5 or 9 (empty pI ≈ 6.0–6.2, MW ≈ 3.6 - 3.8 MDa), [27], 100 µL, 1 × 10¹³ vg/mL, CMV-GFP (when filled capsid pI lowered by about 0.2 – 0.4, MW ≈ 0.78 MDa) or Empty, Virovek) [28] were thawed at room temperature for <2 h and transferred to a low adsorption vial (186009186, Waters Corporation, Milford, MA) which was placed in the autosampler cooled to 6 °C. Such a sample was analyzed without further manipulations. The leftover samples were discarded after >120 h of storage. The performance of the columns was periodically monitored with protein ladder standard (BEH450 SEC Protein Standard Mix, 186006842, Waters).

2.1. Particle characterization

SEM micrographs were taken at 5 kV using a Hitachi SU3900 scanning electron microscope. Particle samples were spread on adhesive carbon tape purchased from Oxford Instrument. SRT SEC-500 and SEC-100 columns were acquired from Sepax (Sepax Technologies, Newark, Delaware). Particle batches obtained for physicochemical characterization were different than those applied in chromatographic testing. BEH SEC Packing Materials were prepared from organic/inorganic hybrid particles with an empirical formula SiO₂(O_1.5SiCH₂CH₂SiO_1.5)_0.25 according to previously reported procedures. The selected batch had an average particle diameter of 2.05 µm, average pore diameter of 444 Å, surface area of 80 m²/g, pore volume of 1.17 cm³/g and surface coverage of 4.64 μmol/m². Pore size and pore volume were measured using an AutoPore™ V9600 porosimeter machine by MicroMeritics (Norcross, GA). Particle size was measured with a Beckman Coulter Multizier 4e (Brea, CA) and the average particle diameter corresponding to 50% volume distribution is reported. Surface area was measured by nitrogen adsorption using a MicroActive instrument (MicroMeritics, Norcross, GA), and percent carbon was measured by a commonly applied combustion method performed on a TruMac instrument (Leco corp, St Joseph, MI). Coverage was thereby calculated according to elemental compositions of the applied silane reagent and carbon percentage found in the particles. Columns were packed to ensure they would be mechanically stable up to the linear velocity of running 0.6 mL/min flow rates on a 4.6 mm ID column.

2.2. UHPLC-SEC-UV-Fluorescence

All experiments were performed with an ACQUITY™ UPLC™ H-Class Bio QSM System equipped with the following in-series detectors: TUV detector (Titanium 5 nm flow cell, Waters) and FLD detector. AAV samples were characterized with fluorescence detection using 280 nm for excitation and 350 nm for detection with a 2 points/s recording rate. The study entailed the evaluation of 2.5 µm diol bonded BEH 450 Å columns constructed with 4.6 × 150 mm h-HST hardware (XBridge™ Premier GTx BEH SEC 450 Å 2.5 µm Columns, 186010584, Waters Corporation, Milford, MA). Additional measurements were also made on prototype columns packed with larger diameter particles (3.5 µm diol bonded BEH 450 Å Column, 4.6 × 150 mm, h-HST hardware) as well as stainless steel hardware versions of the column (2.5 µm BEH 450 Å, 4.6 × 150 mm Column, stainless steel hardware). Reference measurements were made using a 5 µm silica 500 Å, 4.6 × 150 mm, PEEK lined hardware column (Sepax Technologies, Newark, Delaware, US). This column contains particles that are modified with a proprietary hydrophilic modification.

The chemicals and solvents used in the study were used as received (Sigma-Aldrich, Fisher Chemical). Standard mobile phase was comprised of 10 mM K₂HPO₄ adjusted with 6 M HCl to pH 7.4 and added 200 mM of KCl as well as 0.02% NaN₃ as a bacteriostatic, which was run in a 11 min method at 0.25 mL/min and ambient temperature using 1 µL as a standard injection volume. In the case where baseline separation between the aggregates and the monomer was not achieved the overall level of HMWS was approximated by integration until the valley preceding the monomer peak. Water was used a blank injection every 4 AAVs injections but only negligible carry-over (<0.1%) was detected using the abovementioned conditions. All mobile phases were filtered with 0.2 µm membranes (PES, 5660020, Nalgene) before use and were replaced within 48 h of their preparation. No sinkers were used at the end of lines and the instrument was periodically flushed with 70% isopropanol solution to prevent microbial contamination. We found that SEC AAV analysis is extremely sensitive to microbial contamination – even slight bacterial growth led to worsening of the analysis efficiency, problems with reproducibility and increased operated pressure. Therefore, it is of utmost importance to abide by good laboratory practices recommended for chromatography using aqueous buffers [29].

Thermal aggregation experiments were performed directly in the low adsorption vials that were incubated at a given temperature without agitation and periodically moved into the autosampler for SEC analysis.

The kinetic performance limit graph was built based on a variable flow rate experiment and porosity determination. Briefly, the AAV2 analysis was performed at 50, 250 and 600 µL/min flow rate for all of the tested columns (2.5 µm BEH 450 Å Column, 3.5 µm BEH 450 Å Column and 5 µm silica 500 Å conventional column) and the monomer plate number was determined via the Empower™ Software (Waters Corporation) built-in integration tool. The operating pressure of the particular analysis was used to determine the permeability of the columns while the porosity was measured by establishing the elution time of uracil. The model assumed 276 bar as the pressure limit and viscosity of the buffer as 1.0169 cP. The monomer plate height was corrected for extra column volume assuming 9 uL² as the system variance. The details of the calculations can be found in Ref [30].

To determine the limit of detection and quantification (LOD and LOQ), a calibration curve was constructed using a standard AAV5 sample and 5 levels (10X, 7X, 4X, 2X, 1X). Each sample was injected 5 times and linear regression (fixed intercept as 0) of averaged fluorescence peak height afforded fitted line slope (S). In a separate set of injections, baseline noise (n) of blanks at the elution time range of the peak of interest (1 min interval at the elution time of the monomer peak) was calculated with Empower Software. LOD was then estimated as 3 · n/S and 10 · n/S was considered for the LOQ.

Increased pressure experiments were performed by the addition of post column low diameter tubing (PEEKSIL, 50 µm x 600 mm) via zero-volume connectors, with the increase of pressure experimentally verified via the UPLC instrument reading.

2.3. UHPLC-SEC-MALS

SEC-MALS experiments were performed using an ACQUITY UPLC H-Class Bio QSM System connected in line with TUV detector (Titanium Flow Cell, 5 mm, 1500 nL), Wyatt DAWN™ Instrument with a Wyatt QELS™ Dynamic Light Scattering (DLS) Module and RI detector. AAV samples were characterized by UV absorbance at 280 nm, light scattering at 18-angles with Wyatt QELS (quasi-elastic light scattering) DLS Module in a dynamic fashion at 659 nm wavelength and refractive index properties through data analysis using ASTRA™ 8 Software. This three-pronged multidetector approach provides detailed measurements of AAV whose component species scatter light depending on their size and concentration in solution. Size measurements include radius of gyration (Rg also called root mean square radius) and hydrodynamic radius (Rh). Angular (up to 18 angles) dependence of scattered light intensity determines the Rg value, and fluctuations of light scattering intensity due to diffusion of molecules (Brownian motion) determines the Rh value. Leveraging the combined UV, RI, MALS, DLS data, Wyatt ASTRA Software determines the physical attributes of an analyte, including its particle size, molar mass, and compositions (such as Vg/Cp for AAV) [31]. The study entailed the evaluation of 2.5 µm diol bonded BEH 450 Å Columns constructed with 4.6 × 300 mm and 7.8 × 300 mm h-HST hardware (XBridge Premier GTx BEH SEC 450 Å 2.5 µm Columns, Waters Corporation, Milford, MA) using AAV2 (Virovek). Prior to use, AAV2 samples (stored at −80 °C) were thawed to room temperature, mixed gently with a pipette, and centrifuged using a Benchmark Scientific Mini Centrifuge for 1 min at 15,000 rpm. SEC was performed at 0.2 mL/min flow rate using 2x strength PBS mobile phase (20 mM Phosphate, 276 mM NaCl, 5.4 mM KCl pH 7.4) and a 30 °C column temperature.

2.4. Analytical ultracentrifugation

Analytical Ultracentrifugation-Sedimentation velocity (AUC-SV) experiments were conducted on a Beckman Coulter ProteomeLab XL-I instrument (Beckman Coulter, Indianapolis, IN, USA). Absorbance optics were employed, and cells equipped with two-sector Titanium centerpieces and Sapphire windows (Nanolytics, Germany) were utilized. Immediately prior to AUC-SV analysis, the sample was thawed and its absorbance was evaluated by Nanodrop UV–Vis and a 12 mm centerpiece was found appropriate for a wavelength of 280 nm. In a typical experiment, 395 μl of the sample and 400 μL of buffer were loaded in the sample and reference sector, respectively. A 4-hole rotor AN60ti was used for analysis of one sample at a time. After the cell and the rotor were mounted, the sample was inspected briefly at the low centrifugation speed, 1500 rpm, and followed by the rotor being set at rest at the pre-set temperature of 20 °C for an equilibration period of 2 hours. SV data were collected primarily at 20 000 rpm and 20 °C. Additionally, lower speed runs were also performed. The absorbance scans were collected continuously at the radial resolution of 0.003 cm in the radial range spanning from air/liquid interface to the bottom of the cell. The AUC-SV data were analyzed by SEDFIT version 16.36 (NIH, USA) software [32]. The analysis involved fitting the data to the continuous C(s) distribution of the program, leaving the following parameters float: the meniscus, time-invariant noise, radial-invariant noise, and the frictional ratio. Typical values of sedimentation coefficient (s) ranged from 1 to 300 S with a resolution of 250, and a regularization by 2nd derivative at a level of 0.68 was applied. Partial specific volume of 0.68, the density and viscosity of the buffer (measured by Anton Paar DMA 4500 equipped with Lovis 200ME) were kept constant, and these settings were maintained the same for all samples.

...