Developing A Scalable Process For Adenovirus Manufacturing
By Mats Lundgren, Customer Applications Director, Bioprocess, GE Healthcare Life Sciences
Market analysis done by GE Healthcare in 2016 showed that viral vectors are being utilized in a number of different ways. The most commonly used viral vector at that time was the adenovirus, although application of the lentivirus was rising due to its use in cell therapies, where viral vectors are predominantly used to transduce the cells and reinfuse the modified cells into the patient. When looking at other areas of clinical trial application, viral vectors are becoming a valuable tool in the promising area of oncolytics, such as with Amgen’s IMLYGIC®, as well as with the use of adeno-associated virus (AAV) in gene therapy.
As a growing number of approvals inspire more companies to explore and invest in this market, increasing speed and efficiency in the current methods of viral vector manufacturing is critical. Scalability, in particular, becomes a focus as the demand for viral vector supply multiplies. To address this, a team of process development and analytical scientists at GE focused on developing a process for adenovirus manufacturing that would not only fulfill all regulatory requirements but also be inherently scalable up to 2,000 liters in their single-use stirred-tank bioreactors.
Upstream Cell Culture And Virus Propagation
For this project, the recombinant adenovirus serotype 5 encoding green fluorescent protein (GFP) was used as a model virus. Human embryonic kidney cells (HEK293) were selected for virus production. HEK293 cells were suspension adapted in cell culture media from various vendors. The adenovirus was then propagated in cells grown in the different media and virus productivity was subsequently analyzed.
The cell culture medium for HEK293, which is a chemically defined medium, was compared to another commonly used competitor medium (Figure 1). It was determined the HyCloneTM CDM4HEK293 medium could support higher cell density and provide 10 times more virus as compared to the reference medium. This was true in both multiplicities of infection. While both medias provide acceptable cell growth, the difference is evidenced in the virus productivity.
The team optimized the multiplicity of infection (between the range of 0.01 and 10), the time of infection, and time of harvest in shake flasks using different cell densities at the time of infection. The effect of the media exchange before virus infection was also evaluated. For the time of harvest, different time points between 36 and 72 hours were examined.
The next step was to scale up the process using a bioreactor, as all work done up to this point was performed in shake flasks. Figure 2 below shows the virus productivity after three different runs in GE’s single-use, stirred-tank bioreactor, the Xcellerex™ XDR-10, compared to the shake flask control. The cells were expanded for three days and the media dilution was completed on day three. The cells were infected at that point and then harvested 42 hours later.
The infectious virus particles were consistent between each run. The total virus particles are shown in green, whereas the blue bars represent infectious virus particles. The difference between the blue and green bars is due to the intrinsic property of the adenovirus biology that many virus particles are not fully infectious. A similar process was also created for GE’s rocking bioreactor, the ReadyToProcessTM WAVE 25; while the system is robust and easy to use, it is designed for volumes below 50 liters. However, the XDR bioreactors are scalable up to 2,000 liters.
Because the adenovirus is a non-lytic virus, the downstream process had to begin with cell lysis followed by DNA fragmentation with DNase enzyme. There are several ways lysis can be done, but the most common and scalable method is by using detergents. After screening detergents for an effective and harmless option, the Tween 20 detergent was selected. Clarification was done using normal flow filtration, and then concentration and buffer exchange were completed using tangential flow filtration with hollow fiber filters. Depending on the virus size, different filter cutoffs can be used. In this case, a relatively low cutoff was used to ensure good recovery of virus and the remaining impurities were shown to be efficiently removed in the downstream process steps to follow.
Two approaches were used for the capture step. One was to use the Capto™ Q ImpRes chromatography resin with gradient elution. The advantage of this relatively new chromatography resin is that the beads are small, so there is a large surface area available for virus binding, improving the capacity. Even though the beads are small, the virus can be bound only to the resin surface (due to its size). The other alternative was to use ReadyToProcess Adsorber Q membrane with step elution, which also offers certain advantages, such as high binding capacity because the virus can reach a high proportion of the membrane surface and bind to the ligands. However, it was difficult to separate the virus from the DNA and other impurities using the ReadyToProcess Adsorber Q membrane, so Capto Q ImpRes with optimized conditions was ultimately selected for efficient DNA and host cell protein (HCP) removal. Figure 3 shows a multiplex fluorescence SDS-PAGE and Western blot of how the impurities were removed using Capto Q ImpRes. The green bands represent the viral proteins detected with a Cy3-labeled antibody, and the red bands represent labeled protein impurities from the host cells (samples Cy5 pre-labeled before SDS-PAGE). The impurities are successfully removed as the purification process continues from lane 1 (starting material) to lane 8 through 10 (virus particle fractions). The green bands in lanes 2 and 3 are free viral proteins.
Different options could be considered for the polishing steps. One is the traditional approach of size exclusion, which is suitable for separating virus from impurities, but the disadvantage is it is difficult to scale up due to limitations in load capacity, typically resulting in large columns. Instead, GE has developed the Capto Core 700 resin, which has an inert porous surface and a strong multimodal octamylamine ligand inside that binds impurities with both hydrophobic and ionic interactions. Consequently, viruses that are larger than the pore size of the resin will pass on the outside of the resin bead, whereas smaller impurities will be trapped in the resin. When polishing with Capto Core 700 and GE’s size exclusion resin SepharoseTM 4 Fast Flow were compared, the impurity reduction was similar. However, Capto Core 700 enables higher sample-load-volume capacity (0.1- to 0.2-column volume versus up to 30-column volume) and higher total recovery of particles, making it a better resin for scale-up.
Three downstream processes were run as described above but different capture and polishing combinations were evaluated in larger scale (3L) and run in duplicate; the best process was then run in a 10-liter scale (Figure 4).
The results of the final bulk analysis between ReadyToProcess Adsorber Q (membrane process) and the novel resin process developed by GE, as well as an older published process (reference process), are shown in the table below.
The average recoveries of the total virus particles (vp) and infectious virus particles (ivp) were similar for all three process variants in duplicate runs in 3-liter scale or triplicate runs (one additional run with novel resin process in 10-liter scale). The quantitative polymerase-chain-reaction (qPCR) analysis showed low variation for triplicates of the same sample but sometimes varied between analysis occasions of the same sample. Total protein and host cell proteins (HCP) levels were lower for the novel resin process compared to the membrane process using ReadyToProcess Adsorber Q. LOD for the HCP analysis was 1 ng/ml. Genomic DNA (gDNA) was below the LOD (1 ng/mL) for both processes. The dose size assumption was set to 1011 virus particles, and the regulatory demands were reached for all processes, even if reference and the membrane process were slightly higher on total protein. The ratio of total virus and infectious virus titer was under 30, as required by regulatory agencies.
A size exclusion, high-performance liquid chromatography (SEC HPLC) analysis of the starting material and final sample from the new process shows enrichment of the virus and the near complete removal of the impurities (Figure 5). The final bulk product is represented with orange, and blue represents the starting material.
The final purified samples from the process variants were sent to Vironova, a biotech company with expertise in transmission electron microscopy (TEM) analysis, as TEM can identify particulate impurities that protein and DNA assays cannot. Figure 6 shows the TEM analysis for samples purified with Q Sepharose XL and Sepharose 4 Fast Flow on the left (reference process) and with the Capto Q ImpRes in combination with Capto Core 700 on the right (novel resin process).
The results clearly show that there were many more cell debris particles in the samples purified with the reference process. These types of impurities can potentially interfere with the infectivity of adenovirus and are difficult to detect with other analytical assays. The low level of debris after the novel resin process could be due to the optimized gradient elution and the Capto Core 700 polishing step that were more efficient in removing this type of debris.
Virus Titer Assay Development
First, the GE team wanted to develop an alternative to the industry standard TCID50 method, as TCID50 for adenovirus sometimes takes two weeks to perform. A technology was developed based on the IN Cell Analyzer, an automated fluorescence microscopy instrument. In this assay, the virus was serially diluted in a plate with the indicator cells and then the fluorescence was imaged after 42 hours. If you are not using green fluorescent protein (GFP), this can be done by staining with antibodies against viral antigens. Then the infectious virus titer can be calculated. This technology is better in terms of accuracy and reduced time for virus infectious titer determination.
Another titer assay was developed to measure the number of total virus particles, which was based on a BiacoreTM T200 instrument. In this case, recombinant Coxsackie adenovirus receptors (CAR) and Factor X (FX) were used for the detection of virus. These proteins bind the virus in different positions, thereby complementing each other in terms of analytical resolution. By using Biacore, the sample can be passed over the sensor chip to detect the interaction between the virus and the chip (Figure 7).
A comparison of the total virus particles per milliliter achieved by Biacore CAR assay, Biacore FX assay, and the standard technology, qPCR detecting viral DNA, is shown in Figure 8 below.
The disadvantage with the qPCR process is that it takes more time and requires preparation of the DNA before the assay. Biacore assay (CAR and FX), however, offers similar results with a higher level of reproducibility and convenience. The infectious virus particle concentration was measured with the InCell assay. This assay correlates very well with the TCID50 assay (data not shown).
Process Economy Evaluation
For a new manufacturing method to be successful, it must be able to show a payable process economy. In a process economy simulation, the cost per batch was compared between the novel resin process and the reference process, and a sensitivity analysis was performed, which identified yield as a major influence on cost per dose. Since there were issues with yield determination, the team decided to fix the yield of all processes at 48 percent, meaning that each process produced the same number of doses per batch in any given scenario. To make an appropriate comparison between single-use technology (SUT) and stainless steel, the influence of campaign changeover activities like cleaning verification and column packing had to be taken into account.
Figure 9 shows an example of a 12-month campaign that consisted of two bioreactor scales, with the start and end assigned with different amounts of labor for stainless-steel and single-use. All resins, ReadyToProcess-columns, and filters were assumed to be replaced at campaign changeover. In a cost per batch comparison, the new process scales well and has better process economy than the reference process. This is mainly due to higher capacity in Capto Core 700 (15 CV used in this calculation) compared to Sepharose 4 Fast Flow (0,15 CV used in this calculation) and lower labor costs for single-use technologies.
In the hybrid scenario, where stainless-steel process equipment was combined with single-use buffer/media prep and hold, there was a lower batch cost than with single-use in the single-product scenario, but not in the multi-product scenario. This was due to the impact of the campaign changeover activities, as well as the increasing importance of capital costs, as fewer batches are run per year.
The adenovirus process developed here is based on single-use and scalable technologies, such as bioreactors, filtration, and chromatography. The process is cost-efficient and leads to high yields and purity. Since modern cell culture media and purification technologies were used, it fulfills regulatory requirements, which is essential. All of the process technologies used are compatible with large-scale GMP production and can be implemented in flexible manufacturing facilities (e.g., GE’s FlexFactoryTM).
Dr. Mats Lundgren, Ph.D. has more than 25 years of experience in the field of biotechnology. He holds a PhD in Immunology, Cell and Molecular Biology from the Karolinska Institute, Stockholm, Sweden and extensive post-doc training at the MRC Clinical Sciences Centre, Imperial College School of Medicine in London, UK. In his industrial career Mats has had positions as scientist, team manager and VP at Pharmacia, AstraZeneca and smaller biotech companies. In his previous function, Mats was managing both the Cell line and Upstream Process Development teams at a major biotech company. In his current role, Mats works across different viral vector and vaccine application projects as well as general upstream topics, focusing on customer support, applicability of new technologies and manufacturing solutions.
This article was created from the author's presentation at the 2018 Bioprocessing Asia conference, with GE Healthcare as principal sponsor. The BioProcessing Asia Conference series was created to provide a platform to advance the contribution of bioprocessing sciences towards the development and manufacture of affordable biopharmaceutical products in Asia.