In the last decade, adenovirus vectors (AdV) have been the vectors of choice for transferring corrective genes into human cells for transient expression. Up to 600 gene therapy clinical protocols have been reported 1, 28% of which used AdV to deliver therapeutic or marker genes. One of the main advantages of AdV is their ability to replicate at high titers in complementing cell lines. Important developments in molecular medicine and adenovirus vector design have been achieved and have been the subject of excellent reviews 2–7.
The availability of large quantities of clinical-grade AdV is recognized as an important limitation to in vitro experimentations, as well as pre-clinical and clinical studies. It is our aim to review the problems and strategies involved in AdV production and provide an overall view of the technology as well as future trends.
The first objective for developing an efficient, reproducible and scalable AdV production technology is to maximize the specific vector yield on a per cell basis. As a theoretical target, one might consider the specific yield achieved by replicating a wild-type adenovirus on a permissive cell line of interest. Specific titers as high as 30 000 infectious viral particles (IVP) per cell have been reported 8. This being established, increasing cell density for vector production, while maintaining maximal specific production, will truly translate into improving volumetric production. Similarly, increasing production scale should ideally be completed while maintaining maximal specific production and volumetric production identical to what has been achieved at small scales. Then, the theoretical output of the process in terms of mass production of adenovirus vector would simply be obtained by multiplying the specific production by the cell density, then by the total volume. With the current yields achieved 9, hundreds of liters of culture might be necessary for the HIV-1 prophylactic prevention and therapeutic treatment program 10.
In order to quantitate AdV yield and bioactivity, two main parameters are defined: (1) viral particle count (VP) determined by a direct physicochemical method such as UV spectroscopy; and (2) infectious viral particle count (IVP) indirectly determined following infection of a permissive cell line. Figure 1 summarizes the basic principles of replication for the first-generation adenoviral vector. When the E1 protein essential for replication is provided by a production cell line such as HEK-293, replication of viral DNA, encapsidation, maturation and expression of viral particles can be completed. This translates in terms of batch process in cell infection, cell growth cessation, virus particles accumulation in the cells as early as 24 hours post-infection (hpi) with a maximum titer around 40 hpi and cell death thereafter. The viral replication factor of the process based on VP quantitation might range from 103 to 104.
The second objective of successful scale-up and optimization of an AdV production strategy is to maintain the specified quality requirements of the product. For replication-defective recombinant adenovirus vector, this translates to efficient and safe delivery of the therapeutic genes to the nucleus of the target cells without replication of the viral genome. However, there are cancer therapeutic strategies for which the adenovirus replication is desirable. In these approaches, different replication-competent adenovirus vector designs are used 7. AdV are ultimately intended for use in human gene therapy, thus any solution proposed by a process developer should be compatible with good manufacturing practices. As for many new biological therapeutics, product identity relies on a set of specifications because complete physicochemical characterization is not yet possible. Therefore, reproducibility of the process is a key element to guarantee, among other aspects, the identity of the manufactured biological. As a consequence, demonstration of reproducibility needs to be extended to all steps of the process.
We believe that a rational approach to developing a successful recombinant AdV process requires a detailed understanding of basic host cell line physiology and metabolism; vector engineering; mechanism and kinetics of interaction between host cell line and the vector. Identification, quantitation and control of key process parameters during production and purification steps using reliable on-line and off-line process monitoring tools are also essential requirements.
In this report, the different steps of AdV process development are explored following these guiding principles. A predominant set of data, used in this paper to exemplify the approaches, has been generated in the authors' laboratory 11–20 using derived clones from HEK-293 cells 21. However, whenever possible, these data are discussed in light of conclusions from published papers in the field.
Adenovirus vector quantitation
A critical aspect often underestimated in the early stages of developing a new production process is quantitation. This is even more critical in the case of viral vectors for gene therapy. To help worldwide standardization of quantitation methods and facilitate interpretations of pre-clinical and clinical data, a reference material for adenovirus vector has been generated 25.
Primary characterization of AdV lots requires definition of viral particle units (VP) and infectious particle units (IVP). It is essential to have these tools available at the very early stage of process development. The most common method for physical determination of total purified viral particles in an SDS-treated preparation relies on absorbance reading at 260 nm which is converted to total particle number using published extinction coefficients 26. The reproducibility and precision of total particle quantitation are significantly enhanced by using anion-exchange high-performance liquid chromatography (HPLC), prior to detection by UV spectroscopy 22. Moreover, use of this method has been extended to cell lysate samples during the course of a production 15, 19. Viral particle concentrations reported in this paper were measured using an HPLC system consisting of a 600E multisolvent delivery system, a 717 Plus autosampler cooled to 4 °C, a 996 photodiode array detector (PDA), and a UNO Q 0.46 cm i.d. × 1 cm long polishing column. Other methods for total particles quantitation are based on quantitation of virus DNA following fluorescent labeling 27 or viral proteins such as hexons 28.
The determination of infectious particles relies on direct or indirect observations of cell infection signposts following contact of virus particles with permissive cells. The result of the infection process such as cytopathogenicity, expression of a marker gene such as β-galactosidase or the green fluorescent protein (GFP) will then be considered to infer the number of infectious viral particles (IVP), titer, plaque forming units (PFU) or GTU (gene transfer units). As a consequence, the precisions of the titers derived from these assays strongly depend on the biological material and operating conditions 29. Generally, standard plaque assay methods underestimate titers because of diffusion limitations 30. Other assays have been developed to measure subcomponents of the virus such as the DNA binding protein, an early protein encoded by the vector that is detected by quantitative immunofluorescence 31. The infectious virus titration method used in this study has been described in detail elsewhere 11. Briefly, 1 mL of fresh 293S cells are seeded at 1 × 106 cells per well in 6-well plates. A fixed volume of sample supernatant is diluted 1 : 100 in fresh DMEM medium, and different volumes are added to each well. The dilution factor of the sample is chosen in order to obtain a final ratio of GFP-positive cells to viable cells (GPR) between 0.1 and 0.4. After 20 h of incubation, the GFP-positive cells are scored using a flow cytofluorometer equipped with a 15-mW argon-ion laser and the following filters: 488-nm laser blocking, 488-nm long pass dichroic, 550-nm long pass dichroic and 525-nm band-pass.
Viral particle units and infectious units are two critical parameters that allow for comparison of different production processes 9. However, one needs to keep in mind that these values strongly depend on the type of viral construct, the transgene expressed and host cells used for virus replication. Moreover, these values are derived from different types of quantitation assays, as well as different operating conditions for the same assays. Reported values generally range from 104 to 105 adenovirus particle units per cell, while infectious units are roughly lower by a factor of ten. In order to properly assess process performances, it is advisable to compare yields based on specific expression, i.e. on a per cell basis. An example of adenovirus quantitation using both parameters is shown in Figure 2. Two adenovirus constructs expressing GFPs65t (GFP) and GFPsg25 (GFP-Q) mutant under the control of the CMV promoter were compared. Figure 2 shows a two- to threefold lower specific infectious titer (IVP) as well as a lower total viral particle concentration (VP) for similar constructs expressing GFPsg25 compared with GFPs65t. This result is often referred to as the ‘transgene effect’ in conference reports; however, the reason behind this effect is not clear. Therefore, one should be careful about direct extrapolation or comparison of titers for different vectors.
Replication competent adenovirus (RCA) determinations are generally performed on pre-clinical and clinical lots by observation of the cytopathic effect (CPE) on permissive cells such as Hela and A-549 cells after multiple dilutions 32, 33. Using PCR can enhance the sensitivity of the method. The emergence of RCAs is cell line and viral construct dependent. For many therapeutic applications, generation of RCAs remains a safety concern for production of clinical lots. Characterization and quality control of recombinant adenovirus vectors for gene therapy have been recently reviewed 34. A large spectrum of techniques, involving electron microscopy, mass spectrometry, genome characterization, sequencing and a set of bioassays, are used in order to establish the identity, bioactivity and purity of the viral vector lots.
The ‘cell density effect’ and production yield
Commercial-scale cultivation of mammalian cells has been accomplished using different technologies: roller bottles, cell cube, hollow fibers, microcarriers and suspension culture bioreactors (batch, fed-batch or perfusion mode). However, for products needed in large amounts, suspension culture is the most effective method over processes using adherent cells. Suspension culture is characterized by a homogeneous concentration of cells, nutrients, metabolites and products, thereby facilitating scale-up and enabling accurate monitoring and control of the culture. This observation was the basis of our decision to invest efforts in adapting the HEK-293 cell line to suspension culture (293S). They were further adapted to serum-free medium 293 SF 11 and a subclone 293 SF-3F6 was obtained after two serial endpoint dilutions, as described elsewhere 12. Serum-free clones were cultured in low-calcium serum-free medium (LC-SFM) or a low protein proprietary formulation NSFM13 17. For production of biologicals, we strongly advocate the consideration, at the very early stage of process development, of transformation of cell lines that are readily adapted to suspension and serum-free medium to maximize the potential for scalability.
Maximal viable cell densities as high as 5 × 106 cells/mL are routinely achieved in suspension bioreactor cultures at scales ranging from 3 to 20 L (Figure 3). This was possible in our laboratory following an integrated approach that focused on: (1) selecting and improving an HEK-293 clone 11; (2) selecting and developing a robust serum-free medium, based on detailed understanding of the cell line central metabolism 12, 17, 18; and (3) optimizing physicochemical operating conditions in the bioreactor 13, 14, 16, 23.
However, production of adenovirus vectors that maintain a high specific yield in batch operation was limited at cell densities in the range 5 × 105 cells/mL, which is an order of magnitude lower than the maximal cell densities achievable. Results shown in Figure 4 clearly demonstrate a drop in the specific infectious viral particles expressing a GFP-Q marker gene at cell densities higher than 5 × 105 cells/mL. Note that these results were generated while proceeding with infection in fresh NSFM13 medium. Beyond 2 × 106 cells/mL, no infectious particles are detectable. These results strongly suggest a limitation either due to key nutrient depletion or inhibitor byproduct accumulation. The break point related to specific production drop with cell densities at infection is medium dependent. No available commercial medium to date has shown potential to support high yields of viral particles, while maintaining the specific production optimal at cell densities beyond 1 × 106 cells/mL.
Therefore, further process developments were considered consecutively: sequential-batch, fed-batch and perfusion operations. These modes of operation are of incremental difficulty in implementation and operation. Figure 5 shows the basic principles of these methods.
Figure 6 summarizes results from an experiment completed in a controlled 3-L scale bioreactor using HEK-293 cells in serum-free medium NSFM13. Sequential separation was realized using a continuous centrifuge 35. Infection was completed at 2 × 106 cells/mL in fresh medium. Subsequently, medium was exchanged (∼90%) at 22 and 42.5 hpi. Maximum specific yield was obtained at 42.5 hpi with 9400 and 3500 viral particles (VP) and infectious viral particles (IVP) per cell, respectively, while the HEK-293 cell viability dropped to ∼60%. In general, maximum yield of adenovirus vectors is observed around 40 hpi, which is when the product is harvested in production runs. Decrease in VP at 46.5 and 70.5 hpi by 24 and 30%, respectively, is observed. A less pronounced decrease, in the range of 10%, is observed with the specific IVP around 3000 IVP/cell. These trends are systematically observed in our experiments and they have been correlated with the cell viability decrease during the course of the production. However, at this time, the mechanism behind this observation is unclear. Multiple assumptions, including viral particle aggregation, sample preparation, and viral construct instability, are being further investigated in order to better explain this phenomenon.
Easy to operate and readily scalable, the fed-batch mode is employed to extend culture lifetime by supplementing limiting nutrients or reducing the accumulation of toxic metabolites. Fed-batch addition of glucose, glutamine and amino acids allowed infections at cell densities up to 2 × 106 cells/mL; however, the productivities attained at high cell densities were lower than those obtained with infection at cell density of 1 × 106 cells/mL [17,18]. The development of an efficient feeding strategy is intimately linked to the identification of the factors limiting or inhibiting the virus replication beyond the breakpoint cell density. The rational approach undertaken in this laboratory relies on a better understanding of the cell metabolism during the course of viral production through metabolic flux analysis which helps identify key factors for the formulation of feed concentrates.
In the perfusion mode, cells are retained at a relatively high concentration inside the reactor. Many retention devices or reactor configurations have been employed: sedimentation chambers, membrane units, centrifuges, hollow fiber bioreactors (HFBR), ceramic cartridges, immobilized cell systems, spin-filters and acoustic separators; all with their intrinsic advantages and disadvantages. To account for the increased shear sensitivity of the cell during the infection phase, an acoustic filter was used in our experiments. By using a perfusion system, in situ medium exchange was accomplished and a better control of the culture environment was possible. Additionally, toxic byproducts were removed. The fresh nutrient supply and metabolite removal, combined with the cells' retention in the bioreactor, allow a high cell concentration to be achieved (results not shown); however, the specific production could only be maintained by infecting at densities up to 3 × 106 cells/mL. Figure 7 shows that as a result of feeding early in the growth phase and at a rate of two reactor volumes per day, cells were infected at densities up to 2.5 × 106 cells/mL and a maximum specific production of 18 800 VP/cell was attained. Current efforts are being extended towards the optimization of operating conditions such as feed rate, infection time and harvest time and shear reduction (reduced agitation and recirculation).
In summary, nutrient limitations and/or inhibitor accumulations restrict the high cell density infection and viral vector production. In batch operation, AdV-specific production is maintained only at low cell densities ranging from 0.3–0.5 × 106 cells/mL. To date, despite considerable efforts, addition of selected amino acids and glucose in fed-batch cultures during adenovirus production has not translated into significant increases in cell density at infection or high yields of AdV. This approach has led to moderate improvements over the batch process by maintaining AdV-specific production at 1 × 106 cells/mL. Further developments of this strategy will require better understanding of the cell's central metabolism and infection kinetics in order to develop optimal feeding strategies. As an alternative, a perfusion method has been shown to be feasible using a low shear separation system to circumvent metabolic limitations beyond cell densities at infection in the range of 3 × 106 cells/mL. Our current work focuses on designing a better perfusion medium. Again we rely on better understanding the HEK-293 cellular metabolism during the adenovirus replication process using metabolic flux analyses to achieve this task.
On-line monitoring and control
The production of viral vectors can be greatly facilitated by monitoring methods that can quantitatively indicate successful and consistent infections. Implementation and use of on-line monitoring was instrumental in the understanding of adenovirus production kinetics and allowed for advanced control and optimization of the AdV bioprocess 14, 16–18. Several on-line (GFP, oxygen demand, capacitance) and off-line (cell size, metabolite assays, cell viability) methods offer different approaches for simple and robust means to monitor infection and production. Figure 8 shows increased oxygen consumption after the infection followed by a sharp decrease due to the death of the cells after completion of viral synthesis. An on-line fluorescence sensor can also be used as a tool to monitor AdV production, more specifically the viral protein expression kinetics 14. Measurement of the capacitance of the cell culture shown in Figure 9 is a consistent indicator of a successful cell infection due to the increase in cell volume following viral DNA replication 24.
Chromatography purification method
Traditionally, purification of recombinant adenovirus vector was achieved using two rounds of cesium chloride (CsCl) density gradient ultracentrifugation. This technique was useful for generation of small-scale clinical-grade lots. However, with the advent of gene therapy protocols, large quantities of AdV are required. Because the CsCl density gradient method is not scalable, other purification procedures were evaluated, for example, ion exchange, hydrophobic interaction, metal chelate and size-exclusion chromatography. These methods were evaluated for capture and purification of type-5 adenovirus vector 36.
In our laboratory, ion-exchange chromatography was selected as a first purification step and size exclusion as a polishing step. Figure 10 shows the adenovirus vector downstream processing flow sheet, currently used in our process. The flow sheet summarizes the sequence required to achieve a 99% pure adenovirus from a 20-L scale suspension culture production. Cell slurry concentrates are harvested by continuous centrifugation at 48 hpi from culture infected at 106 cells/mL. Cells are lysed for 10 min by osmotic shock. Lysate is treated with DNAase, centrifuged, conditioned and then filtered. The feed applied to a 7.0 cm diameter Fractogel® column has a titer of 1.1 × 1010 IVP/mL. Approximately 1100 mL of semi-purified AdV is eluted with NaCl. Recovery achieved at this stage is in the range of 80%. The semi-purified AdV solution is ∼10× concentrated by ultra-filtration. The polishing step is completed using a Sephacryl® size-exclusion chromatography column 37. As an alternative polishing step, Green et al.38 proposed use of a PolyFlo® flow-through chromatographic method to remove additional host and viral proteins not removed in the first anion-exchange step. A final concentration and formulation is completed thereafter according to the end use. Throughout the processing, concentration of the viral particles by ultra-filtration is required; however, more than 50% of VPs can be lost at this stage. It is believed that very high local concentrations of adenovirus particles in the vicinity of the membrane might be responsible for particle aggregation that contributes to a reduction in the recovery. Therefore, optimization of these steps is critical for maximization of the recovery. Stability of semi-purified adenovirus was evaluated at 4 °C and −80 °C for periods exceeding 20 months without loss of VPs or IVPs. Different buffers have been used for the final formulation of the purified virus. A recent report from an adenovirus reference material working group recommended the following formulation: 2.5% glycerol (w/v) and 25 mM NaCl in 20 mM Tris at pH 8.0 and referred to as GTS formulation 39. The purity of an adenovirus preparation is generally estimated by HPLC profile using UV detection. In our hands 37, the tandem ion-exchange/size-exclusion column generated material with high purity. The HPLC profile showed a single peak and a UV detection ratio at 260 over 280 nm of 1.21. Western analysis with anti-Ad5 polyclonal antibodies was used to confirm identity of the purified material with appropriate hexon, penton and fiber bands while silver-stained reduced SDS-PAGE gel was used to demonstrate successful removal of host proteins. Absence of double-stranded DNA (dsDNA) in the purified material was confirmed by PicoGreen dsDNA assay 27. In all the assays double-banded CsCl-purified material was used as a reference material. For a more detailed description of quality control analyses required for recombinant adenovirus vectors for gene therapy, please refer to the aforementioned publication 34. In summary, the two-step chromatography process provides a high-purity adenovirus vector equivalent, if not better than the CsCl-purified AdV. Operations are currently being scaled up to 100 L.
In recent years, as a result of the steady increase in the number of clinical protocols using recombinant type-5 AdV, a dramatic shift in the production of adenoviral vectors from adherent cell culture to suspension culture has been observed. Large-scale productions were possible only after intense development of medium formulation and feeding strategies. Scalability of adenovirus production is considered achievable; however, many questions remain to be addressed. Important literature was dedicated to adenovirus infection and trafficking processes in vitro and in vivo; however, mechanisms of virus replication and maturation in complementing host cells for production purposes remain largely unknown. Maximization of the per cell virus particle number and stability of active vector during the course of production will certainly rely on better understanding of key replication mechanisms and key kinetic parameters, yet to be determined.
Another significant achievement in the field is related to the use of serum-free medium for manufacturing AdV vector lots. Different formulations of serum-free media are now available from various manufacturers. Also, many companies and organizations have developed their own proprietary formulations. This might have the advantage of allowing rational development and optimization of production processes based on defined medium.
To overcome the so-called cell density effect that limits the production of AdV vectors at cell densities in the range of 5 × 105 cells/mL, fed-batch as well as perfusion mode were evaluated. Clearly, serious attempts in using fed-batch operations to overcome nutrient limitation and/or byproduct inhibition did not translate into significant yield improvements. More work is certainly needed to elucidate basic host cell metabolism in order to rationally define feeding strategies that provide essential nutrients under conditions that prevent accumulation of inhibiting factors. As an alternative, perfusion mode using low-shear cell retention devices are used to overcome these limitations. This is now appearing as a common trend in viral vector conference reports.
Current chromatographic techniques based on ion exchange and size exclusion allow for large-scale purification of AdV vectors with purity comparable to the CsCl gradient method. However, further optimization is needed to minimize ultra-filtration steps where loss of AdV might be encountered following particle aggregation. Also, a better integration and containment of downstream processing steps is desirable to ease operations at larger scales.
The authors would like to thank Danielle Jacob for shake-flasks and bioreactor experiments; Philippe-Alexandre Gilbert for developing the GFP probe; Robert Voyer for the implementation of the capacitance probe; Isabelle Nadeau for developing the HEK-293 metabolic model; and Alice Bernier and Normand Arcand and Parminder Chahal for developing the adenovirus purification process. Thanks are also extended to Marc Aucoin for useful editorial comments.