|Home | About | Journals | Submit | Contact Us | Français|
Lentiviral vectors (LV) represent a key tool for gene and cell therapy applications. The production of these vectors in sufficient quantities for clinical applications remains a hurdle, prompting the field toward developing suspension processes that are conducive to large-scale production. This study describes a LV production strategy using a stable inducible producer cell line. The HEK293 cell line employed grows in suspension, thus offering direct scalability, and produces a green fluorescent protein (GFP)-expressing lentiviral vector in the 106 transduction units (TU)/mL range without optimization. The stable producer cell line, called clone 92, was derived by stable transfection from a packaging cell line with a plasmid encoding the transgene GFP. The packaging cell line expresses all the other necessary components to produce LV upon induction with cumate and doxycycline. First, the study demonstrated that LV production using clone 92 is scalable from 20mL shake flasks to 3L bioreactors. Next, two strategies were developed for high-yield LV production in perfusion mode using acoustic cell filter technology in 1–3L bioreactors. The first approach uses a basal commercial medium and perfusion mode both pre- and post-induction for increasing cell density and LV recovery. The second approach makes use of a fortified medium formulation to achieve target cell density for induction in batch mode, followed by perfusion mode after induction. Using these perfusion-based strategies, the titer was improved to 3.2×107 TU/mL. As a result, cumulative functional LV titers were increased by up to 15-fold compared to batch mode, reaching a cumulative total yield of 8×1010 TU/L of bioreactor culture. This approach is easily amenable to large-scale production and commercial manufacturing.
Lentiviral vectors (LV) offer a number of valuable properties, including stable gene integration into the host genome, the ability to transfer genetic information into dividing and nondividing cells, and a broad tissue tropism via VSV-G pseudotyping.1 They are currently used in clinical trials to treat both rare and more frequent genetic and acquired diseases,2–9 as well as in chimeric antigen receptor T cell cancer therapy.10,11
LVs are typically produced using adherent HEK293 cell lines by multi-plasmid transient transfection in serum-containing media. However, this methodology is labor-intensive, plasmid supply-dependent, and not directly scalable. Adherent processes rather need to be scaled out by increasing the available surface for cell attachment and growth.12 Transient transfection also results in contaminations of the final product due to excess plasmids13 and residual transfection reagent. As a result, stable LV producer cell lines have been generated. Successful approaches involve the use of an inducible system to circumvent cell death caused by the continuous expression of cytotoxic proteins (Gag, Rev, and VSV-G).14 Most reports of stable producer cell lines have described adherent cell lines,5,15,16 imposing constraints for scale-up and large-scale production. These challenges have partially been overcome for other viral vectors through the use of highly intensified adherent cultures using single-use fixed-bed bioreactors.17–19 In parallel, several academic and industrial laboratories are working on the development of suspension-based production processes,12,20,21 which will eventually allow manufacturing of LV in classical stirred tank bioreactor facilities.
According to recently published reviews,22,23 the field is moving toward the use of suspension processes for LV production. Indeed, the use of LV in therapeutic applications requires manufacturing of sufficient amounts of clinical and commercial high-quality supply in a reproducible way. Depending on the application and disease, 1–40×109 infectious units of vector per patient are needed13; this increases the economic pressure and creates the need to develop high-yield production processes.1,22
The present study proposes a LV production strategy using a stable inducible producer cell line derived from a packaging cell line that has been previously described14 and that grows in suspension in serum-free media (SFM). Induction of expression of the elements necessary to produce LV occurs after the addition of cumate and doxycycline. Because of the notorious low stability of LV, the study opted for the development of a production process in perfusion mode, and two different strategies were evaluated.
The HEK293SF-LVP-CMVGFPq-92 cell line (abbreviated clone 92) was grown and maintained in SFM4TransFx293 (Hyclone), or in HyCell™ TransFx-H media (Hyclone), which were both supplemented with 4mM of L-glutamine in all the experiments described in this study. HyCell™ TransFx-H was also always completed with 0.1% poloxamer 188, which is a shear protectant used in high shear stress environments. Cell Boost 5™ (CB5) Supplement (3.5g/L; Hyclone) is a chemically defined feed and was supplemented only if indicated in the text. Cells were grown in suspension in shake flasks (Corning) with an agitation of 110–120rpm using orbital shakers (Infors HT), 5% CO2 at 37°C. Cell counts were performed with automated cell counters (Cedex Automated Cell Counter or NucleoCounter® NC-200™) or using a hemacytometer and erythrosine B. Cells were passaged regularly when reaching densities approaching 2×106 cells/mL.
Plasmids used in this study were purified by chromatography using the maxiprep plasmid purification kit (Qiagen) by following the manufacturer's recommendations. The LV packaging cells used to generate HEK293SF-LVP-CMVGFPq-92 have been described previously and were obtained using the One-Shot strategy (i.e., all the LV components were added simultaneously in a single transfection).14 To generate clone 92, packaging cells (clone 29-6) grown in LC-SFM medium (Life Technologies) supplemented with 1% fetal bovine serum (FBS; Hyclone) were co-transfected using Lipofectamine 2000 reagent (Life Technologies) with plasmid pCSII-CMV5-GFP previously linearized with StuI and with plasmid pCDNA6-hisA (Life Technologies) encoding the resistance for blasticidin previously linearized with Xho1 at a plasmid mass ratio (μg/μg) of 7:1. Two days after transfection, 7μg/mL of blasticidin (InvivoGen) was added to the culture medium, and the cells were grown until a pool of blasticidin-resistant cells was generated (about 3 weeks). The pool was then subcloned by limiting dilution in 96-well plates in medium without blasticidin. Clones were picked and first analyzed for GFP expression. Clones with high levels of GFP expression were expanded and tested for the production of LV; LV production is initiated by inducing the cells through the addition of 1μg/mL of doxycycline and 30μg/mL of cumate. One of the best stable producer clones (clone 92) was adapted to grow in serum-free conditions in SFM4Transfx-293 medium (Hyclone), supplemented with 4mM of glutamine. A research cell bank was made in the same medium supplemented with 10% dimethyl sulfoxide (Sigma–Aldrich).
Clone 92 was thawed in SFM4TransFx-293 and cultivated in suspension culture in 125mL shaker flasks with 20mL of culture volume. At 2, 6, and 10 weeks after thawing, cells were induced at a concentration of 1.0×106 cells/mL for the production of LV. At 48h post induction, the medium was clarified by centrifugation, and the supernatant containing the LV was supplemented with 1% FBS (as cryoprotectant) and then frozen at −80°C. The amount of LV in the supernatant was titrated simultaneously by gene transfer assay (GTA; see section “GTA for lentiviral quantification” below).
In the first set of shake flask experiments, cells were seeded at 0.2×106 cells/mL in SFM4TransFx293 medium. Once the cells reached a cell density of 1–1.5×106 cells/mL, they were split into 50mL aliquots in 250mL shake flasks. Daily medium replacement (DMR) was performed for five consecutive days until induction by pelleting the cells by centrifugation (300 g, 5min) and re-suspending them in different ratios (0–100%) of fresh versus spent medium. Cell counts and viability were determined with a Cedex Automated Cell Counter after treating the cells for 30min with Accumax™ (Sigma–Aldrich) to reduce cell aggregation.
In the second set of experiments, cells were inoculated at a cell density of 0.3×106 cells/mL into 100mL of complete HyCell™ TransFx-H media supplemented with 3.5g/L CB5 and batch cultured in a 500mL shake flask. When cell density reached ≥5E6 cells/mL, the cell suspension was split into 20mL aliquots in 125mL shake flasks, and cells were induced (1μg/mL of doxycycline and 30μg/mL of cumate). Starting at 24h after induction and every 24h thereafter, cells were subjected to DMR at 50% with medium either supplemented with 3.5g/L CB5 or not. The collected medium was filtered through a 0.45μm filter, and LV content was titrated by GTA. Daily cell counts and viability measurements were performed using the NucleoCounter® NC-200™.
The bioreactor setup used for the four batch runs and perfusion run #1 was identical to the one used in a previous study12; similar conditions have been described earlier in detail.24 In brief, a Chemap 3.5L type SG bioreactor vessel (working volume 2.7L) was equipped with probes to measure and control different parameters (temperature of 37°C; pH 7.05–7.1; dissolved oxygen [DO] 40%; agitation 80rpm). In perfusion mode, cells were retained in the bioreactor using a 10L acoustic filter (AppliSens) operated in backflush mode (full recycling of cell suspension into bioreactor at each backflush) with an interval of 30min and a run/stop ratio of 55/5s (Fig. 4). Cultures were grown up to ~1–1.5×106 cells/mL in batch mode. Perfusion was started at 0.5 volume of medium per reactor volume per day (VVD), and increased to up to 1 VVD after induction. The culture was induced after reaching the targeted cell density (5×106 cells/mL) in perfusion mode.
For perfusion #2–4, cells were inoculated at cell densities ranging from 0.25 to 0.35×106 cells/mL into 1L of complete HyCell™ TransFx-H media (Hyclone) supplemented with 3.5g/L CB5, and cultured in the BioFlo® 320 1L stirred tank reactor in batch mode (37°C; pH 7.15; DO 40%; 80rpm). Cells were induced at ≥5×106 cells/mL. After induction, perfusion mode (1 VVD) was started with fresh medium containing inducers for 5–7 days. The miniBioSep (Applikon®) was used for cell retention (1W power setting) during perfusion. Cell counts and viability measurements were performed using the NucleoCounter® NC-200™ after daily sampling from the bioreactor.
In all perfusion runs, harvests were collected, and the LV-containing supernatant was kept on ice or at 4°C until clarification (once daily) and subsequently stored at −80°C until quantification using the GTA assay. As two different bioreactor scales were used in this study, the total cumulative LV yield is normalized per liter of bioreactor culture (TU/L) to facilitate direct comparison.
Functional viral titer (GTA titer) was determined using a flow cytometry–based GTA, as described previously with slight modifications.12,21 In brief, HEK293A or 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 5% FBS and 2mM of L-glutamine. The cells were seeded at a density of 1E05 cells/well of a 24-well plate 5h before transduction with LV. Prior to transduction, LV samples were serially diluted in DMEM supplemented with 8μg/mL of polybrene and incubated at 37°C for 30min. Transduction was performed by removing culture medium, adding 200μL of diluted LV to cells, and incubating over night at 37°C. Next day, 800μL of culture medium was added to each well, and cells were incubated for an additional 48h prior to flow cytometry to quantify GFP expressing cells. Therefore, GFP was measured 3 days post transduction. Note that the GFP signal measured is truly due to transgene expression and not to the presence of episomes, since a similar signal was measured 10 days post transduction in a control experiment (Supplementary Fig. S1; Supplementary Data are available online at www.liebertpub.com/hgtb). The titer (TU/mL) is determined using the formula [(% of GFP positive cells/100)×(number of cells transduced)×(dilution factor)×(1mL/volume transduced)]. An in-house LV standard was used in all measurements to minimize inter-assay variability.
A packaging cell line that contains all the necessary components to produce LV, except for the LV genomic RNA, has been described previously.14 Briefly, all the LV components (Gag/Pol, Rev, and VSV-G) were added simultaneously through a single transfection event to generate a packaging cell line named PacLV. In the packaging cells, transcription of Rev and the envelope protein (VSV-G) is under the control of the tetracycline and cumate switches, which means that addition of doxycycline and cumate in the culture medium is required to induce the production of LV (Fig. 1A). Gag/Pol is under the control of a CMV promoter. Even though the expression of Gag/Pol is constitutive and not directly controlled by the switches, its expression depends on the presence of REV because of a REV responsive element (RRE). A cell line (known as producer) that contains all the elements necessary to produce LV, including the genomic RNA, was then constructed. For ease of titration, a decision was made to construct producer for LV expressing GFP regulated by the strong constitutive CMV promoter. This was done by co-transfecting the packaging cells with the transfer vector pCSII-CMV5-GFPq, which also contains a RRE. Details about the plasmid construction have been previously reported.14 The producer cell line selected was named clone 92. To confirm that the production of LV from clone 92 was stable, the cells were cultivated for 10 weeks in shake flasks, and the production of LV was monitored at different time points after induction. LV titers >3.0×106 TU/mL were maintained throughout this period of time, indicating that clone 92 is stable (Fig. 1B).
In order to evaluate the performance of clone 92 in stirred tank bioreactors, four batch cultivations were performed at the 3.5L scale (Fig. 2) in SFM4TransFx293 media, our starting basal medium for this cell line. The viability profile after induction was showing a reproducible decline, reaching 70–80% at 3 days post induction (dpi; Fig. 2A). Functional LV titer production kinetics exhibit a reproducible peak at 3dpi, with an average titer of 6×106 TU/mL (Fig. 2B). Production was found to be directly scalable from shake flasks to bioreactors in batch mode, with comparable production profiles (Fig. 2C). The total LV yield per liter of culture was of 6×109 TU on average (Fig. 2D), with a specific production of 4.4 TU/cell.
In order to address the low stability of LV, the plan was to use continuous harvesting in perfusion mode to recover produced LV rapidly. As a model, two strategies were explored in shake flasks. First, experiments were performed to determine the effect of DMR during the cell growth phase and to test the impact of higher cell density at induction on production yield. Cells were centrifuged once daily, and 0–100% of the spent medium was replaced by fresh SFM4Transfx293 medium between −3dpi and 0dpi. Results show that the growth phase could be extended through the daily addition of increasing amounts of fresh medium; a cell density of >8×106 cells/mL was reached at 0dpi through complete daily medium replacement over 4 days (Fig. 3A), which is three times more than the batch control. Following induction, LV was quantified at 3dpi. Titer was significantly increased by media replacement going from 1.0×106 TU/mL without DMR to 3.35×107 TU/mL when 75–100% of the spent medium was replaced with fresh medium (Fig. 3B). A better specific production was measured with 75% DMR compared to 100% DMR. This indicates that an optimal cell density at induction for LV production under the conditions used is the one obtained with 75% DMR, which is approximately 5.5×106 cells/mL.
Therefore, a cell density at induction of 5×106 cells/mL was targeted in the second set of small-scale experiments. A hybrid system combining the use of a fortified medium during the cell growth phase and DMR during the production phase was tested. A new medium formulation that does not contain hydrolysates and promises a lower lot-to-lot variability (HyCell™ TransFx-H) was chosen. In our hands, the LV titers were similar between SFM4TransFx293 and HyCell™ TransFx-H when induction took place at 5×106 cells/mL after daily medium replacement of 75% (Supplementary Table S1).
In order to support growth to high cell density in batch mode, HyCell™ TransFx-H was supplemented with Cell Boost 5 (CB5), and cells were grown to 5×106 cells/mL until induction; 50% daily media replacement took place during the production phase with or without CB5 addition (Fig. 3C). Similar functional LV titers were obtained at 50% DMR with or without CB5 from 1dpi to 3dpi. However, supplementing with CB5 improved the daily titer at 4dpi and 5dpi (Fig. 3D). Taken together, these data suggest that expanding cells in a fortified medium (HyCell™ TransFx-H with CB5) results in cell densities similar to what is obtained in perfusion mode with simplified bioreactor operation until induction. In addition, adding CB5 during the production phase prolongs the production phase, presumably by improving cell viability.
Based on the results obtained in shake flasks, parameters were selected for bioreactor productions performed in perfusion mode. A schematic of the setup used to operate in perfusion mode is shown in Fig. 4. The first run (Perfusion 1) was performed using SFM4TransFx293 medium for growth and production in a 3.5L stirred tank bioreactor and is used as a reference. Perfusion was started at 0.5 VVD and increased to up to 1 VVD after induction.
In an effort to simplify the process, three additional runs (Perfusion 2–4) were operated using a hybrid mode similar to what was performed in shake flasks; cells were expanded in 1L of HyCell™ TransFx-H media supplemented with CB5 in batch mode. Cultures were induced at densities ≥5×106 cells/mL, and perfusion mode was started (1 VVD) with fresh medium containing CB5 and inducers. The cell density and viability of Perfusion 1 is significantly different from Perfusion 2–4 after induction (Fig. 5A and B); the viability dropped quickly after induction, similarly to what was observed in cultures without CB5 supplementation (Figs. 2A and 3A and C), and the run was stopped at 5dpi when the viability reached 25%. In contrast, the viable cell density of Perfusion runs 2–4 showed no growth arrest and increased up to 17×106 cells/mL after induction; viability remained >80% until harvest (Fig. 5A and B). The peak in titer (2–3×107 TU/mL) was observed at 3dpi for Perfusion 1, whereas it was at 4dpi for Perfusion 2–4 (Fig. 5C). Cumulative titers of the four perfusion runs are in a similar range of 5–8×1010 TU/L, with the cumulative titer of Perfusion 1 being highest (Fig. 5D). The titers measured in the bioreactor versus the harvest vessel are comparable (Supplementary Fig. S2).
Compared to the batch cultures, the average peak titer measured in the harvest for the four perfusion runs (2.2×107 TU/mL) is nearly fourfold higher than the one obtained in batch mode (6×106 TU/mL; Fig. 2B). Furthermore, on average, 11 times more total LV (up to 8×1010 TU/L) is obtained in perfusion mode compared to batch mode because of the daily harvests (Fig. 5B). The viable cell density at time of induction was between 1 and 1.5×106 cells/mL in batch, whereas it was 5.4×106 cells/mL in perfusion. The specific production was also improved in perfusion with an average of 11.5 TU/cell compared to 4.4 TU/cell in batch mode.
Using a stable producer cell line (clone 92), two strategies were developed for high yield LV production in perfusion mode. The first approach uses the basal medium SFM4TransFx293 and perfusion mode for both increasing cell density and LV recovery (Figs. 3A and B and 5A and B). The second approach made use of a new medium formulation (HyCell™ TransFx-H) in combination with medium fortification through CB5 addition: this approach might be simpler to operate, since the perfusion mode is only started after induction (Figs. 3C and D and 5C and D). Overall, it was possible to increase the cumulative yield of LV to up to 15-fold compared to batch mode. Although it is expected that the absolute LV titers of 8×1010 TU/L are transgene dependent, it is hypothesized that the yield improvements of >1log compared to batch mode will be transferable to other transgenes.
A few adherent stable producer cell lines have been described with LV titers in the range of 1×106–1.5×108.25–27 However, adherent cell lines present significant limitations for large-scale productions. In the strategy employed here, the mother cell line (HEK293SF-3F6) used to generate the packaging cell line and subsequently the stable producer was already adapted to suspension and serum-free conditions.28 A key element for the successful generation of clone 92 was the use of a double switch system; two inducers are required to initiate the synthesis of LV, which offers a tight control on the production of LV cytotoxic elements. Early attempts using only one switch were not successful (results not shown). The producer cell line reported here is grown in suspension, but further process optimization could be pursued. For example, CB5 addition seemed to extend the window of high LV titers (>107 TU/mL) to beyond 4–5dpi and also slowed down the decline in culture viability while maintaining cell growth after induction. CB5 contains several nutrients that support cell viability but may have some components (e.g., growth factors) that may be counterproductive to vector output. Indeed, comparing CB5 addition to its absence during induction in shake flasks, LV titers are at comparable levels but much lower on a per-cell basis with CB5 (Fig. 3C and D). It is envisioned that this system could be manipulated to balance LV production (a cytotoxic process) and cell maintenance to develop a continuous LV production model. In contrast, production in the basal medium SFM4TransFx293 without supplementation led to a faster release of LV until 3dpi (Fig. 4C and D). A combination of CB5 supplementation during the growth phase and removal of CB5 during the production phase could lead to a further increase in production yields at earlier time points. An extensive metabolite analysis and further CB5 titration could also be beneficial and support induction at higher cell densities.
The recovery after purification and downstream processing will also affect the final functional titers.29,30 The authors are currently investigating the use of column/membrane chromatography to decrease the amount of impurities in the final product, as published by other teams.21,31,32 Purification should also aim at removing the two inducers used in this study. Cumate is known for being non-cytotoxic,33 but doxycycline is an antibiotic of the tetracycline class. Therefore, this could be an issue for people with tetracycline allergies if traces are present in the final product.
In conclusion, to the best of the authors' knowledge, this study is the first to describe process development and improvement efforts for a stable LV producer suspension HEK293 cell line. The process described was reproducible and robust from small scale to 1–3L bioreactor scale. Two strategies for improvement of the cumulative LV yield were tested. Both of these employed acoustic cell filter technology to operate bioreactor cultures in perfusion mode. Overall, an increase in cumulative LV titer (TU per L of bioreactor culture) of >1log was achieved. This means that it was possible to produce up to 8×1010 TU/L bioreactor volume in perfusion mode. It is expected that the developed production process will be easily amenable to large-scale production and commercial manufacturing.
The authors would like to acknowledge the contribution of Julia Transfiguracion, Adam McLeod, Nick Timmins, Aaron Dulgar-Tulloch, Mélanie Simoneau, and Nazila Nazemi-Moghaddam to this work. Work performed at CCRM was co-funded by FedDev Ontario (Advanced Manufacturing Fund) and GE Healthcare.
No competing financial interests exist