Packaging cells for lentiviral vectors generated using the cumate and coumermycin gene induction systems and nanowell single-cell cloning

Abstract

Lentiviral vectors (LVs) are important for cell therapy because of their capacity to stably modify the genome after integration. This study describes a novel and relatively simple approach to generate packaging cells and producer clones for self-inactivating (SIN) LVs pseudotyped with the vesicular stomatitis virus glycoprotein (VSV-G). A novel gene regulation system, based on the combination of the cumate and coumermycin induction systems, was developed to ensure tight control for the expression of cytotoxic packaging elements. To accelerate clone isolation and ensure monoclonality, the packaging genes were transfected simultaneously into human embryonic kidney cells (293SF-3F6) previously engineered with the induction system, and clones were isolated after limiting dilution into nanowell arrays using a robotic cell picking instrument with scanning capability. The method’s effectiveness to isolate colonies derived from single cells was demonstrated using mixed populations of cells labeled with two different fluorescent markers. Because the recipient cell line grew in suspension culture, and all the procedures were performed without serum, the resulting clones were readily adaptable to serum-free suspension culture. The best producer clone produced LVs expressing GFP at a titer of 2.3 × 10⁸ transduction units (TU)/mL in the culture medium under batch mode without concentration.

Graphical abstract

Vectors derived from lentivirus are used extensively to modify cells for cell therapy. A method to generate cells (packaging cells) for efficient production of such vector is provided. To increase the product safety and facilitate scale-up, the packaging cells were adapted to serum-free suspension culture.

Introduction

Lentiviruses are complex retroviruses that can efficiently integrate their RNA genome after reverse transcription into the chromosomes of dividing as well as non-dividing cells. This results in stable integration of the gene they carry. For this reason, viral vectors derived from lentivirus (lentiviral vectors [LVs]) are commonly used to genetically modify cells for cell and gene therapy applications. Although encouraging data have been reported for the correction of genetic mutations in vivo,^1,2 the typical application of LV is to modify cells ex vivo for cell therapy.^3,4,5 Several such therapies are currently under investigation in clinical trials and some have been approved for commercialization to correct genetic disorders, such as β-thalassemia and metachromatic leukodystrophy, as well as to treat cancer, based on engineered T cells (CAR-T cell therapy). In light of the potential growth of commercial cell and gene therapy products based on LV, the demand for high titers and good quality LV production is expected to increase. Therefore, it remains important to optimize the LV production process to improve the yield and reduce the manufacturing costs under cGMP-compatible conditions.

Transient transfection of human embryonic kidney cells (HEK293 cells) with plasmids that carry genetic elements necessary for LV assembly is the most straightforward method to produce this vector. In the case of a third-generation LV, these elements, which are normally split into four plasmids, consist of the surface envelope protein (most often the glycoprotein G of vesicular stomatitis virus [VSV-G]), the Gag/Pol gene (encoding the capsid, reverse transcriptase, integrase, and protease), Rev, and the transfer vector (carrying essential cis-acting sequences and the gene of interest).^5,6,7,8 This method is cumbersome and expensive since it requires the production and purification of several plasmids and it is difficult to scale up, especially if the cells grow on an adherent surface. In order to facilitate the manufacturing of LV, packaging cells have been developed that contain all the elements necessary for LV assembly.^5,6,7,8 LV can be produced using packaging cells either by transient transfection with one plasmid (the transfer vector), instead of multiple plasmids, or through the generation of producer clones that have stably integrated the transfer vector. Since there is no need of plasmid and transient transfection, producing LV using producer cells is greatly simplified.

Some gene products necessary for the formation of LV, such as the protease encoded by the Gag/Pol gene and VSV-G, are cytotoxic. Therefore, their expression needs to be reduced or silenced during cell growth and turned on during the production phase. Despite the toxicity of the protease, it has been possible to generate packaging cells that constitutively express the protease,^9,10,11,12 or to use a less active mutated version of the protease.¹³ In such constitutive packaging cells, the VSV-G is replaced with envelope proteins that are less toxic, such as those derived from gammaretrovirus.^9,10,12,13 Despite the difficulty of comparing titers obtained from different laboratories, the highest titers have been reported with inducible packaging cells.⁵ Regulation using the tetracycline-,^{14,15,16,17,18} the ecdysone-,¹⁹ or a combination of the tetracycline- and cumate-inducible systems²⁰ has been employed to generate inducible packaging cells for LV. Ideally to facilitate downstream processing and to streamline approbation by regulatory agencies for clinical applications, it is preferable to produce LV in the absence of serum. Serum is a complex substance that can be contaminated by various adventitious agents and whose ill-defined composition might vary from lot to lot.²¹ The difficulty in ensuring lot consistency is very challenging from a manufacturing perspective because it will affect process reproducibility and may reduce product quality and yield. Another important characteristic of packaging cells would be their capacity to grow in suspension culture because the scale-up of adherent cell-based manufacturing is particularly challenging. Attachment-dependent culture processing generally requires more steps (the required detachment and reattachment of cells during passaging) and time than suspension. In addition, maximum cell density in adherent culture is also limited by surface area in two-dimensional culture techniques (such as cell factory and roller bottles) or microcarrier culture, limiting its volumetric productivity. As a result, the scale-up of adherent cell culture process is not only complicated but also labor intensive, involving higher cost of goods. In contrast, suspension culture simplifies handling of cultures such as inoculation, cell passaging, and process scale-up significantly, and therefore reduces the production cost. Although scaling up adherent cells has been facilitated through the development of fixed-bed bioreactors,²² these are limited to a surface growth area of 500 m² (fixed-bed volume of 60–70 L).

Furthermore, demonstrating that a biological product is produced using a uniform population of cells (also known as monoclonal) is imperative for regulatory filling.²³ The most common approach to clone cells is through limiting dilution in 96-well plates. Although the method is simple and easy, it is tedious, due to the fact the plates are normally scanned manually under a microscope to identify colonies. To ensure selected colonies originated from a single cell (monoclonality), a second round of cloning is usually performed by limiting dilution. To facilitate the generation of cell lines and ensure monoclonality, various methods based on automation, miniaturization, and cell imaging have been developed.²⁴ One of these methods is based on limiting dilution in 6- or 24-well plate format containing an array of 4-nL nanowells, followed by scanning microscopy to identify nanowells containing single cells. The colonies derived from single cells are then picked and transferred into 384-well plates using a robotic system (the CellCelector, Automated Laboratory Solutions [ALS]).

In the current study, we described a relatively simple and straightforward method to generate efficient packaging cells and producer clones for self-inactivated (SIN) LV pseudotyped with VSV-G. The method is based on a novel gene regulation system that is described in the current study, which combines the cumate-²⁵ and coumermycin-inducible²⁶ systems to control expression of the genes needed for LV assembly. An additional novelty of this method is to accelerate clone isolation and ensure monoclonality, by transfecting the cells simultaneously with all the packaging plasmids and then clone them by limiting dilution into nanowell arrays followed by scanning and picking using the CellCelector. The packaging cells and producer clones grow in serum-free suspension culture and are stable for at least 9–10 weeks in continuous culture without selective agent. They are therefore amenable for large-scale production of LV for R&D purposes and clinical applications.

Results

Construction of 293SF-CymR/λR-GyrB cell line

Our first step to construct inducible packaging cells for LV was to test a novel induction system based on the combination of the repressor (CymR) of the cumate-inducible system²⁵ and of the chimeric transactivator (λR-GyrB) of the coumermycin-inducible system.²⁶ For that purpose, we first generated a cell line (referred to as 293SF-CymR/λR-GyrB) that stably expresses CymR and λR-GyrB. The 293-CymR/λR-GyrB cell line was constructed by transfecting HEK293 cells that were previously adapted to serum-free suspension culture (the 293SF-3F6 cell line²⁷) with a plasmid encoding CymR regulated by the strong constitutive CMV5 promoter (Figure 1A). Clones expressing CymR were isolated and characterized for cumate induction by screening them using an LV expression GFP regulated by the CMV5CuO promoter (Figure S1). One of the clones that demonstrated good regulation by cumate was then stably transfected with a plasmid encoding the λR-GyrB transactivator regulated by the CMV5CuO promoter (Figure 1A). Clones were isolated and tested for the presence of λR-GyrB by transducing them with an LV expressing GFP regulated by the coumermycin-regulated promoter (LV-12xlambda-TPL-GFP; Figures 1B and S2). Therefore, transcription of the genes regulated by this system is under the control of the coumermycin promoter, which consists of several copies of the lambda operator (λOp). Transcription from the coumermycin promoter is driven by the λR-GyrB, which, in the presence of coumermycin, dimerizes and binds to λOp to activate transcription (Figure 1D). Transcription of λR-GyrB is itself controlled by the cumate-inducible promoter (CMV5VCuO). CymR (which is constitutively produced by the cells) binds to the cumate operator (CuO) of the CMV5CuO promoter and prevents transcription. Addition of cumate releases CymR from CuO and allows transcription of λR-GyrB. Hence, in this inducible system, transcription of the gene of interest is activated after addition of cumate and coumermycin to the culture medium.

Figure 1.

Diagram of constructs used in this study and mechanism of gene regulation by the cumate/coumermycin gene induction system

(A) Constructs used to make the 293SF-CymR/λR-GyrB cells. (i) The coding sequence for the repressor of the cumate gene switch (CymR) is controlled by a strong constitutive promoter (CMV5). (ii) The coding sequence for the coumermycin transactivator (λR-GyrB) is controlled by the cumate-inducible CMV5CuO promoter. (B) Transfer vectors used for the production of LV. (i) Conditional SIN LV expressing GFP regulated by the coumermycin-inducible promoter (12xlambda-TPL). (ii) SIN LV expressing GFP regulated by the CMV promoter. (C) Genes used to construct the packaging cell line. (i) The Rev gene is under the regulation of the coumermycin-inducible promoter (13xlambda-TPL). (ii) The Gag/Pol gene is regulated by the constitutive hybrid CMV enhancer/β-actin promoter (CAG). (iii) The VSV-G gene is regulated the 13xlambda-TPL promoter. (D) Mechanism of gene regulation by the cumate/coumermycin induction system. The 293SF-CymR/λR-GyrB cell line constitutively expresses CymR and it contains the gene for λR-GyrB under the control of CMV5CuO promoter. In the absence of cumate, CymR binds to CMV5CuO and prevents transcription. Addition of cumate releases CymR from the promoter and the gene for λR-GyrB can be transcribed. In the presence of coumermycin, λR-GyrB forms a dimer that binds to several copies of the lambda operators (λOp; in this example, 12 copies are indicated) to activate transcription of the transgene of interest. 5′ LTR and 3′ LTR, long terminal repeats located at the 5′ and 3′ ends of the lentivirus respectively; R, R region of the LTR; U5, U5 region of the LTR; Tet, tetracycline promoter; ψ, encapsidation signal; RRE, Rev responsive element; cPPT, central polypurine track; WPRE, woodchuck hepatitis virus posttranscriptional regulatory element; SD and SA, splice donor and acceptor respectively; pA, polyadenylation signal; 12xλTPL, 12xlambda-TPL promoter.

Given the cytotoxic potential of some LV elements, the on/off ratio before and after induction is a crucial parameter. The efficacy of the cumate/coumermycin gene induction system of six of the best clones of 293SF-CymR/λR-GyrB was evaluated by testing the induction level before and after addition of cumate and coumermycin after transduction with LV-12xlambda-TPL-GFP. The induction level (on/off ratio after and before induction) ranged between 2,100 and 3,800 for the three best clones, thus demonstrating a very good regulation capacity of this induction system (Figure 2).

Figure 2.

Induction level of the coumermycin/cumate gene expression system

Suspension culture of clones of 293SF-CymR/λR-GyrB (7-2, 7-3, 7-10, 8-59,15-35, and 15-51) were transduced at an MOI of 5 TU with an LV producing GFP regulated by the coumermycin-inducible promoter (LV-12xlambda-TPL-GFP). The next day, the cells were induced by the addition of cumate and coumermycin, and, at 3 days post-transduction (2 days post induction), the level of GFP expression was analyzed by flow cytometry. The relative fluorescent index of the cell population in the absence (off) and in the presence of inducers (on) is indicated. 293SF cells (without the gene induction system) were transduced in parallel. The induction level after induction (On/Off ratio of fluorescence index) is indicated above the bars for each clone. The data are the mean ± SD of two independent experiments.

Cell cloning using nanowell arrays

Our standard procedure to isolate single clones involved limiting dilution into 96-well plates using serum-free medium. This was the procedure employed to create the 293SF-CymR/λR-GyrB cell line. However, sometimes it was challenging to obtain colonies using this approach because the cells had difficulty growing after limiting dilution in 96-well plates. As an attempt to improve the survival rate, we tested limiting dilution using nanowell arrays in a 24-well plate format. Each of the 24 wells (referred to as macrowell) contains 4,300 nanowells that physically separate the cells while enabling diffusion of medium components (Figure 3). The volume of the nanowells, being smaller compared with 96-well plates, might provide a better growth environment, because the culture medium would be conditioned faster by the cells due to its smaller volume. In addition, the low attachment surface of the plate prevents cells from adhering to the well. We therefore compared the efficacy to obtain colonies after limiting dilution of 293SF-CymR/λR-GyrB cells into 96-well plates and nanowell arrays using a cell density of 0.6 cells per wells. As shown in Table 1, the survival rate was improved using the nanowell arrays since 85% of the cells formed viable colonies after limiting dilution into nanowells compared with 56% in the case of 96-well plates.

Figure 3.

Example of cell growth in nanowell arrays in a 24-well plate format

(A) Diagram of nanowells. Each macrowell of the 24-well plate contains 4,300 nanowells (200 × 200 × 100 μm in size) that physically separate the cells while enabling diffusion of components since the medium covers the top of the nanowells. (B) Phase contrast microscopy of HEK293 cells within a nanowell. The growth of a single cell in one nanowell was monitored for 4 days (D0–D4). Colonies were picked and transferred into a 384-well plate at 3 to 4 days post seeding.

Table 1.

Comparison of cell survival after plating 293SF-CymR/λR-GyrB cells into nanowells vs. 96-well plates at 0.6 cell per well

	Number of wells plated per test (N = 3)	% of wells with cells at day 0 (Poisson)a	% of wells with cells at day 0 (Exp)b,c	Number of colonies at 5–10 daysc	% of cells that formed colonies (Poisson)c,d	% of cells that formed colonies (Exp)c,e
Nanowells	2,667	45.2	46.0 ± 5.74	1,043 ± 105	86.8 ± 7.8f	85.2 ± 3.6
96-wells	480	45.2	NA	170 ± 2.8c	58.9 ± 1.0	NA

NA, not available (96-well pates were not scanned on the day of plating).

^aTheoretical value based on Poisson distribution on the day of plating.

^bData obtain after scanning and analysis by CellCelector on the day of plating.

^cValues are the means ± SD (N = 3).

^dValues based on the number obtained with the Poisson distribution on the day of plating.

^eValue based on the experimental number on the day of plating.

^fValue significantly higher compare to 96-well plates (p < 0.01).

Monoclonality demonstration

After limiting dilution into nanowell arrays, the colonies were isolated and transferred into 384- or 96-well plates. Because of the small size of the nanowells, this is done more conveniently using a cell picking system, such as the CellCelector (ALS). To ensure monoclonality, the plate is scanned at day 0 (day of plating) and the scan is compared with a scan of the same well before and after picking the colonies (Figure S3). Because the culture medium completely covers all the nanowells within the wells of the 24-well plate, there is a risk that cells migrate into an adjacent well and generate what is referred to as a ghost well; i.e., a nanowell that contains a colony several days after plating but that was empty on the day of plating (day 0). To evaluate the probability of generating ghost wells and the level of cross-contamination after cell picking, we generated pools of HEK293 cells expressing fluorescent markers (GFP and Discosoma species red [dsRed]). The pools were generated by transducing HEK293 cells with LV expressing GFP and dsRed regulated by the strong constitutive cytomegalovirus (CMV) promoter. We performed three cloning experiments (tests 1–3), where the pool of GFP-positive cells was mixed with the pool of dsRed-positive cells to obtain a mixed population containing between 8% and 15% of dsRed-positive cells. After limiting dilution of the mixed pools into nanowells at a cell density of 0.6 cell/well, a total of 452 dsRed-positive colonies were picked and transferred into 384-well plates. The colonies were analyzed by fluorescence microscopy immediately after picking and 10 days later, and 120 colonies were also amplified and analyzed by fluorescence microscopy and flow cytometry. For these three tests, no GFP-positive cell contaminants were detected within the dsRed-positive clones (Figure 4A).

Figure 4.

Monoclonality demonstration using GFP and dsRed-positive HEK293 cells

(A) A pool of GFP-positive HEK293 cells was mixed with 8% of HEK293-dsRed-positive cells and seeded into nanowell arrays. At 4 days post plating, dsRed-positive colonies (negative for GFP) were picked and transferred into a 384-well plate. Ten days later, the dsRed and GFP fluorescent signal of the cell population was examined by fluorescent microscopy. Only dsRed-positive cells were detected. Bright-field and fluorescent images of the wells are shown. (B) A pool of 50% dsRed- and 50% GFP-positive HEK293 cells was plated into nanowell arrays. At 4 days post plating, GFP- and dsRed-positive colonies were picked alternatively and transferred into a 384-well plate. The cells were then analyzed by fluorescent microscopy 10 days later. Only pure dsRed- and GFP-positive colonies were observed. (C) GFP-positive and dsRed-positive HEK293 cells were seeded into distinct nanowell arrays. Four days later, dsRed- and GFP-positive colonies were transferred alternatively into a 384-well plate. Ten days later, the cells were analyzed by fluorescence microscopy. Only pure dsRed- and GFP-positive colonies were observed.

Next (test 4), we mixed the GFP and dsRed-positive pools together at an equal ratio and then seeded the mixed pool into nanowells at a cell density of 0.6 cell per nanowell. The colonies were picked by alternating each time between GFP- and dsRed-positive colonies and transferred into a 384-well plate. Overall, 364 colonies were isolated and no cross-contamination with GFP- or dsRed-positive cells was observed after picking (Figure 4B). Finally for test 5, the GFP- and dsRed-positive pools were plated into different wells. A total of 384 colonies were isolated 4 days later by alternating each time between GFP- and dsRed-positive colonies (Figure 4C). Once again, no cross-contamination of GFP- and dsRed-positive cells was observed after picking. The presence of ghost nanowells was also analyzed for tests 1 to 5. From the analysis of 16,030 empty wells on the day of plating (day 0), we detected only 11 ghost wells on day 3, thus indicating a probability of monoclonality greater than 99.9% (Table S1).

Development of packaging cells

Limiting dilution into nanowell arrays was then used to isolate clones of packaging cells for LV. The strategy we employed to generate the packaging cells was to transfect simultaneously the four plasmids carrying the genes necessary for the formation of LV (Gag/Pol, Rev, VSV-G, and the hygromycin resistance) into the 293SF-CymR/λR-GyrB cell line (clone 7-2). This one-shot transfection approach has been used successfully in the past to generate packaging cells by us and other researchers.^20,28 Transcription of Rev and VSV-G genes is regulated by the coumermycin-regulated promoter (13xLambda-TPL; Figure 1D), whereas transcription of the Gag/Pol gene is under the control of a strong constitutive hybrid CAG promoter (made by the fusion of the CMV enhancer to the actin promoter²⁹). Because the presence of the Rev protein is needed for the expression of the Gag/Pol gene product (Rev being required for the transport to the cytoplasm of the Gag/Pol mRNA where it is translated into a polyprotein³⁰) expression of the Gag/Pol gene product is therefore indirectly regulated by the coumermycin and cumate induction system.

The hygromycin-resistant cell pool was plated at a density of 1.4 cells per well into nanowells. The cells were plated at a higher cell density (1.4 instead of 0.6 cells per well) to facilitate the isolation of hygromycin-resistant clones. Five days later, 173 hygromycin-resistant colonies were picked and pooled together into one well of a 96-well plate to form what is referred to as a mini-pool. The next day, the mini-pool was resuspended and plated at a density of 0.6 cell per nanowell. Single colonies were then isolated and transferred into 384-well plates. Clones were amplified in 96- and then in 24-well plates and tested for the production of LV by transient transfection with a transfer vector for LV-CMV-GFP (Figure 1B) in conjunction with the addition of cumate and coumermycin. A total 281 clones were analyzed for LV-CMV-GFP production at this stage (data not shown). The best ones were amplified and tested in suspension culture. The titer of LV-CMV-GFP for the best 35 clones of packaging cells is shown in Figure 5A. The average titer was 6.5 × 10⁶ TU/mL, and several clones produced LV at titers above 1.0 × 10⁷ TU/mL. The variability of production by the packaging clones is due to their ability to produce the components for vector generation and the transfection efficacy, which vary between different clones. The stability of five clones was then tested for a period of 10 weeks in culture in the absence of selection agent (Figure 5C). The stability testing consists in keeping the non-induced cells in culture for up to 10 weeks; at different time points, some cells are retrieved, induced, and transfected with GFP to monitor LV production. LV samples from different time points were frozen and quantified at the same time. The five clones maintained their capacity to produce LV for a period of up to 10 weeks in culture. The two best clones were 3D4 and 3G8. Both clones produce LV-CMV-GFP during the stability study at an average titer of 1.3 × 10⁷ TU/mL.

Figure 5.

Production of LV from packaging and producer clones and stability study

Production of LV-CMV-GFP from clones of packaging (A) and producer cells (B). Stability study of clones of packaging (C) and producers cells (D). For (A) and (B) each dot represents the production from an individual and different clone. Cells were grown in suspension culture using serum-free medium. Production of LV was done by transfecting the packaging cells with the transfer vector for LV-CMV-GFP and induction with cumate and coumermycin. Sodium butyrate was added the next day. The LV was harvested and titrated by flow cytometry at 3 days post induction. In the case of the producer cells, LV-CMV-GFP was produced by adding only cumate, coumermycin, and sodium butyrate (no transient transfection). For the stability study, the clones were maintained during 9–10 weeks in culture in the absence of selection agent. Production of LV was tested at the indicated time point. The data are the infectious titer (TU per mL) in the culture medium without concentration. Titration for the stability was performed in duplicate. Data are the means ± SD.

Generation of producer clones

To test the capacity of our packaging cells to generate producer clones for LV (which produce LV without the need of transfection), clone 3D4 was co-transfected with a plasmid encoding LV-CMV-GFP (Figure 1B) and a plasmid encoding the resistance for neomycin. A neomycin-resistant pool was generated. The pool was cloned by plating the cells at a density of 0.6 cell per well into nanowell arrays. GFP-positive colonies (380), derived from single cells, were picked and transferred into 384-well plates. The clones were expanded further in 96-well plates and tested for the production of LV-CMV-GFP by adding the inducers (cumate and coumermycin). The majority (more than 80%) of the clones in 96-well plates were positive for LV production, and 60 promising clones were amplified further and tested again for LV production (data not shown). The 20 best clones were then tested for LV production in six-well plates after induction at a density of 1.0 × 10⁶ cells/mL. A titer above 1.0 × 10⁷ TU/mL was observed for the 20 clones tested (Figure 5B). The stability of LV production by the six best clones was then tested in the absence of selective agent for a period of 9 weeks in culture. The six selected clones maintained their LV productivity above 1.0 × 10⁷ TU/mL for the duration of the study (Figure 5D). The best two producer clones were clones 1E9 and 3E9, which produce, on average during the 9 weeks of culture, 5.0 × 10⁷ and 5.5 × 10⁷ TU/mL, respectively.

Regulation of gene expression of LV packaging elements

To confirm the efficacy of the cumate/coumermycin gene induction system to regulate gene expression in the context of the packaging cells, the expression of Rev, Gag, and VSV-G was analyzed by western blot before and after induction using clone 3D4 at 24, 48, and 74 h post induction (Figure 6A). Expression of Rev, Gag, and VSV-G was clearly induced after addition of cumate and coumermycin. Although Gag/Pol transcription is under the control of a constitutive promoter (the CAG promoter), its expression was increased after induction due to the increased expression of Rev, which is needed for the efficient transport of the Gag/Pol mRNA to the cytoplasm. Because it has previously been shown that the addition of sodium butyrate can significantly increase the LV production,^{14,16,17,31,32} we investigated its effect on the expression level of Rev, Gag, and VSV-G. Addition of sodium butyrate resulted in an increase in the amount Gag, VSV-G, and Rev produced by the packaging cells (Figure 6A), which is consistent with the significant improvement of LV titer observed after addition of this compound (see below). The increase of Gag (Gag precursor protein and p24 fragment) and VSV-G after induction, and the additional increase caused by sodium butyrate, was confirmed by performing a semi-quantitative analysis by scanning the gels (Figure 6B).

Figure 6.

Regulation of expression of LV gene components in the packaging cells

(A) Expression of VSV-G, Rev, and Gag by the packaging cells (clone 3D4) was analyzed by western blots after induction with cumate and coumermycin. The same amount of total protein from the cell lysate was used for the analysis. Cells were harvested before (0) or after 24, 48, and 72 h of induction in the absence or presence of sodium butyrate (+B) (added 18 h post induction). 293SF-CymR/λR-GyrB cells were used as a negative control (CT). The position of molecular weight marker in kDa (MW) is indicated. The position of VSV-G, Rev, the Gag polyprotein (GagPP), and p24 is indicated. Note the presence of non-specific bands (∗) in the negative control when using the anti-VSV-G antibody. (B) Analysis of the induction level for VSV-G, GagPP, and Gag p24. The western signal before and at 48 h post induction in the absence or presence of sodium butyrate was measured using a digital imaging system. The values were normalized to the signal of β-actin obtained after stripping and incubating the membrane with an anti-actin antibody. The experiment was done in triplicate (N = 3). Value are the means ± SD. Means significantly different from each other: ∗∗p < 0.01, ∗p < 0.05.

To evaluate the importance of repressing transcription of Gag/Pol and VSV-G genes during cell culture, the cell growth and cell viability of packaging clone 3D4 were compared under induced and non-induced conditions. As control, the parental cell line (293SF-CymR/λR-GyrB) was cultured in parallel under the same condition (Figure S4A). Although the presence of the inducers (cumate and coumermycin) did not significantly affect the cell viability, induction significantly reduced the growth rate of the packaging cells. This effect was observed as early as 48 h post induction. Furthermore, the packaging cells cultured under induced condition formed large cell aggregates, which were far less abundant in the non-induced cultures (Figure S4B).

Increased productivity at higher cell density in the presence of sodium butyrate

The production of LV described so far was performed at a cell density of 1 million cells/mL. We were interested to investigate if the volumetric titer from one of our top producer clones (clone 3E9) could be improved by increasing the cell density at the time of induction. To test this, the cells were grown to a concentration of 2 million and 5 million cells per mL without medium replacement and then induced with cumate and coumermycin. The LV in the culture medium was titrated without concentration at 3 days post induction. We observed that the volumetric titer was improved 2.5-fold (reaching a titer of 2.3 TU × 10⁸ TU/mL) by increasing the cell density at induction by the same value (Figure 7). We also investigated the effect of sodium butyrate on the titer. When cells were induced at a density of 2 million cells/mL, we observed that the presence of sodium butyrate increased the titer by 11-fold (increasing from 0.8 to 9.0 × 10⁷ TU/mL). The effect of sodium butyrate is not due to an enhancement of GFP expression during titration, because there was no difference in titer when an LV (previously concentrated on sucrose cushion, to remove any trace of sodium butyrate) was titrated in the presence or absence of 0.7 mM sodium butyrate (data not shown). This corresponds to one-tenth the sodium butyrate concentration used for production, whereas the LV is diluted more than 100 times during titration. We have also investigated the effect of sodium butyrate on the quantity of vector RNA present in the producer clone 3E9 by RT-PCR using primers specific for the packaging signal. Addition of sodium butyrate resulted in an increase of 5-fold in the amount of vector RNA at 48 h post induction (Figure S5). Taken together, the data indicate the enhancement of titer by sodium butyrate is multifactorial as it stems from an increase in the amount of structural components of the LV (Gag and VSV-G; Figure 6) as well as vector genomic RNA.

Figure 7.

Effects of cell density and addition of sodium butyrate on LV-CMV-GFP productivity by a producer clone

Cells from producer clone 3E9 grown in suspension culture in shake flasks, with a volume of 15 mL of serum-free medium, were induced with cumate and coumermycin when they reached the indicated cell density. Sodium butyrate (supplemented at 18 h post induction) was added in some flasks. The titer of LV-CMV-GFP in the culture medium was measured at 3 days post induction by flow cytometry. The experiment was done in triplicate (N = 3). The titer is expressed as the mean ± SD; ∗∗p < 0.01 (significantly different).

Discussion

In the current study, we described a novel approach to generate packaging cells for the production of SIN LV pseudotyped with VSV-G. This was achieved through the use of a combination of two gene induction systems (cumate and coumermycin) to control the expression of the packaging elements, in conjunction with cell cloning using nanowell arrays and the CellCelector to isolate colonies and document monoclonality. To speed up the establishment of the packaging cells without having to go through several rounds of cell cloning, the three plasmids encoding the packaging elements (Gag/Pol, Rev, and VSV-G) were transfected simultaneously in the presence of a fourth plasmid encoding a selective agent. The cell line (293SF-3F6) that served as receptor for these genetic elements has been described before²⁷ and was already adapted to serum-free suspension culture. Since all the procedures employed to generate the packaging and producer cells were performed in serum-free medium and suspension culture (except for limiting dilution in nanowell arrays and cell amplification up to 24-well plates, which were conducted in adherence), no additional efforts were needed to adapt the clones to serum-free suspension culture. We have chosen to work in suspension culture to facilitate the scale-up and we employed serum-free medium to facilitate acceptance by regulatory agencies, since serum has an ill-defined composition and can be contaminated by various adventitious agents.²¹ One key advantage of packaging cells is their capability to generate producers that can yield LV without the need of transient transfection. We have shown that our packaging cells are readily amenable to this after the stable integration of a plasmid carrying the sequence of a SIN LV.

The highest titer produced by our best packaging cells (clone 3D4) after transient transfection in suspension culture under our standard conditions (cell density of 1 million cells/mL, without the addition of feed) for LV-CMV-GFP ranged between 1.0 × 10⁷ and 2.5 × 10⁷ TU/mL. In the case of producers, the productivity of the best clones (3E9) was slightly higher, as it was in the range of 6.0–7.0 × 10⁷ TU/mL. One potential reason for the lower productivity of the packaging cells has to do with the transfection efficacy which was between 40% and 60%, in contrast to producer clones, whose entire population can produce LV. In addition, no process development to optimize the culture conditions and transfection procedure to increase the titer of the packaging cells was performed. It is important to note that the selection criteria for the producer clones were based on productivity and clones with the best titers were therefore selected. Using producer clones 3E9, we demonstrated that the volumetric productivity could be increased by inducing the cells at higher cell density. The titer in shake flasks, without medium replacement, was increased 2.5-fold (from 9.0 × 10⁷ to 2.25 × 10⁸ TU/mL) by simply increasing the cell density at induction from 2 million to 5 million cells/mL. As demonstrated previously with producer clones derived from our earlier version of packaging cells (induced with cumate and doxycycline), further titer improvement would most likely be achievable through the use of a perfusion or a fed batch process.^33,34 The packaging cells and producer clones described in the current study were stable for a period of 9 weeks of continuous culture without addition of any selective agents. This would provide sufficient time for cell banking and scale-up production in bioreactors. The doubling time of the packaging and producer cells being around 22 h (data not shown), this would be the equivalent to 68 cell doublings.

As described by other researchers,^{14,16,17,31,32} we have also observed that the addition of sodium butyrate increased significantly the amount of LV produced. In our case, the LV titer for the producer clone was increased by one log (Figure 7) and a similar increased was observed with the packaging cells after transient transfection (data not shown). Sodium butyrate is a histone deacetylase (HDAC) inhibitor and it promotes the decondensation of the chromatin, thus facilitating transcription. This mode of action for sodium butyrate is in agreement with our protein expression analysis, where an increase of Gag, Rev, and VSV-G by the packaging cells was observed. However, another HDAC inhibitor (valproic acid) when tested on the packaging cells did not augment LV production (data not shown). Sodium butyrate also increased the amount LV genomic RNA. This could be due to an enhancement of the transcriptional activity of the CMV promoter that is driving expression of the vector RNA, or because of improved RNA stability or cytoplasmic transport due the presence of additional Rev protein, since the vector genome contains the RRE sequence that interacts with Rev.

One key aspect to successfully generating packaging cells for LV pseudotyped with VSV-G is to use an inducible gene expression system to control VSV-G expression, and to a certain extent the expression of the protease because of their toxicity. The Tet Off,^14,18,28 the Tet On,^17,35 the ecdysome,¹⁹ or a combination of the Tet On and cumate gene expression systems²⁰ have been employed for that purpose. The Tet Off has the disadvantage that the inducer (doxycycline) has to be removed from the culture medium to trigger LV production. This might not be an issue at small scale or using adherent cells, because the culture medium can be readily replaced (by centrifugation or by carefully aspirating the medium from the dish). This is more laborious at large scale (above 10 L, for example), particularly if using suspension culture, since centrifugation, the most common method for medium replacement, cannot be readily scaled up. Furthermore, replacing culture medium with fresh medium adds extra cost to the process. For that reason, to construct our packaging cells, we have chosen a gene expression system that is activated upon addition of inducers without the need of medium replacement. In our previously described packaging cells based on the cumate and Tet On system,²⁰ the induction level of gene expression (ratio of expression level after and before induction; On/Off ratio) was 2,600. This is comparable with the induction level (3,500) observed with clone 7-2 (Figure 2), which was used to construct the packaging cells of the current study. It seems that an induction level in that range is essential to generate efficient packaging for LV pseudotyped with VSV-G, because we have been unable to generate packaging cells by using only the cumate induction system, which is characterized by a lower induction level²⁵ (data not shown). The importance of silencing the expression of VSV-G and Gag/Pol was demonstrated by growing the packaging cells under induced condition. This resulted in a significant reduction of the growth rate as well as in the formation of large cell aggregates. Such undesirable changes would certainly prevent cell culture scale-up (or at least make it difficult) and might have a negative impact on clone stability. This would also reduce transfection efficacy (as cell aggregates are difficult to transfect) and prevent the generation of producer clones (as it would be difficult to clone aggregated cells that grow slowly).

A novel technology we have tested to facilitate clone isolation was to perform limiting dilution into nanowell arrays in conjunction with the scanning and cell picking capability of a robotic cell picker (the CellCelector). Because the scanning system documents that the colonies are derived from single cells, there is no need for subcloning to ensure monoclonality, which, in combination with the one-shot transfection approach employed to integrate simultaneously the packaging genes, has significantly sped up the process of cell line development. It was therefore important to demonstrate that the colonies were indeed derived from single cells. Using pools containing different proportions of GFP- and dsRed-positive cells, we have consistently isolated colonies consisting of a homogeneous population of one type of fluorescent markers, indicating the colonies were derived from single cells. Furthermore, the fact that the proportion of ghost wells (wells without cells on the day of the plating but containing cells a few days later) was lower than 0.01% indicates that the probability of contamination by cells from adjacent wells is low. The clones of the packaging cells were isolated by first generating a mini-pool. The advantage of a mini-pool is that it speeds up by 1 week the cloning procedure. Normally, after transfection and treatment with a selective agent (such as hygromycin), it takes about 3 weeks to generate a pool of resistant cells that can be cloned by limiting dilution. In contrast, it takes only 2 weeks to generate a mini-pool: after transfection, the cells are treated with a selection agent for 1 week and plated into nanowell arrays, then an additional 5 days are needed to allow colony formation and 1 day for cell picking and pooling to make the mini-pool. Due to the presence of cell debris after 1 week of selection, it is difficult to identify single cells after plating into the nanowells. This is the reason the colonies are pooled into a mini-pool, which is then cloned the next day by limiting dilution. Interestingly, the nanowell arrays provided some growth advantages during cloning. When compared with limiting dilution in 96-well plates, the survival rate (as monitored by the percentage of wells with colonies) was 47% higher when performing limiting dilution in the nanowell arrays. The volume of a nanowell is significantly smaller compared with the well of a 96-well plate. Since it is known that various cell types can condition the culture medium by secreting growth factors,^36,37 it is possible that the survival rate was better in the nanowell because the medium was conditioned faster due to its smaller size.

Although several packaging cells for LV have been described (reviewed in Ferreira et al. and Perry and Rayat^5,6), many of them are not inducible and, for that reason, cannot produce LV pseudotyped with VSV-G because of its cytotoxicity. Despite this undesirable property, VSV-G is the most common envelope protein used to pseudotype LV, because it confers a broad tropism³⁸ and improves its stability, thus allowing concentration by ultracentrifugation.³⁹ A few packaging cells for VSV-G pseudotyped LV that grow in adherence and in the presence of serum have been described.^{15,16,18,19,28,35} For the reason mentioned earlier, we favor the development of packaging cells with the capability to grow in suspension culture using serum-free medium. Only two types of packaging cells for VSV-G pseudotyped LV having such properties have been described. One was constructed by our research group using the cumate and Tet On inducible system.²⁰ In contrast to the packaging cells described in the current study, these packaging cells contain a fluorescent gene reporter (GFP) that was used to screen the best clones expressing VSV-G. The other published packaging cell line, which employs only the Tet On inducible system, was constructed more recently by transfecting cells with a single plasmid encoding all the genetic elements necessary to make producer clones, including the gene of interest.¹⁷

The highest titer obtained from our best producer clones for an LV expressing GFP at a cell density of 5 million cells/mL (at the time of induction) and under batch mode was 2.3 × 10⁸ TU/mL, for a specific productivity of 46 TU per cell (at 3 days post induction). Keeping in mind it is difficult to compare titers obtained from different laboratories, using different genes of interest and LV backbones and slightly different titration methods, a titer of 2.3 × 10⁸ TU/mL is comparable with, if not higher than, the best titers reported for a VSV-G pseudotyped LV using either producer cells^{17,18,22,28,33,34,40} or transient transfection.^{8,41,42,43,44,45}

In summary, this study describes a simple approach to generate efficient packaging and producer cells for SIN LV pseudotyped with VSV-G. The cells grow in serum-free suspension culture and are stable for at least 9 to 10 weeks under continuous culture in the absence of selective agent. The cells are therefore amenable for large-scale production of LV for R&D purpose and clinical applications.

Materials and methods

Cells and culture conditions

The 293SF-3F6²⁷ and the 293SF-CymR cells were cultured in SFM4-Transfx-293 medium (Hyclone, South Logan, UT) supplemented with 6 mM L-glutamine (Hyclone). 293SF-CymR/λR-GyrB cells were generated using low-calcium-serum-free medium (LC-SFM) (Gibco, Life Technology Corporation, Grand Island, NY), supplemented with 6 mM L-glutamine and 10 mg/mL rTransferin (Biogems, Westlake Village, CA) and expanded in suspension culture in SFM4-Transfx-293. The packaging cells (293SF-PacLVIIIb) and producer clones were generated using a mixture of 50% LC-SFM supplemented with 6 mM L-glutamine and 10 mg/mL rTransferin and 50% HyCell-Transfx-H (Hyclone) supplemented with 4 mM L-glutamine and 0.1% kolliphor (Sigma-Aldrich, St. Louis, MO). They were maintained in suspension culture in 100% HyCell-Tranfx-H. For suspension culture, the cells were grown in shake flasks (Corning, Oneonta, NY) at 120 rpm with an orbital diameter of 25 mm. The HEK293A cells (ATCC, Manassas, VA) were grown in Dulbecco’s modified Eagle’s medium (DMEM; Hyclone) supplemented with 5% fetal bovine serum (FBS; Hyclone) and 2 mM L-glutamine using tissue culture-treated dishes. All the cell lines were maintained at 37°C in a 5% CO₂ humidified atmosphere.

Plasmids used in the present study

Plasmids were constructed using standard methods of molecular biology. Following amplification in Escherichia coli, the plasmid DNA was purified by chromatography using commercial kits (Qiagen, Hilden, Germany). Plasmid concentration was measured on the Nanodrop spectrophotometer (Thermo Fisher Scientific, Waltman, MA). Plasmid integrity was confirmed by digestion with restriction enzymes, as well as by DNA sequencing (Centre d’expertise et de service Génome Québec, Montreal, QC). The plasmids used for this project were generated as described below.

LV transfer vectors

The transfer vector used to construct an LV expressing GFP regulated by the CMV5CuO promoter (pTet07-CSII-CMV5CuO-GFP) was constructed by replacing the CMV promoter of pTet07-CSII-CMV-GFP²⁰ with the CMV5CuO sequence extracted from pRRL.cppt.CMV5CuO-rcTA²⁵ by digestion with restriction enzymes.

The transfer vector used to construct an LV expressing GFP regulated by the coumermycin-regulated promoter (pTet07-CSII-12xlambda-TPL-GFPq) was constructed by first ordering the sequence of the coumermycin-regulated promoter (12xlambda fused to the CMV-CuO minimal promoter and adenovirus tripartite leader) from a gene synthesis company (GenScript, Piscataway, NJ). This sequence was used to replace the CMVCuO promoter of pKCMVCuO (obtained by removing the Rev gene of pkCMV5CuO-Rev²⁰ by digestion with restriction enzymes), thus generating pKC_12xlambdaCuO-TPL-MCS). The coding sequence for GFP, isolated from pAd-CMV5-GFPq by digestion with restriction enzymes,⁴⁶ was cloned downstream of the 12xlambdaCuO-TPL promoter, thus generating pKC_12xlambdaCuO-TPL-GFP. The 12xlambdaCuO region of the latter plasmid was replaced by the 12xlambda-CMV minimal promoter (without the CuO), which was amplified by PCR using a plasmid encoding this DNA element ordered from a gene synthesis company (GeneArt, Thermo Fisher Scientific). The resulting expression cassette (12xLambda-TPL-GFP) was removed by restriction digestion and was used to replace the CMV-GFP expression cassette of pTet07-CSII-CMV-GFP.²⁰ The nucleotide sequence of 12xLambda-TPL appears in Figure S6.

To construct a transfer vector for a self-inactivating LV encoding GFP regulated by CMV (pNRC-LV1-CMV-GFPq). We first engineered an empty vector (pBV3) using Gibson assembly to combine three synthetic DNA fragments (Integrated DNA Technologies, San Diego, CA) containing a complete backbone of a lentiviral vector (Figure 1B) with a cloning vector derived from pMK (GeneArt, Thermo Fisher). In an effort to optimize the titer potential of this transfer vector, we replaced the XbaI/SalI fragment of pBV3 with a synthetic fragment (CMV5′UTR-HIV-1Ψ-RRE-cPPT, GenBank: FR822201.1) ordered from a gene synthesis company (GenScript) containing the same restriction sites. The resulting empty LV transfer vector was referred to as pNRC-LV1. The CMV-GFP expression cassette was amplified by PCR from pCSII-CMV-GFPq²⁰ and was introduced by Golden Gate assembly using Esp3I sites, thus generating pNRC-LV1-CMV-GFPq.

The transfer vector pCSII-CMV-dsRed was constructed by inserting the coding sequence for dsRed-Express (Takara Bio USA, San Jose, CA) into the multiple cloning site of pCSII-CMV-mcs using restriction enzymes.⁴⁷ It was used to produce an LV encoding dsRed regulated by the CMV promoter (LV-CMV-dsRed), which in turn was employed to generate the 293SF-dsRed pool of cells.

Plasmids used to construct cell lines

A plasmid encoding the cumate repressor (pMPG-CMV5-CymR) was obtained by removing the hygromycin resistance from pMPG-CMV5-CymRopt-Hygro⁴⁸ by digestion with restriction enzymes, followed by ligation to circularize the plasmid.

A plasmid encoding the puromycin resistance gene, pMPG-Puro, was constructed by isolating the pac gene from pTT54⁴⁹ and inserting it into an insert-free plasmid derived from pMPG.⁵⁰

A plasmid encoding the λR-GyrB transactivator regulated by the CMV5CuO promoter, pKCMV5-CuO-λR-GyrB, was constructed by replacing the Rev sequence of pKCMV5-CuO-Rev²⁰ with the λR-GyrB sequence from pGyrb²⁶ using restriction enzymes.

Plasmids encoding the resistance genes for blasticidin (pBlast) and hygromycin (pHygro) cloned into pUC57 were obtained from a gene synthesis company (GenScript). A plasmid encoding the neomycin resistance gene, pMPG/TK∗/neo, was generated by removing the CymR-nls cassette from pMPG/TK∗neo/CymR-nls²⁵ by AscI digestion.

The plasmid encoding VSV-G regulated by the coumermycin-inducible promoter, pVV-13xlambda-TPL-VSVg-Q96H-I57L, was constructed as follows. The sequence of VSV-G (GenBank: ABD73123) was codon optimized for expression in human cells and was synthetized by GenScript. It was subcloned into pkCR5 downstream of the CR5 promoter,²⁵ thus generating pKCR5-VSVg-Q96H-I57L. Plasmid pVV-13xlambda-CMVmin was obtained by replacing the CMV5 promoter of pTT5⁵¹ (with dyad symmetry (DS) and family of repeats (FR) sequences previously deleted) with the 12xlambda-CMV minimal promoter, which had been synthetized by GeneArt (Thermo Fisher) and subsequently amplified by PCR. Sequencing revealed the presence of an additional copy of the lambda operator. A fragment encoding the tripartite leader, adenovirus major late enhancer, and the VSV-G coding sequence from pKCR5-VSVg-Q96H-I57L was cloned into pVV-13xlambda-CMVmin, thus generating pVV-13xlambda-TPL-VSVg-Q96H-I57L. The sequence of the 13xlambda-TPL promoter appears in Figure S6.

To construct pVV-13xlambda-TPL-Rev, the Rev sequence (GenBank: AF033819.3), synthetized by GenScript was first cloned into pkCR5,²⁵ thus generating pkCR5-Rev. The tripartite leader, adenovirus major late enhancer, and Rev sequences were extracted from pKCR5-Rev by digestion with restriction enzymes and used to replace VSV-G gene with similar upstream sequences of pVV-13xlambda-TPL VSVg -Q96H-I57L.

To generate a plasmid encoding the HIV Gap/Pol gene under the control of the CAG promoter, Gag/Pol sequence (accession number EU541617.1) was synthetized by GenScript. It was then cloned into a plasmid derived from pKCMV-B43⁵² that had been previously modified by introducing the CAG promoter and intron⁵³ in place of the CMV-B43 cassette, thus giving rise to pCAG-Gag/polIIIb.

Production and titration of lentivirus

A stock of LV encoding CMV-regulated GFP, LV-CMV-GFP, was produced using a producer clone.³³ The culture medium containing the LV was collected at 48 h post induction with cumate (Sigma-Aldrich) and doxycycline (Sigma-Aldrich) and was clarified by low-speed centrifugation, and subsequent filtration through a 0.45-μm pore size HT Tuffryn membrane (Pall Corporation, Fajardo, Puerto Rico). The LV was then concentrated by ultracentrifugation at 100,000 × g for 2 h at 4°C on a 20% sucrose cushion as described.⁵⁴ The pellet containing LV was resuspended in 1 mL of culture medium supplemented with 1% FBS and was frozen at −80°C. LV-CMV5CuO-GFP and LV-12xlambda-TPL-GFP were produced by transient transfection of packaging cells (293SF-PacLV #29-6) with pTet07-CSII-CMV5-CuO-GFP and pTet07-CSII-12xlambda-TPL-GFP respectively, as described previously.²⁰ The LV was concentrated by ultracentrifugation on a sucrose cushion as described above and was used to transduce 293SF-PacLV #29-6 cells to generate pools of producer cells for each LV. The pools were amplified in suspension culture using LC-SFM + 1% FBS, and LV production was induced by the addition of 1 μg/mL of doxycycline and 30 μg/mL of cumate. The LVs were harvested at 48 and 72 h post induction, concentrated by ultracentrifugation on a sucrose cushion, and frozen at −80°C as described above. The LV-dsRed stock was produced in suspension culture by transfecting the 293SF-PacLVIIIb packaging cells (described below) with 0.4 μg/mL of pCSII-CMV-dsRed using PEI (PEIpro, Polyplus, Illkirch, France) at a ratio of PEI to DNA of 3. The cells were induced by addition of 80 μg/mL cumate and 10 nM coumermycin (Promega, Madison, WI) at 6 h post transfection, and 8 mM sodium butyrate (Sigma-Aldrich) was added 18 h post transfection. The LVs were harvested at 48 and 72 h post induction, concentrated by ultracentrifugation on a sucrose cushion, and frozen at −80°C as described above. LV-CMV-GFP, LV-CMV5CuO-GFP, and LV-CMV-dsRed were titrated by transducing HEK293A cells. LV-12xlambda-TPL-GFP was titrated by transducing 293ArtTA cells.²⁰ For transduction, LV samples were serially diluted in DMEM supplemented with 8 μg/mL of polybrene (hexadimethrine bromide, Sigma-Aldrich) and incubated at 37°C for 30 min. Transduction was performed by removing the culture medium, adding 50 μL of diluted LV to 20,000 HEK293A or 293ArtTA cells previously plated in 96-well plate, and incubating overnight at 37°C. Then 150 μL of culture medium was added to each well and the cells were incubated for an additional 48 h prior to flow cytometry analysis using an LSR Fortessa (BD Biosciences, San Jose, CA) or an FC 500 MPL (Beckman- Coulter, Mississauga, ON, Canada) flow cytometer to quantify GFP or dsRed expressing cells as described previously.²⁰

Generation of 293SF-CymR/λR-GyrB cells

293SF-3F6 cells were transfected with 1 μg/mL of pMGP-CMV5-CymR and pMPG-Puro at a DNA ratio of 9:1 using Lipofectamine 2000 CD (Invitrogen). Both plasmids were previously linearized with MfeI. Forty-eight hours later, the cells were diluted in SFM4-Transfx-293 medium containing 0.4 μg/mL of puromycin (Sigma-Aldrich) and plated in 96-well plates at 5,000 and 10,000 cells/wells. Selected clones expressing CymR (see below) were further subcloned by plating in semi-solid medium at 1,000 cells/mL and 3,000 cells/mL as described.⁵⁵ The semi-solid medium was made by mixing Flex methylcellulose (StemCell Technologies, Vancouver, BC, Canada), with 2× SFM4-Transfx-293, 6 mM L-glutamine, and 2.5% ClonalCell ACF CHO supplements (StemCell Technologies). Colonies were isolated using the CellCelector (ALS, Jena, Germany) and were transferred into a 96-well plate. Clones and subclones were split into two populations: one was used for analysis, while the other one remained untouched for clone expansion and banking. To screen for the presence of CymR, the cells were transduced in 96-well plates with LV-CMV5CuO-GFP and were analyzed by flow cytometry for GFP expression after induction with 100 μg/mL of cumate (Sigma-Aldrich). Clones selected for the highest GFP intensity were subjected to secondary screens for GFP expression with and without cumate.

The 293SF-CymR subclone with the best induction level (clone 198-2) was used to introduce the λR-GyrB gene. The cells were transfected with 1 μg/mL of pKCMV5-CuO-λR-GyrB and pBlast that had been linearized with restriction enzymes at a DNA ratio of 9:1 using 1 μg/mL of PEI. At 48 h post transfection, the cells were diluted with LC-SFM medium containing 7 μg/mL of blasticidin hydrochloride (Enzo Life Sciences, Farmingdale, NY) and were transferred into 96-well plates at a density of 1,000 cells/well. After 1 week, blasticidin concentration was increased to 10 μg/mL. The best clones expressing λR-GyrB (identified as described below) were subcloned in 96-well plates by limiting dilution at 0.3 cell/well and 1.0 cells/well in LC-SFM medium without selection. The clones and subclones were screened for the expression of λR-GyrB by transducing them with LV-12xlambda-TPL-GFP and were analyzed by flow cytometry for GFP expression after induction with 100 μg/mL of cumate and 10 nM coumermycin.

Gene regulation of 293SF-CymR and 293SF-CymR/λR-GyrB clones

The best subclones of 293SF-CymR and 293SF-CymR/λR-GyrB cells were transduced with LV in the presence of 8 μg/mL of polybrene. 293SF-CymR were transduced at an MOI of 20 with LV-CMV5CuO-GFP, and the cells were induced by the addition of 100 μg/mL of cumate the next day. The 293SF-CymR/λR-GyrB were transduced with LV-12xlambda-TPL-GFP at an MOI of 5 and the cells were induced by addition of 100 μg/mL of cumate and 10 nM coumermycin the next day. The cells were fixed and processed for flow cytometry analysis at 72 h post transduction as described above. The fluorescent index (percentage of GFP-positive cells multiplied by the mean GFP signal of the GFP-positive population) was compared in the presence and absence of induction. 293SF-3F6 cells transduced with LV-CMV5CuO-GFP or with LV-12xlambda-TPL-GFP under the same conditions and analyzed simultaneously by flow cytometry were used as control.

Generation of 293SF-GFP and 293SF-dsRed pools

Pools of GFP- (293SF-GFP) and dsRed (293SF-dsRed)-positive cells were generated by transducing 293SF-3F6 cells with LV-CMV-GFP and LV-CMV-dsRed as described below. Five hours before adding the LV suspension, 293SF-3F6 cells were seeded in 24-well plates using 500 μL of LC-SFM/HyCell mixture per well under static conditions. The LV suspension was prepared by mixing 200 μL of LVs with a final concentration of 8 μg/mL of polybrene and incubated at 37°C for 30 min before adding it directly to the cells. The cells were transduced at an MOI of 100 (LV-CMV-GFP) and 50 (LV-CMV-dsRed). The next day, 500 μL of LC-SFM/HyCell mixture was added to the cells. The cells were expanded in six-well plates, T-25 flasks, and finally cultivated in suspension culture in shake flasks using an LC-SFM/HyCell mixture. The pools were analyzed by flow cytometry to determine the percentage of positive cells.

Cell plating into nanowell arrays and clone isolation

To avoid air trapping into the nanowells arrays (Plate S200-100 100K, 24-well; ALS, Jena, Germany), the wells were first filled with 300 μL of anhydrous ethanol. Ethanol was washed four times by adding and removing 2 mL of PBS. Similarly, PBS was then washed twice with 2 mL of a mixture of LC-SFM/HyCell medium, and 400 μL of medium was left in the wells. Before seeding the cells, they were gently resuspended and homogenized by pipetting them up and down. The cells were counted and seeded into the nanowell array in a final volume of 500 μL per well at the appropriate cell density. After plating, the nanowell arrays were centrifuged at 300 × g for 3 min.

Immediately after seeding the cells into nanowell arrays, they were scanned using the bright-field 10× objective of the imaging system (Leica DMIL inverted microscope equipped with a monochrome Retiga Exi camera) of the CellCelector cell picking system (ALS). The wells containing a single cell were identified using the AnalySIS 3.1 software of the instrument. The nanowells were scanned again at 3–5 days post plating for colony identification. Additional scans using the GFP and dsRed fluorescence channels were performed for the experiments using the 293SF-GFP and 293SF-dsRed cells. Colonies derived from single cells (identified on the day of plating, day 0) were transferred into 384-well plate (Corning) using the CellCelector robotic arm equipped with the single-cell tool with an 80-μm capillary. For the experiment involving the 293SF-GFP and 293SF-dsRed cells, colonies transferred into 384-well plates were scanned using the 4× objective in bright field and using the GFP and dsRed channels immediately after picking and 10 days later. Cells were also amplified and fixed with 2% formaldehyde and analyzed by flow cytometry to measure the percentage of GFP and dsRed-positive cells.

Generation of packaging cells

293SF-CymR/λR-GyrB cells (clone 7-2) were transfected in six-well plate suspension culture using PEI and 1 μg/mL of DNA mixture (PEI:DNA ratio 2:1) of pCAG-Gag/polIIIb, pVV-13lambda-TPL-VSVg-Q96H-I57L, pVV-13lambda-tpl-Rev, and pHygro at a proportion of 40%, 25%, 25%, and 10% respectively. Plasmids were linearized by digestion with restriction enzymes before transfection. At 24 h post transfection, the cells were diluted to 0.43 × 10⁶ cells/mL with LC-SFM/HyCell medium containing 80 μg/mL of hygromycin. After 8 days, cells were plated into the nanowell arrays at 1.4 cells per nanowells at a hygromycin concentration of 50 or 25 μg/mL. Selection was continued in nanowells for 4 days. The wells were scanned and 173 colonies were picked and pooled together in the same well of a 96-well plate using the CellCelector. The mini-pool was cloned 6 days later by dilution into nanowell arrays at a cell density of 0.6 cells per nanowell in medium supplemented with 20 μg/mL hygromycin. Wells with single cells were identified, and 348 colonies were isolated 4 days after plating using the CellCelector and transferred into a 384-well plate.

The resulting clones were analyzed for the production of an LV expressing GFP regulated by CMV (LV-CMV-GFP) after transient transfection. LV production during the clone screening in 96- and 24-well plates was performed under static conditions by transfection with PEI and pCSII-CMV-GFP²⁰ at a DNA concentration of 1 μg/mL (ratio PEI:DNA of 2:1) Cells were induced 5 h later by adding fresh medium supplemented with cumate, coumermycin, and sodium butyrate to a final concentration of 80 μg/mL, 10 nM, and 7 mM respectively. LV were titrated after transduction of HEK293A cells at 48 h post transfection as described above. Functional titer in the culture medium produced in 96- and 24-well plates was estimated by observation under a fluorescent microscope or titrated by flow cytometry as described above. Production in six-well plates was performed in suspension culture using 1.0 × 10⁶ cells/mL prepared in 2.7 mL of fresh medium. Cells were transfected immediately with PEI and pCSII-CMV-GFP (ratio PEI:DNA of 2:1) using 0.4 μg/mL of DNA. Cells were induced 6 h later by adding 200 μL of medium supplemented with cumate and coumermycin to final concentration of 80 μg/mL and 10 nM respectively. Sodium butyrate was added 18 h later at a final concentration of 7 mM. LV was titrated by transduction of HEK293A cells at 72 h post transfection by flow cytometry (as described above).

Generation of producer clones

Clone 3D4 from the packaging cells (293SF-PacLVIIIb) was transfected in six-well plates in suspension culture using PEI and 0.8 μg/mL of pNRC-LV1-CMV-GFP and 0.2 μg/mL of pMPG/TK∗/neo, previously linearized by digestion with restriction enzymes (ratio PEI:DNA of 2:1). At 36 h post transfection, the cells were diluted at 0.5 × 10⁶ cells/mL with LC-SFM/HyCell medium containing 400 μg/mL of G418. Cells were maintained under selection for 18 days and were then plated into nanowell arrays at a cell density of 0.6 cell per nanowell in LC-SFM/HyCell medium (without G418). Single cells in nanowell arrays were identified at day 0 and 380 colonies were picked using the CellCelector and transferred into a 384-well plate in LC-SFM/HyCell medium supplemented with 400 μg/mL G418. Clones were analyzed for LV-CMV-GFP production 72 h after induction with 80 μg/mL of cumate, 10 nM coumermycin, and 7 mM sodium butyrate. They were first tested in static conditions in 96-well and 24-well plates and finally in suspension in six-well plates at a density of 1.0 × 10⁶ cells/mL. LV-CMV-GFP was titrated by flow cytometry after transduction of HEK293A cells as described above.

Stability study of packaging and producer clones

Vials of a research cell bank were thawed and cells were maintained in suspension culture using baffled shake flasks. LV-CMV-GFP production was assayed at different time points after 1 to 10 weeks in culture. Production was performed in suspension culture in six-well plates at 1.0 × 10⁶ cells/mL prepared in 2.7–3 mL of fresh medium. For the packaging cells, cells were transfected with pCSII-CMV-GFP and induced as described above. For the producer clones, only induction was performed. LV was titrated at 72 h post induction as described above.

Effect of induction on cell growth of packaging cells

Cells from clone 3D4 and 293SF-CymR/λR-GyrB were inoculated in triplicate in suspension culture in a six-well plate at 0.15 × 10⁶ cells/mL in a final volume of 3 mL in the presence or absence of 80 μg/mL of cumate and 10 nM coumermycin. The cells were split to 0.15 × 10⁶ cells/mL after 4 days and maintained in culture for an additional 6 days. The cells were counted with the automated cell counter CellDrop (DeNovix, USA) using the default protocol with trypan blue to assess cells viability.

LV production from the producer clones in shake flasks

Producer clone 3E9 was seeded into three different groups, each consisting of three 125-mL shake flasks, two groups at 0.3 × 10⁶ cells/mL and one group at 0.6 × 10⁶ cells/mL. Three days later, the first two groups were diluted at 0.8 × 10⁶ cells/mL and the third group at 2.5 × 10⁶ cells/with fresh medium. The first two groups were induced the next day at 2.0 × 10⁶ cells/mL and the third was induced 48 h later at 5 × 10⁶ cells/mL with 80 μg/mL of cumate and 10 nM of coumermycin. Butyrate was added 18 h post induction at 7 mM except for the first group, where no butyrate was added. LV was titrated by transduction of HEK293A cells at 72 h post induction by flow cytometry as described above.

Western blotting for Gag, VSV-G, and Rev

Packaging cells (293SF-PacLVIIIb, clone 3D4) were tested by western blot for the expression of Gag, VSV-G and Rev protein. On the day of induction, cells were centrifuged and resuspended in HyCell medium to a final concentration of 1.0 × 10⁶ cells/mL in 125-mL shake flasks. Induction was done 1 h later using 80 μg/mL of cumate and 10 nM coumermycin. As negative control, 293SF-CymR/λR-GyrB cells were used. Two groups of cells were prepared with and without the addition of 8 mM sodium butyrate at 18 h post induction. Cell cultures were harvested at 0, 24, 48, and 72 h post induction. The cells were harvested by centrifugation and the cell pellet was transferred to −80°C. For western blot analysis, the cells were thawed and lysed with radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 0.1% SDS, 1% NP-40, 0.25% Na deoxycholate). After 30-min incubation on ice, the samples were sonicated and the lysates were clarified by centrifugation. Protein concentration was determined by BIO-RAD DC protein Assay (Bio-Rad Laboratories, Hercules, CA). The same amount of total protein (40 μg) was separated through a NuPAGE 4–12% Bis-Tris Gel, (Invitrogen, Carlsbad, CA) and analyzed by western blotting using a rabbit polyclonal HIV p24 Ab (ProSci, Poway CA, catalog no. 7313), a rabbit polyclonal anti-VSV-G tag antibody (ab83196, abcam, Waltham MA), or anti-HIV1 Rev mouse monoclonal antibody (ab85529, abcam), followed by a horseradish peroxidase-conjugated donkey anti-rabbit immunoglobulin (Ig) G antibody or a horseradish peroxidase-conjugated sheep anti-mouse IgG antibody (GE Healthcare, Buckinghamshire, UK). The signal was revealed by chemiluminescence using the enhanced chemiluminescence (ECL) western blotting detection reagents (PerkinElmer, Boston MA) and analyzed with a digital imaging system (image Quant LAS 4000 mini biomolecular imager, GE Healthcare, Buckinghamshire UK).

Quantification of Gag, VSV-G, and Rev expression

To quantify the induction level for Gag and VSV-G, the experiment described above was repeated in triplicate, but only samples harvested at 48 h post induction (with and without butyrate) were used for the analysis. The samples were processed for western blot as described above and scanned with a digital imaging system (600C Azure Biosystems, Dublin, CA). After scanning the Gag and VSV-G signal, the membrane was stripped (Abcam protocol for mild stripping), blocked as described above, and incubated with a mouse anti β-actin antibody (clone 137CT26.1.1, NSJ Bioreagents, San Diego, CA) followed by a horseradish peroxidase-conjugated sheep anti-mouse IgG. The Azure Spot version 2.2.167 software was used for quantification of the signal. The signal for actin was used to normalize the signal for Gag and VSV-G. For quantification, the signal measured in the negative control (lysate of 293SF-CymR/λR-GyrB) was subtracted from the signal obtained with Gag and VSV-G.

Quantification of LV vector RNA by RT-PCR

Cells from producer clone 3E9 grown in six-well plates at a density of 1.5 × 10⁶ cells/mL were induced or not by supplementing the medium with 10 nM coumermycin and 80 μg/mL of cumate. Sodium butyrate was added in some wells at 18 h post induction to a final concentration of 7 mM. The packaging cell line, clone 3D4, was used as a negative control. Total RNA was extracted at 48 h post induction with and without sodium butyrate, using the RNeasy Plus mini kit (Qiagen, catalog no. 74134) according to the manufacturer’s recommendation. RT was done using the Quantitect reverse transcription kit (Qiagen, catalog no. 205311) starting with 500 ng of RNA. The qPCR reaction was done using 2 μL of the RT reaction, 10 μL of 2× PrimeTime Gene expression MasterMix (Integrated DNA Technologies, IDT, catalog no. 1055771), LV qPCR primers/probe set (from IDT, NRC-LV1-Psi-F, 5′-TGAAAGCGAAAGGGAAACCAG-3′, NRC-LV1-Psi-R; 5′-CACCCATCTCTCTCCTTCTAGCC-3′, NRC-LV1-Psi-Pr [FAM], 5′-AGCTCTCTCGACGCAGGACTCGGC-3′) in a final volume of 20 μL. The qPCR was performed on the QuantStudio 5 (Applied Biosystems, Waltham, MA) using the following thermocycling conditions: initial enzyme activation at 95°C for 3 min followed by 40 cycles of 95°C for 15 s, and 60°C for 1 min. The standard curve was performed with a 10-fold serial dilution of a linearized plasmid for an LV with an identical packaging sequence to pNRC-LV1-CMV-GFP. The data were analyzed using the QuantStudio analysis software, and the number of LV copies was calculated by mapping the samples threshold cycle number to the standard curve. The experiment was performed in triplicate.

Statistical analysis

The data are presented as the means ± standard deviation (SD). The number (n) of independent samples used for the analysis is indicated in the figures. Means were compared using a Student’s t test. The difference between the means was considered significant at p < 0.01 (∗∗) or p < 0.05 (∗).

Data availability

The data supporting the finding of this study can be found in the main text or the supplemental information. Additional information may be made available upon reasonable request to the corresponding authors.