Keywords
Selection system, asparaginase, glutamine synthetase, Chinese hamster ovary cells, biotherapeutic production, mammalian cell line
The integration of exogenous DNA into cultured cells is often facilitated by co-introduction of the desired DNA alongside a selectable element and growth under the corresponding selective pressure (e.g., transfection of the desired gene + a gene imparting antibiotic resistance combined with growth in medium containing the antibiotic). In the biopharmaceutical industry, metabolic selection systems that restore nutritional prototrophies are routinely used to generate mammalian cell lines producing high quantities of life-saving biotherapeutic protein drugs. Dihydrofolate reductase (Dhfr) or glutamine synthetase (Gs) are the most commonly leveraged metabolic selectable markers (Cockett et al., 1990; Kaufman & Sharp, 1982). However, the cell line generation process using these is time-consuming and laborious, often requiring one (Gs) or several (Dhfr) rounds of gene amplification driven by the addition of inhibitory compounds (e.g., methotrexate and methionine sulfoximine) during selection as well as the screening of 100s to 1000s of clones to identify clones with the desired production and quality profiles. It has been shown that utilizing these two systems simultaneously increases the probability that a highly productive cell will be generated–as well as improving the maximum product titer (Li et al., 2010). Recent work has demonstrated the feasibility of additional metabolic selection systems in CHO (Budge et al., 2021; Capella Roca et al., 2019; Pourcel et al., 2020; Sun et al., 2020; Zhang et al., 2020) with one study (Zhang et al., 2022) showing that using 8 selectable markers simultaneously can significantly increase the productivity of the resulting cell lines (although the cells grew very slowly). The orthogonality of these new selection systems (e.g., each requiring the dropout of a different medium component for selective pressure) and/or need for multiple genetic edits led us to explore whether it was possible to increase the selective stringency of glutamine deprivation in a simpler manner, hereby enhancing the selective pressure of one of the most established tools in the clinical cell line generation workflow, without requiring changes from established selection conditions.
Using a CRISPR-Cas9 knockout screen targeting metabolic genes, we identified asparaginase (Aspg) as putatively essential for CHO cell growth in medium lacking glutamine (Karottki et al., 2021). As Gs was also identified as essential in that condition, we hypothesized that a dual selection system based on Aspg together with Gs could enhance selection without requiring alteration of the selective pressure.
To confirm the importance of Aspg for growth in glutamine-free media and assess its viability for use as an additional selectable marker simultaneously with Gs, we generated three clonal knockout (KO) cell lines using CRISPR/Cas9: Gs-KO, Aspg-KO, and Gs/Aspg-KO. Knockouts had verified frameshift insertions or deletions in all alleles and also showed decreased mRNA expression (Supplementary Figure 1). Both clones lacking Aspg showed decreased maximum viable cell density when grown in a glutamine-containing medium (Supplementary Figure 2A) but with comparable growth rates to the Gs KO cell line (Supplementary Table 1). When grown without glutamine, Aspg knockout cells showed negligible growth, but remained viable. Gs/Aspg-KO cells, on the other hand, showed a dramatic decrease in cell viability–even more quickly than Gs-KO cells (Supplementary Figure 2B). This suggested that a double selection system using Gs and Aspg simultaneously would be more stringent than Gs alone–while still using only glutamine deprivation as the sole selective pressure.
We generated Enbrel-producing cells from the different knockout cell lines via 5 different transfections: 1) GS-KO cells with a Gs+Enbrel plasmid, 2) GS-KO cells with both Gs+Enbrel and Aspg+Enbrel plasmids, 3) Aspg-KO cells with Aspg+Enbrel plasmid, 4) Aspg-KO cells with both Gs+Enbrel and Aspg+Enbrel plasmids, and 5) double Gs/Aspg-KO cells with both Gs+Enbrel and Aspg+Enbrel plasmids. We tested selection in both static minipools (192/transfection) and bulk suspension format (duplicates in 6 well suspension, permitting quantification of recovery profiles). Following recovery, surviving minipools were split 1:2 and evaluated for terminal cell count and titer. After 5 days of culture, minipools derived from clones with Aspg knockouts showed lower cell density and product titer, however, minipools derived from the Gs/Aspg double knockout transfected with both plasmids showed ~3-4-fold higher median cell-normalized product titer than Gs knockout derived minipools (Figure 1, Supplementary Table 2). We observed no change in recovery timelines in the bulk suspension format and, after characterization of recovered pools in batch culture, the Gs/Aspg double knockout cells transfected with both plasmids again showed decreased growth, but significantly (~16-fold) higher titer and specific productivity (Supplementary Figure 3B and Supplementary Table 3).
We then tested if minipools could obtain improved performance after being transitioned to suspension culture. Top minipools from all transfections were expanded and characterized in 6 well suspension culture. The trend of higher titer and specific productivity in Gs/Aspg double knockout derived minipools was maintained, but minipools derived from cells with Aspg knocked out showed low VCD (Supplementary Figure 4). We thus continued expanding the top 3 Gs/Aspg knockout derived and Gs knockout derived minipools (based on titer) to test if prolonged time in suspension culture would improve the performance of the former.
After expansion in shake flask culture, the growth and viability of Gs/Aspg double knockout derived minipools were still decreased compared to minipools derived from Gs knockout cells (Figure 2A), but both were improved compared to their performance in 6-well plates (Supplementary Figure 4). We again saw significantly higher production of Enbrel, both in titer (2-4-fold higher in the best performing Gs/Aspg-KO derived minipool) and specific productivity (10-15-fold improvement in Gs/Aspg-KO derived minipools) (Figure 2B and Supplementary Table 4).
The improvement in growth and viability from 6-well to shake flask led us to explore whether further adaptation of minipools derived from Gs/Aspg knockout cells would improve performance in selection conditions designed for Gs knockout derived cells. Following ~1 month of adaptation (Supplementary Figure 5), we evaluated the top minipool from each transfection: The Gs/Aspg KO derived minipool showed significant improvements in growth and viability while still outperforming the Gs knockout derivatives in titer and specific productivity (Figure 3 and Supplementary Table 5). We further assessed the long-term stability of the dual selection strategy and found that after an additional month of passaging, minipool performance remained stable (Figure 4 and Supplementary Table 6).
Finally, we explored the cause of poor growth in the Gs/Aspg-KO derived pools. Cell growth in this selection system depends on the rescue of both knocked out enzymes (Gs and Aspg) through the uptake and integration of both transfected plasmids. It is possible that the observed low cell growth results from low expression of either or both plasmids after selection. However, both Gs and Aspg expression levels in the Gs/Aspg-KO derived pools were at least as high as that of Gs-KO derived pools following selection and recovery, prior to adaptation (Supplementary Figure 6); thus, the expression should be sufficient for robust growth. As adaptation partially recovered growth (and considerably improved viability) we anticipate that additional media and/or platform optimization (e.g., altering the plasmid ratio) could further improve the performance of this system.
The dual Gs/Aspg selection system thus is an intriguing option to generate more highly productive cell lines. As such it only requires a single additional genetic edit to the starting cell line and does not require changes to the traditional Gs-based selection workflow. Furthermore, it has the potential to be used as an alternative system for not only the production of proteins but also the expression of several genes of interest without the use of antibiotics (as seen with Dhfr/Mtx co-selection (Lee et al., 2018)). Continued work with cells generated by this approach, e.g., single-cell cloning, expansion, and characterization in fed-batch bioreactors, will further demonstrate the value of this system for cell line generation for biotherapeutic protein production.