2.1. Plackett-Burman Media Screening
Four Plackett-Burman styled experiments were conducted to rapidly screen media components to support suspension growth in Vero cells. In total, 62 compounds were tested over 96 formulations in the four different Plackett-Burman experiments (see Supplementary Table S1). From this a chemically defined adherent and a suspension medium formulation (CDM) was made for Vero cells. The medium formulations were improved iteratively with each subsequent Plackett-Burman experiment, and the base formulation for the next experiment was designed from the previous set of results. An example of the results from a Plackett-Burman experiment designed for suspension media is illustrated in Fig. 1a, where the effect of each component on growth rate is shown. Through this experiment, the effect of purines and pyrimidines (adenine, guanosine, uridine, and HT supplement) can be seen having overall a positive effect on growth rate, with the exception of the HT supplement which was supplemented at too high of a concentration for the highest factor level. Another category of compounds that improved the growth rate was metals. In addition to metals such as calcium, magnesium, iron, copper and zinc, various trace metals were added to the medium formulation. Specifically cobalt, maganese, molybdenum, silicate, selenite, nickel, tin, and vanadium, which are normally excluded from basal media, but can be found in serum, were critical for cell viability and growth 9. Trace metals are important cofactors in various enzymes, and these metals can be found in deionized water, but not in ultra pure water. One of the drawbacks of adding metals to protein-free media, is that metals are pro-oxidants and can lead to the oxidation of media if antioxidants or metal chelators are not present. Antioxidants such as α-tocopherol (Vitamin E), ascorbic acid (Vitamin C), glutathione, citric acid, and pyruvate were added to the medium. Of these compounds, all of them were beneficial at high concentrations, with the exception of glutathione, which had a detrimental effect on cell growth at concentrations greater than 3.3 µM (Supplementary Figure S3). All the experiments and the concentration of each factor level can be found in the Supplementary Figures S1-S4.
Growth factors, polyamines, vitamins and steroids were investigated to increase the growth rate. Surprisingly, not all growth factors had a positive effect on the growth rate of Vero cells, insulin-like growth factor (IGF) had no poistive effect, while recombinant epidermal growth factor (rEGF) was required for growth (Fig. 1 and Supplementary Figure S1). More recent media formulations developed for Vero cells have used rEGF 10,11 rather than fibroblast growth factor (FGF). Cinatl et al., found that Vero cells could grow well on polyvinyl formal flasks in a protein free media that contained progesterone 12. It is likely that rEGF could be replaced by a combination of steriods to create a protein-free medium for Vero cells.
The adherent chemically defined medium contained 10 ng/mL of rEGF as the only protein present in the final formulation and achieved a doubling time of 32 hours, compared to OptiPRO SFM (35 hours) and NutriVero Flex10 (34 hours) (Fig. 1b). Epidermal growth factor (EGF) is commonly added to chemically defined cell culture media because the EGF receptor (EGFR, also known as ErbB1) is expressed on almost all cell types. Plant hydrolysates have been known to mimic growth factor-like effects for many different cell types at low cost and is animal component-free (ACF), which makes them a good alternative to supplementing with individual growth facotrs 13–15. OptiPRO SFM was chosen as a comparison because it is commonly used in industry as a medium to produce vaccines using Vero cells. This medium is animal component-free but it does contain a very low protein concentration of plant hydrolysates (≤ 10 µg/mL) 16. NutriVero Flex10 is another Vero cell medium that is commercially available, and unlike OptiPro, it is chemically defined, as well as ACF, making it a better comparison to the medium developed in this paper.
Overall, none of the ACF media formulations (NutriVero Flex 10, OptiPRO SFM, or CDM2) grew as fast as serum-containing DMEM/F12 with 10% FBS. Therefore, one could infer that there are still some components missing from ACF media such as cell signaling molecules that are in low concentrations, or growth factors that are missing from serum-free media. A study conducted by Desai et al., tracked the growth factors that were produced by Vero cells 17,18. They found that Vero cells excreted platelet derived growth factor (PDGF), interleukin 6 (IL-6) and leukemia inhibitory factor (LIF), but were unable to detect EGF or active TGF-β. Interestingly, Guo et al., were only able to find EGFR on Vero cells when they conducted a proteomic analysis of the membrane proteins 19. This may have been due to the limited annotation of the Vero genome and surface proteins, or that the EGFR in Vero cells can interact with many different ligands. Nevertheless, rEGF did stimulate proliferation of Vero cells adherently in the chemically defined medium, but the literature suggests that other growth factors could be used in addition, or to replace rEGF.
2.2. Suspension Medium Development
Cells were slowly adapted from DMEM/F12 + 10% FBS medium in adherent culture to the new medium formulations, and as the amount of calcium and magnesium decreased, the cell growth rate was also observed to decrease dramatically. At the lowest concentration range of 0.1 mM calcium, Vero cells were still able to adhere to the tissue-culture treated Tflasks; therefore, cells were transferred to non-treated Tflasks. Through subsequent passaging in non-treated Tflasks, Vero cells detached and formed large aggregates (Fig. 2a). After 90 days of culturing in non-treated Tflasks, the cells were moved onto a shaker at 40 rpm in an effort to break up the cell aggregates. Only one formulation supported viable single cells (Formula 17).
Sodium bicarbonate (NaHCO3) had a large effect on the doubling time, which can be seen in Fig. 2b. There are two major groups in Fig. 2b, one group has a large doubling time (> 1,000 hours), and one with a shorter doubling time (~ 500 hours). The faster growing group contained 2.2 g/L of sodium bicarbonate, while the slower group had 1.2 g/L. For 5% CO2 incubators, the concentration of NaHCO3 should be between 1.2 and 2.2 g/L, and 3.7 g/L for 10% CO2 incubators 20. For our medium formulation, 2.2 g/L of sodium bicarbonate provided extra buffering. Even with the improvements seen from adding more NaHCO3, the doubling time of these cells was very lengthy (doubling time of 20 days), therefore further medium formulations focused on decreasing the doubling time.
One of the major differences between adherent media and suspension media is the concentration of calcium and magnesium. These two ions are important for cell adhesion proteins such as cadherins, which require a certain concentration (> 55 µM) of calcium to maintain the correct rigid conformation to remain active 21. Calcium is also involved in many cellular processes besides cell adhesion 22, such as cell signaling 23, enzyme activity, and apoptosis 24. Healthy cells maintain a large concentration gradient between the cytosol (0.1 µM) and extracellular space (1–2 mM) 22. Magnesium is the second (after potassium) most abundant cation inside the cell and ranges from 17–20 mM in most mammalian cells 25. Most of the magnesium is bound to various cellular structures, and only about 0.8–1.2 mM is free Mg2+ 25. It is essential for the most basic functions in the cell, for example, for ATP to be biologically active. It is most likely that the reduction of these ions in the suspension medium (Formula 17) had detrimental effects on cell signaling and enzymatic activity. Subsequently, Formula 17 was modified to have the same concentration of calcium and magnesium ions as DMEM/F12 for comparison purposes, and was called CDM2. Formula 17 was thus renamed ‘CDM2 low calcium/low magnesium’ henceforth.
2.3. Growth with Other Cell Lines
To investigate if CDM2 could support growth of other cell lines, MDCK, CHO-K1 and HEK293T cells were sequentially adapted to CDM2 from serum-containing basal medium using static Tflasks. HEK293T and MDCK cells were adapted over 2 weeks, while CHO-K1 cells took 3.5 weeks to adapt to CDM2. Given that CDM2 contains the same amount of calcium and magnesium as basal DMEM/F12, this indicated that cells could still grow in suspension with higher levels of the divalent cations. Figure 3 demonstrates how the cells grew in CDM2 compared to commercial media, and serum-containing basal media, along with the maximum cell densities achieved when culturing the cells in suspension. For CHO-K1, HEK293T and MCDK cells grown in CDM2, cells started to lose their viability after approximately 4 days. After day 3, a significant color change in the medium could be seen indicating that the media was becoming acidic. Given that this medium was developed for adherent Vero cells, the ability to grow 3 other cell lines in suspension was a surprising result that led to the question of what is different about Vero cells that prevents them from growing well in suspension.
2.4. RNA-seq experiment to cellular expression changes
To identify the reasons for the reduced growth of the Vero cells in suspension in a defined medium, a transcriptomic analysis was performed to identify expressional changes. This data helped to identify metabolic processes and transcription factors that could lead to a decreased doubling time in the chemically defined medium. The RNA-seq dataset comprised three groups of cells; 1) Vero cells grown adherently in DMEM/F12 + 10% FBS, 2) Vero cells grown adherently in CDM2 with added calcium and magnesium, or 3) Vero cells grown in suspension in CDM2. Cells grown adherently in DMEM/F12 medium (Control, or ‘Con’) and cells grown adherently in CDM2 with additional calcium and magnesium (Adherent, or ‘Adh’) were sequenced and compared to cells grown in suspension in CDM2 (Suspension, or ‘Sus’). A summary of the alignment statistics is provided in Supplementary Table S2. Over 96% of the reads for each sample were uniquely aligned to the Chlorocebus sabaeus (C. sabaeus) genome. After filtering the identified genes for an expression of at least 1 CPM in four samples, 11,135 genes were considered as expressed for further analysis (Supplementary Table S3). Due to the limited information about cellular pathways in C. sabaeus, the Ensembl database was used to identify H. sapiens homologs for the expressed genes. This allowed a more detailed gene set enrichment analysis. The complete results of the differentially expressed gene analysis are in Supplementary Table S4-S7.
The variability within the dataset was analyzed with: 1) a principal component analysis (PCA) of the 500 most variable genes; and 2) a Pearson correlation of 3122 known, expressed housekeeping genes (Supplementary Figure S5). The Pearson correlation of housekeeping genes controlled for outliers with unanticipated changes in housekeeping gene expression. The PCA showed a clear separation between all three conditions with Adh samples being grouped between Con and Sus samples as predicted.
2.5. Changes in regulation of proliferation and apoptosis
GSEA analysis was used to identify the most consistent expression changes in gene sets. An enrichment map of the significant down-regulated GO-terms in biological processes with a stringent FDR threshold of 5% can be seen in Fig. 4 (Supplementary Table S8). Overall, 151 gene sets were significantly down-regulated, and 91 were identified as significantly up-regulated. Most down-regulated gene sets were related to cell cycle regulation, mitosis, DNA-replication and DNA organization which is consistent with the observation of the long doubling time of Vero in suspension (Fig. 5a). Suspensions cells had down-regulated genes from each part of cell cycle progression compared to adherently grown Vero cells in CDM2 (Fig. 5b). The gene c-myc is down-regulated in suspension Vero cells, and this gene has been associated with cell cycle progression, along with tumorogenesis 26. The overexpression of this gene has been shown to allow quiescent cells to reenter the cell cycle and begin to proliferate 26. Up-regulated genes were linked to inflammation, fluid shear stress, migration and endothelial barriers, indicating a stress response during cell adaption to suspension. Importantly, gene sets related to regulation of programmed cell death were not detected as either down- or up-regulated. Genes sets regulating apoptosis were not detected as differentially expressed by GSEA, as seen by the expression patterns of positive regulatory (GO:0043065) and negative regulatory (GO:0043066) gene sets related to apoptotic processes (Fig. 5c). Despite a wider LFC range for pro-apoptotic genes and anti-apoptotic genes, the average LFC for both sets of genes is approximately zero for Adh samples, and only slightly positive for Sus samples.
The highest LFC for anti-apoptotic genes lrp2, pkhd1 and tmigd1 have been shown to specifically protect renal cells from apoptosis 27–29. On the other hand, for the top up-regulated pro-apoptotic genes, tnfsf10 has been proven to trigger apoptosis in renal cells. The expression of apoptogenic genes like bax, bak1, or bad were unaffected, whereas, anti-apoptic genes are either unchanged like mcl1 or bcl2l2 or increase like bcl2 30. Overall, the pattern does not seem conclusive for the Sus samples to indicate a change in the rate of apoptosis due to the widespread expression changes of both pro- and anti-apoptosis regulating gene sets.
2.6. Mitochondrial fatty acid metabolism
The fatty acid (FA) beta oxidation pathway is the main up-regulated gene set in the GSEA amino acid and lipid metabolism cluster (Fig. 4). The product of this pathway, acetyl-CoA, is further converted in the Krebs cycle for energy production. The expression changes of FA beta oxidation in Sus samples are presented in Supplementary Figure S6. The heatmap in Figure S6b illustrates the upregulation of 29 beta oxidation associated genes. The schematic diagram in Figure S6a illustrates that for each step in fatty acid beta oxidation at least one gene is up-regulated. This upregulation of enzymes for each main step might relate to an increased energy consumption. One key regulator of fatty acid metabolism in renal cells is tgfb131 (encodes for transforming growth factor-b1 protein), which is down-regulated in Sus cells whereas its receptor tgfb1r is up-regulated. The expression of the tgfb1 gene inhibits beta oxidation and promotes dedifferentiation, along with the epithelial-to-mesenchymal transition 32. Differentiated tubular epithelial cells use FA beta oxidation as their main method to produce energy 31. Interestingly, in non-cancerous cells the addition of transforming growth factor beta (TGF-b) will cause the down-regulation of c-myc and stop cell proliferation 26, but once the cancerous phenotype is already established, TGF-b is unable to stop them from proliferating. In summary, this data supports the idea that suspension Vero cells are becoming quiescent or senescent cells by using fatty acid beta oxidation, along with the large doubling times and down-regulated cell cycle genes. A possible method to overcome this senescence may be to up-regulate key genes such as c-myc 26, bcl2 33, or the addition of TGF-b 34.
2.7. Upregulation of kidney related genes
An enrichment map for up-regulated GO-terms for biological processes showed three large clusters of upregulated gene sets (Fig. 4). These gene sets are related to ion and small organic molecule transport as well as amino acids and lipid metabolism. To identify if the changes in the Sus group were connected to tissue specific pattern and tissue profiles for enriched genes were defined from the human protein atlas (HPA, available from http://www.proteinatlas.org) RNA-seq data 35. The significant profiles (p-adj < 0.05) of the gene set enrichment analysis with the tissue profiles can be seen in Fig. 6a, which shows that kidney and liver enriched genes were up-regulated. There were 17 genes associated with the liver expression profile and 12 of them contributed to the enrichment score, while the liver profile from HPA contains 242 enriched genes. The kidney profile contains 59 genes from which 25 were identified in the dataset and 22 contributed to the enrichment score. The 25 genes are further shown in the mean-difference (MD) plot in Fig. 6b. Except for slc13a3, npr3, and lhx1, all genes were significantly up-regulated with an FDR < 0.001. Since Vero cells were originally derived from a kidney, and the expression profile appears to specifically match the HPA kidney profile, further analysis was done on kidney-related genes.
Based on this up-regulation of renal tubule associated genes, expression changes were investigated to see if they may be related to even more specific segments of the renal tubule. To build profiles for the different segments bulk RNA-seq of dissected rat kidneys were used 36. Thirteen segment profiles were built with TissueEnrich of which eight passed the threshold of three detected genes in the RNA-seq dataset. The results are summarized in Fig. 6c (FDR threshold of 0.1). The complete results are in Supplementary Table S9. The enrichment analysis presents an up-regulation of multiple sets referring to different parts of the renal tubule with proximal tubule segment S3, cortical thick ascending limb which is part of the Loop of Henle as well as the inner medullar part of the collecting duct (IMCD). In contrast, only the glomerulus related gene set was detected as downregulated. From the 25 genes that were identified from the dataset, twelve of these genes are specific for renal tubule according to HPA (red points in Fig. 6b). Together, these findings suggest a cellular change towards renal tubule like expression patterns during the adaption to suspension conditions used in this study.
Additionally, the analysis of transcription factor target gene sets revealed the up-regulation of multiple kidney associated transcription factors (Supplementary Table S10). The three strongest enriched genes sets were foxi1, pax2, and hnf4a. The foxi1 gene is known to be important in regulating the expression of vacuolar-type H+-ATPase subunits in the kidney collecting duct 37; while, pax2 is an important factor for nephron differentiation in kidney development 38. The main function of hnf4a is the control over the expression of drug metabolizing enzymes and transporters in the proximal tubule39. The regulatory network of these transcription factors strongly supports the directed changes towards renal tubule cells.
2.8. Vero Transition to Renal Tubule-Like Cells
2.8.1. Membrane transporter
A closer investigation of common renal tubule functions reveals even more similarities with the Sus samples. Membrane transport proteins like those from the solute carrier (SLC) superfamily are important in renal tubule cells to transport diverse molecules, from ions to lipids or amino acids, to and from the bloodstream or tubular lumen. 246 genes of this superfamily were identified as expressed. The heatmap in Fig. 7a shows the changes for solute carriers. The increased expression of many SLC genes indicate a change of the Sus cells to increased absorption and secretion similar to renal tubule cells.
V-ATPases form another group of transporters whose expression increased. V-ATPase are proton pumps, which contain two domains. The heatmap in Fig. 7b emphasizes the strong up-regulation of subunits for these domains in Sus samples, especially for atpv6v1, which is an ATP hydrolysis domain. For atp6v1, five of seven subunits are upregulated whereas for atp6v0, two of the eight are up-regulated. The clustering of the samples shows again a strong difference between Sus samples against the other samples. V-ATPase is important for proton secretion into the tubular lumen and thus urine acidification, and it is located in the apical membrane 40,41.
A similar pattern can be seen for drug secreting ATP-binding cassette (ABC) transporters. Ten members, including abca, abcb, and abcc were upregulated. Again, the transporters are strongly upregulated in Sus samples as can be seen in Fig. 7c. The heatmap also shows that the increase in expression occurred progressively from Con to Adh to Sus cells. In renal tubules, ABC transporters help to secrete drugs as well as a range of macromolecules from lipids to bile salts or insulin into the tubular lumen 42–44.
To study the transport direction, the renal tubule associated carriers are represented in Fig. 7d with substrate and orientation. The diagram shows a strong direction for transport from many molecules into the cell and from there into the blood stream. The transported target molecules that are facilitated from the highly expressed genes that are also associated with transport from the lumen are diverse and include bicarbonate, uric acid, magnesium ions and amino acids. Although the Sus samples had no separation between a potential blood stream and a tubular lumen, the RNA-seq data show an up-regulation of direction-specific transmembrane transporters for Mg2+ or amino acids (Fig. 7d). This suggests that when these Vero cells were placed in suspension, the physiological conditions caused the cells to revert to a more kidney-like cell that focused on filtering the medium rather than proliferating.
2.8.2. Barrier function between tubular lumen and blood stream
To promote single cell suspension, CDM2 had low amounts of calcium and magnesium to prevent calcium-dependent cell adhesion molecules from properly functioning. Many cell adhesion molecules require calcium or magnesium to be functional such as cadherins, integrins and selectins. Although, even with lower calcium and magnesium concentrations, this did not prevent all cell adhesion since there are other adhesion molecules that do not require calcium or magnesium like claudins and vascular adhesion molecules (VCAM-1). Claudins are transmembrane proteins that specifically bind cells together at tight junctions. Vero cells that were grown adherently had down-regulated expression of various claudins, whereas suspension Vero cells up-regulated many cell adhesion proteins including claudins (except cldn15) (Fig. 8). Tight junctions are especially important in the kidney tubule to control passive reabsorption of particular compounds between the blood stream and the tubular lumen, which also indicates that the suspension Vero cells were behaving more like senescent kidney cells 45. The upregulation of these proteins, along with other cell adhesion molecules such as itgb2, vcam1 and jam1 may be the reason why even with low calcium and magnesium levels, the cells would still form aggregates (Fig. 2a).
3.7.3. Conclusion
A chemically defined media was designed that supports the growth of various mammalian cell lines in suspension. Vero cells were unable to effectively proliferate in suspension in this medium, and transcriptomic work revealed that these cells were behaving more as mature kidney cells. Overall, the RNA-seq data shown here demonstrated that suspension Vero cells became senescent rather than apoptotic, and that genes associated with cell-cycle progression were down-regulated and genes associated with fatty acid beta oxidation were up-regulated. In addition to the arrest in growth, suspension Vero cells up-regulated kidney tissue-associated genes. Suspension Vero cells specifically up-regulated genes associated with solute membrane transporters and proteins that form tight junctions which is important for the epithelial barrier and function of the tubule of the kidney for filtering blood. This work offers insights as to why Vero cells have been troublesome to adapt to suspension in the past and elucidates some possible avenues to establish a robust suspension cell line to support vaccine production using Vero cells.