Influence of cytoskeleton organization on recombinant protein expression by CHO cells

Abstract In this study, we assessed the importance of cytoskeleton organization in the mammalian cells used to produce therapeutic proteins. Two cytoskeletal genes, Actin alpha cardiac muscle 1 (ACTC1) and a guanosine triphosphate GTPase‐activating protein (TAGAP), were found to be upregulated in highly productive therapeutic protein‐expressing Chinese hamster ovary (CHO) cells selected by the deprivation of vitamin B5. We report here that the overexpression of the ACTC1 protein was able to improve significantly recombinant therapeutic production, as well as to decrease the levels of toxic lactate metabolic by‐products. ACTC1 overexpression was accompanied by altered as well as decreased polymerized actin, which was associated with high protein production by CHO cell cultured in suspension. We suggest that the depolymerization of actin and the possible modulation of integrin signaling, as well as changes in basal metabolism, may be driving the increase of protein secretion by CHO cells.


| INTRODUCTION
Improving recombinant therapeutic protein manufacturing using mammalian Chinese hamster ovary (CHO) cell hosts is a major challenge for biopharmaceutical companies. Thus, many research are currently ongoing to unravel and overcome the remaining unknown metabolic bottlenecks of these cells for high-level protein production.
We previously described how the deprivation of vitamin B5 in CHO cell culture medium, coupled to the cotransfection of an expression vector for a vitamin B5 transport protein with the gene of interest, allowed the identification of cell variants that are capable of expressing at very high levels recombinant therapeutic proteins . However, the cellular properties that allow very high-level recombinant protein secretion remain unknown.
In this study, we wished to identify specific gene expression alterations that accompany efficient therapeutic protein expression, with the hope to construct CHO cell line derivatives that would be permanently more efficient for production. Among these alterations, two genes involved in cytoskeleton organization were significantly induced after vitamin B5 selection, namely actin alpha cardiac muscle 1 (ACTC1) and a guanosine triphosphate GTPase-activating protein (TAGAP). ACTC1 is the major protein of the cardiac sarcomere thin filaments, which are responsible for the muscle contraction function of the heart. Consistently, ACTC1 deficiency has been mainly linked to heart diseases (Debold et al., 2010;Wang et al., 2016). TAGAP is a signaling protein that was involved in thymocyte loss of adhesion and thymocyte and T cells cytoskeleton reorganization (Connelly et al., 2014;Duke-Cohan et al., 2018). Alterations of the TAGAP gene have been associated with various autoimmune diseases (Eyre et al., 2010).
Recent studies have shown that suspension CHO cells have evolved from adherent cells by disruption of the extracellular attachment matrix accompanied with major changes in the cytoskeleton, such as increased actin filament expression, which is required for proper interaction with integrins, resistance to shear stress, and cell proliferation in suspension (Walther, Whitfield, & James, 2016). Therefore, cytoskeleton organization and modulation of actin filament levels may impact suspension cell fitness and recombinant protein expression, from messenger RNA (mRNA) translation to protein secretion.
Here, we studied the effect of ACTC1 and TAGAP on recombinant protein expression. We show that ACTC1 overexpression improves the production of easy-to-express and hard to express recombinant proteins. We further identify a link between an overall decrease of actin polymerization and an increase in recombinant protein secretion, which we hypothesize is mediated by the cytoskeleton structure and/or stability.

| Analyses of stable cell pools and cell lines
Fed-batch performance evaluation, IgG cell surface staining, IgG cell secretion assay, and vitamin B5 metabolite quantification, were performed as previously described .
Briefly, IgG secretion performances in fed-batch culture were performed as previously reported (Le Fourn et al., 2014). The assay of cell surface IgG was as reported previously (Brezinsky et al., 2003), and cell pools secreting high levels of recombinant IgG protein were subcloned using ClonePix™ FL Imager from Molecular Devices. For vitamin B5 metabolite quantification, cell pellets were extracted with 1 ml of cold MeOH:H 2 O (4:1, vol/vol) solvent mixture, then probe-sonicated. The supernatant obtained after 1 hr incubation at −20°C, followed by 15 min centrifugation at 13,000 rpm at 4°C were collected and evaporated to dryness then reconstituted in 100 µl MeOH:water (4:1) and injected into the liquid chromatography-mass spectrometry (LC-MS) system.

| RNA RT-PCR and sequencing RNA-seq analysis
For RNA reverse transcription and real-time quantitative polymerase chain reaction (RT-qPCR) analysis, total RNA was extracted from 10 6 cells and reverse-transcribed into cDNA using polyT primers.
Transcripts accumulation was quantified by qPCR using the SYBR Green-Taq polymerase kit from Eurogentec Inc, and ABI Prism 7700 PCR machine (Applied Biosystems). Transcript levels were normalized to that of the GAPDH housekeeping gene. RNA-seq analysis of the B5-and puromycin-selected CHO cell was as previously described .

| Protein sample preparation and immunoblotting
Total actin content was evaluated as follows. Protein extraction was performed from 10 7 cells washed in phosphate-buffered saline (PBS), after which the cell pellet was resuspended in radioimmunoprecipitation assay lysis buffer (150 Mm NaCl, 50 mM Tris-HCl

| Analysis of actin polymerization by cytofluorometry and confocal microscopy
The organization and quantification of polymerized actin (F-actin) were assessed on cell cultures initiated by seeding 2 × 10 5 cell/ml in SFM and culturing for 3 days at 37°C with 5% CO 2 . For microscopy analysis, cells were plated on coverslips placed in a 24-well plate, fixed with 3.7% formaldehyde in PBS for 10 min and then permeabilized with 0.1% Triton X-100 in PBS for 30 min. Samples were blocked with 2% bovine serum albumin in PBS for 1 hr and the low, medium, and high fluorescence). These cell populations were expanded and maintained at 37°C, 5% CO 2 until further analysis.

| Statistical analysis
The results are expressed as means ± standard error of mean.
Statistical analyses were performed using Student's t-test, with variance equality depending on sample variance F-test. Asterisks in the figure panels refer to statistical probability p values of less than .05, which were considered as statistically significant.

| Effects of cytoskeleton-related genes on therapeutic protein secretion by CHO cells
We recently devised a metabolic selection method whereas polyclonal cell pools and monoclonal cell lines mediating very high transgene expression levels can be more reliably isolated using a vitamin B5 deprivation method . Using this approach, CHO cells were cotransfected with expression vectors encoding an "easy-to-express" (ETE) trastuzumab or a "difficult-toexpress" (DTE) infliximab or Enbrel therapeutic protein, together with the vitamin B5 transporter SLC5A6 or with an antibiotic resistance gene as a control. Cells were then selected for their aptitude to survive in a B5-deficient culture medium or for antibiotic resistance, respectively, and differentially expressed cellular genes were identified by RNA sequencing. After antibiotic selection, the expression of both ACTC1 and TAGAP was lower than in nontransfected cells, while they were increased after B5 selection ( Figure 1a,b). The increase of TAGAP expression following SLC5A6 expression and vitamin B5 starvation was validated using three independent recombinant CHO cell pools isolated using either antibiotic or B5 selections (Figure 1c).
Gene induction after B5 selection may be caused either by B5 starvation occurring during the selective process, as found in a previous study , by the overexpression of SLC5A6 itself, as it mediates higher vitamin B5 intake into the cell (Figure 1d), or by a combination of both effects. B5 is an F I G U R E 1 Expression of the ACTC1 and TAGAP genes following vitamin B5 selection. (a and b) Transcriptomic RNA sequencing (RNA-seq) analyses of ACTC1 and TAGAP mRNA levels, comparing nontransfected nonselected parental control cells (C) with transfected cell pools submitted to antibiotic selection or to B5 selection and expressing trastuzumab (ETE; Panel a) and with cell lines expressing interferon β (DTE; Panel b). Transfected cells were cultured in nonselective complete culture medium, and total mRNA was isolated and submitted to highthroughput sequencing to identify genes upregulated in cell populations submitted to the B5 selection process. The relative mRNA levels correspond to normalized read counts from RNA-seq analyses. (c) Effect of SLC5A6 overexpression and selection by B5 deprivation on ACTC1 and TAGAP gene expression. Parental CHO cells were cotransfected with the Tras or GFP expression plasmid, and with the puromycin resistance and SLC5A6 expression vectors, after which the cultures were selected either in the presence of puromycin (antibiotic selection), or in B5-deficient medium (B5 deprivation), respectively. The resulting selected cell pools were transferred to a nonselective culture medium followed by the quantification of ACTC1 and TAGAP mRNAs by RT-qPCR. mRNA levels of cells selected by B5 deprivation were normalized to those of antibiotic-selected cells. (d) The vitamin B5 content of cells transfected and selected as described for Panel C was measured by LC-MS after 6 days of batch culture. (e) Comparison of the ACTC1 and TAGAP mRNA levels of cell pools transfected with the antibiotic resistance gene without or with the SLC5A6 expression vector and submitted to antibiotic selection. Relative mRNA levels were determined by RT-qPCR and normalized to those of antibiotic-resistant cells. Data are mean ± SEM of 3-5 biological replicates. *p ≤ .05; **p ≤ .02 with respect to antibiotic selection (t-test; one tail). CHO, Chinese hamster ovary; DTE, difficult-to-express; ETE, easy-to-express; LC-MS, liquid chromatography-mass spectrometry; mRNA, messenger RNA; RT-qPCR, real-time quantitative polymerase chain reaction; SEM, standard error of mean essential cofactor for acetyl CoA, a key element in central metabolism and energy metabolism, which could be linked to cytoskeleton regulation. To distinguish between these possibilities, cell pools overexpressing the SLC5A6 transporter were generated without any B5 deprivation, which indicated that increased SLC5A6 expression suffices to upregulate significantly the ACTC1 gene, whereas a nonsignificant increase of TAGAP expression was noted (Figure 1e). Therefore, the B5 selection process might activate ACTC1 gene expression by the increased B5 intracellular import mediated by SLC5A6 overexpression, whereas a significant increase of TAGAP expression required a combination of both SLC5A6 overexpression and B5 starvation. Interestingly, we also observed that TAGAP overexpression increased ACTC1 mRNA and protein accumulation ( Figure S1), suggesting that the increased ACTC1 | 1121 ACTC1 high expressor clones were randomly picked for further analysis ( Figure S3a). ACTC1 protein overexpression was validated by western blot (Figures 3a and S3b). Among the four ACTC1overexpressing clones, three showed the highest IgG titers after 13 days of fed-batch culture as compared with the empty vector clones (Figures 3b and S3c).
To determine if the increased therapeutic protein secretion elicited by ACTC1 overexpression may result from cellular metabolic alterations, we measured primary metabolism markers by MS analysis of pools of ACTC1-overexpressing cells. Notably, we assessed the accumulation of lactate, a toxic by-product of the early steps of glycolysis, which has been well documented as a bottleneck for therapeutic protein production (Lao & Toth, 1997).
This revealed a strong reduction of lactate accumulation by ACTC1-overexpressing cells after 3 days in batch culture when compared with control cells (Figure 2e). Overall, we, therefore, concluded that ACTC1 gene overexpression significantly improved the secretion of various therapeutic proteins and that this effect may be linked to a decrease in the accumulation of the toxic lactate metabolic by-product.

| Implication of actin filament organization and level in recombinant protein secretion
We next assessed whether the actin polymerization status may be affected by ACTC1 overexpression. Phaloidin staining of F-actin was performed on independent cell lines expressing ETE proteins, revealing structural differences. Negative control cells expressing only the ETE proteins displayed round actin filament structures, as observed for the untransfected parental CHO cells (Figures 4a, S4, and S5; Movie S1). However, sharp actin structures appeared when ACTC1 was overexpressed, and cells displayed a somewhat bloated shape (Figures 4a, S4, and S5; Movie S2). F-actin staining was abolished when cytochalasin, an actin polymerization inhibitor, was added ( Figure S5).
To further assess actin polymerization levels, we relied on SiRactin staining, which specifically binds to F-actin (Lukinavičius et al., 2014), yielding fluorescence level of stained cells that are proportional to actin polymerization. Comparison of SiR-actin-staining of the ACTC1 and trastuzumab-expressing clones relative to control clones revealed higher fluorescence in the control clones than in the ACTC1-overexpressing ones, implying that actin polymerization levels are reduced by ACTC1 overexpression (Figures 4b,c and S6).
To assess whether actin polymerization may affect recombinant protein expression, two independent CHO cell pools expressing the trastuzumab protein, but not submitted to an ACTC1 vector transfection, were stained with SiR-actin. Stained cells were then sorted in three independent cell batches according to their low, medium, or high fluorescence level (Figures 5a and S7 F I G U R E 3 Characterization of the productivity of ACTC1-overexpressing cells. A trastuzumab-expressing line was stably retransfected with the CHO ACTC1 or with an empty expression vector, and cell clones were isolated for further analysis. (a) Immunoblots of total protein extracts labeled with ACTC1 or GAPDH mouse antibodies. The histogram shows the ratio of the ACTC1 signal relative to that of GAPDH, as assessed using the ImageJ. Ponceau red-stained membranes are shown in Figure S3a.  protein synthesis and that this activated the cellular stress response (Silva, Sattlegger, & Castilho, 2016 We, therefore, conclude that several mechanisms may well explain the positive effect of actin overexpression on protein production by CHO cells, which may pertain both to the basic metabolism of CHO cell and energy production by glycolysis, as well as by a potential activation of protein secretion. Further work will be required to determine the contribution of these mechanisms on recombinant protein accumulation by cultured CHO cells. In any case, this study shows that ACTC1 overexpression and/or the assay for spontaneous alterations in F-actin polymerization using SiR-actin staining and cell sorting can both be used to facilitate the isolation of high expressor CHO cells from stable cell pools. As for ACTC1, TAGAP overexpression improved recombinant protein secretion. TAGAP enzymatic activity was showed to regulate the organization of actomyosin fibril, to release integrin-mediated adhesion in thymocytes, which is necessary for directed-thymocyte migration from the cortex to the medulla (Duke-Cohan et al., 2018).
In suspension CHO cells, TAGAP could function as a mediator for intracellular cytoskeleton signal to cell surface integrins, hence improving cell proliferation, viability, and adaptation to suspension.
(a) (b) (c) F I G U R E 5 Sorting of therapeutic protein-producing cell pools according to their F-actin polymerization level. Representative histograms of flow cytometry analyses of a trastuzumabexpressing cell pool treated by SiR-actin staining. Unstained control and other analysis are depicted in Figure S4. A total of 5 × 10 5 cells were analyzed per acquisition, among which 0.4-1.4 × 10 5 cells were sorted by cytofluorometry according to their low, medium, or high actin polymerization (pol.) levels, as depicted. (b) Selected cells were transferred to an antibiotic-containing culture medium followed by the analysis of IgG cell surface display by immunofluorescence staining cytofluorometry. (c) IgG secretion assays of the sorted cells of Panel B. Histograms represent the average values ± SEM from six cell pools. **p ≤ .02 (t-test; two tails; paired) relative to the low actin polymerization category. IgG, immunoglobulin gamma; SEM, standard error of mean [Color figure can be viewed at wileyonlinelibrary.com] Overall, we conclude that cytoskeletal proteins and the modulation of the cytoskeletal organizations may be used to improve protein production for biotechnological purposes. It will be interesting to assess the ability of actin monomer-binding agents that mediate transient actin depolymerization to modulate protein translation and secretion in CHO cells, and whether this may provide easy approaches towards increasing the therapeutic production capacity of these cells.