Harnessing metabolic plasticity in CHO cells for enhanced perfusion cultivation

Chinese Hamster Ovary (CHO) cells have rapidly become a cornerstone in biopharmaceutical production. Recently, a reinvigoration of perfusion culture mode in CHO cell cultivation has been observed. However, most cell lines currently in use have been engineered and adapted for fed‐batch culture methods, and may not perform optimally under perfusion conditions. To improve the cell's resilience and viability during perfusion culture, we cultured a triple knockout CHO cell line, deficient in three apoptosis related genes BAX, BAK, and BOK in a perfusion system. After 20 days of culture, the cells exhibited a halt in cell proliferation. Interestingly, following this phase of growth arrest, the cells entered a second growth phase. During this phase, the cell numbers nearly doubled, but cell specific productivity decreased. We performed a proteomics investigation, elucidating a distinct correlation between growth arrest and cell cycle arrest and showing an upregulation of the central carbon metabolism and oxidative phosphorylation. The upregulation was partially reverted during the second growth phase, likely caused by intragenerational adaptations to stresses encountered. A phase‐dependent response to oxidative stress was noted, indicating glutathione has only a secondary role during cell cycle arrest. Our data provides evidence of metabolic regulation under high cell density culturing conditions and demonstrates that cell growth arrest can be overcome. The acquired insights have the potential to not only enhance our understanding of cellular metabolism but also contribute to the development of superior cell lines for perfusion cultivation.

Since the food and drug administration first approved a biopharmaceutical compound produced by Chinese Hamster Ovary (CHO) cells more than 35 years ago (Kaufman et al., 1985), CHO cells have developed into the industry's favorite mammalian expression platform due to their human like post translational modifications (MacDonald, Nöbel, Roche Recino, et al., 2022;Walsh & Walsh, 2022).Over the last two decades, these cells were primarily cultivated under fed-batch conditions.However, recent developments in the biopharmaceutical industry have renewed interest in perfusion cultivation (MacDonald, Nöbel, Roche Recino, et al., 2022).
Given its high volumetric productivity, perfusion cultivation can be a significant contributor to meet the increasing demand for biopharmaceuticals, especially considering the needs of an aging population.
This shift toward perfusion systems signifies an important evolution in biopharmaceutical manufacturing strategies, highlighting the need for further research and optimization.However, our understanding of CHO cell line's behavior in perfusion cultures remains in its infancy (MacDonald, Nöbel, Roche Recino, et al., 2022).The complex dynamics of continuous perfusion systems and the unique stresses they impose on cells (Zamani et al., 2018) necessitate a more profound understanding to optimize cell performance and, ultimately, enhance the productivity and efficiency of perfusion bioprocessing.There is a critical need for comprehensive studies exploring CHO cells' cellular and metabolic responses to perfusion conditions to drive future improvements in this rapidly evolving field.
To date, the industry is often focused on bioprocess optimization through improved media formulation (Lin et al., 2017;Mayrhofer et al., 2020) or process setup and parameters (Clincke et al., 2013;Villiger-Oberbek et al., 2015).Targeted metabolic engineering is rare and a yet untapped source for improvement of cell lines specifically for perfusion systems.Engineering strategies are more common in conventional fed-batch and batch setups to omit the effects of apoptosis (Henry et al., 2020), improve glycosylation (Tejwani et al., 2018), and most commonly increase productivity (Golabgir et al., 2016).Numerous genetic modifications are transferrable across cell lines; however, differences in process environment can reflect at the metabolic level (Templeton et al., 2017;Walther et al., 2019;Zamani et al., 2018).Therefore, optimising cell lines specifically for given bioprocess conditions, like perfusion bioprocesses is desirable.
In our preceding research, we investigated the performance of an apoptosis resistant CHO cell line, deficient of the BAK1, BAX, and BOK genes (MacDonald, Nöbel, Martínez, et al., 2022).Notably, we observed a growth arrest that was not induced by external factors.Such unusual cellular behavior under perfusion conditions underscores the complex interplay of genetic and metabolic factors influencing cell performance in continuous bioprocessing.The finding has set the stage for a deeper exploration into the mechanisms behind the observed growth arrest and the potential strategies to overcome it, thereby enhancing the productivity and robustness of perfusion cultures.A reduction in growth rate can be advantageous in high-production states to prevent wastage of media and products, and prior research has investigated inducing such a slowdown through cell cycle arrest via temperature shifts and valeric acid administration (Chen et al., 2004;Coronel et al., 2020;Wolf et al., 2018).However, the growth arrest we observed in the apoptosis resistant CHO cell line under perfusion conditions was distinctly different.It was more profound and occurred without the alteration of any process conditions.This intrinsic growth arrest suggests complex, endogenous regulatory mechanisms at play in the cells, offering a novel perspective on managing cell growth and yield in perfusion cultures.
The continuous nature of the perfusion setup provides an opportunity to study the long-term effects of the triple knockout on cell metabolism, unimpeded by deteriorating cell environment conditions due to the accumulation of toxic by-products (Pereira et al., 2018).However, other stresses are still experienced by the cell, most prevalent, oxidative stress.Linked to the generation of reactive oxygen species (ROS) from oxidative phosphorylation, β-fatty-acid degradation or peroxisome, the ROS response is complex, and a range of different proteins and pathways are involved (Espinosa-Diez et al., 2015).Cells contain several radical-accepting metabolites to counteract ROS damage, with glutathione being the most prominent (Chevallier et al., 2020).Glutathione's ability to alleviate the adverse effects of increased oxidative phosphorylation-a process directly linked to high energy production for recombinant protein production-has associated it with enhanced productivity (Orellana et al., 2015).
Here, we further characterized the apoptosis-resistant triple knockout cell line capable of maintaining high viabilities in a perfusion system without applying a bleed stream.We link the observed phenotypic changes to alterations at the proteome level and discuss their implications and potential for future optimized cell lines for perfusion cultures.This aims to enhance our understanding of CHO cell metabolism under perfusion conditions, thereby facilitating the development of novel CHO cell lines that can meet the evolving demands of both the industry and regulatory bodies.Through this work, we aspire to highlight the potential of cell lines specifically designed for conditions present in perfusion systems.Thereby, inspire future efforts for the development of process specific cell lines and contributing to the optimization of perfusion bioprocessing, advancing the capabilities of this crucial platform for biopharmaceutical production.

| Cell line development
Proapoptotic genes BAK1, BAX, and BOK were knocked out in a CHO-K1-derived cell line, which stably expresses the human mAb m104.2 against the Hendra and Nipah viruses (Playford et al., 2020;Zhu et al., 2008).The antiapoptotic properties of the cell line have been confirmed previously (MacDonald, Barry, et al., 2022).
Control and knockout cell lines were adapted to CHOKO Cellca SMD media (Sigma-Aldrich) supplemented with 0.4% anti-clumping agent (Gibco).Cells were thawed from a working cell bank and expanded in shake flasks (Corning).Cultures were passaged every NÖBEL ET AL.

| Perfusion culture
Perfusion cultures were performed in a 3 L stirred tank (Sartorius, DE) with 1.2 L working volume.An ATF2 with 0.2 µm hollow fibre membrane (Refine Technologies) was used for cell retention.Cells were seeded at 0.5 × 10 6 viable cells/mL in CHOKO Cellca SMD media supplemented with 0.4% Anti-clumping agent.A more detailed overview of the cultivation method has been previously described by MacDonald et al (MacDonald, Nöbel, Martínez, et al., 2022).

| Sampling and analyses
Samples were taken daily and cell density, viability and cell diameter were determined using a Vi-CELL BLU Cell Viability Analyzer (Beckman Coulter).A BioProfile ® FLEX2™ Automated Cell Culture Analyzer (Nova Biomedical) was used to quantify glucose, lactate, ammonia, glutamine, glutamate, and osmolarity.Additionally, samples of approximately 5 × 10 5 cells were taken, spun down at 500g for 3 min, discarded of their supernatant, and a pellet was frozen at −80°C for later proteomics analysis.Three subsequent days were chosen from each growth phase to

| Proteomics sample preparation
Cells were lysed using 10% SDS in 50 mM triethylammonium bicarbonate buffer (TEAB) before reduction with 0.2 M dithiothreitol.Following incubation at 60°C for 30 min and sonication for 15 min.Samples were stored at 4°C overnight to maximize disulphide bond reduction.
The following day, 0.02 M iodoacetamide was added before samples were incubated at RT for 30 min.Post incubation, 1.2% phosphoric acid was added before the sample was diluted 6 times in S-Trap binding buffer (90% MeOH, 100 mM TEAB, pH 7.1) and pipetted into S-Trap™ Micro Spin Columns installed on a vacuum manifold.The flow-through was discarded, and the column was washed 4 times with S-Trap binding buffer.Proteins were then digested by adding 21 µL of 50 mM ammonium bicarbonate at pH 8.0 and 4 µL of 0.5 µg/µL Trypsin, followed by 2 h of incubation at 42°C.
Peptides were eluted three times by the addition of 35 µL of 50 mM ammonium bicarbonate, 35 µL of 0.2% aqueous formic acid and 50 µL of 50% acetonitrile in 0.2% formic acid, respectively.The peptide containing flowthrough was transferred into an analytical vial and dried using a vacuum centrifuge.Finally, samples were resuspended in 15 μL of analytical buffer A (5% ACN + 0.1% FA).RSLCnano HPLC system (Thermo Fisher Scientific).0.1% formic acid in MQ-H 2 O was used as mobile phases A and 80% acetonitrile in 0.1 aqueous formic acid as phase B for elution over 60 min at a gradient of 8% to 95% mobile phase B with a flow rate of 0.5 µL/min.Mass spectra were generated using the Thermo Fisher Scientific Q-Exactive HF-X (Thermo Fisher Scientific).A full MS scan was done at a resolution of 30,000 at 200 m/z and a collision energy of 30%.A precursor mass range of 400−1100 m/z was scanned in 10 m/z isolation windows with 1 m/z overlap.

| Data analysis
Samples were normalized and protein groups were identified using Spectronaut v. 13.15.200430 (Bruderer et al., 2015).p were estimated using a Kernel Density Estimator.Proteins were mapped against the Cricetulus griseus genome assemble GCF_003668045.3.
A hypergeometric test was used to identify enriched pathways as per Equation 1, with k being the number of differentially expressed proteins in the data with specific pathway annotation and M representing all proteins from the genome annotated to a pathway.
The number of differentially expressed proteins detected and annotated to a pathway is n and N is the number of proteins in the genome with specific pathway annotation.
Proteins were plotted on KEGG pathways as their log2 fold change in relation to the start of the stationary phase using the path view extension package (Luo & Brouwer, 2013).
The cell-specific productivity was calculated as published previously (MacDonald, Nöbel, Martínez, et al., 2022).Specific consumption and production rates were estimated by integrating the following equations: where s i is the substrate concentration in the reactor, p is i the concentration of product in the reactor, D the perfusion rate, X cell density, q i the specific rates and c i the concentration in the feed.
After integration, Equations 4 and 5 were used: where The cell number updated by biomass was estimated based on an average cell of diameter 14.95 and 15.26 µm, and dry weight of 566 and 408 [pg/cell], for triple knock-out cell and control cell lines, respectively.

| Gene copy number
A quantitative polymerase chain reaction was performed as described elsewhere (Jiang et al., 2016) to determine the number of genes encoding for the mAb present throughout batches.Samples were selected from the end of growth, stationary, and secondary growth phase.Primers used are listed in the Supporting Information Material.

| A reversible and noninduced arrest in growth
We The same parental cell line described previously served as a control (MacDonald, Nöbel, Martínez, et al., 2022).The control did not show any growth arrest and was controlled by a bleed set to maintain a constant VCD between Days 11 and 17.
Based on the observed phenotype, the culture was split into 3 phases, growth, stationary and second growth (Figure 1a encircled 1, 2, and 3, respectively).The volumetric titer stayed relatively constant throughout the culture, but cell-specific productivities changed (Figure 1e), dropping for the knockout cell line during the second growth phase.
Key process parameters such as cell diameter, lactate and ammonia concentrations and osmolarity, followed the trends previously observed.
(MacDonald, Nöbel, Martínez, et al., 2022).At the time of growth arrest, ammonia concentrations were 2.39 mM and 3.24 mM for replicate A and B, respectively and lactate concentrations were <2 mM for both cultures (Supporting Information Material S1: Figure 1).Given that the growth arrest could not be explained by the accumulation of inhibitory metabolites such as ammonia or lactate, we proceeded to analyze amino acid consumption (Figure 1d).We focused specifically on the changes in amino acids consumption patterns at transition points in the growth trajectory, namely the initiation of growth arrest and the end of the second growth phase (Figure 1d).Asparagine was the only amino acid consumed at >90% in both replicates.Although asparagine is not traditionally classified as an essential amino acid for growth, its high consumption rate suggests a potentially significant role in the metabolic adaptation of this specific CHO cell line in response to growth changes under perfusion conditions (Carrillo-Cocom et al., 2015;Hefzi et al., 2016).Across all amino acids, the percentage of consumption dropped over the time of growth arrest.By the end of the second growth phase, the essential amino acids leucine, valine, isoleucine, cysteine and arginine and the nonessential amino acids glutamine and asparagine were consumed to more than 90% of their feed concentrations.This consumption pattern strongly suggests that the second growth phase ended due to amino acid depletion, even though sufficient nutrients were still available at growth arrest.Moreover, there was no significant loss in gene copy number throughout the cultivation, ruling out genetic instability as a reason for the decrease in cell-specific productivity (Supporting Information S1: Figure 2).An additionally performed nitrogen balance strongly suggests that the secondary growth was founded on redirecting resources from mAb production toward biomass (Supporting Information S1: Table 2−9 and Supporting Information S1: Figure 3).These findings underscore the need for a comprehensive understanding of the complex interplay between nutrient availability and cellular metabolism for optimized CHO cell performance in perfusion cultures.

| Growth arrest is linked to cell cycle arrest and upregulation of central carbon metabolism
Under normal conditions, cell death is an essential part of the cell's life cycle.Therefore, the inhibition of apoptosis, as in the case of our triple knockout CHO cell line, is expected to impact cellular metabolism significantly.This was observed as a phenotypic shift after 10 days of culture.When we analyzed the global clustering of proteomics samples taken from three consecutive time points during the growth phase and at a constant VCD, we observed a close clustering between the growth phases of the knockout and control cultures, independent of time (Figure 2a).This indicates that while the knockout behaved similarly to the control during growth, a shift in its proteome occurred when growth arrest was initiated at the onset of the stationary phase.Whilst not all homologs were detected, SODs and catalase (CAT) indicated increased oxidative stress during the stationary phase, particularly in mitochondria.Glutathione and TXN responses, both converting hydrogen peroxide to water, did show upregulation of some homologs but no global upregulation.

| DISCUSSION
In this study, we conducted an in-depth examination of the phenotypic and proteomic changes occurring in a CHO triple knockout cell line, deficient in BAK1, BAX, and BOK, under highdensity perfusion culture conditions.We have observed an unexpected resurgence in growth following the previously described cell growth arrest, and we have delved into the underlying metabolic implications of this phenomenon.(MacDonald, Nöbel, Martínez, et al., 2022).While CRISPR engineering can lead to off-target effects, the risk for those was kept minimal in the cell line design (MacDonald, Barry, et al., 2022).Further, the by the knockout Relative abundance of proteins associated to the response/reduction of ROS over time.Error bars represent the standard deviation with n = 3 for all phases except for the second growth were n = 2. CAT, catalase; GCL, glutathione ligase; GPX, glutathione peroxidase; GSH, glutathione; GSS, glutathione synthetase; GSSG, glutathione disulphide; GSR, glutathione reductase; Keap1, Kelch-like ECH-associated protein 1; NFKB1, NF-kappaB; PRDX, peroxiredoxins; SOD1 and SOD2, superoxide dismutase 1 and 2; TXN, thioredoxin; TXNR, thioredoxin reductase.
| 1377 targeted mitochondrial outer membrane permeabilization is, to the best of our knowledge, not known to cause unintended downstream effects on other pathways.Thus, we expect the here-described observations to be solely based on the intended knockout.
In traditional scenarios, cells typically exit the growth phase due to nutrient depletion or the accumulation of inhibitory metabolites, with lactate and ammonia being the most prevalent in CHO cell cultures.(Martinez et al., 2013;Mulukutla et al., 2017;Pereira et al., 2018).Whilst lactate is almost entirely consumed, ammonia concentrations of 3.2 mM at the time of growth arrests are well below the concentration expected to be inhibitory (Kurano et al., 1990).Although other inhibitors have been described, their inhibitory effect is often masked unless there is a specific reduction in lactate and ammonia levels (Mulukutla et al., 2017).
Similarly to lactate and ammonia, the concentrations of amino acids observed in our study did not solely account for the observed growth arrest.This suggests that other factors are contributing to this phenomenon, potentially interplaying with the unique metabolic characteristics of our apoptosis-resistant CHO cell line.
(MacDonald, Nöbel, Martínez, et al., 2022).CHO cell culture media are complex, with components extending beyond just glucose and amino acids.Vitamins, lipids, trace elements, and other components can also influence cell behavior, potentially playing a role in the observed growth arrest.(Ritacco et al., 2018).The metabolic requirements of our apoptosis-resistant CHO cell line might be unique, leading to an underestimation of certain nutrient needs based on traditional CHO cell culture paradigms.Understanding these complex interactions and requirements will be crucial to unravel further the metabolic underpinnings of the observed growth phases in our study.
Despite the increase in biomass, the maintained overall titer during the second growth phase further supports the notion that the cell cycle arrest is not simply a result of nutrient depletion.Instead, it suggests a shift in the cells' metabolic state, where the cells may redirect resources toward maintaining viability and other essential cellular functions rather than dividing.The performed nitrogen balance supported the notion that the decrease resource flow toward product partially supported the regrowth.This shift might be an adaptation mechanism of our apoptosis-resistant CHO cell line to the long-term culture conditions, potentially indicative of a greater metabolic flexibility in response to changing environmental cues.
Understanding this metabolic adaptability might offer valuable insights for optimizing perfusion culture conditions and ultimately enhancing productivity in biopharmaceutical production.
In the knockout cell line, the shift to a constant, nonbleed driven VCD coincides with a strong downregulation of growth and cell cycle pathways, indicating an arrest in the cell cycle.Such link between the blockage of apoptosis and cell cycle arrest is likely facilitated through a shared pathway in the activation cascade, e.g. the p53 pathway (Chen, 2016).However, the p53 pathway revealed an inconclusive picture due to its multitude of interactions and complex network of posttranslational modifications (Kruse & Gu, 2009).P21 as a link between the cell cycle and p53 pathway did show clear downregulation during the stationary phase supporting this theory.
P27 on the other hand, which has, previously linked to cell cycle arrest in apoptosis resistant cells, was not found to be upregulated (Janumyan et al., 2008).
Upregulated pathways during the stationary phase were linked to the central carbon metabolism, the supply of its intermediates, and oxidative phosphorylation.An increase in flux toward these pathways has previously been reported as a characteristic of high antibody production due to the increased supply of ATP, once cells exit the growth phase (Pan et al., 2017;Templeton et al., 2013).A redirection of pyruvate toward oxidative phosphorylation instead of lactate has been reported for CHO cells overexpressing Bcl-2, which is responsible for the suppression of mitochondrial pore formation by BAK and BAX (Templeton et al., 2014).
Despite the notable changes in other pathways, there was no significant difference in the TCA cycle activity between the apoptosis-resistant and control cell lines.This lack of differential activity aligns with the notion that the upregulation of central carbon metabolism is intrinsically connected to cell cycle arrest.In other words, when cells enter a state of nongrowth, or cell cycle arrest, they redirect their metabolic resources toward the central carbon metabolism, presumably supporting other cellular activities rather than focusing on replication and growth.
The transition from the growth phase to cell cycle arrest is a common strategy in CHO cell cultures to enhance productivity and delay the onset of apoptosis.This strategy exploits the fact that, under certain conditions such as temperature shifts, cells cease dividing and instead concentrate their energy on protein production, commonly leading to an increase in cell size (Kaufmann et al., 1999;Xu et al., 2018).Principally, cell cycle arrests are transient and can be reversed upon repair of damaged DNA, which is standard practice in the synchronization of cells (Tobey et al., 1990).Only once the DNA damage becomes irreparable, cells transition into irreversible senescence (Kumari & Jat, 2021).
In traditional fed-batch systems, mechanisms that trigger cell cycle arrest usually persist until the end of the culture phase.In such systems, inhibitory substances accumulate over time, potentially leading to irreparable cell damage and subsequent apoptosis (Hiller et al., 2017).This is in contrast to perfusion cultures where there is a continuous exchange of media, which tends to improve the cell environment, allowing the cells to revert from the cell cycle arrest potentially.A minor example of this phenomenon can be seen in cells arrested by the presence of 1 mM valeric acid in the media.In this case, growth rates declined over 7 days, but eventually began to increase again, indicating an adaptation to the stress caused by the valeric acid (Wolf et al., 2018).However, it is important to note that growth rates did not fully recover to initial levels, demonstrating the long-term impact of such stress factors.This underscores the complexity and adaptability of cell cultures, and the various factors that can influence their growth and productivity.
The described phenomenon, where the entire culture undergoes cell cycle arrest for an extended duration of around 20 days without external process changes, is markedly distinct from typical scenarios.
Such behavior is likely specific to this apoptosis-resistant cell line.
This suggests that the mechanisms employed by these cells to resist apoptosis also lead to a prolonged cell cycle arrest under certain conditions.This unique behavior has implications for the overall cell culture dynamics and productivity, and emphasizes the need for more in-depth understanding of apoptosis-resistant cell lines and their behavior under various culture conditions.
Increasing oxidative phosphorylation is usually beneficial in biopharmaceutical manufacturing due to the resulting increased ATP supply.However, an upregulation of oxidative phosphorylation is linked to the generation of ROS from Complex I and III, which can trigger an oxidative stress response, potentially leading to apoptosis (Chevallier et al., 2020).Additionally, the proteins associated to the peroxisome as the second main contributor of ROS was also found to be upregulated during the stationary phase.A mixed picture emerged when observing the relative abundance of key proteins involved in the oxidative stress response (Espinosa-Diez et al., 2015).The oxidative stress was expected to increase as the oxidative phosphorylation increased.Yet, a range of proteins showed a downregulation compared to both growth phases.Upregulation was strongest for SOD2 and, compared to other PRDX, PRDX5.Both are known to be present in mitochondria (Karnati et al., 2013;Knoops et al., 1999), where oxidative stress during the stationary phase would be expected to be the most profound.Downregulation of glutathione synthesis associated proteins was unexpected and a potential target for genetic engineering as it has been linked to increased productivity (Orellana et al., 2015(Orellana et al., , 2017)).Overall, data indicated a strong compartmentalization of ROS stress experienced, which is strongest in mitochondria and peroxisomes during cell cycle arrest when oxidative phosphorylation is upregulated.This might help to guide the engineering of specific oxidative stress resistance in manufacturing cell lines.
This study presents, to the best of our knowledge, the first description of a second growth phase in mammalian cell culture.
This phenomenon appears to be facilitated by the cell cycle arrest, unlike the typical patterns of continuous growth or apoptosis seen in nonapoptosis resistant cell lines.This discovery naturally prompts an investigation into its underlying causes.Conventionally, the termination of a cell cycle arrest is linked to a decrease in cellular stress, but the specific triggers in this case remain to be identified.(Kumari & Jat, 2021).A multitude of inhibitors have been described (Mulukutla et al., 2017;Pereira et al., 2018), most of them capable of causing the arrest in cell cycle by its accumulation.As both replicates showed signs of a regrowth within short time of one another, a spontaneous adaptation event is unlikely.The second growth phase, as revealed by changes in its proteome, shows a partial reversal of the upregulated and downregulated pathways observed during cell cycle arrest.
Despite this, the pace of the second growth phase is notably slower, suggesting that the initial pressures causing the cell cycle arrest may still be present, albeit to a diminished degree, or new ones may have superseded them.Importantly, this cell line originated from a single clone (MacDonald, Barry, et al., 2022).
However, clonality does not always ensure a uniform population at a reactor stage (Ko et al., 2018).While the observed secondgrowth could potentially be attributed to a subpopulation within the cell line, this notion remains speculative.
Pinpointing a definitive pathway responsible for the here observed cell behavior was challenging due to the high complex- account for biological variability.Proteomics samples were separated into growth (Days 8−10), stationary start (Days 17−19), stationary end (Replicate A: Day 25−27 and replicate B: Day 29−31) and 2nd growth (Replicate A: Day 32−34 and replicate B: Day 38−40).

2. 5 |
Liquid Chromatography/Mass spectrometry (LC/MS) Samples were analyzed in data-independent acquisition (DIA) mode. 2 µL of the sample was injected into a C18 nanoEaseTM M/Z CSH C18 1.8 µm (150 µm × 100 mm) column (Waters) in an Ultimate 3000 cultured the apoptosis-resistant CHO cell line in a 3 L perfusion bioreactor, maintaining a constant media exchange of 1 reactor volume (RV)/day.The cell line was previously shown to be effective in perfusion bioreactors without bleeding while maintaining product quantity and quality(MacDonald, Nöbel, Martínez, et al., 2022)    (Figure1a−c).During the initial 11 days, both replicates exhibited exponential growth, reaching 50 and 61 × 10 6 cells/mL densities, respectively.Upon reaching growth arrest, the cells entered a previously described stationary phase, which persisted until Day 25 for replicate A and Day 29 for replicate B. Interestingly, without any modification to process parameters, the cells transitioned into a second growth phase post-stationary phase, where cell densities exceeded 10 8 cells/mL.Remarkably, no bleed was employed throughout the culture duration, yet cell viability was consistently maintained above 89%.This behavior underscores the potential of the apoptosis-resistant triple knockout CHO cell line for high-density, high-viability perfusion cultures without the need for a bleed stream.

Focusing
on time-dependent changes in the knockout cell line, the arrest in growth and later onset of a second growth phase, indicate two major metabolic shifts.Principal Component Analysis (PCA) demonstrated a tight clustering between the stationary and second growth phases as compared to the initial growth phase.The dominating Principal Component 1 (PC1) grouped the second growth phase between the initial growth and the stationary phase (Figure2b), suggesting a partial reversion of the changes associated with the first metabolic shift.As the PCA did indicate differences between growth phases, pathways involved in the metabolic shift were further explored through pathway enrichment using a hypergeometric test.We compared differentially expressed pathways (those with an adjusted p < 0.1) in terms of the log2 fold changes of associated proteins, relative to the initial growth phase (Figure2c).Compared to the control, this analysis corroborated our prior observation of tighter clustering between stationary and second growth phases.Downregulated pathways included DNA replication, Spliceosome, Nucleocytoplasmic transport and Proteasome(Kito et al., 2020), all associated to cell growth.On the other hand, upregulated pathways were either directly linked to the central carbon metabolism (tricarboxylic acid [TCA] cycle, Glycolysis/Gluconeogenesis) or indirectly associated by supplying intermediates to the TCA cycle (propanoate metabolism, Pyruvate metabolism, valine, leucine, and F I G U R E 1 (a) VCD (circle) in 10 6 cells/mL and the percentage of viable cells (triangle) during perfusion cultivation for replicate A (grey) and replicate B (black).The control is pictured as white circles and triangles, respectively, which has been published previously but are included for reference(MacDonald, Nöbel, Martínez, et al., 2022).For the control, a bleed was applied from Day 11−17 at 65 × 10 6 cells/mL.Bleed and growth rate over time are shown in panel (b, c), respectively.(d) Amino acids consumed as a percentage of the amino acid concentration in the feed media for replicates A and B at the end of the first growth on Day 13 (light grey), the start of the 2nd growth on Day 28 for replicate A and Day 33 for replicate B as well as the end of the second growth on Day 38 and 42 for replicates A and B, respectively.Glycine, Alanine and Glutamine were produced and are displayed in their absolute concentration.Amino acids are marked as essential and as nonessential for growth(Hefzi et al., 2016).Bars represent the mean between replicate A and B (n = 2) plus one standard deviation.(e) Cell-specific productivities over the different phases of culture are shown as boxplots representing first to the third quantile, mean marked by a black line and whiskers showing maximum and minimum in the bottom right.The different phases are indicated in panel A as encircled number (1) growth, (2) stationary and (3) second growth in white and grey for the control and knockout strain, respectively.F I G U R E 2 Results of (a) PCA plot for proteomics samples of the knockout (circular dots) and control (triangles) cell lines during growth (dark red) and start of the stationary (grey).(b) PCA plot for the knockout cell line over the time of culture, grouped into growth, stationary start, stationary end (purple) and 2nd growth (light red).(c) The log2-fold changes of pathways identified as differentially expressed with an adjusted p < 0.1 compared to the initial growth.PCA, Principal Component Analysis.isoleucine degradation), or by providing reduction equivalents (oxidative phosphorylation).These findings suggest that the triple knockout CHO cell line orchestrates a complex rearrangement of metabolic pathways in response to growth arrest, characterized by a reduction in growth-related processes and an upregulation of energyproducing pathways.Reaffirming the stability of the genetic modifications in our cell line throughout the culture period, BAK, BAX or BOK were not detected by proteomics at any stage of the culture, indicating that no partial reversal of knockout was present.The potential of false negatives in pathway enrichment studies was of concern, as partial up or down regulation of pathways are challenging to detect, whilst being of metabolic importance.Based on the result of the pathway enrichment above, we hypothesized an arrest in cell cycle and manually assessed the pathway in its expression over time using KEGG pathview (Figure 3).A downregulation of cell cycle proteins during growth arrest was observed.Single proteins like cyclin dependent kinase 7 (CDK7) or cyclin B (CycB) were upregulated during the stationary phase or showed a less pronounced down regulation and possibly contributed to the pathway not being identified during enrichment.However, other cyclin dependent kinases (CDKs) and most other proteins showed clear downregulation during the stationary phase.3.3 | Phase dependent response to oxidative stressGiven the observed upregulation of oxidative phosphorylation and the increased cellular age, it is plausible that ROS could accumulate over time.ROS are a by-product of cellular metabolism, particularly oxidative phosphorylation, and can cause damage to cellular components if they accumulate beyond a certain threshold(Dickinson & Chang, 2011).This ROS accumulation could potentially impact our observed growth patterns, prompting further exploration into the cell's oxidative stress responses and their effects on cell behavior in perfusion culture.Additionally, our findings showed a major upregulation of peroxisome activity and fatty acid degradation pathways during the stationary phase.These pathways and oxidative phosphorylation are among the primary contributors to intracellular ROS generation.Therefore, their increased activity potentially coincides with heightened ROS production.(Supporting Information Material S1: Figure4 and 5).Therefore, key proteins in the response F I G U R E 3 Cell cycle pathway as obtained from KEGG, showing the log2 fold change of proteins in order of growth, stationary end and second growth compared to the stationary start.Green, grey and red color illustrate down regulation, constant expression and upregulation compared to the start of the stationary, respectively.to oxidative stress were analyzed in their expression profiles over time (Figure4).Our investigation revealed intriguing distinctions based on the cellular localization of proteins with identical functions.For instance, the enzyme superoxide dismutase (SOD), which catalyzes the conversion of superoxide radicals to hydrogen peroxide, displayed significant differential regulation.During the stationary phase, the mitochondrial form of this enzyme, SOD2, was notably upregulated, much more than its cytoplasmic counterpart, SOD1.This trend was also evident for other antioxidant proteins such as glutathione peroxidase, peroxiredoxins (PRDX), and thioredoxin (TXN), where mitochondrial isoforms exhibited a stronger upregulation during the stationary phase.These findings suggest an increased mitochondrial oxidative stress during the stationary phase, underscoring the role of mitochondria in managing oxidative stress during growth arrest.Glutathione synthesis (GSS and GCL) was downregulated during the stationary phase, as was TXN reductase and Kelch-like ECHassociated protein 1 (Keap1), one of the key signalling proteins for oxidative stress.Further, both NADPH regenerating enzymes, glutathione reductase and TXN reductase (txnR), stayed either constant or decreased in abundance during the stationary phase.
ity of the CHO cell metabolism.Unraveling the metabolic puzzle is expected to hold tremendous capacity for future cell line engineering efforts focusing to on the generation of perfusion specific cell lines.Similarly, studying the long-term changes of cells undergoing a second growth phase would help to understand the potential for industrial application of the here presented work.In summary, the present work provides important insights into the complex cellular responses of CHO cells under high-density perfusion conditions, with potential implications for the biopharmaceutical industry.These findings are anticipated to pave the way for the generation of superior CHO cell lines, capable of enhancing biopharmaceutical production while maintaining robustness and viability in perfusion cultivation systems.AUTHOR CONTRIBUTIONS Matthias Nöbel and Esteban Marcellin conceptualised the study, Matthias Nöbel and Michael A. MacDonald performed the experiments, Matthias Nöbel and Craig Barry performed the proteomic analysis, Matthias Nöbel analysed the data, Verónica S. Martínez performed the mass balance analysis, Matthias Nöbel and Esteban Marcellin wrote the manuscript, all authors gave input throughout the work as well as read and approved manuscript.ACKNOWLEDGMENTS This work was supported by the Australian Research Council Training Centre for Biopharmaceutical Innovation (CBI) (project IC160100027).Elements of this research utilized equipment and services provided by the Queensland node of the National Biologics Facility (NBF) and the Queensland node of Metabolomics and Proteomics (Q-MAP).NBF is supported by Therapeutic Innovation Australia (TIA).Q-MAP is supported by Bioplatforms Australia TIA and BPA are supported by the Australian Government through the National Collaborative Research Infrastructure Strategy (NCRIS) program.Lars K. Nielsen was supported by Novo Nordisk Foundation grants NNF10CC1016517 and NNF14OC0009473.Verónica S. Martínez was supported by the Advance Queensland Women's Research Assistance Program (WRAP213-2019RD1).Open access publishing facilitated by The University of Queensland, as part of the Wiley -The University of Queensland agreement via the Council of Australian University Librarians.