Comprehensive Lipidomic Analysis of Recombinant Protein-Producing CHO Cell Cultures: Insights for Media and Process Development

Abstract

Chinese hamster ovary (CHO) cell expression systems are widely utilized for biologics manufacturing due to their efficient recombinant protein expression and human-like post-translational modifications. With an increasing demand for bio-therapeutic proteins, the development of high-yielding CHO cell lines and refining bioprocess parameters are key to achieving reliable, efficient, and cost-effective therapeutic protein production. Many lipidomic studies on CHO cells have highlighted the crucial roles of lipids in supporting cell growth and specific productivity, offering valuable insights for optimizing culture media and metabolic pathways. Among these lipid classes, sterols such as cholesterol and its downstream intermediates have been investigated in CHO cells, though their functions remain less thoroughly explored compared to overall lipid metabolism. In this study, comprehensive lipidomic profiling, including cholesterol and esterified cholesterol, was performed on two distinct basal media and feed conditions, which identified lipid composition changes that offer mechanistic insights into how culture conditions potentially influence membrane dynamics. These findings provide a foundation for future development of culture media and bioprocess control strategies that can ultimately improve yield and process consistency.

INTRODUCTION

Chinese hamster ovary (CHO) cells are widely regarded as the leading mammalian host cell expression system for commercial biologics production, owing to their ability to efficiently express recombinant proteins with human-like post-translational modifications. Enhancing their productivity remains central to reducing production costs and achieving scalable supply. This imperative has led to the development of high-producing CHO cell lines and optimization of bioprocess parameters to achieve robust, cost-effective therapeutic protein production. Evaluating different culture conditions in CHO cells using traditional metrics for evaluating cell culture performance, such as viable cell density (VCD), integral viable cell density (IVCD), and product titer, remains essential to characterize growth kinetics and productivity.

Lipids play a fundamental role in cell growth by providing structural components for membranes, which allow the cells to divide. Additionally, lipids are also energy resources, and signaling molecules such as phosphatidylinositol (PI), sphingolipids, and sterols act as secondary messengers in key growth-related signaling pathways.^1–3) These signals help to regulate nutrient uptake, anabolic metabolism, and ensure proper cell cycle progression. In mammalian cultures, it is known that cells rely on fatty acids (FAs) and other lipids for energy, especially in the phase of exponential growth. Lipid supplementations of species such as FAs and cholesterol have also been shown to improve viability and proliferation in CHO cells.^4,5) Many studies have demonstrated the dynamic nature of the lipidome in CHO cells, especially in growth and productivity.^1,6) One study reported that the reduction in diacylglycerol (DG) as the culture progresses has led to a decrease in growth as a negative impact on cell membrane composition.^6–8) Additionally, the increase in both PI and triacylglycerol (TG) species has shifted the cell metabolism from energy consumption to energy storage.^7–10) Ultimately, integrating lipidomics in process monitoring has led to the identification of lipid biomarkers such as TG (16:1/18:1/18:1), a specific lipid that correlates with higher productivity, allowing for the selection of high-yielding clones.⁶⁾ These have shown that lipidomics data from biologics-producing cell lines can provide insights directly relevant to productivity, critical quality attributes (CQAs), protein yield, and cell growth, as well as supporting broader bioprocess development and quality control (QC) efforts.

While previous studies have focused mainly on total lipid content, FA composition, or broad lipid classes, high-resolution mapping of TG chain length and saturation, cholesterol and esterified cholesterol (CE) within cell lines used for biologics has not been widely reported, which makes this study a distinct advancement in bioprocess characterization. This study aims to provide new mechanistic insights into lipid remodeling linked to cell growth and the transition between growth and production phases, with the potential for enhancing predictive process models. Additionally, the rapid quantitative assessment of cholesterol and CE has been made feasible with an optimized liquid chromatography-mass spectrometry (LC-MS) analytical workflow.

It is widely reported that many sterol intermediates involved in the cholesterol biosynthesis pathway are responsible for playing regulatory roles in cholesterol homeostasis maintenance. Cholesterol homeostasis is tightly regulated by the interplay between complex networks of cellular processes in the body to ensure there is an ample supply of cholesterol while preventing toxic accumulation of cholesterol in the membranes. Cholesterol and its downstream intermediates, such as CE and oxysterol derivatives, play an important role in the regulation of cholesterol efflux, striking the dynamic balance between biosynthesis, uptake of lipoproteins, export, and esterification.¹¹⁾ Cholesterol esterification is the cell’s mechanism of storing and transferring cholesterol while preventing toxicity from excess unesterified cholesterol. It was previously reported that the involvement of the cholesterol esterification pathway could play a key role in the regulation of cell growth and division.¹²⁾

Mass spectrometric methods for sterols analysis have conventionally relied on protocols involving saponification and/or sample derivatization, followed by gas chromatography-MS or LC-MS detection techniques.¹³⁾ Such methods, however, require tedious sample preparation and the use of complex solvents. To the best of our knowledge, there are currently limited methods reported for the combined detection of various sterol classes, including CEs, using LC-MS technology. Most present methods are focused on the analysis of a single class of sterols, either cholesterol and its precursors,^14,15) oxysterols,^16–18) or CE alone,^19–21) or a combination of two sterol classes.^22,23) Having an analytical method that can simultaneously detect various classes of sterols in a single workflow enables direct comparison of different sterol classes to identify key trends that are correlated with cell culture performance.

This study provides a comprehensive comparison of two distinct basal and feed media conditions applied to a recombinant biologics-producing CHO-K1 cell line. Cell growth kinetics were characterized by VCD and integrated IVCD, while product titer was quantified to assess productivity. Although both culture conditions supported robust growth, Media A demonstrated better IVCD and titer than Media B, indicating enhanced cell line productivity. Importantly, we observed that many lipid species exhibit opposing correlations with growth and productivity in CHO cells. We propose a unified, flux-based model whereby cells undergoing exponential growth prioritized phospholipid synthesis and free cholesterol incorporation for membrane biogenesis, whereas cells in the stationary/productive phase redirected metabolic processes towards TG/CE and lipid droplet (LD) biogenesis, polyunsaturated fatty acid (PUFA)-enriched membrane remodeling, and controlled ceramide (Cer) signaling, collectively enabling Golgi expansion and enhanced protein secretion. When this anticipated exponential-to-stationary growth phase lipid metabolism shift was not clearly observed under one of the media conditions, we instead noted elevation of selected glycosphingolipid levels consistent with membrane repair, suggesting that the cells could have entered a repair-dominant state. This framework mechanistically links lipid dynamics to culture phenotypes and identifies possible mechanisms behind media-dependent differences, which potentially impact recombinant protein production in CHO cells.

MATERIALS AND METHODS

CHO-based cell culture for monoclonal antibody production

Suspension CHO-K1 cells expressing an anti-Her2 monoclonal antibody (mAb) were grown in 14-day fed-batch cultures using the Ambr 250 micro-bioreactor system (Sartorius, Göttingen, Germany). Cells were inoculated at a density of 3 × 10⁵ cells/mL in 200 mL of HyClone ActiPro cell culture media (GE media; GE Healthcare, Logan, UT, USA) or EX-CELL Advanced CHO Fed-batch media (SAFC media; Sigma-Aldrich, St. Louis, MO, USA) supplemented with 6 mM L-glutamine (Sigma-Aldrich), herein denoted as Media A or Media B, respectively. The cultures were maintained at 37°C with dissolved oxygen controlled at 50% through air and oxygen sparging. The pH was regulated at 7.1 using 7.8% sodium bicarbonate as the base and carbon dioxide gas as the acidifying agent. Mixing was achieved using dual pitch-blade impellers operating at 300 rpm. For cultures in Media A, 6% Cell Boost 7a and 0.6% Cell Boost 7b feeds were added to the culture on Days 3, 5, 7, 9, and 11. For cultures in Media B, 10% of EX-CELL Advanced CHO Feed 1 (with glucose) was fed into culture on Days 3, 5, 7, 9, and 11. A specific volume of 45% glucose stock was added to the cultures when needed to maintain a final glucose concentration of 6 g/L. Three biological replicates were conducted for each fed-batch culture.

Extraction of intracellular lipid and sterol metabolites

Fed-batch CHO-K1 cell culture media were collected on Days 5, 7, and 11 of each culture for intracellular metabolomics analysis. For each sample, 1 × 10⁷ cells were collected and quenched in five volumes of cold 150 mM sodium chloride solution. The cell pellet was obtained through centrifugation at 4°C, 3400 × g, and the resulting supernatant was removed. The cell pellet was then subjected to a two-phase liquid extraction procedure as described previously.²⁴⁾ Briefly, methanol (Optima grade; Thermo Fisher Scientific, Waltham, MA, USA), 3.8 mM tricine (Sigma-Aldrich) solution, and chloroform (Gradient grade; Merck, Darmstadt, HE, Germany) (1:0.5:1 v/v/v, total 2 mL) were used to separate polar metabolites (aqueous fraction) from lipid species (organic fraction). The lower organic layer containing lipid and sterol metabolites was collected and purged with nitrogen gas before storing at −80°C. The whole extraction process was kept cold on ice.

LC-MS analysis of lipid metabolites

Analyses of the lipid extract were performed in triplicate using an ultra-performance liquid chromatography (UPLC) system (ACQUITY UPLC System; Waters, Milford, MA, USA) coupled to a Xevo G2 QTof Mass Spectrometer (Waters). An ACQUITY UPLC CSH C18 column (1.0 × 50 mm, 1.7 µm; Waters) was used for chromatographic separation. The solvent system consisted of Solvent “A”: acetonitrile (Merck), methanol, and water (2:2:1); Solvent “B”: isopropanol (Thermo Fisher Scientific); and both solvents contained 0.1% acetic acid and 0.025% ammonia solution as additives. The column oven was maintained at 45°C. The elution gradient used was as follows: 1% B for 1 min (0.1 mL/min), followed by 1% B–82.5% B for 9 min (0.1 mL/min), before B was increased to 99% for a 5-min wash (0.15 mL/min) and re-equilibration for 2.2 min at 1% B (0.1 mL/min).

Mass spectrometric analysis was performed using electrospray ionization (ESI) in positive and negative modes. The mass spectrometer settings used were as follows: mass range, 100–1800 m/z; resolution, 10000; cone gas flow, 40 L/h; desolvation gas flow, 600 L/h; desolvation temperature, 600°C; ESI capillary voltage, 2.0 kV (positive mode) and 1.0 kV (negative mode). Mass calibration was performed using sodium formate. QC samples consisting of equal aliquots of each sample were run at regular intervals during the batch LC-MS runs. Prior to LC-MS analysis, the lipid extracts were dried under nitrogen gas and reconstituted with 50% Solvent “A” and 50% Solvent “B.” The injection volume was 2 µL.

LC-MS analysis of sterols

Analyses of the lipid extract were performed in triplicate using the ACQUITY UPLC System coupled with a SYNAPT G2-Si Mass Spectrometer (Waters). Chromatographic separation was achieved using the same ACQUITY UPLC CSH C18 column as described above. The solvent system consisted of Solvent “A”: methanol and water (1:1) with 5 mM ammonium formate (Merck) and Solvent “B”: isopropanol with 5 mM ammonium formate. The column oven was kept at 30°C, and the flow rate was 0.1 mL/min. The elution gradient used was follows: 1% B for 2 min, followed by 1% B–95% B over 10 min, before B was increased to 99% for a 3-min wash and re-equilibration for 2 min at 1% B.

Mass spectrometric analysis was performed using positive ESI mode. The mass spectrometer settings used were as follows: mass range, 100–1800 m/z; resolution, 10000; cone gas flow, 40 L/h; desolvation gas flow, 600 L/h; desolvation temperature, 200°C; ESI capillary voltage, 2.0 kV. Mass calibration was performed, and QC samples were run as described above. The lipid extracts were dried and reconstituted with Solvent “B” prior to LC-MS analysis. The injection volume was 2 μL.

Data preprocessing and metabolite identification

The raw LC-MS data obtained from the lipid extracts were processed using an XCMS-based peak-finding algorithm.²⁵⁾ The QC samples were used for instrumental drift adjustment, and total ion count normalization was performed. Detected mass peaks were assigned putative metabolite identities by matching the respective masses (<10 ppm error) with the Kyoto Encyclopedia of Genes and Genomes (KEGG)²⁶⁾ and Human Metabolome Database (HMDB),²⁷⁾ and subsequently, metabolite identities were confirmed based on mass spectral comparison with available metabolite standards or with online mass spectral libraries.²⁸⁾ Unsupervised principal component analysis (PCA) was performed using Pareto scaling to assess the intracellular lipid profiles across Days 5, 7, and 11 using SIMCA software (version 13.0.3.0; MKS Umetrics, Umeå, Sweden). Subsequently, orthogonal projections to latent structures discriminant analysis (OPLS-DA) and the corresponding S-plot were employed to compare the profiles of Days 7 and 11 during the stationary/productive phase of Media A and Media B. Pearson’s correlation analysis was used to identify linear associations between the intracellular lipid profiles and culture outputs, including the specific growth rate, titer, and IVCD of the CHO cell cultures. The LC-MS analysis of the polar metabolites for the detection of reduced glutathione (GSH) and oxidized GSH (GSSG) and calculation of GSH:GSSG ratio is provided in Supplementary Method 1.

RESULTS AND DISCUSSION

In this study, we performed comprehensive lipidomic and sterol analyses on bioreactor cultures of CHO cells producing an anti-Her2 mAb. The results are presented in three parts: First, we focus on the LC-MS workflow for the detection of sterol classes (cholesterol and CE); second, multivariate and univariate analyses of combined intracellular lipid and sterol species profiles of the CHO cells cultured under two media conditions, Media A and Media B, are presented; and finally, we performed a detailed analysis of lipid trends that correspond to specific growth, titer, IVCD, and GSH:GSSG ratio.

Development of LC-MS analytical workflow for the detection of sterol classes

A screening platform for the detection of a variety of sterol classes, including cholesterol and its precursors, oxysterols, and CE using a single chromatographic LC-MS workflow without the need for derivatization is advantageous. This platform enhances analytical efficiency by simplifying sample preparation, offering broader analyte coverage, and achieving high-throughput performance in a shorter timeframe. In the field of bioprocessing, it is important to gain insights into and understand key differences in physiological processes across different cell lines, media, and process conditions, with the goal of improving cell growth and recombinant biotherapeutic protein production. To date, most such studies have focused largely on metabolites involved in amino acid, vitamin, and energy metabolism, with some reports on the trends of broad lipid classes and their associations with growth and productivity. The sterols analytical workflow reported in this study can profile multiple sterol classes in a single run without the use of a derivatizing agent. First, a mixture of eight isotopic standards was spiked into CHO cell extracts to determine the limit of detection (LOD), limit of quantification (LOQ), and linear range on the high-resolution LC-MS system. The list of isotopic sterol standards used with their corresponding mass-to-charge ratio (m/z) and retention times can be found in Supplementary Table S1.

Overall, good linearity (R² >0.98) was achieved for calibrants constructed with at least 5 points over linear dynamic ranges as indicated in Supplementary Table S1 and Supplementary Figures S1A–S1H. The LOD, LOQ, and linear range were determined in 20-µL aliquots of CHO-K1 cell extracts. LOD and LOQ were defined based on a signal-to-noise ratio of more than 3 and 10, respectively.

PCA, hierarchical clustering, and Pearson’s correlation analyses of combined lipid species

CHO cells from Days 5, 7, and 11 of fed-batch cultures were sampled, extracted, and analyzed as described in the “Materials and Methods (Extraction of intracellular lipid and sterol metabolites)” section. Sampling at Days 5, 7, and 11 corresponded to the mid-exponential phase, late-exponential phase, and stationary phase of the culture, respectively (Supplementary Figure S2). LC-MS workflows of lipids and sterols in CHO intracellular samples were performed separately. Processed data were normalized by total ion count and analyzed together. PCA and hierarchical clustering analyses were performed for CHO cell extracts under the two media conditions and are presented in Fig. 1.

Fig. 1. (A) Scores plot from the PCA analysis and (B) heatmap derived from hierarchical clustering for CHO cell samples obtained at different culture time points across the two media conditions, Media A and Media B. (A) The PCA scores plot provides an overview of the similarities and differences between cellular lipid profiles from both media at different culture phases, with green-colored circles representing Media A and blue-colored circles representing Media B samples. (B) The heatmap depicts the z-scores of selected lipid species (rows) across two media conditions (columns). Color intensity represents scaled z-score values from low (blue) to high (red). Hierarchical clustering was performed to group lipids based on similarity, revealing distinct lipidomic profiles associated with media conditions. CHO, Chinese hamster ovary; PCA, principal component analysis.

From the PCA in Fig. 1A, it is observed that cellular lipid composition changed significantly as the cultures progressed for both media conditions (as described by PC1, 61.0%). Additionally, cell samples from Media A and Media B exhibited distinct differences (PC2, 25.2%), particularly as the cultures progressed into the late-exponential and stationary phases. To gain more insights into the variation of individual lipid species, hierarchical clustering analysis (HCA) was performed.

In the heatmap derived from HCA (Fig. 1B), a general decrease in the levels of several key phospholipid species, including phosphatidylcholine (PC), phosphatidylethanolamine (PE), and cardiolipin (CL), was observed as the cultures progressed from Days 5 to 11 under both Media A and Media B conditions (cluster C1 in the figure). Additionally, other lipid species such as TG, lyso-phospholipids, ether-bond phospholipids, and CE showed increasing abundances from Days 5 to 11 under Media A; in contrast, specific TG and CE species were at the highest at Day 7 under Media B (cluster C2 in the figure).

To further study these time- and media-dependent changes, the trends of individual lipid species were then correlated with specific growth rate, titer, IVCD, and GSH:GSSG ratio for each culture, using Pearson’s correlation analysis.

According to the experimental data obtained, the specific growth rates and mAb productivities were linked but not linearly (Supplementary Figure S3). Specific growth rate calculates how fast biomass accumulates, while mAb productivity measures the efficiency of the CHO cell line in producing the mAb. The relationship between specific growth rate and mAb productivity can be broadly categorized into “growth-associated” and “non-growth-associated”. “Growth-associated” refers to culture conditions in which mAb production is positively correlated with cell growth, while under “non-growth-associated” conditions, mAb production is decoupled from growth. Under these conditions, a high growth rate first results in the accumulation of a large cell population; subsequently, cells then shift resources from growth to maximize antibody production by each cell.

Pearson’s correlation analysis partitioned the lipidome into membrane lipids and storage lipids, linking growth and the transition between growth and production phases in broad key patterns: (1) higher levels of core membrane components such as PC, PE, CL, and sphingomyelin (SM) species supported membrane synthesis during the rapid growth phase; (2) accumulation of storage lipid species such as TG during the late-stage exponential or stationary culture phases represented a possible indication of energy surplus.²⁹⁾ In addition, (3) alterations in cholesterol and CE levels could be associated with membrane fluidity and protein trafficking processes, and could serve as potential lipid biomarkers for titer levels, and (4) differences in the levels of selected glycosphingolipid species were potentially linked to membrane repair/turnover.

First, many membrane lipid species from phospholipid classes such as PC, PE, CL, and selected SM species exhibited strong positive correlations with specific growth rate (Pearson’s correlation coefficient, r ≥0.7) under both media conditions, suggesting that membrane biogenesis and remodeling are central to supporting cell proliferation (Supplementary Table 2). Additionally, the majority of DG species, as precursors to phospholipid synthesis, were also observed to be positively correlated with the specific growth rate. These observations supported the notion that the cells prioritize biomass synthesis during the rapid growth phase and corroborate with the findings by Ali et al., which showed similar lipidomic profiles and trends in CHO cells.⁶⁾ These systematic changes in the levels of various lipid classes in Media A and Media B, with growth phase and production (Supplementary Table 2), suggest that the tracking lipid metabolism could provide useful insights into CHO intracellular behavior.

In Media A, a greater number of TG, unsaturated PC/PE/PI, cholesterol, and CE species were negatively correlated with specific growth rate but positively correlated with titer and IVCD. In particular, these TG and CE species were present as storage lipids, suggesting that active lipid storage and mobilization processes were taking place.³⁰⁾ The higher abundance of these species during the late-exponential and stationary phases may also provide a buffer to changes in membrane lipid demand to maintain the homeostatic flux of unsaturated PC/PE/PI and cholesterol, in response to an increase in protein secretion through the membrane, hence improving antibody titer and maintaining productivity.^30,31) On the other hand, in Media B, a weaker correlation of the storage lipid levels with the specific growth rate was observed. These differences in the association between specific lipid subclasses, such as TGs and sterols, with respect to growth for Media A and Media B suggest the presence of media-dependent effects, where cell growth could be tightly coupled to the supply and turnover of structural, energy-related, and protective metabolites. It is hypothesized that these differences could be attributed to redox and potential stress responses that are unique to the cells under the two different media conditions.

The GSH:GSSG ratio served as the central redox axis for this lipidomic analysis. Pearson’s correlation analysis effectively partitioned media conditions into protective (positive correlation, r ≥0.7) versus stressed (negative correlation, r ≤−0.7) states. In this analysis, lipids that were positively correlated with GSH:GSSG ratio in Media A included selected PC/PE species (17 in total) containing 0 to 2 double bonds, which are less susceptible to oxidation damage.³²⁾ In Media B, positively correlated species were predominantly lipids from classes such as TG and CE (15 in total), suggesting cells could be actively leveraging a lipid storage strategy to maintain redox balance and at the same time protecting against lipid-induced damage³³⁾ (Supplementary Table 2).

Conversely, lipids with negative correlations to GSH:GSSG ratio in Media A, including FA(22:2), unsaturated TG, PC, PE, and PI species, as well as selected Cer/glycosylated-Cers and cholesterol species (34 in total), are consistent with previous reports on unsaturated lipids being more susceptible to lipid peroxidation, thus driving higher GSSG production.³⁴⁾ In Media B, a smaller number of negatively correlated lipid species to the GSH:GSSG ratio was observed (21 in total)—these comprised similarly of unsaturated PC, PE, and PI species, but no TGs and only one unsaturated FA (FA 24:1); additionally, several lyso-phospholipids were also identified, which suggested that the cells could be engaging in active membrane lipid remodeling to remove oxidized phospholipids³⁵⁾ (Supplementary Table 2).

Collectively, these observations suggest that cells cultured in Media A experienced higher oxidative stress in comparison to those cultured in Media B. For both conditions, the GSH:GSSG ratio also served as a key regulator of redox-dependent lipid homeostasis.³⁶⁾

Additionally, in Media A, the primary damage markers (FA(22:2) and Cer(d34:1)) that showed strong negative correlations with GSH:GSSG ratio could act as physiological stressors, triggering adaptive pathways that boost protein output.³⁷⁾ This observation mirrored the trends observed with temperature and osmotic hormesis in CHO literature, including a study that described how temperature-downshift-induced controlled stress and improved productivity by three times via the mTORC1 modulation and stress-response genes.³⁸⁾

Evaluation of Media A and Media B differences through supervised multivariate analysis

To further investigate possible reasons for the differences in culture performance between both media, particularly as the cultures progressed towards the late-exponential and stationary phases, we performed an OPLS-DA between Day 7 and 11 samples obtained from Media A and Media B (Supplementary Figure S4). From the resulting S-plot, lipid metabolites contributing the most to class separation between Media A and Media B samples were identified. These included DGs, FAs, and lyso-phospholipids, which were all present at higher levels in Media A samples (p(corr) <−0.7). As precursors for phospholipid synthesis, the elevation of DGs and FA species is in agreement with the higher cell growth rate seen in Media A. Additionally, increased lyso-phospholipid levels have previously been associated with Golgi membrane tubule formation,³⁹⁾ which might in turn lead to increased intracellular trafficking and hence protein productivity for cultures grown in Media A.

Another interesting trend is the elevation of different glycosylceramide (GlcCer) species in the two media—in Media A samples, lactosylceramide (LacCer) levels were found to be significantly higher than in Media B samples. LacCer is the downstream product of GlcCer and was previously reported to induce the growth of human aortic smooth muscle cells, potentially through LacCer-enriched glycosphingolipid signaling domains that activated cell proliferation.⁴⁰⁾ This seems to be aligned with the observation of LacCer levels being higher in Media A cultures, which had a relatively faster growth rate compared to Media B.

Conversely, GlcCer levels were elevated in Media B (p(corr) >0.7). The conversion of Cer to GlcCer has been reported to act as a buffer against Cer-induced apoptosis,⁴¹⁾ which might drive cells towards repair to improve stress adaptation and survival. While this enabled the culture to maintain a high cell viability (from Days 7 to 11), the potential bottleneck in converting GlcCer to LacCer, and to more complex glycosphingolipids, could impair membrane trafficking, which in turn might have negatively affected growth and productivity for cells cultured in Media B.

Through identifying lipid species and metabolic pathways that correlate with enhanced growth and productivity, this study provides a basis for uncovering key lipid mechanisms that are potentially important for process optimization and quality control in biologics manufacturing.

Opposing correlations of lipid subclasses with growth versus production are media dependent

Across 37 TG species and multiple sterol and SM subclasses identified in this study, a consistent pattern in Media A was observed with TGs, CEs, selected Cers (e.g., Cer(d18:1/16:0)), and PUFA-containing lipids negatively correlated with specific growth rate, yet positively correlated with titer and IVCD (Pearson’s r ≥0.7; Supplementary Table 2). In Media B, correlations for these species were weaker or reversed, indicating media-dependent regulation of lipid allocation. These data suggest a phase-dependent redistribution of lipid flux rather than independent, contradictory behaviors of each lipid class.

We next present representative lipid species from key subclasses that showed strong negative correlations <−0.7 with specific growth rate and strong positive correlations >0.7 with both titer and IVCD in Media A and present them in Figs. 2–4. In the same graphs, the trends of the same specific lipid species present in Media B were also shown to demonstrate the contrasting behavior of these lipids between the two media conditions. In particular, Cers, represented by Cer(d34:1), showed a strong negative correlation with specific growth rate in Media A (r = −0.89), while having a moderately positive correlation with the same variable in Media B (r = 0.64) (Fig. 2). Controlled Cer accumulation may support trafficking and stress adaptation, where excessive Cer is proapoptotic and therefore constrained in high-productivity cultures.⁴²⁾ This corroborates with the above finding where Cer was shown to be negatively correlated with GSH:GSSG and acted as a primary damage marker for oxidative stress. Additionally, FA (FA 22:2, r = −0.81) and TGs (TG o-56:2, r = −0.82) showing strong negative correlations with the specific growth rate in Media A (Fig. 2) are highlighted for their involvement in lipid storage and mobilization pathways. The strong negative correlations with growth potentially reflect a reduction in lipid storage during higher growth conditions, as lipid precursors are actively utilized for membrane lipid synthesis.

Fig. 2. Representative specific lipid species in lipid subclasses that demonstrate strong negative correlations with specific growth rate in Media A. Trends for the same species are provided for Media B as a basis for comparison. Red and green denote specific growth rates; black and blue indicate normalized peak areas (total ion count normalization) in Media A and Media B, respectively. Error bars indicate standard deviation, where n = 3.

Fig. 3. Representative specific lipid species in lipid subclasses that demonstrate strong positive correlations with titer in Media A. Trends for the same species are provided for Media B as a basis for comparison. Red and green denote titer; black and blue indicate normalized peak area (total ion count normalization) in Media A and Media B, respectively. Error bars indicate standard deviation, where n = 3.

Fig. 4. Representative specific lipid species in lipid subclasses that demonstrate strong positive correlations with IVCD in Media A. Trends for the same species are provided for Media B as a basis for comparison. Red and green denote specific growth rates; black and blue indicate normalized peak area (total ion count normalization) in Media A and Media B, respectively. Error bars indicate standard deviation, where n = 3. IVCD, integrated viable cell density.

Next, cholesterol (r = −0.74) and its ester derivatives, CE species (CE 18:1, r =−0.91), were also shown to be negatively correlated with specific growth rate in Media A (Fig. 2). These compounds are the major membrane constituents together with phospholipids and their availabilities in cell culture have direct effects on cell viability⁴³⁾ and membrane fluidity. Interestingly, polyunsaturated, ether bond-containing phospholipids (PC, PE) with carbon chain lengths of 36–40 also exhibited relatively high negative correlations with growth in Media A and Media B (e.g., PE(P-38:2), r = −0.93 in both media) (shown in Fig. 2). Such lipids are reported to be important for maintaining the stability of membrane regions that are rich in cholesterol,⁴⁴⁾ which could explain their relatively similar trends with cholesterol and CE species in Media A (Fig. 2).

In summary, changes in the abundance of TG, Cer, FAs, and overall membrane lipid diversity (cholesterol and CE), including plasma membrane and intracellular membrane vesicles, are tightly correlated with the specific growth rate of CHO cells, most likely coupled to the supply and turnover of structural metabolites. Conversely, the differences in their trends between media conditions suggest media-dependent effects on regulatory lipid networks and cellular adaptability.¹⁾

In addition to the specific growth rate, lipid metabolism differences observed between the two media conditions for CHO cells have been shown to be significant for titer. In this instance, the same representative specific lipid species depicted in Fig. 2 show strong positive correlations with titer (Fig. 3). FA 22:2, a PUFA species, could increase membrane fluidity by introducing kinks in acyl chains that prevent tight packing⁴⁵⁾ This enhanced fluidity can improve membrane dynamics, including vesicle formation, trafficking, and fusion, which are key processes for efficient protein secretion and export in CHO cells. The TG species, represented by TG o-56:2, where TG/CE-rich LD buffer lipotoxic stress and provide lipid precursors for Golgi remodeling, facilitating proper protein folding and hence secretion.⁴⁶⁾ These species are once again elevated in CHO cells cultured in Media A, corresponding to higher titer than those cultured in Media B (Fig. 3).

Another interesting observation is the similarly high positive correlations of long-chain ether bond-containing PE and PC species (containing 36- and 38-carbon molecules) with titer in both media (Supplementary Table 2). Specifically, our observations on long-chain ether-bond PE species, particularly plasmalogens, are in agreement with a previous study on CHO cells, with these species reported to improve vesicular protein secretion.⁴⁷⁾ These trends suggest that long-chain ether bond-containing lipid species alter vesicular trafficking and secretion pathways that are crucial to biologics production in CHO cells, although specific functional studies in this area are still emerging. Additionally, shorter chain ether-bond-containing PE and PC, including PE(P-34:2), PC(O-32:1), and PC(O-34:1) also showed positive correlations with the GSH:GSSG ratio, further suggesting that these compounds may act as endogenous antioxidants and can protect neighboring polyunsaturated lipids from peroxidation; it is speculated that these lipid species may also enhance membrane stability and resistance to oxidative stress encountered during bioprocess fermentation.⁴⁸⁾

Recent literature has shown the profound link between cholesterol and CE levels to the growth rate and productivity in CHO cells, primarily by supporting membrane integrity and key signaling and trafficking pathways essential for high-yield bioprocesses,⁴⁹⁾ which corroborates with our findings presented in Fig. 3, demonstrating strong positive correlations of cholesterol (r = 0.79) and CE (CE(18:1), r = 0.89) with titer in Media A. These trends are also in agreement with another previous study on CHO cells, which reported the important role that cholesterol plays in efficient endoplasmic reticulum to Golgi transport of proteins.⁵⁰⁾

Another critical cell indicator that contributes to the overall volumetric productivity of CHO cells is IVCD. In this study, the majority of long-chain (38 carbons and above) polyunsaturated phospholipid species appear to be positively correlated with IVCD in both Media A and Media B (Supplementary Table 2), potentially suggesting the need for increased vesicular transport to accommodate higher levels of recombinant protein production. However, other lipid species highlighted in Fig. 4 (FA 22:2, r = 0.84; cholesterol, r = 0.85; CE18:1, r = 0.85; Cer (d18:1/16:0), r = 0.92; and TG o-56:2, r = 0.81) showed much stronger positive correlations with IVCD in Media A but not in Media B. This difference in trends suggests that the availability of lipids may not be the only factor limiting cell growth and productivity in Media B CHO cultures.

For Media A, the increased abundance of PUFAs such as FA 22:2 and TGs may be an indication of accumulation of LDs as an energy reserve during growth and transition to the stationary phase, correlating with increased IVCD and productivity duration.⁵¹⁾ Cholesterol, as a major component of the membrane lipid bilayer, supports cell viability and productivity at high IVCD, while CE serves as storage forms of cholesterol, buffering free cholesterol during cell growth and stationary phase maintenance.⁵¹⁾ The balance between cholesterol and CE is important for membrane homeostasis during high IVCD. Cers, as represented by Cer(d34:1), show a positive correlation with growth, titer, and IVCD for cells cultured in Media A, but not in Media B. As discussed earlier, the slight decrease in Cer(d34:1) levels over time in Media B cultures could be potentially attributed to the relatively higher levels of GlcCer in the latter culture, which acted as a buffer against Cer-induced apoptosis,⁴¹⁾ and might drive cells towards repair to improve stress adaptation and survival. Overall, the balance and remodeling of these lipid species influence membrane and cell energy metabolism, directly impacting IVCD and associated titer profiles in CHO cell cultures.

Influence of lipid unsaturation on CHO cell productivity

Unsaturation in FAs, especially oleate (FA 18:1), has been shown to promote TG synthesis and neutral lipid droplet formation, protecting cells from lipotoxicity and enhancing cell survival during culture.⁵²⁾ Another study showed that increased desaturase enzyme activity in CHO cells elevated unsaturated FA content in lipid pools, correlating with higher maximum viability and extended culture longevity.¹⁾ These findings collectively show how modulating unsaturation in lipid pools could be a potential strategy for optimizing CHO cell bioprocess performance. In our study, the degree of unsaturation in phospholipids was observed to be highly correlated with titer in Media A CHO cultures. As shown in Fig. 5, up to 21 phospholipids with 0 or 1 unsaturation were found to be negatively correlated, while only one phospholipid was highly positively correlated with titer. On the other hand, phospholipids with higher degrees of unsaturation (≥2) were found to have highly positive correlations with titer (total of 15 IDs). An increase in the degree of unsaturation in phospholipids was previously reported to contribute to increased membrane fluidity secretion⁵³⁾ and vesicular dynamics, which might be required to accommodate higher levels of recombinant protein production in CHO cells.

Fig. 5. Distribution of unsaturation in phospholipids showing positive and negative correlations with titer in Media A.

Unified mechanistic model of lipid-flux redistribution observed under different media conditions

To explain the different correlations we observed between specific lipid species in Media A and B with growth and titer, we propose the following two-phase lipid-flux model, as described in Fig. 6.

Fig. 6. Lipid-flux redistribution from membrane synthesis to storage pathways underlies the opposing correlations of lipid species with growth versus productivity. Left panel: Exponential/fast growth phase where FA and DG are channeled to phospholipid biosynthesis and free cholesterol to membranes, limiting TG/CE synthesis and LD formation (dashed arrows). The levels of TG/CE and PUFA-rich species are therefore reduced as growth rates increase. Right panel: Stationary/productive phase where flux switches to TG, CE, and LD biogenesis. PUFA-rich species enhance membrane fluidity and vesicle budding/fusion, supporting Golgi remodeling and secretion. These changes link higher TG/CE/PUFA-rich levels with titer and IVCD (solid arrows). In impaired stationary phase shift, phase-transition efficiency is reduced. Weak fluxes from DG to TG (dotted arrow) and cholesterol to CE (dotted arrow) are observed, and LD formation is limited. Instead, Cer is routed to GlcCer, which favors repair over secretion,⁵⁴⁾ leading to a weaker coupling between lipid changes and secretion phenotypes. CE, esterified cholesterol; DG, diacylglycerol; ER, endoplasmic reticulum; FA, fatty acid; IVCD, integrated viable cell density; LD, liquid droplet; PC, phosphatidylcholine; PE, phosphatidylethanolamin; PI, phosphatidylinositol; PUFA, polyunsaturated fatty acid; SM, sphingomyelin; TG, triacylglycerol.

In the exponential phase, FA is converted to DG, which is then used as a precursor for phospholipid and SM synthesis for plasma and organelle membrane formation. Additionally, free cholesterol is used directly for membrane synthesis instead of being converted to CE. As such, there is less conversion of DG to TG species, as well as cholesterol to CE species, accounting for lower TG and CE levels during the exponential phase. These trends also suggest that LDs remain scarce during the high growth phase.⁵⁵⁾ On the other hand, during the stationary/productive phase, where proliferation slows down as culture progresses, FA/DG species are not highly utilized for growth and are instead converted to TG, while cholesterol is esterified to form CE species. Under these circumstances, LDs start to accumulate.^56,57) The incorporation of PUFAs into phospholipids also enhances membrane fluidity, further supporting LD to ER lipid exchange, potentially leading to an expansion of the Golgi for vesicle trafficking. This redistribution increases secretion capacity and extends culture longevity by buffering lipotoxic and oxidative stress. In short, cells cultured in Media A appear to promote this shift with strong TG/CE/PUFA gain; in contrast, cells cultured in Media B do not appear to sufficiently redirect FA and DG flux toward TG and CE during the stationary phase, accounting for the weaker correlations between TG and CE with growth and titer. Instead, cells cultured in Media B appear to divert Cer to GlcCer synthesis, an indication that they could be favoring membrane repair over protein secretion. This potentially reduces Golgi expansion and dampens the dependence of secretion on TG/CE/PUFA-rich pools, thereby weakening correlations. Together, this framework supports a view that media composition modulates the efficiency and timing of lipid flux shift proposed in Fig. 6.

CONCLUSION

In summary, the distinct lipidomic and sterol metabolite profiles driven by culture media composition are observed to be associated with cell growth, viability, and protein secretion efficiency in CHO-K1 cells. This work established that specific lipid species and sterols in plasma and intracellular membranes are tightly linked to CHO cell growth, titer, IVCD, and GSH:GSSG ratio, highlighting how lipid dynamics are closely linked to overall cellular metabolism and adaptability. By integrating these results into a two-phase lipid-flux framework, we show that media-dependent shifts in key lipid subclasses, including TGs, sterols, PUFA-rich phospholipids, and Cers, reflect differences in how cells transition from a growth-driven membrane biogenesis state to a secretion-driven lipid-storage and remodeling state. These media-specific lipid-flux behaviors provide mechanistic insights into how culture conditions influence cellular processes and key factors that may affect process efficiency and product quality. Leveraging these lipid metabolic differences through tailored media and process design may enable further improvements in the growth and production characteristics of recombinant protein-producing CHO cells. Another key aspect of future work should focus on investigating the potential effects of lipid metabolism changes under different media conditions on recombinant protein quality.

ABBREVIATIONS

CE, esterified cholesterol; Cer, ceramide; CHO, Chinese hamster ovary; CL, cardiolipin; DG, diacylglycerol; FA, fatty acid; IVCD, integrated viable cell density; LC-MS, liquid chromatography-mass spectrometry; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; SM, sphingomyelin; TG, triacylglycerol; VCD, viable cell density.

SUPPORTING INFORMATION

Supporting Information 1

Supplementary Method 1, Supplementary Tables 1 and 3, Supplementary Figures S1–S4

Supporting Information 2

Supplementary Table 2

Mass Spectrom (Tokyo) 2026; 15(1): A0192

This article was contributed by a Keynote Speaker of the 10th Asia-Oceania Mass Spectrometry Conference (AOMSC2025).