Comparative Proteomic Characterization of Serum-Free Cultivated HepG2 Cells Reveals Upregulated Drug Metabolism and Increased Oxidative Stress Protection

Abstract

Fetal bovine serum (FBS) is commonly used in cell culture models despite being considered critical from both an ethical and scientific standpoint. Since it has been used for many years with a variety of different cells and cell types, research has become dependent on its use, and scientists hesitate to change their working model without solid data demonstrating the usability of serum-free cultivated cells in their desired applications. To shed light on the highly complex issue of dietary demands of cells in culture and the impact of media supplements on cellular behavior, we performed a comparative proteomic characterization between HepG2 cells cultivated in serum-free or FBS-containing conditions with a focus set on drug metabolism and oxidative stress response, the main application field of the HepG2 cell line. We observed the predicted upregulation of multiple pathways associated with drug metabolism and oxidative stress protection, as well as a strong overexpression of multiple antioxidative enzymes such as glutathione peroxidase and glutathione S-transferase at the protein expression level. We confirmed that increased enzyme expression correlates with higher enzyme activity in vitro and linked increased glutathione peroxidase activity to selenium supranutrition under serum-free conditions.

Introduction

With an increasing focus in science, politics, and society on animal welfare, much progress has been made to reduce, replace, and refine (3R) the use of animals in research and development. One major area for the implementation of the 3R principles in testing pipelines is the use of advanced in vitro cell culture models. These include tissue and organ cultures, microfluidic systems, organs-on-a-chip, and other models to better approximate the human in vivo situation and achieve reliable data without the use of animal experimentation. − Besides these advanced cell culture models, simpler in vitro models using cell lines and primary human cells are still of high importance, especially in early drug development, where high-throughput screening of compounds is necessary to assess the efficacy, pharmacology, and safety of a large number of potential drug candidates.

However, despite the recognition of human-derived in vitro models as important animal replacement and reduction tools, animal-derived components are still prevalent in these models. − This causes a continuous discussion about the ethical and scientific problems arising from the use of animal components for cell maintenance and proliferation, thereby driving efforts to replace animal components in cell culture. −

One of the more controversial animal components in in vitro cell culture is fetal bovine serum (FBS). FBS is currently the most important cell culture media supplement, supplying all of the necessary nutrients for cells to grow and proliferate. However, FBS has never been fully characterized since its introduction to cell culture in the 1950s. Recent advances in proteomics have shed light on this complexity. More than 3,000 proteins have been reported to be present in FBS, with only a fraction identified. , Due to its undefined nature, a potential influence of FBS components on scientific outcomes cannot be ruled out. Conversely, studies have concluded that FBS may alter the phenotypic stability of cells and interact with test substances, thus influencing experimental outcomes. − Furthermore, due to its variable nature, FBS likely contributes to the reproducibility crisis in biomedical research and, therefore, adds to the high clinical attrition rate. , Nevertheless, since FBS has been used for many years with a variety of different cells and cell types, research has become dependent on its use, leading to hesitation to change the working model despite knowing that the influence of FBS on experimental outcomes is far from understood.

Therefore, the use of serum-free media to avoid FBS is highly recommended. However, switching to serum-free medium can be challenging. Cells cultivated under serum-free conditions need to exhibit growth characteristics and morphology similar to those cultured under serum-containing conditions while also maintaining their cellular functions and signaling pathways. Due to the complexity of FBS, identifying the necessary factors for cellular survival is not an easy task, and protocols have been developed to simplify the process of serum-free media development. Additionally, several serum-free media are commercially available for various cells and cell lines, and well-working serum-free media formulations have been published to promote the use of animal-derived component-free media.

In a previous study, the HepG2 cell line was used to compare different commercially available serum-free media. Media were compared based on their ability to maintain HepG2 cells with their typical growth characteristics and morphology. Subsequently, as HepG2 cells are commonly used as an early screening tool for hepatotoxicity and cytotoxicity, the cells were compared in different toxicological assays focusing on viability, mitochondrial toxicity, oxidative stress, and intracellular drug response. Two of the compared media maintained HepG2’s typical growth rates, morphology, and marker expression, as well as albumin and urea secretion. When compared in early toxicology screening assays, serum-free cultivated cells generally showed a slightly higher sensitivity in cytotoxicity assays. When assessing compounds categorized as most DILI concern by the FDA, HepG2 cells in serum-free conditions identified 6 out of 12 compounds correctly, while serum-free cells identified 7 or 8, respectively. In contrast, a lower oxidative stress response was observed under serum-free conditions.

To further understand these changes and investigate potential alterations in the proteomic landscape occurring in HepG2 cells upon changing from medium containing FBS to serum-free medium, a comparative proteomic study was conducted using media suitable for either long-term or short-term serum-free culture. Based on our previous data, a particular focus was set on drug-metabolizing enzymes and NRF2-regulated proteins. ,,

Materials and Methods

Materials

Media were obtained from different suppliers and tested for long-term culturing capacity and preservation of morphology and cellular function as previously described.

Cell Culturing

HepG2 cells (HB-8065, ATCC) were derived from ATCC, and cell banking was performed prior to use including quality control measures such as mycoplasma testing, sterility assessment, STR-analysis, and cross-contamination assessment. After thawing, cells were cultured in Dulbecco’s modified Eagle medium/nutrient mixture F12 (DMEM/F12 with HEPES, 31330) supplemented with 10% fetal bovine serum and 1 mM sodium pyruvate (all Gibco) at 37 °C and 5% CO₂. Subcultivation was performed twice per week using Tryp-LE (Gibco, 12604013) for detachment, and cells were seeded at 8 × 10⁴ cells/cm². To ensure full recovery after the thawing process, HepG2 cells were maintained in standard medium for three passages before starting the experiment. To generate proteomic samples, cells were adapted to different serum-free media using the inside adaptation strategy proposed by van der Valk et al. or kept in standard medium containing FBS for comparison (Table ). Media were supplemented with HEPES, sodium pyruvate (1 mM), or GlutaMax (2 mM) if not present in the medium already. Vitronectin coating (using VTN-N, A14700, Thermo Fisher, 0.5 μg/cm²) was used to facilitate attachment. Cells were seeded at 8 × 10⁴ cells/cm² and counted during each subcultivation process to monitor the cellular growth. From adaptation, cells were kept in their respective condition for a maximum duration of 20 passages, corresponding to 10 weeks in culture. If cells showed a significant reduction in growth rates, the culture condition was terminated early.

1. Media Characterization.

Product	Manufacturer	Classification	Animal and human-derived components	Concentration of animal-derived components
Standard medium (DMEM/F12 + 10% FBS)	Thermo Fisher	serum containing	FBS	10%
Advanced DMEM/F12	Thermo Fisher	serum-free	lipid-rich BSA	0.4 g/L (lipid-rich BSA)
TCM serum replacement	MP Biomedicals	serum-free	BSA, bovine transferrin	<0.1% (at final concentration)
Serum replacement 3	undisclosed	serum-free	BSA, bovine transferrin	0.77 g/L (total protein at final concentration)
Serum-free medium 2	undisclosed	serum-free, xeno-free	human serum albumin	-

Sample Preparation

For sample preparation, cells were seeded at 8 × 10⁴ cells/cm² in biological triplicate and cultured for 96 h. Cells were washed three times with ice-cold PBS, scraped off the plate in 1 mL of ice-cold PBS, and transferred to an Eppendorf microcentrifuge tube. Cells were centrifuged, the supernatant was discarded, and the pellet was resuspended in 300 μL of cell lysis buffer (see Table S1). Cells were ultrasonicated for 5 min and subsequently incubated with 100 U benzonase nuclease for 30 min at 37 °C. Subsequently, the lysate was cleared by centrifugation at 16,000g for 10 min at 4 °C and transferred to a new precooled microtube, and protein concentrations were determined using the BCA protein assay kit from Thermo Fisher. For each sample, a lysate containing 200 μg of protein was transferred for PAC digestion.

Proteolytic Digestion

Proteolytic digestion was performed using the Protein Aggregation Capture (PAC) protocol on the KingFisher Flex system from Thermo Fisher according to the manufacturer’s instructions. , Digestion was performed in digestion buffer (see Table S1) using sequencing-grade modified trypsin (1:20 w/v) for 1 h at 42 °C. The digestion process was quenched and acidified to 1% (v/v) trifluoroacetic acid (TFA). The supernatant was transferred for subsequent LC/MS/MS analysis.

Liquid Chromatography Electrospray Ionization Tandem Mass Spectrometry (LC-ESI-MS/MS)

Dried peptides were eluted in 150 μL of 50 mM HEPES/0.1% TFA and analyzed using an UltiMate 3000 RSLCnano UHPLC system (Thermo Fisher) coupled online to a hybrid TIMS quadrupole TOF mass spectrometer (Bruker timsTOF Pro) via a CaptiveSpray nanoelectrospray ion source. From each sample, 2 μL were loaded on a 25 cm C18 column (75 μm ID, 1.7 μm particles, Aurora) from IonOpticks (Australia) corresponding to 200 ng of peptides. Peptides were separated with 30 min linear gradients from 5% solvent A (1% ACN, 0.1% FA)–35% solvent B (99% ACN, 0.1% FA) at a flow rate of 300 nL/min. The TIMS-TOF was operated in the DIA-PASEF mode (m/z range: 100–1700; 1/K₀: 0.60–1.60 Vs/cm²).

Data Analysis

Bruker (TIMS-TOF Pro) *.d result files were analyzed using DIA-NN (1.8.1). The following settings were used: maximum mass accuracy tolerances set to 10 ppm for both MS1 and MS2 spectra, enzyme: trypsin; fixed modification: carbamidomethylation at cysteine; number of allowed missed cleavage sites: 1; minimum peptide length: 7. The relaxed-prot-inf option was used for library-free processing of the data. Data were searched against UniProt (entries: 20,365, release date: 05/2020) in combination with the common repository of adventitious proteins (cRAP) database containing common contaminants. Match between runs (MBR) was enabled, and the normalization option was switched on. The software output was filtered at a precursor q-value of <1% and a global protein q-value of <1%. Although the global protein q-value filter in DIA-NN does not mandate a minimum peptide count per protein, we observed high proteome coverage and redundancy in peptide identifications. Specifically, 109,959 unique peptides were detected across 8907 proteins, from which 8347 proteins were identified with ≥2 unique peptides, with a mean and median of 12/8 peptides, respectively, per protein (see Figure S1)

The output (.csv) file containing protein IDs and quantities was further processed and visualized by using different software packages. Proteins present in all conditions were preprocessed using Pareto scaling and mean centering and subsequently analyzed using principal component analysis (PCA) in SIMCA. Quantitative values were calculated for either the full experimental timeline (20 passages) or the first three passages only, and protein expression fold changes were submitted to Ingenuity Pathway Analysis (IPA). The significance of canonical pathways, diseases, and functions derived from IPA was determined using the right-tailed Fisher’s exact test. For comparison analysis, p-values were corrected using the Benjamini–Hochberg correction for multiple testing.

Additionally, data were log₂ and median-normalized (equal median normalization function), and selected proteins were compared on the passage level. Statistics on the passage level were determined by an unpaired two-sided Student’s t-test and corrected using the Benjamini–Hochberg correction. Proteins with an absolute log₂ fold change of ≥0.58 and a false discovery rate (FDR) of < 0.05 were defined as significantly regulated. Statistical significance levels are depicted in Table .

2. Statistical Significance Levels.

p-value	Description	Summary
> 0.05	not significant	ns
0.01 to 0.05	significant	*
0.001 to 0.01	very significant	**
0.0001 to 0.001	extremely significant	***
< 0.0001	extremely significant	****

Visualization was performed using different software, including R, IPA, STRING, InstantClue, and GraphPad Prism.

Data Availability

All of our raw DIA-PASEF files and the 2020-based search outputs have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD056319. We encourage the community, and ourselves in planned future work, to research these data against an updated, focused FASTA incorporating splice variants or proteoforms of particular biological interest, ideally combined with the latest DIA-NN versions.

Enzyme Activity Measurement

The activity of glutathione S-transferase was measured using the Glutathione S-Transferase (GST) Assay Kit (CS0410-1KT) according to the manufacturer’s instructions. Conjugation of l-glutathione to 1-chloro-2,4-dinitrobenzene (CDNB) was monitored by measuring absorbance at 340 nm in kinetic mode immediately after preparing the reaction tests and every minute thereafter. Ten measurements were conducted.

Glutathione peroxidase activity was measured using the Glutathione Peroxidase Activity Assay Kit (Fluorometric, ab219926) from Abcam, according to the manufacturer’s instructions, with the exception that bovine serum albumin was replaced with human serum albumin to avoid the use of animal-derived components. Fluorescence (excitation 420 nm; emission 480 nm) was measured in kinetic mode for 60 min, with measurements taken every 3 min. A standard curve was conducted using a sigmoidal, four-parameter logistic curve (4PL) to interpolate enzyme activities in cell lysates.

For both assays, cell lysates from cells cultivated for 10 passages in their respective conditions were used, and enzyme activity was normalized to the protein concentration to allow for comparison between samples.

Statistical significance was assessed using one-way ANOVA with Dunnett’s correction for multiple comparisons. Results were considered significant if adjusted p < 0.05. Statistical significance levels are depicted in Table .

Ethical Statement

The materials used for this research do not meet the standard of being completely free of animal-derived components. We are committed to continuously improving our sourcing and manufacturing processes to meet our high ethical standards and societal expectations.

Results

Sample Comparison

To investigate the influence of serum-free media on the proteomic landscape, we conducted a comparative proteomic study, comparing HepG2 cells cultivated under standard (FBS-containing) conditions to four different serum-free media (Figure A). Of those four media, two media (Advanced DMEM/F12 and TCM Serum Replacement) were selected as suitable for HepG2 long-term culture in previous experiments. Additionally, two serum-free media that previously showed either a low (Serum Replacement 3, SR3) or limited (Serum-free media 2, SF-M2) long-term culture capacity were selected. Three of these media (Advanced DMEM/F12, TCM Serum Replacement, and SR3) contain animal-derived components (bovine transferrin or bovine serum albumin), while one medium (SF-M2) contains only human-derived and human recombinant supplements (see Table ).

1.

Experimental workflow and sample characteristics. (A) Experimental workflow consisting of sample collection, preparation, measurement, and analysis. Cells were seeded in biological triplicate and cultivated in their respective media for up to 20 passages (10 weeks). Proteomic samples were subsequently prepared by cell lysis, protein quantification, and proteolytic digestion. Measurement was performed on a TIMS-TOF Pro MS instrument operated in the DIA-PASEF mode. Created in BioRender. Pfeifer (2025) BioRender.com/r29b883. (B) Characteristics of proteomic samples. Viability (rhombuses) and cell density (circles) were determined and compared to cells cultivated in an FBS-containing medium (standard). (C) Long-term culturing capacity of serum-free cultivated cells throughout the experiment. Termination was based on a significant reduction in cell growth. (D) Total number of proteins identified in each condition and passage across the different replicates.

Cells were adapted to the different serum-free media via an immediate transition to serum-free conditions (inside adaptation) and cultivated for up to 20 passages in their respective media. After 1, 3, 5, 10, and 20 passages, proteomic samples were collected for analysis. If cellular growth or viability was significantly diminished, then the culture condition was terminated. Early termination was performed for SR3 after three passages and for SF-M2 after 12 passages (see Figure C). Cells adapted to either Advanced DMEM/F12 or TCM Serum Replacement were kept for 20 passages in their respective medium before disposal.

To assess both viability and growth during the experiment, one additional biological replicate was carried out for each condition throughout the whole experiment, since sample preparation did not allow for cell counting (Figure B). For cells adapted to Advanced DMEM/F12 and TCM Serum Replacement, cell density remained comparable to cells cultivated under FBS-containing conditions throughout the whole experiment, and viability remained over 90%. Cells adapted to SR3 showed a strong decrease in cell density after three passages (12.76 × 10⁴ cells/cm²), accompanied by a slight decrease in viability (88.35%). Lastly, cells adapted to SF-M2 showed lower cell densities across the first five passages, again with a decrease in viability (83% in passage 5).

Throughout the experiment, cells in all media displayed an epithelial-like morphology. However, cells cultivated in serum-free media displayed cytoplasmic extensions, lower density within their cell aggregates, and rough edges in early passages (see Figure S2). This was not observed in mid- and late passages.

Global Proteomic Comparison of Serum-Free Cultivated HepG2 with Cells Cultured under Standard Conditions

From all measured samples, a total of 8907 unique proteins were identified at a 1% false discovery rate. Numbers of identified proteins across samples were comparable and ranged from 7877 to 8369 (Figure D). Initial data set analysis consisted of normality assessment using log₂ intensity histograms, sample correlation analysis, and distribution analysis of protein focus groups (drug-metabolizing enzymes and transporters). Intensity distribution was comparable throughout the different samples, with nearly Gaussian curve progressions. This is shown exemplarily for one replicate of Advanced DMEM/F12, TCM Serum Replacement, and Standard media for passages 1 and 20 (Figure A and see Figure S3 for all histograms). As a result, normality was assumed when applying statistics.

2.

Initial data set analysis. (A) Normality assessment using log₂ intensity histograms, shown exemplarily for Standard media, Advanced DMEM/F12, and TCM Serum Replacement for either passage 1 or passage 20. (B) Exemplary protein rank plot. Distribution of all proteins found in Standard media, passage 1, demonstrating the dynamic range and the signal distribution of drug-metabolizing enzymes and transporters. (C) Sample correlation matrix using the Pearson correlation. (D) Exemplary pairwise correlation plots to demonstrate biological and technical reproducibility.

Protein rank plots were used to depict the dynamic range of the signal intensity. Protein focus groups, consisting of drug-metabolizing enzymes and transporters (DMETs), were distributed throughout the whole range, with drug-metabolizing enzymes showing generally higher signal intensities compared to transporters (Figure B). Interestingly, when comparing the different conditions, medians for drug-metabolizing enzymes in serum-free conditions were generally higher, while medians of transporters were comparable throughout the different conditions and passages (Figure S4).

For a global comparison, all samples were compared using the Pearson correlation. Generally, a high correlation was observed between all samples, with all Pearson correlation coefficients higher than 0.96 (Figure C). The highest correlation was observed between biological replicates, indicating biological reproducibility and minimal technical variance (Figure D). Furthermore, a high correlation was observed between different conditions in the early passages. Strongest differences were observed for comparisons between SR3Passage 3 with all other samples and SF-M2Passage 5 with all standard conditions.

Additionally, to assess proteome stability within a condition, Pearson scores between early and late passages (passage 1 vs 20 and passage 3 vs 20) were closely examined. Interestingly, the highest correlation scores were observed for TCM Serum Replacement, showing that the HepG2 proteome remains stable after long-term culture under serum-free conditions (Figure S5). Pearson scores in Standard medium and Advanced DMEM/F12 were slightly lower, albeit still higher than 0.98.

To assess variances driving changes among sample groups, we performed principal component analysis (Figure ). A significant separation was observed between cells cultivated in standard medium containing FBS and cells cultivated under serum-free conditions, emphasizing distinct protein expression patterns between these subgroups. Furthermore, samples of cells cultivated under serum-free conditions displayed dimensional separation in early and late passages, with early passages exhibiting high correlation within their respective clusters, while late serum-free conditions showed higher variability. To identify key driving proteins for differences between serum-containing and serum-free conditions, we examined the loading plot for proteins with the highest absolute values in M1.p[2] (Figure B and Table )

3.

Principal component analysis of serum-free cultivated cells and cells cultivated in a standard medium. (A) Scores plot of principal component analysis. Three clusters were observed, consisting of standard cells, early serum-free passages, and late serum-free passages. (B) Loading plot of principal component analysis. Proteins with the highest absolute M1.p[2] loadings are marked and annotated in Table .

3. Proteins with the Highest Absolute M1.p[2] Loadings.

Protein	UniProt AC	M1.p[2]	Protein	UniProt AC	M1.p[2]
histone H1.2	P16403	0.119635	fatty acid-binding protein, liver	P07148	–0.15698
histone H4	P62805	0.115685	delta-1-pyrroline-5-carboxylate synthase	P54886	–0.116324
aldo-keto reductase family 1 member C2	P52895	0.104147	cathepsin D	P07339	–0.0997433
l-lactate dehydrogenase A chain	P00338	0.0776691	isocitrate dehydrogenase [NADP], mitochondrial	P48735	–0.0988487
histone H1.5	P16401	0.0757374	α-1-antitrypsin	P01009	–0.0869023
aldo-keto reductase family 1 member C3	P42330	0.0718159	squalene synthase	P37268	–0.0833412
tubulin α-4A chain	P68366	0.0652538	keratin, type I cytoskeletal 19	P08727	–0.0830891
cytochrome c oxidase subunit 6C	P09669	0.0596803	fatty acid synthase	P49327	–0.0807452
eukaryotic initiation factor 4A-I	P60842	0.0588053	microsomal triglyceride transfer protein large subunit	P55157	–0.0770185
long-chain-fatty acid-CoA ligase 4	O60488	0.056403	cocaine esterase	O00748	–0.0770162

Interestingly, three histones were found to strongly contribute to differences between serum-free and standard cultivated cells, with positive M1.p[2] loadings, indicating higher expression in cells cultivated in FBS-containing medium compared with cells cultivated serum-free. As a result, other histones were analyzed as well. In total, 16 histones and histone variants were detected, with 11 in the M1.p[2] top 10 percentile (Table S3). Additionally, two members of Aldo-keto reductase family 1 showed high M1.p[2] loadings. In contrast, 7 out of 10 proteins with the lowest M1.p[2] loadings participate in metabolic pathways, and four are involved in lipid metabolism or biosynthesis.

To further analyze differences between serum-free and standard medium cultivated cells, significantly regulated proteins were determined by calculating quantitative values of each protein with a subsequent comparison of serum-free conditions to cells cultivated in standard medium. Since cells were cultured for different numbers of passages, two comparisons were made. To capture the whole cultivation time, quantitative values were calculated for all 20 passages to compare cells cultivated in either Advanced DMEM/F12 or TCM Serum Replacement with cells cultivated under FBS-containing conditions. Additionally, quantitative values were calculated based on the first three passages only, to compare all serum-free media with the standard conditions and compare those serum-free media with high long-term culturing capacity to those with lower long-term culturing capacities. Proteins with an absolute log₂ fold change equal to or higher than 0.58, corresponding to a 1.5-fold change, with q-values below 0.05 were considered significantly regulated.

As for the comparison between serum-free cultivated cells and cells cultivated under FBS-containing conditions across 20 passages, 383 proteins were differentially regulated in Advanced DMEM/F12 cultured cells, while 346 proteins were differentially regulated in TCM Serum Replacement cultured cells (Figure A,B). Of those proteins, 235 proteins were significantly regulated in both Advanced DMEM/F12 and TCM Serum Replacement. When analyzing the first three passages only, 224 proteins were significantly regulated in Advanced DMEM/F12, 267 proteins in TCM Serum Replacement, 425 in SR3, and 475 in SF-M2 (Figure C). Interestingly, SR3 and SF-M2 displayed a high number of differentially expressed proteins specific to either or both of these two conditions, with 148 proteins differentially expressed solely in SR3 and 141 proteins differentially expressed in SF-M2. Additionally, 63 proteins were differentially expressed in SR3 and SF-M2.

4.

Comparison between serum-free cultivated cells and cells cultivated under FBS-containing conditions. (A) Volcano plots generated from log₂ fold changes and −log₁₀(p-values) for either 20 or 3 passages. (B) Venn diagram showing overlap between significantly regulated proteins in Advanced DMEM/F12 and TCM Serum Replacement when compared to cells cultivated under standard conditions based on 20 passages. (C) Venn diagram showing overlaps between significantly regulated proteins in all serum-free media when compared to cells cultivated under standard conditions based on the first three passages. (D) Heatmap of all proteins identified as significantly regulated in any of the analyzed serum-free conditions. Hierarchical clustering was performed using the Euclidean distance metric with the average linkage.

Hierarchical clustering of log₂-normalized and median-centered signal intensities of all significantly regulated proteins (n = 900) was used to visualize changes between the different conditions across the passages (Figure D). In general, changes between conditions were visible but subtle. The strongest differences were again observed between cells cultivated under FBS-containing conditions with SR3 and SF-M2, as expected based on the number of differentially regulated proteins identified in each condition. Correlation values calculated for comparisons between serum-free and FBS-containing conditions based solely on these 900 identified proteins ranged between 0.86 and 0.96 (Figure S6). Generally, TCM Serum Replacement showed the highest correlation with cells grown under standard conditions.

Taken together, we were able to demonstrate proteome stability under serum-free conditions across passages for Advanced DMEM/F12 and TCM Serum Replacement, and a high correlation of samples derived from serum-free cultivated cells with those cultured under FBS-containing conditions.

Canonical Pathway Analysis Suggests Improved Oxidative Stress Protection in Serum-Free Cultivated Cells

To investigate differentially regulated pathways, comparisons between cells cultivated serum-free and cells cultivated under FBS-containing conditions were analyzed using Ingenuity Pathway Analysis Software (IPA) from Qiagen. The core analysis tool was used to analyze each serum-free to standard comparison separately. Comparison analysis was performed to investigate similarities and differences between different serum-free media when compared with the standard condition. The obtained information from IPA consisted of p-values of overlap, activation z-scores, and protein ratios. The p-value of overlap is based on the null hypothesis that molecules in the data set do not overlap with a particular pathway, disease, or function and is calculated using the right-tailed Fisher’s exact test. For comparison analysis, p-values were corrected using the Benjamini–Hochberg correction for multiple testing. Activation z-scores predict activation or inhibition of a certain pathway, function, or disease based on a comparison with information in the Ingenuity Knowledge Base. The protein ratio describes the overlap of analysis-ready proteins (proteins passing the threshold) with the total number of proteins in a pathway.

A general canonical pathway analysis revealed multiple pathways significantly upregulated in Advanced DMEM/F12 and TCM Serum Replacement, while fewer pathways were significantly downregulated (Figure A,B). To focus only on pathways with strong prediction of activation or inhibition, pathways were filtered according to their activation z-score (absolute z-score ≥ 2) and p-values (<0.05). Clustering of canonical pathways into networks showed multiple upregulated pathways related to metabolism, biosynthesis, or degradation/utilization/assimilation. Downregulated pathways were related to cellular immune response and cytokine signaling (only Advanced DMEM/F12), cellular response to stimuli and gene expression (only TCM Serum Replacement), or generation of precursor metabolites and energy, metabolism, and protein localization (both).

5.

Canonical pathways analysis. (A) Bubble plot of canonical pathways differentially regulated in Advanced DMEM/F12 when considering all 20 passages. Only pathways with absolute z-scores ≥ 2 and p-values <0.05 are shown. Pathways are clustered according to functionality. Size indicates the protein ratio of regulated to unregulated proteins, while color indicates z-scores. (B) Bubble plot of canonical pathways differentially regulated in TCM Serum Replacement when considering all 20 passages. Only pathways with absolute z-scores ≥ 2 and p-values <0.05 are shown. Pathways are clustered according to functionality. Size indicates the protein ratio of regulated to unregulated proteins, while color indicates z-scores. (C) Bar plots visualizing p-values, z-scores, and protein ratios of canonical pathways related to oxidative stress protection or drug metabolism for the comparison of cells cultivated in Advanced DMEM/F12 in comparison to cells cultivated in FBS-containing medium based on 20 passages in culture. (D) Bar plots visualizing p-values, z-scores, and protein ratios of canonical pathways related to oxidative stress protection or drug metabolism for the comparison of cells cultivated in TCM Serum Replacement in comparison to cells cultivated in FBS-containing medium based on 20 passages in culture. (E) Heatmap of canonical pathways related to oxidative stress protection or drug metabolism for comparisons between serum-free media and the standard based on the first three passages only.

Of particular interest, and taking into account previously generated in vitro data, were pathways of detoxification, xenobiotic metabolism, cellular response to stimuli, cellular stress and injury, ingenuity toxicity list pathways, and metabolism networks (Table S2). In this case, pathways with absolute z-scores <2 were also considered. A total of 98 different pathways were found, and the pathways of highest interest were further analyzed (Figure C–E).

The highest predicted pathway inhibition was observed for “cytoprotection by HMOX1” for both media, while the highest predicted activation was observed for NRF2-mediated oxidative stress response (Advanced DMEM/F12 vs Standard medium, z-score: 2.449; TCM Serum Replacement vs Standard medium, z-score: 2.829). Additionally, high predicted activation was observed for “phase IIconjugation of compounds” (Advanced DMEM/F12 vs Standard medium, z-score: 2.236; TCM Serum Replacement vs Standard medium, z-score: 2.646) as well as “glutathione redox reactions I” (z-score: 2 for both comparisons). These results support previously generated in vitro results showing a higher sensitivity of serum-free cultivated cells toward cytotoxicity as well as lower susceptibility toward oxidative stress.

To further investigate the possibility of increased oxidative stress protection, proteins involved in the NRF2-mediated oxidative stress response were examined (Figure A). In total, 237 proteins belong to this pathway according to IPA, of which 131 were detected in the analyzed samples. Only a subset of proteins was significantly regulated; nevertheless, a clear trend toward upregulation of NRF2-mediated oxidative stress response was observed. When considering 20 passages, a total of 79 out of 131 proteins had log₂ fold changes >0 when comparing Advanced DMEM/F12 with Standard medium, while slightly more proteins had positive log₂ fold changes when assessing TCM Serum Replacement (82 out of 131; Figure B). Additionally, 10 proteins were identified as significantly differentially regulated in Advanced DMEM/F12, with two downregulated and eight upregulated proteins, while in TCM Serum Replacement, nine proteins were differentially regulated, all of them upregulated.

6.

NRF2-mediated oxidative stress response. (A) Simplified schematic depiction of NRF2-mediated oxidative stress response overlaid with measurement and prediction data for Advanced DMEM/F12 vs Standard medium based on the whole experimental timeline (20 passages). Pink-labeled molecules were upregulated in the data set. Orange (arrows and molecules) indicates a predicted upregulation. Gray molecules were found in the data set but did not pass the threshold. The pathway was adapted from IPA. (B) Hierarchically clustered heatmap visualization of 133 proteins involved in NRF2-mediated oxidative stress response for Advanced DMEM/F12 versus Standard medium and TCM Serum Replacement versus Standard medium based on the whole experimental timeline. (C) Hierarchically clustered heatmap visualization of 131 proteins involved in NRF2-mediated oxidative stress response for all serum-free media versus standard medium based on the first three passages. (D) Log₂ fold changes of proteins determined to be up- or downregulated in at least one passage in Advanced DMEM/F12. (E) Log₂ fold changes of proteins determined to be up- or downregulated in at least one passage in TCM Serum Replacement.

When considering only the first three passages, a similar trend was observed (Figure C). Most proteins displayed positive log₂ fold changes when compared to the standard condition (Advanced DMEM/F12:79 proteins; TCM Serum Replacement: 98 proteins; SR3:75 proteins; SF-M2:81 proteins out of 131). Of these, seven proteins were significantly upregulated in Advanced DMEM/F12, eight in TCM Serum Replacement, nine in SR3, and ten in SF-M2.

Significantly regulated proteins based on 20 passages were additionally analyzed on the passage level (Figure D,E). For most upregulated proteins, a similar progression was observed, with an initial increase in the log₂ fold change until passage 5 and subsequently either a stable progression or a slight decrease. Downregulated proteins showed a stable progression across all 20 passages.

Altogether, we observed a general upregulation of drug-metabolic and detoxification pathways under serum-free conditions, with the strongest predicted upregulation for NRF2-mediated oxidative stress response.

Glutathione S-Transferase and Glutathione Peroxidase Activity Upregulated in Serum-Free Media

As a major part of oxidative stress protection, the expression of antioxidant enzyme families was more closely investigated. Glutathione peroxidases (GPX), catalase (CAT), and superoxide dismutases (SOD) are considered first-line defense antioxidants and are therefore of particular interest. Additionally, the glutathione S-transferase (GST) family was included. This enzyme family belongs to the Phase II detoxification enzymes, which are important for the detoxification of electrophilic xenobiotics and electrophilic metabolites.

In total, two members of the superoxide dismutase family, 15 members of the glutathione S-transferase family, and 5 members of the glutathione peroxidase family as well as catalase were detected throughout the different conditions (Figure A,B). Of these proteins, most but not all are involved in NRF2-mediated oxidative stress response based on the IPA pathway analysis (SODs, CAT, 9 GST members, and 2 GPX members).

7.

Expression analysis of antioxidant enzymes under serum-free conditions compared to cells cultivated under FBS-containing conditions. (A) Log₂ fold changes of antioxidant enzyme families for Advanced DMEM/F12 vs Standard. (B) Log₂ fold changes of antioxidant enzyme families for TCM Serum Replacement versus Standard.

Strong upregulation was observed for two members of the glutathione S-transferase family (GSTA1/P08263 and GSTA2/P09210) in both Advanced DMEM/F12 and TCM Serum Replacement. Additionally, four members of the glutathione peroxidase family (GPX1/P07203, GPX2/P18283, GPX4/P36969, and GPX8/Q8TED1) showed strong upregulation in both media. A moderate upregulation was observed for catalase. To investigate whether increased expression leads to higher enzyme activity, we measured glutathione S-transferase activity and glutathione peroxidase activity in cell lysates of cells cultivated for 10 passages in serum-free media (Advanced DMEM/F12 or TCM Serum Replacement) or Standard medium. For both enzyme families, increased activity was observed for cells cultivated under serum-free conditions (Figure ). As we observed increased abundances for multiple enzymes of these families in the proteomics experiments, it is likely that the increased activity is due to increased enzyme expression.

8.

Antioxidant enzyme activity in cell lysates after 10 passages in a serum-free culture. Enzyme activity was normalized to protein concentration to allow for comparison between samples. Ordinary one-way ANOVA with Dunnett’s correction for multiple comparisons was used to assess statistical significance (N = 3; n = 2). (A) GST activity in nmol/mL/mg protein. Serum-free cultivated cells displayed higher GST activity. The highest enzyme activity was observed for cells cultivated in TCM Serum Replacement. (B) GPX activity in nmol/mL/mg protein. Serum-free cultivated cells show a significantly increased GPX activity. The highest enzyme activity was observed for cells cultivated in Advanced DMEM/F12.

Glutathione S-transferase activity was moderately higher in Advanced DMEM/F12 (fold change of 1.57 ± 0.1), while a strong increase was observed for cells cultivated in TCM Serum Replacement (fold change of 3.16 ± 0.37). Glutathione peroxidase activity was highest in cells cultivated in Advanced DMEM/F12 (fold change of 2.31 ± 0.41), while enzyme activity was moderately increased in cells cultivated in TCM Serum Replacement (fold change of 1.79 ± 0.13).

Selenium Supranutrition in Serum-Free Medium Drives Increased GPX Activity

As it is crucial to understand which changes in media composition drive changes in proteomic profiles under serum-free conditions, available literature was screened for media supplements influencing the expression of GST and GPX. Furthermore, molecular activity prediction, BioProfiler, and upstream regulator prediction available in IPA were used based on data for Advanced DMEM/F12 versus Standard to identify direct and indirect relationships between media supplements and antioxidative enzymes.

As FBS has been shown to significantly reduce intracellular reactive oxygen species (ROS) levels in HepG2 cells as well as other models, the simplest reason for increased enzyme expression might be a higher concentration of ROS under serum-free conditions, leading to increased antioxidative enzyme expression. ,

Nevertheless, different media supplements were shown to have direct effects on antioxidant enzyme expression as well, with the most prominent candidates being selenium. Analysis of the mammalian selenoproteome has identified five members of the glutathione peroxidase family (GPX1, GPX2, GPX3, GPX4, and GPX6) to contain selenium in the form of selenocysteine, and the bioavailability of selenium was demonstrated to directly influence the expression of selenoproteins. − Additionally, selenium concentrations were shown to directly affect GPX1 and GPX4 expression as well as GPX activity in HEK293 cells, and selenium supranutrition was shown to increase GPX activities in rats, pigs, and chickens. −

As the increased bioavailability of selenium should also influence other selenoproteins, proteomic data were analyzed for expression of selenoproteins. Besides the GPX members, nine additional selenoproteins (SELENOF/O60613, SELENOH/Q8IZQ5, SELENOT/P62341, SELENON/Q9NZV5, SELENOP/P49908, SELENOS/Q9BQE4, SELENOO/Q9BVL4, SELENOK/Q9Y6D0, and DIO1/P49895) were detected. From these selenoproteins, increased expression of eight proteins was observed in Advanced DMEM/F12, while seven proteins were upregulated in TCM Serum Replacement (Figure ). Additionally, when analyzing upstream regulator predictions, a higher selenium activity was predicted for all serum-free conditions when compared to cells cultured in standard medium (Figure S7).

9.

Expression analysis of selenoproteins expressed in serum-cultivated HepG2 cells in comparison to cells cultured in standard medium. (A) Log₂ fold changes of selenoproteins for Advanced DMEM/F12 versus Standard. (B) Log₂ fold changes of selenoproteins for TCM Serum Replacement versus Standard.

Based on available literature, the results obtained from molecular activity predictions (see Figure A), upstream regulator predictions, and the increased expression of multiple other selenoproteins, the possibility of selenium as a key driver for increased expression and activity of the GPX family member was experimentally validated.

10.

Molecular activity prediction and the impact of selenium nutrition on GPX and GST activity. Ordinary one-way ANOVA with Dunnett’s correction for multiple comparison was used to assess statistical significance for GPX and GST activity (N = 3; n = 2). (A) Molecular activity prediction conducted in IPA based on data from Advanced DMEM/F12 versus Standard. Red and green colors indicate up- and downregulation; orange and blue indicate activity prediction. (B) Influence of selenium nutrition on GPX activity. (C) Influence of selenium nutrition on GST activity.

Therefore, cells were adapted to a self-made serum-free medium based on the formulation of Advanced DMEM/F12 and treated with different concentrations of sodium selenite, the source of selenium commonly used in serum-free media formulations.

After 5 days in culture, cells were assessed morphologically. Lower selenium concentrations led to a slight impairment in growth, which was displayed by lower confluency. Additionally, cells cultured in the absence of selenium displayed altered morphology, and a higher number of apoptotic cells was observed, highlighting the importance of selenium for cellular survival (Figure S8).

Subsequently, the activity of GPX was measured in cell lysates. Additionally, GST activity was determined as well. Enzyme activity was normalized to protein concentration (mU/mg protein), and subsequently fold changes were calculated in comparison to cells cultivated under FBS-containing conditions.

A clear dose-dependent response was observed after treatment with different concentrations of sodium selenite when measuring GPX activity (Figure B). Higher sodium selenite concentrations led to a significant increase in activity, with a fold change of 4.1 ± 0.97 in the highest concentration when compared to the standard condition. In contrast, the GPX activity was strongly diminished when no sodium selenite (0.33 ± 0.13) or only low concentrations (0.51 ± 0.24) were present. Additionally, a nonlinear regression model was used to calculate the EC₅₀ value for GPX activity (0.0053 mg/L) as well as the fold change between the lower and upper plateaus of GPX activity (fold change 10.66; Figure S9).

In contrast, no dose response was observed for GST with the tested selenium concentrations (Figure C). A depletion in activity to ∼50% was observed for all conditions, except in the absence of sodium selenite, where activity dropped to 9.8%.

Taken together, we were able to show that multiple antioxidant enzymes are upregulated under serum-free conditions, correlating with increased enzyme activity, and identified selenium as a key driver of increased GXP activity.

Discussion

The in vitro culture of cells and tissues is an important tool in cellular and molecular biology and is used extensively to study the physiology of cells and disease pathways or to investigate the safety and metabolism of novel drug candidates. Especially with the increasing focus on the 3R principles, in vitro systems are gaining particular interest as potential replacement tools for animal experimentation. , As such, more advanced models are being developed to approximate the human in vivo situation to generate more reliable/translational data.

In this respect, it is necessary to apply the 3R principles not only to animal experimentation but also to those systems intended to replace animal studies. − Unfortunately, animal-derived components such as FBS, Matrigel, or growth factors are often used to maintain and proliferate cells in vitro, thereby compromising the true potential of in vitro systems as “pure” animal replacement tools. ,, Besides this, animal-derived components, especially variable products such as FBS or Matrigel, have the ability to significantly influence experiments due to unknown interactions with the target cells and therefore bias any generated data. Especially as FBS is still ill-defined, it is difficult to estimate the impact FBS may have on experimental outcomes.

Previously, we successfully adapted the HepG2 cell line to serum-free conditions for the performance of several in vitro safety assays. Interestingly, we observed a higher sensitivity in cytotoxicity assays, while the oxidative stress response was reduced. To understand the underlying causes for these changes in cellular behavior, a comparative proteomic analysis was conducted to compare cells cultivated under either serum-free or FBS-containing conditions.

As expected, we observed significant differences between serum-free cultivated HepG2 cells and cells cultivated in the presence of FBS. However, and most importantly, taken as a whole, these changes did not affect the global proteome as correlations between different samples were shown to be high throughout all conditions and passages. As it is crucial that cells retain their key characteristics upon adaptation, these findings are of high importance to those scientists considering moving from FBS toward serum-free cell culture. Furthermore, using the Pearson correlation, we were able to demonstrate the stability of the HepG2 proteome comparing early and late passages. Most interestingly, the highest correlation scores were observed for TCM Serum Replacement.

The driving factors that cause differences between samples were analyzed by principal component analysis and investigated more deeply on the pathway as well as protein expression level. Principal component analysis revealed two primary factors separating samples: aging and the availability of serum.

Interestingly, when examining key driving proteins for differences between serum-containing and serum-free conditions, we found that multiple histones contributed significantly. Therefore, we analyzed histones and histone variants on the protein expression level and compared those to cells cultivated under FBS-containing conditions. For Advanced DMEM/F12 and TCM Serum Replacement, changes in protein expression were generally below the threshold. In contrast, 5 out of 16 histones were found to be significantly downregulated in SR3 and SF-M2 (Table S3). A general downregulation of histones may be associated with histone degradation, a process that has been linked to cellular homeostasis, DNA damage, and cellular stress. − Furthermore, degradation of histone H2B was observed under insulin/IGF signaling-mediated nutrient stress conditions in C. elegans epidermal stem cells. Therefore, nutrient-related stress might indeed be a plausible explanation for the poor performance of these two media.

These observations were supported by the findings from our pathway analysis. Generally, the most significantly regulated pathways across all serum-free media were related to biosynthesis and metabolism. As serum-free media contain only a fraction of the supplements present in serum, these results were to be expected. A downregulation of oxidative phosphorylation, the main pathway for ATP generation in higher animals and plants, as well as electron transport, ATP synthesis, and heat production by uncoupling proteins after three passages in culture was only observed for SR3, albeit to a limited extent (Table S4). After 20 passages in culture, these pathways were also found to be downregulated in Advanced DMEM/F12 and TCM Serum Replacement, accompanied by a prediction of mitochondrial dysfunction. Mitochondrial energy production is dependent on reduced nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FADH2), which are generated from the oxidation of carbohydrates, fatty acids, and to a lesser extent amino acids. −

Furthermore, multiple micronutrients, such as B vitamins, ascorbic acid, α-tocopherol, selenium, zinc, coenzyme Q10, and lipoic acid, acting as either cofactors or antioxidants, are important for mitochondrial function. − Therefore, we hypothesize that energy production is altered under serum-free conditions due to a decreased/altered availability of substrates for utilization or micronutrients. As nutrient-related stress and energy generation were identified as key factors for survival under serum-free conditions by PCA and pathway analysis, further investigations should focus on identifying which media supplements are missing in SR3 and SF-M2 and how energy generation can be improved in Advanced DMEM/F12 and TCM Serum Replacement.

Within this study, our aim was to better understand changes in drug metabolism and oxidative stress response; therefore, we investigated pathways and proteins involved in these functionalities. Multiple pathways were significantly regulated, with most pathways showing positive z-values predicting upregulation, including “phase Ifunctionalization of compounds,” “phase IIconjugation of compounds,” and “NRF2-mediated oxidative stress response.” Additionally, we confirmed that increased antioxidant enzyme expression correlates with the higher activity of glutathione S-transferase and glutathione peroxidase. Given our previously generated in vitro results, these results are of high importance. We were able to show that the aforementioned higher sensitivity of HepG2 cells toward cytotoxic compounds and higher oxidative stress protection indeed have underlying mechanistic causes and are not a mere consequence of lesser protective agents (e.g., albumin) and potentially altered concentrations of antioxidants able to directly scavenge reactive oxygen species (e.g., glutathione, vitamin C, or vitamin E) present in the serum-free media. Nevertheless, it is important to understand how the media composition induces the observed changes in drug metabolism and oxidative stress response pathways and what implications these findings have for the use of HepG2 cells in early toxicity screening assays.

Given the inexpensive costs as well as the consistent and reproducible assay performance, HepG2 cells are most often used in early assessments and high-throughput screenings of drug candidates as a Tier 1 model for DILI assessments. Considering this, the upregulation of proteins related to drug metabolism and oxidative stress response is particularly interesting, especially since it is known that HepG2 cells have low expression and activity of drug-metabolizing enzymes, such as the CYP3A and CYP2C families. ,−

An increase in drug-metabolizing and NRF2-controlled enzymes might lead to an overall better performance and improved identification of potential DILI risks during early compound screens. A comparison between our data and the findings of Sison-Young et al. revealed that multiple proteins with low expression levels in HepG2 when compared to cryopreserved primary hepatocytes (cPHH) were significantly upregulated in serum-free cultivated HepG2 cells (see Table S5). However, the changes caused by serum-free conditions are considerably smaller when compared to the differences between HepG2 and cPHH. Therefore, while it is possible that serum-free conditions contribute toward a slightly more hepatocyte-like behavior in in vitro toxicity studies, the effect is most likely limited, and HepG2 cells should still not be used as a replacement tool for human primary hepatocytes. Nevertheless, additional studies comparing serum-free cultivated HepG2 cells and HepG2 cells cultivated in the presence of FBS with primary human hepatocytes should be conducted to gather additional insights.

Additionally, the underlying causes driving the upregulation of drug-metabolizing and antioxidant enzymes should be addressed. However, as the observed findings can be influenced directly or indirectly by absent supplements or supranutrition, this matter is rather complex.

We have shown that increased glutathione peroxidase activity is driven by selenium supranutrition under serum-free conditions. Interestingly, while Advanced DMEM/F12 contains 5 μg/L sodium selenite corresponding to ∼2.28 μg/L selenium, FBS contains selenium levels between 8.14 and 68.72 μg/L, corresponding to 0.81–6.87 μg/L if used at 10%. , However, despite the similar concentrations, selenium in serum is mainly incorporated into the selenoproteins GPX3 and SELENOP in the form of selenocysteine or bound to albumin. Therefore, selenium might be more readily available in serum-free media, leading to increased levels of selenoprotein expression and GXP activity. Interestingly, selenium has also been shown to impact the expression and activity of certain CYP450 isoforms. , Selenium supplementation was shown to prevent Aflatoxin B1 induced upregulation of CYP1A1, CYP1A2, CYP2A6, and CYP3A4 in chick liver. Furthermore, the increased activity of CYP1A2 and CYP2D25 and the decreased activity of CYP3A29 have been shown in Landrace pigs after selenium supplementation, while selenium deficiency led to decreased activity of CYP1A2, CYP2D25, and CYP2E1. Therefore, additional studies should include CYP450 isoforms when investigating the impact of selenium supplementation in a serum-free cell culture.

In contrast, the increased activity of glutathione S-transferase and catalase expression remain unresolved. One might argue that as the oxidative stress response is a complex system aiming to keep the production of reactive oxygen and the detoxification of reactive intermediates in balance, one possible explanation is a higher concentration of ROS due to the absence of multiple protective agents that are usually present in FBS. FBS has already been shown to significantly reduce intracellular ROS levels in HepG2 cells as well as other models. , In comparison to serum-free media, FBS contains a wide range of antioxidants that are able to protect cells from oxidative stress. In comparison, serum-free media generally contain only a subset of antioxidants such as ascorbic acid and glutathione but often miss other agents, especially fat-soluble antioxidants such as tocopherol, retinol, or lipoic acid. Furthermore, FBS contains a significantly higher concentration of serum albumin, which has been shown to reduce intracellular ROS production. , Therefore, cells might respond to higher ROS levels by the overexpression of antioxidant enzymes such as GST and catalase in a compensatory mechanism to ensure cellular survival.

Additionally, different media supplements were shown to influence antioxidant enzyme expression and activity directly. While selenium did not influence GST activity after 5 days of treatment, Coskun et al. were able to demonstrate that selenium increased GST activity in larvae. As increased GST expression was observed only at later passages, especially for Advanced DMEM/F12, selenium might still be a contributing factor, albeit not identified in our experiments due to the short-term treatment. Additionally, GSTA1 was identified as a potential regulator of lipid accumulation, with increased enzyme expression levels after treatment with free fatty acids, while catalase expression was shown to increase in cardiac mitochondria of mice fed a high-fat diet. , Taken together, additional experiments are necessary to unravel which changes in media composition are driving the changes in the antioxidant enzyme expression.

Lastly, it is necessary to address the remaining issues of other animal-derived components besides FBS. Both Advanced DMEM/F12 and TCM Serum Replacement still contain animal-derived components: bovine serum albumin (both media) and bovine transferrin (only TCM serum replacement). Although the concentrations of both supplements are low, they should still be considered critical from not only an ethical but also a scientific standpoint, as especially bovine transferrin was shown to contain a xenoantigen (Neu5Gc) contamination, which might impact assay outcomes. − Therefore, the next steps should include the optimization of the media formulation based on the known formulation of Advanced DMEM/F12, taking into account the generated proteomic results for optimal performance of HepG2 cells in early toxicology applications. Furthermore, an optimized medium should be tested for other hepatic cell lines, such as HepaRG, which are also commonly used in DILI assessments as well as primary human hepatocytes.

Overall, we showed that the use of serum-free media increases the drug metabolism and oxidative stress protection in HepG2 cells. This can be considered beneficial when assessing the safety of novel drug candidates in vitro. Especially when measuring oxidative stress response, an adaptation of timelines and assay-specific thresholds might be necessary, and further experiments should focus on experimentally evaluating the underlying causes of changes in enzyme expression.

Glossary

Abbreviations

3R

reduce, replace, and refine

Adv.

advanced DMEM/F12

CAT

catalase

DMETs

drug-metabolizing enzymes and transporters

DMEM/F12

Dulbecco’s modified Eagle medium/nutrient mixture F12

FBS

fetal bovine serum

FDA

false discovery rate

GPX

glutathione peroxidase

GST

glutathione S-transferase

IPA

ingenuity pathway analysis

PCA

principal component analysis

SF-M2

serum-free medium 2

SOD

superoxide dismutase

SR3

serum replacement 3

TCM

TCM serum replacement

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.5c00100.

• Buffers (Table S1); z-scores of canonical pathway analysis focused on selected pathways related to detoxification, xenobiotic metabolism, cellular response to stimuli, cellular stress and injury, ingenuity toxicity list pathways and metabolism (Table S2); AUC fold changes of histones (Table S3); z-scores of canonical pathways, related to energy, ATP synthesis, and oxidative phosphorylation (Table S4); comparison of proteins involved in drug metabolism and oxidative stress response with literature values comparing HepG2 with cryopreserved human hepatocytes (PHH) (Table S5); histogram showing the number of unique peptide sequences identified per protein across all 8907 proteins (Figure S1); microscopic images of cells throughout the whole experimental timeline (Figure S2); log₂ intensity histograms of all measured samples (Figure S3); signal intensity distribution of drug-metabolizing enzymes and transporter (Figure S4); Pearson scores of early vs late passages within a condition (Figure S5); Pearson of all proteins significantly up- or downregulated in at least one condition (Figure S6); predicted upstream regulator activity, derived from IPA analysis (Figure S7); microscopic images after selenium treatment (Figure S8); and nonlinear regression model for GPX activity (Figure S9) (PDF)

#.

J. Singh and A.Z. contributed equally. Conceptualization: LM.P., J. Singh, A.Z., M.K., F.F., J. Sensbach, F.P., D.W., and P.H. Funding acquisition: F.P. Project administration: D.W. Sample analysis: M.K. and J. Singh. Data analysis: L.M.P., J. Singh, and A.Z. Writingoriginal draft: L.M.P. Writingreview and editing: L.M.P., J. Singh, A.Z., M.K., F.F., J. Sensbach, F.P., D.W., and P.H.

The authors declare the following competing financial interest(s): Authors LM.P., J. Sensbach, and P.H. were employed by Merck Healthcare KGaA. Authors J. Singh, A.Z., F.F., M.K., F.P. and D.W. were employed by Merck KGaA.