Simultaneous detection of nicotinamide adenine nucleotides and adenylate pool to quantify redox and energy states in mAb-producing CHO cells by capillary electrophoresis

Chinese hamster ovary (CHO) cells are predominant in the production of therapeutic proteins to treat various diseases. Characterization and investigation of CHO cell metabolism in a quick and simple way could boost process and cell line development. Therefore, a method to simultaneously detect seven redox- and energy-related metabolites in CHO cells by capillary electrophoresis has been developed. An on-line focusing technique was applied to improve the peak shape and resolution by using a 50 μm × 44 cm uncoated fused silica capillary. Key parameters and their interactions were investigated by design of experiments (DoE) and optimized conditions were determined by desirability function as follows: 24 °C, 95 mM, and pH 9.4 of BGE. The method was validated to ensure sensitivity, linearity, and reproducibility. The limits of detection (LODs) ranged from 0.050 to 0.688 mg/L for seven metabolites, and correlation coefficients of linearity were all greater than 0.996. The relative standard deviations (RSD) of migration time and peak area were smaller than 0.872% and 5.5%, respectively, except for NADPH, and the recoveries were between 97.5 and 101.2%. The method was successfully applied to analyze the extracts from CHO cells under two different culture conditions.

Keywords: Capillary electrophoresis; On-line focusing; Design of experiments (DoE); CHO cells; Nucleotides; Redox

Jiaqi Wang and Chen Wang contributed equally to this work.

Graph: Graphical abstract

Chinese hamster ovary (CHO) cells are the most widely used hosts for the production of biopharmaceuticals with a global market approaching $100 billion per year [[1]]. However, a deep insight into this host is still missing. Process development often consists of time-consuming screening works of cell clones or medium components. On the other hand, the market further requires therapeutics with lower cost by approaching the limits on protein yields and shortening the time line of process development. Therefore, it would be valuable to understand the metabolic characteristics of CHO cells to provide targets for genetic engineering and investigate their correlations with protein yields and critical quality attributes, which are highly cell line–specific and process-specific. Moreover, monitoring robust processes also ensures efficacy and safety of the therapeutics. In the bioprocess industry, capillary electrophoresis (CE) has been widely used to analyze proteins, sugars, amino acids, organic acids, metabolites, etc. [[2]]. Compared with the commonly used high-performance liquid chromatography (HPLC), gas chromatography (GC), and colorimetric method, CE possesses the advantages of smaller sample size and solution volume, higher efficiency and speed, and lower cost [[3]]. Moreover, it is an orthogonal technology to separate a wide range of compounds [[4]], making it a suitable high-throughput tool to analyze intracellular metabolites in CHO bioprocessing to achieve higher titer and quality of therapeutic proteins.

Nicotinamide adenine nucleotides (NADH, NAD+, NADPH, and NADP+) are coenzymes for many intracellular reactions involved in various metabolic pathways. NADH and NADPH are electron donors for oxidative phosphorylation and intracellular biosynthesis, respectively, while NAD+ and NADP+ are electron acceptors in various catabolic reactions and in pentose phosphate pathway (PPP), respectively. It was reported that NADH/NAD+ balance might be a key factor to tightly control the fluxes of glycolysis and TCA cycle in a CHO cell fed-batch process [[5]] and high-producing clones were associated with enhanced net NADH production [[6]]. On the other hand, NADPH/NADP+ coenzyme pair reflects significant cellular metabolic changes involving cell growth and therapeutic protein production [[7]]. Moreover, NADH/NAD+ and NADPH/NADP+ are also known as redox couples to defend intracellular oxidative stress [[8]]. Research on copper deficiency in a CHO cell culture attributed cell death to redox imbalance caused by insufficient NADPH and reduced glutathione to scavenge reactive oxygen species (ROS) [[9]].

Similar to nicotinamide adenine nucleotides, adenosine phosphates (AMP, ADP, and ATP) play a central role in metabolism as many intracellular reactions are energy-requiring. However, it is adenylate energy charge (AEC) that correlates more closely with cell growth and protein production rather than the individual concentration of AMP, ADP, and ATP [[10]]. Unlike the quick response of intracellular ATP throughout the course of CHO cell cultivation, AEC was relatively stable [[11]]. Therefore, it would be helpful, no matter in the process development or manufacture, if nicotinamide adenine nucleotides and adenosine phosphates could be measured in CHO cells to elucidate the metabolic state of this host.

The measurement of nicotinamide adenine nucleotides and adenosine phosphates is commonly conducted by colorimetric method and ion-pair reversed-phase HPLC, respectively. Different sample preparation and analysis processes are time-consuming. Recently, CE has been used to measure adenosine phosphates in a method designed for nucleotides and nucleotide sugars in CHO cells [[13]], and nicotinamide adenine nucleotides could be analyzed using a coated capillary developed to discriminate them from their derivatives [[14]]. However, CE is still underused in the biopharmaceutical industry where its main usage has been focused on proteins and carbohydrates in quality control [[15]]. Few works have explored nicotinamide adenine nucleotides and adenosine phosphates simultaneously in CHO cells by CE.

Therefore, a method on simultaneous detection of these seven metabolites has been developed in this work. An on-line focusing technology [[16]] was applied for sample enrichment by using an uncoated capillary and sodium tetraborate as background electrolyte (BGE). Moreover, a design-of-experiment approach was used to optimize capillary temperature, pH, and ionic strength of BGE and investigate their interactions. Linearity, limit of detection (LOD) and limit of quantification (LOQ), and recovery were also identified. Finally, AEC, catabolic reduction charge (CRC), and anabolic reduction charge (ARC) [[18]] were calculated under different conditions to elucidate the metabolic states of the CHO cells. To our best knowledge, this is the first report on simultaneous detection of nicotinamide adenine nucleotides and adenosine phosphates in CHO cells and we hope it will be helpful to more thoroughly understand redox balance and energy metabolism to ultimately improve quantity and quality of therapeutic proteins.

Nicotinamide adenine dinucleotide (NAD+, 98%) and triphosphopyridine nucleotide disodium salt (NADP+, 98%) were purchased from Solarbio (Beijing, China). Dihydronicotinamide adenine dinucleotide (NADH, 92%) and nicotinamide adenine dinucleotide phosphate (NADPH, 93%) were purchased from Beyotime (Shanghai, China). Adenosine 5-monophosphate sodium salt (AMP, 99%), adenosine 5-diphosphate disodium salt (ADP, 98%), and adenosine 5-triphosphate disodium salt (ATP, 98%) were purchased from Aladdin (Shanghai, China). All other reagents including tetraborate as BGE were purchased from Sigma (St. Louis, MO, USA). Individual stock solutions of standards (1.5 g/L NADH, 2.5 g/L NAD+, 2.4 g/L NADPH, 1.4 g/L NADP+, 1 g/L AMP, 1 g/L ADP, and 2 g/L ATP) were prepared in Milli-Q water and stored at − 80 °C.

Two glutamine synthetase–deficient CHO cell lines producing mAbs were generously provided by Hisun Pharmaceutical Co., Ltd. (Hangzhou, China) and cultivated in our in-house chemically defined medium, respectively. Exponentially growing cells were seeded in 250-mL shaking flasks (Corning, NY, USA) with 80 mL initial medium at 1 × 106 cells/mL for fed-batch cultures in a humidified 5% CO2 shaking incubator with a rotation speed of 130 rpm. The cultures were carried out by delivering in-house feed medium daily at 2% (v/v) of the initial working volume from day 1 to the end of the processes. Cell line A was initially cultured at 35 °C and shifted to 31 °C on day 5, while 3 mM oxidative glutathione (GSSG) was supplemented to cell line B on day 4 without temperature shift.

The extraction was conducted according to a previous report [[19]] with slight modifications. Briefly, 1 × 107 CHO cells were quickly mixed (v/v, 1:4) with quenching solution (0.9% NaCl, 4 °C) prior to pelleting for 1 min using a centrifuge at 1000g at 0 °C. Then, the cell pellets were resuspended with acetonitrile solution (50%, v/v, 4 °C), fully vortexed, and ice-incubated for 10 min. The mixtures were centrifuged for 10 min at 20,000g; then, the supernatant was collected and stored at − 80 °C. Eventually, the frozen supernatant was dried by SpeedVac (Thermo Savant, NY, USA) and stored at − 80 °C for further analysis.

The separations were performed on a Beckman PA800 Plus CE system with a photodiode array (PDA) detection system and 32 Karat Station as a control software (Beckman, Fullerton, CA, USA). An uncoated fused silica capillary (50 μm × 44 cm) was used for separation and 75 mM phosphate (pH 6.0) was used as sample matrix to enhance resolution. Different separation voltages and temperatures, along with concentrations and pH of BGE, were tested. Before the daily runs, the capillary was rinsed with 1 M NaOH for 10 min, water for 10 min, and running buffer for 10 min. To ensure the reproducibility, before each injection, the capillary was washed with 1 M NaOH for 30 s, 0.1 M NaOH for 30 s, methanol for 30 s, water for 30 s, and followed by BGE for 1 min; BGE in the vials were replaced after each run. In all flushing processes, the pressure of liquid and aeration was 20 psi and 100 psi, respectively. All the involved solutions including the samples were filtered with 0.22-μm membrane filters.

Concentration (95 mM, 110 mM, 125 mM) and pH (8.6, 8.8, 9.0, 9.2, 9.4) of BGE and separation temperature (22 °C, 24 °C) were optimized, using the D-optimal design by Design-Expert software (Stat-Ease Inc., Minneapolis, USA). Peak resolution (R) and separation time were set as responses to evaluate the results. Evaluation of the selected critical factors was carried out by ANOVA statistics for experimental design.

The linearity of the method was calculated using external standard curves. The LOD and the LOQ were determined from the ratio of peak signal and baseline noise level (S/N) as three- and tenfold, respectively. The precision of the method was reflected by within-day variation which was determined by repeated analysis of a single volume of spike standard compounds for six times. The recovery was performed by adding spike standards into cell extract. The quantity of each analysis was obtained from the relative calibration curve.

AEC, CRC, and ARC are calculated as follows:

$$\mathrm{AEC}=\left(\left[\mathrm{ATP}\right]+0.5\left[\mathrm{ADP}\right]\right)/\left(\left[\mathrm{ATP}\right]+\left[\mathrm{ADP}\right]+\left[\mathrm{AMP}\right]\right)$$

Graph

$$\mathrm{CRC}=\left[\text{NADH}\right]/\left(\left[\text{NADH}\right]+\left[\mathrm{NA}{\mathrm{D}}^{+}\right]\right)$$

Graph

$$\mathrm{ARC}=\left[\text{NADPH}\right]/\left(\left[\text{NADPH}\right]+\left[\mathrm{NAD}{\mathrm{P}}^{+}\right]\right)$$

Graph

The data were presented as mean ± standard deviation (S.D.), and the values obtained from CHO culture processes were means of three independent experiments. Meanwhile, Student's t test was used for statistical analysis, and p value < 0.05 demonstrated to be statistically significant.

We first tried to separate seven standards using water as sample matrix. As shown in Fig. 1a, poor performance was observed. The peak widths were too wide to discriminate NADP+ from NADH (nos. 2 and 3), and ADP from NADPH (nos. 6 and 7) as well. Hence, an on-line focusing technique was applied to enhance the sensitivity and resolution (see Electronic Supplementary Material (ESM) Fig. S1). Theoretically, a discrete pH step will be generated between an acidic sample matrix and a basic BGE, which causes analyte mobility changes to finally narrow the peak width. Therefore, 75 mM phosphate buffer (pH 6.0) was used as sample matrix according to the previous reports [[20]] to enhance the ionic strength of the analytes and large injection plug (10% of capillary length, 0.7 psi × 70 s) was used. It took good effects on the separation of all seven standards by narrowing the peak widths (Fig. 1b). Peak resolution between NADP+ and NADH (nos. 2 and 3) was well improved, while ADP and NADPH (nos. 6 and 7) were fully separated. We also tested the volume ratio of the sample matrix and the standards. The less phosphate participated, the worse peak resolutions were found. Eventually, 86.5% was determined to be the limiting percentage of sample matrix (data not shown).

Graph: Fig. 1Electropherograms of nicotinamide adenine nucleotides and adenosine phosphates in a pure water, b 75 mM phosphate, pH 6.0 as sample matrix. The BGE used was 125 mM tetraborate buffer at pH 9.2. Identification: (1) NAD + , (2) NADP + , (3) NADH, (4) AMP, (5) ATP, (6) ADP, (7) NADPH

We scanned the standard mixture under wavelength ranging from 190 to 300 nm via PDA detector under 20, 25, and 30 kV separation voltages. Increasing the separation voltage significantly reduced the integral migration time without significant impact on peak resolution (data not shown). Figure S2 (see ESM) shows the contour graph of the absorbance spectra under 30 kV. Generally, 250–280 nm was used for nucleotide analysis by CE-UV [[22]]; however, it did not work in this study as it was hard to identify the standard mixture when the wavelength exceeded 210 nm. Therefore, 200 nm was selected as detection wavelength in order to achieve appropriate sensitivity with reduced background noise.

BGE properties such as concentration and pH, along with separation temperature, are key factors which need to be optimized. Moreover, it is also critical to understand the role of each factor and the interaction with each other. Therefore, optimal factorial design (D-optimal) was used due to its accommodation of factor types and levels; parameter settings are presented in Table 1. The levels of temperature, buffer concentration, and pH were set on 2, 3, and 5 gradients, respectively. Peak resolution for NADP+ and NADH (R1), peak resolution for ADP and NADPH (R2), and separation time were set as responses. The design contained 27 runs and the mathematical model was analyzed by software as quadratic model. The model significance and significant factors in D-optimal design are presented in Table 2. ANOVA implied all the models were significant and there was only a 0.01% chance that it could occur due to noise. The factors with p < 0.05 were deemed to have significant effect on the responses. pH and ionic strength, as well as their interactions, were key factors for R1 and R2, while separation time was more prone to be affected by all the factors we set.

Factors and factor levels for the optimization

The significance of the model and corresponding significant factors

The detailed relationships between factors and responses are shown in Fig. 2. Increasing the pH and/or ionic strength of BGE would remarkably improve R1 and R2 (Fig. 2a and b); however, it would also lengthen the separation process (Fig. 2c). For instance, it would take 80 min to fully separate all seven metabolites under the conditions of pH 9.4, ionic strength 125 mM, and temperature 24 °C (Fig. 2c), which was unacceptable for high-throughput analysis. In general, pH of BGE determines the ionic charge on each analyte, which is crucial for the selectivity and migration order of separations, and alkaline BGE has been widely used in the analysis of nucleotides due to the stable electroosmotic flow. We could increase the alkaline BGE concentration to avoid the occurrence of wide peaks when the ionic strength of sample matrix is greater than that of BGE [[23]]. However, based on the Henderson-Hasselbalch equation for acids, the overall negative charge on metabolites is increased with more alkaline BGE, which leads the metabolites move to the cathode more slowly and increased the migration time [[24]]. Elevating separation temperature is a generally applied method to shorten the separation time, which was the same in this study (Fig. 2d).

Graph: Fig. 2Model graphs of different factors with different responses. a Effect of pH and ion strength of BGE on peak resolution 1. b Effect of pH and ion strength of BGE on peak resolution 2. c Effect of pH and ion strength of BGE on separation time. d Effect of temperature on separation time

Finally, the desirability function was employed to achieve the most appropriate factor levels with improved peak resolutions and separation time. The lower limits for R1 and R2 were set as 1; the upper limits for R1 and R2 were set as 2. Separation time was minimized in the range from 15 to 50 min. Only one solution was given: 24 °C, 95 mM, and pH 9.4 of BGE with the desirability value of 0.437. The R1, R2, and separation time under this condition were 1.029, 1.508, and 34.72 min, respectively. We performed the separation according to this condition and found R1 = 0.906 and R2 = 0.998, and the whole process took 40 min (Fig. 3), which was in accordance with the result given by desirability function. This condition was then used in the following experiments.

Graph: Fig. 3Electropherograms of nicotinamide adenine nucleotides and adenosine phosphates using the optimized conditions: 24 °C, 95 mM, and pH 9.4 of BGE. Identification: (1) NAD + , (2) NADP + , (3) NADH, (4) AMP, (5) ATP, (6) ADP, (7) NADPH

The optimal simultaneous separation and quantification of the seven metabolites were achieved by using 95 mM sodium tetraborate buffer at pH 9.4 under 24 °C. Linearity was checked by double dilution of standards to obtain six concentration points for the mixture standards (ESM Fig. S3). As shown in Table 3, the regression coefficients for calibration curves of each analyte were all greater than 0.996. The average retention time of analyte and its relative standard deviation (RSD) reflected good reproducibility (< 1%, except NADPH), along with the RSD of peak area (1.483–5.500%, 15.006% for NADPH). The LODs and LOQs of the analytes ranged from 0.050 to 0.688 mg/L and from 0.167 to 2.292 mg/L, respectively. The LODs of AMP, ADP, and ATP were much lower than those in a previous report [[13]] due to the application of preconcentration technique. The recoveries of metabolites were between 97.5 and 101.2%, indicating that this method was efficient in separation and for further analysis of these metabolites.

Summary of linearity, reproducibility, and sensitivity of the method (n = 6)

aStandards were spiked into sample solution before CE separation

Figure 4 shows the separation of the target metabolites in CHO cell extracts. By spiking the standards in the extracted samples and UV comparison, all seven metabolites were identified in spite of the complexity of the intracellular components. Retention times of the seven metabolites were in accordance with those of the standard mixture (Fig. 3), indicating the reliability of this method when analyzing real cell samples.

Graph: Fig. 4Electropherograms of CHO cell extracts using the optimized conditions: 24 °C, 95 mM, and pH 9.4 of BGE. Identification: (1) NAD + , (2) NADP + , (3) NADH, (4) AMP, (5) ATP, (6) ADP, (7) NADPH. All seven metabolites were identified by standard spiking

We then chose two specific culture processes to study the metabolic characteristics of the CHO cells. mAb producing phase was deemed to be quite different from its prior rapid growth phase in metabolic states, which might be reflected in AEC, CRC, and ARC by measuring the seven metabolites. Moreover, redox state was getting more and more attention due to its impact on cell growth, mAb production, and even product quality. Therefore, in another culture, quantification of the redox state was expected by disturbing the redox balance with oxidant agent.

As shown in Fig. 5a, AMP, ADP, ATP, and AEC were not affected by the culture phases changing from rapid growth phase (D3) to mAb producing phase (D8). AEC is an indicator of the degree of phosphorylation of the ATP-ADP-AMP system and it is commonly used to compare the energy states of the cells under different conditions. Cells under starvation could also maintain AEC at a value of approximately 0.9, while ATP content declined from 3 to 1 nmol/106 cells [[25]]. Therefore, AEC is a stable parameter and its decline often accompanied by cellular metabolic turnovers such as the drop of cell viability [[11]]. In this case, cell viability on D3 and D8 was 97.7% and 97.6%, respectively, while ATP level was not decreased, indicating a healthy metabolic status for the cells to produce mAb.

Graph: Fig. 5Detection of nicotinamide adenine nucleotides and adenosine phosphates in CHO cell line A in different culture phases. a Concentrations of AMP, ADP, and ATP and AEC value. b Concentrations of NAD + and NADH and the ratio of NAD + /NADH. c Concentrations of NADP + and NADPH and the ratio of NADP + /NADPH. d CRC and ARC value. * p < 0.05, ** p < 0.01, *** p < 0.001

The pairs of NADH/NAD+ and NADPH/NADP+ were significantly affected in different culture phases (Fig. 5b and c). In the production phase, the intracellular content of NADH and NADPH was significantly elevated, while that of NADP+ decreased. The ratios of NADH/NAD+ and NADPH/NADP+ were increased in mAb producing phase. The significant metabolic characteristic of this cell line is the shift of lactate production (D0~D4) to consumption (D5~D12), which often indicates a desired process for high production of mAbs. Therefore, the conversion of lactate to pyruvate was associated with the elevated NADH level as a cofactor. Moreover, it was reported that the flux of PPP was increased in the production phase [[26]–[28]], in accordance with the increased level of NADPH (Fig. 5c) which might be mainly involved in biomass synthesis for mAbs and redox balance with increased oxidative stress. Besides, the decreased flux of fatty acid from citrate in production phase [[7], [29]] would also lead to the increased level of NADPH due to the highly NADPH-consumed process of fatty acid synthesis. Finally, ARC was increased from 0.289 ± 0.035 to 0.551 ± 0.008 in production phase, indicating a highly reactive process for protein synthesis, while the increase of CRC demonstrated the shift of lactate metabolism to enter TCA cycle for more efficient catabolism (Fig. 5d).

Oxidative glutathione (GSSG), as an oxidizing agent, inevitably influences the redox states of the CHO cells, and the effect of GSSG on the intracellular energy metabolites was also tested. With the addition of GSSG, the levels of AMP, ADP, and ATP were not affected, and the AEC was maintained above 0.8 (Fig. 6a). As demonstrated in the above case, AEC was relatively constant to provide driving force for energy-demanding reactions. The levels of NADH and NAD+ were not affected by GSSG addition (Fig. 6b); however, the content of NADP+ was significantly elevated (Fig. 6c). The supplementation of GSSG might consume a lot of NADPH to convert to GSH as a major antioxidant molecular [[30]], thus disturbing the balance of NADPH/NADP+ pair. Moreover, the decline of ARC, together with the unchanged CRC (Fig. 6d), might suggest the intracellular responses of the cells to compensate redox imbalance, with the minimal impact on cell growth, cell viability, and mAb production (data not shown).

Graph: Fig. 6Detection of nicotinamide adenine nucleotides and adenosine phosphates in CHO cell line B with GSSG addition. a Concentrations of AMP, ADP, and ATP and AEC value. b Concentrations of NAD + and NADH and the ratio of NAD + /NADH. c Concentrations of NADP + and NADPH and the ratio of NADP + /NADPH. d CRC and ARC value. ** p < 0.01, *** p < 0.001

In this work, a CE method to simultaneously detect four nicotinamide adenine nucleotides (NADH, NAD+, NADPH, and NADP+) and three adenosine phosphates (AMP, ADP, and ATP) in CHO cells was established. An on-line focusing technique was applied to greatly improve the separation, and the whole process was optimized by using design-of-experiment methodology. Increasing the pH and/or ionic strength of BGE could improve peak resolution but lengthening the separation process and elevating the temperature could shorten the time. Therefore, the optimized conditions were given by desirability function as follows: 24 °C, 95 mM, and pH 9.4 of BGE. The method was validated to ensure sensitivity, reproducibility, and resolution and applied to two CHO cell culture processes. Taking the advantages of small sample size, sensitivity to low intracellular amount, convenience for preparation and conduction, and low cost of regents and materials, this method is promising in making full use of CE apparatus to get insight into CHO cells in process development of biopharmaceuticals.

The authors declare that they have no conflict of interest.

Graph: (PDF 358 kb)

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.