Impact of iron raw materials and their impurities on CHO metabolism and recombinant protein product quality

Abstract Cell culture medium (CCM) composition affects cell growth and critical quality attributes (CQAs) of monoclonal antibodies (mAbs) and recombinant proteins. One essential compound needed within the medium is iron because of its central role in many cellular processes. However, iron is also participating in Fenton chemistry leading to the formation of reactive oxygen species (ROS) causing cellular damage. Therefore, this study sought to investigate the impact of iron in CCM on Chinese hamster ovary (CHO) cell line performance, and CQAs of different recombinant proteins. Addition of either ferric ammonium citrate (FAC) or ferric citrate (FC) into CCM revealed major differences within cell line performance and glycosylation pattern, whereby ammonium was not involved in the observed differences. Inductively coupled plasma mass spectrometry (ICP‐MS) analysis identified varying levels of impurities present within these iron sources, and manganese impurity rather than iron was proven to be the root cause for increased cell growth, titer, and prolonged viability, as well as altered glycosylation levels. Contrary effects on cell performance and protein glycosylation were observed for manganese and iron. The use of low impurity iron raw material is therefore crucial to control the effect of iron and manganese independently and to support and guarantee consistent and reproducible cell culture processes.

CCM are amino acids, carbohydrates, vitamins, lipids, inorganic salts, and trace elements. [4][5][6] Each of these components can have a tremendous effect on cellular performance and critical quality attributes (CQAs) of the final protein and a good understanding of each of these component's role in cellular metabolism and protein expression is required. 6 For instance, amino acids have been shown to increase final titer and modulate glycosylation profile of the produced protein when optimized in CCM, [7][8][9] and several trace elements have been reported to impact CQAs of the final product, such as glycosylation, charge, and aggregation. 2 Iron is an essential element for cellular processes. Many enzymes involved in energy metabolism, deoxyribonucleic acid (DNA) biosynthesis, and repair, or antioxidant functions use iron as a cofactor. 10 Iron uptake in cell culture is regulated by two main mechanisms, the transferrin-bound iron (TBI) uptake and the non-transferrin-bound iron (NTBI) uptake, depending on the availability of transferrin. In the presence of transferrin, two atoms of ferric iron are bound by it to form an iron-transferrin complex. This complex is then recognized by transferrin receptor 1 and a receptor-mediated endocytosis of TBI follows. Within the endosome, ferric iron is released from transferrin due to acidification and the iron is subsequently reduced to ferrous iron by an endosomal ferrireductase. In the NTBI uptake, several surface ferrireductases are involved, reducing NTBI to its ferrous form before iron is imported to the cell by transporters such as the divalent metal-ion transporter 1 (DMT1). 11,12 The capability of iron to take part in redox reactions makes it a crucial transition metal for cellular functions. However, iron can catalyze Fenton reactions resulting in the formation of reactive oxygen species (ROS) that may lead to DNA, protein, or membrane damage. 11,13 The formation of ROS due to the presence of iron, which is commonly added in form of iron complexes such as ferric ammonium citrate (FAC), ferric citrate (FC), or ferric ethylenediaminetetraacetic acid (EDTA), can increase the degradation rate of media components in CCM. [14][15][16] Altering the iron concentration within CCM can have different effects on cell performance and final CQAs of the product. Increasing iron concentrations were shown to improve overall cell growth and productivity of a CHO cell culture. 15,17 However, an increase in CCM iron concentration was also demonstrated to correlate with increased color formation of the recombinant proteins produced in CHO cells, which also correlated with an increased level of acidic charge variants. Both observations are likely to be related to oxidation effects within the recombinant protein caused by iron-generated ROS. 14,18 ROS production due to iron was also identified to cause oxidative stress enhancing protein aggregation, for example in neuronal or lens crystallin proteins. 19,20 In other studies, the effect of iron on glycosylation macroheterogeneity was investigated.
The addition of iron to CCM was reported to increase site-occupancy of the glycoprotein tissue plasminogen activator. 21 Moreover, a constant glycosylation pattern was detected for interferon gamma (IFN-γ) upon iron citrate addition to the medium compared to the non-supplemented version, in which increased levels of non-glycosylated IFN-γ were observed. 22 Furthermore, the addition of iron to CCM was shown to significantly increase galactosylation of the recombinant glycoprotein, which has been disclosed in US Patent No. 9598667B2. 23 In this study, the impact of iron in CCM on cell performance and product quality (aggregation and glycosylation profile) was investigated.
Fed-batch data for a CHO K1 cell line revealed major differences in cell performance and glycosylation profile of the recombinant monoclonal antibody (mAb) upon usage of different iron sources. Inductively coupled plasma mass spectrometry (ICP-MS) characterization of the sources was performed to determine elemental impurity levels within the raw material. Among all impurities, manganese was identified as the root cause for improved cell performance and altered glycosylation level of the recombinantly produced protein. The effect of manganesecontaminated iron source was further studied for another CHO K1 cell line as well as for a CHOZN® clone. Altogether, the results demonstrate the importance and need of low impurity iron sources in CCM to ensure consistent and reproducible cell culture processes.

| Cell culture experiments
Fed-batch experiments were performed in 50 ml spin tubes (TPP, Trasadingen, Switzerland) with vented cap at 37 C, 5% CO 2 , 80% humidity and with an agitation speed of 320 rpm. Iron deficient Cel-lvento® 4CHO medium was used, whereas different amounts of iron in the form of either FAC or FC were added. Seeding density was 2 × 10 5 cells/ml in a working volume of 30 ml. Cellvento® 4Feed was added on day 3, 5, 7, 10, 12, and 14 with a feeding strategy of 1.5, 3, 3, 3, 3, and 3% (v/v), respectively (cell line 1 and 3). For cell line 2, 3% (v/ v) were added on day 3, 5, 10, 12, and 14 and 6% (v/v) on day 7. Glucose (400 g/L) was fed on demand to up to 6 g/L during the week and up to 13 g/L before the weekend. Viable cell density (VCD) and viability were measured with the Vi-CELL™ XR 2.04 (Beckman Coulter, Fullerton, CA). Glucose, titer, iron and ammonium concentrations were analyzed with the Cedex Bio HT (Roche, Mannheim, Germany) after centrifugation of the sample for 5 min at 4500 rpm (2287 g).

| Antibody purification and CQAs analysis
Antibodies and fusion proteins were purified from the cell culture supernatant by using protein A PhyTips® (PhyNexus Inc., San Jose, CA). Aggregation profile was analyzed using size exclusion chromatography coupled to an UV detector (SEC-UV) using an Acquity ultraperformance liquid chromatography (UPLC) (Waters, Milford, MA) and a TSK gel SuperSW series column (Tosoh Bioscience, Griesheim, Germany) at room temperature. 10 μl of sample, adjusted to 1 mg/ml with storage buffer (85% (v/v) of 30 mM citric acid pH 3.0 and 15% (v/v) of 0.375 M Tris Base pH 9.0), were applied to the system at a flow rate of 0.35 ml/min. 0.05 M sodium phosphate, 0.4 M sodium perchlorate solution adjusted to pH 6.3 was used as mobile phase. Glycosylation analysis was performed either by capillary gel electrophoresis with laser-induced fluorescence (CGE-LIF) for the two antibodies or by UPLC coupled to a mass spectrometer (UPLC-MS) for the fusion protein as described elsewhere. 24

| Iron source characterization
The detection and quantification of trace elements within iron sources was performed using ICP-MS. Sample preparation was carried out by microwave-assisted digestion involving nitric acid. Indium was used as internal standard throughout all analyses. Either a semiquantitative elemental screening method was performed involving the adaption of the system response curve as intensity per ng/ml of analyte by a single point calibration using an ELAN 6000 (PerkinElmer Inc., Waltham, MA), or a quantification by external calibration using an ELEMENT 2™ high resolution (HR)-ICP-MS (ThermoFisher, Waltham, MA) was done.
In both cases, the recovery rate was determined by spiking the samples with a known amount of analyte. 25,26

| Statistical analysis
Data are expressed as means ± standard deviation (SD) of six biological replicates unless stated otherwise. Graphic analysis was performed with GraphPad Prism 8 Software (GraphPad Software Inc., San Diego, CA).
Statistical analysis was performed by the non-parametric Kruskal-Wallis test for multiple-group comparison with subsequent Dunn's test.
p-values smaller than 0.01 or 0.001 were considered significant.

| Effect of increasing iron amounts in CCM on cell performance and IgG quality attributes
To understand whether increasing amounts of iron in CCM impact cell performance and the quality of the produced recombinant antibody, a small-scale iron dose response fed-batch experiment was performed in spin tubes with cell line 1. Therefore, increasing amounts of FAC, a commonly used iron source in CCM, were spiked to iron deficient Cel-lvento® 4CHO to obtain iron concentrations of 2, 10, 50, and 100 mg/L. As shown in Figure 1, the cell performance was significantly impacted by the addition of increasing FAC concentrations. For instance, VCD on day 10 was higher for 10, 50, and 100 mg Fe/L compared to 2 mg Fe/L (absolute increase of 19.2%, 20.0%, and 16.8%, respectively) ( Figure 1a). However, the increase in cell growth led to a faster decrease in viability (Figure 1b). The highest final titer (D17) was detected for 2 mg Fe/L even though the IgG concentration was higher for 50 mg Fe/L until D14 (Figure 1c), suggesting that 50 mg Fe/L may be more suitable for commercial processes where the harvest is commonly performed at viabilities above 80%. The iron concentration decreased significantly during the course of the fedbatch process for 10, 50, and 100 mg Fe/L, and the absolute decrease in iron concentration over time increased with increasing iron concentration ( Figure 1d). Since FAC was used as iron source, higher ammonium concentrations were detected at day 0 for higher applied iron concentrations in the medium. However, the produced metabolic ammonium amount for all tested conditions was similar for the following days (Figure 1e), indicating that the starting concentration is not likely to impact cell metabolism. Iron deficient medium led to no cell growth and no titer production highlighting the necessity of iron in CCM to maintain essential cellular functions (data not shown).
To investigate the effect of increasing FAC concentrations on mAb1 CQAs, aggregation and glycosylation profiles were determined by SEC-UV and CGE-LIF, respectively. Therefore, samples of day 10 of the fedbatch process were analyzed for which cell culture viability was compara-  detected upon ammonium addition to CCM. Absolute differences for HMW and main peak species were below 0.2% and for glycosylation species below 0.9% for the tested conditions. Overall, these data suggest that ammonium present in FAC is neither responsible for increased cell performance, nor does it lead to elevated levels of HMW or terminal galactosylation species.

| Impact of different iron sources, FAC and FC, on cell performance and IgG quality attributes
Since increasing amounts of ammonium did not account for the observed differences in cell performance and CQAs upon increasing FAC concentrations, a second iron source, namely FC, was investigated. FC was selected due to the similar chelation strength and the absence of ammonium. Chosen iron concentrations were 2, 10, 50, and 100 mg/L and were added to iron deficient Cellvento® 4CHO medium. As shown in Figure 4, usage of FC as iron source in CCM caused a lower cell growth with a reduced maximal VCD of 7.4%, 28.9%, and 3.5% for 10, 50, and 100 mg Fe/L, respectively, compared to the corresponding FAC condition (Figure 4a). For both iron sources, a faster decrease in viability was detected with increasing iron concentrations, whereas an even faster decline in viability was observed To investigate whether FC has a different impact on mAb1 CQAs than FAC, glycosylation and aggregation profiles on day 10 of the fedbatch process were analyzed. These samples were used to directly compare them to the previously analyzed samples upon FAC usage, although viability was already significantly different for FC on day 10. Table 1 summarizes the mAb1 glycosylation pattern of the two main species found upon usage of either FAC or FC in CCM. Whereas increasing levels of FAC in the CCM led to increased levels of terminal galactosylation with 13.3% for 2 mg Fe/L and up to 25.6% for 100 mg Fe/L, FC caused a decrease in terminal galactosylation species from 11.5% for 2 mg Fe/L to 8.0% for 50 mg Fe/L, which was inversely correlated with terminal GlcNAc levels. These data suggest again that the observed differences in glycosylation between FAC and FC are independent of the iron concentration and impacted by unidentified attributes between both iron sources.
Aggregation profile upon FC usage showed the same small dosedependent increase in HMWs for increasing iron amounts like FAC (data not shown).
Altogether, the results obtained with increasing amounts of FAC and FC indicate that attributes other than iron and ammonium levels impact growth, viability, and IgG glycosylation.

| Analysis of trace element impurities in FAC and FC
Since it is known from literature that iron sources can be contaminated with trace elements, 27 the next step focused on trace element characterization of both iron sources because trace elements in general are known to influence cell performance and product quality. [27][28][29][30] Trace element impurity levels of FAC as well as FC were determined by ICP-MS analyses. Table 2  higher levels were measured in FAC compared to FC. Since manganese is well described as a modulator of antibody galactosylation 30 and the manganese amount present as impurity in the FAC iron source contributed to more than 94% of the total manganese concentration in the CCM formulation, this element was considered as highly relevant. Other impurities such as copper and zinc, that are also known to impact cell culture performance and CQAs, 28,29 contributed to less than 15% of the final medium concentration and were thus deemed less impactful and were not studied further. Based on these assessments, the increased manganese impurity in FAC raw material compared to FC iron source was hypothesized to be the major trigger for increased galactosylation and prolonged viability upon usage of FAC.

| DISCUSSION
Iron is an essential trace element needed in CCM to maintain and regulate cellular functions. It plays a key role as a cofactor in many enzymatic processes due to its capability to take part in redox reactions.
However, the redox cycling properties of iron may lead to the formation of ROS, which induce oxidative stress to the cells. 10,13 In this study, we therefore evaluated the impact of iron in CCM on cell performance and CQAs. FAC iron source dose response in CCM for cell line 1 in a fed-batch process indicated significant changes in cell growth, viability, and glycosylation profile of mAb1. Those differences were proven to be independent of ammonium introduced by different amounts of FAC. Furthermore, comparison of FAC and FC iron source also demonstrated significant differences in cell performance and mAb1 glycosylation profile indicating that other factors than iron con- Comparing the effects of manganese and iron on cell performance, the data suggest a possible opposite impact of both metals on cellular oxidative stress regulation affecting cell performance.
Although both, manganese and iron, are metals involved in redox chemistry, both metals have different potentials to cause oxidative stress. Whereas iron is known to generate very oxidizing species through the Fenton reaction, 39 manganese has a higher reduction potential than iron and is thus less reactive. 33 Figure S1).
Whereas manganese was demonstrated to increase terminal galactosylation and terminal sialylation of mAb1/mAb2 and fusion protein, respectively, which is a well characterized effect in literature, 7,24,30,45 iron revealed an opposite effect on glycosylation.
Usage of increasing amounts of low impurity iron source FC Synt led to a slight decrease of terminal galactosylation species for mAb1 and mAb2. An effect of iron on glycosylation was already described in literature, where the spiking of iron and copper salt solution to a bioreactor process decreased IgG3:κ terminal galactosylation level. 46 The underlying mechanism has not been established but in general, metals are known to impact enzymes in the glycosylation pathway. For instance, ß-1,4-galactosyltransferase was shown to have two metal binding sites with high affinity for manganese 47 and addition of metal ions such as zinc or cobalt resulted in the formation of an enzymemetal-complex that reduced ß-1,4-galactosyltransferase activity. 48 Thus, in this study, it is possible that iron also caused the formation of an inhibitory enzyme-metal-complex leading to a decrease in terminal galactosylation species.
Contrary to the observed effect of iron on mAb1 and mAb2 terminal galactosylation, usage of increasing amounts of low impurity FC Synt had a different impact on the glycosylation profile of the fusion protein presenting significant levels of sialylation. Upon increasing FC Synt concentrations, a significant decrease in terminal sialylated species and increase in terminal galactosylation level was detected. One hypothesis is that the sialylation level decreased with higher FC Synt concentration due to a loss of cell viability. Such a correlation was already reported as a consequence of released sialidase enzymes into the extracellular space due to cell lysis. 49 In another study, the sialylation level of a CHO Fc-fusion protein was decreased with increased oxidative stress and higher glycolytic metabolism, which also correlated with decreased cell culture viability. 50 Another hypothesis for the decreased terminal sialylation level is that iron might inhibit sialyltransferases, similarly to the effect proposed for galactosyltransferase. Since the inhibitory effect was predominantly observed for sialic acid, the inhibitory potential of iron seems to be higher for sialyltransferase compared to galactosyltransferase. However, the effect of iron on galactosyltransferase and sialyltransferase activity, and its binding affinity to both enzymes needs to be further investigated. Altogether, the glycosylation results presented within this study indicate that manganese and iron caused different effects on galactosylation and sialylation. Thus, both elements need to be decoupled from each other to be evaluated and controlled independently.
For decoupling both elements, the usage of low impurity iron sources within CCM is needed and requires a careful understanding and management of the used iron sources. Several root causes for raw material impurities have been described. For instance, manufacturing processes for iron salts may use undefined starting materials coming from other industries and thus do not fit for a biopharmaceutical application. Utilization of impure solvents as well as the usage of contaminated equipment or leaching packaging during the manufacturing process might as well lead to raw material impurities and lot-to-lot inconsistencies. 51,52 Since raw material variability is very well known to impact CCM performance, 27,46,53 appropriate control systems for synthesized raw materials are essential to reduce the variability due to trace metals. 54 Besides control steps within the engineering process, quality test controls such as ICP-MS methods may help to characterize purchased raw materials, allowing a better risk assessment of their putative impact on the cell culture process. 46,52,54 However, designing and developing new iron sources with low levels of impurities to decrease the risk of variability might be even more effective and must be considered.

| CONCLUSION
This study describes the impact of iron raw material and its impurities on cell culture performance and recombinant protein glycosylation.
While manganese was identified as the impurity in iron sources contributing to an overall improved cell performance, increasing levels of iron caused a decrease in cell growth, viability, and titer. Likewise, a contrary effect of iron and manganese was identified for the glycosylation profile of the tested recombinant proteins. Thus, this study highlights the need for low impurity iron sources in order to decouple the effects of iron and its impurities within cell culture experiments. If this is done appropriately, each element concentration can be controlled and adjusted individually, allowing the development of a consistent and stable cell culture process leading to reproducible quality attributes of the recombinant product. This may be particularly relevant within the field of biosimilars, whereby the drug substance release requires to match the product quality attributes within a very narrow range.
source characterization with ICP-MS and Thomas Kilian and Frank