Accelerating Monoclonal Antibody Quality Control: The Role of LC–MS in Upstream Bioprocessing

Monoclonal antibodies (mAbs) play a major role in the modern biopharmaceutical industry. Ensuring sufficient product safety as well as process efficiency demands advanced analytical methods. Liquid chromatography coupled with mass spectrometry (LC–MS) offers multi-attribute methods for quality control, surpassing the limitations of conventional techniques. Yet the implementation of LC–MS analysis in the upstream process to validate critical quality attributes (CQAs) during production has rarely been demonstrated. The comparison between crude and purified mAb samples has revealed the capability of the established LC–MS method to effectively detect and analyze mAbs with minimal sample preparation. Purification of the sample allowed for a more detailed investigation of the mAb by reducing impurities and lowering the signal-to-noise ratio, but was not necessary to quickly assess important quality attributes. The analysis allowed the monitoring of glycoform distributions during the production process. Failure modes concerning varied carbon source feeding regimes were detected within 1–2 days. This study highlights the promising potential of LC–MS as a powerful tool for mAb quality control within the context of upstream processing.

Antibody (Ab), also known as an immunoglobulin (Ig). 3d vector © sakurra - stock.adobe.com

Novel monoclonal antibody (mAb)-based drugs enable targeted therapies with applications including a wide range of diseases, such as cancer, autoimmune disorders, and viral infections (1,2). As a result of the remarkable success and increasing demand for mAb-based therapeutics, stringent and rapid quality control throughout the development and manufacturing process has become inevitable to ensure safe and effective application to patients. Regularly monitored critical quality attributes (CQAs) involve the structural integrity, purity, and potential bioactivity (3,4). Conventional quality control methods for recombinantly produced therapeutics were based on a combination of analytical techniques, including capillary isoelectric focusing (cIEF), sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), high performance liquid chromatography (HPLC), and enzyme-linked immunosorbent assays (ELISA) (5–7). While these approaches showed a measure of efficacy, they can lack the required specificity, sensitivity, and capacity to unravel nuanced structural information of the investigated mAb (8).

In recent years, the application of liquid chromatography–mass spectrometry (LC–MS) has emerged as a promising analytical tool to overcome the inherent limitations of conventional methodologies (9,10). LC–MS offers a comprehensive assessment of mAb quality attributes by merging the separation capability of liquid chromatography with the analytical precision afforded by mass spectrometry. The utility of LC–MS builds upon qualitative analysis, extending it to quantitative evaluation of mAb attributes and the determination of minute compositional alterations (11,12). Detection of subtle changes is central to assessing mAb uniformity, long-term stability, and cross-batch comparability (13). This analysis can include the detailed measurement and identification of attributes such as the intact mass, distribution of glycoforms, charge variants, and impurities, which encompass product- and process-related variants, host cell proteins (HCPs), and other contaminants (6,14). The LC–MS analysis therefore builds up a toolbox for fast-paced quality analysis over the entire production chain of mAbs, including the analysis of crude samples, which requires minimal volumes, as well as minor sample preparation (15).

Material and Methods

Chemicals and Materials

Buffer chemicals and eluents were purchased from Carl Roth GmbH + Co. KG if not stated otherwise. Mass standards, NIST mAb, and MS cleaning solutions were provided by Waters Corporation. The deionized H₂O was prepared with Arium pro VF system (Sartorius Stedim Biotech).

Cultivation Process

The analyzed IgG1 mAb was produced using an established CHO DG44 cell cultivation process, described by Boehl et al. (16).Additional glucose feeds were implemented from day five to adjust the glucose culture concentration for standard conditions to 5 g/L. For the mannose substituted condition, the glucose feed was replaced with 400 g/L mannose solution. The glucose deficiency conditions were obtained by omitting the additional feed, leading to complete glucose consumption by the end of day seven. All conditions were conducted in triplicates. Metabolite concentrations and mAb titers were determined using the Cedex Bio (Roche).

Purification of mAbs

mAbs were purified using an Äkta pure 250 system (Cytiva Life Sciences) utilizing affinity chromatography. The mAb was diluted to 1 mg/mL in the equilibration buffer (20 mM NaHPO₄, 150 mM NaCl, pH 7.5) and filtered using a syringe filter (pore size 0.45 μm, Polyvinylidenfluorid [PVDF] membrane [Avantor Inc.]) and single-use syringes (10 mL, B. Braun SE). All steps of the subsequent separation methodology were executed with a flow rate of 1 mL/min and a preparative column (Cytiva Life Sciences) (HiTrap Protein A HP, Column volume [CV] = 1 mL).

Sample Preparation for LC–MS

All samples were diluted with deionized H₂O to meet the recommended concentration range of the 2.1 × 100 mm, 2.7-μm 450 Å BioResolve RP mAb column (Waters Corporation) of 2.0 to 20 μg mAb. Applied dilution factors are summarized in Table S1 in the supplementary information. All diluted samples were filtered with a 0.2 μm syringe filter (PVDF, 4 mm diameter, Thermo Fisher Scientific Inc.) and single use syringes (1 mL Omifix-F, B. Braun SE).

Analytical Methods

The LC–MS measurements were conducted using a BioAccord LC–MS system (Waters Corporation). Reversed-phase liquid chromatography (RPLC) was performed to separate the mAb from other sample components such as fragments or HCPs, with a gradient approach for 8 min. This method included a column washing step to avoid overlay and transfer effects caused by complex samples or previous measurements. Separated peaks were passed on to the MS between time points 1.8–3.0 min. Each sample was measured in duplicate, and each sample set was completed with blank measurements, with ddH₂O and quality controls with NIST reference mAb (4 μL injection). Eluents are listed in Table S2, applied parameters in Table S3, and details of the gradient in Table S4 in the supplementary section.

Measurements were planned and results obtained via IntactMass software (Waters). All data were processed further using the analyzing software UNIFI (Waters) to receive the calculated mass data and mass errors, chromatograms, spectra, and assigned peak identifications. Deconvolution of the raw m/z data was executed automatically with a MaxEnt-algorithm (17,18).

Statement of Human and Animal Rights

This article does not contain any studies with human or animal subjects.

Results

Figure 1: Comparison of observed mass spectra of crude sample (blue, top) and protein-A purified sample (green, bottom) for the mAb monomer peak of the LC.

LC–MS method protocols require mostly purified mAb to ensure definite results and in-depth protein analysis (19–22). However, purification is an additional step, which can be time- and material-intensive, as well as lead to a loss of sample and process information by removal of mAb fragments or other impurities. The direct measurement of crude samples is therefore advantageous for implementing LC–MS analysis as process quality control, but it has to be tested regarding its suitability. Therefore, the harvested supernatant was partly purified and both the supernatant and the obtained purified mAb were analyzed via LC–MS. The generated observed mass spectra of the mAb monomer peak detected in the LC system are depicted in Figure 1.

The LC–MS results confirmed that the produced mAb can be detected and analyzed as a crude sample with minimal sample preparation according to the theoretical mass. While the peak resolution was improved and impurities, possibly related to polyethylene glycol variants (below 2000 m/z), were removed after purification, all main charge states were identified in both samples. The measured spectra were further analyzed to determine the glycosylation pattern and the reliability of the method.

Figure 2: Comparison of glycoform distribution for supernatant and purified sample. Assignment of the glycoforms on the basis of the brightness gradations.

Both approaches can be evaluated regarding their applicability for process control and identification of CQAs. The detected relative glycoform distribution is shown in Figure 2 for comparison.

The glycosylation pattern of human IgG is biantennary as one oligosaccharide chain can bind to the asparagine in the C_H2 domain of the heavy chain. One fucose monosaccharide can bind to the inner N-acetylglucosamine (GlcNAc). In addition, one galactose molecule can bind to each of the external GlcNAcs. It is possible that no, one, or two galactose molecules are bound. These possibilities to bind further saccharides to the core structure result in the following nomenclature: GXF, where G describes the binding of galactose and the X indicates the number of galactose molecules. The F stands for the binding of fucose, which was detected in all glycoforms (23,24).

The glycoforms were identified with small mass errors less than 10 ppm, confirming precise and reliable classifications. While the general distributions of the glycoforms were similar for both samples, with G0F/G1F as dominant form, the most significant difference was observed for the standard deviation. It was very low and consistent for the purified sample, whereas the standard deviation of the crude sample exceeded for each glycoform. Particularly noticeable was the standard deviation of the G2F/G2F form of the crude sample. It was the overall highest with ± 3.1 and can be traced back to an irregular detection of this form.

The reason behind this was the slightly increased measurement noise due to the complexity of these samples, which can cause an overlay of minor peaks in the spectra. This does not apply to the purified samples as media components, host cell proteins, and other residual molecules are already separated by the preparative chromatography process.

Figure 3: Cultivation results for the varied feeding scenarios with (a) viable cell densities and (b) cell viabilities.

To demonstrate the suitability of LC–MS analysis as a process control tool, the standard process was monitored. To determine the ability to detect irregularities, two different process failure scenarios were conducted: either the glucose feed was replaced by a mannose solution, or it was omitted completely. Both approaches should induce a nutrient substitution/deficiency regarding the available carbon source and consequently were suspected of producing process failures with potentially different glycoform distributions. The cultivation results for the viable cell density (VCD) and the cell viability over the process duration are plotted in Figure 3.

The cultivations with varied nutrient conditions showed similar cell growth and viabilities until day 7. All conditions reached peak VCDs around 23 × 10⁶ cells/mL. The mannose-substituted cultivations further matched the standard references, with VCDs around 15 × 10⁶ cells/mL and viabilities over 95%. The glucose deficiency cultures showed a significant drop in VCDs as well as viabilities after depletion of glucose around day 7, with VCDs around 5 × 10 ⁶ cells/mL and viabilities under 75%.

Figure 4: Changes in glycosylation pattern after process failure simulation. Assignment of the glycoforms on the basis of the brightness gradations.

To equalize the cultivation effects of the manipulations on the resulting mAb titer, the observed MS signal intensities were normalized to the mAb amount injected into the system. The resulting glycoform distributions are visualized in Figure 4.

Despite the normalization, the glucose deficiency sample showed lower signal intensity, which can be attributed to higher noise as a result of process-related impurities. For this measurement, the reference and the mannose substitution sample resulted in a similar glycosylation pattern, both for the glycoform distribution and the MS signal intensities. G0F/G0F and G1F/G1F occurred nearly to the same extent, and G0F/G1F was the most prominent form. G1F/G2F showed half the signal intensity of G0F/G1F, whereas G2F/G2F was a minor form. It can therefore be assumed that mannose could be efficiently utilized by the CHO cells for cell proliferation and maintenance, while still ensuring mAb production with correct glycosylation patterns.

The glucose deficiency sample showed differences for both factors. G1F/G1F rather than G0F/G1F was the main glycoform, while G2F/G2F was not detectable at all—likely a result of concentrations below the limit of detection. The minor form was G0F/G0F. Moreover, the glucose deficiency sample showed lower signal intensities compared to the other samples, despite the applied normalization. This can be attributed to higher signal to noise due to the process-related impurities. Investigation of the mass errors for the MS measurements further elucidated the challenges with analyzing samples associated with extreme failure modes. The mass errors are summarized in Table S5, with red cells marking errors above the internal threshold for accurate data.

Figure 5: Glycoform distributions of mAb produced with standard and glucose deficiency conditions over the cultivation duration (Day 7–Day 12). Assignment of the glycoforms on the basis of the brightness gradations.

To assess the feasibility of detecting process failures during production, samples from intermediate cultivation days were analyzed for their glycosylation patterns. The resulting distributions for samples from day 7, day 9, and day 12 are visualized in Figure 5.

Since process variations related to glucose deficiency became apparent around day 7, any potential shifts in glycoform distribution compared to the standard reference were expected only after this point. While the glycoform patterns of the investigated process conditions still closely matched at day 7, the first deviations were already detectable by day 9. G1F/G1F became the most prominent form in the glucose deficiency sample and the ratio of G1F/G2F grew around 10%. This trend intensified up to the final distribution at day 12. These data indicate that failure modes related to metabolite availability can be rapidly detected during production using the established LC–MS analysis. Early determination of process deviations can be used to enable advanced feedback strategies for process control or support decisions regarding premature termination of the production process to save resources, time, and money.

Conclusion and Outlook

This study demonstrates the role of LC–MS as a promising tool to overcome the limitations of conventional analytical methods. LC–MS offers a comprehensive approach to qualitative mAb analysis, allowing the investigation of multiple critical quality attributes. This comparison of crude and purified samples confirmed the capability to detect and analyze mAbs with minimal sample preparation, although further downstream processing enhanced peak resolution and reduced impurities of the sample. Nearly the same glycoform distributions could be identified in both cases, with lower standard deviations for the purified sample. The decision to purify samples before LC–MS therefore depends on the specific aim of the analysis, with a trade-off needed between detail and efficiency. The established multi-attribute method was furthermore effective in monitoring changes in glycosylation patterns in response to varied carbon source feeding regimes.

As demonstrated by these results, LC–MS in the field of mAb quality control is expected to evolve and become an increasingly important tool for research and development as well as for biopharmaceutical manufacturers. The technique highlights the importance of monitoring glycoform distributions and can be used to optimize production processes to ensure consistency in mAb products. Moreover, the ability to quickly detect process failure modes and deviations in product quality will contribute to enhancing process control during the upstream process. Even more detailed process knowledge can be gained by the analysis of small molecules for control of nutrient consumption, which can be used to create feedback loops for advanced feeding regimes. In conclusion, LC–MS represents a significant advancement in the quality control of mAbs, offering a powerful tool to quickly enhance the spectrum of investigated quality attributes.

Acknowledgments

This research was funded by the German Research Foundation (DFG) via the Emmy Noether program (project ID 346772917). The authors would like to further acknowledge the support by Waters Corporation for this study.

Conflict of Interests

The authors have no conflicts of interest to declare. All co-authors have seen and agree with the contents of the manuscript and there is no financial interest to report. We certify that the submission is original work and is not under review at any other publication.

References

(1) Kinch, M. S.; Kraft, Z.; Schwartz, T. Monoclonal Antibodies: Trends in Therapeutic Success and Commercial Focus. Drug Discov. Today 2023, 28, 103415. DOI: 10.1016/j.drudis.2022.103415

(2) Kaplon, H.; Crescioli, S.; Chenoweth, A.; Visweswaraiah, J.; Reichert, J. M. Antibodies to Watch in 2023. mAbs 2023, 15, 2153410. DOI: 10.1080/19420862.2022.2153410

(3) ICH, Q6B Specifications: Test Procedures and Acceptance Criteria for Biotechnological/Biological Products, ICH Harmonised Tripartite Guideline (1999).

(4) Wohlenberg, O. J.; Kortmann, C.; Meyer, K. V.; et al. Optimization of a mAb Production Process with Regard to Robustness and Product Quality Using Quality by Design Principles. Eng. Life Sci. 2022, 22, 484–494. DOI: 10.1002/elsc.202100172

(5) Zhang, E.; Xie, L.; Qin, P.; et al. Quality by Design-Based Assessment for Analytical Similarity of Adalimumab Biosimilar HLX03 to Humira®. AAPS J. 2020, 22, 69. DOI: 10.1208/s12248-020-00454-z

(6) Alt, N.; Zhang, T. Y.; Motchnik, P.; et al. Determination of Critical Quality Attributes for Monoclonal Antibodies Using Quality by Design Principles. Biologicals 2016, 44, 291–305. DOI: 10.1016/j.biologicals.2016.06.005

(7) Flatman, S.; Alam, I.; Gerard, J.; Mussa, N. Process Analytics for Purification of Monoclonal Antibodies. J. Chrom. B 2007, 848, 79–87. DOI: 10.1016/j.jchromb.2006.11.018

(8) Ezan, E.; Bitsch, F. Critical Comparison of MS and Immunoassays for the Bioanalysis of Therapeutic Antibodies. Bioanal. 2009, 1, 1375–1388. DOI: 10.4155/bio.09.121

(9) Chen, G.; Pramanik, B. N. LC-MS for protein Characterization: Current Capabilities and Future Trends. Expert Review of Proteomics 2008, 5, 435–444. DOI: 10.1586/14789450.5.3.435

(10) Camperi, J.; Goyon, A.; Guillarme, D.; Zhang, K.; Stella, C. Multi-dimensional LC-MS: The Next Generation Characterization of Antibody-based Therapeutics by Unified Online Bottom-up, Middle-up and Intact Approaches. Analyst 2021, 146, 747–769. DOI: 10.1039/D0AN01963A

(11) Liu, Y.; Zhang, C.; Chen, J.; et al. A Fully Integrated Online Platform For Real Time Monitoring Of Multiple Product Quality Attributes In Biopharmaceutical Processes For Monoclonal Antibody Therapeutics. J. Pharm. Sci. 2022, 111, 358–367. DOI: 10.1016/j.xphs.2021.09.011

(12) Graf, T.; Heinrich, K.; Grunert, I.; et al. Recent Advances in LC-MS Based Characterization of Protein-based Bio-therapeutics - Mastering Analytical Challenges Posed by the Increasing Format Complexity. J. Pharm. Biomed. Anal. 2020, 186, 113251. DOI: 10.1016/j.jpba.2020.113251

(13) Zheng, K.; Bantog, C.; Bayer, R. The Impact of Glycosylation on Monoclonal Antibody Conformation and Stability. mAbs 2011, 3, 568–576. DOI: 10.4161/mabs.3.6.17922

(14) Zheng, K.; Yarmarkovich, M.; Bantog, C.; Bayer, R.; Patapoff, T. W. Influence of Glycosylation Pattern on the Molecular Properties of Monoclonal Antibodies. mAbs 2014, 6, 649–658. DOI: 10.4161/mabs.28588

(15) Pais, D. A. M.; Carrondo, M. J. T.; Alves, P. M.; Teixeira, A. P. Towards Real-time Monitoring of Therapeutic Protein Quality in Mammalian Cell Processes. Curr. Opin. Biotech. 2014, 30, 161–167. DOI: 10.1016/j.copbio.2014.06.019

(16) Böhl, O. J.; Schellenberg, J.; Bahnemann, J.; et al. Implementation of QbD Strategies in the Inoculum Expansion of a mAb Production Process. Eng. Life Sci. 2021, 21, 196–207. DOI: 10.1002/elsc.202000056

(17) Ferrige, A .G.; Seddon, M. J.; Green, B. N.; et al. Disentangling Electrospray Spectra with Maximum Entropy. Rapid Commun. Mass Spectrom. 1992, 6, 707–711. DOI: 10.1002/rcm.1290061115

(18) Ferrige, A .G.; Seddon, M. J.; Jarvis, S.; Skilling, J.; Aplin, R. Maximum Entropy Deconvolution in Electrospray Mass Spectrometry. Rapid Commun. Mass Spectrom. 1991, 5, 374–377. DOI: 10.1002/rcm.1290050810

(19) Rogers, R. S.; Nightlinger, N. S.; Livingston, B.; et al. Development of a Quantitative Mass Spectrometry Multi-attribute Method for Characterization, Quality Control Testing and Disposition of Biologics. mAbs 2015, 7, 881–890. DOI: 10.1080/19420862.2015.1069454

(20) Dong, J.; Migliore, N.; Mehrman, S. J.; et al. High-Throughput, Automated Protein A Purification Platform with Multiattribute LC-MS Analysis for Advanced Cell Culture Process Monitoring. Anal. Chem. 2016, 88, 8673–8679. DOI: 10.1021/acs.analchem.6b01956

(21) Tharmalingam, T.; Wu, C.-H.; Callahan, S.; Goudar C. T. A Framework for Real-time Glycosylation Monitoring (RT-GM) in Mammalian Cell Culture. Biotechnol. Bioeng. 2015, 112, 1146–1154. DOI: 10.1002/bit.25520

(22) Doherty, M.; Bones, J.; McLoughlin, N.; et al. An Automated Robotic Platform for Rapid Profiling Oligosaccharide Analysis of Monoclonal Antibodies Directly From Cell Culture. Anal. Biochem. 2013, 442, 10–18. DOI: 10.1016/j.ab.2013.07.005

(23) Xu, J.; Shao, Z.; Wang, Z.; et al. Developing a Medium Combination to Attain Similar Glycosylation Profile to Originator by DoE and Cluster Analysis Method. Sci. Rep. 2021, 11, 7103. DOI: 10.1038/s41598-021-86447-0

(24) Batra, J.; Rathore, A. S. Glycosylation of Monoclonal Antibody Products: Current Status and Future Prospects. Biotechnol. Progress 2016, 32, 1091–1102. DOI: 10.1002/btpr.2366

Carlotta Kortmann is a PhD student at the Institute of Technical Chemistry at Leibniz University Hannover, in Hannover, Germany.

Ole Jacob Wohlenberg is a postdoc at the Institute of Technical Chemistry at Leibniz University Hannover.

Charlotte Hauschildt is a PhD student at the Institute of Technical Chemistry at Leibniz University Hannover.

Sascha Beutel is a professor and group leader at the Institute of Technical Chemistry at Leibniz University Hannover.

Dörte Solle is group leader with the Institute of Technical Chemistry at Leibniz University Hannover.

Janina Bahnemann is a professor and group leader with the Institute of Physics, Centre for Advanced Analytics and Predictive Sciences (CAAPS) at the University of Augsburg in Augsburg, Germany.