Automated Workflow Paves Way for High-Throughput Reproducible Operations—Trastuzumab Glycosylation Profiling as a Case Study

Glycosylation is a critical quality attribute of monoclonal antibodies (mAbs), influencing pharmacokinetics and effector function. We compared manual and automated N-glycan profiling workflows for an in-house trastuzumab biosimilar using seven independent replicates. Both workflows consistently identified the canonical IgG1 glycan repertoire, with G0F as the predominant glycoform. Manual profiling showed higher galactosylated species (G1F: 22.49%; G2F: 2.26%) but greater inter-replicate variability (mean coefficient of variation [CV]: 7.95%). Automation produced higher G0F abundance (~63% vs ~56% manual), lower G1F (14.99%) and G2F (1.13%), and improved reproducibility (mean CV: ~5.70%). These distributional differences likely reflect workflow-dependent biases in glycan recovery and quantitation rather than biological variation. Overall, automation improves robustness and reproducibility while maintaining glycan repertoire coverage, supporting its use for scalable, high-throughput glycoanalytics in biopharmaceutical development.

Monoclonal antibodies are one of the most successful classes of biotherapeutics, with applications spanning oncology, autoimmune disorders, and infectious diseases.^1–3 Their expanding use in oncology, including HER2-targeted antibodies and antibody-drug conjugates, further highlights the clinical impact and versatility of antibody-based modalities.^4,5 As these molecules progress from discovery to late-stage development and commercialization, robust analytical characterization becomes increasingly important. N-glycan profiling generally comprises sample preparation, enzymatic glycan release and labeling, followed by chromatographic separation and detection; the manual and automated routes evaluated here are summarized in Figure 1.

Glycosylation is widely recognized as a critical quality attribute because it directly affects antibody effector functions, pharmacokinetics, stability, and immunogenicity.^6,7 The conserved Fc glycosylation site at Asn297 modulates antibody-dependent cellular cytotoxicity, complement-dependent cytotoxicity, and Fc receptor binding^.8–10 Even subtle changes in galactosylation, fucosylation, or high-mannose content can translate into measurable differences in biological activity and clinical performance.⁸ As a result, regulatory agencies require detailed glycan characterization and consistent monitoring throughout product development and manufacturing.^11,12

Unlike amino acid sequence variants, glycosylation is not genetically encoded. Instead, it arises from host-cell–specific biosynthetic pathways that are sensitive to culture conditions, nutrient availability, and process parameters. This inherent variability places significant pressure on analytical laboratories to generate glycan data that are not only accurate, but also reproducible, scalable, and suitable for long-term trending and biosimilar comparability.

Limitations of Traditional Manual Glycan Workflows

Conventional N-glycan analysis typically involves enzymatic release of glycans, fluorescent labeling, chromatographic separation, and mass spectrometric detection.¹³ While these methods are analytically robust and well established, they rely heavily on manual handling and expert interpretation. Each step introduces opportunities for operator-dependent variability, particularly when workflows are applied across multiple analysts, laboratories, or sites.

Manual workflows often provide reliable quantitation for dominant glycan species, but reproducibility can deteriorate for low-abundance glycans such as sialylated or high-mannose structures. These minor species, despite their low abundance, are frequently biologically relevant and may influence clearance, immunogenicity, or effector function. Variability in their measurement complicates comparability exercises, stability studies, and regulatory submissions.

As development timelines shorten and sample volumes increase, the scalability limitations of manual workflows become increasingly apparent. High-throughput applications such as clone selection, process optimization, and biosimilar similarity assessments demand analytical approaches that can deliver consistent results across large data sets.

Automation as an Enabler of Scalable Glycoanalytics

Automated glycan preparation platforms have been developed to address these challenges by standardizing critical sample preparation steps and minimizing operator intervention.^14,15 Cartridge- or plate-based systems integrate antibody capture, denaturation, enzymatic digestion, labeling, and cleanup into a single controlled workflow. Reaction conditions are tightly regulated, reducing variability and improving consistency across samples and runs.

In addition to improved reproducibility, automation provided a clear operational advantage in this study. Sample preparation time decreased from approximately 5 h for the manual workflow to 2.5 h per 96-well plate for the automated workflow, representing an approximately 50% reduction. Standardized liquid handling also reduced hands-on manipulation and enabled parallel processing, supporting higher throughput without changing the downstream chromatographic and mass spectrometric conditions. These advantages are particularly relevant when large sample sets must be processed consistently under regulated or multisite conditions.

Case Study: Trastuzumab Glycosylation Profiling

To assess the impact of automation on glycan profiling outcomes, we performed a direct comparison of manual and automated N-glycan workflows using an in-house trastuzumab biosimilar as a model system. Trastuzumab is a humanized IgG1 antibody with a well-characterized Fc glycosylation profile, making it a suitable benchmark for evaluating both qualitative coverage and quantitative reproducibility.

Both workflows were evaluated across seven independent replicate preparations. The manual workflow used conventional deglycosylation, labeling, cleanup, and liquid chromatography–mass spectrometry (LC–MS) with analyst-guided peak integration. The automated workflow used a robotic liquid-handling platform for glycan release, fluorescent labeling, and cleanup, followed by standardized data processing. Identical chromatographic and mass spectrometric conditions were used at the detection stage, allowing the comparison to focus on differences introduced during sample preparation and data processing.

Glycan Repertoire Coverage and Quantitative Differences

We first analyzed the N-glycan repertoire of in-house trastuzumab using a conventional manual workflow. LC–MS profiling reproduced the expected IgG1 glycosylation pattern, dominated by the core-fucosylated agalactosylated form (G0F, ~56%). Galactosylated forms were also detected, including G1F (22.49%) and G2F (2.26%), while non-fucosylated glycans were observed at lower levels (G0, 4.87%; G1, 1.21%). Minor subgroups included high-mannose glycans (Man, 2.89%), sialylated species (0.69%), and a pooled complex group (3.61%) composed of multiple low-abundance structures (e.g., G0F-N, G0-N, G1F-N, H4N3, H4N5F1, H6N3).

Although the manual workflow captured the expected glycan repertoire, its reproducibility was limited, with CVs ranging from 1–22%. (Figures 2a–c, Table 1). Replicate profiles within single experiments were acceptable, but variability increased across independent runs, particularly for galactosylated glycans. Coefficients of variation for minor species ranged from 1.25–21.53%, reflecting operator-dependent variability during glycan release, labeling, and peak integration.

To determine whether automation could improve analytical consistency and throughput, we implemented an automated liquid-handling workflow that combined IgG capture, denaturation, PNGase F digestion, fluorescent labeling, and cleanup in a single robotic pipeline.The automated workflow reduced sample-preparation time from approximately 5 h to 2.5 h per 96-well plate. It reproduced the glycan repertoire detected manually but yielded a different quantitative distribution: G0F increased from 55.62% to 62.58%, whereas G1F and G2F decreased from 22.49% and 2.26% to 14.99% and 1.13%, respectively. The automated workflow also reported G0 (7.22%), G1 (0.53%), high-mannose glycans (3.20%), sialylated glycans (1.76%), and complex glycans (6.01%). These results demonstrate preserved qualitative coverage but also show that quantitative abundance values are workflow dependent. (Figures 2a–2c and Table 1). Statistical comparisons indicated differences for G0F (p < 0.01), G1F (p < 0.001), and G2F (p < 0.05).

Direct comparison showed that the manual workflow reported higher total galactosylation (G1F + G2F, 24.75% versus 16.12%), whereas the automated workflow reported higher G0F abundance (62.58% vs 55.62%). Total non-fucosylated G0 + G1 abundance was also higher with automation (7.75% vs 6.08%), although the individual G1 value was lower (0.53% vs 1.21%). (Figure 2a–b and Table 1).

High-mannose glycans showed minor differences (Man: ~ 3.15% in automation vs ~ 2.89% manual), whereas sialylated species were enriched in automation (1.76% vs 0.69%). Complex glycans, particularly G0F-N and G0-N, were consistetly higher in automation (6.01% vs 3.61%), reflecting redistribution from low-abundance trace glycans observed in the manual workflow such as H4N3, H4N5F1, H6N3.

Taken together, the two workflows provided comparable qualitative glycan coverage but different quantitative distributions. The principal advantage of automation was not identical abundance values; rather, it improved consistency and operational efficiency. Mean CV decreased from 7.95% with the manual workflow to 5.70% with automation, an approximately 28% relative reduction, while preparation time decreased by approximately 50%. These findings support automation for high-throughput and longitudinal applications, provided that workflow-specific performance is validated and the same method is used consistently for comparability assessments. (Figures 2a–2c).

Based on seven replicate analyses, the average CV across the eight glycan groups was lower for the automated workflow than for the manual workflow (5.70% versus 7.95%), representing a 28.1% relative reduction in average variability. The greatest reductions were observed for high-mannose glycans (1.12% versus 13.43%), sialylated glycans (2.80% versus 15.26%), and complex glycans (4.11% versus 5.87%). Despite workflow-dependent differences in the relative abundances of individual glycans, the two methods showed strong linear agreement (R²=0.9964, r≈0.983), indicating that 96.64% of the variation in automated measurements was explained by the corresponding manual measurements. Automation also reduced sample-preparation time by approximately 50%. Collectively, automation preserved the qualitative glycan profile while improving processing efficiency and average repeatability, although its effect on variability differed among glycan groups. (Figures 2a–c and Table 1).

Closer inspection of the G0F peak showed a reproducible workflow-dependent difference in relative abundance. Manual processing yielded a lower G0F estimate (Figure 3a), whereas the automated workflow yielded a higher estimate. (Figure 3b). Because both sample preparation and data-processing steps can affect glycan recovery and peak integration, this difference should be interpreted as a method-dependent quantitative shift rather than biological variation. Workflow-specific validation and consistent use of methods are therefore essential when comparing batches or establishing long-term trends.

Implications for Biopharmaceutical Development

The advantages of automated glycan profiling extend beyond individual experiments. A 50% reduction in preparation time, standardized handling, and lower replicate variability can increase analytical capacity for clone selection, process optimization, biosimilar comparability, and long-term process monitoring. Automation can also improve harmonization across analysts, laboratories, and sites by reducing the number of operator-dependent sample-preparation steps.

From a regulatory standpoint, automated workflows enhance data integrity by minimizing manual intervention and enabling standardized, auditable processes. Structured and consistent data sets further support advanced data analysis approaches, including multivariate analysis and digital quality monitoring initiatives.

Conclusion

This comparative evaluation demonstrates that automated N-glycan profiling provides a more efficient and reproducible alternative to conventional manual processing. By reducing processing time and operator-dependent variability, automation offers a robust platform for high-throughput glycoanalytical applications.

However, the workflow-dependent shifts observed in G0F, G1F, and G2F abundances indicate that results generated using manual and automated methods should not be considered directly interchangeable without appropriate bridging studies. Consistent application and validation of the selected workflow are therefore essential for longitudinal monitoring and comparability assessments.

References:

1. Udeabor, S. E. Monoclonal Antibodies in Modern Medicine: Their Therapeutic Potential and Future Directions. Trends Pharm Biotechnol. 2024, 2 (2), 12–20. DOI: 10.57238/tpb.2024.153196.1019
2. Rawat, S.; Hussain, M. S.; Mohapatra, C.; et al. An Overview of Monoclonal Antibodies and Their Therapeutic Applications. Nat Volatiles Essent Oils 2021, 8 (06), 4121–4130.
3. Ndoja, S.; Lima, H. Monoclonal Antibodies. In Current Developments in Biotechnology and Bioengineering: Human and Animal Health Applications; Elsevier, 2017, pp 71–95. DOI: 10.1016/B978-0-444-63660-7.00004-8
4. Swain, S. M.; Shastry, M.; Hamilton, E. Targeting HER2-Positive Breast Cancer: Advances and Future Directions. Nat Rev Drug Discov. 2023, 22, 101–126. DOI: 10.1038/s41573-022-00579-0
5. Sabit, H.; Abbas, S.; El-Safoury, M. T.; et al. Antibody–Drug Conjugates in Breast Cancer: Navigating Innovations, Overcoming Resistance, and Shaping Future Therapies. Biomedicines 2025, 13 (9), 2227. DOI: 10.3390/biomedicines13092227
6. Higel, F.; Seidl, A.; Sörgel, F.; et al. N-glycosylation Heterogeneity and the Influence on Structure, Function, and Pharmacokinetics of Monoclonal Antibodies and Fc Fusion Proteins. Eur J Pharm Biopharm. 2016, 100, 94–100. DOI: 10.1016/j.ejpb.2016.01.005
7. Majewska, N. I.; Tejada, M. L.; Betenbaugh, M. J.; et al. N-glycosylation of IgG and IgG-like Recombinant Therapeutic Proteins: Why Is It Important and How Can We Control It? Annu Rev Chem Biomol Eng. 2020, 11, 311–338. DOI: 10.1146/annurev-chembioeng-102419-010001
8. Boune, S.; Hu, P.; Epstein, A. L.; et al. Principles of N-linked Glycosylation Variations of IgG-based Therapeutics: Pharmacokinetic and Functional Considerations. Antibodies 2020, 9 (2), 22. DOI: 10.3390/antib9020022
9. Pleass, R. J. The Therapeutic Potential of Sialylated Fc Domains of Human IgG. MAbs 2021, 13 (1), 1953220. DOI: 10.1080/19420862.2021.1953220
10. Golay, J.; Andrea, A. E.; Cattaneo, I. Role of Fc Core Fucosylation in the Effector Function of IgG1 Antibodies. Front Immunol. 2022, 13, 929895. DOI: 10.3389/fimmu.2022.929895
11. Kavur, M. M.; Lauc, G.; Pezer, M. Systems Glycobiology: Immunoglobulin G Glycans as Biomarkers and Functional Effectors in Aging and Diseases. Compr Glycosci. 2021, 1, 439–478. DOI: 10.1016/B978-0-12-819475-1.00086-9
12. Delobel, A. Glycosylation of Therapeutic Proteins: A Critical Quality Attribute. Methods Mol Biol. 2021, 2271, 1–21. DOI: 10.1007/978-1-0716-1241-5_1
13. Helali, Y.; Delporte, C. Updates of the Current Strategies of Labeling for N-glycan Analysis. J Chromatogr B Analyt Technol Biomed Life Sci. 2024, 1237, 124068. DOI: 10.1016/j.jchromb.2024.124068
14. Schilling, M.; Feng, P.; Sosic, Z.; et al. Development and Validation of a Platform Reduced Intact Mass Method for Process Monitoring of Monoclonal Antibody Glycosylation During Routine Manufacturing. Bioengineered 2020, 11 (1), 1301–1312. DOI: 10.1080/21655979.2020.1842651
15. Dutt, M.; Duong, M. N.; Bringans, S.; et al. Semi-Automated Lectin Magnetic Bead Array (LeMBA) for Translational Serum Glycoprotein Biomarker Discovery and Validation. Methods Mol Biol. 2023, 2628, 395–411. DOI: 10.1007/978-1-0716-2978-9_25