As biologics pipelines expand beyond conventional monoclonal antibodies, developers are facing new questions about how best to manage product quality, consistency, and performance. One area drawing increased attention is glycosylation – a critical quality attribute that can influence how antibody-based therapies behave.
In the following interview, Sojeong Lee, Director of Upstream Development at Samsung Biologics, discusses why glycosylation control is moving upstream, how early analytical insights can support development decisions, and what capabilities may be needed as biologics become more complex.
How is the growing complexity of biologic modalities changing the way developers think about glycosylation control?
In the past, many contract development and manufacturing organizations (CDMOs) operated with a “fix it later” approach, because traditional antibodies afforded them the time to adjust their processes at a later stage, even into Phase 2.
As rising numbers of next-generation antibodies like bispecifics, antibody-drug conjugates (ADCs), and Fc-fusions, as well as immunotherapies, enter the development pipeline, there is no room for error. These modalities are much more prone to developing heterogeneous glycans, a factor that can compromise clinical efficacy and batch-to-batch consistency. Even a single glycan variant can derail clinical efficacy. Bispecifics with asymmetric glycosylation, for instance, will lose target engagement.
This kind of glycan specificity means that glycosylation must now be a design parameter in the initial production of next-generation antibodies, not just a downstream attribute.
What are some of the limitations of trying to address glycosylation issues through downstream processing alone?
Downstream processing often acts as a band-aid for glycosylation issues. It can cover the small wounds of traditional antibody therapeutics like monoclonal antibodies (mAbs), but not the great damage that can be wrought by improper glycosylation in more complex antibody therapeutics. Modern molecules with glycosylation considerations, therefore, require more foresight.
For example, glycans can affect the drug-antibody ratio (DAR) stability of ADCs, and glycan heterogeneity can cause ADC payload loss. These issues cannot be solved with downstream corrections alone.
From an industry perspective, how does early clone selection influence glycosylation consistency and antibody functionality?
It has a great effect. Around 80 percent of glycan variability is clone-dependent. Even minor differences in genes like MGAT1 and FUT8 in a clone can alter its Golgi enzyme expression, which in turn can affect the branching/fucosylation of subsequent antibodies, thereby affecting antibody-dependent cellular cytotoxicity (ADCC). A single glycan variant can therefore destroy a therapeutic’s mechanism of action. Consider an Fc-fusion immunotherapy targeting the interleukin-6 receptor. If produced from a clone with high mannose levels, the eventual therapeutic’s half-life could drop from an expected 21 days to three days. Only clone selection would eliminate the high-mannose producers early and avoid such a scenario.
What key quality attributes should developers be evaluating alongside productivity during early screening?
The key quality attributes, especially glycosylation patterns, are different depending on the mode of action of the molecule. Fucosylation level, for one. This variable directly controls ADCC potency and is particularly critical for oncology bispecifics. It can be screened for using rapid hydrophilic interaction liquid chromatography-ultra performance liquid chromatography (HILIC-UPLC).
Galactosylation symmetry is another key quality attribute to evaluate, because any asymmetry can lead to batch failure. It can be measured using LC-MS glycopeptide mapping.
Sialic acid linkage is another quality attribute to monitor. These terminal modifications typically involve an α-glycosidic bond connecting the second carbon of sialic acid to the glycan’s underlying sugar chain. The position of these linkages can have varying effects. An α2,6-sialylation can extend a glycan’s half-life, for instance, while an α2,3-sialylation can trigger inflammation. The quality can be measured via HILIC-UPLC with fluorescence detection or high-performance liquid chromatography (HPLC).
As previously mentioned, high mannose levels can also influence the efficacy of a potential therapeutic. An increase of 5 percent can lead to a rapid clearance of mannose receptors and, in turn, an increased immunogenicity risk. During screening, this quality should be measured using capillary electrophoresis.
How are high-throughput analytical methods reshaping early-stage decision-making in cell line development?
While early-stage clone selection historically prioritized titer alone, with glycosylation outcomes treated as secondary, we at Samsung Biologics have since evolved our approach. For the past five years, starting at cell line development, we have proactively selected clones using molecule- and target-specific analytics to ensure that both titer and glycan quality meet specifications from the outset. This integrated strategy eliminates the need to depend on glycan consistency, as high productivity and profile fidelity are co-optimized by design.
What advantages do early glycosylation insights provide compared to later-stage characterization?
Consistency from day one. Rather than expecting clones to exhibit a "perfect fit" glycan profile from the outset, glycosylation insights focus on selecting clones with the closest possible initial fit, followed by minimal small-scale modifications to refine the profile to the desired specifications.
If a near-identical match is achieved early, maintaining this consistency during scale-up through targeted, minimal adjustments then becomes key. The goal is to prioritize clones that inherently align as closely as possible with the target profile, reducing the need for extensive re-engineering while ensuring that scalability and product quality remain uncompromised. This approach balances initial clone selection rigor with pragmatic, small-scale optimization to achieve robust, reproducible outcomes across scales.
How can upstream process parameters be intentionally designed to influence glycosylation outcomes?
Upstream, two key parameters can be modulated for greater control of glycosylation. The first is temperature. By lowering the temperature early in the glycosylation process, cell cycle progression can be slowed, thereby reducing metabolic burden and providing sufficient time for enzymatic reactions. The second parameter is pH. By optimizing extracellular pH, a CDMO can influence the transport of ionized species across cellular membranes, indirectly affecting intracellular conditions for glycosylation enzymes.
Substrate modulation can also have a large impact on glycosylation. This cellular environment directly controls the intracellular availability of ingredients for the Golgi, where glycans are built. Therefore, even slight adjustments in a substrate’s mix of sugar nucleotide precursors, amino acids, and lipids can influence glycan branching.
What role do structured experimental approaches play in balancing productivity with quality?
Human interpretation alone is insufficient due to the complexity of experimental data. Tools such as design of experiments (DoE) are therefore essential to reduce experimental burden and analyze multiple quality metrics (e.g., titer, glycan profiles) simultaneously. CDMOs should strategically apply screening DoEs and optimization DoEs to identify critical process parameters. The application of these methodologies should be carried out by statistics-trained scientists.
Upstream monitoring also directly fulfills ICH Q8/Q11 guidelines as it defines glycosylation as a critical quality attribute, uses DoE to establish a design space for process parameters, and implements real-time monitoring for continuous quality assurance.
How does this approach support consistent therapeutic performance at scale?
If not properly monitored and controlled, the many variables of therapeutic production, from oxygen transfer rates to mixing times, will lead to microenvironmental differences between bioreactor runs. These differences will naturally impact glycan consistency, even across identical cells, leading to batch-to-batch variability.
Upstream glycosylation control mitigates these risks. To enact it at scale, CDMOs should rigorously review the quality analytics of their bioreactors during scale-up.
How does integrating quality considerations earlier help mitigate late-stage risk or regulatory challenges?
By design, upstream considerations reduce the risk of missteps in therapeutic quality and consistency. Early focus on glycosylation control, for instance, can prevent costly late-stage failures, such as batch rejection due to inconsistent ADCC. By demonstrating proactive quality assurance, the protocol can also reduce the number of queries from regulators.
How do you expect upstream glycosylation control strategies to evolve over the next few years?
The main changes will be twofold: automation and advanced analytics. Most CDMOs will already have integrated both qualities into their workflows in some way; the change over the coming years will be the technologies’ increasing sophistication.
Regarding automation, we can expect the further integration of online monitoring tools with automated feedback loops to dynamically adjust bioreactor conditions. Full, seamless automation will likely remain aspirational for the foreseeable future, but progress towards it will be visible.
As for advanced analytics, artificial intelligence (AI) and machine learning will be used for more rapid and accurate data interpretation, perhaps linking process parameters to glycan outcomes. The challenge for CDMOs will be providing large, reliable data on novel molecules, but technological progress may surmount this.
What capabilities will biopharma companies need as functional requirements become more demanding?
The main priority for any biopharma company should be scaling up capabilities to integrate vast data at high speed. As a result, analytical data on glycans could be translated into actionable process adjustments within hours, not weeks.
This quality will be crucial as advanced AI analysis becomes more embedded across production workflows. For CDMOs, these AI models should be trained on multi-batch datasets to predict optimal “golden batch” profiles. When using AI, data security and compliance must be considered.
Integration capabilities should be flexible to maintain quality regardless of molecule diversification or facility scale-up.
Newsletters
Receive the latest pharmaceutical news, personalities, education, and career development – weekly to your inbox.
