Simplifying glycan monitoring of complex antigens such as the SARS-CoV-2 spike to accelerate vaccine development

The COVID-19 pandemic spurred the rapid development of various SARS-CoV-2 vaccines, including mRNA vaccines, virus-like particles, and protein subunits. While mRNA vaccines garnered significant attention, recombinant protein antigens remain prominent in preclinical and clinical stages, with several recently approved. These proteins also serve in antibody serology assays for detecting SARS-CoV-2 antibodies. The SARS-CoV-2 spike glycoprotein, a key vaccine antigen candidate, possesses up to 22 N-glycosylation sites per protomer, resulting in a complex glycosylation profile. This profile varies across sites and depends on the producing cell line, leading to batch-to-batch variation even under consistent conditions. Critically, glycosylation significantly impacts immunogenicity; for instance, altered glycosylation in the spike protein can affect its protective efficacy. Despite this, glycan analysis is not routinely performed for protein antigens, with glycosylation often assumed to be typical of the expression cell line. This lack of consistent glycosylation monitoring hinders the assessment of glycosylation's impact on immunogenicity and cross-study comparisons. Given glycosylation's influence on critical product attributes in glycoprotein-based therapeutics, reliable monitoring assays are crucial. While established methods like HILIC-FLD and LC-MS are valuable for simpler glycoproteins, their application to complex proteins like the SARS-CoV-2 spike yields intricate datasets challenging to analyze. Therefore, a fast, reliable, and easily implemented assay is needed for routine glycosylation monitoring of complex biotherapeutics.

Existing methods for glycoprotein analysis include hydrophilic interaction chromatography with fluorescence detection (HILIC-FLD) and liquid chromatography-mass spectrometry (LC-MS). HILIC-FLD, while effective for simpler glycoproteins with few glycosylation sites, becomes complex and challenging to interpret with multiple glycosylation sites. LC-MS provides detailed site-specific glycosylation information but involves time-consuming sample preparation and analysis, producing substantial datasets requiring sophisticated interpretation. Studies have already utilized HILIC-FLD and LC-MS to characterize SARS-CoV-2 spike protein glycosylation, but the resulting complex profiles highlight the need for a simpler approach. The literature emphasizes the lack of routine glycosylation analysis in protein antigen production, often overlooking the impact of glycosylation variations on immunogenicity. Studies have demonstrated that alterations in glycosylation can significantly influence the immunogenicity of viral antigens, including the SARS-CoV-2 spike protein and influenza hemagglutinin.

This study investigated the use of high-performance anion-exchange chromatography coupled with pulsed amperometric detection (HPAEC-PAD) for rapid glycosylation monitoring. Two cohorts of spike proteins were analyzed: one with major glycosylation differences and another exhibiting minor variations representative of typical production batch differences. The first cohort comprised spike proteins produced in engineered Chinese hamster ovary (CHO) cell lines with varying glycosylation patterns: wild-type CHO cells, FUT8 knockout (afucosylated), ST3GAL4 knockout (asialylated), double knockout (afucosylated and asialylated), and wild-type CHO cells treated with kifunensine (high-mannose glycans). The second cohort comprised three independent batches of spike protein from the same stable CHO pool, expected to show subtle glycosylation variations. Each cohort was analyzed using three assays: HPAEC-PAD (for neutral and acidic monosaccharide content), HILIC-FLD (for released N-glycans), and LC-MS (for glycopeptides). HPAEC-PAD involved separate analyses for neutral sugars (Fuc, GalNAc, GlcNAc, Gal, Glc, Man) and sialic acids (Neu5Ac, Neu5Gc). For HILIC-FLD, N-glycans were released, labeled with 2-aminobenzamide, and separated by UPLC with fluorescent detection. LC-MS utilized bottom-up proteomics with four proteases to analyze glycopeptides, followed by data processing with an in-house algorithm, GlycoPIQ. The results from the three assays were compared to assess the suitability of HPAEC-PAD for rapid glycosylation monitoring.

HPAEC-PAD effectively detected major glycosylation differences in the engineered CHO cell line-derived spike proteins, aligning with HILIC-FLD and LC-MS results. Specifically, the absence of fucose in FUT8 knockout lines and the reduced sialic acid in ST3GAL4 knockout lines were clearly identified. The kifunensine-treated spike protein showed a predominantly high-mannose glycan profile, as expected. Importantly, HPAEC-PAD also detected minor glycosylation variations among the three batches from the same stable pool. These subtle differences, also observed by HILIC-FLD and LC-MS, demonstrate HPAEC-PAD's sensitivity in detecting lot-to-lot variations. The HILIC-FLD analysis, though providing a complex profile, supported the observations from HPAEC-PAD by showing differences in fucosylation and sialylation. LC-MS confirmed the site-specific glycosylation patterns, corroborating the findings from both HPAEC-PAD and HILIC-FLD. For example, the increased sialylation observed in PRO1-468 by LC-MS was reflected in the higher Neu5Ac content detected by HPAEC-PAD. Similarly, the shift towards glycans with more antennae in PRO1-468 was mirrored by a greater GlcN concentration in HPAEC-PAD results. These results strongly suggest that HPAEC-PAD provides a good correlation with the more detailed but far more complex analysis by HILIC-FLD and LC-MS. The comparison between three independent spike protein batches (PRO1-392, PRO1-394, and PRO1-412) produced from the same CHO pool further underscored HPAEC-PAD's sensitivity. While Fuc and Man levels were consistent, significant differences were found in GlcN, Gal, and Neu5Ac content. These differences were consistent with the HILIC-FLD and LC-MS data, demonstrating the ability of HPAEC-PAD to detect subtle glycosylation variations between batches produced under supposedly identical conditions. Cell viability and culture duration were explored as possible factors impacting Neu5Ac content differences among the batches.

Established assays, such as HILIC-FLD and LC-MS, struggle with routine monitoring of complex glycosylation profiles in highly glycosylated antigens. The HPAEC-PAD monosaccharide assays offer a significant advantage due to their speed and simplicity. Sample preparation is minimal, requiring no purification or derivatization steps. Results are obtained quickly, without intricate processing or calculations, making it highly suitable for quality control settings. This study aimed to determine if HPAEC-PAD could effectively monitor batch-to-batch glycosylation variations in SARS-CoV-2 spike glycoprotein. The results show that HPAEC-PAD effectively monitors both major and minor glycosylation variations, aligning with the more complex HILIC-FLD and LC-MS data. HPAEC-PAD is faster and provides quantitative, reproducible, and easily interpretable results. While HILIC-FLD and LC-MS provide deeper insight into complex glycan structures, HPAEC-PAD excels in the rapid monitoring of batch-to-batch consistency.

This study demonstrates the efficacy of HPAEC-PAD monosaccharide assays for monitoring batch-to-batch glycosylation variations in SARS-CoV-2 spike glycoprotein antigens. The results correlate well with data obtained by HILIC-FLD and LC-MS, highlighting HPAEC-PAD’s speed, simplicity, and quantitative accuracy. Its suitability for quality control settings is evident, offering a valuable tool for efficient vaccine development, particularly crucial in pandemic response scenarios. Future studies could explore HPAEC-PAD's applicability across a broader range of complex glycoproteins and investigate the correlation between specific monosaccharide variations and the resulting impact on immunogenicity.

The study primarily focused on N-linked glycosylation, with limited insights into O-linked glycosylation due to the analytical methods employed. While HPAEC-PAD detected some minor glycosylation variations, it may not capture the full spectrum of subtle changes detectable by LC-MS, which provides more detailed site-specific information. The interpretation of HILIC-FLD data was limited by the complexity of the spike protein glycan profiles, hindering the identification of all glycans. The study did not directly investigate the correlation between glycosylation variations and immunogenicity, which requires further investigation.