(M1430-09-57) Using Enzymes to Accelerate PS80 Degradation and Elucidate Mechanistic Impacts of PS80 Degradants on Protein Stability via Thermal- and Agitation-Induced Aggregation

Purpose: Polysorbate 80 (PS80) is commonly added to biologics formulations for protection against agitation stresses. However, PS80 can degrade through various pathways that result in different combinations of degradants that impact the stability of the biologic. Recent studies indicate an impact of individual PS80 degradants on protein stability, but the combined effects of all the degradants that result from a particular PS80 degradation pathway is not well understood. This work demonstrates a method to enable mechanistic understanding on the correlation between specific PS80 degradation patterns on biologic product stability, specifically focusing on enzymatic PS80 hydrolysis.

Methods: 32 different esterases, lipases, and proteases immobilized on beads were screened for PS80 hydrolysis activity by incubating the enzymes with 2% PS80 in formulation buffer for 24 h at 30 °C on a 200 RPM shaker. PS80 degradation profiles were determined by using high performance liquid chromatography with a charged aerosol detector. Three enzymes were selected for unique degradation profiles that were confirmed by mass spectrometry. The enzymes were removed from solution via membrane filtration and the resulting hydrolyzed PS80 was used to formulate mAb1 samples. To probe the particular protein aggregation mechanisms these PS80 degradants can impact, samples were subjected to the following stresses: acoustic mixing at 20 g for up to 24 h to accelerate air-water interface-mediated aggregation; incubation at 40 °C for up to 3 months to accelerate bulk-mediated aggregation; and room temperature orbital shaking at 300 rpm for up to 8 weeks to probe a combination of interface- and bulk-mediated aggregation. The resulting samples were analyzed for subvisible particles using backgrounded membrane imaging, and fragmentation and aggregation using size exclusion chromatography (SEC). To extract further insight into particle formation mechanisms, images of subvisible particles were passed through a pre-trained deep learning model to extract feature vectors that were consequently reduced into two dimensions using principal component analysis. This allows visualization of particle image features on a 2D map.

Results: PS80 degraded by Candida Antarctica lipase B (CALB), Rhizomucor miehei lipase (RML), and Bacillus subtilis protease (BSP) were selected to formulate MAb1 samples as they exhibited distinguishable PS80 degradation profiles via mass spectrometry (data not shown). Under all stress conditions, SEC measurements demonstrated that BSP-degraded PS80 induced noticebly higher levels of fragmentation compared to other samples (Fig. 1). Samples containing CALB- and RML-degraded PS80 did not exhibit noticeably different levels of aggregation or fragmentation compared to control samples. These results suggest that, of the PS80 degradants probed in this study, only those formed by BSP-hydrolysis impact chemical degradation of mAb1. Under acoustic stress, subvisible particle measurements indicated that BSP-degraded PS80 induced significantly higher levels of particle formation compared to all other PS80-containing samples. However, this was not the case under thermal incubation, where the samples containing BSP-degraded PS80 exhibited even less particle formation than the control sample containing PS80 (Fig. 2). Instead, CALB- and RML-degraded PS80 induced high levels of particle formation that were not observed during acoustic mixing. These observations suggest that CALB- and RML-degraded PS80 are effective at mitigating air-water interface-induced protein aggregation, but accelerate protein particle formation in bulk. Conversely, BSP-degraded PS80 protects against protein particle formation in bulk but is ineffective at preventing interface-mediated protein particle formation. Interestingly, the particle formation trends observed during orbital shaking mirror those observed during thermal incubation, suggesting that bulk-mediated protein particle formation driven by exposure to the room temperature environment for as little as two weeks (the first intermediate time point measured) dominated over the air-water interface turnover promoted by the orbital shaking. To confirm this theory, images of the sub-visible particles measured in each sample were passed through a pre-trained deep learning model and visualized on a 2D principal component map (Fig. 3). Indeed, the 2D maps of samples containing BSP-, CALB-, and RML-degraded PS80 illustrate that the particle populations formed during orbital shaking more closely resemble those formed during thermal incubation compared to acoustic mixing, corroborating the hypothesis.

Conclusion: These results indicate that different PS80 degradation profiles can have with unique mechanistic impacts on protein stability. Overall, this work demonstrates a method to investigate the impacts of combinations of PS80 enzymatic degradants on the stability of a biologic. This tool will enable a better understanding of the relationship between PS80 degradation and biologics stability, and more robust therapeutic products.

References: Larson, Nicholas R., et al. "Comparison of polysorbate 80 hydrolysis and oxidation on the aggregation of a monoclonal antibody." Journal of Pharmaceutical Sciences 109.1 (2020): 633-639.
Zhang, Jingtao, Jiayi He, and Katelyn J. Smith. "Fatty Acids Can Induce the Formation of Proteinaceous Particles in Monoclonal Antibody Formulations." Journal of Pharmaceutical Sciences 111.3 (2022): 655-662.
Salami, Hossein, Wood, Caitlin, Ouyang, Hanlin, Zhao, Xi, Skomski, Daniel. “Deep Learning Analysis Models Pre-trained on Daily Objects are Useful for the Preliminary Characterization of Particulate Pharmaceutical Samples.” Biotechnology and Bioengineering doi: 10.1002/bit.28488

BSP-degraded PS80 induced increased fragmentation in mAb1 samples

Subvisible particle formation trends observed during orbital shaking more closely resemble those observed during acoustic mixing than during thermal incubation

2-dimensional principal component maps demonstrate the particle populations formed during orbital shaking more closely resemble those formed during thermal incubation than during acoustic mixing. Dashed boxes highlight map areas to be compared between thermal incubation, acoustic mixing, and orbital shaking.