Thesis entitled “Upstream process development for production of biotherapeutics in mammalian cell culture.”
Date : 11th February 2025
Venue : Block II Room 276 (Chemical Engineering Seminar Room)
Time : 1030 am
Abstract: Monoclonal antibodies (mAbs) remain dominant in the category of biotherapeutics in terms of approvals of genuinely new biopharmaceuticals coming to market and market value contribution during 2018 – 2022. Chinese hamster ovary (CHO) cell is the default mammalian expression system with 89% of total products expressed in mammalian system because of its capabilities for post translational modifications (PTMs) of mAbs which closely resembles human mAbs. Despite being most efficient therapeutic molecules, production of mAbs in CHO cells imposes certain challenges like making them cost effective, maintaining batch-to-batch consistency, understanding process variability to make them suitable for a continuous production platform. Cell culture media forms the immediate environment of cells where they grow and propagate and eliminate their waste. Media is the water-based solution in which many components such as amino acids, minerals, vitamins, and salts are dissolved to make it suitable for cell growth. In this study we explored the role of the media in manipulating the critical quality attributes (CQAs) of mAb therapeutics. We have examined the use of pyridoxine hydrochloride, biotin, folic acid, choline chloride, thiamine hydrochloride, and D-calcium pantothenate, as well as metal ions like Fe, Zn, Cu, Co, Mg, Mn, and Ni as addition to cell culture media. The findings suggest that pyridoxine enhances volumetric productivity, while Zn, Cu, Fe, Mn, and biotin influence charge heterogeneity. Fe, Mn, and Ni increase the production of acidic variants, whereas Cu and biotin inhibit it. Zn decreases the formation of basic variants, while biotin increases them. Further we demonstrate how we can achieve the charge variant profile of the reference molecule Herceptin® by utilizing metal ions as charge variant modulators. The in-house produced trastuzumab exhibited significantly lower acidic variants (17.64 ± 1.07% acidic and 12.86 ± 0.43% basic) compared to the reference product (24.97 ± 0.54% acidic and 11.41 ± 1.44% basic). After a thorough investigation, zinc was selected to regulate basic charge variations and iron was used to modify acidic charge variants. With optimal metal ion supplementation, the charge variant profile for in-house monoclonal antibody (24.7 ± 0.83% acidic and 14.4 ± 0.64% basic) closely matched that of the reference product. Although a minor reduction in viable cell density and product titer (~7%) was observed, there was no significant impact on other quality attributes, including glycans and specific cell productivity. Next, an extensive screening of various media additives (including metal ions, vitamins, sugars, and nucleosides) was conducted to evaluate their effects on glycosylation. This was followed by optimizing the concentration of the shortlisted components to control the galactosylation, afucosylation, mannosylation, and sialylation of monoclonal antibodies (mAbs). A one factor at a time (OFAT) approach was employed to screen the media additives for their effects. Six additives were shortlisted based on their impact and used to modulate the glycosylation profile of an in-house produced mAb (G0 2.38 ± 0.08%, G0F 75.58 ± 0.45%, G1F 10.07 ± 0.04%, G2F 0.54 ± 0.01%, G0F-N 5.84 ± 0.32%, sialylation 1.60 ± 0.33%, mannosylation 1.56 ± 0.39%) to match the glycan profile of a commercially available reference product (G0 2.49 ± 0.07%, G0F 37.83 ± 0.37%, G1F 34.77 ± 0.03%, G2F 4.87 ± 0.01%, G0F-N 2.34 ± 0.12%, sialylation 9.84 ± 0.30%, mannosylation 2.86 ± 0.29%). The proposed approach resulted in a glycan profile (G0 2.10 ± 0.07%, G0F 38.00 ± 0.49%, G1F 31.92 ± 0.09%, G2F 5.26 ± 0.54%, G0F-N 1.92 ± 0.02%, sialylation 10.28 ± 1.68%, mannosylation 3.12 ± 0.29%) that was nearly identical to the reference product. Importantly, other quality attributes, including charge variants, aggregates, titer, and cell viability, were not significantly impacted by the addition of these additives. Next, an end-to-end machine learning framework for media component selection, CQA prediction, and optimization is presented. The preliminary dataset for feature selection was generated by conducting CHO-GS(-/-) cell culture in media formulations with varying metal ion concentrations. The acidic and basic charge variant composition of the reference product (24.97 ± 0.54% acidic and 11.41 ± 1.44% basic) was chosen as the target variable to evaluate the media formulations. Pearson's correlation coefficient and random forest-based techniques were used for feature ranking and selection to predict acidic and basic charge variants. Additionally, a global interpretation analysis using Shapley Additive exPlanations (SHAP) was utilized to select optimal features by evaluating the contributions of each feature in the extracted vectors. Finally, fifteen different regression models were employed to predict medium combinations, with grid search and random search cross-validation used for hyperparameter optimization. Experimental results demonstrate that Fe and Zn significantly impact the charge variant profile. Finally, a near-infrared (NIR) spectroscopy-based quantification method has been developed for media additives known to be potential glycan modulators. A convolutional neural network (CNN)-based chemometric model was created to estimate galactose and uridine concentrations in various media formulations. By leveraging data augmentation, the CNN model demonstrated strong predictive performance (test R² > 0.9) for both analytes in real-time. This model was further combined with a Design of Experiments (DoE) based statistical model to predict glycosylation using the concentrations of media additives as inputs. The predicted glycosylation distributions aligned closely with the actual distributions, showing no significant differences in the investigated media formulations. The proposed approach can be utilized for media characterization and the implementation of process analytical technology (PAT) based control of mammalian cell culture raw materials. Overall, this thesis aims to address the key challenges faced by the biopharmaceutical industry in upstream process development (USP) more specifically in media optimization. Results of these studies could be utilized for maintaining consistent CQAs (charge variant or glycosylation).