Multi-component, multi-scale Raman spectroscopy modeling results from a monoclonal antibody producing CHO cell culture process including data from two development scales (3 L, 200 L) and a clinical manufacturing scale environment (2,000 L) are presented. Multivariate analysis principles are a critical component to partial least squares (PLS) modeling but can quickly turn into an overly iterative process, thus a simplified protocol is proposed for addressing necessary steps including spectral preprocessing, spectral region selection, and outlier removal to create models exclusively from cell culture process data without the inclusion of spectral data from chemically defined nutrient solutions or targeted component spiking studies. An array of single-scale and combination-scale modeling iterations were generated to evaluate technology capabilities and model scalability. Analysis of prediction errors across models suggests that glucose, lactate, and osmolality are well modeled. Model strength was confirmed via predictive validation and by examining performance similarity across single-scale and combination-scale models. Additionally, accurate predictive models were attained in most cases for viable cell density and total cell density; however, these components exhibited some scale-dependencies that hindered model quality in cross-scale predictions where only development data was used in calibration. Glutamate and ammonium models were also able to achieve accurate predictions in most cases. However, there are differences in the absolute concentration ranges of these components across the datasets of individual bioreactor scales. Thus, glutamate and ammonium PLS models were forced to extrapolate in cases where models were derived from small scale data only but used in cross-scale applications predicting against manufacturing scale batches.

Mitigating risks to biotherapeutic protein production processes and products has driven the development of targeted process analytical technology (PAT); however implementing PAT during development without significantly increasing program timelines can be difficult. The development of a monoclonal antibody expressed in a Chinese hamster ovary (CHO) cell line via fed-batch processing presented an opportunity to demonstrate capabilities of altering percent glycated protein product. Glycation is caused by pseudo-first order, non-enzymatic reaction of a reducing sugar with an amino group. Glucose is the highest concentration reducing sugar in the chemically defined media (CDM), thus a strategy controlling glucose in the production bioreactor was developed utilizing Raman spectroscopy for feedback control. Raman regions for glucose were determined by spiking studies in water and CDM. Calibration spectra were collected during 8 bench scale batches designed to capture a wide glucose concentration space. Finally, a PLS model capable of translating Raman spectra to glucose concentration was built using the calibration spectra and spiking study regions. Bolus feeding in mammalian cell culture results in wide glucose concentration ranges. Here we describe the development of process automation enabling glucose setpoint control. Glucose-free nutrient feed was fed daily, however glucose stock solution was fed as needed according to online Raman measurements. Two feedback control conditions were executed where glucose was controlled at constant low concentration or decreased stepwise throughout. Glycation was reduced from ∼9% to 4% using a low target concentration but was not reduced in the stepwise condition as compared to the historical bolus glucose feeding regimen.

Accumulation of lactate in mammalian cell culture often negatively impacts culture performance, impeding production of therapeutic proteins. Many efforts have been made to limit the accumulation of lactate in cell culture. Here, we describe a closed loop control scheme based on online spectroscopic measurements of glucose and lactate concentrations. A Raman spectroscopy probe was used to monitor a fed-batch mammalian cell culture and predict glucose and lactate concentrations via multivariate calibration using partial least squares regression (PLS). The PLS models had a root mean squared error of prediction (RMSEP) of 0.27 g/L for glucose and 0.20 g/L for lactate. All glucose feeding was controlled by the Raman PLS model predictions. Glucose was automatically fed when lactate levels were beneath a setpoint (either 4.0 or 2.5 g/L) and glucose was below its own setpoint (0.5 g/L). This control scheme was successful in maintaining lactate levels at an arbitrary setpoint throughout the culture, as compared to the eventual accumulate of lactate to 8.0 g/L in the historical process. Automated control of lactate by restricted glucose feeding led to improvements in culture duration, viability, productivity, and robustness. Culture duration was extended from 11 to 13 days, and harvest titer increased 85% over the historical process. Biotechnol. Bioeng. 2016;113: 2416-2424. © 2016 Wiley Periodicals, Inc.