Mammalian cell culture requires complex media design, specific bioreactor design, and robust process control systems (Pohlscheidt et al., 2012). Among the many operating parameters that control the cell culture environment, appropriate mixing and aeration in large-scale bioreactors are not only crucial to meet cellular oxygen demand in both aerobic microbial and mammalian production processes (Gelves et al., 2014), but are also important for local fluid hydrodynamics related to the homogeneous distribution of energy dissipation, pH, bubble size, and oxygen concentration, all of which impacts cell behavior and product synthesis (Bylund et al., 1998, Koynov et al., 2007, Restelli et al., 2006, Serrato et al., 2004). Despite the use of Pluronic F68 (PF68) and other nonionic surfactants in cell culture media to protect cells from aeration associated shear forces (Chalmers and Bavarian, 1991, Murhammer, 2009, Murhammer and Goochee, 1990, Sieblist et al., 2013), cell damage from shear stresses remain as a factor to consider in large-scale cell cultures (Chaudhary et al., 2020). In addition, CO2 removal from the culture is a challenge in large-scale mammalian cell cultures due to the use of relatively low power inputs, low sparging air flow rates (deZengotita et al., 2002, deZengotita et al., 1998, Gray et al., 1996), and longer gas residence times (Sieblist et al. 2011) compared to their microbial counterparts. Therefore, modeling tools that include all of these scale-up considerations would be helpful in the design of agitation and air flow rates for manufacturing facility fit.

Application of computational fluid dynamics (CFD) modeling to characterize local hydrodynamics at different bioreactor scales has helped in the successful scale-up from laboratory to manufacturing bioreactors (Zhang et al., 2009). To properly capture the complexity of mammalian cell culture processes, a gas-liquid flow model that is capable of handling both agitation and gassing is needed (Dhanasekharan et al., 2005). There are two classes of gas-liquid flow models, namely the single bubble size model (SBS) and the population balance model (PBM). PBM captures the distribution of bubble sizes in a stirred bioreactor (Sarkar et al., 2016), where bubbles experience both breakage and coalescence under turbulent flow conditions (Dhanasekharan et al., 2005). However, available experimental data has not been adequate to calculate the parameters appearing in the coalescence and break-up kernels for a PBM model (Khopkar et al., 2006). The SBS gas-liquid flow model has been applied in simulating fermentation and cell culture processes. This model has been shown to reduce computational time by one order of magnitude while maintaining a local description of the two-phase flow details (Kazemzadeh et al., 2018, Khopkar et al., 2003, McClure et al., 2014, Morud and Hjertager, 1996, Ranade and Deshpande, 1999, Rathore et al., 2012).

Agitation related cell shear stress, either directly from impeller mechanical shear (Varley and Birch, 1999) or indirectly from small gas bubble bursts (Chisti, 2000) was found to be insignificant under “normal” operating bioreactor agitation rates for power input per unit volume (P/V) of 10–100 W/m3 (Hu et al., 2011, Nienow, 2010, Chalmers, 2015, Varley and Birch, 1999). However, the shear stress that stems from high gas entrance velocity (GEV) emitting from the sparger holes into the culture was reported to impact cell culture performance. A GEV of < 30 m/s was reported to be safe with respect to cellular damage in microalgae cultures (Barbosa et al., 2004) and NS0 cell culture (Zhu et al., 2008), while a GEV level of < 20 m/s was recommended for Chinese hamster ovary (CHO) cell cultures, because no negative impact on cell health was found at both small and large scales (Chaudhary et al., 2020).

Application of a carbon dioxide (CO2) stripping model to simulate bioreactor CO2 profiles is particularly of interest in process development and scale up. Historically, CO2 stripping models for bioreactors have been based on the CO2 mass transfer coefficient (kLa) (Mostafa and Gu, 2003, Matsunaga et al., 2009, Xing et al., 2009). These models were applicable for cell cultures where the gas bubbles had not saturated with CO2 before entering headspace, due to either low viable cell densities or short gas residence times. Recently, it was reported that gas bubbles need only ~ 4 seconds to become saturated with CO2 under representative cell culture conditions (Sieblist et al. 2011). In large bioreactors, the bubble’s residence time is much longer than a few seconds, for example, > 16 s in 5,000-L bioreactors (Xing et al. 2017), thus all bubbles would be saturated with CO2 upon leaving the culture into the bioreactor’s headspace. This makes CO2 stripping no longer dependent on CO2kLa in the new generation cell culture processes characterized by longer exponential growth phase, higher peak viable cell densities, and higher specific production rates, which produce more CO2 to be removed from the culture. An alternative CO2 stripping model was applied to cell culture manufacturing-scale bioreactors, which is independent of CO2kLa but accounts for the gas-residence and gas CO2 saturation times. Such a model was applied to CHO cell culture processes in 5,000-L and 25,000-L bioreactors (Xing et al., 2017).

In the present work, a modeling tool package was applied to design scale-dependent operating parameters for facility fit during the process scale-up from the laboratory to manufacturing. The CFD model was applied to assess homogeneous distribution in the bioreactor and provide or verify the kLa required for other models. The oxygen demand, GEV, and CO2 stripping models were then applied to screen agitation speeds and gas flow rates needed for sufficient oxygen supply, CO2 control, and mitigating risks of cellular damage from shear stress, foaming, and potential fire hazards of the bioreactor outlet gas. The design condition was operated and tuned using manufacturing development runs. Overall, the application of modeling tools in manufacturing facility fit aims to seamlessly scale-up and minimize the number of at-scale development runs needed for cell culture process commercialization.