Scale-up analysis of geometrically dissimilar single-use bioreactors

Cell culture scale-up is a challenging task due to the simultaneous change of multiple hydrodynamic process characteristics and their different dependencies on the bioreactor size as well as variation in the requirements of individual cell lines. Conventionally, the volumetric power input is the most common parameter to select the impeller speed for scale-up, however, it is well reported that this approach fails when there are huge differences in bioreactor scales. In this study, different scale-up criteria are evaluated. At first, different hydrodynamic characteristics are assessed using computational fluid dynamics data for four single-use bioreactors, the Mobius® CellReady 3 L, the Xcellerex™ XDR-10, the Xcellerex™ XDR-200, and the Xcellerex™ XDR-2000. On the basis of this numerical data, several potential scale-up criteria such as volumetric power input, impeller tip speed, mixing time, maximum hydrodynamic stress, and average strain rate in the impeller zone are evaluated. Out of all these criteria, the latter is found to be most appropriate, and the successful scale-up from 3 to 10 L bioreactor and to 200 L bioreactor is confirmed with cell culture experiments using Chinese Hamster Ovary cell cultivation.

Impact of Post-Freeze Annealing on Shrinkage of Sucrose and Trehalose Lyophilisates

Freeze-drying of pharmaceuticals produces lyophilisates with properties that depend on both the formulation and the process. Characterisation of the lyophilisate in terms of appearance is necessary not only to produce a visually appealing product, but also to gain insight into the freeze-drying process. The present study investigates the impact of post-freeze annealing on the volume of lyophilisates. For this purpose, sucrose and trehalose solutions were freeze-dried with different annealing conditions and the resulting lyophilisates were analysed with a 3D structured light scanner. The external structure of the lyophilisates was found to be dependent on the bulk materials as well as the choice of vials, while the volume was influenced by the annealing time and temperature. Additionally, differential scanning calorimetry was used to determine glass transition temperatures of frozen samples. As a novelty, the volumes of the lyophilisates and their corresponding glass transition temperatures were compared. This resulted in a correlation supporting the theory that the shrinkage of lyophilisates depends on the amount of residual water in the freeze-concentrated amorphous phase before drying. Understanding the volume change of lyophilisates, in combination with material properties such as glass transition temperature, forms the basis for relating physicochemical properties to process parameters in lyophilisation.

Quantitative Analysis of Glassy State Relaxation and Ostwald Ripening during Annealing Using Freeze-Drying Microscopy

Supercooling during the freezing of pharmaceutical solutions often leads to suboptimal freeze-drying results, such as long primary drying times or a collapse in the cake structure. Thermal treatment of the frozen solution, known as annealing, can improve those issues by influencing properties such as the pore size and collapse temperature of the lyophilisate. In this study we aimed to show that annealing causes a rearrangement of water molecules between ice crystals, as well as between the freeze-concentrated amorphous matrix and the crystalline ice phase in a frozen binary aqueous solution. Ice crystal sizes, as well as volume fractions of the crystalline and amorphous phases of 10% (w/w) sucrose and trehalose solutions, were quantified after annealing using freeze-drying microscopy and image labelling. Depending on the annealing time and temperature, the amorphous phase was shown to decrease its volume due to the crystallisation of vitreous water (i.e., glassy state relaxation) while the crystalline phase was undergoing coarsening (i.e., Ostwald ripening). These results allow, for the first time, a quantitative comparison of the two phenomena. It was demonstrated that glassy state relaxation and Ostwald ripening, although occurring simultaneously, are distinct processes that follow different kinetics.

CFD-Based and Experimental Hydrodynamic Characterization of the Single-Use Bioreactor XcellerexTM XDR-10

Understanding the hydrodynamic conditions in bioreactors is of utmost importance for the selection of operating conditions during cell culture process development. In the present study, the two-phase flow in the lab-scale single-use bioreactor XcellerexTM XDR-10 is characterized for working volumes from 4.5 L to 10 L, impeller speeds from 40 rpm to 360 rpm, and sparging with two different microporous spargers at rates from 0.02 L min-1 to 0.5 L min-1. The numerical simulations are performed with the one-way coupled Euler-Lagrange and the Euler-Euler models. The results of the agitated liquid height, the mixing time, and the volumetric oxygen mass transfer coefficient are compared to experiments. For the unbaffled XDR-10, strong surface vortex formation is found for the maximum impeller speed. To support the selection of suitable impeller speeds for cell cultivation, the surface vortex formation, the average turbulence energy dissipation rate, the hydrodynamic stress, and the mixing time are analyzed and discussed. Surface vortex formation is observed for the maximum impeller speed. Mixing times are below 30 s across all conditions, and volumetric oxygen mass transfer coefficients of up to 22.1 h-1 are found. The XDR-10 provides hydrodynamic conditions which are well suited for the cultivation of animal cells, despite the unusual design of a single bottom-mounted impeller and an unbaffled cultivation bioreactor.

Influence of particle size and blender size on blending performance of bi-component granular mixing: A DEM and experimental study

The effect of particle size enlargement and blender geometry down-scaling on the blend uniformity (BU) was evaluated by Discrete Element Method (DEM) to predict the blending performance of a binary granular mixture. Three 10 kg blending experiments differentiated by the physical properties specifically particle size were performed as reference for DEM simulations. The segregation behavior observed during the diffusion blending was common for all blends, while the sample BU, i.e., standard deviation of active ingredient content % was different among the three blends reflecting segregation due to the particle size differences between the components. Quantitative prediction of the sample BU probability density distribution in reality based on the DEM simulation results was successfully demonstrated. The average root mean square error normalized by the mean of the mean sample BU in the blends was 0.228. Beside the ratio of blender container to particle size, total number of particles in the blender and the number of particles in a sample were confirmed critical for the blending performance. These in-silico experiments through DEM simulations would help in setting a design space with respect to the particle size and in a broader sense with respect to the physical properties in general.