Achievement of a robust and scalable cell retention device remains a challenge in perfusion systems. Of the two filtration systems commonly used, tangential flow filtration (TFF) systems often have an inferior product sieving profile compared to alternating tangential flow filtration (ATF) systems, which is typically attributed to the ATF's unique alternating flow. Here, we demonstrate that observed performance differences between the two systems are a function of cell lysis and not the alternating flow as previously thought. The peristaltic pump used in typical TFF perfusion systems is shown to be the single major contributor to shear stress and cell lysis. Replacing the peristaltic pump with a low shear centrifugal pump brought cell growth, cell lysis, particle concentration, and product sieving in a TFF perfusion system to levels comparable with that of an ATF. These results provide a correlation where poor product sieving can be partially explained by high shear in cell retention systems and demonstrate that low shear TFF systems are a feasible alternative to ATF systems.

Perfusion medium was successfully developed based on our fed-batch platform basal and feed media. A systematic development approach was undertaken by first optimizing the ratios of fed-batch basal and feed media followed by targeted removal of unnecessary and redundant components. With this reduction in components, the medium could then be further concentrated by 2× to increase medium depth. The medium osmolality was also optimized where we found ∼360 mOsm/kg was desirable resulting in a residual culture osmolality of ∼300 mOsm/kg for our cell lines. Further building on this, the amino acids Q, E, N, and D were rebalanced to reduce lactate and ammonium levels, and increase the cell-specific productivity without compromising on cell viability while leaving viable cell density largely unaffected. Further modifications were also made by increasing certain important vitamin and lipid concentrations, while eliminating other unnecessary vitamins. Overall, an effective perfusion medium was developed with all components remaining in the formulation understood to be important and their concentrations increased to improve medium depth. The critical cell-specific perfusion rate using this medium was then established for a cell line of interest to be 0.075 nL/cell-day yielding 1.2 g/L-day at steady state. This perfusion process was then successfully scaled up to a 100 L single-use bioreactor with an ATF6 demonstrating similar performance as a 2 L bioreactor with an ATF2. Large volume handling challenges in our fed-batch facility were overcome by developing a liquid medium version of the powder medium product contained in custom totes for plug-and-play use with the bioreactor. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:891-901, 2017.

Tangential flow filtration (TFF) and alternating tangential flow (ATF) filtration technologies using hollow fiber membranes are commonly utilized in perfusion cell culture for the production of monoclonal antibodies; however, product retention remains a known and common problem with these systems. To address this issue, commercially available hollow fibers ranging from several hundred kilo-Daltons (kDa) to 0.65 μm in nominal pore size were tested and were all demonstrated to undergo moderate to severe product retention. Further investigation revealed accumulation of particles in the same size range (approximately 20-200 nm) as the pores. Based on the assumption that these particles contribute to product retention and membrane plugging, a hollow fiber with an unconventionally larger pore size was subsequently identified and demonstrated to drastically reduce product retention with no impact to cell clarification, Furthermore, these hollow fibers demonstrated surprisingly high membrane capacities, making them an attractive solution to the problem of product retention in perfusion reactors.

We designed, built or 3D printed, and screened tubular reactors that minimize axial dispersion to serve as incubation chambers for continuous virus inactivation of biological products. Empirical residence time distribution data were used to derive each tubular design's volume equivalent to a theoretical plate (VETP) values at a various process flow rates. One design, the Jig in a Box (JIB), yielded the lowest VETP, indicating optimal radial mixing and minimal axial dispersion. A minimum residence time (MRT) approach was employed, where the MRT is the minimum time the product spends in the tubular reactor. This incubation time is typically 60 minutes in a batch process. We provide recommendations for combinations of flow rates and device dimensions for operation of the JIB connected in series that will meet a 60-min MRT. The results show that under a wide range of flow rates and corresponding volumes, it takes 75 ± 3 min for 99% of the product to exit the reactor while meeting the 60-min MRT criterion and fulfilling the constraint of keeping a differential pressure drop under 5 psi. Under these conditions, the VETP increases slightly from 3 to 5 mL though the number of theoretical plates stays constant at about 1326 ± 88. We also demonstrated that the final design volume was only 6% ± 1% larger than the ideal plug flow volume. Using such a device would enable continuous viral inactivation in a truly continuous process or in the effluent of a batch chromatography column. Viral inactivation studies would be required to validate such a design. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:954-965, 2017.

Insufficient mixing in laminar flow reactors due to diffusion-dominated flow limits their use in applications where narrow residence time distribution (RTD) is required. The aim of this study was to design and characterize a laminar flow (Re 187.7-375.5) tubular reactor for low pH viral inactivation with enhanced radial mixing via the incorporation of curvature and flow inversions. Toward this aim, the reactor described here, Jig in a Box (JIB), was designed with a flow path consisting of alternating 270° turns. The design was optimized by considering the strength of secondary flows characterized by the Dean No., the corresponding secondary flow development length, and the reactor turn lengths. Comprehensive CFD analysis of the reactor centerline velocity profile, cross-sectional velocity, and secondary flow streamlines confirmed enhanced radial mixing due to secondary flows and changes in flow direction. For initial CFD and experimental studies the reactor was limited to a 16.43 m length. Pulse tracer studies for the reactor were computationally simulated and experimentally generated to determine the RTD, RTD variance, and minimum residence time for the tracer fluid elements leaving the reactor, as well as to validate the computational model. The reactor was scaled length wise to increase incubation time and it was observed that as the reactor length increases the RTD variance increases linearly and the dimensionless RTD profile becomes more symmetrical and tighter about the mean residence time.