Quantitative interpretation of protein breakthrough curves in small-scale depth filter modules for bioprocessing

Highlights

• Protein binding behavior evaluated in small-scale Supracap™ 50 depth filters.

• CFD analysis shows presence of four distinct flow paths through capsule.

• Multiple flow paths lead to broad and complex breakthrough curve.

• Computational/experimental results provide insights into performance of depth filters.

Abstract

Depth filters are used throughout downstream bioprocessing, with specialized filter chemistries developed for removal of host cell proteins, DNA, and protein aggregates. These filters often show very broad breakthrough curves, but there have not been any previous analyses of this phenomenon. Protein breakthrough curves were evaluated for Ribonuclease A and α-chymotrypsin in two small-scale depth filter modules using PDH4 media containing diatomaceous earth. Computational fluid dynamics (CFD) simulations were developed, with flow in the porous media described using the modified Navier-Stokes equations incorporating a Brinkman flow term and a Langmuir binding model. The breakthrough curve in the stainless-steel module was much sharper than that in the commercial Supracap™ 50 capsule due to the more uniform axisymmetric flow field. The behavior in the Supracap™ 50 capsule was more complex, with multiple inflection points observed due to the different flow paths and the time-dependent saturation of the two layers of the depth filter media. These results provide important insights into the factors controlling protein binding/breakthrough in commercial depth filter modules.

Introduction

Depth filtration is used extensively in downstream bioprocessing for removal of cells and cell debris in the clarification of cell culture fluid [1,2]. Depth filters can also increase the capacity and performance of other downstream unit operations, e.g., chromatography columns [3] and virus filtration membranes [4], by removing aggregates and other foulants. In addition, appropriately designed depth filters can provide high levels of removal of host cell proteins [5], DNA [6], and viruses [7], significantly reducing the required level of purification that must be obtained in the rest of the downstream process.

A number of studies have examined the adsorptive capacities of depth filter media with different physical characteristics, including different filter aids (e.g., diatomaceous earth or perlite), polymeric fibers (e.g., cellulose or polypropylene), and binders. For example, Singh et al. [8] found dynamic binding capacities for a model protein (bovine serum albumin) at low conductivity (5 mS/cm) ranging from <1 mg/mL for the Zeta Plus 60SP05A to more than 50 mg/mL for the Emphaze ST-AEX, both of which are positively-charged media. Similarly, Charlton et al. [9] found DNA binding capacities in salt-free solutions ranging from <0.05 μg/mm² for the Sartobind Q to more than 1 μg/mm² for the Zeta Plus 90SP.

In addition to the large range of dynamic binding capacities, many depth filters have very broad breakthrough curves. Yigzaw et al. [5] evaluated the concentration of Chinese Hamster Ovary host cell proteins (CHOP) in the effluent from a Millistak mini-cartridge with A1HC media and found 10% breakthrough at a volumetric loading of 30 L/m² while 90% breakthrough didn't occur until more than 150 L/m². A similarly broad breakthrough curve has been reported for DNA, with 10% breakthrough at <40 μg DNA loading and 90% breakthrough at >400 μg for the Zeta Plus 10SP [9]. Because of these broad breakthrough curves, it is very difficult to realize the full binding capacities of these depth filters, particularly if they are sized for breakthrough at around 10% of the feed concentration. Note that the behavior of these depth filters is very different than commercially available membrane adsorbers which typically have very sharp breakthrough curves, with dynamic binding capacities approaching the equilibrium binding capacity due to the minimal mass transfer limitations in these systems [10].

Although there have yet to be any detailed analyses of the breakthrough curves in commercial depth filters, a number of groups have used computational fluid dynamics (CFD) to evaluate the breakthrough behavior in membrane adsorbers. For example, Ghosh et al. [11] developed a CFD model that successfully predicted protein breakthrough curves for membrane chromatography devices from Sartorius with bed volumes ranging from 0.08 to 1200 mL. The large-scale system showed a sharper breakthrough curve than the laboratory device due to the more uniform flow distribution across the membrane. Geerling et al. [12] developed a multi-scale model to optimize the porous structure of depth filtration media, but their analysis was focused entirely on the particle capture behavior without any consideration of the adsorptive removal properties of these depth filters.

The objective of this study was to develop a detailed mathematical description of the protein breakthrough curves for the Pall PDH4 media in a commercial Supracap™ 50 capsule and in a custom-designed stainless-steel holder. The flow distributions in both systems were evaluated using CFD as discussed by Kim et al. [13]. The breakthrough curves were then evaluated assuming Langmuir binding kinetics with the equilibrium binding constants determined from static adsorption experiments with the PDH4 media. The results provide important insights into the factors governing the protein breakthrough curves in depth filtration systems.