Robert van Reis

Membrane separations in biotechnology.

Membranes have always been an integral part of biotechnology processes. The sterile filtration of fermentation media, purification buffers, and protein product pools is standard practice in industry. Microfiltration is also used extensively for medium exchange and harvest. Ultrafiltration can be found in virtually every biotechnology process. A significant number of mammalian cell processes use filtration as an integral part of the overall strategy for viral clearance. Depth filters have also seen widespread use for the clarification of both mammalian and bacterial feed streams. Improvements in membrane technology are now focused on high-resolution applications, including improved protein-virus separation, protein purification by high-performance tangential flow filtration and enhanced membrane chromatography. These developments will allow membranes to play an important role in the evolution of the next generation of biotechnology processes.

High performance tangential flow filtration

Conventional tangential flow filtration (TFF) has traditionally been limited to separation of solutes that differ by about ten-fold in size. Wide pore-size distributions, membrane fouling, and concentration polarization phenomena have commonly been cited as reasons for this limitation. The use of TFF in the biotechnology industry has therefore been restricted to cell-protein, virus-protein, and protein-buffer separations. A multi-disciplinary team with industrial and academic members was formed to overcome these limitations and enable protein-protein separations using High Performance TFF (HPTFF) systems. Pore-size distributions have been improved with the development of new membrane formulation and casting techniques. Membrane fouling has been controlled by operating in the transmembrane pressure-dependent regime of the filtrate flux curve and by carefully controlling fluid dynamic start-up conditions. Concentration polarization was exploited to enhance, rather than limit, the resolution of solutes. Concentration polarization has also been controlled by operating a co-current filtrate stream that maintains transmembrane pressure constant along the length of the TFF module. High yields and purification factors were obtained even with small differences in protein sieving. IgG-BSA and BSA monomer-oligomer mixtures have successfully been separated with these systems. HPTFF technology provides a competitive purification tool to complement chromatographic processing of proteins.

Optimization diagram for membrane separations

A novel methodology has been developed which enables optimization of membrane separations. In multi-component separation processes, sieving coefficients for the individual solutes, defined as the ratio of the filtrate and feed concentrations, tend to reach optimum values under different process conditions. It is not possible to determine a priori the pair of sieving coefficients which will give the best combination of product yield and purification for a given application. A purification factor-yield diagram for such an optimization has been developed which utilizes a family of curves representing two dimensionless numbers plotted on yield versus purification-factor coordinates. Analysis can be performed with knowledge of only three experimental variables: the filtrate flux and the two solute sieving coefficients. Complete optimization of membrane processes can be achieved by combining these variables with membrane area, process time, and retentate-volume constraints. The methodology should be applicable to ultrafiltration, microfiltration, and high-performance tangential flow (selective) filtration processes.

Linear scale ultrafiltration

Tangential flow filtration has traditionally been scaled up by maintaining constant the filtrate volume to membrane surface area ratio, membrane material and pore size, channel height, flow path geometry and retentate and filtrate pressures. Channel width and the number of channels have been increased to provide increased membrane area. Several other parameters, however, have not been maintained constant. A new comprehensive methodology for implementation of linear scale up and scale down of tangential flow filtration processes has been developed. Predictable scale up can only be achieved by maintaining fluid dynamic parameters which are independent of scale. Fluid dynamics are controlled by operating parameters (feed flow rate, retentate pressure, fed batch ratio and temperature), geometry (channel length, height, turbulence promoter and entrance/exit design), materials (membrane, turbulence promoter, and encapsulant compression), and system geometry (flow distribution). Cassette manufacturing procedures and tolerances also play a significant role in achieving scale independent performance. Extensive development work in the aforementioned areas has resulted in the successful implementation of linear scale up of ultrafiltration processes for recovery of human recombinant DNA derived pharmaceuticals. A 400-fold linear scale up has been achieved without intermediate pilot scale tests. Scale independent performance has a direct impact on process yield, protein quality and product economics and is therefore particularly important in the biotechnology industry. (c) 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 55: 737-746, 1997.

Ion exchange chromatography of monoclonal antibodies: Effect of resin ligand density on dynamic binding capacity

Dynamic binding capacity (DBC) of a monoclonal antibody on agarose based strong cation exchange resins is determined as a function of resin ligand density, apparent pore size of the base matrix, and protein charge. The maximum DBC is found to be unaffected by resin ligand density, apparent pore size, or protein charge within the tested range. The critical conductivity (conductivity at maximum DBC) is seen to vary with ligand density. It is hypothesized that the maximum DBC is determined by the effective size of the proteins and the proximity to which they can approach one another. Once a certain minimum resin ligand density is supplied, additional ligand is not beneficial in terms of resin capacity. Additional ligand can provide flexibility in designing ion exchange resins for a particular application as the critical conductivity could be matched to the feedstock conductivity and it may also affect the selectivity.

Exploration of overloaded cation exchange chromatography for monoclonal antibody purification

Cation exchange chromatography using conventional resins, having either diffusive or perfusive flow paths, operated in bind-elute mode has been commonly employed in monoclonal antibody (MAb) purification processes. In this study, the performance of diffusive and perfusive cation exchange resins (SP-Sepharose FF (SPSFF) and Poros 50HS) and a convective cation exchange membrane (Mustang S) and monolith (SO(3) Monolith) were compared. All matrices were utilized in an isocratic state under typical binding conditions with an antibody load of up to 1000 g/L of chromatographic matrix. The dynamic binding capacity of the cation exchange resins is typically below 100 g/L resin, so they were loaded beyond the point of anticipated MAb break through. All of the matrices performed similarly in that they effectively retained host cell protein and DNA during the loading and wash steps, while antibody flowed through each matrix after its dynamic binding capacity was reached. The matrices differed, though, in that conventional diffusive and perfusive chromatographic resins (SPSFF and Poros 50HS) demonstrated a higher binding capacity for high molecular weight species (HMW) than convective flow matrices (membrane and monolith); Poros 50HS displayed the highest HMW binding capacity. Further exploration of the conventional chromatographic resins in an isocratic overloaded mode demonstrated that the impurity binding capacity was well maintained on Poros 50HS, but not on SPSFF, when the operating flow rate was as high as 36 column volumes per hour. Host cell protein and HMW removal by Poros 50HS was affected by altering the loading conductivity. A higher percentage of host cell protein removal was achieved at a low conductivity of 3 mS/cm. HMW binding capacity was optimized at 5 mS/cm. Our data from runs on Poros 50HS resin also showed that leached protein A and cell culture additive such as gentamicin were able to be removed under the isocratic overloaded condition. Lastly, a MAb purification process employing protein A affinity chromatography, isocratic overloaded cation exchange chromatography using Poros 50HS and anion exchange chromatography using QSFF in flow through mode was compared with the MAbs commercial manufacturing process, which consisted of protein A affinity chromatography, cation exchange chromatography using SPSFF in bind-elute mode and anion exchange chromatography using QSFF in flow through mode. Comparable step yield and impurity clearance were obtained by the two processes.

Use of the steric mass action model in ion-exchange chromatographic process development

Despite the preponderance of models in the literature, chromatographic process development in industrial protein purification has traditionally been based on heuristic knowledge and expertise. In this work, we explore the feasibility of using a mathematical model to guide process development and optimization. The development of an anion-exchange step to separate an antibody from its dimer is used as a paradigm to test this approach to process development. In the approach involving models, we show that the work required may be reduced to the task of determining conditions for adequate selectivity. Once these conditions are obtained, the steric mass action formalism is used to predict the preparative experimental results. Our results indicate that this model is able to accurately predict experimental results under high loadings with minimal parameter estimation. Finally, we identify ways in which such models can be used to increase productivity and process robustness.

Application of high‐performance tangential flow filtration (HPTFF) to the purification of a human pharmaceutical antibody fragment expressed in Escherichia coli

High‐performance tangential flow filtration (HPTFF) is shown to successfully enable concentration, purification and formulation in a single unit operation. This is illustrated with feedstreams comprising recombinant proteins expressed in Escherichia coli (E. coli). Using positively charged cellulosic membranes of 100 kDa molecular weight cut‐off and operating under a selected range of buffer pH and ionic strength at a filtrate flux of 100 L m−2 h−1, a 10‐fold removal of E. coli host cell proteins (HCP) was obtained with an overall process yield of 98%. The HPTFF performance was shown to be robust and reproducible. In addition, the novel charged membrane was regenerated and re‐used seven times without loss of selectivity or throughput. When compared with a conventional purification scheme, the proposed process results in the elimination of one chromatographic step, a 12% yield improvement and a significant reduction in purification cost of goods. Biotechnol. Bioeng. 2008;100: 964–974.

Surface extenders and an optimal pore size promote high dynamic binding capacities of antibodies on cation exchange resins.

Increased recombinant protein expression yields and a large installed base of manufacturing facilities designed for smaller bulk sizes has led to the need for high capacity chromatographic resins. This work explores the impact of three pore sizes (with dextran distribution coefficients of 0.4, 0.53, and 0.64), dextran surface extender concentration (11-20mg/mL), and ligand density (77-138 micromol H+/mL resin) of cation exchange resins on the dynamic binding capacity of a therapeutic antibody. An intermediate optimal pore size was identified from three pore sizes examined. Increasing ligand density was shown to increase the critical ionic strength, while increasing dextran content increased dynamic binding capacity mainly at the optimal pore size and lower conductivities. Dynamic binding capacity as high as 200mg/mL was obtained at the optimum pore size and dextran content.

Modeling electrostatic exclusion effects during ion exchange chromatography of monoclonal antibodies

Recent experimental studies have shown a reduction in dynamic‐ binding capacity for both monoclonal antibodies and antigen‐binding fragments at very low conductivity, conditions that should generate the greatest electrostatic attraction. This behavior has been attributed to the steric and electrostatic exclusion of the charged protein from the entrance of the resin pores. This manuscript presents a quantitative mathematical description of this phenomenon. The protein partition coefficient was evaluated using models for the partitioning of a charged sphere into a charged cylindrical pore, with the pore size distribution evaluated by inverse size exclusion chromatography. The results were in very good agreement with experimental data for batch protein uptake and dynamic‐binding capacity over a range of pH and conductivity. This theoretical framework provides important insights into the behavior of ion exchange chromatography for protein purification. Biotechnol. Bioeng. 2009;102: 1131–1140.