Klaus Tarrach

Biomanufacturing of a biopharmaceutical blockbuster product such as a monoclonal antibody or a plasma-derived product is characterized by high production costs. These costs are mainly attributed to the high capital investments into a current good manufacturing practices (CGMP) biotechnology facility, which can exceed $1 billion (1). There is, however, ongoing downward pressure on treatment costs as governments and managed-care groups focus on ways to reduce spending on healthcare. In addition, there is downward pressure on the cost of goods (COGs) from companies marketing such products to enhance revenue streams (2).   

New manufacturing concepts and platforms are therefore required to meet commercial expectations for existing and upcoming new drugs. Product yields must be improved and the process must use manufacturing space and resources more efficiently. Reevaluating existing purification schemes and redefining the process may reduce the COGs by as much as 50% (2).   

The concept of a generic technology platform for the purification of multiple drug candidates enables faster drug development and earlier definition of the commercial process and facilitates the effective utilization of manufacturing space. A key element to consider for any new biopharmaceutical manufacturing platform is the use of disposable equipment (disposables). This practice, together with the requirement for a viral clearance strategy, has led to the development of disposable nanofiltration technology. Using disposable nanofilters within the downstream process has been proven to reduce capital expenditure for housings and reduce labor, cleaning, and validation costs. Disposable nanofilters are capable of turning fixed costs into variable costs with single-use membrane technology. These costs become relevant as cash-out only when the plant is in operation and the nanofiltration step is up and running.  

The data presented in this article outline the challenges of integrating a large-scale SP Sepharose column ("Resolute" 1000 mm DAP/M column, Pall UK, Portsmouth, UK) with a disposable nanofiltration setup, to meet flow expectations of 1000 L/h and batch volumes of 10,000 L. This chromatography and nanofiltration process is used to purify an ovine-derived polyclonal antibody product at Protherics UK  (3). The technology platform for purification of this polyclonal antibody fragment is outlined in Figure 1. 

Figure 1: Purification process overview. (Figure 1 courtesy of the authors.) 

The IgG is captured by an MEP "Hypercel" chromatography gel (Pall UK) in a Resolute 1600 mm DAP/M column (Pall UK) (see Figure 2). The captured IgG is then digested with papain and diafiltered. The resulting Fab is further purified and polished by Q-Sepharose FF (GE Healthcare, Chalfont St. Giles, UK) and SP Sepharose chromatography steps. The purified Fab fragment is then subject to nanofiltration, followed by concentration and drug formulation. 

Figure 2: The authors with a Pall "Resolute" 1600 mm DAP/M column. (Figure 2 courtesy of Protherics UK.) 

For this ovine-based product, the nanofiltration step is the cornerstone of the virus clearance strategy, targeting small nonenveloped viruses. Regulatory expectations are that the nanofiltration step provides robust and efficient removal of small nonenveloped viruses (porcine parvo virus [PPV] is the model virus being used) and achieves a 4-log

When developing a manufacturing platform for this product, disposability was a prime consideration to ensure simplification of the process, reduction of labor, increased flexibility, and elimination of cross-contamination. The 20-nm nanofiltration was introduced at the end of the purification process where the purity of the product is the highest and filter blockage due to contaminants is the lowest. Protein concentration at this stage is 2–6 mg/mL, and pH is 7.0–7.8. 

The technical challenge in combining the SP chromatography step with the 20-nm nanofiltration step comes from matching flow requirements and pressure capabilities of the chromatography process with the flow and capacity requirements of the nanofiltration technology. At the same time, the nanofiltration technology must be used at ultra-large scale because the batch volumes of product at this stage reach 10,000 L. Flow and capacity capabilities of 20-nm nanofiltration technologies currently available vary significantly because of the use of different membrane types and pore geometries, or different filtration modes (dead end versus tangential flow).   

Therefore, initial screening studies were undertaken to select the most appropriate 20-nm technology for this application. Nanofilters from three suppliers were used in the screening process; each meets the technical requirements for a large-scale disposable. The screening of these technologies was based on the following parameters:  

The screening study involved spiking a sample of the material to be nanofiltered with a preparation of PPV, at a virus concentration recommended by each supplier.  Using a lab-scale nanofilter from each supplier, attempts were made to filter a volume equivalent to the required full-scale manufacturing volume of 400 L/m

Table I outlines data from the non-GLP screening study conducted at Catalent Pharma Solutions (formerly Cardinal Health). 

Table I: Results of PPV screening studies at Catalent Pharma Solutions. 

For filter B, a volume of 250 mL was obtained using a "run–spike–run" method in which only the last 50 mL was spiked. All other filtrate volumes were achieved using a "spike–run" method in which the full volume was spiked. The screening studies demonstrated that a 20-nm polyethersulphone (PESU)-based membrane (Sartorius Stedim Biotech, Aubagne, France) provided the best flowrate and capacity and met the virus clearance requirements. Once this nanofilter was selected, more detailed analysis of the virus clearance capability of the Sartorius virus-filter technology ("Virosart CPV") was initiated. 

The first goal was to determine the capability of the filter to retain viruses throughout the filter lifetime. Because it was not feasible to spike the product with viruses at manufacturing scale, the filtration process had to be accurately scaled down to ensure that good laboratory practice (GLP) spiking studies were representative of the full-scale process.   

The first phase of screening studies was carried out using bacteriophage PP7 as a model virus for small nonenveloped viruses. The PP7 bacteriophage assay has previously been described (4), and the use of PP7 as a model virus for mammalian viruses has been shown in previous studies (5, 6). 

The screening studies were performed using "Virosart CPV Minisart" filters at the Virology Department of Sartorius Stedim Biotech GmbH in Germany. Eight runs were performed, each using a 2.1 × 10

/mL spike concentration of PP7 bacteriophage particles. All runs showed the same log reduction, and the results of a typical run are summarized in Table II.  

The loading values are presented in Table II as L/m

  to reflect the volume to be filtered at manufacturing scale (maximum 400 L/m

  of filter area is required to meet the full batch volume). The results demonstrated that a good log reduction could be achieved throughout the filtration process, and this was replicated in all scaled-down PP7 runs.  

As a next step, non-GLP spiking studies using PPV and the same scaled-down model were performed at Catalent Pharma Solutions to verify the PP7 data obtained. As required by regulatory guidance, the scaled down model used during the GLP validation run should reflect process conditions as closely as possible unless a justification to deviate from this can be provided (7, 8).  

Consequently, the volume to be filtered using the scaled-down model is derived from the required filtration volume at process-scale. This volume was initially established by testing the virus filter without a virus spike to determine the flow rate and capacity. It has been reported that virus spike preparations can lead to significant fouling of nanofiltration membranes in spiking studies, and properties of the virus spike itself are an often-overlooked source of contaminants that can affect filter performance (9). Potential contaminants in virus stocks may include serum proteins, host-cell proteins, host-cell DNA, and lipids (10, 11). 

To determine the effect of the virus spike on product filterability, nonpurified and purified spike preparations of PPV were filtered using the same scaled-down model. The purified virus was prepared using a Q-Sepharose chromatography step. 

Figure 3 outlines the filtration characteristics of the Fab product using Virosart CPV, comparing a 1% virus spike of chromatography-purified PPV with a 1% spike of non-purified PPV. A nonspiked control is also shown. The purified PPV samples gave a better flow rate and capacity than nonpurified virus, which demonstrated that contaminants within the viral spike were affecting filter performance. As a result of these studies, a Q-chromatography purified PPV preparation was used for subsequent validation of the nanofiltration step to achieve the required scaled-down volume of product (400 L/m

Figure 3: Filtration characteristics of purified versus nonpurified PPV at 1% spike. (Figure 3 courtesy of the authors.) 

 A series of scaled-down runs was then performed using purified PPV at different concentrations, to determine the viral spike concentration that would give the best log reduction without affecting filter performance. Table III outlines PPV retention data for three runs using different spike concentrations of purified PPV.  

Table III: Non-GLP PPV test results for Virosart CPV. 

All studies used a test pressure of 2 bar (30 psi) and filter lot number 0650773R50Z3/2. For all runs, a scaled-down filtrate volume equivalent to 400 L/m

  was obtained to reflect the full-scale filter capacity requirements. 

All three viral spike concentrations resulted in acceptable viral reduction. A range of 0.5–1% was therefore selected for validation runs to maximize the flow rate and capacity of the filter, and to minimize the amount of virus required. 

To validate the nanofiltration step, GLP runs were performed using the following panel of viruses as model viruses: PPV, murine leukemia virus (MuLV), bovine viral diarrhea virus (BVDV), and reovirus type 3 (Reo 3) virus. The overall GLP spiking study results are outlined in Table IV.  

Table IV: Overall GLP spiking study results for Virosart CPV. 

All studies used a viral spike concentration of 0.5%, a test pressure of 2 bar (30 psi), and filter lot number 0650773R50Z3/2. For all runs, a scaled-down filtrate volume equivalent to 400 L/m

  was obtained, to reflect the full-scale filter capacity requirements. 

GLP studies also included robustness studies using PPV as the model virus. The robustness studies incorporated anticipated worst-case protein concentration based on laboratory data for the SP Sepharose process, and worst-case pressure. Results are outlined in Table V. 

Table V: Results from GLP robustness spiking study for Virosart CPV. 

All studies used filter lot number 0650773R50Z3/2. For all runs, a scaled-down filtrate volume equivalent to 400 L/m

The small-scale studies described above were carried out to support the manufacture of full-scale batches. A large-scale nanofiltration rig capable of handling as much as 5000 L of product was used. The filter rig was designed so that it can be directly connected to the outlet of the chromatography rig. This has enabled the entire setup to be controlled and monitored by GE Healthcare BioProcess equipment and Unicorn control software, and has combined the SP Sepharose and nanofiltration steps into a one-unit operation. 

The filters remain in the rig for integrity testing, which reduces the manual handling requirements and therefore risk of damage to the filters. Integrity test scenarios for such large-scale systems have been presented (12) and include the use of test systems capable of handling multiple filters simultaneously in a cleanroom environment. Such test systems enable the operator to test more than 20 filter elements independent of the pore size in less than 1 h.  

 With respect to the overall process design for the large-scale manufacture of this ovine Fab fragment, the disposable approach is applied wherever possible. Sterilizing-grade filters, prefilters, product collection bags, and the majority of tubing is irradiated and disposable. In addition, all processing solutions are manufactured in single-use disposable bags.    

Design-space studies performed by the process development group at Protherics UK provided the parameters within which the SP Sepharose chromatography step could be operated at process scale. A flow rate of ?1000 L/h was consistently achieved during validation runs, without exceeding the maximum column pressure. This in turn delivered a pressure well within the validated nanofiltration pressure of 30 psi. 

The challenge in handling such a large amount of liquids in a Class 10,000 environment exists in finding disposable filter and bag technologies that will not only meet process flow requirements but also meet cost requirements and provide a realistic level of manual handling. The nanofiltration filtrate was therefore collected into 1000 L bags, each having an in-line sterilizing-grade 0.2 µm filter to ensure bioburden control.  

The integration of large-scale chromatography with nanofiltration can be effectively achieved using disposable technologies wherever possible. The use of disposable tubing and bag assemblies plus the integration of the nanofiltration operation with the chromatography step all help reduce the amount of cleaning validation. 

The direct connection of the nanofiltration set up to the chromatography system offers the benefit of using the same software to control and monitor both operations. With accurate process control, the integration of two technology platforms has given a robust, convenient, and combined process step. 

Articles

The authors examine the challenges of integrating a large-scale chromatography and nanofiltration process for purification of a polyclonal antibody.

Integration of Large-Scale Chromatography with Nanofiltration for an Ovine Polyclonal Product

Monoclonal antibodies and recombinant proteins have increased their importance and gained success as therapeutic agents in the treatment of various diseases.

Biomanufacturing a recombinant product by cell culture using animal cells follows a main route from a biology-driven upstream process to a downstream process strictly guided by technical and biochemical principles.

The biology-driven upstream process is the pacemaker and latest advances in cell culture revealed product titres of up to 5 mg/mL of an recombinant protein. Furthermore, both cell densities and cell viabilities have undergone significant changes, which have brought new challenges for the initial recovery process (IRP) of such a biopharmaceutical product.  

Today's technologies in the IRP of the downstream process require the principal capability to handle cell densities of at least 10 mio cells/mL, product titres of up to 5 mg/mL and cell viabilities of <50%.

Such recovery steps aim to remove cells and cell debris; however, this is now being challenged with cell culture processes increasingly featuring low cell viabilities. Instead, the focus has shifted to the recovery of the target protein out of cell debris such as feed stream, with a significantly increased amount of cell-derived contaminants, such as Chinese hamster ovary protein (CHOP), host cell DNA, lipoproteins and endogenous viruses. DNA levels in direct bioreactor offloads have increased up to 10

  ppm and pose a major challenge to the entire chromatography strategy applied to a biopharmaceutical product.

Figure 1 Principle positioning of multilayer technologies. 

Cellulose-based filtration techniques have been used from early on in the IRP and are now advancing into the spotlight as the contaminant level needs to be tackled earlier to enable better intermediate purification and polishing of the target protein. 

Traditionally, one or more cellulose-based depth filter operations have been used in IRP.

  Depth filter operations can feature either single layer depth filter media or a combination of various depth filter media and/or membranes into one process step.

  Combining two or more depth filter media composites into one process element reduces the filtration steps from two to one. However, driven by the large amount of cells to be removed and cell debris-induced fouling mechanisms, multilayer technologies can act as single-stage operations that can be positioned after each other to tackle the overall challenge. In detail, bioreactor offloads with turbidities in excess of 1000 FNU might require a two-stage filtration each featuring multilayer depth filter configurations (Figure 1). Alternatively, a continuous centrifuge might be considered for the first removal step followed by a multilayer filtration step. The higher the cell density within a cell culture process the more likely it is a centrifuge will have to be positioned as the first recovery step taking into account that ultra large-scale volumes of >10000 L require handling by the IRP. Recent publications address this topic and discuss the use of two different multilayer depth filters in a post centrifuge application.

Figure 2 Filterability data for single layer and double layer depth filter technologies; post bioreactor offload application; 5 mio cells/mL and 75% viability; 11 Î¼m/4 Î¼m and 8 Î¼m/1 mm filters are from Sartorius-Stedim Biotech. 

The depth filtration operation in a post centrifuge application remains challenging as larger amounts of submicron particles are induced by either the cell culture process or the centrifuge process by shear forces or other mechanical stresses applied to the biological fluid.  

Initial recovery is more than just applying filtration or centrifugation techniques to remove cells and cell debris. Bioburden control is another major task and thus 0.2 μm sterile grading filtration is integrated into the process chain, usually as a disposable technique.  

The overall capacity of the sterile grading filtration operation is a direct function of the depth filter capability to reduce the bioreactor offload or post centrifuge turbidity.  

The turbidity in both applications is a function of larger and submicron particles, which must be removed prior to sterile grade filtration.  

Multilayer depth filter technologies offer higher bed heights than single layer technologies of up to 11 mm and, therefore, significantly better control of breakthrough effects of bioreactor-derived contaminants. Thus, the protection of a subsequent 0.2 μm filter is improved and overall process costs are reduced. 

Table 1 Comparison of filter effluent turbidity; post bioreactor offload application; 5 mio cells/mL, 75% viability, 1000 FNU filter inlet turbidity. 

The main goal of using cellulose-based depth filters in the post bioreactor application is to remove cells and cell debris, and allow for a subsequent bioburden reduction filtration. Figures 2–4 present data based on both low cell density (5 mio cells/mL) and high cell density (19 mio cells/mL) bioreactor offloads.  

In Figure 2, it is shown that a double layer depth filter used in a bioreactor offload application provides similar capacity to a single layer filter.  

All double layer depth filters tested provided significantly lower effluent turbidity and thus higher sterile grade filter capacity than any of the single layer depth filters tested.  

The lowest single layer effluent turbidity (8 μm filter) has been measured at 272 FNU, whereas the lowest effluent turbidity for the double layer filter is <5 FNU, which is summarized in Table 1. However, all filters tested showed a protein transmission of >95%.  

Figure 3 Breakthrough curves for single and double layer depth filter technologies; post bioreactor offload application; 5 mio cells/mL and 75% viability; 11 Î¼m/4 Î¼m filter is from Sartorius-Stedim Biotech. 

Figure 3 outlines breakthrough curves for single and double layer technologies used for the low cell density post bioreactor application described previously.  

For all single layer depth filter technologies tested the breakthrough curve starts earlier and ends in a higher final point and, thus, average effluent turbidity is greater than the values obtained for all double layer depth filters tested. 

Table 2 Comparison of depth filter effluent turbidities, taken from Figure 5. 

The data shown in Figure 4 outline decreased capacities for double layer technologies of <100 L/m

 . This is because of very high cell density (19 mio cells/mL) and ultra low cell viability (6%).  

For such applications featuring a cell density of >10 mio cells/mL and a cell viability of <15%, it is suggested to position a centrifuge as the first cell removal and clarification step.  

However, cell viabilities of <15% pose a major challenge to the subsequent purification regime as larger amounts of host cell-derived contaminants are loaded into the feed stream, which will have to be cleared by chromatography unit operations. 

Figure 4 Direct comparison of two double layer depth filter combinations used for the filtration of a high cell density (19 mio cells/mL) cell derived bioreactor offload fluid; 11 Î¼m/4 Î¼m filter is from Sartorius-Stedim Biotech. 

Within the post centrifuge application, the main goal of using cellulose-based depth filters is to remove residual cell debris and ideally other cell-derived contaminants.  

Figures 5–7 present filterability data based on both initial low cell density (3.8 mio cells/mL [Figure 5]) and high cell density (19.5 mio cells/mL [Figures 6 and 7]) bioreactor offloads subject to centrifugation and depth filtration at a tested capacity of 200 L/m

Within the post centrifuge application, the single layer depth filter does not match the capability of the double and multilayer filters with respect to filter effluent turbidity. Thus, protection of a subsequent sterile grading 0.2 μm membrane filter is improved using a double or multilayer depth filter. 

Figure 5 Comparison of filterability data of single layer, double layer and multilayer depth filter technologies in a post centrifuge low cell density (3.8 mio cells/mL) application; all filters are from Sartorius-Stedim Biotech. 

In this application, the post centrifuge inlet turbidity was measured at 128 FNU and the single layer 0.3 μm depth filter tested reduced this turbidity to an average of 59 FNU at a target capacity of 200 L/m

Changing from the single layer 0.3 μm depth filter to a double layer 0.3 μm/0.3 μm depth filter further decreased the turbidity to 9.3 FNU.  

Incorporating a 0.2 μm membrane into the double layer depth filter combination provides the lowest effluent turbidity of 4.5 FNU, which is outlined in Table 2. 

Results from another post centrifuge application with a filter inlet turbidity of 86 FNU (high cell density 19.5 mio cells/mL [Figure 6]) confirm the reduction of the depth filter effluent turbidity, significantly increasing the sterile grade membrane filter capacity by incorporating a membrane into the double layer depth filter unit operation.  

A combination of depth filters with a 0.1 μm membrane provides the lowest filter effluent turbidity value observed of 1.8 FNU. 

Figure 7 outlines breakthrough curves for double and multilayer technologies used for the high cell density post centrifuge application from Figure 6. 

Table 3 Comparison of depth filter effluent turbidities, taken from Figure 6. 

For the multilayer depth filter with a 0.1 μm membrane, the breakthrough curve ends in a lower final point and thus presents a lower average effluent turbidity than the multilayer depth filter with a 0.2 μm membrane and the double layer depth filter without a membrane.  

Figure 6 Direct comparison of double layer and multilayer depth filter combinations in a post centrifuge application (high cell density 19.5 mio cells/mL and 82% viability); all filters except product AXXX are from Sartorious-Stedim Biotech..

For post centrifuge applications, combining depth filter configurations with membranes significantly reduces the turbidity and effectively controls breakthrough effects.  

Determining the breakthrough point is essential as excessive breakthrough observed on any depth filter operation will lead to significant reduction in the subsequent membrane filter unit operation.  

Furthermore, excessive initial differential pressure or flow rates on the depth filter will significantly influence the overall depth filter capacity, as well as the depth filter effluent turbidity.

Therefore, it is important to carefully monitor the pressure differential and filter effluent turbidity during a process optimization run.  

With respect to the process design for the IRP, it is essential to consider all parameters that might significantly influence the process design.  

Cell density and cell viability are crucial and need to be considered as limiting parameters for the post bioreactor application.  