Selecting a platform filter for high concentration mAbs

A collaboration between Fujifilm Diosynth Biotechnologies and Cytiva

This application note stems from our collaboration with Fujifilm Diosynth Biotechnologies as they continually look to optimize and refine formulation and filling.

Fujifilm Diosynth Biotechnologies, a CDMO with expertise in the formulation and final filling stages of drug product manufacture, chose to evaluate our Supor™ Prime filter with challenging simulant solutions that were both highly viscous and liable to filter fouling. The goal was to simulate filtration steps during a typical biopharmaceutical formulation and filling process that includes bioburden filtration post-formulation and a final, redundant sterile filtration closer to the point of filling.

Fujifilm Diosynth Biotechnologies’ experimental data confirmed that Supor™ Prime sterilizing grade filters had twice the throughput and average flux compared to PVDF filters, making them an excellent candidate for use in both the bioburden reduction steps and the final filtration stages of highly viscous protein solutions.

Introduction

For multiproduct manufacturing and filling facilities, the application of a filtration platform technology can simplify process development and control process risk. As process trends change, these platform choices benefit from review to accommodate the changing filtration challenge associated with new modalities and new formulations.

The growth in high concentration mAbs is one such trend. High concentration mAbs, typically >100 g/L and their associated product attributes such as higher viscosity (up to 50 cP), result in very different filtration challenges than lower concentration formulations. This can lead to oversized filters being used, which has a direct impact on cost of goods and can increase product loss due to greater non-recoverable volumes.

Fig 1. Typical sequence of filtration steps for bulk drug product and final sterile filtrations.

Considerations when selecting sterilizing-grade filters used in high-viscosity applications

Typically, purified drug substance is brought from downstream processing for final formulation, where a first, bulk bioburden filtration of formulated drug product occurs, followed by a final sterilization step that typically incorporates redundant filtration (Fig 1).

Post formulation bulk filtration aims to reduce potential bioburden and particle load and subsequently minimize fouling in final sterile filtration. When sizing filters for use in high-viscosity applications at this stage in a process, consideration should be given to filtration throughput, and the selected filter should be characterized by a high resilience to blocking.

Final drug product is transferred to a filling line, where a redundant, double sterile filtration often occurs. Due to the bioburden reduction step, final fill solution is virtually free of aggregates, so the filter size is determined on the flow rate required for the filling process.

In addition to filter throughput and flow characteristics, other factors to consider when choosing the right filter for formulation and filling are:

• A low extractables and leachables profile to mitigate potential toxicity administered to the patient.
• A low binding active ingredient and excipients.
• Operational conditions such as pre-use, post-sterilization integrity testing (PUPSIT).
• In process validation of bacterial retention.

Purpose of this study

In this study, Fujifilm Diosynth Biotechnologies compared the performance of Supor™ Prime, Fluorodyne™ EX EDF, and Fluorodyne™ II DFL filters in the sterile filtration of protein solutions with varying viscosity and fouling characteristics.

The goal was to determine which filter was most effective in terms of throughput and flux in a typical biopharmaceutical formulation and filling process.

Materials and methods

Test filters

Three Cytiva sterilizing-grade, 0.2 µm filters were tested during this study (Table 1).

Table 1. Filters used in this study

25 mm membrane discs were installed in single-use Whatman™ Swin-Lok filter holders, giving an effective filtration area (EFA) of 3.7 cm².

Test solutions

The use of high sucrose solutions is a common approach to generating high viscosity solutions (1). Due to high costs and relative scarcity of drug substance, bovine serum albumin (BSA) and sucrose simulant solutions were used.

Solutions of 100 g/L BSA with varying sucrose concentrations were formulated with 20 mM L-histidine buffer pH 6.0 to create simulants covering a range of different viscosities (Table 2).

Table 2. Preparation and characterization of simulant solutions

*Measured after filtration

Test method

Two consecutive filtrations were performed at constant pressure to simulate bulk filtration and final sterile filtration of drug product. At constant pressure, the flow rate decreases with increasing load or filtration throughput due to membrane fouling or blocking. The degree of blockage will vary depending upon the volume filtered, and the amount of particulates and/or aggregates of the fluid being filtered.

Each test fluid was kept in a pressure vessel at a constant pressure of 0.5 bar (7.3 psi, 0.05 MPa). The pressure vessel was then connected to test filters using platinum-cured silicone tubing. Filter membranes were fully wetted with water prior to product filtration. Filtrates were collected in beakers on scales. Pressure monitoring and scale data collection were performed with a PendoTECH NFF system. Data from scales was collected at fixed intervals of 10 s. Results are given in volume units after correction with solution density. Filtration tests were conducted at a temperature of 20°C.

Bioburden filtration

Test solutions were used as is after preparation for the first bioburden filtration. Filter throughput was measured by assessing flux decay during the experiment. The standard pore blocking model (V_max) was applied to facilitate data comparison and filter sizing, as well as extrapolation of expected filter capacity at 90% flux decay (V₉₀).

Final filtration

The effluent pool from the bioburden filtration step for each filter type was utilized with the identical filter type for the final filtration test, simulating a filter train in which the filter type used in bioburden filtration was the same as the filter type used in the final filtration.

Flux decay in this second filtration was negligible or too small to adapt the standard pore blocking model. Instead, the average flux for each filter and simulant fluid was measured in experiments lasting for at least 15 min.

Results

Impact of filtration in drug formulation (first/bioburden filtration)

This experiment aimed to simulate initial bioburden filtration of the formulated product in which the fluid is more likely to foul the filter and so concentrates on filter throughput. This fouling, coupled with the process volumes and desired process times, typically dictates filter selection and sizing.

Figure 2 compares filter throughput using V90 with three different filters (Supor™ Prime, Fluorodyne™ EX EDF and Fluorodyne™ II DFL) at viscosities ranging from 2 to 140 cP.

Fig 2. Throughput in bulk filtration of high-concentration protein solutions with varying viscosities.

Table 3 shows the EFA required to filter a 200 L batch, for Supor™ Prime and Fluorodyne™ EX EDF.

Table 3. Filtration area in m² required for 200 L of BSA simulant solution

Different degrees of fouling were observed with the BSA-sucrose simulant solutions, but in all instances Supor™ Prime filters showed a significantly higher capacity than Fluorodyne™ EX EDF. The Fluorodyne™ II DFL filter showed the fastest filter fouling, indicating that traditional PVDF filters do not provide optimal performance for more challenging applications in which aggregates and particulates are higher.

Impact of filtration in final fill (second/sterile filtration)

This experiment aimed to simulate an initial final filtration step of the formulated product in which the fluid is more likely to be low fouling and is focused on filter flux. This filter flux, coupled with the process volumes and desired process times, are typically used for filter selection and sizing.

Table 4 shows average flux for all three filter types (Supor™ Prime, Fluorodyne™ EX EDF and Fluorodyne™ II DFL) at viscosities from 2 to 140 cP and the calculated time required to filter a 200 L batch. No appreciable flux decay was observed, hence average fluxes are given.

Table 4. Average flux and required time to filter a 200 L batch per m² of filter area.

Figure 3 shows the relative average flux of Supor™ Prime and Fluorodyne™ EX EDF when compared to Fluorodyne™ II DFL at varying viscosities.

Fig 3. Flux increase relative to Fluorodyne™ II DFL (taken as a baseline) for the different viscosity solutions.

Supor™ Prime shows the highest average flux in all cases and presents itself as a good option for filtration in low fouling, high viscosity applications requiring a high flux.

Effect of viscosity on flux

Data from the second filtration was also used to assess the effect of viscosity on flux.

Figure 4 shows the dependency of flux with increasing viscosity for Supor™ Prime, Fluorodyne™ EX EDF and Fluorodyne™ II DFL.

Fig 4. Effect of fluid viscosity in filtrate flux.

The relationship between flux and viscosity was found to be nonlinear, unlike what happens with filtration of non-fouling fluids. It is therefore crucial to test filter membranes in lab scale using formulated drug product to properly evaluate expected flow characteristics at large scale.

Discussion and conclusion

The process simulants used during testing confirmed a significant performance difference between the tested filters in both the bioburden and final filtration steps.

Bioburden filtration

PVDF-based filters showed a low throughput compared to the other two candidates making them a less-than-optimal choice for high fouling feeds. Both Supor™ Prime and Fluorodyne™ EX EDF showed good throughput for high fouling feeds making them more effective choices. However, Supor™ Prime filters showed up to two-fold throughput increase when compared to Fluorodyne™ EX EDF, demonstrating that Supor™ Prime filters are an ideal candidate for high fouling bioburden reduction steps.

Final sterile filtration

Supor™ Prime filters provide a consistently higher average flux (at least two-fold) when compared to a PVDF membrane, regardless of the viscosity of the fluid. This offers the opportunity for a reduction in filter size while still supporting the required minimum fill rate.

Final considerations

The performance advantages of Supor™ Prime filters combined with the higher filtration area in Kleenpak™ Spectrum capsules, lead to a significant reduction in the size of the filter capsule and associated single use systems. This provides opportunities for a reduction in the cost of goods and for improved product recovery due to smaller hold-up volumes in filtration assemblies.

Traditionally many pharmaceutical manufacturers have tended to select PVDF-based filters for formulation and filling applications due to their favorable leachable and extractables profile and their low adsorption properties. However, reducing both filter size and product contact time is also a key consideration in this equation.

Filters currently used for low concentration formulation platforms may not be the ideal choice for high-concentration or high-viscosity mAb formulations. This work demonstrates that Supor™ Prime filters provide a high throughout for the filtration of high-fouling fluids. This enables the use of smaller filters and provides faster process times, making them an excellent candidate when considering a platform filter for the filtration of high concentration mAb product.

Reference

1) Telis VRN, Telis-Romero J, Mazzotti HB, Gabas AL. (2007). Viscosity of aqueous carbohydrate solutions at different temperatures and concentrations. Int. J. Food Prop. 2007;10:185-195. doi.org/10.1080/10942910600673636

CY48644