Filter Design and Validation
Figure 1a: Hollow fiber cross section. (All figures courtesy of the authors.)
To overcome current limitations in virus filtration—and as a step toward the development of common standards—a high-performance,
hollow-fiber parvovirus filter has been developed, which demonstrates robust retention at high transmembrane pressures. The
unique structure of the membrane and its chemically modified surface address many of the limitations of current retentive
The Virosart HF filter (Sartorius-Stedim) features a surface-modified asymmetric polyethersulfone (PES) hollow-fiber membrane
optimized for the manufacture of monoclonal antibodies (see Figure 1a). The membrane is characterized by a funnel-like pore size gradient designed to achieve the robust retention of parvoviruses
under challenging conditions (such as high blockage or pressure release) without impeding the efficient transfer of high-molecular-weight
proteins such as monoclonal antibodies. The membrane is surface-modified with a hydrogel-forming, low-binding polymer, to
reduce the adsorption of soluble proteins and protein aggregates. The pore size gradient and the hydrogel are unique aspects
of the membrane that contribute to its high performance. The hollow fibers can be packed densely into modules ranging in capacity
from 5 cm² to 2.4 m², the latter presented as a presterilized 10-in single-use device (see Figure 1b). The capacity of the
filter can be extended by combining it with the Virosart MAX adsorptive pre-filter, featuring an optimized polyamide microfiltration
flat-sheet membrane in a homogeneous triple-layer configuration, with a nominal pore size of 0.1 µm.
Figure 1b: Comparison of a 0.8/2.4-m² process module and a 5-cm² laboratory module with vent filter for contained flushing.
Validation of Virosart HF has proven the consistent performance of the product family. Figure 2 shows selected validation data (permeability of 2.4-m² process modules) for illustration. However, measures have been taken
to ensure future product quality from lot-to-lot. Validated in-process as well as release tests are performed during the manufacture
and release of Virosart HF membranes and modules according to pre-defined sampling plans to measure and monitor critical performance
attributes of all product components.
Figure 2: Water flow rate in Virosart HF capsules taken from three individual lots of the 2.4-m² module. Each capsule lot
was built from a different membrane lot. The average permeability is 170 liter/m²h bar.
Membrane testing includes in-process and lot release tests. Membrane performance release tests are executed on laboratory
modules that have experienced the same manufacturing steps as laboratory or process modules that would be shipped to customers.
Virosart HF modules are in-house integrity tested by air-diffusion as well as gamma irradiated. Three membrane release tests—bacteriophage
PP7 retention in buffer, bacteriophage PP7 retention in human IgG (grab sample at 75% flux decay), and water permeability—are
consequently performed on lab modules, which have also been flushed with water, dried, and then exposed to gamma irradiation.
Protein filtration capacity is monitored while PP7 retention in buffered human IgG solution is determined. These release tests
ensure that membrane performance items meet expected and validated levels.
In addition, Virosart HF modules are released based on a 100% inspection scheme. Water flow rate and integrity of each module
is tested prior to shipment. Integrity testing is based on an air-diffusion test at 4.5 bar and modules subsequently released
based on a correlation between diffusive flow rate and PP7 retention.
The performance of the Virosart HF module family was tested for scalability using the same batch of a buffered human IgG at
2 g/L. Three different 5-cm² laboratory modules and one 0.8-m² process module were challenged at 2 bar differential pressure
until 95% flux decay was achieved (see Figure 3). The volume vs. time filtration data for the three laboratory modules was averaged and compared to the corresponding data
for the process module according to the filtered protein mass per filtration area. Figure 3 confirms that (based on the performance
data gathered using 5-cm² devices) laboratory modules can be scaled-up to larger feed stream volumes and filter areas.
Figure 3: The performance of the Virosart HF module family was tested for scalability using the same batch of a buffered human
lgG model protein stream (highly blocking) until 95% flux decay was achieved.
The retentive capabilities of the Virosart HF filter were tested under worst-case conditions, by challenging with a 10-20-g/L
monoclonal antibody solution (pH 6-7, conductivity 4-8 mS/cm) spiked with 0.5% MMV. Experiments were carried out at constant
flux at 120 L/m²h. To implement worst-case load conditions, the membrane was challenged with more than 5.5-kg antibody/m²
resulting in a permeability decay of more than 70%. Two spike trials and one control trial were conducted for comparison.
Figure 4: Virosart HF laboratory modules were challenged with a 10-20-g/L monoclonal antibody solution (pH 6-7, conductivity
4-8 mS/cm) spiked with 0.5% MMV. Experiments were carried out at constant flux at 120 L/m2h. The membrane was challenged with
up to 7.4-kg antibody/m2 resulting in a permeability decay of more than 70%. A log reduction value of greater than 5 was achieved
in both spike trials (Run B and Run C).
In all three trials, the transmembrane pressure increased over time but the pressure profiles varied slightly from run to
run with the mass throughput ranging from 5.8 to 7.4 kg/m² (see Figure 4). The transmembrane pressure did not increase above 2.7 bar in any of the trials and thus remained below operating pressure.
Breakthrough was observed in one of the spike trials but the other achieved complete retention. The log reduction values for
the pooled permeate were 5.19 and 5.21, respectively (see Table I).
These data show that Virosart HF achieves robust log reduction values of greater than 5 even under challenging conditions,
thus meeting the retentive requirements of a high-performance parvovirus filter with minimal lot-to-lot variability.
Table I: Summary of the filtration and MMV retention data for the control trial (Run A) and the two spike trials (Run B and