Introduction
Virus filtration is widely used as
a necessary virus removal step for biotherapeutics and protein drug
products derived by plasma fractionation (Besnard et al., 2016; Grein et
al., 2014; Miesegaes et al., 2010; Strauss et al., 2017). Virus filters
have multi-layered membranes that allow for separating viruses from
protein solutions, which have only a small size difference (Manabe et
al., 1987, 1989). While virus removal from protein solutions is made
possible by the more sensitive separation membrane,
larger sized proteins and
aggregates in particular have been confirmed to cause clogging in virus
filters (Barnard et al., 2014; Bieberbach et al., 2019; Chen & Chen,
2015; Genest et al., 2013). To improve the capacity of the virus
filtration step, processing the feed solution through one of several
commercially available prefilters has been confirmed to be effective in
removing substances that cause clogging and has been widely implemented
(Brown et al., 2010; Genest et al., 2013; Roederstein & Thom, 2013).
Alternatively, column chromatography has also been shown to effectively
remove aggregates (Yigzaw et al., 2009).
Examination of filtration behavior and, in particular, changes in
filtration behavior due to clogging can be used to elucidate the
clogging mechanisms and to extrapolate filtration capacity in well
characterized systems. The theoretical relationship between filtration
volume and filtration throughput for a given feed solution can be fitted
to one of four established clogging models for filter membranes (Grace,
1956): cake filtration, intermediate blocking filtration, standard
blocking filtration and complete blocking filtration. Maximum filtration
volume (Vmax) was theoretically determined using the
standard blocking model based on filtrations conducted with a
microfilter at constant pressure, and the data were then used for
facility scale-up recommendations as reported in Badmington et al.
(1995). Following this same strategy, the standard blocking model has
been applied in nanofiltration to determine Vmax of
virus filters under constant pressure (Bolton et al., 2004). Thus, small
volumes of feed solution samples can be used to effectively evaluate the
filtration volume or filter size that will be used in the process and
additionally quantitatively compare the filterability of process
solutions. Further, there have been attempts to combine several clogging
models into a theoretical equation in order to match complex filtration
behavior results to theoretical values (Ho & Zydney, 2000; Bolton et
al., 2006; Bolton & Apostolidis, 2017).
In this report, monoclonal
antibody (mAb) and plasma IgG each spiked with aggregates prepared from
the respective solution were processed by various chromatography resins
or prefilter and run on a virus filter under constant pressure to
determine the relationship between filtration volume (throughput) and
flow rate (flux). In order to analyze the experimental results,
filtration behavior results for solutions spiked with aggregates as well
as several column chromatography outputs were applied to the clogging
models of cake filtration, intermediate blocking, standard blocking and
complete blocking, the model that best fits the filtration behavior data
was identified, and clogging models were used to extrapolate filtration
behavior beyond the experimental range.