Results
Pressure characterization of TFF, rTFF and HPTFF
Measurement of pressure gradients within the hollow fiber and on the
filtrate side along the filter length for the TFF, rTFF and HPTFF system
was performed with a special characterization setup enabling the
operation of all three systems (Figure 1A ). This setup not only
allowed us to measure the retentate loop and filtrate pressures at the
inlet of the filter module (\(\text{PT}_{R1}\) and \(\text{PT}_{F1}\))
and at the outlet of the filter module (\(\text{PT}_{R2}\) and\(\text{PT}_{F2}\)), but enabled us to get additional measurements of
the filtrate pressure along the filter module length
(\(\text{PT}_{A1-5}\)). A crossflow ramping from 0 – 1500 mL/min to
simulate the TFF system or one of the rTFF phases demonstrated that the
axial pressure drop within the retentate loop increased with increasing
crossflow as seen in the diverging retentate inlet pressure\(\text{PT}_{R1}\) and retentate outlet pressure \(\text{PT}_{R2}\)(Figure 2A ). All remaining pressure sensors on the filtrate
side, irrespective of the crossflow, indicated the average pressure of\(\text{PT}_{R1}\) and \(\text{PT}_{R2}\). Aligning the pressures at a
crossflow of 650 mL/min according to their position revealed positive
TMP at the filter inlet and negative TMP at the filter outlet. The TMP
was zero in the middle of the filtration module (Figure 2C ). A
similar crossflow ramping was performed to characterize pressures for
the HPTFF system with activated delta pressure control to match the
filtrate inlet pressure \(\text{PT}_{F1}\) with the retentate inlet
pressure \(\text{PT}_{R1}\) (Figure 2B ). In contrast to the TFF
and rTFF, filtrate pressures along the filter were not identical anymore
but matched the retentate pressure gradient along the entire filter
length (Figure 2D ). The only filtrate pressure sensor with a
discrepancy to the respective retentate pressure was \(\text{PT}_{F2}\).
This discrepancy was negligible for crossflows below 400 mL/min but
increased slightly with larger crossflow (Figure 2B ). Required
co-current filtrate flows for varying crossflows using the lab-scale
filter are provided in the supporting information (SI Figure
1A ).
Time resolved pressure recordings for the operation of a TFF and rTFF
system at a crossflow of 650 mL/min are provided in Figure 2E .
TFF is represented by only considering the forward crossflow phase. The
rTFF is described by adding a reverse crossflow phase, and thereby
alternating the crossflow. The only changing pressures upon crossflow
reversal were the retentate pressures, with \(\text{PT}_{R1}\) taking
the previous value of \(\text{PT}_{R2}\) and vice-versa. HPTFF operation
at 650 mL/min crossflow was achieved upon delta pressure control
activation with a co-current filtrate flow of approximately 1400 mL/min
(Figure 2F ). Immediately, filtrate pressures align with the
retentate pressure gradient and are stably maintained at the target
values.
A schematic representation of the pressure characterization experiments
summarizes the findings for the TFF system (Figure 3A ), rTFF
system (Figure 3B ) and for the HPTFF system (Figure
3C ). The schematic pressure plots demonstrate the TMP differences along
the fiber length for the TFF and the rTFF system. A zoomed view into a
hollow fiber at the beginning, in the middle and at the end of the
filter module further highlights the Starling recirculation indicated by
arrows. Compared to the TFF and rTFF system, the filtrate pressures in
the HPTFF system are well aligned with the retentate pressure thereby
generating a uniform TMP of only slightly above zero along the length of
the filtration module. Small arrows from retentate to filtrate indicate
that the entire filtration area is utilized for filtration by avoiding
Starling recirculation.
Pressure characterization of scTFF
Whereas HPTFF operation focused on matching filtrate pressures with the
retentate pressure gradient and thereby removing Starling recirculation
completely, we also examined a novel operating mode for unidirectional
TFF defined as stepping co-current TFF (scTFF). The scTFF consists of
two phases, a first phase with lower co-current filtrate flow than
required for HPTFF, and a second phase with higher co-current filtrate
flow, resulting in a step profile for the co-current filtrate flow
(Figure 4A ). To demonstrate the impact of co-current filtrate
flow on the pressure profiles, a co-current filtrate flow ramping was
performed by fixing the crossflow to 650 mL/min (Figure 4B ). At
0 mL/min co-current filtrate flow, the system basically corresponded to
a standard TFF operation. With increasing co-current filtrate flow, the
pressure aligned more and more to the retentate pressure gradient and
matched it at about 1400 mL/min, corresponding to the situation in HPTFF
operation. Further increasing the co-current filtrate flow led to higher
filtrate pressures in the first half of the filter and lower filtrate
pressures in the second half of the filter compared to the retentate
pressure gradient. The filtrate pressure at the outlet\(\text{PT}_{F2}\) was not plotted as similar discrepancies to the
retentate pressure gradient as seen in Figure 2B were observed.
Selecting a co-current filtrate flow of 870 mL/min for phase 1 (blue
vertical dashed line) and 1890 mL/min for phase 2 (red vertical dashed
line) of the scTFF operation, a delta pressure between\(\text{PT}_{R1}\) and \(\text{PT}_{F1}\) of -10 mbar and +10 mbar,
respectively, was achieved. Pressures recorded for the two phases of
scTFF were then plotted according to their position along the filter
(Figure 4C ). The black line represents the retentate pressure
gradient, the blue dashed line represents the pressure drop on the
filtrate side for scTFF phase 1 and the red line represents the pressure
drop on the filtrate side for scTFF phase 2. A common intersection of
all three lines was located in the middle of the filter length, meaning
the absolute TMP is zero in the middle of the filter and gets larger the
closer to one of two filter ends.
By switching between scTFF phase 1 and scTFF phase 2 with defined phase
times, a scTFF system with unidirectional crossflow but reversing
Starling recirculation was obtained (Figure 4C ). Red areas
represent the flux of filtrate back into the retentate due to higher
filtrate pressures compared to the retentate pressures, whereas blue
areas represent flux from retentate to filtrate due to higher retentate
pressures compared to filtrate pressures. As such, filtrate pressure\(\text{PT}_{A1}\) positioned at 5.5 cm from the filter inlet was lower
than the corresponding retentate pressure at 5.5 cm filter length (black
dashed line) during scTFF phase 1 and got larger than the corresponding
retentate pressure during scTFF phase 2. Similar, but reversed,
behaviour was observed for pressure \(\text{PT}_{A5}\) positioned on the
second half of the filter at 64.5 cm filter length. In this case,\(\text{PT}_{A5}\) was larger than the retentate pressure during scTFF
phase 1 and smaller than the retentate pressure during scTFF phase 2. A
combined HPTFF-scTFF operation is also possible by integrating a
sweeping into the HPTFF operation. The sweeping was achieved by lowering
the co-current filtrate flow (scTFF phase 1) and subsequently increasing
the co-current filtrate flow (scTFF phase 2). After the sweep, the
system was again operated at HPTFF conditions (Figure 4D ).
Characterization of performance in perfusion cell culture
processes
Cell culture parameters and product retention were compared for TFF,
rTFF, HPTFF and scTFF operation in steady-state perfusion processes. For
all four cell retention setups (Figure 1B-D ), steady-state
operation was achieved after approximately 5 days and culture viability
was not impacted by the cell retention operating mode (Figure
5A ). Target process run time of 30 days was achieved for all runs
except TFF_1 and HPTFF_2. These runs were terminated at day 19
(TFF_1) and day 21 (HPTFF_2) due to a sudden decrease in crossflow
caused by inlet blocking of the fibers. Cell diameter increased slightly
with runtime for all of the cell retention systems (Figure 5B )
and pH stayed within the defined range of 7.07 ± 0.17 for all runs
(Figure 5C ). Cell debris increased for most runs until day 25,
after which a slight decrease in cell debris was observed. In general,
TFF and rTFF runs showed slightly higher debris levels compared to HPTFF
and scTFF runs especially after day 13 (Figure 5D ). The harvest
titer plot (Figure 5E ) and the product sieving plot
(Figure 5F ) revealed significantly reduced product sieving of
around 80% for the TFF operation after only a few first days of
steady-state operation. Product sieving further decreased down to 60%
or lower for TFF. Product sieving for rTFF stayed above 90% for the
entire experiment for run rTFF_1 and remained above 80% for run
rTFF_2. HPTFF operation resulted in similar or even higher product
sieving with yields above 95% for the entire run.
Cell culture bioreactors must be oxygenated to support cell growth by
sparging air or oxygen. Centrifugal pumps in unidirectional crossflow
operations (TFF, HPTFF and scTFF) tend to accumulate gas bubbles coming
into the cell recirculation loop. This problem was solved by stopping
the pumps for 3 seconds every 3 minutes to release the air from the pump
head. With activated delta pressure control during HPTFF operation
controlling delta pressure to 0 mbar, stopping the crossflow for 3
seconds caused a sharp change in the pressure profile along the filter
length (Figure 6A ). Due to some delay of the PI controlled
co-current filtrate flow regulation, the filtrate pressure\(\text{PT}_{F1}\) was higher than the retentate
pressure\(\ \text{PT}_{R1}\) immediately after crossflow stopping, which
resulted in a negative delta pressure up to -14 mbar (Figure 6Bred area). After reactivation of the crossflow, the co-current filtrate
flow was reduced and the PI control required some more time to establish
HPTFF conditions. During that time interval, a positive delta pressure
of up to 5 mbar was seen at the filter inlet (Figure 6B blue
area). Taken together, stopping the crossflow during HPTFF operation
resulted in a slight membrane sweeping. In rTFF operation, gas bubble
trapping was alleviated by positioning the pumps such that they are
pointing towards each other, with air removed from the pump head by
alternating activation of the retentate pumps.
Large-scale filter pressure
characterization
Large-scale experiments using TFF, HPTFF and scTFF operation confirmed
results from the lab-scale experiments. Crossflow ramping from 0 – 45
L/min showed a continued increase in the pressure gradient along the
filter length in TFF operation while the permeate pressures were
independent of position (Figure 7B ). Considering similar
crossflow conditions with 29 L/min corresponding to a shear rate of 1470
s-1 as applied during lab-scale perfusion cell culture
runs resulted in a fiber inlet pressure of 71 mbar \(\text{PT}_{RC1}\),
a fiber outlet pressure of 31 mbar (\(\text{PT}_{RC2}\)) and an average
filtrate pressure (\(\text{PT}_{A1-5}\)) of 51 mbar (Figure
7D ). Filtrate pressures on the filtrate inlet and outlet
(\(\text{PT}_{A1-5}\)) showed similar values as the filtrate pressure
sensors on the backside of the filter module (\(\text{PT}_{AB1-5}\)).
HPTFF operation could be achieved by controlling pressure\(\text{PT}_{F1}\) so that it was 6 mbar above the pressure\(\text{PT}_{R1}\) for all evaluated crossflows up to 45 L/min
(Figure 7D ). Filtrate
pressures on the filtrate inlet and outlet (\(\text{PT}_{A1-5}\)) were
aligned with the corresponding filtrate pressures on the backside
(\(\text{PT}_{AB1-5}\)) (Figure 7F ). In contrast to TFF
operation, filtrate pressures matched the pressure gradient of the
retentate loop, with pressure sensors \(\text{PT}_{A1}\) and\(\text{PT}_{AB1}\) slightly lower than the corresponding retentate
pressures. The filtrate outlet pressure \(\text{PT}_{F2}\ \)was
significantly lower compared to the other pressure sensor readings.
Co-current filtrate flows to achieve HPTFF at varying crossflows are
provided in the supporting information section (SI Figure 1B).
scTFF operation to generate controlled Starling recirculation was
further demonstrated with a large-scale filter module and the data are
provided in the supporting information section (SI Figure 2 ).
As such, a filtrate loop ramping at constant crossflow (SI
Figure 2A ) and pressure distribution along the filter length for scTFF
phase 1 (SI Figure 2B ) and for scTFF phase 2 (SI Figure
2C ) are provided. In addition to changing the filtrate flow, a
crossflow stop whilst keeping the filtrate PI control active was able to
achieve effective membrane sweeping (SI Figure 2D ).