The approach to new filter implementation at CMC Biologics involved filter evaluation at small-scale in process development
followed by scale-up to a full-scale engineering run, then use in cGMP production. After selecting Sius as the candidate single
use TFF filter for PF1, a monoclonal antibody (MAb X) was chosen as the initial experimental model (see Figure 3). MAb X had
previously been manufactured using a traditional re-useable TFF filter and was known to tolerate both the standard TFF process
and final product concentration without issues. It also provided a baseline for performance parameters such as flux and step
recovery. However, a detailed comparison between the two filters was not a focus of the evaluation, as switching to single-use
technology was a strategic decision. Using vendor-recommended operating parameters, a simple buffer exchange process was performed
in the Development laboratories with a bench-scale 0.01 m2 Sius-LS Prostream mPES TFF cassette. Product load was 270 g/m2 , which is typical for many TFF processes, although this was 2.5-fold higher than the original manufacturing process that
had used oversized filters for faster operations. Buffer exchange with the Sius filter, as measured by conductivity, matched
predictions for a zero retention coefficient, the expected result for low-molecular-weight buffer components using a 30 kDa
MWCO membrane. Product yield was high at 96%, which was equivalent to the original manufacturing process. Product quality
attributes, such as percentage of aggregates, were also unaffected by the TFF operation using either filter. Flux was slightly
higher with the Sius, averaging 53 L/m2 /h (LMH) compared with 46 LMH with the re-usable filter in manufacturing, despite the higher protein load and lower transmembrane
Figure 3: Evaluation with MAb X. A) Process comparison. The re-useable tangential flow filter from the original PF1 manufacturing
process and laboratory-scale and PF1 manufacturing-scale Sius filters were all 30 KDa molecular weight cut-off and used cross-flow
rates of 360 L/m2/h. Flux for the concentration and diafiltration stages of the process are shown in the upper graph, and operating parameters
in the lower table. B) Buffer exchange for laboratory-scale Sius. Conductivity measurements at various points in diafiltration
are plotted against the theoretical conductivity assuming 0% retention of buffer components. Predicted % residual buffer is
Mab X was then used to address the second objective of process scale-up to PF1 for a 500-L production-scale engineering run.
Product was loaded at 210 g/m2 , somewhat less than in the lab-scale trial but close to twice the load of the original process with the re-useable membrane.
Because the single-use system eliminates several pre- and post-use operational steps (see Figure 1), this time savings more
than compensates for longer processing times during UF/DF from the larger load. Both flux and yield for the single-use filter
were comparable between lab-scale and GMP manufacturing. Likewise, product quality was consistent. Operationally, no issues
were encountered implementing the system in cGMP manufacturing, and the design was sufficiently similar to traditional multi-use
TFF that little additional operator training was required.
Sius TFF cassettes were next evaluated with two problematic proteins to determine whether the system might provide a platform
buffer exchange process for the wide variety of client projects performed at CMC Biologics (see Figure 4). Protein Y was a
large multimeric protein that was prone to aggregation, and it was evaluated using the same membrane chemistry as had been
used for MAb X (30 kDa molecular weight cut-off, Prostream). Again, the filter was first tested at small-scale in development
followed by a full-scale engineering run before use in cGMP manufacturing. Performance was similar at both scales. To minimize
risk of aggregation, the final target concentration in the TFF step was only slightly higher than the target for final drug
substance. Thus, very little buffer volume could be used for a final system flush, providing a challenge for product recovery.
TFF performance was good simply using vendor-recommended conditions, with essentially quantitative recovery in PF1 cGMP manufacturing.
More importantly, no increase in aggregation was observed as a result of the operation. Protein Z was a fusion protein for
which a final target concentration of 50 mg/mL was evaluated in the development group. For this application, the more hydrophilic
Hystream mPES membrane at 50 kDa MWCO was used, as it is the vendor-recommended filter for high protein concentration applications.
Only minor process optimization was required to achieve an average flux of 96 LMH during diafiltration for rapid processing
time, which included performing the diafiltration step at 20 mg/mL protein concentration at an increased cross-flow rate of
900 LMH before a final concentration step to 50 mg/mL. Product quality was maintained through buffer exchange. This high
concentration process was not scaled up because of a change in project requirements.
Figure 4: Evaluation with Protein Y and Protein Z. A) Tangential flow filtration (TFF) of protein Y at manufacturing scale
in PF1. The table on the left shows process parameters, flux during concentration and diafiltration are in the graph on the
right. B) TFF of protein Z in a small-scale experiment. The table on the right shows process parameters, flux and cross flow
rates during concentration and diafiltration are in the graph on the left.
The final stage of evaluation was to assess full-scale process consistency in cGMP manufacturing. Six batches of MAb X and
five of Protein Y were produced using the Sius TFF system for final buffer exchange (see Table I). Data were obtained from
routine cGMP operations as real-world examples. As such, flux and TMP are simply averages of data points periodically recorded
manually in manufacturing batch records. One objective of this evaluation was to determine whether lot-to-lot variation in
filter cassettes might be an issue for the single-use system. A total of four different lots of Sius filters were used over
the course of the MAb X campaign and two lots for the Protein Y campaign. Product yield was consistently high across all manufacturing
batches and processing times for UF/DF were rapid and predictable, within ranges expected for manually performed operations.
Table I: Process performance of single-use buffer exchange step in cGMP manufacturing.