Scaling up production to 500 L for a highly concentrated monoclonal antibody

Cell culture

We ran a fed-batch culture in Xcellerex™ XDR-500 bioreactor to produce a monoclonal antibody using HyClone™ ActiPro™ cell culture media, HyClone™ Cell Boost™ 7a supplement, and HyClone™ Cell Boost ™ 7b supplement. The culture had a mAb titer of 4.6 g mAb/mL and a VCD (viable cell density) of 35.8 million cells/mL at harvest.

Clarification using depth filters

Before harvest, we rinsed the Stax™ depth-filter capsules with deionized water and primed with PBS. They were then drained to minimize product dilution.

The primary depth filtration for cell removal was performed on ten 1 m² PDK7 Stax™ L capsules followed by five × 1 m² PDCX Stax™ L capsules for the secondary clarification. The selected flux setpoint was 50 LMH (liters per square meter per hour) for the primary depth filters and 100 LMH for the secondary depth filters. The primary and secondary clarifications were run separately to allow for sampling and turbidity measurement in between.

After the depth filtration, we filtered the product through a 0.2 µm, Supor™ EKV sterilizing grade filter, 30 in. capsule.

Chromatography purification and viral inactivation

The clarified depth filtrates were purified through three different chromatography steps. The capture was performed on a ReadyToProcess™ MabSelect PrismA™ chromatography column at a load of 50 to 58 g/L; the first polishing was performed on a on a ReadyToProcess™ Capto™ S ImpAct chromatography column at a load of 70 g/L and the second polishing was performed in flowthrough mode on a Mustang™ Q XT chromatography membrane at a load of 2000 g/L.

We used an ÄKTA ready™ chromatography system (using a low-flow kit for capture and polishing 1 and high-flow kit for polishing 2) for all three purification steps.

The chromatography method used for capture, polishing 1 and polishing 2 are shown in Tables 1 to 3 below.

Table 1. Capture MabSelect PrismA™ chromatography method

The capture MabSelect PrismA™ eluate pool was viral inactivated for 60 min at pH 3.7 using 4 M acetic acid. We then adjusted the pH to 5.0 with 2 M Tris base and the product solution was filtered through a 0.2 µm Supor™ EKV sterilizing grade filter before polishing 1.

The viral inactivation (VI) was performed in Xcellerex™ XDUO-100 mixer.

Table 2. Polishing 1 method using a ReadyToProcess™ Capto™ S ImpAct chromatography column

We conditioned the Capto™ S ImpAct eluate pool using a 100 mM sodium acetate buffer, pH 8.3, before final polishing on a Mustang™ Q XT chromatography membrane. This achieved an adjustment of the pH to about 6.0 and the conductivity to less than 11 mS/cm in preparation for the next step, which was a flowthrough polishing step.

We performed the conditioning in Xcellerex™ XDUO-500 mixer.

Table 3. Polishing 2 method using Mustang™ Q XT chromatography membrane

Table 4 below summarizes the output data from the chromatography purification steps at the 500 L scale.

Table 4. Chromatography purification and viral inactivation (VI) output data

Virus filtration

We filtered the flowthrough eluate pool from Mustang™ Q XT chromatography membrane using a 0.2 µm Supor™ EKV sterilizing grade filter (1.8 m2). We then applied the eluate pool to a filter train consisting of Pegasus™ Protect prefilter and Pegasus™ Prime virus filter connected to an ÄKTA ready™ chromatography system (using the low-flow kit). The filter train was primed with the Mustang™ Q XT chromatography membrane equilibration buffer before filtering the product.

We performed the filtration at constant delta column pressure of 2.0 bar (29 psi, 0.2 MPa) through a UNICORN™ chromatography software method. We set a flow rate high enough in the method for UNICORN™ software to allow a delta column pressure of 2.0 bar while controlling the flow. The UNICORN™ method used is shown in Table 5 for the 500 L scale.

The system maintained a steady delta filter pressure and a slight decrease in flow rate was seen over time.

Table 5. Virus filtration method using Pegasus™ Protect and Pegasus™ Prime virus filters at the 500 L scale

Final formulation through ultrafiltration/diafiltration (UF/DF) TFF

We performed the final formulation TFF step on a 30 kDa Delta membrane of 2.5 m² using an ÄKTA readyflux™ 3/8 in. flow kit. Two sublots were run with the maximum system flow being 18 L/min as running only one sublot would have required a recirculation flow of 30 L/min using two membrane units of 2.5 m² each.

Before the UF/DF step, we integrity tested the membranes at a pressure of 4 bar (58 psi, 0.4 MPa) with a passing result. We performed membrane cleaning in place (CIP) using 0.25 M NaOH followed by a flush with water to remove the CIP agent. A normalized water permeability (NWP) test was performed before priming the membrane with formulation buffer.

During the initial concentration of the viral filtered product, the volume in the recirculation bag was kept at about 15 L during the initial concentration, for the ÄKTA readyflux™ system and was reduced slightly below 15 L to reach the titer of 50 g/L prior the start of the DF phase.

The UF and DF ran as expected at a transmembrane pressure (TMP) of 1.5 bar (21.8 psi, 0.15 MPa) and a recirculation feed flux of 360 LMH. This corresponded to a recirculation flow of 15 L/min on the ÄKTA readyflux™ system.

After the DF, we opened the retentate valve completely and the recirculation feed flux was lowered to 120 LMH. We closed the permeate valve and the product was allowed to recirculate for 15 min to increase the recovery by gentle mixing and enhanced diffusion of mAb from the membrane surface back into the solution. The retentate was emptied and collected separately.

Two flushes were performed using a flush volume of just above the system hold-up volume for each flush (hold-up for ÄKTA readyflux™ system was 1.22 L).

We cleaned the TFF membranes in place after use and flushed with water. We performed a post-UF/DF NWP test, which showed full recovery of the membranes.

Before the final concentration on SPTFF, we filtered the product pool on Supor™ EKV sterilizing grade filter.

Final concentration through SPTFF

The final concentration was performed on a single-pass TFF (SPTFF) module. A preassembled 4-in-series unit consisting of seven × 0.1 m² T12 membranes was used. We ran the step using a Quattroflow 150S pump. We controlled the retentate flow through a valve, preventing unstable performance, which most likely could lead to a too high concentration as the mAb concentration increases over the SPTFF unit. Another prerequisite was to start with a lower product concentration, and this was why we did not concentrate the mAb solution above 50 g/L in the previous TFF step.

Before starting the process, the SPTFF unit was cleaned in place with 0.25 M NaOH, flushed with water, integrity tested at 4 bar with a passing result, and an NWP test was performed, confirming successful cleaning. The hold-up weight in the system was 250 g for the unit used for the 500 L batch. At the end of the SPTFF step, we performed three flushes with one hold-up volume for each flush. We pooled the SPTFF retentate and flush 1.

Filtration using Supor™ Prime membrane of the final high-concentration mAb product

The Supor™ Prime filter capacity has previously been determined to be 390 L/m² when filtering at a flux of 400 LMH at a max. pressure setting of 2.0 bar (29 psi, 0.20 MPa) on a 3 cm² filter.

For the final SPTFF retentate at the 500 L scale we had 6.7 kg of product solution. We therefore filtered on a 240 cm² Supor™ Prime filter, despite the overcapacity, as about 9.4 kg of product solution could be filtered on this unit.

Although we did not have enough product solution to achieve 390 L/m² on the 240 cm² Supor™ Prime filter, we achieved about 280 L/m² with no increase in pressure. Therefore, we believe the determined filter capacity at 390 L/m² will hold valid for the 240 cm² filter unit as well.

The cell culture used for this study was a fed batch CHO cell culture. At the point of harvesting the 50 L batch, the viability was at 95%, the mAb titer was 5.1 g/L, the maximum cell density reached 51 × 10⁶ cells/mL, and the turbidity at harvest was 3020 FNU. The corresponding figures for the 500 L cell culture at harvest were 97% viability, a mAb titer of 4.6 g/L, a maximum cell density of 48 × 10⁶ cells/mL, and the turbidity at harvest was 3385 FNU.

The depth filtration on primary PDK7 filter and secondary PDCX filter went as expected. These grades are slightly tighter than their Stax™ max counterparts and offers improved fine particle removal from this challenging high cell density feed stream.

Figure 2 below shows the throughput vs the differential pressure for the 500 L batch for the PDK7 and PDCX filtration. The differential pressures are very similar at the two process scales, indicating a correct scale-up of the depth filtration. The results from the 50 L study described previously are included for comparison.

Fig 2. Depth filtration throughput vs. differential pressure for the 50- and 500 L batch. (A) Stax™ PDK7 and (B) Stax™ PDCX filter capsules.

The turbidity from the depth filtration and the 0.2 µm Supor™ EKV filtration is shown in Figure 3 for both the 50- and 500 L scale. As expected, the tighter depth filter train of PDK7 plus PDCX generated a filtrate pool with a low turbidity at similar levels for both process scales, confirming that the depth filtration has been scaled up well.

Fig 3. Turbidity after depth filtration with Stax™ PDK7 and PDCX filters and sterilizing grade filtration with Supor™ EKV filter.

The capture chromatography using MabSelect PrismA™ resin, as well as, both polishing steps on Capto™ S ImpAct resin, and Mustang™ Q XT membrane performed as expected. This demonstrates the suitability of running Mustang™ Q XT membranes on an ÄKTA pure™ 150 system during process development as well as running on an ÄKTA ready™ system at the 500 L batch scale. Examples of the chromatograms are shown in Figure 4.

The VI step at low pH also performed as expected and for the 500 L batch we demonstrated the suitability of using the Xcellerex™ XDUO-100 mixer for the virus inactivation. It is not uncommon to see a slight increase in product cloudiness when increasing the pH after the low pH viral inactivation, and pharmaceutical manufacturers may include a depth filtration step after the VI to remove the precipitated impurities. We observed a slight increase in cloudiness, and used a 0.2 µm Supor™ EKV filter only for precipitate removal.

Fig 4. Chromatograms from (A) MabSelect PrismA™ chromatography resin, 50 L batch; (B) MabSelect PrismA™ chromatography resin, 500 L batch; (C) Capto™ S ImpAct resin, 50 L batch; (D) Capto™ S ImpAct, 500 L batch, (E) Mustang™ Q XT chromatography membrane capsule, 50 L batch; (F) Mustang™ Q XT membrane capsule, 500 L batch.

A virus filtration step is traditionally performed at constant pressure, that is, through pressurizing a stainless-steel vessel with the product solution being transferred through tubing into the virus filter. However, we have shown that virus filtration can be performed with a filter train consisting of Pegasus™ Protect prefilter and Pegasus™ Prime virus filter using both an ÄKTA pilot™ 600 chromatography system for the 50 L batch and on an ÄKTA ready™ system for the 500 L batch.

The chromatograms from the virus filtrations are shown in Figure 5. They show how the chromatography system with the UNICORN™ software method can maintain a steady delta column pressure throughout the virus filtration and adjust the flow rate accordingly. In this case, we could not see any flow decay during the virus filtration for the 50 L scale, but a slight decrease was seen for the 500 L scale.

(A)

Fig 5. Data from the virus filtration using the Pegasus™ Protect/Pegasus™ Prime filter train at the (A) 50 L process scale and the (B) 500 L process scale.

The viral filtration output data is summarized in Table 6 below. The difference in load on Pegasus™ Protect and Pegasus™ Prime is a consequence of product weight and the size of the filters.

Table 6. Viral filtration output data

During the final formulation, we showed that the mAb solution could be formulated on T-series cassettes with Delta membrane of 30 kDa quickly through UF and DF using an ÄKTA flux™ 6 chromatography system for the 50 L scale and scaled up using an ÄKTA readyflux™ at the 500 L scale.

The scale-up to the ÄKTA readyflux™ system also showed the possibility of running an entire UF/DF process through a controlled UNICORN™ method. We also showed robust performance at both scales concerning steady TMP and similar process times.

Table 7 summarizes the TFF data and only the first sub-lot for the 500 L scale is shown as the second sub-lot showed data in the same range. Noted weights and titers for the 500 L scale are given for the pool of the two sub-lots.

Table 7. Final formulation TFF data

Figure 6 below shows the TMP and permeate flux during the UF and DF. Here, we see that the permeate flux decreases as the product is concentrated during the UF. The drop in flux and TMP after about 75 to 80 min was due to lowering the recirculation flow (since the ÄKTA flux™ 6 system is a manually operated system) for rearranging the buffer vessel and permeate collection vessel before the DF started for the 50 L scale. This is not seen for the 500 L scale using the ÄKTA readyflux™ as the UF/DF was run through a UNICORN™ method.

Once the DF started, we see that the permeate flux increases initially, which may be an effect of the viscosity change due to the buffer exchange. The TMP is seen to be stable during the UF and DF.

Fig 6. TMP and permeate flux during the UF and DF of the TFF step: (A) 50 L scale; (B) 500 L scale.

Following the UF/DF TFF concentration, we ran the SPTFF step for final concentration with a 4-in-series SPTFF unit with a Delta membrane of 30 kDa MWCO. SPTFF final concentration step data comparing the 500 L scale with the previous 50 L scale process are summarized in Table 8.

The data show almost identical volumetric concentration factors (VCF) of 5.4 and 5.1, respectively at steady feed and retentate pressures throughout the entire process step.

Figure 7 shows the accumulated retentate- and permeate weights throughout the SPTFF process and the profiles look very similar for both the 50- and 500 L scale.

Fig 7. Accumulated retentate- and permeate weights throughout the SPTFF step, as well as the feed and retentate pressure. (A) 50 L scale, (B) 500 L scale.

The final titer in the SPTFF retentate reached well above 200 g/L and ended around 225 g/L for both process scales. The results are summarized in Table 8.

Table 8. Final concentration SPTFF data

We filtered the final concentrated product through a 240 cm² 0.2 µm Supor™ Prime filter. The throughput capacity was previously determined to 390 L/m² at a flux of 400 LMH on a 3 cm² Supor™ Prime filter unit.

For the final concentrated product from the 50 L scale, we filtered the product quickly at constant pressure, whereas for the final product from the 500 L scale, all the product was filtered at a constant flux instead (Fig 8), but not enough product was present to achieve 390 L/m² on the 240 cm² unit.

We achieved a throughput capacity of 280 L/m² for the 500 L process scale and at that point, we had not observed an increase in pressure. The final titer in the Supor™ Prime filtrate became 222.9 g/L.

The final data is summarized in Table 9.

Fig 8. Supor™ Prime filtration data at 500 L scale.

Table 9. Final Supor™ Prime filtration data

We monitored the mAb production process at each process step using several different analytical methods.

Analytical results are listed in Table 10 for the final product and are as expected. We found that the level of aggregates was slightly high. However, we expected these aggregation levels at mAb titers of over 200 g/L considering that the formulation buffer was not optimized at all. The formulation buffer was only a 50 mM acetate buffer at pH 5.0 without any stabilizing excipients.

Table 10. Analytical results for the final bulk product

The step yields in Table 11 are as expected. The mAb amount in the 50 L cell harvest was 229.5 g and post-clarification, 208 g of mAb remained, which gave a step yield of 90.6%. The mAb amount in the 500 L cell harvest was 1984 g, and post-clarification the titer determined by a CEDEX analyzer gave a mAb amount of 2104 g. This is a falsely high value as mAb cannot be generated over the clarification step. An assumed step yield of 90% for a midstream clarification step is more accurate and would instead give a mAb amount of 1786 g post-clarification. This value was used as the starting value before the capture step.

The mAb titer was determined with the CEDEX analyzer before the capture and with a spectrophotometer post-capture. Calculating the step yield based on two different analytical methods will most likely generate different values with an impact on the calculated step yield.

The step yield for the MabSelect PrismA™ step is usually about 95%. The low step yield of 85.7% for the 50 L scale is most likely an effect of determining the titer with two analytical methods. If the step yield for the capture step were based on CEDEX analyzer titer values only, the step yield would have been 102.8% which is not possible either. The same was observed for calculating the step yield for the MabSelect PrismA™ at the 500 L scale, which is also noted in Table 11 below.

The step yield for Capto™ S ImpAct was expected to be around 90% since we ended the peak collection at 500 mAU. However, it was slightly higher for the 50 L batch and lower for the 500 L batch. This could be due to analytical variations, and also an effect of using different chromatography systems with differences in delay volumes for the peak cutting.

The step yields for the virus filtration, TFF, and SPTFF were as expected.

The step yield for filtration with the Supor™ Prime filter on the final product is also most likely a false value as it is somewhat difficult to pipette a viscous high-concentration mAb.

Table 11. Step yields (%) for the 50- and 500 L scale process