OR WAIT null SECS
Benedicte Lebreton is a scientist, early-stage purification, at Genentech, Inc.
Amber R. Cothran is an engineer in late-stage purification at Genentech, Inc.
Sandy Sargis is a senior research associate in early-stage purification at Genentech, Inc.
Melody Trexler-Schmidt is a scientist in late-stage purification at Genentech, Inc.
Binh Q. Thai is an engineer in late-stage purification at Genentech, Inc.
Brian Kelley is vice president, bioprocess development at Genentech, Inc.
Gregory S. Blank is the director of late-stage purification at Genentech, Inc.
Stefanie Sze-Khoo is an engineer in late-stage purification at Genentech, Inc.
Altering the order of operations, using new resins, and increasing dynamic binding capacity can obviate the need for major facilty changes.
Monoclonal antibody purification processes have been challenged recently to recover 5 g/L cell culture titers in existing manufacturing facilities. This requires the modification of a platform purification process comprising Protein A chromatography as the robust capture step followed by two ion exchange chromatography steps. Because buffer and pool tank volumes are the primary facility bottlenecks, process development efforts have focused primarily on the order of unit operations, the dynamic binding capacity of current and new resins, reducing pool manipulations and buffer consumption, and elution buffer selection. This article presents examples in which new operating conditions or purification technologies can help to accommodate increases in titer without extensive changes to the manufacturing facility or equipment requirements.
Over the past five years, there has been a dramatic increase in the titers of recombinant proteins, particularly in the production of IgG monoclonal antibodies (MAbs) using Chinese hamster ovary (CHO) cells. Improvements in cell lines, media composition, and cell culture operating conditions have all contributed to higher expression levels. In most cases, these changes have had little or no impact on manufacturing equipment and facilities. However, as titers have increased by an order of magnitude or more using existing cell culture bioreactors, there has been a concomitant increase in the starting mass of protein entering downstream purification. This development has resulted in a shift in emphasis from increasing volumetric productivity for low-titer processes to handling significant increases in the amount of protein during downstream processing, particularly in existing facilities. Although incremental increases in titers can be handled by increasing the scale of purification unit operations (e.g., using larger chromatography columns and filters), at some point linear scaling exceeds the physical limits of existing facilities.
Lonza, Ltd., Basel, Switzerland
To address the purification challenges of high antibody titers, it is imperative to determine the specific limitations of a company's manufacturing facility or facilities. This can best be accomplished by using a model that performs a facility fit using an existing process. Such a model can identify the bottlenecks for each unit operation at a specific manufacturing facility as a function of antibody titer. In some cases, it may be necessary to use multiple model iterations to find a bottleneck, model a change to the process or facility, and then determine where the next bottleneck may be. For the purposes of this article, we assume that one cell culture batch is purified to produce one bulk. Alternative process options such as split batch processes may be considered, and these could potentially allow higher recoverable titers from one cell culture batch.
Almost all current antibody purification platform processes use Protein A affinity chromatography as the capture step because it is robust, widely applicable, and facilitates the implementation of a successful standard purification process.1,2 Figure 1 shows two platform purification processes that we have implemented. After cell and product separation during harvest operations, Protein A affinity chromatography is used as the initial capture step with low-pH viral inactivation in the Protein A pool. Depending on the platform, this is followed by either the cation exchange chromatography step in bind-and-elute mode3 or the anion exchange chromatography step in flow-through mode. The virus filtration operation is placed between the two ion exchange chromatography steps. After the third chromatography step, ultrafiltration–diafiltration is performed using tangential flow filtration (TFF) for final product formulation.
Potential volume bottlenecks caused by increased titers are included for each unit operation of each process shown in Figure 1. Underneath each potential bottleneck are process options that should be considered for optimization to avoid facility bottlenecks, and those having the greatest impact are shown in red. We have found that the primary bottlenecks in existing facilities are buffer and pool tank volumes, whereas secondary bottlenecks include pump capacities or pipe diameters that limit flow rates. For these reasons, the use of larger chromatography columns or filter surface areas is not sufficient to handle higher titers.
One approach that allows greater amounts of antibody to be purified without increasing column volume is to increase the binding capacity of the chromatography step. Increasing the load capacity means the overall amount of processed antibody can be increased without increasing column size or volume. Since buffer volumes are usually based on column volume, increasing the load mass without increasing the column volume requires no increase in buffer volumes. The same applies for elution conditions where the pool volume will only be affected slightly by overall protein mass when using a final optical density (OD) cut off, or not affected at all when using constant volume pooling criteria.
For Protein A chromatography, Table 1 shows the increase in titers that can be tolerated in an existing facility without increasing the size of the chromatography column. At a dynamic binding capacity (DBC) of 15 g/L, a maximum titer of 1 g/L can be accommodated in the facility. If the DBC can be doubled to 30 g/L, then the maximum recoverable titer increases to 3 g/L and a DBC of 45 g/L accommodates titers of up to 5 g/L. An increase in DBC can be achieved in several ways: by lowering the flow rate at constant bed height, increasing the bed height at constant flow rate to increase residence time, and/or by changing to a different Protein A resin with higher capacity.4 In addition, increasing the maximum manufacturing load density from low breakthrough levels (0.5–2% DBC) to slightly higher levels (3–5% DBC) also enables the purification of >5 g/L titers and most likely has an insignificant impact on step yield. Differences in DBC resulting from packing or resin lot-to-lot variability are considered negligible.
Table 1. Maximum expression level of a MAb that can be purified as a function of Protein A dynamic binding capacity. This assumes no other modifications were made to the facility, equipment, or operating conditions.
Protein A resins all demonstrate higher DBCs with decreasing flow rate, although the slope of the curve may vary from resin to resin. As shown with the DBC curve for resin A in Figure 2, the DBC can be increased from 17 g/L at 40 column volumes (CV)/h to 38 g/L by decreasing the flow rate to 10 CV/h, therefore doubling the amount of antibody recovered with no other process changes. Processing time and facility throughput need to be assessed to determine whether reducing the flow rate also reduces overall capacity.5 This may be addressed in part by reducing the flow rate only for the loading phase of the Protein A step and then increasing the flow rate for the other phases such as equilibration, wash, elution, and regeneration. Unpublished data from our laboratories indicates that step yield is not affected by operating these other phases at higher flow rates. Another potential strategy is to conduct the load phase at two flow rates, and decrease the flow rate during the final stages of the load phase.6
Table 2 shows how increasing the DBC affects a cation exchange chromatography step for a particular antibody and highlights the bottlenecks in an existing facility. In this example, a DBC of 50 g/L can accommodate a 1 g/L titer without constraints. A titer of 2 g/L is constrained by the buffer tank volume (the equilibration buffer in this case), and if this limitation is removed (perhaps by in-line dilution), the constraint becomes the pool tank volume at a titer of 3 g/L. Increasing the binding capacity of the cation exchange resin to 75 g/L or 100 g/L allows titers of 3 g/L and 4 g/L to be accommodated, respectively. In both cases, the buffer tank limitation occurs at lower titers than the pool tank volume limitation. This assumes that all other process parameters remain unchanged when the capacity increases. However, one key consideration during process development is the separation of the antibody and impurities at these higher load densities or with different (higher capacity) resins that may have different selectivities.
Table 2. Maximum expression level of a MAb that can be purified as a function of the dynamic binding capacity of cation exchange chromatography. This assumes no other modifications were made to the facility, equipment, or operating conditions.
As discussed above, it is important to take a holistic view of the process and carry out multiple model iterations that allow bottlenecks to be identified and removed, and subsequent bottlenecks to be identified. In examining the entire purification process, Table 3 shows the maximum titer that can be purified downstream by each unit operation. The baseline Protein A step can accommodate a titer of 1.3 g/L, but modification of the equilibration and various wash phases to minimize the Protein A buffer volumes increases this to 1.6 g/L. Table 4 shows an example of one equilibration phase and three separate wash phases in each Protein A chromatography cycle, where wash 1 and wash 3 also consist of equilibration buffer. In the baseline process, five column volumes are used for equilibration, three for wash 1, and three for wash 3. Studies have shown that this can be reduced to three column volumes for equilibration, two for wash 1, and two for wash 3 without affecting yield, purity, or column re-use, resulting in an overall reduction of four column volumes of equilibration buffer per cycle. For example, for a 140-L column operating eight cycles to process each cell culture batch, the revised process would reduce the consumption of equilibration buffer by 4,480 L per batch.
Table 3. Maximum expression level of a MAb that can be purified based on different process or facility modifications
Since buffer tank volumes are a significant constraint throughout the purification process as antibody mass increases, one solution is to use concentrated buffers in volume-constrained tanks and then dilute each buffer to the desired concentration in-line to the chromatography column. This method uses various types of equipment (pumps, metering devices, and pH and conductivity sensors) to blend a concentrated buffer with water to achieve the desired concentration, which has been shown to attain the target concentration ±2%.7 The primary limitations of in-line buffer dilution are the ability to concentrate buffers sufficiently to fit into available tank volumes and the installation and validation of the in-line dilution equipment. As shown in Table 3, the maximum titer accommodated in the Protein A step increases from 1.3 g/L in the baseline case to 1.9 g/L using optimized phase durations and a 1.25x in-line buffer concentration factor. Using 2x concentrated buffers for the Protein A step would increase the capacity still further, accommodating titers of up to 2.9 g/L. The cation and anion exchange chromatography steps are also able to process more antibody through the use of in-line dilution and large column volumes. Although this approach requires capital expenditure and plant down time for installation, in-line dilution enhances the capacity and flexibility of a manufacturing facility. One additional benefit may be the option to move away from using fixed, stainless steel tanks for buffer storage and move to concentrated buffers in disposable bag systems.
Table 4. Operational sequence of a Protein A chromatography step and example phase durations in units of column volume (CV).
Once buffer tank volume constraints have been addressed, pool tank volumes can become the next significant bottleneck. Pool tank constraints can arise during a given step through two mechanisms: a) the volume of the elution pool exceeds the tank capacity, or b) the pH and/or conductivity adjustments to condition the pool in preparation for the next chromatography step exceed the tank capacity. Increasing the resin DBC can address the first limitation, but the second limitation requires downstream unit operations to be investigated. Depending on the elution conditions and the target conditions for the subsequent step, a large volume of water may be necessary to lower the conductivity to the desired level for optimal antibody binding or impurity removal.
When the cation exchange chromatography step is placed after the Protein A step, specific development considerations are needed to ensure both acceptable process performance and pool tank fit at different facilities. In some of our facilities, the bottleneck does not lie in the cation exchange pool tank itself, but rather in the pool tanks of the subsequent viral filtration and anion exchange steps. The cation exchange elution phase must be optimized to remove host cell protein (HCP), leached Protein A, and aggregates, while also minimizing pool volume and downstream pool conductivity. The anion exchange step functions to clear HCP and often requires a low load conductivity for acceptable process performance. The available operating space is often defined by a combination of factors affecting purity, product quality, product stability, yield, and manufacturing facility fit, as shown in Figure 3 for a cation exchange process. In this example, facility fit sets the upper limit for salt concentration, while both yield and purity are constrained by the interactive combination of elution buffer salt concentration and pH. One approach for debottlenecking the plant fit is to decrease downstream pool conductivities by eluting the cation exchange step at higher pH, thus lowering the salt concentration, assuming no negative impact on product quality at higher pH conditions.
Newer anion exchange resins on the market have been designed for improved process performance at higher load conductivities. Table 5 compares two anion exchange resins, Q Sepharose Fast Flow (QSFF, GE Healthcare, Uppsala, Sweden), which has been a standard in antibody purification for some time, and Capto adhere (GE Healthcare), a new multimodal anion exchange resin introduced in 2007. In the flow-through process modeled in the platform process A example, QSFF requires a conductivity of 11 mS/cm or less to achieve optimal yield and requires 1,700 mL of water for each liter of the input pool to achieve this load conductivity. In contrast, Capto adhere was designed to have an optimal operation window at higher conductivities than previous anion exchange resins. At a load conductivity equal to that of QSFF (11 mS/cm), Capto adhere has a significantly poorer yield, but as the conductivity increases, the yield improves and is equal to that of QSFF at a conductivity of 22 mS/cm. The benefit to plant fit is the reduced volume of water (from 1,700 mL/L to 0 mL/L) needed to condition the load conductivity, removing the bottleneck caused by the pool tank volume limitation. As an alternative, TFF can be used to diafilter the pool, achieving the desired pH and lower conductivity, but this involves adding a new unit operation to the process. In addition, this increases the capital costs of installation, plant down time, and running costs. There must also be space available to install a TFF system in the existing plant.
Table 5. Impact of water required to adjust the conductivity for loading on an anion exchange chromatography resin used in platform process A
The Protein A process plays an important role in purification when the anion exchange chromatography step is next, as in platform process B. Protein A process parameters such as load density, elution buffer, and pooling criteria can have a significant effect on downstream pool volumes. The pH of the Protein A pool must meet low pH criteria for viral inactivation (pH ≤3.6),8 ideally without requiring additional acid titration. Process optimization must focus on minimizing the number of cycles, the pool volume per cycle, the volume of titrant required for pool pH adjustment, and the adjusted pool conductivity. Although maximizing the Protein A load density reduces the number of cycles, modifying the elution buffer components to minimize the pH and conductivity adjustment before the next chromatography step presents another alternative to address pool tank limitations. Figure 4 shows the impact of four different Protein A elution buffers on volume of titrant (Tris base) required to titrate the Protein A pool to pH 7.5 for loading onto the anion exchange step, and the resulting pool conductivity. There is a considerable difference in the amount of Tris base needed to adjust the pool pH to 7.5, ranging from 7 to 54 mL of Tris base per L of pool, depending on the Protein A elution buffer. Even more significantly, the acetic acid elution buffer resulted in the highest pool conductivity (~7 mS/cm), whereas the two glycine-based elution buffers resulted in pool conductivities close to 3 mS/cm. The lower conductivity load for the anion exchange step resulted in a reduction of HCP from 800 ng/mg to <5 ng/mg in the anion exchange pool for the two glycine buffers but only to 40 and 620 ng/mg, respectively, for the acetic acid and citrate elution buffers. Therefore, the glycine-based buffers offer the potential for a two-chromatography step process while the citrate buffer was eliminated because of its interference with the anion exchange process.9
The need to model the entire downstream process to understand the effects on the performance of each unit operation is illustrated in Table 6. These results show that there is a slight impact on the overall pool volume for the final adjusted Protein A pool (~3,300 L compared to ~3,000 L) and that this increase is not constrained by available tank volumes. Because the anion exchange step is operated in the flow-through mode, there is an overall increase in volume in the anion exchange pool tank with all conditions resulting in ~4,000 L, which again is not constrained by tank capacity. However, due to the higher ionic components present in the acetic acid elution pool, a significantly large volume of water is needed to adjust this pool (450 mL/L) for loading onto the subsequent cation exchange step compared to either of the glycine-based elution buffer pools where no conductivity adjustment is needed. This may represent a bottleneck for a manufacturing facility because the adjusted anion exchange pool requires a tank with twice the capacity. Therefore, the glycine-based buffers deliver a robust process with lower impurity levels while also removing downstream pool tank bottlenecks by reducing pH titrant volumes and pool conductivities.
Table 6. Impact of Protein A elution buffer in platform process B on conditioning operations required for downstream chromatography operations. Facility bottleneck is the load pool to the cation exchange step (marked in red).
The optimization of operating conditions in platform purification processes allows titers of up to 5 g/L to be accommodated in existing manufacturing facilities. We and others have shown that downstream bottlenecks can be eliminated through the use of conventional purification technologies and the optimization of minor facility equipment.5,10 These platform processes use conventional technologies, including affinity and ion exchange chromatography, without the need for novel unit operations as previously discussed.1,10,11 Chromatography and other standard unit operations offer the benefit of consistent process robustness, high productivity, facilitation of platform processes, low raw material costs compared to the overall cost of manufacturing, and regulatory convenience. Since each antibody may bring its own unique challenges to a platform process, such processes should be flexible where needed while still permitting the use of standard conditions wherever possible.
The authors would like to acknowledge Chris Dowd and Nuno Fontes for developing the facility fit model used in sections of this article and Jean Bender for discussions on process fit at multiple manufacturing facilities.
MELODY TREXLER-SCHMIDT, PhD, is a scientist, STEFANIE SZE-KHOO is an engineer, AMBER R. COTHRAN is an engineer, BINH Q. THAI is an engineer, and SANDY SARGIS is a senior research associate, all in late-stage purification; BENEDICTE LEBRETON, PhD, is a scientist, early-stage purification, BRIAN KELLEY, PhD, is vice president, bioprocess development, and GREGORY S. BLANK, PhD, is the director of late-stage purification, all at Genentech, Inc., South San Francisco, CA, 650.225.1956, firstname.lastname@example.org
1. Jagschies G, Grönberg A, Björkman T, Lacki K, Johansson HJ. Technical and economical evaluation of downstream processing options for monoclonal antibody (Mab) production. BioPharm Int. 2006 June (supplement); 10–19.
2. Shukla A, Hubbard B, Tressel T, Guhan S, Low D. Downstream processing of monoclonal antibodies—Application of platform approaches. J Chromatogr B. 2007;848:28–39.
3. Fahrner RL, Knudsen HL, Basey CD, Galan W, Feuerhelm D, Vanderlaan M, Blank GS. Industrial purification of pharmaceutical antibodies: Development, operation, and validation of chromatography processes. Biotechnol Genet Eng Rev. 2001;18:301–27.
4. Swinnen K, Krul A, Van Goidsenhoven I, Van Tichelt N, Roosen A, Van Houdt K. Performance comparison of protein A affinity resins for the purification of monoclonal antibodies. J Chromatogr B. 2007;848:97–107.
5. Gadam S, Bergevin J, Peers D. Basey C. Reconfiguration of a commercial biologics facility to a launch facility for new product introductions: Balancing process and engineering solutions. BioProcess International Conference; 2008 September 23–26; Anaheim, CA.
6. Ghose S, Nagrath D, Hubbard B, Brooks C, Cramer SM. Use and optimization of a dual-flowrate loading strategy to maximize throughput in protein-A affinity chromatography. Biotechnol Prog. 2004;20:830–40.
7. Matthews T, Wolk B. Inline buffer dilution and bioprocess bags: An integrated approach. IBC Biopharmaceutical and Development Manufacturing Summit; 2006 December 6–7; Orlando, FL.
8. Norling L, Chen Q, Blank, G. Establishing Robust Conditions for Clinical and Marketed Products, presented at ACS National Meeting, 2006 September 10–14; San Francisco, CA.
9. Kundu A. Strategies for maximizing clearance of impurities across AEX in the context of an integrated process. ACS National Meeting; 2007 August 19–23, Boston, MA.
10. Kelley B. Very large scale monoclonal antibody purification: The case for conventional unit operations. Biotechnol Prog. 2007;23:995–1008.
11. Kelley B, Blank G, Lee A. Downstream processing of monoclonal antibodies: Current practices and future opportunities. In: Gottschalk U, editor. Process scale purification of antibodies. Hoboken, NJ: John Wiley & Sons; 2009.