
Single-use membrane chromatography as a polishing option during antibody production is gaining momentum.
ABSTRACT
Monoclonal antibodies (MAbs) continue to dominate the biopharmaceuticals market and their demand is predicted to increase dramatically over the next decade. This demand generates pressure on manufacturers to develop robust and economical large-scale processes that deliver an extremely pure product. Increasingly, manufacturers are turning to disposable chromatography modules for their polishing operations, including membrane adsorbers as an alternative to packed-bed columns. These will significantly reduce downtime by eliminating the need for cleaning and validation. However, economic competition still remains between traditional chromatography and disposable systems. This article discusses the basis of membrane adsorber technology and compares the performance of Q membrane adsorbers and packed-bed columns in terms of contaminant removal, viral clearance, and process economy in antibody manufacture. With new and improved scale-down models, disposable Q membranes now offer competitive advantages over traditional packed-bed systems.
The monoclonal antibody sector is the fastest-growing segment of the biopharmaceuticals market, with a current value of $12 billion, which is expected to increase to at least $20 billion by the year 2010.1 There are currently 23 approved antibodies and antibody-related proteins on the market with hundreds more in preclinical and clinical development.1-3 Despite their hegemonic market position, MAbs are among the most complex and expensive pharmaceutical products to manufacture due to the strict regulatory principles applied to the production of protein-based pharmaceuticals. Over 40% of downstream purification systems are designed for MAbs produced in recombinant mammalian cells.1 Therefore, there is significant interest in finding ways to improve the economy and scalability of downstream processing without compromising purity and reliability.
MAbs represent a special product class within the biopharmaceuticals sector because similar purification steps can be used for all antibodies, regardless of their specificity. This generally involves the use of Protein A affinity chromatography as a primary capture step, because this molecule binds strongly and specifically to the constant region of most IgG class antibodies.4-5 The use of immobilized recombinant Protein A achieves 98% of product-related purity in a single step, significantly reducing the risk of protease contamination, and considerably decreasing the process volume. Thus, all subsequent steps are regarded as polishing operations (Figure 1).6 These downstream steps comprise a mixture of chromatography and microfiltration modules to remove degradation products and any remaining trace contaminants, such as host cell proteins (HCP), nucleic acids, pyrogens, leached Protein A, and viruses.5-7
The most popular choice among the polishing steps in current MAb manufacture is flow-through anion exchange chromatography (FT-AEX). This is because many contaminants, including viruses, nucleic acids, endotoxins, and some HCPs, tend to be negatively charged at neutral pH, and are therefore retained by the resin, whereas antibodies are positively charged under the same conditions and flow through efficiently.8-9 However, a major disadvantage of packed-bed FT-AEX is that to cope with the high-throughput of today's MAb manufacturing processes, the diameter of the columns needs to be very large to facilitate the necessary volumetric flow rates. As the production scale increases, certain capital and operational costs need to be considered (e.g., resin life studies, packing, cleaning validation, and storage validation studies). Thus, alternative polishing technologies become increasingly attractive.8,10-11
Principles of Membrane Chromatography
Unlike traditional column chromatography, which uses packed beads as the resin, membrane chromatography involves the use of thin, synthetic microporous or macroporous membranes that are chemically activated to fulfill the same function. An important mechanical difference between the two processes is that in membrane chromatography, the transport of solutes to their binding sites occurs mainly by convection, whereas pore diffusion is minimal compared with packed beads (Figure 2). This diminishes mass transfer resistance, reducing the time required for adsorption, washing, elution, and regeneration.8 Therefore, membranes potentially allow increased flow rates in large-scale processes, leading to an overall reduction in process time.8,11-12 The comparatively large pore size of membranes provides a higher binding capacity for large biomolecules such as viruses and DNA, even at high flow rates. Another advantage of using disposables is that exhausted membrane modules can be replaced with new ones rather than the traditional requirement to clean the columns, validate the cleaning procedure, and repack with resin. Various approaches for the purification of proteins using resins and membranes have already been reviewed.5,10
Unfortunately, although efficient membrane adsorption has been demonstrated for laboratory and pilot scale protein purification, several limitations remain to be overcome before such devices can be used routinely in processscale operations. Some of these limitations reflect structural artifacts, such as uneven pore-size distribution and variations in membrane thickness, which can be addressed through the use of multiple-layer configurations. However, one attribute that is proving intractable is the low binding capacity of such membranes for smaller molecules compared to column resins, which results from the lower surfaceto-bed volume ratio and flow distribution problems. Q anion exchange adsorber devices have been used for the preparation of vaccines and gene therapy vectors, as well as for endotoxin removal on a process scale.13-18 The low binding capacity is a serious disadvantage when using membranes in retention mode, but this problem can largely be ignored for processes where the flow-through mode is used. Flowthrough Q membrane adsorbers therefore provide an ideal solution to the requirement for a robust, highthroughput polishing step in antibody production, removing impurities, and including viruses, which are present at less than 1% concentration.10-11
Performance Comparisons
Several companies have carried out evaluation studies to compare the performance of membrane adsorbers and traditional packed-bed columns, demonstrating the competitive nature of Q membranes compared to Q columns. The standards used to measure the performance of each system are not directly comparable and care needs to be taken when reviewing the performance data. One important aspect is the measurement of process capacity, which in the case of column chromatography is usually expressed as grams of product per liter of resin, with a typical value of 50-100 g/L. A direct comparison between packed-bed columns and membrane adsorbers means comparing the volume of membrane to the volume of resin, even though the volume of a membrane must take into account its thickness, which has little impact on binding capacity and can be difficult to calculate accurately due to uneven thickness within a membrane sheet. Membrane vendors use the area of membrane to express process capacity, and also sell membranes by unit area rather than by unit volume. Therefore, it is more convenient and appropriate to express process capacity in grams of product per square meter of membrane.10-11
Since one of the great advantages of membrane adsorbers is fast processing time, flow rate is also an important comparator between column and membrane modules. In the case of columns, flow rate is calculated as linear velocity (cm/h), whereas several different units have been used for membrane chromatography; each is based on flux rate, analogous to the system used in cross-flow filtration. As for process capacity measurements, flow rate measurements such as membrane volume per minute and device volume per minute depend on membrane thickness. Therefore, it is able to measure flow rate as a linear velocity, calculated according to the cross-sectional area of the membrane.10-11 The maximum linear flow rate reported for packed-bed columns in process-scale AEX procedures is about 100-200 cm/h using Q-Sepharose Fast Flow (GE Healthcare).10 In contrast, linear flow rates of up to 600 cm/h can be achieved with functionally equivalent Q membranes. This occurs with no compromise in efficiency, since the large membrane pores provide a high binding capacity for viruses and nucleic acids even at these high flow rates. Another advantage is that much less buffer is used with membrane chromatography compared to conventional packed-bed chromatography.
A number of studies published since 2001 have demonstrated the potential of Q membranes for largescale processes. In the first such study, a 10-layer Sartobind Q membranestacked Swinny 13-mm-folder scaledown model (Sartorius AG) for an antibody process showed that the Sartobind Q membrane provided a 2.3 log reduction value (LRV) for murine leukemia virus (MuLV) at a flow rate of 620 cm/h and a capacity of 2000 g/L (or about 550 g/m^sup 2^).8 The LRV decreased as the antibody load increased. For example, 1.0 LRV was achieved at a capacity of 1100 g/m^sup 2^, 0.5 LRV was obtained at a capacity of 1650 g/m^sup 2^, and 0.3 LRV was reported at a capacity of 2755 g/m^sup 2^. In contrast, Q-Sepharose Fast Flow provided a viral clearance power of >5.1 LRV at a capacity of 50 g/L.
In a similar study, a Sartobind Q capsule module (Q-10 in.) was used for antibody production and a Sartobind Q 75 15-layer module was used for viral clearance during large-scale antibody production.19 The capacity was estimated at 450 g/m^sup 2^ with better viral clearance rates than described above. Approximately 4.0 LRV was reported for respiratory enteric orphan virus III (Reo-3), MuLV, and pseudorabies virus (PRV).
More recently, the use of Sartobind Q in a 15-layer format was reported in an antibody pilot plant producing material for a toxicity study.12,20 The scale-down model used the Sartobind Q75. The detailed study demonstrated that Sartobind Q75 has the capacity to remove HCP and viruses at a neutral pH and conductivity below 3 mS/cm. Under these conditions, the virus clearance results were excellent: 5.57 LRV for MuLV, 7.28 LRV for Reo3, 6.77 LRV for minute virus of mice (MVM), and 5.67 LRV for PRV.12,20 However, the process capacity was estimated at 480 g/m^sup 2^ or 1750 g/L membrane volume.
The performance studies outlined above suggest that the process capacity of Q membranes (up to 2 kg/L or 550 g/m^sup 2^) is significantly higher than that of Q resin (up to 100 g/L). From an economic perspective, however, this comparison does not take into account the relative cost of the media.10 A direct comparison of resins and membranes based on volume show that membranes are on average 15 times more expensive, and are in any case priced by area rather than volume. Ultimately, cost analysis indicates membrane adsorber process capacity needs to exceed 2 kg/m^sup 2^ or 8 kg/L to compete with packed-bed resins (i.e., a fourfold increase), as explained below.
Cost Analysis with the Original Scale-Down Model
The original scale-down model process capacity indicates that Q membrane chromatography is impressive in performance but economically unfeasible. To extend that analysis, consider a cost comparison between Q-Sepharose Fast Flow and Sartobind Q using a throughput of 13.5 kg of antibody, and assuming a water-for-injection (WFI) cost of $3/L (Table 1). The cost of 1 m^sup 2^ of membrane is about $2,400 but this would drop to approximately $2,000 at larger scales. The device is disposable and not reusable. Although this saves on cleaning and reuse validation, it does increase the raw material cost. To process 13.5 kg of antibody, the cost of membrane alone is about $54,000 at a loading capacity of 500 g/m2. This, together with the labor cost of buffers, generates a total cost for each cycle to be $58,588. The process capacity of Q-Sepharose Fast Flow commonly used for antibody purification is about 70 g/L. Thus, at $500/L, the initial cost is about $110,000 for 220 L of resin, which would be packed into a 100-cm column to a bed height of 25 cm. The Q resin can be reused for at least 100 cycles, thus, the resin cost per cycle is only $110,010. However, the cost of the resin is only one part of the true overall cost of the unit operation. Additional costs include labor and buffers (equilibration, wash, regeneration, and storage). The total cost is approximately $18,500 per cycle. Thus, the Q column unit operation is more cost-effective than the Q membrane unit operation at a process capacity of 500 g/m^sup 2^.
A careful comparison of the items listed in Table 1 shows how we arrive at the target process capacity of 2 kg/m^sup 2^ required to make Q membranes competitive with resins in a large-scale process. At this capacity, the total cost for one cycle, including the Q membrane ($13,500), instillation ($700), equilibration buffer ($1820), and wash buffer ($620), is about $16,640. This makes the membrane adsorber very competitive against the column unit operation cost per cycle of $18,500. This comparison does not include the column packing and validation costs incurred with resin reuse. Therefore, it is necessary to demonstrate a process capacity of 2 kg/m^sup 2^ on a large scale and validate this with a scale-down model while still achieving excellent quality parameters such as DNA and viral clearance.
A New Scale-down Model
Currently, scale-down models (including Sartobind Q75) are not suitable for viral clearance studies because of process capacity limitations generated by operational backpressure with the scale-down devices. The liquid flow path in scale-down models is not the same as that found in the large-scale capsule, and thus tends to generate extremely high operational backpressure when a high flow rate is applied (450 cm/h). The scale-down model, such as Sartobind Q75, is made with stacks of several flat sheets, and the axially directed velocity of a mobile solution is much faster in the center of the membrane unit than near the edges of the adsorptive bed.10-11 In contrast, radial flow adsorbers, prepared by spirally winding a flat sheet membrane over a porous cylindrical core, are used in a capsule for large-scale processes.10 Distorted or poor inlet flow distribution has less of an impact in the scaled-up capsule, thus leading to a smaller operational backpressure drop for large-scale operations.10
Scalability and operating backpressure are major issues for the membrane absorber scale-down model. Extremely high operational backpressure in the scale-down model reduces the capacity considerably, resulting in the use of an oversized membrane at process scale. This may cause errors in economic calculations and undervalue the economic benefits of membrane adsorbers. Large-scale membranes do not have backpressure issues. However, the use of large membranes in a scale-down model is also impractical for viral clearance studies because of the high cost of viruses and feedstock.
If Q membrane adsorbers are made from diethylaminoethyl (DE) cellulose material, the conductivity of the operation solution directly affects operational backpressure. It is also important to determine the effects of operation temperature on membrane backpressure drop because a relatively low operation temperature would increase the viscosity of an antibody solution. In addition, some buffers are temperature-sensitive. At a low temperature, the low conductivity can generate a high operational backpressure leading to a low process capacity. These factors have necessitated a redesign of the scale-down model.21
The new scale-down model mimics the liquid flow path found in the large-scale modules. When the newly designed Q125 was examined at a flow rate of 450 cm/h with a process capacity of 3 kg/m^sup 2^, excellent viral clearance was achieved with four model viruses (Table 2). A maximum operational backpressure of <16 psi was observed. This new device also achieves performance parameters comparable to those of the larger Sartobind Q units, with a process capacity of >3.6 kg/m^sup 2^ (total membrane surface) or >13.2 kg/L (membrane volume) at a maximal flow rate of 600 cm/h with an operational backpressure of <18 psi. A LRV of >5.6 was achieved for MuLV21. Since such high process capacity required a significant amount of feedstock and virus, the clearance studies with PRV, MVM and Reo-3 viruses were not carried out. The clearance studies with the four model viruses were run at a flow rate of 450 cm/h and process capacity of 3 kg/m^sup 2^.
The fact that the viral clearance performance at high flow rate and high capacity is similar to the MuLV LRV at 450 cm/h and a capacity of 3 kg/m^sup 2^ suggests that a similar viral clearance value could be achieved at a higher operational flow rate and capacity for the other three model viruses.
Mass balance data in a viral clearance study is extremely important to demonstrate efficient virus removal by a membrane adsorber device. For example, 100% recovery was obtained for PRV (n = 2), Reo-3 (n = 2) and MVM (n = 2) virus particles with particle sizes of 18-200 nm when the membrane was stripped with 1 M NaCl, demonstrating efficient charge capture for the three model viruses. When the membrane was treated with high salt, an average of only 70% viral recovery (n = 3) was achieved for MuLV. This may reflect the possibility that 1 M NaCl inactivated some MuLV particles, they irreversibly bound to the membrane or were inactivated by mechanical stress during the process of adsorption and desorption.
These concepts and parameters have been successfully applied to purify four lots of recombinant human antibody at the 2000-L scale. The impurities, including HCP, DNA, and leachable rProtein A, were below the limit of detection following Q membrane purification. Figure 3 shows the typical MVM viral clearance power when the Sartobind Q 125 device is used. These data support the replacement of Q packedbed columns with membrane adsorbers as a polishing step in flow-through chromatography for process-scale antibody production.
Cost Analysis with the New Scale-down Model
At a process capacity of 3 kg/m^sup 2^, Q membrane chromatography is very competitive with resin chromatography when using $3/L as the cost of WFI in the calculations (Table 1). Although this is the cost currently utilized by contract manufacturers, the estimated WFI costs for large biotech firms can be as low as $0.2/L Table 3 compares the total cost for membrane operation at a capacity of 3 kg/m^sup 2^ versus Q column operation at a capacity of 70 g/L using only $0.2/L for WFI cost. In this scenario, column unit operation is more cost efficient ($9,168 per cycle) than membrane unit operation ($11,738) even at the high process capacity. The membrane price dropped only $400 with the decrease in WFI cost whereas the column saved more than $9,000. This highlights the difficulty in comparing the real cost of a unit operation and the danger of minimizing something as common as WFI. This calculation also did not take into account the cost of tanks to contain the extra buffer needed for the column.10
For commercial manufacturing, column chromatography has a supplemental upfront cost that can be grouped into three main categories: development, manufacturing, and validation costs. Development and validation are performed once for the Q column, and these studies are not needed subsequently as long as the columns are run in exactly the same format, including bed width and height, buffers, and operation conditions such as flow rate and cleaning-in-place. Upfront manufacturing costs include column hardware and packing equipment. In the current work, 40 lots per year over a 10-year production time were used for the comparison between Q columns and Q membranes. This is a reasonable time period given product patents and process changes. The 40 lots per year are designed for production efficiency or product demand and four columns are required for the 10-year production based on 100 runs (cycles) per column. Obviously, some products are run more frequently and some less frequently. If more cycles are run between process changes, the initial investment would be averaged over more runs.10
For a ten-year production time or 400 cycles, development activities are not required for membrane chromatography and about 1,800 m^sup 2^ of Q membrane is required based on a process capacity of 3 kg/m^sup 2^. This disposable system does not incur most of the initial investment costs listed above because the membrane is single use only. Moreover, the validation cost is minimal and no packing studies are required. After summarizing all the costs for column chromatography such as resins, buffers, labor, and validation studies, industry has the potential to spend more money on column operations than Q membrane operations.10
Potential Limiting Factors of Membrane Adsorbers
Even with improved scale-down devices, the operational backpressure for Q membrane chromatography will always influence process capacity because of the high viscosity of the solution. However, viscosity depends on factors such as operational temperature, pH, conductivity, and protein concentration if the solution contains protein.
Temperature is likely to be the most important factor because viscosity increases as the temperature falls, and low temperatures are preferred to maintain product stability. The viscosity of an antibody solution is higher than the viscosity of the control buffer solution at the same low temperature. At 4�C, the viscosity of an antibody solution (3 mg/mL at pH 7.2 and about 4 mS/cm) is much higher than at 8�C or at 13�C. 21 High viscosity at low temperature significantly increases operational backpressure and results in lower flow rates. At a low temperature, even smaller particles can generate significant backpressure. Indeed, at room temperature, the operational backpressure is independent of virus particle size (a similar backpressure curve is found for Reo-3, PRV, and MuLV particles, which range in size from 18-200 nm). Although a low operational temperature reduces membrane process capacity because of the higher operational backpressure created by the higher viscosity of protein and buffer solutions, our data suggest that the low temperatures have very minimal impact on viral clearance with the backpressure.21
Another limiting factor is found when the buffer pH is close to the pI value of protein. The viscosity appears to peak at the pI of the molecule studied.21 Other limiting factors, including conductivity, need to be monitored. Low conductivity causes more molecular interaction to occur between antibodies, thereby increasing the solution viscosity.
Conclusions
The use of disposable Q membrane chromatography modules is becoming increasingly attractive in largescale antibody processing as a result of impressive performance and economic comparisons with traditional anion-exchange chromatography. Membranes benefit from faster processing times due to convective flow, higher process capacity, and up to a 95% reduction in buffer use, all helping to reduce costs in head-to-head comparisons. Single-use membrane chromatography is robust and simple to use, with no column packing or subsequent cleaning validation required. In particular, the Q membrane is an effective and robust unit operation to remove viruses.21 Q membranes have a relatively short history and we can anticipate further improvements with a new generation of even smaller scale-down models showing better host-cell protein removal capacity. We conclude that Q membrane chromatography in flowthrough mode is a viable alternative to Q column chromatography as a polishing step for late-stage processscale antibody production.
[Sidebar]
A benefit of using disposables is that exhausted membrane modules can be replaced with new ones instead of cleaning the columns.
[Sidebar]
Because one of the great advantages of membrane adsorbers is the fast processing time, flow rate is an important comparator between column and membrane modules.
[Sidebar]
Mass balance data in a viral clearance study are extremely important to demonstrate efficient virus removal by a membrane adsorber device.
[Sidebar]
Single-use membrane chromatography is robust and simple to use, with no column packing or subsequent cleaning validation required.
[Reference]
References:
1. Daemmrich A, Browden ME. A rising drug industry. Chem Eng News. 2005; 6:28-42.
2. Walsh G. Biopharmaceuticals: approvals and approval trends in 2004. BioPharm Int. 2005; 18(10): 58-65.
3. Chadd HE, Chamow SM. Therapeutic antibody expression technology. Curr Opin Biotechnol. 2001; 12:188-194.
4. Fahrner RL, Whitney DH, Vanderlaan M, Blank GS. Performance comparison of protein A affinity-chromatography sorbents for purifying recombinant monoclonal antibodies. Biotechnol Appl Biochem. 1999; 30(2):121-8.
5. Fahrner RL. Industrial purification of pharmaceutical antibodies: development, operation, and validation of chromatography process. Biotechnol Gen Eng Rev. 2001; 18:301-327.
6. Zhou J, Tressel T, Hong T, Solamo F, Dermawan S. Utilize current separation technology to achieve a cost-effective robust process for recombinant monoclonal antibody production. Antibody World Summit. Jersey City, NJ; 2005.
7. Tressel T. Development of a generic platform and use of statistically designed experiments to enable rapid development of several antibodies and increase throughput for first in human antibodies. Antibody production and downstream processing, IBC conference. San Diego, CA; 2004.
8. Knudsen H, Fahrner R, Xu Y, Norling L, Blank G. Membrane ion-exchange chromatography for process-scale antibody purification. J Chrom A. 2001; 907:145-154.
9. Kemp G, O'Neil P Large-scale production of therapeutic antibodies: considerations for optimizing product capture and purification. Antibodies (Subramanian G, Ed.). 2004; 1.2 vols, Plenum Publishers, New York.
10. Zhou J, Tressel T. Basic concepts in Q membrane chromatography for largescale antibody production. Biotechnol Prog. 2006; 22:341-349.
11. Zhou J, Tressel T. Q membrane chromatography as a robust purification unit for large-scale antibody production. Bioprocess Int. 2005; 3:32-37.
12. Zhang R, Bouamama T, Tabur P, Zapata GA. Q membrane chromatography application for human antibody purification process. BioProduction. 2004; http://www.abgenix.com/ documents/ Bioproduction2004. pdf.
13. Deshmukh RR, et al. Large-scale purification of antisense oligonucleotides by high-performance membrane adsorber chromatography. J Chrom A. 2000; 890:179-92.
14. Grunwald AG, Shields MS. Plasmid purification using membrane-based anion-exchange chromatography. Anal Biochem. 2001; 296:138-41.
15. Zhang S, Krivosheyeva A, Nochumson S. Large-scale capture and partial purification of plasmid DNA using anion-exchange membrane capsules. Biotechnol Appl Biochem. 2003; 37: 245-9.
16. Teeters MA, Conrardy SE, Thomas BL, Root TW, Lightfoot EN. Adsorptive membrane chromatography for purification of plasmid DNA. J Chrom A. 2003; 989:165-73.
17. Kuiper M, Sanches RM, Walford JA, Slater NK. Purification of a functional gene therapy vector derived from moloney murine leukaemia virus using membrane filtration and ceramic hydroxyapatite chromatography. Biotechnol Bioeng. 2002: 80:445-53.
18. Specht R, et al. Densonucleosis virus purification by ion exchange membranes. Biotechnol Bioeng. 2004; 88:465-73.
19. Gallher R Fowler E. Validation of impurity removal by the CAMPATH-1H biomanufacturing process. IBC's Biopharmaceutical Production Week; 2001, http://www.sartorius.com/ index.php?id=532&v=l&title=&year= &scale=&format=&membrane=&mole cules=&type=.
20. Zhang R, et al. Viral clearance feasibility study with Sartobind Q membrane adsorber for human antibody purification. BioProduction; 2004, http://www.sartorius.de/en/ biotechnology/laboratory/products_a pplications/membrane_adsorbers/ literature/pdfs/Zhang_2004_Viral_cle arance_feasabiltity_study_Sartobind_ Q.pdf.
21. Zhou J, et al. Scale-down model design and viral clearance study of Q membrane for antibody process capacity. J Chrom A. 2006; 1134:66-73.
[Author Affiliation]
JOE X ZHOU is principal scientist, PPD, TIM TRESSEL is director, PPD, SAM GUHAN is executive director, process and analytical sciences, at Amgen Inc., One Amgen Center Drive, Thousand Oaks, CA, USA. 805.313.4065. joez@amgen.com.