Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Improvements in upstream production have boosted productivity in the biomanufacturing industry, but this is leading to bottlenecks in downstream processing as current technology platforms reach their limits of throughput and scalability. Although chromatography remains an indispensible component of downstream processing due to its simplicity and high resolving power (The Good), there is virtually no economy of scale effect so more product translates almost linearly into greater production costs. Bind-and-elute processes (such as the initial capture step in antibody manufacturing) are volume-driven and therefore have knock-on effects that impact on the entire production facility since the space required for preparation, storage, and cleaning steps has to be similarly adapted (The Bad). During long-term operations with multiple cycles, thorough cleaning is necessary to prevent progressive fouling and microbial contamination (The Ugly). Innovative solutions are required, which may include revisiting simpler and less expensive separation technologies, the use of disposable modules, and the integration of improved processes that are scalable to cope with increased demands. Among the alternatives that have been put forward, membrane adsorbers are beginning to make a real impact on the industry, particularly for flow-through applications such as polishing and viral clearance.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Remarkable progress has been made in the biomanufacturing industry over the past 20 years, with the yield of recombinant proteins increasing from tens of milligrams per liter of cell culture to up to 10 grams per liter for some monoclonal antibodies in early stages of development (1). As batch volumes match this upward trend, we face the likelihood of 50- to 100-kg batches of protein becoming a realistic prospect in the not so distant future (2). This has created a bottleneck in downstream processing, where technological advances have failed to keep up with productivity increases upstream.
The problem here is two-fold. First, upstream production depends predominantly on biological processes such as gene expression, cell growth, and protein stability, which have been improved largely through empirical testing and the selection of better performing cell lines and media in a manner analogous to Darwinian evolution. In contrast, downstream processing depends on chemical and physical interactions that cannot be selected in the same manner; instead they must be modeled, tested, and developed from first principles (3). Second, the upstream phase of production still has much room for further improvement, as long as cell density and protein expression/solubility do not exceed their physical limits. There is an economy of scale here because greater productivity upstream does not necessarily imply commensurate increases in costs, and this also applies to some parts of downstream processing (e.g., centrifugation, depth filtration, and cross-flow filtration). However, it does not apply to all downstream steps, including chromatography.
Since the aim of downstream processing is to purify the active pharmaceutical ingredient and separate it from any potentially harmful chemical or biological residues, chromatography−due to its high selectivity−is currently indispensible as a unit operation and most process trains involve at least two distinct, orthogonal chromatographic steps. Nevertheless, these specific operations have become important technological and economic bottlenecks as companies strive for increased manufacturing productivity while simultaneously aiming to reduce the cost of goods (4).
The shift in perspective from upstream to downstream processing has become more evident in the past few years as experts begin to acknowledge that improved productivity cannot rely on more efficient upstream production alone and that improvements downstream will not result from tinkering with single unit operations but must apply to the whole integrated process. If any particular step is left out from process development, then it will almost certainly become the next production bottleneck, as has already happened with process-scale chromatography. In 2006, a leading survey of industry opinion revealed that most companies considered downstream processing to be more important than the upstream production phase for productivity and process economy, a reversal of the results in 2005 (5−7). The same survey highlighted disposable membrane technology as one of the most promising ways to address this challenge, at least for flow-through operations (6). Modern process development must follow a rational strategy that allows no redundancy, includes the use of orthogonal separations wherever possible, and maximizes integration, each step serving a defined purpose (8). Generic processes should be compiled from modules that are predeveloped at scale and allow process train assembly in weeks rather than years (9).
Column Chromatography: A Question of Scale
Downstream processing would never have developed as an individual sector of the bioprocessing industry without chromatography, whose inherent simplicity and selectivity has made it the key enabling technology in all bioseparation processes, hence “The Good” (10). In traditional chromatography, a glass or steel column is packed with a resin (stationary phase) comprising porous beads made of a polysaccharide, mineral, or synthetic matrix derivatized with specific functional groups. The functional groups exploit different separation principles, such as size, charge, hydrophobicity, or affinity for particular ligands, to resolve target proteins. The resin may bind the target protein while eluting contaminants (bind-and-elute or capture mode) or may bind contaminants while eluting the target protein (flow-through or polishing mode) (11). The highest resolution and therefore the highest purity is achieved when orthogonal principles are applied. For the purification of monoclonal antibodies, this usually involves a native or recombinant Protein A capture step combined with a cation exchange column for intermediate purification and an anion exchange column operating in flow-through mode to remove negatively charged impurities such as DNA, host cell proteins, endotoxins, and endogenous and adventitious viruses (Figure 1). However, companies such as Medarex Inc. have developed a non-Protein A platform for antibody purification with optimized yields and minimized cycle times (Figure 2). The evolution of this process involved replacement of the Protein A step with an equally efficient cation exchange resin and reduction of the overall number of column chromatography steps from three to one by replacing the anion exchange flow-through column with a disposable anion exchange membrane (12).
What makes the combination of resin and membrane chromatography steps such a good idea? On closer inspection, capturing and polishing, although following different principles, are subject to the same limitations. However, the consequences are different as discussed below.
Capturing. Despite the essential role of packed-bed chromatography in biomanufacturing, high-titer processes impose practical limitations that suggest that the true bottleneck in recovery processes is the first adsorptive column, reflecting poorer performance with increasing scale (13). Very large columns can be as robust and reliable as smaller ones, but there is no economy of scale with such devices because the additional cost of resins, buffers, and other consumables outstrips any savings made by increasing the productivity. Indeed, the bottleneck in process-scale chromatography negates any advantages of scaling up earlier process units, since capture steps are driven by mass rather than volume and savings made upstream therefore do not translate into increased productivity during purification. Larger columns also impact directly on facility layouts, costs and infrastructure because the space and buffer volumes for all steps, including preparation and cleaning, must be adapted to compensate. As a consequence, pool and buffer volumes are serious limitations when it comes to the introduction of high-titer processes into existing facilities, hence “The Bad”.
The physical constraints of process-scale chromatography must also be considered. The largest biochromatography columns in use today are 2 m in diameter and are operated at a 10- to 20-cm bed height, which is fast becoming a limitation in large-scale processes. The requirements to capture 100 kg of monoclonal antibody on a Protein A affinity column in a single cycle may serve as an example. If one assumes an optimal loading capacity of 50 g/L, this would require 2,000 L of resin, which would need to be packed in a column with a diameter of 3.2 m at a bed height of 25 cm (4). The need for oversized columns can be circumvented if several cycles are used to process a single batch (and this is common industrial practice for the Protein A capture step for antibodies). This, in return, reduces the operational lifetime of the resin and requires more buffer, therefore introducing additional costs. A column containing 2,000 L of resin would require 50,000 L of buffer for equilibration, elution, and cleaning, and these buffers need to be prepared, stored, and disposed of in an appropriate manner (4). As well as productivity and economic issues, large columns also suffer from scale-related packing problems such as hysteresis, edge-effects, and resin compression, which result in unpredictable fluid distribution and pressure drops. Hysteresis is a phenomenon in which uneven resin packing leads to differences in fluid flow rates throughout the column, especially around the edges where looser packing density encourages the mobile phase to flow faster and can, in the worst cases, result in the so-called “cheese-cake” effect where the resin collapses and leaves large gaps. Resin compression reduces its porosity and thus influences the pressure drop across the column. Other risks not so openly discussed include progressive fouling of the resin due to the aging of feed material during long load cycles and/or insufficient cleaning accompanied by microbial contamination, hence “The Ugly”.
Although there is considerable debate about this issue (13−15), one can conclude that bind-and-elute operations for monoclonal antibodies are approaching their physical limits. For the time being, there is no alternative to packed-bed chromatography for such operations due to its selectivity and ease of use. In the long run, limitations in productivity and thus process economy may better be addressed by low-technology alternatives as discussed later.
Polishing. The limitations of column chromatography outlined above also apply to flow-through applications, where the objective is to capture impurities rather than the product. Flow-through anion exchange chromatography as used in the purification of monoclonal antibodies illustrates the problem well. In resin-based anion exchange media, the transport of solutes to their binding sites relies on pore diffusion, but the contaminants are often large molecules, e.g., DNA and viruses, which do not readily diffuse into the pores (Figure 3a). This causes mass transfer resistance and lowers the column efficiency because large molecules can only bind to the outer surface of the bead and longer residence times are required to find binding ligands inside the resin particle (16, 17). This challenge may be addressed by using a greater column bed height and/or reducing the linear flow rate, both of which impact on overall productivity. Therefore, to keep up with process demand, most traditional polishing steps operate at a flow rate of between 100 and 150 cm/h and use significantly oversized columns to accommodate this. For flow-through applications, these limitations can be overcome by changing the chromatographic support structure from a resin to a membrane, a topic explored in detail later.
What Are the Alternatives?
Various alternatives have been put forward either to replace column chromatography or to reduce the load of impurities in the feedstream so that one or more chromatography steps can be eliminated. Some of these alternatives apply to capture steps whereas others represent innovate filtration and chromatography formats that apply to polishing operations.
Examples of simple technologies that have been revisited recently include flocculation, precipitation, two-phase extraction, and crystallization (4, 13, 18, 19). Flocculation and precipitation can be used in combination with conventional cell separation techniques such as centrifugation and microfiltration to remove residual particulates and soluble impurities that might otherwise increase the burden on downstream polishing steps, therefore allowing the number of chromatography processes to be reduced. Flocculation is particularly useful for removing sub-micrometer-sized particulates that increase the turbidity of the feedstream and usually need to be removed with an interstitial dead-end filtration step to prevent column fouling. Polymers can also persuade small particles to clump together, facilitating their removal by centrifugation, but innocuous inorganic alternatives such as calcium chloride and potassium phosphate have also been used successfully (20, 21). Crystallization is another inexpensive technology that in some cases can simultaneously purify, concentrate, and stabilize a recombinant protein and provide a useful delivery mechanism (22). Several commercial processes for therapeutic protein manufacture involve a crystallization step that replaces chromatography (23, 24), although applications in antibody purification are currently limited because of low yields, the inherent complexity of the process, and difficulties with process control. Some engineering-based solutions have also been implemented, including the deployment of radial flow chromatography and simulated moving bed chromatography to increase throughput while reducing buffer usage (25).
Examples of higher-end technologies for the replacement of column chromatography include the use of charged ultrafiltration membranes and membrane-adsorbers. The separation of proteins by charged ultrafiltration membranes was first reported by Nakao and colleagues (26) who used polyethersulfone ultrafiltration membranes bearing either positive or negative charges to separate myoglobin and cytochrome c by setting the buffer pH near the pI of one or other of the proteins, allowing efficient separation even though the proteins were of similar sizes. The basis of protein separation using charged filtration devices has been studied intensively by Zydney and van Reis (27, 28), who have developed positively charged ultrafiltration membranes that can separate antibodies from CHO cell impurities including host cell protein (29). The most obvious progress has been made in the development of alternative chromatography formats such as monoliths and membrane absorbers (30, 31). Here we concentrate on the use of membrane chromatography (32) as a potential solution to circumvent the limitations of resins for flow-through applications.
Membranes are already integral to many bioprocesses because they can be used as disposable modules, but thus far their principal application has been filtration rather than chromatography. Disposables are becoming more important in bioprocessing as confirmed in the industry survey discussed above (6). For many unit operations, particularly filtration and media/buffer storage, disposable devices have been in common use for quite some time because they save on cleaning and validation costs. However, disposables have other benefits: they save time, provide flexibility, and streamline process development (33). The ability to replace each module completely makes it easier to assemble process trains for new products in existing premises without worrying about cross-contamination, although there can be additional validation burden because of leached materials.
Interest in membrane chromatography is growing because of the success of disposable membrane filters, but there is still a lack of appreciation of the many advantages membrane devices offer in downstream processing (34, 35). In contrast to the resin-based flow-through processes described above, membrane chromatography involves the use of thin, synthetic microporous or macroporous membranes that are stacked 10−15 layers deep in a comparatively small cartridge. The footprint of such devices is much smaller than columns with a similar output. Membrane devices in a range of sizes are available from suppliers such as Millipore (Intercept), Pall (Mustang), and Sartorius-Stedim Biotech (Sartobind), with functional groups equivalent to the corresponding resins, e.g., membranes containing activated quaternary ammonium groups for anion exchange. The use of membrane devices results in the complete elimination of cleaning and validation, a major expense in downstream processing. FDA regulations require the cleaning, maintenance, and sanitization of fixed equipment and piping at appropriate intervals to prevent malfunctions and contamination, but this is unnecessary when fouled or exhausted modules can simply be swapped out and replaced with new ones.
Another significant functional advantage of membranes over resins is that the transport of solutes to their binding sites occurs mainly by convection, while pore diffusion is minimal (Figure 3b). Because of these hydrodynamic benefits, membrane adsorbers can operate at much greater flow rates than columns, considerably reducing buffer consumption and shortening the overall process time by up to 100-fold (36). The use of membrane adsorbers can be viewed as the equivalent of shortening traditional columns to near zero length, allowing large-scale processes to run with only a small pressure drop at very high flow rates. For example, polishing with an anion exchange membrane can be conducted with a bed height of 4 mm at flow rates of more than 600 cm/h, providing a high frontal surface area to bed height ratio (Figure 4). Small-volume disposable membrane chromatography devices can now handle more than 10 L/min/bar/m2. Even at these high flow rates, the membrane pores provide adequate binding capacity for large biomolecules such as viruses and DNA, and they play an important role in the overall viral clearance strategy for antibody purification (37−39). Functionality and integrity tests demonstrate that even with bed heights of less than 1 cm, membrane stacks are robust and reliable.
As shown in Table 1, an important advantage of membrane chromatography is the linear scale-up for important parameters such as frontal surface area, bed volume, flow rate, and static binding capacity, while normalized dynamic capacity remains fairly constant at 10% or complete breakthrough (Figure 5). It is thus apparent that membrane devices used for flow-through applications can be scaled up with none of the attendant disadvantages of column resins, making the goal of polishing 100-kg batches of antibody entirely possible without oversizing.
Table Table 1.. Scale-Up with SingleSep Q Membrane Chromatographya
frontal surface area (cm2)
scale-up factor for flow rate
rec flow rate (L/min)
bed vol (mL)
min static binding capacity (g) (release test)
dynamic capacity at 10% (mg/mL)
dynamic capacity at 100% (mg/mL)
a Parameters such as frontal surface area, bed volume, flow rate and static binding capacity scale up in a linear fashion (assuming constant bed height of 4 mm). Normalized dynamic BSA binding capacity remains constant at a given breakthrough (values shown at 10% and 100%; see also Figure 5). Data from Sartorius-Stedim Biotech.
Membrane Chromatography Performance
The concept of membrane chromatography arose from the combination of high mass transfer and low-pressure resistance conferred by an open membrane structure. The first membrane adsorbers suffered from problems related to both adsorptive capacity and device performance, e.g., low loading capacity, membrane fouling and suboptimal fluid distribution leading to a substantial performance loss during scale-up (40, 41). However, these issues have been largely addressed by the development of more suitable membranes, improved surface chemistries and the optimized design of membrane devices. In process-scale operations, 15-layer devices are commonly deployed and these achieve excellent contaminant removal and viral clearance results, routinely reducing DNA and host cell protein (HCP) below detection levels and achieving log reduction values (LRVs) of >5 in virus spiking studies. For example, a flow-through membrane chromatography case study designed to reflect process-scale conditions and performed with a 3.5 mL/125 cm2 spiral wound scale-down device achieved LRVs of >5 for four model viruses (Table 2). In this experiment the mass balance could be closed by quantitative elution of virus in three of four cases, proving the adsorptive mechanism of virus removal as opposed to a size exclusion effect. As part of the study, the virus-binding capacity was investigated by increasing the sample load up to 3 kg antibody/m2 (10.9 kg antibody/L) and there was no visible compromise on LRV for minute virus of mice (MVM) during the run (34). In a follow-up experiment, the load was increased to 10 kg antibody/m2 (36 kg antibody/L) and LRVs of >6.81 and >7.21 were achieved for MVM (34). Comparable LRVs have been demonstrated at pH 7 with a 5% virus spike (42). LRVs of >5 have also been achieved for four model viruses using Q-membranes at high conductivity (up to 15 mS/cm) (43). In this test, anion exchange membranes were used in addition to affinity capture and low pH inactivation, and were followed by a nanofiltration step, resulting in a cumulative LRV of >21 for MuLV and >7 for Reo3 and PPV (43).
Table Table 2.. Membrane Chromatography Spiking Study with Four Model Virusesa
LRV run 1
LRV run 2
virus recovery (%)
a Test substance was a human monoclonal antibody (5−9 g/L), pH 7.2; 4 mS/cm; 1% spike; 450−600 cm/h. Data from ref 50.b MVM: minute virus of mice. Reo-3: reovirus Type III. MuLV: murine leukemia virus. PRV: pseudorabies virus.
6.03 ± 0.21
6.03 ± 0.20
7.00 ± 0.31
6.94 ± 0.24
5.35 ± 0.23
5.52 ± 0.27
5.58 ± 0.28
5.58 ± 0.22
These performance studies confirm that packed columns and membrane chromatography devices are both capable of trace contaminant removal and virus clearance in polishing applications. Apart from the handling concept the main difference between the two formats is load capacity at flow rates acceptable for large-scale manufacturing, with multilayer Q membranes achieving a much greater productivity compared to equivalent volumes of resin with no loss of performance in contaminant and virus removal. For this reason, the volume of a disposable membrane device is typically 5% of that required for a conventional column, which needs to be oversized to accommodate the volumetric flow rate.
Retrospectively, the most serious limitation for the implementation of membrane chromatography has been unrealistic expectations and unclear positioning. It should be pointed out that membrane chromatography is restricted to simple on-or-off binding scenarios that are typically found in niche applications such as polishing or capturing large molecules from diluted feed streams. Packed-bed chromatography is still the preferred unit operation for capturing molecules of <200 kDa, especially when peak cutting, gradients, etc. are required for the separation of closely related species. Selectivity should be similar for both concepts, as long as the same ligand is used. However, for larger molecules (including most of the anticipated contaminants in antibody manufacture) there is higher capacity and faster processing, which is why membranes are better for flow-through applications in biomanufacturing (Figure 6).
Head to Head: Counting the Cost
The important remaining question is how membrane devices compare with columns in terms of cost, both fixed (capital) costs and variable (running) costs. Capacity and disposability are critical factors to consider when calculating unit operation costs for new processes. Although membrane devices clearly have a higher throughput, a direct comparison of resins and membranes based on volume shows that disposable membranes are currently more expensive. This must be balanced, however, against the reduced size of membrane devices, which also reduces buffer requirements, makes the process time shorter, and brings along all the other benefits of a disposable technology (33).
A 10-year cost model (44) showed that Q-membrane chromatography was economically unfeasible compared to Q-resin columns at a process capacity of 500 g/m2 (equivalent to about 1.8 kg/L) mainly due to the cost of membrane devices. The model was based on an upstream CHO platform featuring a 15,000-L bioreactor with a yield of 1 g/L antibody. This generates a load of 13−15 kg of antibody per batch, which would require a 220-L column or a 1.6-L membrane device based on typical performance standards. The model assumed that up to 40 batches could be run in a year, with the column resin replaced after each 100 cycles. Therefore, the column would need to be repacked with resin four times during the process lifetime, whereas 400 membrane devices would be required over the same period. The model suggested that capacity would need to increase above 2 kg/m2 (7.2 kg/L) to become competitive. Capital costs for chromatography hardware were not considered.
Another cost model (45, 46) suggests that membrane chromatography could break even with resins at a load of just 2 kg/L. With a load capacity of 10 kg/L, the membrane-based process costs only one-fifth as much as an equivalent operation using resins (Figure 7). This cost of goods model was based on the use of 10 in. Q ion-exchange membrane devices (180 mL volume). The values of 10 kg/L and 2 kg/L were considered typical for a relatively pure (late stage) feed stream after intermediate polishing and a less pure (earlier stage) feed stream, after clarification and capture by Protein A chromatography. The unique aspect of the model was its consideration and separation of all direct and indirect costs into four major categories: capital equipment, consumable equipment and media, consumable chemicals and materials, and labor. The fixed capital cost was the most significant in the case of column chromatography (nearly 19,000 EUR per batch) while that of membrane chromatography is less than 2,500 EUR. Consumables were important in both forms of chromatography, with resins accounting for the bulk of costs in column chromatography and disposable devices accounting for most of the costs in membrane chromatography. As might be anticipated, the cost of consumable equipment and media is higher for membrane chromatography because the membrane device needs to be replaced after each batch while the column resin can be cleaned and regenerated and the costs therefore spread over 100 cycles. However, the consumption of membrane is lower at the higher loading capacity since fewer capsules need to be used. So, whereas at 2 kg/L load the consumable equipment represents nearly 60% of the entire process cost, when capacity is increased to 10 kg/L this falls to less than 20%. Overall, disposables are only 11% more expensive than column chromatography at the highest loading capacity considered in the model. The use of other consumables is much higher in the case of column chromatography because the resin needs to be washed and regenerated and the large size of the column demands higher volumes, and the cost of labor is approximately four-fold higher because of buffer preparation, cleaning, validation, maintenance of equipment, and quality work.
Higher upstream productivity presents a challenge to chromatographic unit operations, and there is no unified solution for the capturing and polishing steps. While capturing will remain the domain of packed-bed columns with little (but growing) substitution pressure from low-technology alternatives, polishing is currently in a technology transition phase. Membrane chromatography offers a cost-effective alternative to traditional packed-bed chromatography in flow-through operations, such as polishing for the removal of viruses and contaminants in antibody manufacture. They are particularly attractive at larger scales, where columns suffer from the rising costs of resins and buffers and fall foul of scale-related packing issues that reduce column efficiency. With devices up to 5 L, membrane adsorbers can polish 100-kg batches of antibodies and can thus be operated in high-titer processes while providing other advantages through their disposability. The impact of disposable membrane chromatography on virus clearance procedures represents an important technological advantage, since studies have shown LRVs of >5 with model viruses even at high loads without the need for carry-over studies. Virus elution studies demonstrated an adsorptive mechanism that is orthogonal to any filtration process based on size exclusion. Membrane chromatography thus provides an advantage that is now increasingly important given the new EMEA guidelines on virus clearance studies, which indicate that at least two orthogonal steps for the clearance of enveloped and non-enveloped viruses are mandatory for phase 1 studies (47).
Since the advantages of membrane chromatography have only been realized in polishing applications so far, the open question is how to address other bottlenecks, particularly the capture step in antibody manufacture. More specifically, what is the role for packed-bed chromatography in future processes? Four reviews (4, 8, 13, 18) published over the past 2 years come to virtually the same conclusion on this topic: yes, packed bed chromatography is still the workhorse in bioseparation and will remain as the standard for at least the next 5 years in this conservative industry, but it cannot really cope with the challenges provided by increasing fermentation titers. Because of its selectivity, packed-bed chromatography has never been a “necessary evil”. As the central enabling technology, it has laid the foundations for downstream processing as an independent discipline. However, although it is here to stay for the foreseeable future, it is now slowly becoming part of the problem and not the solution. Even if we are able to bind 100 kg of a monoclonal antibody to a single column, it is questionable whether this is the best solution from an economic and quality standpoint, since older enabling technologies such as precipitation could now provide an equally acceptable and much more economic alternative. Following good initial recovery of the product from the biomass, downstream unit operations could concentrate on polishing. In the future, whole processes could focus more on the contaminants than the product, leading to a quantum leap in bioprocessing (48). With the range of alternative technologies now becoming available, packed-bed chromatography may not occupy such a hegemonic position in the future but may instead be one of a selection of equally viable solutions to tomorrow's downstream processing challenges.