In recent years, most pharmaceutical companies have focused on the development of monoclonal antibodies (mAbs). Increasing upstream titers and shrinking development timelines have posed several challenges to downstream process development of mAbs. Some of the major strategies and tools to address these challenges include the development of highly efficient platforms, high-throughput screening (HTS) tools, and reduction of the number of unit operations to help with facility fit. In the future, mAbs may represent a smaller percentage of the pipeline as portfolios concentrate more on development of antibody fragments, nanobodies, biosimilar protein therapeutics, conjugated proteins and vaccines, Fc-fusions, and nonantibody protein scaffolds. In addition to high cell density mammalian expression, expanded utilization of other expression systems such as microbial and yeast will support these newer biotherapeutics (BioTx). The diversity of BioTx and expression systems will pose unique challenges for downstream development. Novel tools, approaches, and/or platforms will be required to enable rapid development. Furthermore, with the increasing emphasis on Quality by Design (QbD), there is a need to develop paradigms to apply QbD not only to mAbs but also to other BioTx.

Early biotechnology products in the US and Europe were characterized by low titers and low cell densities using various cell hosts (1, 2). Most companies had few recombinant products licensed or in development, including insulin, somatotropin, interferon, tissue plasminogen activator, erythropoietin, Factor VIII, and Factor IX. There were no benefits or need for platform development or expanding manufacturing capabilities. Cell culture was performed using numerous methods, including roller bottles and stirred tank bioreactors operated in perfusion, batch re-feed, or fed-batch mode (1).  

Starting in 1986 with the licensure of Orthoclone Okt3 (muromonab-CD3), monoclonal antibodies (mAbs) such as Rituxan (rituximab), Herceptin (Trastuzumab), and Remicade (infliximab) have dominated the BioTx market (3). Most company pipelines saw a dramatic increase in the number of therapeutic mAb candidates, which in turn resulted in a desire to shorten development timelines. Due to the large doses required, antibodies are typically expressed at high titers in high cell density mammalian cell culture processes. This past decade has seen both the titers and cell densities increase by an order of magnitude to meet the clinical and commercial demands (4–6). Further, companies have made significant investments in manufacturing facilities, resulting in cell culture bioreactor capacities of 12,000–20,000L. The increasing volumes, cell densities, and protein masses have posed numerous new challenges for downstream processing.  

Shrinking timelines have necessitated developing tools for faster process development. Because of the similarity of mAbs, which mainly differ from each other in their complementarity determining regions (CDR), these biomolecules are amenable to platform development and manufacturing processes. The standardization of cellular expression systems, bioreactor conditions, purification processes, manufacturing hardware, and disposables has resulted in faster clinical development and lower costs. Most companies have developed standardized purification processes typically consisting of a cell harvest method followed by a capture chromatographic step and one or  polishing steps. The approach used at Pfizer has been described elsewhere (7–12). Weak partitioning chromatography (WPC) and a high throughput screening (HTS) tools have been essential for development of a robust two-column platform purification process. The benefits of a two-column purification process include reduced capital, footprint, quality systems, validation, cleaning, development costs, solutions, and water consumption. The HTS method has also been applied to the development of other purification processes for unique biotherapeutic modalities where screening of a large number of resins may be required.  

Another consideration in mAb processing is reducing the cost of goods (COGs). The cost of the virus-retaining filter, which can approach that of the protein A resin, can be reduced by using prefilters or optimizing protein concentration, temperature, buffer composition, and solution pH (see Figure 1) (13). The costs of freezing and storing bulk drug substance can be reduced by targeting high concentrations. Concentrations above 250 g/L have been achieved using ultrafiltration at elevated temperatures (14) and over 500 g/L using a wet-ultrafiltration membrane evaporation method (15) (see Figure 2). 

 Figure 1: mAb total mass-throughput and volumetric flux through the Viresolve Pro filter as a function of time.  An X0HC and a Viresolve Pro prefilter were used with 93 g/L mAb in 3 mM histidine, pH 5.0 at 35 Â°C.			

A significant consequence of the transfer of products between facilities is the generation of in-process pools that exceed the capacity of the storage vessels. Linking unit operations through tandem processes would eliminate this constraint and additionally may reduce process time, process costs, and documentation requirements. Feasibility has been demonstrated for a tandem downstream process for the purification of mAbs employing an affinity Protein A capture step followed by a flow-through anion-exchange (AEX) step and a virus filtration step (VRF) (16). 

 Figure 2: Viscosity versus concentration for an Fc-fusion protein and a mAb concentrated using a wet-ultrafiltration membrane evaporation method.  			

FUTURE CHALLENGES IN DOWNSTREAM PROCESSING

Because of the broad range of new molecules, hosts, and delivery methods being investigated, recent biopharmaceutical development has resembled that from the early stages of the industry. A broad range of alternative BioTx modalities have obtained regulatory approval and are currently in clinical trials (17–21) including antibody fragments, single-domain mAbs, Fc-fusions, vaccines, antibody-drug conjugates, nonantibody protein scaffolds, PEGylated proteins and peptides, and viral vectors.  

Some of the newer BioTx use nonmammalian expression systems such as yeast and 

if there is no requirement for post-translational modifications. These alternate expression systems may offer cost advantages. However, other issues may arise, including undesired mannosylation in a transferrin-exendin-4 fusion molecule expressed in yeast (22). This unwanted modification required an extra process step using hydrophobic interaction chromatography (HIC) to separate these modified molecules from the desired product. 

 expression often yields high levels of protein production, often in the form of inclusion bodies, which require extra steps involving their isolation and the subsequent extraction and refolding to obtain the desired product (23).  

For a difficult-to-refold protein such as neurotrophin-4, an additional step involving sulfonation of the cysteines was also required to obtain a useful refold yield (24). The use of high throughput screening with Design of Experiments has been used to accelerate the determination of optimal conditions for refolding difficult proteins (25).			

Each of these BioTx modalities poses a unique challenge in downstream processing because of the lack of a platform process resembling that used for mAbs. Examples of the purification challenges involved with two Fc-fusion proteins will be described in the following sections.  

Because each Fc-fusion protein has a different protein sequence fused to an antibody Fc region, the protein amino acid sequence, charge, size, and hydrophobicity vary more than those of antibodies. In addition, they are often thermally, chemically, or enzymatically unstable, which can lead to high levels of aggregates, clips, or inactive species.  

A purification process was developed for an acidic Fc-fusion protein that was unstable at ambient temperatures and low pH and that quickly formed high molecular weight oligomers (HMW1) and dimers (HMW2).			

To improve stability, the protein A chromatography step was performed at 2–8 

 C and the low pH eluate was neutralized immediately. The capacity of the protein A resin for the molecule was less than 15 mg/mL which is typical for Fc-fusion proteins.  

A combination of HTS methods and gradient chromatography was employed to identify resins and operating conditions for high molecular weight (HMW) species removal. Anion exchange, cation exchange, and HIC resins were screened in 96 well plates in an HTS format. Additionally, ceramic hydroxyapatite (CHA), immobilized metal affinity chromatography (IMAC), and HIC methods were screened with elution gradients. 

Due to the low pI of this Fc-fusion protein, a WPC–AEX step was not successful. Based on the resin screening, CHA chromatography was selected for optimization. An initial test of CHA was performed using a phosphate gradient. This method allowed identification of step elution conditions that reduced the HMW levels from 29% to 10.1% with 87% yield. The HMW1 flowed through the column and the peak contained primarily dimer species (HMW2). The CHA step had a capacity of 20 g/L.  

The HTS analyses also indicated that a HIC resin under certain loading and elution conditions could provide reduction of HMW. An initial experiment using protein A peak pool and a salt gradient indicated the resin had a relatively low capacity for the product but was effective at reducing both the levels HMW species and less active monomer variants. 

Further development was performed using the protein A pool as a load to the HIC step to identify conditions where the product would flow through the column while the HMW and less active monomer would bind. A successful method was developed but the HMW1 species began to flow through the column at about 10 mg/mL of load. 

Because the CHA step was successful at removing the HMW1, but not HMW2 species, the CHA pool was used as a load for the HIC step. The HIC step was loaded with the CHA pool containing primarily HMW2 at 20 mg/ml. The HIC step provided a six-fold reduction in HMW2 levels. In addition, the HIC step reduced the level of host cell proteins by 1.4 log

 . Based on these data, the final process used a Protein A capture step with elution at low pH followed by immediate neutralization. This was followed by a bind-elute CHA step to remove HMW1 followed by a flow through HIC step to remove HMW2 (see Figure 3). The first two chromatography steps were performed in the cold to maintain product stability. 

Figure 3: Yield and percentages of HMW1 and HMW2 versus step in the Fc-fusion purification process described in case study 1. 			

During development of an early-phase purification process for an acidic Fc-fusion protein expressed in a CHO cell culture system, it was determined that the protein displayed a range of 24–100% activity.  Several types of chromatographic resins, including cation exchange, anion exchange, HIC, CHA, and IMAC, were evaluated to remove the less active molecules along with other product and process-related impurities. The process and conditions described above for Fc-fusion protein 1 were not applicable to Fc fusion protein 2.   

The CHA is a bimodal resin with two primary functional groups: phosphate and Ca

 . The negatively charged phosphate groups serve as cation exchangers and the Ca

  groups serve as chelators to proteins.  For purification of basic and neutral proteins, the CHA column is usually equilibrated with phosphate buffer at a neutral pH to utilize a cation exchange mechanism for the separation of protein monomers from HMW species.   

The CHA column, when equilibrated with phosphate, resolved fully active protein from partially active protein.  The product was in the unbound fraction while the less active molecules were bound to the column. However, the load capacity was low and removal of high molecular weight species was poor using the phosphate-charged CHA column in flow-through mode. 

A calcium-charged CHA column operated in bind-elute mode successfully removed the less active monomer species, as well as high molecular weight species, from the process stream.   In this case, the CHA column was equilibrated with a low concentration of CaCl

 in a neutral pH buffer, and then loaded with a protein mixture also containing CaCl

.  After washing the column with a neutral pH buffer, the column was eluted with a phosphate buffer and stripped with a high phosphate buffer.  A high yield of the product was obtained in the elution peak.  The less active protein and HMW species were present in the strip fraction.  These data demonstrate that although Fc-fusion proteins are more challenging to purify than mAbs, other chromatography resins can be used to yield high-quality product. 

These two examples illustrate the challenges posed by the different Fc-fusion molecules.  In one case, calcium charging of the CHA column was required for removal of a lower activity monomer species.  In another case, a bind-elute CHA column followed by a flow through HIC column was required for adequate HMW removal, resin capacity, and product yield. The sequence of the polishing steps is also important for both purity and capacity.  

Numerous antibody drug conjugates (ADCs) are in clinical trials with most of them being used in cancer treatment with a toxin payload (26). The major downstream challenge is separation of multiple species containing different levels of the attached drug to the mAb. This results from the limited specificity for chemical conjugation of the drug to either exposed lysines or cysteines (generated by limited reduction) on the mAb. To enable straightforward downstream processing of ADCs, two main strategies have been used to generate more homogeneous molecules. These are the incorporation of additional free cysteines that can then be chemically targeted specifically or to mutate out some of the native cysteines to serines so that they are less available for coupling after a limited reduction reaction (27, 28).  

One approach to generate smaller and simpler BioTx for downstream processing is to replace full-length mAbs with antibody fragments such as single-chain variable fragments (scFvs) and antigen-binding fragments (Fabs). Cimzia and Lucentis are two recently approved Fabs produced in 

29) and although no scFv has yet received regulatory approved, there are a large number of both types of Ab fragments in clinical trials (30–32). Purification of these Ab fragments utilizes normal chromatographic procedures unless the H-chain is from the VH3 IgG gene family in which case it will bind to protein A. After purification these small fragments are usually modified to improve their pharmacokinetics using technologies, such as PEGylation. PEGylation dramatically changes the chromatographic properties of proteins such that loading capacity is severely compromised thus requiring larger columns (33, 34). Additionally PEGylation can dramatically increase the viscosity of concentrated drug substances. 

In addition to mAb fragments, another approach to obtain smaller less complex antibody molecules is to use either Domain Abs (human derived) or nanobodies (Llama derived), which are composed of a single variable domain of ~	14kDa. These small single domain antibodies can be easily captured from fermentation broth using Protein A since they can be derived from the VH3 IgG gene family (35, 36). Subsequent downstream purification has been accomplished using standard chromatographic techniques (36).  

Using viral vectors as BioTx for gene therapy offers many exciting opportunities to inactivate new targets. Their production presents numerous challenges because of the complex nature of their composition and large size (2–5 x 10

 kDa). They are usually composed of multiple proteins that encapsulate the nucleic acid payload. Additionally, there may be empty capsids that do not contain the gene of interest but are similar in physicochemical properties and can be difficult to separate from the desired product. CsCl density gradient centrifugation has long been the method of choice for their isolation; however, for clinical testing, larger scale production has recently been accomplished using standard column chromatographic methods. However, because of their large size, their binding capacity is usually very low and chromatographic media with normal pore sizes used for proteins offer no advantages since the pores are not available. Smaller beads with more surface area is the most effective manner to increase capacity (37,38). Other processes that are being utilized include tangential flow filtration that takes advantage of the large particle size for separation from smaller cellular contaminants. 

Vaccines can be produced using numerous technologies and have been generated against proteins, polysaccharide, and small molecule targets. Pneumonia is the largest single infectious disease causing infant mortality (39). To date, three vaccines have been licensed for treating diseases caused by pneumococcus-containing polysaccharides conjugated to proteins (40).  

Because prophylactic vaccines are given to a large healthy population ranging from infants to the elderly, the regulatory requirements during development and for licensure tend to be more stringent as compared with other biotherapeutic products. This also results in more early investment and upfront development work for a vaccine candidate than for a model biotherapeutic agent. Most commercial prophylactic vaccines are not well-characterized biologicals from a regulatory perspective because of their inherent complexity and/or poorly understood mechanism of action. In addition, if an 

animal potency model is available, the results don't usually predict the human response to the product with regards to the protein structure-immunogenecity/antigencity relationship. These reasons force sponsors to initiate extensive process characterization studies and to lock the production processes early in the clinical development program (41).  

One new approach to generate vaccines is to use virus–like particles (VLP) as potent immunostimulaters with covalent attachment of many copies of the desired antigen to their surface (42). Similar to viral vectors, VLPs present unique challenges because of their large size (2.5 x10

  kDa). Q  (14 kDa) is the most commonly used VLP. It is derived from the structural coat protein of this virus and it naturally assembles into 180 copies of itself during expression in the cytoplasm of 

 Subsequent purification is accomplished by classical column chromatography techniques with size-exclusion chromatography cited most frequently, which is not a desired step for scale-up (41). The challenges to develop alternative process steps are similar to the viral vector BioTx because of low capacity for most resins as well as the fact that there may be heterogeneous levels of antigen attached to the VLPs. Another new approach is the delivery of plasmid DNA to cells, which leads to the transcription of antigens and a subsequent immune response (43).  

An additional complexity is the fact that many vaccines are multivalent. The manufacturing processes for these vaccines involve many different fermentation, purification, and conjugation trains for the polysaccharides and carrier proteins. Developing processes that are similar between antigens can improve manufacturability and facility fit (44). 

The FDA and European Medicines Agency (EMA) have indicated that the clinical testing requirements for a biosimilar drug can be reduced if it can be demonstrated that the biosimilar candidate does not meaningfully differ from the innovator drug (45, 46). Currently developed biosimilars, such as human growth hormone, insulin, and erythropoietin, are chemically far simpler than mAbs. It is unlikely that blood or plasma-derived products or complex vaccines will be acceptable as biosimilars in the US or Europe due to their complexity (47). However, eight categories of biosimilar molecules have been manufactured in India, including monoclonal antibodies and recombinant hepatitis B vaccine (48). The first monoclonal antibodies that will lose patent protection in Europe and the US were made using more stringent impurity level targets than exist now because the technology is more mature and the safety and efficacy is well established. For example, early mAbs had levels of high molecular weight species below 0.5% (49). It is possible that mAbs in the future will target levels below 5% (50).  

Articles

Features

Newer classes of biotherapies will require innovations in processing technology.

Addressing the Challenges in Downstream Processing Today and Tomorrow

Anion exchange (AEX) products are commonly used as a polish step in product flow-through (FT) mode to bind impurities. Comparing resins with membranes and membranes to membranes is challenging due to the complicated and unique scale-down model formats of the products offered by different vendors. A novel approach to AEX FT using a short, 5 cm length, packed bed format with a faster operating flow rate is explained. The performance of commonly used AEX resins and membrane adsorbers, detailing dynamic binding capacity performance, efficiency, and a new disposable option for packed bed chromatography are compared. The data presented will show that a high performance AEX resin competes well with the performance of membranes. It provides similar processing times and the added benefits of reusability, ease of packing at different scales in various column formats, and the ability to implement initial process design from early phase manufacturing to commercial manufacturing, reducing overall costs and time to market.

AEX chromatography resins and membrane adsorbers are frequently used for downstream purification in the biotechnology industry. AEX products are used for polish chromatography in product FT mode to bind product-related and process impurities. Multiple product formats are commercially available, including chromatography resins, which are packed into chromatography columns, and membrane adsorbers, which are supplied in self-contained plastic housings. Comparing performance across product types and formats, including resins to membranes and membranes to membranes, is challenging due to the complicated and unique formats for scale-down models of the different product types.  

In FT mode, membranes have shown advantages over traditional soft gel packed beds due to faster operating flow rates, reduced buffer requirements, and disposability. In most cases, traditional soft-gel FT columns are sized for the optimization of volumetric throughput to improve operating flow rate and decrease process bottlenecks, rather than sized for actual impurity binding capacity. This results in packed columns with larger diameter, and therefore volume, than optimal, and this improper column sizing results in greater resin requirements and increased buffer usage, both of which impact operating time and cost of goods. Although membranes may be simpler to implement for early phase manufacturing, where the process scale is typically smaller, these products are not always cost effective at the larger, late phase or commercial manufacturing scales due to high material costs and lack of reusability. In addition, the available product formats are limited, posing linear scale-up challenges, adding to the challenge of maintaining process continuity and proving process equivalence between scales. Late-phase process redesign or a complete switch to packed bed chromatography may be required due to high consumable costs or limited product formats not allowing for linear scale-up.  

On the other hand, the re-use of resins is well established for traditional packed bed chromatography and this facilitates lower overall material costs. In addition, chromatography columns are easily scalable. Unit operations can be defined and locked for early phase manufacturing and then simply scaled up in a linear fashion as manufacturing scale increases reducing the need for later phase redesign. High performance, rigid resins with a higher volumetric throughput capability provide an advantage over soft-gel resins allowing for properly sized columns with smaller footprints similar to membranes. These rigid resins allow for convective flow through the bead, improving mass transfer and increasing efficiency at higher linear flow rates, thereby improving process productivity. 

 Figure 1: Pressure-flow curve for POROS HQ 50 in 8 cmD Go-Pure Pre-Packed column format at bed heights of 5 cm and 20 cm (5 um frits, 0.1 M NaCl, system pressure subtracted).			

Demonstration of performance using a scale-down model

An alternative and beneficial approach to AEX FT is the use of a short, 5-cm length, packed bed format with a faster operating flow rate. This format enables volumetric throughput capability that is similar to the membrane format, and increases flexibility when designing a purification scheme. The scale-down model used to compare the performance of the different AEX products/formats is summarized in Table I. We evaluated five commonly used AEX products: three resins and two membrane adsorbers for dynamic binding capacity of DNA and protein using bovine serum albumin (BSA) to mimic common contaminants, such as host cell proteins, removed during a FT step. The resin target operating flow rate was the maximum flow rate defined per the manufacturer's operating instructions. The membrane target operating flow rate was based on the common industry design space of 10 membrane volumes (MV) per minute. Although POROS chromatography resins can be operated at 2000 cm/h or faster, 1000 cm/h was evaluated as the upper limit for this evaluation. A 5 cmL POROS column can be operated at 1000 cm/h with a low pressure drop allowing for the use of high operating flow rates in conventional low pressure chromatography columns and systems (see Figure 1). 

Table I: Scale down approach for AEX chromatography resin and membrane adsorber comparison.			

DNA Dynamic Binding Capacity: The AEX product was pre-charged with 20 mM sodium phosphate, 1 M NaCl, pH 7.0 followed by an equilibration with 20 mM sodium phosphate, 50 mM NaCl, pH 7.0 (7.8 mS/cm). Each column/membrane was loaded with 2 mg/mL Herring Sperm DNA (Sigma D3159) in equilibration buffer (titrated to pH 7.0 with 0.2 M sodium phosphate dibasic, anhydrous) with a final conductivity of 8.8 mS/cm per the flow rates listed in Table I. Binding capacities at 5% (C5) and 50% (C50) breakthrough were determined based on UV absorbance.

Table II: DNA and BSA dynamic binding capacities of five AEX products.			

BSA Dynamic Binding Capacity: The AEX product was pre-charged with 20 mM Tris, 1 M NaCl, pH 8.0 followed by an equilibration with 20 mM Tris, pH 8.0 (1.1 mS/cm). Each column/membrane was loaded with 10 mg/mL BSA (Sigma A7906, pI 4.7-5.3, MW 66 kDa) in equilibration buffer with a final conductivity of <2 mS/cm per the flow rates listed in Table I. C5 and C50 breakthrough were determined based on UV absorbance.  

POROS HQ Viral Clearance: Polyclonal human IgG (Sigma G4386, MW:155–160 kDa; pI: ~6.9) was used for the model process. The salt concentrations evaluated and the corresponding conductivity values are summarized in Table III and Figure 2. The column format was 0.46 cmD x 5 cmL, 0.83 mL or 0.46 cmD x 20 cmL, 3.3 mL. Viral clearance was assessed at 1000 cm/hr at room temperature for the 25 mM, 50 mM and 150 mM runs and 300 cm/hr for the 100 mM NaCl run. The studies were all run at pH 7.0 using 20 mM bis-tris propane for buffering. The column was loaded with 5 mg/mL IgG with a 5% xenotropic murine leukemia virus (xMuLV, retrovirus, enveloped, ssRNA, 80-120 nm) or murine minute virus (MVM, parvovirus, non-enveloped, ssDNA, 18-26 nm). Spike and column FT samples were taken to determine viral clearance. 

The DNA and BSA binding capacity data of five anion exchangers is summarized in Table II. POROS HQ 50 demonstrated the highest DNA binding capacity of the five AEX products assessed. The C5/C50 ratio is a measure of the mass transfer capability of the resin/membrane and a way to characterize the efficiency of binding. The more efficient the mass transfer capability of a product, the less effect flow rate has on capacity. As the C5/C50 ratio approaches 1.0 the resin/membrane becomes more efficient. The POROS HQ C5/C50 ratio is as efficient as the membranes and significantly more efficient than Fractogel TMAE and Q Sepharose FF for DNA binding. The capacity at 5% breakthrough on the 5 cmL column at 1000 cm/hr was compared to the capacity on a 20 cmL column run at 300 cm/hr. POROS HQ had minimal change in performance in the two formats as compared to the other two resins. In addition, with POROS HQ, the DNA capacity is high under a wide range of operating conditions (30-35 mg/mL at pH 6.0–9.5 with 150 mM NaCl and >20 mg/mL at pH 7.0 with up to 400 mM NaCl, data not shown). POROS HQ ranked second highest for BSA C5 dynamic binding capacity and shows similar capacity and mass transfer efficiency similar to the membrane products.

Table III: Viral clearance on POROS HQ (1000 cm/hr, 5-cm bed height). 			

POROS HQ in FT mode demonstrated good viral clearance capability for XMuLV up to 150 mM NaCl (18 mS/cm) at pH 7.0, as summarized in Table III and Figure 2. The MVM model virus showed good clearance up to 50 mM NaCl (8 mS/cm) in this new AEX FT format suggesting that a shorter column run at a faster operational flow rate can achieve good viral and impurity clearance. The conductivity of the load appears to have an effect on both viruses. MVM is a poorly charged virus so minimal salt is needed to neutralize the charge and decrease the binding. However, it is a small virus and can easily access the pores, so binding performance is flow rate independent. XMuLV, on the other hand, is significantly larger and highly charged. With the higher salt (18 mS/cm), the hydrodynamic radius of the virus is most likely changing, allowing for more optimal perfusion into the bead.

Figure 2: Effect of load conductivity on viral clearance capability of POROS HQ. 			

Table IV presents a cost model comparing POROS HQ 50 in this short-bed format to a traditional resin and membrane process. The POROS HQ product load time in the new format is seven times faster than the traditional resin step and three times faster than the membrane. The total process time is six times faster than the traditional resin process and almost two times faster than the membrane. In addition, the optimized HQ format uses four times less buffer than the traditional resin step. One of the benefits of using a resin is the reusability at commerical scale, and POROS HQ allows for aggressive cleaning and sanitization, yielding excellent cycling and reuse performance. This study shows the cost difference for one cycle compared to 50 cycles and the cost benefit of reuse compared to a single-use membrane.

Table IV: Large scale operating cost model for POROS HQ versus conventional AEX resin (2700 L at 5 mg/mL, 13.5 kg monoclonal antibody process). 			

The novel approach of using a short, disk-like column with a resin capable of operating at high flow rates for AEX flow-through polish chromatography delivers increased flexibility when designing a purification scheme. This format is ideal for rigid resins with flow rate independent performance driven by the convective properties of the base bead. The properties of POROS HQ, for example, drive increased throughput and smaller column sizes, and ultimately a column that can be sized based on capacity for impurities. In addition, POROS HQ has been shown to have high impurity binding capacity and clearance over a range of process conditions, including high conductivity conditions (data not shown). This decreases the need for dilution of the feed stream or inclusion of a diafiltration step prior to loading on the HQ column, making the process more efficient and cost-effective. If disposability is a factor, Life Technologies now offers Go-Pure Pre-Packed Chromatography Columns for maximum convenience, providing faster time to process and faster time between processes.  

This new approach to AEX FT polish chromatography increases process development flexibility and offers a more cost effective approach to this process step as compared to improperly, oversized soft gel columns, and membranes with limited sizes and expensive formats at larger scale. This study shows that a high performance AEX resin competes well with the performance of membranes, providing similar processing times and the added benefits of reusability, which decreases material costs with increased cycles as modeled in Table 4. In addition, ease of packing at different scales in various column formats and the ability to implement initial process design from early phase manufacturing to commercial manufacturing reduces overall process development costs and decreases time to market. 

For Research Use Only. Not intended for animal or human therapeutic or diagnostic use. 

The authors would like to thank Susanne Alexander, Roger Decker, and Elliot Haimes for assistance with the execution of these studies.  

 is a senior field applications scientist in the POROS Applications group at Life Technologies, 

 is the director of POROS R&D and Applications at Life Technologies, Bedford, MA, 

A high-performance anion exchange resin performs well compared with membranes. In addition, the resin offers greater flexibility and cost savings.

Benefits of a Revised Approach to Anion Exchange Flow-Through Polish Chromatography

Few people would claim that process developers and artists have much in common. But at the boundary between art and technology it is clear that there exists a shared appreciation of beauty, simplicity and clarity, in one case generating aesthetic appeal, and in the other taking a complex challenge and crafting from it a simple and inexpensive solution by implementing more efficient design principles. 

Such is the value of innovation in downstream processing, a concept that unites the articles presented in this special edition of 

. Innovation is a difficult concept to define, and an even more difficult concept to capture and apply in bioprocessing, particularly given the industry's reliance on long-established technologies that have evolved incrementally. Process chromatography is a prime example of incremental improvement in action. This bedrock of biopharmaceutical manufacturing processes has evolved from a solution to achieve product quality and therefore regulatory approvals to a more recent incarnation that also focuses on productivity and process efficiency. Unfortunately, this is not a limitless resource. Eventually, the steep hill of improvement flattens out, and developers are forced to realize that no more efficiency can be squeezed from their current processes. Process chromatography has physical limits in terms of dynamic binding capacity and column size, above which there are no further cost savings and therefore no further benefits. 

Luckily there is a second form of innovation which is rightly regarded as the lifeblood of the industry. This involves the advent of disruptive technologies: those offering game-changing improvements over a short timescale, initially serving niche markets and then expanding to challenge the hegemonic position of established platforms.  

Recent examples include disposable media/buffer bags, the introduction of membrane chromatography for flow-through operations, and the replacement of Protein A capture steps with less expensive alternatives. Only by embracing such innovations when they occur can the industry hope to meet the multiple conflicting demands of a changing market, with tighter regulations, customers expecting lower costs of goods, and increased competition from manufacturers outside Europe and North America.

We are witnessing a revolutionary change in manufacturing no less impressive in scale than the revolution in art and literature during the Renaissance. This was regarded as a rebirth of ancient traditions while acknowledging more modern developments and applying contemporary methods.  

Likewise, process developers are going back to the drawing board to look at ways in which older technologies can be merged with new technologies in more productive ways, in the context of new development tools, such as quality by design and process analytical technology. Examples here include the inclusion of low-cost precipitation steps and high-technology membrane chromatography devices in the manufacture of monoclonal antibodies, the ancient and modern combined to reduce the number of steps and the amount of buffer required, without sacrificing either productivity or quality, or increasing the footprint. Such innovations not only allow production to keep up with demand, but allow new processes to be installed in existing facilities rather than requiring upfront investment in new infrastructure.  

 presents three articles from industry leaders presenting a case for innovation in biomanufacturing. We open with a thoughtful article by Ranga Godavarti (Pfizer) asking whether we really need further development in downstream processing or whether incremental improvements in current methods will be enough. Improvements in chromatography are discussed by Shelly Cote (POROS), focusing on polishing applications for anion exchange flow-through chromatography. Finally, Shwu-Maan Lee (Baxter) presents an interesting case study in how procedures for cell harvesting can affect the final product.  

The art of downstream processing is to let innovation create opportunities for process developers to explore. In the words of Scott Adams: "Creativity is allowing yourself to make mistakes. Art is knowing which ones to keep." 

 is vice-president of Purification Technologies at Sartorius Stedim Biotech GmbH. He is also a member of BioPharm International's Editorial Advisory Board and can be reached at 

This summary provides an introduction to our special issue on Separation and Purification.

Innovation and the Art of Downstream Processing

The meningitis vaccine NeisVac-C is a group C meningococcal polysaccharide conjugated to tetanus toxoid. The polysaccharide is recovered from the 

 cell capsule and is purified by base treatment with subsequent diafiltration to remove hydrolyzed cell impurities. Purified polysaccharide is clear to slightly cloudy. A recent group of successive lots contained large amounts of precipitate that had not been observed in 10 previous years of commercial manufacturing. The precipitate was mostly composed of the sodium salt of palmitic acid (C16:0) with lesser amounts of palmitoleic (C16:1), oleic (C18:1), stearic (C18:0) and myristic (C14:0) sodium salts. The elevated fatty acid levels that formed the precipitate were linked to a damaged pump used during harvest. Replacement of the damaged pump corrected the issue and >15 lots have since been produced without precipitation.

NeisVac-C is a vaccine that prevents the invasive disease caused by 

 serogroup C. The active ingredient is a polysaccharide–protein conjugate. Each dose contains 10 µg of de-

-acetylated group C meningococcal polysaccharide (GCMP) conjugated to 10-20 µg of tetanus toxoid protein and adsorbed onto 0.5 mg of aluminum as aluminum hydroxide in saline. 

 The GCMP is isolated from the culture medium of 

 by microfiltration. The microfiltration permeate is concentrated and diafiltered to remove small soluble fermentation components. The major purification step is a saponification reaction in which the GCMP is refluxed with high concentrations of base for several hours. After diafiltration, the mixture typically appears clear to slightly cloudy. A recent cohort of successive GCMP lots contained a precipitate that had not previously occurred in ten years of commercial manufacturing (see Figure 1). This article describes the problem, the root cause analysis, and the corrective actions. 

Figure 1: Visible precipitation during GCMP purification.			

The antigenic component of the vaccine, the de-

-acetylated GCMP, was purified from the culture supernatant of 

 serogroup C, strain C11. The cells and spent culture medium were circulated through 0.2 µm hollow fiber cartridges using a circumferential piston pump. The permeate from the filtration contained the polysaccharide (see Figure 2). 

This filtrate was concentrated and diafiltered across a 300 kilodalton (kDa) nominal molecular weight cutoff (NMWCO) ultrafilter (UF), (Millipore Pellicon 2) (see Figure 2, step 4), which retains the GCMP. The concentrated GCMP was then chemically modified with a saponification reaction to remove acetyl groups (Figure 2, step 5). Base treatment is the major purification step with high temperature incubation for several hours in NaOH.  

Deacetylation removed all the acetyl groups from the 

-positions and most of the acetyl groups from the 

-positions of the GCMP. It also hydrolyzed cell impurities and saponified any fatty acids, which were removed by subsequent diafiltration with water for injection (WFI) across a 50 kDa NMWCO UF (Pellicon 2, Millipore) (see Figure 2, step 6). The GCMP remain in the UF retentate.  

-acetyl groups are believed to be immunologically important, and were restored in a subsequent chemical reaction (see Figure 2, step 7). After reacetylation, the GCMP was diafiltered and concentrated with a 30 kDa NMWCO UF (Pellicon 2, Millipore) that retain the GCMP. The GCMP was then tested for concentration and purity. 

The GCMP content was determined by a colorimetric resorcinol-HCl method (1). This method measured GCMP monomer (sialic acid) using N-acetyl neuraminic acid as a standard.The protein imputity content was determined by the Bradford method (2) using BSA as a standard. 

The nucleic acid impurity content was determined by the absorbance at 260 nm, assuming an absorbance of 1 (1-cm light path) for 50 µg/mL of nucleic acid (3).			

The white, waxy precipitate was analyzed by Energy dispersive x-ray spectroscopy (EDXS), micro-Fourier transform infrared (FT-IR) spectroscopy, proton  nuclear magnetic resonance (NMR), and liquid chromatography-mass spectrometry (LC-MS) to determine its composition. LC-MS analysis of the precipitate was performed on a Waters 2695 HPLC system with a C18 column and a Waters Q-TOf API-US mass spectrometer. 

The fatty acids were identified at Microbial ID (MIDI). Samples were saponified in NaOH at elevated temperature and methylated. The fatty acid methyl ester was extracted in an organic solvent prior to injection into the gas chromatograph. Fatty acid identification is based on their retention times compared to a library of standard (4). 

Eight successive purification lots produced in a recent campaign had an atypical appearance (see Figure 1). These lots were cloudy liquids with a white, waxy precipitate in the 50K retentate (see Figure 2, step 6) and/or the 30K retentate (see Figure 2, step 8). The precipitate was identified as mostly the sodium salt of palmitic acid (C16:0). This amount of precipitation had not been noted in past commercial production and its appearance caused cessation of manufacturing while this issue was investigated. 

It is believed that most of the GCMP isolated from the fermentation is a lipidated molecule that is able to aggregate either with itself or with other macromolecules such as lipopolysaccharide (5). The aggregates are small enough to pass through the 0.2 µm harvest filters but are retained by the 300 KDa UF (see Figure 2, step 4).  

In the deacetylation, many macromolecules are hydrolyzed and the sodium salts of fatty acids are generated. The most common fatty acids in 

are palmitic and palmitoleic acids (see Table I). The fatty acid analysis of our working cell banks was consistent with the literature (6-8) (see Table I) and allowed the authors to conclude that the source of the precipitate was not exogenous but was derived from the cells in our fermentation.

Table I: Fatty acid content (%) of Neisseria meningitidis.			

During early development, fatty acid characterization was part of the product impurity profile. Purified GCMP prior to conjugation contained <0.5% (w/w) fatty acids (palmitic, oleic, and 3-OH myristic). After the initial characterization, fatty acids were not routinely analyzed, although residual protein and nucleic acid levels were. These data confirmed that most fatty acids present were palmitic acid, the most prevalent fatty acid in 

. Since the precipitate appeared in the 50K and 30K retentates, it was initially thought that there were flaws in the saponification and diafiltration steps. A significant portion of the root cause investigation examined the purification process, but nothing unusual about the deacetylation or saponification was uncovered. The in-process testing for residual protein and nucleic acid gave acceptable and typical results (see Table II). Since no protein or nucleic acids could be detected, and purification of GCMP was achieved, abnormalities in GCMP purification were ruled out as a root cause and the focus of the investigation shifted to the fermentation and harvest steps.

Figure 3: Damage to the harvest pump mating surfaces.			

Many of the fermentation inputs can affect cell metabolism and as a consequence, GCMP yield. Prior to the appearance of precipitation, the authors had seen increasing GCMP yields for several months. It was hypothesized that fermentation medium or operational parameters had changed, increasing GCMP yield in the form of lipidated-GCMP and therefore overwhelming the purification system. Several test runs were made to alter the fermentation process and reduce the GCMP yield, but had no apparent effect on the precipitation problem.

 Table II: In-Process testing of 30K retentate for purity of the precipitation lots: residual protein and nucleic acid			

The GCMP harvest was more closely examined. This process involved circulating the contents of the fermentation vessel through 0.2 µm hollow fiber cartridges using a circumferential piston pump. This pump is designed with moving part tolerances tighter than most similarly-sized rotary lobe pumps. Since the product is in the filter permeate, the fermentation medium is continuously circulated until the retained volume was low. We estimate that each cell passes through the harvest pump ~300 times. Upon examination, the harvest pump showed damage on mating surfaces in the lobes and rotor housing (see Figure 3). This damage had occurred when a catastrophic event scored these surfaces, rather than being the result of normal wear and tear. The pump continued to deliver expected volumes and pressures but the authors decided to replace it with an identical pump. The abnormal precipitation in the downstream process immediately ceased.

Figure 4: Fatty acid/GCMP % (w/w) in fermentation lots.			

The authors concluded that the harvest pump, while operational, was damaged in such a way that it was also acting as a cell disruptor. Not only was more lipidated-GCMP sheared from the cell capsule, but the cell membranes were disrupted, forming fragments small enough to pass though the 0.2 µm harvest filters (see Figure 2, steps 2 and 3), but large enough to be retained in the initial GCMP capture step (see Figure 2, step 4). Fatty acids in this material would be saponified in the deacetylation step (see Figure 2, step 5). If the sodium salts were soluble, they would be removed from the system during diafiltration. If they were not soluble, such as the sodium salt of palmitic acid, they would precipitate during this process. The ability of the damaged pump to mechanically extract fatty acids from the cells must have been increasing for several months before a critical point was reached where the amount or type entering the downstream process overwhelmed its ability to remove it. 

Table III: Concentration of palmitic acid entering the saponification step.			

Analysis of the 300K retentates (see Figure 2, step 4) by GC showed a general trend of lower total fatty acid/GCMP ratios in lots that did not generate precipitation downstream (see Figure 4). No specific level could predict if a lot would precipitate, but 30-40 % (w/w), could be used as an alert level. This interpretation was enhanced when the specific concentration of palmitic acid entering the deacetylation/saponification process was examined. Deacetylations with palmitic acid >0.7 g/l were likely to precipitate (see Table III). 

Table IV: Fatty acid composition (%) at different stages of processing in lots that precipitated fatty acid.			

The 300K retentate was the starting material for the saponification reaction. There is a mixture of saturated (palmitic, myristic) and unsaturated (palmitoleic, oleic) fatty acids in this retentate (see Table IV). The precipitate in both the 50K and 30K retentates was composed primarily of sodium palmitate. This contrast is explained by the differences in solubility of the fatty acids. As reported by McBain et al9, the sodium salts of unsaturated fatty acids were more soluble than the sodium salts of saturated fatty acids. Another factor that affects the fatty acid composition of the precipitate is the length of the fatty acid hydrocarbon chain: the longer the hydrocarbon chain, the less soluble the fatty acid was in water. Sodium oleate and sodium laurate are soluble in water at temperatures under 45°C. Sodium myristate, palmitate, and stearate have much lower solubility at the same temperature. This property may be the reason for the oleic acid to be the more abundant unsaturated fatty acid in the precipitate, despite the fact that palmitoleic acid is one of the most abundant fatty acid in the 300K retentate.			

Since an increase in fatty acid concentration in the 300K retentate was associated with the use of a damaged pump, it is reasonable to conclude that the damaged pump affected the integrity of the cells by increasing shear forces due to metal-to-metal contact. The issue described here highlights the importance of preventive maintenance. The pump had been in place for several years and appeared to work properly, in that it maintained normal flow rates and pressures during operation. Maintenance technicians performed preventive maintenance at routine intervals, focusing on electric current demand and replacing hydraulic oil in the gear case. The pump head was not routinely examined. Given what was learned about the sensitivity of bacterial cells to shear, the authors now visually inspect the pump more often and replace it at any sign of damage. 

The authors wish to thank Liqiong Fang, PhD, research scientist, Kirk Ashland, PhD, senior research scientist, Frank Hua, PhD, research scientist, and Catherine Quinn, research associate, all at Baxter Healthcare, Round Lake. IL, who provided critical analyses in support of this study. 

 is a principal scientist, all at Baxter Healthcare, 8000 Virginia Manor Road, Suite 140. Beltsville, MD 20705. 

 1. L. Svennerholm, Biochimica et Biophysica Acta, 24, 604-11 (1957). 

 2. M. Bradford, Anal. Biochem. 72, 248–54 (1976).  

 3. C.E. Frasch, Production and control of Neisseria meningitidis vaccine, Adv Biotechnol Processes.  13, 123–145  (1990). 

 4. Microbial Identification by Gas Chromatographic Analysis of Fatty Acid Methyl Esters (GC-FAME), MIDI Technical note #101. 

 5. E.C. Gotschlich, et. al.,  J. Biol. Chem. 256, 8915–8921 (1981). 

 6. V.J. Lewis, R.E. Weaver, and D.G. Hollis, J. Bacteriol. 96, 1–5 (1968). 

 7. C.W. Moss, et. al., J. Bacteriol.  104, 63–68 (1970). 

 8. M.M. Rahman, V.S.K. Kolli, C.M. Kahler, G. Shih, D.S. Stephens, and R.W. Carlson, Microbiol. 146, 1901–1911 (2000). 

 9. J.W. McBain, and W.C. Sierichs,  J. Amer. Oil Chemist's Soc.  25, 221–225 (1948). 

Careful analysis of an unusual precipitate identifies defects.

BioPharm International-04-02-2011

Addressing the Challenges in Downstream Processing Today and Tomorrow

Benefits of a Revised Approach to Anion Exchange Flow-Through Polish Chromatography

Innovation and the Art of Downstream Processing

Meningitis Vaccine Manufacturing: Fermentation Harvest Procedures Affect Purification

BioPharm International, April 2011 Supplement (PDF)