The affinity capture paradigm that dominates industrial IgG purification has proven unsuitable for IgMs because, in most cases, they are affected adversely by harsh elution conditions. The large size of IgMs is also a challenge because it limits the operating conditions and performance of traditional porous-particle-based chromatography media. This article describes how these challenges can be overcome with available technology to develop effective manufacturing procedures for IgM monoclonal antibodies. 

Recent reports that IgM monoclonal antibodies offer promising anticancer activity have created a strong interest in their therapeutic potential.

  IgMs occur naturally in a variety of forms, represented dominantly by cyclic pentamers with a molecular weight of about 960 KDa, and cyclic hexamers with a molecular weight of about 1.15 MDa.

  IgMs are more heavily glycosylated than are IgGs, with a range of 8–12% carbohydrate. The extinction coefficient for polyclonal IgM is 1.18, but as with IgG, that value can vary from one monoclonal IgM to another.

The primary challenge to purification process development is that IgMs tend to be soluble in a narrower range of conditions than IgGs, and they are more susceptible than IgGs to denaturation.

  Turbidity is the usual consequence of exposure to unsuitable conditions. Light turbidity is often reversible on restoration of more moderate conditions, but heavy or persistent turbidity may precede aggregate formation or precipitation. The extreme pH values used to elute immunoaffinity and mixed-mode affinity columns sometimes appear tolerable at laboratory scale, but these frequently cause recovery problems at process scale.

  Protein-based affinity ligands such as mannan-binding protein and C1q can be eluted under more moderate conditions, but they are susceptible to proteolysis and cannot withstand sanitization by sodium hydroxide.

Low conductivity tends to compound the sensitivity of IgMs to pH. Conditions routinely used for ion exchange purification of IgG, such as pH 8.5 or 4.5, may result in precipitation. IgMs are generally tolerant of high salt concentrations but are sensitive to denaturation on exposure to strongly hydrophobic surfaces. Hydrophobic interaction chromatography (HIC) media commonly used for purification of IgG often denature IgMs.

  Size exclusion chromatography (SEC) is gentle and provides good fractionation, but it is undesirable for manufacturing applications because of its low productivity. 

The large size of IgMs poses another challenge. Large size is associated with slow diffusion constants, and slow diffusion constants limit both capacity and resolution on traditional porous-particle-based chromatography media. Table 1 lists diffusion constants for several antibody classes and fragments. The diffusion constant for IgM is approximately twice as slow as for IgG; this means that the flow rate would have to be twice as slow to achieve similar capacity and separation performance, assuming that equivalent surface area was accessible. But equivalent surface area is not accessible; pore diameters for a given gel type span a characteristic distribution of values, and the larger the protein, the lower the proportion of pores accessible to it. This compounds the capacity limitation already imposed by the slow diffusion constant.  

 Table 1.  Diffusion constants for antibody classes and fragments 

These limitations have collectively engendered the misconception that IgMs are difficult to purify. Indeed, their chemical sensitivities must be accommodated, but IgMs manifest a range of chemical characteristics that enable development of effective orthogonal purification procedures under conditions that avoid unnecessary stress.

  Most IgM monoclonals are highly charged and retained strongly enough by ion exchangers to support high binding capacities at moderate pH values.

  Hydroxyapatite binds IgM strongly at physiological pH and conductivity.

  HIC on weakly hydrophobic ligands provides good retention without risk of denaturation.

  Screening may reveal aggregate separation on any of these methods.  

In addition, the emerging generation of industrial ion exchangers includes solid phase supports that do not rely on diffusion. Both membranes and monoliths exploit convection for mass transport, and both have been applied effectively to IgM purification.

  Convection is independent of molecular size and flow rate, so dynamic binding capacity and resolution are unaffected by either parameter.

  Convective supports thereby offer a solution to the productivity bottleneck that afflicts traditional chromatography media. This solution has already had a positive impact with industrial purification of IgG, where anion exchange membranes are increasingly exploited for flow-through removal of host cell protein, DNA, endotoxin, and virus.

Membranes support less effective peak separation in bind–elute applications because of dispersion in membrane housings and between membrane layers. This dispersion produces a high degree of peak spreading, which erodes resolution and dilutes the eluted proteins. Monoliths lack these dispersion zones, and they further lack the void volume that truncates separation efficiency on porous particle media. These efficiencies result in sharper elution peaks, with the practical benefits of higher resolution and higher eluted product concentration. Capacities for large molecules are very high: dynamic binding capacity for endotoxin is more than 10 times higher on monoliths than on porous particle anion exchangers, and DNA capacity is nearly 50 times higher.

  These features all favor highly effective purification of IgM. 

Table 2.  Initial screening conditions for hydroxyapatite 

IgM cell culture supernatants (CCS) were filtered to 0.22 μm and stored at 4 °C. Buffer components were obtained from Sigma-Aldrich (St. Louis, MO). Buffers were made with water for injection (WFI) and filtered to 0.22 μm before use. Ceramic hydroxyapatite (CHT) Type II, 40 μm, was obtained from Bio-Rad Laboratories (Hercules, CA) and packed by Atoll GmbH (Weingarten, Germany) into 1 mL (5 x 50 mm) and 10 mL (11.3 x 100 mm) MediaScout columns. Monolithic convection interaction media (CIM) QA (quaternary amino) anion exchangers and CIM SO

 cation exchangers were obtained from BIA Separations GmbH (Klagenfurt, Austria). For initial screening and method development, 0.34 mL monolithic disks were used; 8 mL radial flow monoliths were used for process modeling. One mL RESOURCE ETH (ether) and PHE (phenyl) columns for HIC were obtained from GE Healthcare (Piscataway, NJ).  

Table 3.  Initial screening conditions for ion exchange 

An experimental large-pore poly-propylene glycol (PPG) HIC resin from the Tosoh Resin Innovation Program (TRIP) was provided by Tosoh Bioscience (Montgomeryville, PA). It was packed in a 1 mL (5 x 50 mm) column. Screening buffers and conditions for hydroxyapatite are described in Table 2, for anion exchange and cation exchange in Table 3, and for HIC in Table 4. Dynamic binding capacity was determined as described in Gagnon.

  IgM breakthrough was detected with the monolith-based anion exchange assay described in Gagnon, Richieri, Zaidi, et al.

  All chromatography experiments were performed on an ÄKTAexplorer 100 from GE Healthcare. Polyacrylamide gel electrophoresis (PAGE) was performed on Bio-Rad Criterion gels (10–20% gradient). 

Table 4.  Initial screening conditions for hydrophobic interaction chromatography 

A few milligrams of highly enriched product are a valuable asset in early method development because they allow product and contaminant behavior to be evaluated visually from chromatograms. This evaluation permits initial screening and process optimization to be done without secondary testing, and it accelerates development. Enriched product is easily obtained by Protein A affinity chromatography for human IgG monoclonals. Hydroxyapatite provides a useful alternative for IgM. Figure 1 illustrates screening results for an IgM on hydroxyapatite. The IgM peak is clearly identifiable, and the majority of contaminants flow through the column during sample application. Elution between 200 and 300 mM phosphate is the norm.

Figure 1.  Hydroxyapatite screening profiles at different monitor settings. Conditions as in Table 2  

Collecting the IgM peak from an initial hydroxyapatite screening run typically provides antibody of 65–80% purity. This purity is sufficient to proceed with initial screening of anion and cation exchange, but only a few more experiments are required to produce a reference sample that better reflects antibody purity under process conditions, which is typically approximately 90%. Figure 2 compares the initial screening profile with the profile from an optimized gradient. Conditions were set so that most contaminants were either eliminated in the pre-elution wash, or were left on the column after the antibody had eluted. The 1-mL column was subsequently loaded with 50 mL of CCS containing approximately 12.5 mg of IgM, then eluted with the optimized gradient. IgM from this run was used to evaluate other methods.  

Figure 2.  Comparison of screening and optimized profiles on hydroxyapatite. Conditions as in Table 2 

Figure 3 illustrates screening results for an IgM on cation exchange at pH 7.0. The profile is reminiscent of hydroxyapatite in the sense that the IgM binds much more strongly than is usual for IgG antibodies, and the majority of host cell proteins fail to bind. Figure 4 compares retention characteristics at pH 6.0, 7.0, and 8.0. Binding is weaker at pH 8.0, but purification is improved remarkably. These results suggest a process strategy where the IgM can be bound at pH 6.0 to maximize capacity, then washed and eluted at pH 8.0 to maximize contaminant removal.  

Figure 5 illustrates screening results for an IgM on anion exchange at pH 7.0. The upper trace was produced from hydroxyapatite-purified reference, the lower trace from CCS. Again, the IgM binds much more strongly than is typical for IgG monoclonals, and the degree of purification appears to be excellent; however, a substantial proportion of contaminants also bind. Strong binding by IgMs may support a process strategy such as that discussed above for cation exchange: Bind at the pH that supports the highest capacity, and wash and elute at the pH that supports the most effective contaminant removal. 

Figure 3.  Cation exchange screening profile at pH 7.0 with different monitor settings. Conditions as in Table 3 

IgM typically elutes from weakly hydrophobic columns in a single well-defined peak, as shown by the ETH profile in Figure 6, but this antibody was severely denatured by the more strongly hydrophobic PHE column. The small first peak was the only remnant of native IgM. The later eluting dominant peak and shoulders represent denatured forms. They were turbid on elution and precipitated overnight at 4 °C. Although weakly hydrophobic columns avoid the denaturation problem, they leave another challenge in its place: The IgM elutes at very high salt concentrations, which can be difficult to accommodate in later process steps.  

Figure4.  Cation exchange screening profiles at pH 6.0, 7.0, and 8.0. Conditions as in Table 3 

HIC media with intermediate hydrophobicity offer a practical solution. As shown in Figure 7, the IgM elutes from the PPG column as a well-defined  peak near the end of the gradient. Conductivity of the eluted IgM pool is sufficiently low to support binding to an ion exchanger with moderate dilution. Another benefit of HIC on moderately hydrophobic supports is that the high salt required for binding can dissociate ionic complexes that may exist between IgM and contaminants of opposite charge.

  This fact can be important, because the same positive charges that cause IgM to bind strongly to cation exchangers can just as easily bind DNA fragments. This is highlighted by the use of immobilized DNA for affinity purification of IgM.

  As with hydroxyapatite and cation exchange, the majority of contaminants flow through HIC columns on sample application (Figure 7).

Figure 5.  Anion exchange screening profiles of CCS and hydroxyapatite-purified reference. Conditions as in Table 3  

Loading sufficient IgM to obtain antibody for characterization is simple on hydroxyapatite because it accommodates filtered CCS with only minor adjustments. If the CCS contains less than 5 mM phosphate, then phosphate should be added to that concentration to stabilize the hydroxyapatite during large-volume sample applications. Titration may also be required to bring the sample to operating pH. Dynamic binding capacity (5% breakthrough at 200 cm/hr with a 10 cm bed height) was approximately 23 mg/mL for one IgM, and approximately 19 mg/mL for another. Dilution to reduce conductivity increases binding capacity for purified IgM. For CCS, however, it may allow more contaminants to bind, and so it may potentially reduce net product binding capacity. Dilution also increases preparative sample application time, which already accounts for several hours at 200 cm/hr. IgM capacity increases at slower flow rates, or with longer residence time at the same flow rate on a taller bed. 

Figure 6.  HIC profiles on weak and strong hydrophobic media. Blue profile: RESOURCE PHE (GE Healthcare). Green profile: RESOURCE ETH, 1 mL, 2 mL/min (approximately 600 cm/hr). Equilibrate with 1.5 M ammonium sulfate, 50 mM sodium phosphate, pH 7.0. Inject 20 Î¼L (200 Î¼g) hydroxyapatite-purified mouse IgM. Wash 2.5 Cv with equilibration buffer. Elute with a 10 Cv linear gradient to 50 mM sodium phosphate, pH 7.0.The colored areas represent native fully active IgM.  

Sample loading is more complicated for ion exchangers because many IgMs begin to develop turbidity soon after they are equilibrated to the low conductivity conditions required to support high binding capacities. Turbidity may be absent initially, but it tends to form progressively over time. This fact not only risks product integrity, it also creates a source of process variation, because sample composition varies over the duration of the load. Turbidity can be measured conveniently on a spectrophotometer at 600 nm. Solubility limitations generally disqualify bulk offline sample equilibration of IgMs for ion exchange. Sample application by inline dilution provides an alternative.

  Sample is titrated to target pH and loaded through one inlet line. Diluent buffer at target pH is loaded through another. Residence time of the IgM at dilution, defined as the elapsed time from the point of mixing to column contact, represents seconds or fractions of a second, depending on flow rate and configuration of the chromatography system. The short duration is generally insufficient time for turbidity to develop. An equally important benefit is that sample composition is uniform for the duration of the load, no matter how long that duration might be.

Inline dilution factors vary according to the salt concentration of the sample, the charge characteristics of the individual antibody, and the desired binding capacity. A dilution ratio of 1 part sample to 2 parts diluent is a reasonable starting point with feed-streams at roughly physiological conductivity, although higher dilution is likely to support higher binding capacity. Samples with higher conductivities may require still higher dilution factors. Any dilution invites criticism for water or buffer consumption and increased column loading time, but it is important to consider that inline dilution eliminates the need for diafiltration with its attendant equipment, equilibration, and chase buffers; preparation, process, and maintenance time; cleaning validation; and inevitable product losses. Moreover, monoliths support such high volumetric flow rates that loading time ceases to be a consideration. Capacities for hydroxyapatite-purified reference IgM on monoliths ranged from 30 to 40 mg/mL for both anion and cation exchangers at flow rates of 12 Cv.

Inline dilution is necessary for HIC as well, because the concentration of salt required to support good binding capacity on moderately hydrophobic columns is typically sufficient to precipitate the antibody.

  Development of inline dilution conditions for HIC on such ligands is discussed in depth by Gagnon, Grund and Lindback.

SEC supports effective aggregate removal for IgMs, but its low productivity is unattractive for manufacturing operations.

  The primary handicap is the sample volume, typically in the range of 2–5% Cv. This limitation is compounded by flow rates potentially as low as 20 cm/hr.

  A theoretical benefit of SEC is that it simultaneously equilibrates the sample for the next step, but this benefit is moot for most IgMs because equilibrating a sample to ion exchange conditions on an SEC column is likely to result in formation of turbidity and to jeopardize product quality. It may result in the IgM precipitating on the column. This is the basis of an IgM purification technique called euglobulin partitioning chromatography, in which the antibody is retained by SEC media at conductivity values too low to support solubility.

  Soluble contaminants flow through on sample application. The IgM is subsequently eluted by increasing the salt concentration. 

Many high-capacity alternatives are available for aggregate removal. Although it is impossible to predict which will best serve a particular IgM, it is likely that the screening conditions suggested above will reveal one or more methods that offer useful levels of aggregate reduction. Hydroxyapatite has frequently proven effective for removal of aggregates and polymers from IgG, IgA, and IgM monoclonal antibodies.

  HIC and ion exchange frequently provide worthy aggregate fractionation of IgG aggregates, and can be reasonably expected to do so with IgMs as well. Figure 7 illustrates promising separation of IgM aggregates on a PPG HIC column.  

If the native abilities of the methods in a process are not sufficient to achieve adequate aggregate reduction, it may be possible to enhance them. Recent investigations have demonstrated the ability of 3.75–7.5% polyethylene glycol (PEG) to promote effective IgG aggregate removal on ion exchangers and hydroxyapatite, even when no separation is apparent in the absence of PEG.

  Larger proteins are more responsive than IgG to the effects of PEG, suggesting that this treatment should be even more effective for IgMs.

  PEG is economical, is protein-stabilizing, and is an approved inactive ingredient in a large number of parenteral formulations.

The combination of good capacity, excellent contaminant removal, minimal sample preparation, and consistent applicability for most IgMs makes hydroxyapatite a good default candidate for capture. Typically, the eluting salt concentration is sufficiently low, with dilution, to make ion exchange practical as a second step. The high salt tolerance of HIC makes it a good candidate for a second step, but HIC is poorly suited to capture because of the large amounts of salt that are required. This would be an impediment even if the salt could be added directly to the sample, but the usual need for inline dilution multiplies salt requirements to a prohibitive level. Assuming that a concentration of 1.2 M ammonium sulfate was required for binding, and that this was achieved by inline dilution of 1 part sample with 4 parts 1.5 M ammonium sulfate, 0.99 kg ammonium sulfate would be required for every liter of CCS. Column loading time would be an issue, since the dilution factor would quintuple sample volume. 

The strong binding of the IgM shown in Figure 4 suggests the feasibility of cation exchange capture. Published data, however, indicate that cation exchange retention is highly variable for IgMs, so it should not be relied on as a default method.

  Under the best of circumstances, mild reduction of pH and 3–5 fold dilution of the CCS will probably be necessary to obtain good binding capacity. Another positive feature of cation exchange capture is that it removes CCS contaminants that can affect hydroxyapatite, principally including metal ions and chelators. Iron binds to and discolors hydroxyapatite, but published studies indicate that separation performance is unaffected.

  Continuous presence of at least 5 mM phosphate and a minimum pH of 6.5 generally stabilizes hydroxyapatite, but how well it does so in the presence of chelators in CCS remains to be evaluated. One important restriction with cation exchange as a capture method is that it cannot use citrate buffers if hydroxyapatite is the next step: citrate is a calcium chelator. 

Most IgM monoclonals bind strongly to anion exchangers. Anion exchange capture is handicapped, however, because contaminating proteins consume a significant portion of the binding capacity.

  Stronger binding contaminants are an even greater liability for strong anion exchangers, such as Q, quaternary amino (QA), quaternary amino ethyl (QAE), and trimethylaminoethyl (TMAE), because they maintain sufficient positive charge in 1.0 M NaOH to retain DNA through multipoint binding of numerous phosphoryl residues. Combinations of 1–2 M NaCl with NaOH provide more effective cleaning than NaOH alone, but only column treatment with DNase has proven adequate to achieve quantitative removal.

  If capture by anion exchange is desired, it is prudent to use a weak anion exchanger, such as diethylaminoethyl (DEAE). The liability of binding competition by contaminating proteins is still present, but DEAE loses its charge in NaOH, thereby releasing DNA and other strongly bound contaminants. 

Figure 7.  HIC screening profiles on PPG at different monitor settings. Conditions as in Table 4. Red arrow indicates aggregates. 

Articles

Manufacturing challenges surround the use of IgM monoclonal antibodies, but these can be overcome with current technology.

Purification of IgM Monoclonal Antibodies

In human monoclonal antibody (HuMAb) processes, nonaffinity purification schemes have been developed as viable alternatives to affinity schemes, especially when cell culture expression levels have been increased significantly. Primary recovery by tangential flow filtration (TFF) for concentration and diafiltration plays a crucial role in a nonaffinity purification platform with cation exchange as a capture step. A nonaffinity purification scheme, including an optimized primary recovery TFF step, is comparable to Protein A-based purification processes in overall process efficiency, product recovery, and quality.  

 Tangential flow filtration is a common unit operation either for cell culture clarification and concentration or just for concentration of the cell culture supernatant in biopharmaceutical manufacturing where affinity or nonaffinity capture is used. As a platform approach, exploring nonaffinity capture as a viable alternative to Protein A has yielded efficient purification schemes. These nonaffinity schemes with three, two, or one chromatography steps have been successfully developed for HuMAbs.

  Cell culture harvest needs to be preconditioned by concentration and buffer exchange to the appropriate pH and conductivity to have a uniform and maximum load on CEX column. For this primary recovery step to fulfill both concentration and diafiltration, a TFF step is generally optimized for different process parameters. During the diafiltration step, significant slow precipitation from the cell culture harvest is formed as the pH and conductivity are decreased. Development and optimization of TFF is crucial for the nonaffinity purification platform not only for the optimal loading condition on CEX, but also for determining the cost efficiency of the purification schemes. 

The Role of Primary Recovery TFF in Non-Protein A Purification Schemes

Binding to CEX resin requires below-neutral pH and low conductivity for HuMAbs. The lower the pH, the higher the binding capacity, thereby, reducing capture chromatography cycles. However, the low pH requirement can lead to challenges because of gradual precipitation during diafiltration. Therefore, assessing the influence of several process parameters is essential to optimize the TFF step during primary recovery. 

Primary recovery TFF plays a more significant role in the non-affinity scheme than in the affinity scheme not only because it is a batch-volume reduction and feed conditioning step for the capture, but it is also a partial purification step. 

During the primary recovery TFF, the clarified harvest bulk is concentrated several fold, which will reduce batch processing time on capture chromatography especially for high-binding CEX resins. With more high-binding CEX resins competing in the market, development scientists have the luxury of evaluating and choosing the highest binding resins with high processing flow rates.

  Even for the resins that can be operated at very high flow rate, the load time would be significant without volume reduction at primary recovery step.   

Feed Conditioning to Capture Chromatography 

During TFF, the diafiltration endpoint for pH and conductivity are typically kept equivalent to those at the CEX load to avoid further dilution or titration between the primary recovery operations and capture chromatography step. Usually, the diafiltration end-point is when the conductivity and pH of the retentate reach those of the diafiltration buffer and it typically takes about five diafiltration volumes (DV), which involves most of the processing time in this unit operation. Instead, the diafiltration time can be reduced if the diafiltration buffer with lower pH and conductivity than the end-point requirements is used. For example, if the load for CEX is set at pH 5.9 and 4.3 mS/cm, it takes only 3 DVs by using a diafiltration buffer of pH 5.7 and 3.5 mS/cm (Figure 1) instead of 5 DVs with a diafiltration buffer of pH 5.9 and 4.3 mS/cm. Because it takes a longer time for 1 DV as flux decreases towards the end of diafiltration, buffer with lower pH and conductivity significantly shortens the diafiltration time (~114 minutes in Figure 1). Buffer costs can be reduced by 40% when diafiltration is decreased from 5 to 3 DVs which translates into  savings of process time as well as raw material.  

 Figure 1.  Reduction of diafiltration time by using a lower pH and conductivity buffer 

Partial Purification of Host Cell Contaminants

Primary recovery TFF is also a partial purification step. As the pH moves lower during diafiltration, Chinese hamster ovary (CHO) host cell contaminants, especially DNA, gradually precipitate out. As the pH passes through the wide pI range of host cell proteins, it causes gradual precipitation of these contaminants. CHO DNA is significantly reduced as the diafiltration progresses, from starting ~10

 pg/mg at the end of the primary recovery step, while CHO host cell proteins (HCP) are reduced at a modest level (Figure 2). Figure 3 shows a robust purification step for CHO DNA reduction for different Hu-MAbs. DNA partial purification at TFF generally benefits non-Protein A purification schemes, especially for high cell density and high titer cell cultures. This also prolongs capture cycle lifetime because of cleaner and uniform feed. 

Figure 3. Reduction of CHO host cell DNA during primary recovery (PR) of different HuMAbs 

HuMAbs in the diafiltered buffer condition at relatively lower pH also exhibit better stability than the clarified cell culture supernatant (Figure 4). TFF bulk typically has a three- to four-week storage time, while clarified bulk can last only several days. The prolonged stability could be because of the removal of proteases from the cell culture supernatant during buffer exchange. This could be beneficial because the cell densities tend to increase for hightiter cell culture processes and more proteases are produced during the cell culture and harvest operations. Also, the secondary modifications of proteins, such as deamidation and isomerization of antibodies, are better controlled at the lower pH as a result of diafiltration compared to the high pH of cell culture supernatant.

  This provides a convenient in-process hold point, which can potentially benefit flexible manufacturing schedules in commercial scales. 

The primary recovery TFF processes complex, clarified cell culture bulk, and diafiltration can result in enhanced precipitation and gel layer formation, thereby influencing the flux rate. Optimizing the operating conditions to maximize the flux and minimize the operating time should consider several process parameters. Choosing a suitable diafiltration buffer and a TFF membrane is essential to decrease operating time. Testing load protein amount and concentration fold of the feed, as well as optimal combination of inlet and retentate pressures, is important to define the maximized flux rate and optimized throughput. 

Figure 4. Monoclonal antibody stability in clarified cell culture bulk versus primary recovery (PR) tangential flow filtration bulk 

The foremost step in optimization is to identify a suitable buffer system for TFF because the chemical composition of the buffer plays a key role in product recovery. Several aspects should be considered when choosing an optimal buffer system. First, the diafiltration buffer should be compatible with the product during exchange without affecting product recovery. Second, the buffer needs to be compatible with the capture load conditions. For instance, in the case of a HuMAb process, the best binding condition for CEX was found to be sodium phosphate buffer with sodium chloride. However, the process performance at TFF stage was compromised by using this CEX binding buffer as diafiltration buffer, which resulted in poor product recovery. To accommodate best conditions for both TFF and CEX, diafiltration can be performed in a phosphate buffer without NaCl to prevent product loss, and added before binding to CEX.  

Figure 5.  Diafiltration flux curves from different tangential flow filtration membranes 

Performance differences could exist in membranes from various manufacturers probably because of variation in pore size distribution even when they are made up of the same material and with similar molecular weight cut-off (MWCO). To identify the most suitable membrane for HuMAbs, evaluation of three membrane sources and two different pore sizes, 30 kD versus 50 kD, with similar process streams were performed (Figure 5). Better recyclability and consistent flux were noted with 50 kD membranes, whereas with 30 kD, the second cycle performance was compromised by decreased flux rate. However, similar product recoveries were obtained by both 50 kD and 30 kD membranes. Among three different membranes from different suppliers, one of them outperformed for process consistency during two cycles in terms of process flux rate.  

Figure 6.  Concentration and diafiltration time profiles versus load volume 

Determining Load Amount and Concentration Fold

A series of studies on the load volume were performed and the results are plotted in Figure 6. At constant concentration factor, as the load volume increases from 40 L/M

 , the concentration  and diafiltration times increase. Though the overall processing time is increased as load volume is changed from 40 L/M

 , even at the highest load amount, TFF operation could be completed by 6.5 hours. This translates into a five-fold lower membrane cost for processing the same amount of material. 

Figure 7.  Concentration and diafiltration time profiles versus concentration fold 

As concentration fold increases from 5X to 15X, there is a moderate increase in concentration time. However, the major impact is seen on diafiltration step, which occupies maximum TFF operation time (Figure 7). In fact, the diafiltration time at 10X to 15X is less than half of 5X concentration fold. This results in decreasing membrane and buffer usage and processing time, affecting overall raw material and labor costs. 

The permeate flux rate can decrease significantly, especially during the diafiltration step because of gradual precipitation of host cell contaminants from concentrated cell culture supernatant. Therefore, range studies of feed and retentate pressures are essential for smooth TFF operation. The experiments examined the effects from changes of transmembrane pressure (TMP) as well as ΔP, which is the pressure difference between feed and retentate (Figures 8 and 9). 

Figure 8. Concentration and diafiltration time profiles versus ÎP 

When TMP was kept constant (15 psi), the increase of ΔP from 20 psi to 30 psi prolonged the concentration time very moderately, while the diafiltration time was reduced significantly from 4 to 2.5 hours.

Figure 9.  Concentration and diafiltration time profiles versus transmembrane pressure 

However, under constant ΔP (20 psi), elevating TMP reduced the concentration time and had very little impact on diafiltration time.   

Figure 10.  Cost comparison of disposable versus reusable tangential flow filtration membrane 

By optimizing ΔP and TMP ranges, an additive effect on overall TFF process time reduction can be obtained. 

Fouling of TFF membrane by gel layer formation from complex composition of cell culture is common during the diafiltration stage. Normalized water permeability (NWP) is monitored at the beginning and end of each cycle to evaluate the cleanliness of the membrane and performance consistency. Typically, NWP decreases sharply during the first couple of cycles (10–20%) and stays in the range of 60–80% in subsequent runs. Using NWP as the sole evaluation criteria can terminate membrane lifetime earlier than necessary. NWP can be used as an early indication of membrane modification, and is not necessarily directly related to process performance.

  Further, membrane performance can be measured by monitoring fluxversus-time curve, process recovery, process time, and product purity.

  The TFF membrane life can be at least 10 cycles in a typical HuMAb process as assessed by the flux curve, process time, and product recovery in a scale-down TFF system. In this context, an alternate approach can be to use a single-use membrane, which can  reduce or eliminate cleaning and cleaning validation.

  Fully disposable crossflow systems are available in the market, which require extra costs for disposable adapter plates, pressure gauges, a flow meter, and a conductivity meter.

  But savings are more from eliminating cleaning validation and reductions in cleaning solutions, labor cost, membrane storage, process waste disposal, and analytical requirements. Disposable units will have significant cost advantage, especially when the membrane cycle usage is limited (Figure 10). 

Comparison of Affinity Versus Nonaffinity Schemes

Even though primary recovery TFF adds a unit operation to the whole process, this step results in significant batch volume and overall capture cycletime reduction. Loading time reduction is advantageous for high-binding resins where bed volumes can be relatively smaller. For example, for a 5,000-L batch, the unit operation includes one TFF and two cycles of cation exchange versus six cycles of Protein A (Table 1).  

Table 1.  Comparison of operation cycles for nonaffinity and affinity schemes 

As primary recovery TFF is a partial purification step, it aids in reducing the demand of host cell contaminant removal during CEX chromatography. These two-step operations provide efficient purification fold. Primary recovery TFF with ~90% or above process step recovery, along with CEX capture, were comparable to Protein A schemes in overall process recovery. 

Assuming the polishing steps are similar for both CEX and affinity schemes, which is the case for HuMAbs, there are more cost advantages for processing the same amount of HuMAbs for a nonaffinity scheme over a protein A scheme. By developing optimal process conditions, the binding capacity of current leading cation exchange resins can reach several-fold over Protein A resins (Table 1). With a much cleaner load on the cation exchange, the resin lifetime is well maintained to easily reach more than 100 cycles. Cation exchange capture also provides hidden advantages if there is a need to change the resin because of longer schedule interruption during product campaigns or an unforeseen new resin need during storage in large-scale manufacturing.  

An alternative approach to traditional Protein A schemes is comparable in overall efficiency, product recovery, and quality.

Optimizing the Primary Recovery Step in Nonaffinity Purification Schemes for HuMAbs

Increasing titers in mammalian cell culture has recently shifted the focus of process development efforts towards improving the economics of product recovery and purification processes. Some argue that a paradigm shift may be required to ensure timely and cost-effective delivery of future antibody candidates. This paper discusses the key process economic drivers and the impact of scale and titer on downstream processing (DSP) economic trade-offs.  

The success of therapeutic monoclonal antibodies (MAbs) in the treatment of indications such as cancer and autoimmune diseases has fuelled clinical trial activities; in recent years MAbs have become the fastest growing segment of the biopharmaceutical industry.

  Attractive returns coupled with the potential losses in revenue resulting from delays in product approval have made companies focus on speed-to-market rather than on improving process economics.

  However, MAbs are amongst the most expensive drugs—the annual cost per patient can reach $35,000 for antibodies that treat cancer conditions. With intensified competition on the way and increasing pressure from healthcare providers, the ultimate success of the next generation of MAb therapeutics will increasingly depend on economic factors.

  As a result, production costs and capacity use are becoming critical success factors for the industry.

The relative importance of key process parameters on the overall economic feasibility of a process varies with both the scale of operation and the titer assumed. As annual output and scale increase at a given titer or combined with titer increases, the relative importance of different cost categories are expected to change as follows (Figure 1):  

Overall cost of goods/gram (COG/g) decreases

 - Material costs rise considerably and dominate COG/g

 - Labor and capital-dependent costs (overheads) represent a less significant proportion of COG/g

 - Downstream processing costs become a major component of COG/g.  

If the increase in annual output is also accompanied by titer increases, then the trends are expected to become even more pronounced. At small scales, fixed costs tend to dominate and any changes in raw materials will have minimal impact. However, as scale increases, the role of raw materials becomes more critical. It is therefore important to look at the distribution of the costs (upstream to downstream, and direct to indirect costs) and at the scale and titer of interest to prioritize optimization efforts. 

 Figure 1.  Typical cost trends as scale increases for (a) total COG/g, (b) material, labor, and indirect (capital-related) costs, and (c) upstream and downstream processing operating costs. 

Efforts to lower COG/g must be targeted at decreasing the overall batch costs (e.g., reducing raw material costs) or increasing the overall productivity (e.g., increasing process yields). This section discusses specific downstream processing (DSP) drivers, which are of prime importance when handling multigram per liter titers.  

As titers increase to 10–15g/L, it is expected that this will have a profound effect on reducing COG/g, as long as the purification costs do not negate the cell culture gains. With increasing titers, the ratio of upstream to downstream costs shifts so that the downstream costs become more dominant. For example, Sommerfeld and Strube calculated that increasing the fermentation titer 10-fold from 0.1 to 1 g/L caused the ratio of upstream to downstream costs in their process to drop from 55:45 to 30:70. This shift reflects the fact that the upstream costs are inversely proportional to titer but the same is not true for the downstream processing costs.

  Increasing the titer to satisfy larger market demands increases the protein load on chromatography steps resulting in an increase in the number of cycles or additional investment in larger columns; this also produces larger volume loads on any subsequent filtration steps leading to longer filtration times or a need for larger areas.  

All these factors increase the downstream operating costs per batch. However, the overall COG/g can still fall if the increase in overall productivity outweighs the increase in downstream costs. Accordingly, as titers increase further, the downstream processing steps will become major contributors to the overall COG/g and offer greater potential for improvements and cost savings. Consequently, the downstream yield and material costs become significant cost drivers. DSP bottlenecks can lead to increased investment in larger equipment and longer batch durations. This results in increased running costs and decreased productivity and suboptimal COG/g values.  

The overall DSP yield is a function of the individual step yields and the number of downstream processing  steps, as has often been demonstrated using the plot in Figure 2. Improvements in step yields and the reductions in the number of steps have contributed to typical overall process yields increasing from 40 to 75% in recent years, with savings in cost of goods and investment, and allowing for higher facility throughputs.

The impact of increasing step yields has been illustrated by Sommerfeld and Strube where increasing the average step yield in a seven-step downstream process from 85 to 95%, which increases the overall yield from ~30 to 70%, results in a 40% reduction in the downstream COG/g.

  Increasing step yields actually increases the equipment size or number of cycles required and, hence, the cost of the DSP steps, because each step needs to handle a larger load (chromatography) or volume (filtration). However, because more product is produced per batch, the COG/g typically falls with increasing yields. 

Figure 2.  Overall yield as a function of individual step yields and the number of steps  

To maximize productivity and minimize investment and running costs, it is advisable to keep the number of downstream processing steps to a minimum.

  For antibody processes, this has encouraged the elimination of buffer exchange steps (diafiltration) that add little purification value by designing each chromatography step so that it can take the material eluted from the previous step where possible.

 Some companies have recently adopted processes that use only two chromatography steps while maintaining the desired purity levels.

  This requires an anion-exchange step to have additional selectivity to replace the intermediate purification and polishing steps. In cases where the contaminant profile, pH, and conductivity provide the opportunity to adopt this simpler process, time and cost savings can be achieved.  

In downstream processing, the distribution of raw material costs is highly dependent on whether or not resins and filters are reused. When treating these material components as disposables, resins and filters tend to dominate the material costs. Similar patterns can be seen in clinical manufacturing because these materials remain product specific. The impact of this operating strategy on the overall COG/g can depend on both the scale and the phase of development. The use of downstream consumables such as resins and membranes in a disposable fashion for a 200-L antibody process supplying early-phase clinical trials can provide both financial and operational savings.

  It has been reported that downstream processing using disposables can become a major disadvantage at the 10,000-L scale.

  This can be attributed to the fact that economies of scale result in a disproportionate effect on raw materials.

  Consequently, raw materials savings become more important for any process as the scale increases. 

The reuse of resins and filters involves a trade-off between reduced material costs and increased cleaning validation costs to determine the number of reuses with consistent performance. The higher the component cost or number of process steps, or the lower the validation costs, the greater the incentives to adopt filter or resin reuse.

  Chromatography resins, in particular Protein A, are often quoted as dominating purification raw material costs, owing to the high cost of the resin which is higher than ion-exchange resins. Large bioreactor scales of 10,000 L operating with a titer of 1 g/L, can result in Protein A resin costs of $4–5 million.

  Consequently, Protein A resins tend to be used in smaller quantities with multiple cycles, despite the complications of reuse validation.

  The reuse of Protein A resins can dramatically reduce their relative contribution to costs, making the costs associated with filters much more prevalent. In particular, virus filtration can represent a large contributor to purification material costs because of the costly membranes that are often used in a disposable fashion.

As the reuse of filters and resins increases, the cost of made-up buffers [chemical reagents and water for injection  (WFI)] can account for a surprisingly large proportion of the costs, which can, in some cases, be greater than the cost of resins and filters. For example, at a fermentation scale of 20,000 L, approximately 140,000 L of buffer is  required.

  The difficulty in estimating buffer costs partly reflects the  large differences in the estimated costs of WFI, with values of approximately $0.20/L suggested for in-house generation and $3/L for vendor or contract manufacturing organization charges.

  Buffer costs have been quoted as varying between $2/L and $12/L.

  Efforts to reduce the volume and number of buffers required can lead to savings, because this naturally occurs when moving from a three-step chromatography platform to a two-step one. 

As increasing titers demand the use of larger downstream equipment or additional cycles, the requirement for buffers and WFI will also increase. In existing facilities, this can present retrofit challenges if demands exceed the capacity of the buffer preparation suites and WFI storage tanks or the rate of WFI generation. The use of buffer concentrates and in-line buffer mixing can help to reduce the tank size and floor space required for buffers.

The greater the mass load during chromatography, the greater the likelihood that the practical capacity of chromatography columns will be exceeded where the current limit of column diameters is 2 m.

  Under these circumstances, multiple cycles may be required and there is a risk that the downstream processing time will exceed the bioreactor time. This will reduce the potential throughput of the facility and impact on the COG/g. These large columns can also pose installation challenges in existing facilities if there is insufficient floor space and if they cannot fit through doors.

Process Economic Trade-offs for DSP Bottlenecks

Current efforts to avoid downstream processing becoming a bottleneck when handling larger masses include intensifying existing processes by enhancing capacity and speed. Such approaches can mitigate the need for extra investment in equipment and improve the process economics. However, with each of these approaches there are trade-offs and uncertainties that need to be evaluated to assess the impact on overall process economics.  

Chromatography Resin Dynamic Binding Capacity

Increasing resin binding capacity reduces column size requirements with concomitant drops in resin volumes required and buffer consumption [for equilibration, washing, elution, regeneration, and clean-in-place (CIP)] per batch. Novel resins have capacities that are more than twice those of the first generation resins.

  This can lower consumables costs which are a major component of COG/g at higher titers and demands,

  although the impact on the consumable costs will depend on the price differential between first and second generation resins. Increasing the binding capacity of affinity resins can have a greater influence on COG/g because affinity resins are more expensive than ion-exchange resins. With the cheaper ion-exchange resins, Sommerfeld and Strube highlight that a trade-off exists between the less pronounced drop in consumables cost and the increase in labor costs, which becomes more important at higher binding capacities, because of the longer processing times. However, given that most of the novel resins also allow higher flow rates, this may not be an issue. 

The first generation of cell-culture–based MAb processes used compressible chromatography resins. However, these impose severe limits on usable bed heights and flow rates when considering the expected increases in titer and scale.

  A move away from compressible resins and towards rigid resins that can handle higher flow rates reduces process cycle times and turnaround time and increases productivity.

  Flow rates with rigid resins can be three to five times faster than conventional compressible agarose resins.

  However, their use can also lead to increased buffer demands, higher pressure drops, packing complications, and shear stresses.

  New ion-exchange resins have recently been developed that can handle high flow rates (700 cm/h) and low back pressures (<3 bar). If the productivity increases outweigh the increased buffer demands, this will positively influence the COG/g.  

As mentioned earlier, resin and filter re-use can have a significant impact on the process economics at higher scales if the materials are expensive, despite the CIP and cleaning validation costs. Re-use also reduces the frequency of column packing which is time-consuming and costly when carried out on a large-scale.

  New resins are becoming available with increased stability when exposed to the harsh chemicals used for CIP; hence, they have longer cycle limits and can contribute to reducing the raw material costs.  

Platform processes provide a generic approach to antibody production that greatly reduces the development time while streamlining the regulatory aspects of processing. They represent a useful starting point for customization depending on the antibody being manufactured. Advances in resin properties have also allowed platform processes to emerge with two rather than three chromatography steps.

  This can help to alleviate DSP bottlenecks in existing facilities because a two-chromatography–step process occupies less floor space and consumes less buffer. Through such process intensification methods, Kelley predicted that a platform consisting of two chromatography steps with high capacity resins would be able to handle an annual output of 10 tons.

  Newer resins with the combined attributes of longer lifetimes, higher flow rates, and improved dynamic binding capacities will lead to improved platform processes for antibodies and contribute to significant reductions in downstream costs.

Research into alternatives for column chromatography focuses on methods that have the potential to effectively handle increased amounts of both the product and impurities (e.g., host cell proteins and antibody aggregates or isomers). Ideally, these alternatives should achieve a separation power equal to that of column chromatography while reducing the COG/g.

  When assessing the cost-effectiveness of these alternatives, it is important to consider not only the equipment sizes and resource consumption, but also the development and validation costs required.  

Membrane chromatography operating in flow-through mode is emerging as a popular alternative to anion-exchange chromatography steps in MAb purification,

  because of its rapid operation, ease of scale-up, and cost savings (Table 1). The dominant component in the distribution of raw material costs shifts from buffer costs in packed-bed chromatography to membrane costs; a membrane suitable for processing a batch of several thousand liters can cost several thousands of dollars and is disposable and not reusable. The key process economic trade-offs for anion-exchange applications therefore depends on whether the savings in buffer, labor, and overheads outweigh the high cost of the membranes. Critical variables that will affect the outcome of this cost comparison are the relative differences in the handling capacities assumed between anion-exchange membranes and resins, which dictate the sizes required, and the assumed WFI and buffer costs; higher values of these variables increase the economic attractiveness of membrane chromatography.

  The pace at which resin and membrane capacities improve will contribute to which operation secures its place in future platform processes. In cases where packed-bed and membrane chromatography offer similar COG/g, the real cost advantages may be in the development and validation costs that are significantly reduced with membrane chromatography because there is no column packing or cleaning validation.

Table 1.  Example of downstream process economic trade-offs 

The future of therapeutic MAbs lies in the development of economically feasible downstream processes.

Economic Drivers and Trade-Offs in Antibody Purification Processes

Alternatives to standard chromatography for the purification of MAbs have been analyzed at the industrial scale. This article discusses an alternative use of precipitation to remove cell culture broth impurities, including DNA and host cell protein, while leaving the antibody in the supernatant. It evaluates several types of precipitation agents that are suitable for use in the manufacture of a parenteral therapeutic agent. The applications of this technology at scale are also discussed. 

 Many current research initiatives are focused on finding alternatives to column chromatography for the purification of monoclonal antibodies (MAbs). The industry-standard capture resin, Protein A, is expensive, has relatively low capacity compared to ion exchange (IEX) resins, and does not tolerate cleaning agents, which limits the resin lifecycle. Attempts to develop alternative capture resins have met with some success, but they generally do not offer the same platform potential as Protein A. An initial purification step with the purification capability and generic nature of Protein A is required, but at lower cost and increased capacity. 

In their recent review, Przybycien and colleagues discuss how other industrial applications have used processes without chromatography for years because of the uniqueness of the target molecule.

  The paper classifies the alternative technologies into three distinct areas—bulk, field-based, and adsorptive separation—which are categorized according to their industrial maturity versus resolution potential. Precipitation is described as a mature technology with low resolution potential. However, MAbs have unique properties compared to other components of the broth, which could result in more selective precipitation. 

This article will discuss an alternative use of precipitation to remove cell culture broth impurities, including DNA and host cell protein (HCP), while leaving the antibody in the supernatant. It will also present the results of recent experiments determining the efficacy of potential precipitants in impurity removal, and discuss the applications of this technology at scale. 

Precipitation of DNA and Protein—Other Applications

A comprehensive search of the literature reveals that precipitants have been used extensively in other industrial applications, and some of these could potentially be adapted for MAb purification. The precipitants fall into two categories—those used to purify DNA and proteins, and those used as a precipitating agent to purify antibodies, which could be optimized to remove impurities while leaving the antibody in the solution. 

Several precipitants described in the literature are used to purify DNA in aqueous solutions. These include, but are not limited to, indium chloride, polyaluminum chloride, manganese chloride, and zinc.

  In addition, several groups have reported on the use of polyamines, such as spermine and spermidine, for DNA precipitation.

  While these techniques are appropriate for precipitation at the laboratory scale as research techniques, they are not suitable for various reasons for industrial-scale manufacture. 

There are, however, precipitants referenced in the literature that could be adapted for use in therapeutic MAb manufacture. One such class of compounds is charged polymers, such as polyethyl-eneimine (PEI) and polyacrylic acid (PAA), which have proven effective in precipitating both proteins and nucleic acids because of the interaction between the charges on the proteins and the polymers, creating an insoluble complex.

  The use of PEI for the precipitation of nucleic acids has been studied in detail by Cordes and colleagues and by Dissing and Mattiason.

In addition to charged polymers, cationic detergents such as cetyltrimethyl ammonium bromide (CTAB) and domiphen bromide (DB) have been used to precipitate cellular DNA.

  Goerke and colleagues were able to demonstrate the use of DB as a DNA precipitant in the purification of adenovirus.

Persson, et al., used 6,9-diamino-2-ethoxyacridine lactate (ethacridine or ethodin) as a precipitant for both DNA and HCP from an 

 homogenate containing antibody fragments.

  The use of ethacridine (a highly aromatic, cationic dye used as an antiseptic) as a precipitant in antibody purification was first demonstrated by Horejsi and colleagues in 1956.

Of particular interest are the numerous descriptions of the use of short-chain fatty acids, such as caprylic acid, for the precipitation of plasma proteins and DNA, as first described by Chanutin and Curnish in 1960.

  Several groups have developed conditions for the use of caprylic acid in antibody purification from various starting materials, such as ascites fluid and mammalian serum.

  Parkinnen and Lundblad also showed that caprylic acid could inactivate enveloped viruses.

Evaluating Potential Precipitants for Impurity Removal

If certain engineering challenges can be addressed, precipitation may prove to be a valuable tool for antibody purification.

BioPharm International-03-02-2008

Purification of IgM Monoclonal Antibodies

Optimizing the Primary Recovery Step in Nonaffinity Purification Schemes for HuMAbs

Economic Drivers and Trade-Offs in Antibody Purification Processes

Process-Scale Precipitation of Impurities in Mammalian Cell Culture Broth

Solutions for Purification of Fc-fusion Proteins