PEG Precipitation: A Powerful Tool for Monoclonal Antibody Purification

ABSTRACT

Various precipitation techniques have been used in the industrial purification of proteins for many years. Precipitation processes can be separated into two main categories: impurity precipitation and product precipitation. Impurity precipitation is operationally simpler but carryover of the precipitants can challenge subsequent unit operations. Product precipitation may have a higher risk of damage to the target molecule, but in addition to purifying the product, product precipitation also enables buffer exchange and concentration during the resolubilization step. An antibody precipitation step has been developed using a recombinant antibody produced in PER.C6 cells and statistical design of experiments to optimize product yield and host cell protein (HCP) removal. After appropriate precipitation conditions were developed, two methods to capture the antibody pellet were evaluated: depth filtration and microfiltration. A wash step was incorporated in both methods to reduce soluble impurities. The final process resulted in a product yield of 90% and HCP reduction of approximately 1 LRV.

The method of pellet capture was shown to have a significant impact on the purity of the redissolved product. The precipitation step is readily scalable and fits a fully disposable downstream process.

Percivia LLC

Efforts are ongoing to identify alternatives to packed bed chromatography to reduce the time and cost of processing high-titer product streams. Although many efforts focus on membrane adsorbers that directly replace columns of the same or similar chemistries, some older technologies are beginning to gain ground in recombinant protein manufacturing. One attractive alternative to chromatography is precipitation, which has been used in the plasma protein industry for many years.¹ A simple method of precipitation involves titrating the process fluid to the isoelectric point of the protein that is to be precipitated.² Lyotropic salts, such as ammonium sulfate, also have a long history of use in precipitation processes.³ Short-chain fatty acids, such as caprylic acid, are well known for their ability to precipitate DNA and host cell proteins (HCPs).⁴ Polyionic species also are useful precipitants for capturing a product of interest or removing contaminating proteins.^5,6

Polyethylene glycol (PEG) has been used for product and impurity precipitation.^7,8 It also can be combined with isoelectric precipitation to improve the efficiency of the separation.^9,10 After precipitation, centrifugation or filtration can be used to perform solid–liquid separation.¹¹ Although centrifugation is a well-established method to achieve this separation, washing the product pellet to remove impurities could be problematic, and it is not suited to a single-use process. Filtration—normal- or tangential-flow—requires more development, but washing the pellet is simpler, and it is readily adaptable to a single-use process.

In the present work, a product precipitation step was developed using PEG to recover a monoclonal antibody (MAb) from clarified PER.C6 cell culture media. Appropriate precipitation conditions were identified through the use of full factorial experimental designs. Two filtration steps were evaluated for the capture and washing of the precipitated product, and the superior method was scaled-up 10-fold. The total precipitation process resulted in yields of approximately 90% and HCP reduction of 1 LRV with no significant increase in the aggregate level of the redissolved MAb. Finally, the impact of the precipitation step on the subsequent cation exchange (CEX) capture step was investigated.

Materials and Methods Reagents

USP grade salts, Tween 20, hydrochloric acid, acetic acid, and sodium hydroxide were purchased from JT Baker (Phillipsburg, NJ). PEG was of reagent grade and purchased from JT Baker or EMD Chemicals (Gibbstown, NJ). All buffers were prepared using MilliQ-grade water (Millipore, Billerica, MA) and were filtered by 0.22-µm filtration before use.

Feedstock

A human MAb (IgG₁, pI = 8.3, 150 kDa) was produced at Percivia, LLC using a PER.C6 cell line. PER.C6 cells are human embryonic retinal cells immortalized by the adenovirus E1 gene, as described in US patent 5,994,128.¹² The cells were cultured in a standard fed-batch process or the XD process, both using chemically defined media.^13,14 The fed-batch media were clarified by sedimentation and depth filtration, and the XD media were clarified by the enhanced cell settling (ECS) method followed by depth filtration.¹⁵ During ECS, Silica-PEI resin was used to enhance cell settling and also reduce DNA and HCP.

Precipitation Condition Optimization

The conditions used to precipitate the MAb—PEG molecular weight, PEG concentration, and pH—were optimized by full factorial experimental designs using Minitab software (State College, PA). The pH of the clarified XD media was adjusted to the desired level with 2-M Tris in a 15-mL conical tube. The PEG was added as a 40% (w/w) stock solution to the desired final concentration. The tube was then centrifuged at 1,000g and the supernatant decanted. Finally, the pellet was redissolved in phosphate-buffered saline (PBS).

Pellet Capture by Depth Filtration or Microfiltration

Depth filtration was performed with various grades of filter media. Millistak+HC D0HC, C0HC, and X0HC were purchased from Millipore Corp. (Billerica, MA), and ZetaPlus 60SP02A was purchased from Cuno (Meriden, CT). Precipitation was carried out using a 40% (w/w) stock solution of PEG-3350 and the precipitated media was loaded at a feed flux of 50 L/m² /h until all of the material was loaded or the transmembrane pressure (TMP) was 15 psid. The filters were then washed with 20–30 L/m² of 20 mM Tris pH 8.5 + 14.4% (w/w) PEG-3350. After washing, 80 L/m² of resolubilization buffer was passed through the filters at 100 L/m² /h, and the permeate was recirculated through the device at 600 L/m² /h until the A280 of the permeate pool was stable indicating complete MAb dissolution. Finally, any held-up product was recovered with a 20 L/m² buffer flush and air blowdown of the filter module. In some tests, the filter media was subsequently washed with 85-mM acetate pH 5.3 followed by 1-M NaCl. Pressure and flow data were collected using a custom engineered system from ARC Technology Services (Nashua, NH).

Microfiltration was performed with a 0.22-µm hollow fiber membrane from GE Healthcare Life Sciences (Piscataway, NJ). The PEG-3350 was added as a 40% (w/w) stock solution for the small-scale experiment and in powder form for the scale-up work. The feed flux was 710 L/m² /h and the retentate and permeate were unrestricted. The precipitate was first concentrated 10- to 14-fold and then washed with three diafiltration volumes of 20 mM Tris pH 8.5 plus 14.4% (w/w) PEG-3350. Finally, the precipitate was redissolved in 85 mM sodium acetate pH 5.3 or 20 mM Tris plus 50 mM NaCl pH 7.5. Pressure, flow, and conductivity data were collected using a Slice 200 benchtop system (Sartorius, Gottingen, Germany) for small-scale testing and a SciPro system (SciLog, Middleton, WI) for the scale-up experiment.

Cation Exchange Chromatography

Toyopearl GigaCap S-650 was procured from Tosoh Bioscience (Montgomeryville, PA) in the Toyoscreen 5-mL format. This resin has been previously demonstrated as a high capacity capture step for MAbs.¹⁶ The column was equilibrated with 74-mM sodium acetate pH 5.3 and loaded to 90–95 mg-MAb/mL-resin using either clarified media or clarified and PEG-treated material, each adjusted to the same pH and conductivity as the equilibration buffer. The column was then washed with equilibration buffer and the antibody eluted with 50 mM sodium acetate pH 5.3 plus 90 mM NaCl.

Analytical Techniques

The MAb concentration in media-containing samples was determined by analytical Protein A HPLC (Applied Biosystems, Foster City, CA). Aggregate levels were measured by size-exclusion chromatography (SEC) using a TSKgel G3000SWXL column from Tosoh Bioscience (Montgomeryville, PA), with peak detection by UV absorbance at 280 nm. HCP levels were quantified by a PER.C6-specific ELISA from Cygnus Technologies (Southport, NC). SDS-PAGE was performed with NuPAGE 4–12% Bis-Tris gels and staining was done with SimplyBlue SafeStain, both from Invitrogen (Carlsbad, CA).

Results and Discussion Precipitation Conditions

For both molecular weights of PEG, 3,350 and 6,000 Da, the PEG concentration was the dominant factor in the recovery and purity of the redissolved MAb. Precipitation with PEG-3350 resulted in the highest recovery (Figure 1). However, the higher recovery came at the expense of higher HCP burden in the redissolved MAb. In the case of aggregated MAb, the levels were not significantly different from the starting material, but it should be noted that the particular MAb used in this work is not prone to forming aggregates. The improvement of the HCP reduction with the use of PEG-6000 was offset by the reduction in product yield. The final condition selected was 14% (w/v) PEG-3350 (equivalent to 14.4% w/w) and pH 8.5.

Figure 1

Pellet Capture by Depth Filtration

In the first experiment, ECS-clarified XD media was precipitated and the pellet was captured by depth filtration with X0HC media. No substantial increase in the transmembrane pressure (TMP) was observed during the loading of 361 g-MAb/m² (data not shown). After washing, resolubilization of the immobilized pellet was done with 85 mM acetate pH 5.3. Even after recirculation for 2 h, the antibody had not completely redissolved, so 0.1% v/v Tween 20 was added to the resolubilization buffer. After an additional 30 min of resolubilization, the MAb was fully dissolved.

Table 1. Yield and impurity removal for the PEG precipitation operation using depth filtration with X0HC to recover the product

Mass balance data are summarized in Table 1. The HCP reduction was lower than in the previous experiment (Figure 1), where only 5,000 ppm of HCP was in the final pool as compared to 8,300 ppm in this experiment. Some depth filters, however, are known to have hydrophobic and anion exchange (AEX) adsorptive characteristics.¹⁷ The precipitated media was loaded at a relatively high pH (8.5) and low conductivity (<10 mS/cm), which may have induced binding of acidic proteins on an AEX matrix. Furthermore, in the presence of PEG, the protein binding capacity of ion-exchange matrices has been shown to increase.^18,19 Therefore, it is probable that some of the HCPs that remained in solution after precipitation bound to the depth filter media and eluted into the product during resolubilization. This hypothesis also is supported by the difference in the HCP content of the precipitation supernatant and the depth filter flow-through (9,900 versus 240 mg) which indicates HCP removal from the solution by the depth filter. Following immobilization of the antibody onto the depth filter, it was redissolved at a significantly lower pH (5.3) resulting in the elution of bound HCP, and reducing the purity of the final product. This phenomenon could be mitigated by using less adsorptive depth filter media, such as Millistak+HC D0HC/C0HC and ZetaPlus SP filters like 60SP02A, or by redissolving the product in a low-salt, higher-pH buffer, like 20-mM Tris pH 7.5.

Figure 2

These strategies were tested by loading precipitated fed-batch media onto D0HC, C0HC, X0HC, and 60SP02A filters. All filters were washed identically with the PEG/Tris buffer, but the resolubilization was done with 20 mM Tris pH 7.5 plus Tween 20 for the Millipore filters and 85 mM acetate pH 5.3 plus Tween 20 for the Cuno filter. The Millipore filters also were stripped with a low pH buffer (85 mM acetate pH 5.3) and high salt (1 M NaCl) after the product was recovered to determine if any bound HCPs could be eluted. Figure 2 shows that there was substantial fouling of all filters tested. Because the fed-batch media was not clarified by the ECS method, which has been shown to significantly reduce DNA levels in XD harvests, there may be more DNA present, which precipitates in high PEG concentrations.²⁰ The higher DNA content in the precipitate may have reduced filter capacities, but this has not been investigated. Table 2 shows that each of the experiments resulted in better HCP reduction (84–88% versus 46%), and even though the starting HCP burden was higher (49,000 ppm versus 13,000 ppm), the redissolved MAb pools generally had lower HCP contents (6,000–7,800 ppm versus 8,300 ppm). The use of low-adsorptive filters and redissolving in a higher pH buffer appear to solve the problem of HCPs eluting from the depth filter media. SDS-PAGE of the strip fractions revealed that both HCPs and product were bound to the filters—including the less adsorptive D0HC and C0HC filters—and confirmed that X0HC media were more adsorptive than D0HC and C0HC media (Figure 3).

Table 2. Yield and impurity removal for the PEG precipitation operation using depth filtration with various filter media to recover the product

Pellet Capture by Microfiltration

Although the HCP burden of the redissolved MAb was reduced by using a buffer with a higher pH and less adsorptive depth filter media, the capacities of the depth filters were low for fed-batch media (<400 g-MAb/m² ). Microfiltration (tangential flow filtration, MF TFF) was tested with the aim of improving capacity with the added benefit of being able to use any buffer for resolubilization because of the low binding characteristic of the hollow fiber. The hydraulic performance of the MF TFF concentration and washing of fed-batch precipitate is shown in Figure 4. The permeate flux was about 100 L/m² /h and the TMP was between 1.0 and 2.5 psid for the entire operation. The mass loading of 475 g-MAb/m² was better than that achieved in any of the depth filtration operations. The product recovery and HCP reduction were both around 90%, slightly better than that achieved in depth filtration (Table 3).

Figure 3

The resolubilization was much simpler and faster than for the depth filtration; rather than recirculating buffer through the device, the buffer was pumped through the inside of the lumens into the retentate vessel with the permeate closed and allowed to mix for about 60 min. This was sufficient to redissolve the antibody, and no excipients were needed. Because of the difficulty in resolubilization with depth filtration, the product was redissolved at or near the concentration in the feed media. The final pool from the MF TFF process was nearly two-fold more concentrated than the starting material.

Figure 4

Because the retentate and permeate were open to atmospheric pressure, minimal instrumentation was required: a crossflow pump, one pressure sensor, and a transfer pump for the diafiltration. The combination of high capacity, high recovery and HCP clearance, and operational simplicity make MF TFF a preferable option for pellet capture as compared with depth filtration. The precipitation and MF TFF step was scaled-up 10-fold to a 0.12 m² hollow fiber device, and the performance was comparable to the small scale (Table 4).

Table 3. Yield and impurity removal for the PEG precipitation operation using microfiltration to recover the product at bench scale

Cation Exchange Chromatography

High capacity cation exchange (CEX) chromatography was evaluated with feeds pretreated with or without PEG precipitation. The precipitation step did not have any significant impact on the step yield or the percentage reduction of HCPs, but the eluate resulting from the precipitated load material had seven-fold less HCP (Table 5).

Table 4. Yield and impurity removal for the PEG precipitation operation using microfiltration to recover the product at pilot scale

Conclusions

Precipitation has long been used in the plasma protein industry to purify proteins at large scales. The technique has been adapted here to the initial MAb purification from clarified fed-batch and XD media in a scalable manner. Two single-use filtration steps have been developed to capture and wash the precipitated product, eliminating the need for centrifugation. It was shown that the precipitation operation did not negatively affect the yield of the CEX capture step, and it reduced the HCP content of the eluate by a factor of seven.

Table 5. Yield and HCP removal for GigaCap S-650 loaded with precipitated (+PEG) and nonprecipitated (–PEG) antibody

The ability to reduce the impurity burden so far upstream in the purification train is key to truncating the downstream process or replacing traditional chromatography with other single-use technologies. Lower impurity burdens can improve the loading capacity of flow-through membrane adsorbers and possibly virus filters, which are generally very expensive items in a process.

An added benefit is the ability to redissolve the antibody in a buffer that facilitates the subsequent unit operation. For example, cell culture media typically requires extensive titration and dilution or a UF–DF step to prepare for capture chromatography with a cation exchanger. Here, the precipitated antibody can be dissolved in equilibration buffer at high concentration, thus shortening the processing time. This can be important for products that do not tolerate long exposure to low pH/conductivity conditions. In the case of this particular antibody, the clarified media requires a more than two-fold dilution to be loaded onto a CEX column, whereas the redissolved MAb could be loaded directly at nearly two-fold the concentration of the unadjusted media. This is at least a four-fold reduction in the load volume, which can result in substantial time savings for modern, high-capacity CEX resins.

Acknowledgments

The authors would like to thank the Percivia protein and analytical sciences for assay support in this work and the Percivia upstream process development group for supplying the cell culture media.

Trademark Note

PER.C6 is a registered trademark of Crucell Holland BV Corporation; XD is a registered trademark of DSM NV; and ECS is a registered trademark of DSM NV. All other brand names are trademarks of their respective owners.

MICHAEL KUCZEWSKI is scientist I, EMILY SCHIRMER, PhD, is scientist II, BLANCA LAIN, PhD, is a senior scientist, and GREGORY ZARBIS-PAPASTOITSIS, PhD, is a senior director, all in downstream process development, Percivia LLC, Cambridge, MA, 617. 301.8821, mkuczewski@percivia.comeschirmer@percivia.com

References

1. Cohn EJ, Strong LE, Hughes WL, Mulford DJ, Ashworth JN, Melin M, et al. Preparation and properties of serum and plasma proteins. IV. A system for the separation into fractions of the protein and lipoprotein components of biological tissues and fluids. J Amer Chem Soc. 1946;68(3):459–75.

2. Van der Wielen L, Hofland G, Ottens M, Gulobovic M, Witkam G-J. Manufacturing protein-based structures using a volatile acid. In: 224th National Meeting of the American Chemical Society. Boston, MA; 2002.

3. Habeeb A, Francis R. Preparation of human immunoglobulin by ammonium sulfate precipitation. Vox Sanguinis. 1976;31:423–34.

4. Wang J, Diehl T, Aguiar D, Dai X-P, Arunakumari A. Precipitation of process-derived impurities in non-Protein A purification schemes for antibodies. BioPharm Int. 2009 Oct suppl, Downstream Processing 2010: Embracing Innovation;4–10.

5. McDonald P, Victa C, Carter-Franklin JN, Fahrner R. Selective antibody precipitation using polyelectrolytes: A novel approach to the purification of monoclonal antibodies. Biotechnol Bioeng. 2009;102(4):1141–51.

6. Moya W, Jaber J, Moya W, inventors; Millipore Corp., assignee. Purification of Proteins. United States patent US WO/2008/079280. 2006.

7. Polson A, inventor; South African Inventions Development Corporation, assignee. Fractionation of mixtures of proteinaceous substances using polyethylene glycol. United States patent US 3415804. 1968.

8. Giese G. Polyethylene glycol precipitation of monoclonal antibodies and the impact on column chromatography. In: BioProcess Int. Raleigh, NC; 2009.

9. Thrash S, Otto J, Deits T. Effect of divalent ions on protein precipitation with polyethylene glycol: mechanism of action and applications. Protein expression and purification. 1991;2(1):83–9.

10. Ramanan S, Stenson R, Ramanan S, inventors; Amgen, assignee. Method of isolating antibodies by precipitation. United States patent US 2008/0214795 A1. 2008 2/13/08.

11. Venkiteshwaran A, Heider P, Teysseyre L, Belfort G. Selective precipitation-assisted recovery of immunoglobulins from bovine serum using controlled-fouling crossflow membrane microfiltration. Biotechnol Bioeng. 2008;101(5):957–66.

12. Fallaux FJ, Hoeben RC, Eb AJvd, Bout A, Valerio D, Fallaux FJ, Hoeben RC, inventors; IntroGene B.V, assignee. Packaging systems for human recombinant adenovirus to be used in gene therapy. United States patent US 5994128. 1997.

13. Golden K, Bragg C, Chon J. The XD process: development of a high titer process for PER.C6 cells. In: 21st Meeting of the European Society of Animal Cell Technology. Dublin, Ireland; 2009.

14. Chon J. Advances in platform fed-catch and XD production processes using the PER.C6 human cell line. Wilbio Waterside Conference; 2009 Apr 20–22; South San Francisco, CA.

15. Schirmer EB, Kuczewski M, Golden K, Lain B, Bragg C, Chon J, et al. Primary clarification of extreme-density cell culture harvests by enhanced cell settling. Bioprocess Int. 2010;8(1):32–9.

16. Lain B, Cacciuttolo MA, Zarbis-Papastoitsis G. Development of a High-Capacity MAb Capture Step Based on Cation-Exchange Chromatography. Bioprocess Int. 2009;7(5):26–34.

17. Yigzaw Y, Piper R, Tran M, Shukla AA. Exploitation of the adsorptive properties of depth filters for host cell protein removal during monoclonal antibody purification. Biotechnol Prog. 2006;22:288–96.

18. Milby KH, Ho SV, Henis JMS. Ion-exchange chromatography of proteins: the effect of neutral polymers in the mobile phase. J Chromatogr. 1989;482:133–44.

19. Gagnon P, Godfrey B, Ladd D. Method for obtaining unique selectivities in ion-exchange chromatography by addition of organic polymers to the mobile phase. J Chromatogr A. 1996;743:51–5.

20. Paithankar KR, Prasad KS. Precipitation of DNA by polyethylene glycol and ethanol. Nucleic Acids Research.1991;19(6):1346.