Mixed-Mode Chromatography in Downstream Process Development

ABSTRACT

Mixed-mode chromatography materials contain ligands of multimodal functionality that allow protein adsorption by a combination of ionic interactions, hydrogen bonds, and/or hydrophobic interactions. Complex mixtures like fermentation supernatants or cell lysates can be applied directly at relatively high conductivity, and elution is usually achieved by electrostatic charge repulsion. We used mixed-mode materials for capturing and intermediate purification of several recombinant therapeutic proteins from various expression systems like yeast, Escherichia coli, and mammalian cells. Product-related impurities as well as process-related impurities from fermentation media were efficiently removed while the desired product was bound with high selectivity. Because these purification protocols can be scaled up easily to production scale, mixed-mode materials are being considered as potential elements of a general purification platform for recombinant therapeutic proteins produced in various expression systems.

The impurity profile of a starting material for downstream processing depends mainly on the conditions of the fermentation process, such as the expression system, medium composition, fermentation regimen, time point of harvest, or shear forces during isolation. This leads to a variety of process- and product-related impurities which must be removed by an efficient, orthogonal, and robust purification process. Modern downstream processes often consist of only two or three separation steps, and usually avoid conditioning steps like ultrafiltration/diafiltration (UF/DF) for buffer exchange to reduce the total number of process steps. However, this strategy significantly increases the risk that contaminants are not completely removed during the purification process, especially if complex microbial feedstocks or E. coli lysates are used as starting materials. In these cases, single-mode chromatography materials like ion exchange or hydrophobic resins often fail, in particular if the feedstock contains large amounts of complex colored impurities like melanoidins in combination with a relatively high conductivity of 15–30 mS/cm. Mixed-mode resins may help to overcome these problems because multimodal ligands offer excellent selectivity in combination with a salt-tolerant adsorption of the target protein.

(Bayer Schering Pharma)

The ligands of mixed-mode materials typically contain a combination of multiple binding modes like ion exchange, hydrogen bonding, and hydrophobic interactions.¹ The target protein itself is also a multimodal molecule. This situation results in a variety of protein–ligand interactions and often leads to unique selectivity, which in some cases includes pseudo-affinity. Furthermore, binding of the desired protein often is achieved at the conductivity of the feedstock without further dilution or addition of lyotropic salts. That feature makes mixed-mode resins an excellent choice for direct capture steps.

Commercially available mixed-mode materials are based on several principles. Resins containing hydrocarbyl amine ligands (e.g., PPA Hypercel, HEA Hypercel, Pall Corporation) allow binding at neutral or slightly basic pH values by a combination of hydrophobic and electrostatic forces. Elution usually is achieved by electrostatic charge repulsion when the pH value is lowered below the isoelectric point (pI) of the target protein and the pKa of the ligand.² Another ligand, 4-mercapto-ethyl-pyridine (MEP Hypercel, Pall), is based on a similar principle, but hydrophobic interaction is achieved by an aromatic residue and the sulphur atom facilitates binding of immunoglobulins by thiophilic interaction.^3–6 Another group of mixed-mode materials (Capto MMC and Capto adhere, GE Healthcare) contains ligands with hydrogen bonding groups and aromatic residues in the proximity of ionic groups, which leads to the salt-tolerant adsorption of proteins at different conductivities.^7,8 However, in contrast to electrostatic charge repulsion, which is driven mainly by changes in the pH value, the elution of proteins from Capto MMC usually requires an increase in the pH value and salt concentration. It also should be noted that many other chromatography materials that have been around for decades, like affinity resins with dye ligands, hydroxyapatite, and some ion-exchange resins, such as Amberlite CG 50 (Rohm & Haas) or Lewatit CNP 105 (Lanxess) often show a unique specificity that is based mainly on a multimodal protein–ligand interaction.^1,3

Mixed-mode materials were used in our laboratories to purify a variety of different therapeutic proteins, like cytokine muteins, protease inhibitors, and antibody fragments (Table 1). Although these resins are particularly useful for capture steps because of their selectivity and salt tolerance, they can be used for intermediate purification or polishing as well. In the following sections, we describe two case studies in which mixed-mode materials showed significant advantages compared with single-mode resins for removing a product-derived impurity containing a post-translational modification, and for purifying a therapeutic protein with challenging physicochemical properties.

Table 1. The use of mixed-mode materials in the downstream processes of various therapeutic proteins

Purifying Recombinant Protease Inhibitors from Yeast

Mixed-mode chromatography was used successfully to purify several recombinant Kunitz-type protease inhibitor muteins from yeast. Mutein I was expressed in Saccharomyces cerevisiae and secreted into the fermentation supernatant using a complex medium with a conductivity >20 mS/cm that contained a large amount of colored impurities (i.e., melanoidins). The protease inhibitor (pI ~6.3) showed a reasonable solubility of ~13 mg/mL at pH <3.2 that decreased significantly to <0.5 mg/mL when the pH value was increased only slightly (Figure 1). At a neutral or slightly basic pH value, an acceptable solubility could be achieved only at high-salt conditions.

Figure 1

The design of a suitable purification strategy focused on the following objectives: 1) identifying chromatography steps that keep the protein solution at a pH value below 3 or at high salt concentration and 2) using the low solubility of mutein I at neutral pH to find a suitable crystalline storage form for the bulk drug substance. Because the use of conventional ion exchange (IEX) resins was not possible given the unusual solubility profile of mutein I and the high conductivity of the starting material, the salt-tolerant properties of two mixed-mode resins (Capto MMC and HEA Hypercel), which are orthogonal with respect to their binding and elution pH levels, proved to be particularly useful. The purification process is shown schematically in Figure 2. The protein concentration of the feed stock was <0.5 mg/mL, and the clarified fermentation supernatant was applied at >20 mS/cm directly onto Capto MMC at a pH value of 5.5. Elution was achieved at a pH value of 8.5 under high-salt conditions (2 M NaCl), and, subsequently, the protein was applied directly onto HEA Hypercel without any conditioning steps. Elution at a pH value of 2.7 made it possible to keep the protein in solution under low-salt conditions. Lastly, increasing the pH value to more than 4 in a controlled mode of operation initiated crystallization, which led to a stable storage form of the bulk drug substance.

Figure 2

Capturing mutein I with Capto MMC was the key chromatography step of the purification process (Tables 2 and 3). The clarified fermentation supernatant was applied directly at relatively high conductivity (>20 mS/cm) without further dilution. The maximum capacity of 17 mg/mL was only moderate but sufficient for an early-phase production process. Process- and product-derived impurities were removed efficiently, and the purity of mutein I was >90% after this chromatography step, which indicates that Capto MMC acts as a pseudo-affinity resin in this case. This was particularly possible by using stringent washing conditions (Tables 2 and 3). High-salt washing conditions at a pH value of 5 and a low-salt washing step at a pH value of 8.5 led to an almost complete removal of the colored impurities (melanoidin) from the fermentation medium, and the Capto MMC eluate was colorless and clear. De-colorization by Capto MMC is mainly a pH-driven process. During the development of a production process for another mutein of a protease inhibitor (mutein II, see Table 1), we were able to show that colored impurities from complex yeast media bind to Capto MMC at pH values from 3 to 6, but can be completely removed at a pH value of more than 6.5. Thus, the binding of the desired protein to the resin under these conditions is of particular importance for a successful de-colorization strategy using Capto MMC.

Table 2. Experimental conditions for a typical Capto MMC capture run of the recombinant protease inhibitor mutein I

Several approaches for intermediate purifications were tested during process development. Reversed-phase chromatography under strongly acidic conditions worked well for purification of research-grade material, but using high-pressure equipment was not considered suitable for the large-scale manufacturing process. Because mutein I was eluted from Capto MMC at 2 M NaCl, pH 8.5, we screened for a resin that allowed binding at high-salt conditions as well as elution at a low pH value under low-salt conditions. The mixed-mode Hypercel-based resins from Pall meet these requirements because they allow salt-tolerant binding at neutral or slightly basic pH value, and elution is usually achieved at a low pH value by electrostatic charge repulsion. When binding mutein I on the Pall mixed-mode resins was tested, HEA Hypercel showed the highest capacity (27 mg/mL) compared to PPA Hypercel (19 mg/mL) and MEP Hypercel (9 mg/mL). We decided, therefore, to use HEA Hypercel for intermediate purification. The Capto MMC eluate was applied directly without prior buffer exchange, and elution at a pH value of 2.7 enhanced the solubility of mutein I under low-salt conditions (Tables 4 and 5). Product purity was increased only marginally during this step; only traces of product-derived impurities were separated. However, HEA Hypercel enabled us to remove large amounts of salt from the Capto MMC eluate and to immediately lower the pH value, which was necessary to prepare the protein for the crystallization step.

Table 3. Purification data from a typical Capto MMC capture run of the recombinant protease inhibitor mutein I

All unit operations were successfully scaled up by a factor of 5 to 10 and manufacturing was performed using a 5-L Capto MMC and a 2-L HEA Hypercel column, respectively. Crystallization was carried out in a 5-L glass reactor. The overall process yield was 70%.

Table 4. Experimental conditions of a typical HEA Hypercel intermediate step of the recombinant protease inhibitor mutein I

In summary, chromatography resins with multimodal ligands enabled us to purify a recombinant mutein of a protease inhibitor with an unusual solubility profile in a three-step process. Ligands combining ionic interactions with hydrogen bonding and hydrophobic interactions showed a unique selectivity as well as salt-tolerant binding and allowed for efficient removal of major product- and process-derived impurities under conditions where classical chromatography methods like IEX or hydrophobic interaction chromatography (HIC) failed.

Table 5. Purification data from a typical HEA Hypercel intermediate step of the recombinant protease inhibitor mutein I

Purifying a Recombinant Cytokine Mutein from E. coli

Ceramic hydroxyapatite (CHT, Bio-Rad Laboratories) is a crystalline mineral with multimodal functionalities that are significantly different from the mixed-mode ligands described above. The amino groups of proteins bind to the phosphate ions of the mineral by a classical cation exchange mechanism, and elution is typically achieved by increasing salt or phosphate buffer concentration. In contrast, carboxylic side chains of proteins bind to the Ca²⁺ ions of the resin by a combination of calcium metal affinity and anion exchange. The combination of different binding modes leads to excellent selectivity, which allows for efficient removal of aggregates and multimers in production processes of antibodies or antibody fragments.^7,9 We used CHT type I (particle size: 80 µm) to separate a closely related, product-derived impurity with a post-translational modification during clinical manufacturing of a recombinant cytokine mutein from E. coli.

The production process of the cytokine mutein is shown schematically in Figure 3. The protein was expressed in E. coli as an inclusion body and refolded. After pH adjustment and depth filtration, reversed-phase chromatography was performed to reduce the volume of the protein solution after the refolding step and to decrease the salt concentration. CHT type I (80 µm) was used as intermediate step to remove an N-formylated product-derived impurity. Finally, polishing was performed by reversed phase chromatography to separate several product-derived disulfide isoforms. The overall process yield was 50%.

Figure 3. Flow scheme of the cytokine mutein purification process

Analytical characterization of the cytokine mutein intermediate after refolding and reversed-phase chromatography revealed an N-formylated product-derived impurity that had to be removed from the desired nonformylated isoform. A broad screening of commercially available ion exchange resins showed that a separation of the formylated and nonformylated species could be achieved only with a 15-µm Source 15S cation exchanger operating at a pressure of 1–2 MPa. To perform intermediate purification at low pressure, several CHT and ceramic fluorapatite (CFT) resins were tested. Purification with CHT type I (80 µm) resulted in a reasonable separation of the formylated and nonformylated species (Figure 4a) when a column length of 9.5 cm and a column load of 17.5 mg/mL were used. To further improve the removal of the formylated impurity, an optimization program using Design of Experiments (DoE) was performed. Standard conditions for binding were 5-mM K-phosphate at pH 6.5, and elution was achieved using a linear 5–500 mM K-phosphate gradient. Five process variables were selected as factors: column length (5–14 cm), column load (5–30 mg protein/mL resin), flow rate (10–60 column volumes/h), gradient steepness (1–2 % eluent B/column volume), and pH value (6.0–7.0). A chromatogram-response function (CRF) was selected as a response variable to reflect successful separation of the formylated and nonformylated species.¹⁰ Purity and yield of the cytokine mutein were defined as additional response variables. A linear d-optimal design with center points was used that was later augmented to a full response surface design (i-optimal design). Column length and column load showed the largest effects, and several two-factor interactions were identified. Calculating the optimal factor settings (column length: 14 cm, column load: 5 mg/mL) resulted in a significant improvement of the separation of formylated and nonformylated species (Figure 4b). The optimized unit operation was successfully scaled-up (to a 2.8-L column) and used to manufacture large quantities of material. The overall yield of the process step was 85%, and the purity of the cytokine mutein increased from 60 to 75%.

Figure 4a-b. Separation of an N-formylated cytokine mutein species by CHT type I chromatography. The figure shows the elution profile of the native cytokine mutein (blue) and the N-formylated impurity (red) before (a) and after (b) optimization by Design of Experiments

A Generic Strategy for Screening and Optimizing Mixed-Mode Resins

The interaction of mixed-mode ligands with the desired protein is often complex, and extensive optimization is required during process development. Many of the mixed-mode chromatography steps described in Table 1 were optimized successfully without using DoE. However, DoE offers considerable advantages for developing complex separation steps because it enables the identification of main effects as well as two-factor interactions and critical/noncritical parameters. Furthermore, the model makes it possible to calculate the impact of selected factors for each point in the design space, which offers an enormous flexibility for process optimization in both early and late stages of development. We decided, therefore, to develop a generic strategy for optimizing mixed-mode purification steps (Figure 5). This approach includes screening various resins (Capto MMC, Capto adhere, MEP Hypercel, PPA Hypercel, and HEA Hypercel) using batch binding conditions, which allows for a selection of suitable resins based on binding and elution of the desired protein at different pH values and salt concentrations. Then, resins are selected and further optimized by DoE using small column chromatography (2 mL). We usually conduct a linear full-factorial design using four factors (loading pH value, elution pH value, salt concentration in the elution buffer, and column load), a design that can be augmented later to a full response surface design, if necessary. This approach was used successfully to optimize mixed-mode steps of an antibody fragment expressed in Pichia pastoris and a monoclonal antibody expressed in mammalian cell culture. The complete experimental program takes approximately four to five weeks and the total amount of protein needed is usually 1 to 2 g, which is acceptable even in early stages of process development.

Figure 5. Flow scheme of a generic approach for screening and optimization of mixed-mode chromatography steps

Conclusion

Salt-tolerant adsorption and unique selectivity are the major advantages of mixed-mode materials over single-mode resins. These features enabled us to meet a variety of challenges in downstream processing, such as capturing proteins from complex feedstocks, de-colorizing feed streams, purifying difficult proteins, and removing special product- and process-derived impurities. Moreover, the overall number of unit operations was significantly lower in these cases when mixed-mode steps were implemented in the downstream process. Thus, mixed-mode materials offer considerable flexibility for designing efficient downstream processes, in particular if no generic affinity ligand such as Protein A is readily available. The complexity of interactions between the ligand and the desired protein might be considered as a limiting factor. This can, however, be overcome easily by using a combination of resin screening and statistical design of experiments, which makes it possible to optimize complex separation steps in a generic and systematic fashion.

Acknowledgements

The contributions of Torsten Minuth and many other colleagues from Bayer Schering Pharma Biotech Development are gratefully acknowledged. Solubility experiments were performed by Dirk Havekost, Bayer Technology Services. We thank Hans-Dietrich Hoerlein for his support and many fruitful discussions.

FELIX OEHME, PhD, is a senior scientist in downstream process development and JOERG PETERS, PhD, is the pilot plant manager for biotechnology, both in Biotech Development, Global Biologics, at Bayer Schering Pharma AG, Wuppertal, Germany, +49 202 36 7743, felix.oehme@bayerhealthcare.com

References

1. Zhao G, Dong X-Y, Sun Y. Ligands for mixed-mode protein chromatography: Principles, characteristics and design. J Biotechnol. 2009;144(1):3–11.

2. Brenac Brochier V, Schapman A, Santambien P, Britsch L. Fast purification process optimization using mixed-mode chromatography sorbents in pre-packed mini-columns. J Chromatogr A. 2008;1177:226–33

3. Burton SC, Harding DRK. Salt-independent adsorption chromatography: new broad-spectrum affinity methods for protein capture. J Biochem Biophys Methods. 2001;49: 275–87.

4. Lees A, Topping A, Razzaq A, Reiter K, Acosta A-L. Purifying a recalcitrant therapeutic recombinant protein with a mixed-mode chromatography sorbent. Bioprocess Int. 2009;7(2):42–8.

5. Guerrier L, Flayeux I, Boschetti E. A dual-mode approach to the selective separation of antibodies and their fragments. J Chromatogr B. 2001;755:37–46

6. Arakawa T, Kita Y, Sato H, Ejima D. MEP chromatography of antibody and Fc-fusion protein using aqueous arginine solution. Protein Expr Purif. 2009;63:158–63.

7. Chen J, Tetrault J, Zhang Y, Andy Wasserman A, Conley G, DiLeo M, et al. The distinctive separation attributes of mixed-mode resins and their application in monoclonal antibody downstream purification process. J Chromatogr A. 2009; doi:10.1016/j.chroma.2009.09.047

8. Kaleas KA, Schmelzer CH, Pizarro SA. Industrial case study: evaluation of a mixed-mode resin for selective capture of a human growth factor recombinantly expressed in E. coli. J Chromatogr A. 2009; doi:10.1016/j. chroma.2009.07.023

9. Gagnon P, Beam K. Antibody aggregate removal by hydroxyapatite chromatography. Curr Pharm Biotechnol. 2009;10:440–6.

10. Siouffi AM, Phan-Tan-Luu R. Optimization methods in chromatography and capillary electrophoresis. J Chromatogr A. 2000;892:75–106.