Solutions for Purification of Fc-fusion Proteins

ABSTRACT

Platform processes are valuable tools in process development. Applying them to complex fusion molecules made up of different structures, however, can lead to unexpected results. EA2 is a unique engineered molecule, and it did not fit well into a platform process. It does, however, have three potential affinity and pseudo-affinity sites, each target site having a platform purification unit operation. One site is the antibody Fc region for Protein A binding, but the product was too sensitive to low pHs for typical Protein A elution conditions, so a high pH elution process was developed instead. The other two pseudo-affinity sites proved to be unusable, as did ion exchange chromatography. The final process that was developed added hydrophobic interaction and mixed mode (HIC and AEX) chromatography polishing, with a solvent and detergent (S/D) treatment instead of the typical low pH virus inactivation. This product points out the need for innovation and flexibility in cases where platforms do not work.

Platform processes are powerful and useful tools in process development. Applying them to complex fusion molecules made up of different structures, however, can lead to unexpected results. The process described below is a good example.

The product, designated EA2, is a unique engineered protein and it does not fit into an overall process platform. However, it has three well-known affinity and pseudo-affinity targets, each with well-known binding and elution conditions. Purifications for these targets are unit operations that are used in platform processes. Thus, the development of a purification process was anticipated to be reasonably rapid and simple.

Capture Step

The three targets are the Fc antibody domain target for Protein A, a target for binding to hydroxyapatite, and an enzyme whose active site binds to a dye ligand. Protein A was selected for the initial capture step because it is widely used in bioprocessing.

The process started with concentration and diafiltration because of its fairly low harvest titer. The concentrated crude bound readily to Protein A, and total recovery of A280 was good. However, recovery of activity was variable, and usually fairly low (Table 1).

Table 1. Small-scale low pH Protein A recoveries

Because low pH sensitivity is a common cause of loss of yield off Protein A, a pH stability study was performed (Figure 1a). Based on these results, a pH minimum of 5 was set to provide a safety margin. Even for relatively short exposures, pH levels of 4 and below are to be avoided. This ruled out using low pH for virus inactivation, and makes using low pH elution from Protein A a nonrobust option, even with a short column residence time and immediate neutralization.

It is known that eluate pH will vary between Protein A resins.^1,2 Thus, a number of Protein A resins were screened (Table 1). However, in no case was the eluate pH high enough for practical use.

The next step was to test alternate resins (i.e., Protein A mimetics) and alternate elution conditions.^3–8 The Protein A mimetic tested did not exhibit adequate binding capacity. Chaotropic elution with potassium thiocyanate (KSCN) was physically successful, but destroyed the product's activity. However, elution with Tris at pH 11 has proven to be moderately successful and repeatable, and the product was shown to be stable at this pH (Figure 1b).

Figure 1. EA2 pH stability. Concentrated purified enzyme was incubated at 21 ± 3 oC for various times at the indicated pH. Treatment was stopped by pH neutralization and samples were stored at 2–8 °C prior to measuring enzymatic activity. A) low pH stability, B) high pH stability

One downside to this procedure was that the elution peak tailed badly, and a modest amount of product did not elute until the low pH strip condition. Decreasing the elution buffer flow rate decreased the elution buffer usage and peak width. It seems that the product has a slow elution off rate. Given that Tris has essentially no buffer capacity at pH 11, these phenomena are perhaps not surprising. However, it seems that pH is not the sole driver of elution in this case. Glycine at pH 10 has a reasonable buffer capacity, but it failed to elute product when tested, doing even worse than Tris at pH 10 (Figure 2). Further investigation of this effect and optimization of the step are pending. We hypothesize that using a buffer with buffering capacity at pH 11 along with Tris would allow complete product elution.

Figure 2. High pH Protein A elution screening

Viral Inactivation

With the loss of low pH as a viral inactivation step, an alternative robust viral inactivation step was needed. Solvent/detergent (S/D) processes have a long history, and thus were the first to be examined. We decided that performing this unit operation just before the Protein A step was the best location in the overall process. At this point, the product was reasonably concentrated, minimizing the amount of S/D needed. Being early in the process, just before an affinity step, should afford good clearance of the S/D chemicals.

First, it was necessary to prove that the product was resistant to the chemicals. For this, partially purified product (Protein A, low pH elution with immediate neutralization) was incubated with two different detergents at two different Tri-n-butyl phosphate (TNBP) solvent concentrations for up to 24 hours (Figure 3). The detergents tested, Tween 80 and Triton X-100, were those most commonly cited in the literature and shown to be effective.

Figure 3. EA2 solvent/detergent stability. Final detergent concentration is 1.0% of Tween 80 or Triton X-100. At indicated time points, samples were taken and snap frozen, and stored frozen until enzymatic activity assay.

Although the results were somewhat noisy, they indicated that the product was reasonably stable to the treatment conditions. Also, the Tween-treated samples seemed more stable than the Triton samples over the long term, so Tween was the detergent of choice.

After this, we evaluated whether the S/D conditions would interfere with the subsequent Protein A chromatography. As Figure 4 shows, this proved not to be the case. As activity recovery with 1.0% TNBP and 0.3% TNBP seemed to be essentially equivalent (Figures 3 and 4), the 0.3% TNBP level was selected.

Figure 4. Small-scale Protein A purification of S/D-treated EA2. Load samples incubated in 1.0% Tween 80 plus indicated TNBP% for 60 minutes before loading onto column.

Polishing: Ion Exchange

The use of ion-exchange chromatography for polishing is very typical. Ion exchange is well known, well characterized, efficient, and economical. One typical modality is a flow-through anion exchange, more recently in a membrane format. This membrane format is disposable, has high throughput, and has good efficiency in removing negatively charged impurities such as nucleic acids, virus, and if needed, endotoxin, and some host cell proteins.

The pI of the product is variable, depending on the degree of sialylation. However, it is in the range of pH 5.2–6.5. Nonetheless, at pH 5.5, even with high salt, only 4% of the product did not bind to a Q membrane. This behavior was attributed to existence of a concentrated negatively charged patch built into the product molecule. Based on this, flowthrough anion exchange was not pursued.

The use of anionic Q resin in a binding mode was also examined. At first, this seemed modestly successful. The high salt elution peak tailed significantly, and recovery was only modest, in the 80–90% range. However, analysis of the product determined that the material that was lost was concentrated in the highly sialylated product, which was the most desirable form. Thus, this modality was also ruled out.

Given the high pH stability of the product, we tried the weak anion exchanger DEAE, eluting with high pH. It was thought that the combination of the different ligand (DEAE versus Q), and different elution modality (high pH versus high salt) could impact the product recovery. However, even at pH 11, elution recovery was poor.

Finally, cation exchange chromatography was examined. Given the behavior of the product on anion exchange, it was not anticipated that cation binding would be possible at a pH high enough to be stable. Surprisingly, at pH 6 with low salt, only partial flow through was achieved. The bound product was eluted at pH 6.8. The recovery and the sialylation of the product was good. However, there was no clearance of host cell proteins, so this modality was also shelved.

Polishing: Pseudo-Affinity

As mentioned earlier, the product has two pseudo-affinity sites other than the Protein A binding Fc domain. Resins to take advantage of these sites are less well known and are more expensive than ion-exchange resins. However, they have the potential for very powerful purification, so with the failure of the ion exchangers, these were investigated as potential post Protein A polishing steps.

One of these sites targets hydroxyapatite. The product bound quite well to commercially available chromatographic supports based on ceramic hydroxyapatite, Types I and II (Bio-Rad). However, neither high salt nor high phosphate could effectively elute the product (Table 2).

Table 2. Hydroxyapatite (HAP) recovery

The other potential site is an enzyme, alkaline phosphatase, whose active site is known to be a binding target of dye ligands. However, the ligand recommended for this enzyme only binds to some variants of the enzyme.⁹ This proved to be a case where binding of the enzyme to the ligand was of inadequate strength to be suitable for chromatography.

Polishing: HIC and Mixed Mode

Given the high and persistent charge on the protein, in order for the product to bind to an HIC resin, a strongly hydrophobic resin and a high salt concentration were needed. A butyl resin was successfully tested, using slightly over 1 M ammonium sulfate to affect binding.

Activity recovery off the column was generally good. A280 recovery was low because of the recovery of an unidentified colored material which co-eluted with the product off the preceding Protein A column (Figure 5).

Figure 5. Hydrophobic interaction chromatography (HIC) recovery and host cell protein (HCP) clearance

The HIC was also effective in removing biological impurities (Figure 5). However, a third chromatography step was desired to further reduce impurity levels. Given the success of the HIC, and the near success of the Q binding step, a mixed-mode resin with hydrophobic and anion exchange characteristics was tested (Capto adhere, GE Healthcare). The choice eventually proved successful, but only after some unexpected results.

Figure 6. Capto adhere chromatograms. Gradients from 0 to 100% B. A) pH 6 run; B) pH 8 run

The manufacturer recommends loading conditions between pH 6 and 8. Typically, the manufacturer notes that pH 6 gives better recovery, but that pH 8 gives better impurity clearance. Initial gradient elution experiments at both pH 6 and pH 8 confirmed the recovery observation, with 93% recovery at pH 6 and 86% at pH 8 (Figure 6).

Figure 6. Capto adhere chromatograms. Gradients from 0 to 100% B. A) pH 6 run; B) pH 8 run

Because impurity removal was the primary target of this step, development work first targeted pH 7 to 8. It became clear that in this range, the binding affinity for some of the product was low, and thus significant losses would occur if operated in a binding mode. On the other hand, a salt concentration of nearly 1 M would be necessary to operate this resin as a flowthrough column, and it seemed unlikely that this would enhance impurity clearance. Thus, we operated this column as a binding column at pH 6, as was done in the initial experiments, where complete binding was seen under the conditions used. Recovery was generally good, and impurity clearance was consistent (Figure 7).

Figure 7. Capto adhere recovery and HCP clearance

Process

At this stage, the process therefore has shaped up into eight unit operations. First, an initial ultrafiltration/ diafiltration (UF/DF) concentrates the crude material and removes a portion of the low molecular weight impurities. This is followed by S/D virus inactivation. The first major purification step is Protein A chromatography. This is followed by HIC chromatography. A second UF/DF is then needed to reduce the salt concentration for the subsequent mixed-mode chromatography. The seventh unit operation is nanofiltration for virus removal. The process is completed with a final UF/DF step into the final formulation.

Individual unit operations, especially the chromatography steps, would benefit from additional characterization and optimization. Nonetheless, this process has been successfully scaled up, and is being used to produce material for preclinical and early clinical trials.

Summary

EA2 is a unique, engineered Fc-fusion protein, and as a result, cannot be purified following a fully platformed process. However, its purification process, as initially envisioned, was to take steps from various platform processes targeted toward the affinity and pseudo-affinity binding sites on the EA2 molecule. However, what we learned is that with a molecule this complex, none of the three potential steps functioned as expected. The process that was developed was quite different from that initially envisioned, for a truly unexpected purification.

Acknowledgements

Our thanks to the following scientists for their contributions to this paper: Dr. Hong Li, Dr. Peychii Lee, and Munir Nahri at Laureate Pharma, Downstream Process Development: the Analytical Development group, and Eric Leblanc at Enobia Pharma.

DOUGLAS W. REA is senior specialist of downstream process development, MICHIEL E. ULTEE, PhD, is senior director of process sciences, and SHARON X. CHEN, PhD, is senior scientist of analytical and downstream process development, all at Laureate Pharma, Inc., East Princeton, NJ, 609.919.3395, info@laureatepharma.comThomas P. Loisel, PhD, is group leader of biology at Enobia Pharma, Inc.

REFERENCES

1. Ghose S, Allen M, Hubbard B, Brooks C, Cramer, S. Antibody Variable Region Interactions with Protein A: Implications for the Development of Generic Purification Processes. Biotechnol Bioeng. 2005;92(6):665–673.

2. Grönberg A, Monie E, Murby M, Rodrigo G, Wallby E, Johansson H. A strategy for developing a monoclonal antibody purification platform. BioProcess Int. 2007 Jan;48–56.

3. Bywater R. Elution of Immunoglobulins from Protein A-Sepharose Cl-4B columns. Chromatog Synth Biol Polym. 1978;2:337–340.

4. Bywater R, Eriksson G, Ottosson T. Desorption of Immunoglobulins from protein A Sepharose Cl-4B under mild conditions. J Immunol Methods. 1983;64(1-2):1–6.

5. Fornstedt, N. Affinity chromatographic studies on antigen-antibody dissociation. FEBS Lett. 1984;177(2):195–199.

6. Croze E, inventor; E.R.Squibb and Sons, assignee. Alkaline pH elution in improved method for purification of monoclonal antibody. Eur Pat Appl. 453,767. 1991 Oct 30.

7. Hober S, Nord K, Linhult M. Protein A chromatography for antibody purification. J Chromatogr B Analyt Technol Biomed Life Sci. 2007 Mar 15;848(1):40–7.

8. Pierce. Optimization elution conditions for immunoaffinity purification. Technical resource. 2004 Nov. www.piercenet.com/files/TR0027dh4-Elution-conditions.pdf.

9. Lindner M, Jeffcoat R, Lowe C. Design and applications of biomimetic anthraquinone dyes. Purification of calf intestinal alkaline phosphatase with immobilized terminal ring analogues of C.I. reactive blue 2. J Chromatogr. 1989 Jun 28;473(1):227–240.