Session 3.1: Protein A, Hydroxyapatite, and Mixed-Mode Chromatography

3.1.1. Viral Clearance of the Capto Adhere Chromatography Step In Monoclonal Antibody (mAb) Purification Processes (Junfen Ma; Pivotal Purification Process Development Amgen Inc., Cambridge, MA)

The use of mixed-mode chromatography in mAb purification processes provides enhanced removal of product-related and process-related impurities. One such resin is Capto Adhere with ligand chemistry of N-benzyl-N-methyl ethanol amine, exhibiting multiple functionalities including ionic, hydrogen bonding, and hydrophobic interactions. The Capto Adhere resin has been widely used by the biopharmaceutical industry because of its high selectivity for removing impurities (aggregate species, HCP, and host cell DNA), easy facility fitting, and process streamlining due to its salt tolerance feature.

In this study, the Capto Adhere resin was evaluated for its viral clearance capability using multiple mAbs with different physiological properties. These mAbs are either IgG1 or IgG2 with an isoelectric point (pI) range of 6.8 to 8.5. The Capto Adhere chromatography step was operated in a flow-through (FT) mode, where the impurities (aggregates, HCP, etc.) bind to the column and the mAb doesn't.

Figure 1 shows the chromatograms for four different mAbs with the loading and wash phases. It is noted that the loading level (g/Lr) and conditions (including pH and conductivity) vary among the four mAbs. The log reduction values (LRV) for both xenotropic murine leukemia virus (XMuLV) and mouse minute virus (MMV) are summarized in Table I. The Capto Adhere chromatography step, when operated in a FT mode, can reliably provide >4 LRV of XMuLV clearance and the LRV achieved are rather consistent among the four mAbs. On the other hand, MMV clearance by the Capto Adhere chromatography step varied significantly among the four mAbs, ranging from 2.5 to >5 LRV.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Capto Adhere chromatograms for different mAbs.

In addition to evaluating the viral clearance capacity for different mAb molecules, further study focused on understanding what factors affect LRV. Side-by-side comparison revealed that the presence of mAb could adversely affect the viral clearance as shown in Figure 2. The reduced LRV seen in the presence of the mAb is likely due to the competitive binding between viruses and impurities in the mAb load. This observation is consistent with other types of chromatography operated in FT mode in the mAb purification processes. The other possibility is due to the interaction between mAb molecules and viruses, which allows viruses to flow through with the mAb molecules, resulting in lower LRV.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Impact of mAb product on XMuLV clearance. XMuLV was spiked into either mAb A or the buffer with the same composition matrix and then loaded onto the Capto Adhere column for comparison. Open symbol indicates no virus detected.

To summarize, in addition to its high selectivity of removal of impurities and salt tolerance features, Capto Adhere resin, when operated as a FT step, can provide good viral clearance (>4 LRV for XMuLV and >2 LRV for MMV). Further studies are needed to better understand the underlying mechanism and key process parameters for virus clearance by Capto chromatography (e.g., aggregate species, HCP and host cell DNA, role of interaction between mAb and viruses, etc.).

3.1.2. Clearance of XMuLV and Minute Virus of Mice (MVM) by Ceramic Hydroxyapatite (HAP), Mixed-Mode and Anion Exchange (AEX) Chromatography for Pharmaceutical Antibodies with Low pI (Seiryu Ujiie; Chugai Pharmaceutical Co., Ltd, Tokyo, Japan)

We share here the MuLV and MVM clearance data obtained from different types of chromatography used in the purification process for several pharmaceutical antibodies with relatively low pI (pI <8). Since these antibodies and viruses have similar pI (e.g., pI of MVM = 6.2, pI of MuLV = 5.8), they are expected to have comparable surface charge at a given pH. Therefore, it is difficult to achieve good virus clearance and removal of other impurities using AEX chromatography in FT mode. Consequently, alternative chromatographic modes are needed to achieve effective virus clearance and impurity removal for antibodies with relatively low pI.

The virus clearance by HAP chromatography is not well understood (1). Table II shows clearance of MuLV and MVM by HAP chromatography step for two different antibodies with pI <6. HAP chromatography was applied in the 2nd or 3rd column chromatography as a polishing step and each elution step was operated with increasing in-salt or ionic strength. In these conditions, moderate LRV (LRV = 3.40–4.03) for MuLV was achieved. In contrast, poor MVM clearance (LRV = 1.24 or less) was observed. Although the two viruses have similar pI, a better clearance for MuLV was observed than for MVM. It remains unclear what are the determining factors for the significantly different LRV between MuLV and MVM. Further studies are needed to better understand the underlying virus clearance mechanism by HAP chromatography.

Robust clearance of MuLV and MVM using a mixed-mode AEX resin in FT mode was reported in the past symposium (2). We evaluated the effect of pI using the same type of mixed-mode resin for an antibody that has a pI <6. This chromatography was operated in FT mode. As shown in Figure 3, MuLV bound tightly to the resin during loading and wash steps and robust MuLV clearance (LRV >5) was obtained. While the MVM partitioning pattern was similar to that of MuLV, only moderate MVM clearance (LRV = 2.90) was achieved.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

MuLV and MVM partitioning patterns by mixed-mode and AEX chromatography. The asterisk symbol denotes no virus was detected.

Finally, we evaluated whether AEX chromatography operated in bind-and-elute (BE) mode could provide better virus clearance for molecules with low pI in comparison to FT mode. In general, BE mode AEX chromatography for low-pI product requires an elution buffer with high conductivity. Table III shows the MuLV and MVM clearance results from three different antibodies. Poor MVM (LRV <1) and moderate MuLV (LRV = 2.13–3.44) clearance were observed under high-conductivity washing conditions. However, MVM and MuLV clearances were significantly improved by changing the salt and decreasing conductivity. The results suggest that both buffer types and conductivity are important for virus clearance by AEX chromatography operated in BE mode for products with low pI.

3.1.3. Virus Clearance Robustness for Multimodal AEX Chromatography (Sherrie Curtis; Genentech Inc., South San Francisco, CA)

Protein purification processes implement multimodal resins with ligands that exhibit a combination of ionic, hydrophobic, and hydrogen bonding functionalities. It has been shown that virus removal by multimodal AEX resin is robust across wide ranges of pH and conductivity (3). By using lab-scale column chromatography with six Genentech purification processes, we have evaluated multimodal AEX chromatography for virus clearance in product FT mode and confirmed that it provides robust virus removal.

For one mAb in which the multimodal AEX chromatography is operated in product FT mode, two studies were designed to statistically evaluate the process parameter impact on virus clearance. One full factorial study was designed to evaluate column dynamics on virus clearance by systematically studying across the ranges of bed height and flow rate. As shown in Table IV, MMV and Simian vacuolating virus 40 (SV40) clearance as measured by LRV was not affected. Statistical evaluation confirmed that the range of XMuLV LRV is within normal experimental variations (bed height effect estimate 0.12; flow rate effect estimate 0.26).

A second study was designed to evaluate three parameters known to effect chromatogram shape: pH, conductivity, and temperature. Changes in chromatogram shape indicate alterations in product association with the resin and, as such, could also affect virus binding to the resin. Because the chromatography is operated in product FT mode, the pH and conductivity of the load and buffer were moved in the same direction. The study was designed in blocks such that if no impact is observed in one-half of the study, then the analysis can be completed. In one-half of full factorial, no impact to virus clearance was observed (Table V). Even at conditions known to induce the most change to chromatogram shape, virus removal was robust and effective (Table V).

Multimodal AEX chromatography is robust for virus clearance when operated in product FT mode. Two designs of experiments, conducted with one mAb, confirmed the robustness of the chromatography step to remove viruses. Virus clearance is not affected when bed height and flow rate are varied or when pH, conductivity, and temperature are varied.

3.1.4. Evaluation of the Viral Clearance Capacity and Robustness of the Manufacturing Process for the Recombinant Factor VIII (rFVII) Protein, Turoctocog Alfa (Søren Kamstrup; NovoNordisk A/S, Gentofte, Denmark)

Clotting factor concentrates sourced from human blood have been associated with blood-borne infections and pathogens such as the hepatitis B (HBV) and C (HCV) viruses, and the human immunodeficiency virus (HIV). Turoctocog alfa is a B-domain truncated rFVIII protein produced in the Chinese hamster ovary (CHO) cell line. As a third-generation rFVIII product, turoctocog alfa offers potential improvement over previous products through the elimination of the use of human- or animal-derived raw materials during its production. The manufacture of turoctocog alfa includes a five-step purification process developed to obtain a highly purified product with minimal viral contamination risk.

This presentation summarizes the virus clearance results obtained (expressed as LRV as defined in ICHQ5A) and presented to regulatory authorities for registration of turoctocog alfa. Originally, four model viruses were selected for the study:

• Ecotropic murine leukemia virus (eMuLV) as model for RVLPs expressed by CHO cells.

• MVM as model for small, highly resistant viruses, potentially present in raw materials.

• Reovirus, as medium-size, medium-resistant non-enveloped virus.

• Bovine enterovirus (a picornavirus) as an additional model for small, non-enveloped viruses.

During the registration process we were requested to supplement this panel with an additional enveloped virus (only one enveloped virus in original panel), and in response to this request we determined to study the clearance of infectious bovine rhinotracheitis virus (also known as bovine herpesvirus type 1) as a model for large, enveloped viruses. Four of the five purification steps were tested for virus clearance.

The capture step is a mixed-mode chromatography step and includes washing with detergent Triton X-100. As expected, this step is highly effective toward enveloped viruses (Triton X-100 inactivates enveloped viruses through disruption of lipid bilayer envelope) and less effective toward non-enveloped viruses (Table VI). The second purification step is an affinity chromatography step, using an anti-FVIII antibody to capture the FVIII molecule. The effectiveness of this step varied from moderate to high (LRV of 2.0 to >4.9; Table VI). A nanofiltration step including two Planova 20N filters in series was highly effective in reducing all model viruses: LRVs from 5.3 to >7.4 were observed (Table VI). The Capto mixed mode chromatography (MMC) and anti-FVIII chromatographic resins were tested both as new and as used at the maximum limit. In no case was virus clearance compromised by the use of aged resin.

In summary, the turoctocog alfa purification process was highly effective and robust with respect to clearance of viruses and resulted in the inactivation and removal of both enveloped and non-enveloped viruses of different sizes.

3.1.5. Evaluation of Agarose-Based Protein A Chromatography as a Reliable Virus Removal Step: Modeling and Outlier Analysis (John Ruppino and John Mattila; Regeneron, Tarrytown, NY)

Protein A chromatography was assessed for removal of XMuLV for 27 mAb manufacturing processes, resulting in a statistical model for clearance of virus by Protein A based on cell culture mAb titer as the primary input variable. The statistical analysis was performed to investigate an apparent trend of decreasing clearance as a function of bioreactor titer as shown in Figure 4, with clearance ranging from approximately 1 to 5 log₁₀ reduction factors (LRF). As the biotech industry experiences increases in bioreactor titers, there may be a need to move away from protein A chromatography as a consistent viral clearance unit operation and look to other orthogonal steps for added clearance.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

XMuLV clearance (LRF) by Protein A chromatography versus normalized mAb titer for 27 Regeneron mAb programs. Titer was normalized by dividing all titer values by the highest titer.

An initial main effects linear regression model was constructed based on normalized cell culture titer¹, resin type (traditional recombinant protein A versus alkaline stable), isotype (IgG1 and IgG4), column loading (25–45 g/L), and virus load (4–7 log₁₀ genome copies in load). The regression model was refined by removing nonsignificant terms (p >0.1), resulting in the simplified model shown in Equation 1. The simplified model shows that normalized cell culture titer is the only significant factor (with 90% confidence); the model accounts for 57% of the variation in the data set (R² = 0.57) with a standard deviation of 0.59 log₁₀.

The simplified model residuals were compared for each mAb to identify cases not explained by the regression model (Figure 5). Across 27 mAb programs, all have a mean LRF within 1 log₁₀ of the predicted value. Despite the range in actual XMuLV clearance of 1–5 LRF, the model fits 25 of the 27 programs. Only mAbs F and S showed statistically different mean residuals than other molecules, and these cases were assessed for root cause. Monoclonal antibody F was evaluated at a contract testing organization other than the one used for the other 26 programs, possibly implicating changes in virus prep or impurity load. There was no assignable cause for the high clearance seen in mAb S. While the model may not be applicable to broad “platform claims” without clearer understanding of the mechanism, it could be used to guide viral clearance study design such as exclusion of protein A in cases where minimal clearance is expected.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Plot of residual (actual-predicted) Protein A XMuLV LRF for each of 27 Regeneron mAb programs. All 27 programs have mean LRF values (center of diamond) that are within 1 log₁₀ of predicted value. Only mAbs F and S showed statistically different means than the other 25 mAbs.

Monte Carlo simulations were performed to develop an understanding of conditions with reliable viral clearance to predict removal for any “new” mAb, thus informing the decision to evaluate Protein A. Monte Carlo simulations showed that mAbs with a normalized titer of 0.4 ± 0.1 result in predicted >2 LRF of XMuLV by protein A with 99% confidence. The model predicts only 1–2 LRF of clearance as normalized mAb titer increases beyond that range. In these cases, it is important to maximize clearance demonstrated by effective steps due to minimal contribution from protein A.

It is unclear from this analysis if cell culture titer directly affects XMuLV LRF or if titer is correlated with another factor (cell density, impurity challenge, etc.) that is not included in the analysis. To support platform claims, further characterization is needed to expand the constructed model to account for more of the variation within the data set and to define the mechanism.

Summary

In this session, five companies shared their results and experience of using chromatography unit operations for virus clearance. The resin types used in the virus clearance evaluation in these studies are highly diverse, including mixed-mode Capto, ceramic HAP, mixed-mode and single-mode AEX, Protein A affinity, and antibody-based affinity. The LRV or LRF achieved for retroviruses vary broadly from ∼1 to >6. Depending on the mechanism of partitioning, virus removal by chromatography unit operations could be affected by various process parameters as well as biochemical/biophysical properties of the product molecules. Therefore, it is essential to understand the mechanism of action, key process parameters, possible interactions between virus particles and product, and potential limitations when selecting a given chromatography unit operation for virus clearance. In almost all cases, use of chromatography unit operations for virus clearance is secondary to the purification need (impurity removal and product yield) and chromatography unit operations usually are not optimized for effective virus removal. Thus, it is critical for virologists and purification scientists to work together during early stages of process development if it is deemed that one or more chromatography unit operations are needed for virus clearance in order to adequately ensure the product safety.