Viral Clearance Using Traditional, Well-Understood Unit Operations: Session 1.2. Anion Exchange Chromatography; and Session 1.3. Protein A Chromatography

Background

Process performance for both AEX (1, 2) and AEX membrane absorbers (3, 4) is generally effective and robust when performed within a bracketed set of operating conditions (1). The key process parameters that could affect viral clearance for the AEX step include pH, conductivity, resin loading, and reuse cycles.

Most applications of AEX chromatography in monoclonal antibody (MAb) purification are in the flowthrough (F/T) mode, where the MAb flows through (does not bind) while the process residuals bind. Hence, the relative levels of process residuals to MAb in the AEX feed should be considered when determining the MAb loading on the resin. The potential impact of competitive adsorption of impurities and viruses resulting in reduced viral clearance was discussed during the 2011 Viral Clearance Symposium (2). Further investigation into the mechanisms of impurity binding and impact of impurity levels on viral clearance in AEX F/T mode over a broader range of MAbs and viruses would be useful to refine the bracketed operating conditions and provide an explanation for outliers—cases where low log reduction values (LRVs) are achieved. One key question slated for discussion at the 2013 Viral Clearance Symposia was, what is the impact of host cell residuals (including complex with MAb) on the LRV for the AEX step? Since the mechanism of viral clearance by AEX chromatography has been previously demonstrated to be predominantly electrostatic (1), another pending question was, what is the impact of charge distribution of the virus on LRV, in addition to the isoelectric point (pI)?

Session Overview

During the 2013 Viral Clearance Symposia, LRV data for AEX F/T for a total of 42 MAbs purified over a range of operating conditions was reviewed. Genentech presented data on 26 MAbs (17 with AEX as the third chromatography step, nine with AEX as the second chromatography step), and Pfizer on 16 MAbs (AEX as the second chromatography step). In all cases independent of the location of the AEX chromatography step (second vs third), effective viral clearance (LRV > 4) was achieved for Xenotropic murine leukemia virus (XMuLV). For 25 of the MAbs, effective clearance was also observed for mouse minute virus (MMV) under the same operating conditions (data not available for MMV clearance for the other 16 MAbs), which is consistent with the clearance levels previously reported in the literature including the 2011 Viral Clearance Symposia (2,4). However, some variability in LRV was observed when AEX F/T was the second chromatography step when higher levels of process residuals were present.

One specific case study from Merck provided insights into the potential impact of competitive adsorption when a MAb is weakly bound to AEX operated in F/T mode. The case study also evaluated the impact of higher capacity resins on LRV. The results of this case study extend the previous observation (5) of the sensitivity of resin dynamic binding capacity to impurity load to MAb/virus systems where both the protein and virus properties can have a significant impact.

One common theme among the presentations is that a breakthrough in process residuals (Chinese hamster ovary (CHO) host cell protein (CHOP), and DNA) in AEX F/T mode correlates with a breakthrough of virus [XMuLV, MVM, and Pseudorabies virus (PrV)]. Operation of AEX F/T under “generic” operating conditions without host cell residual breakthrough results in robust and effective viral clearance.

Evaluation of Anion Exchange Chromatography Position in Purification Process on Virus Clearance (Sherrie Curtis and Mark Garcia, Genentech)

Many MAb purification processes include Anion Exchange (AEX) chromatography operated in F/T mode. When the AEX chromatography is positioned as the final chromatography step and the feedstock is highly pure, complete removal [no detectable virus signal by polymerase chain reaction (PCR)] or near complete removal (detectable virus PCR signal below the limit of quantitation) is reproducibly achieved for 17 MAbs. Effective virus clearance of >4 LRV has been achieved for 9/9 MAbs when the AEX is positioned directly post-affinity chromatography, where there are likely more product- and process-related impurities in the AEX feedstock (Table I). In one case, X-MuLV clearance was variable when evaluated multiple times (Figure 1). An aliquot of MAb U2, which had the variable LRV, was further purified through the purification process to be essentially free of impurities, and then processed through affinity chromatography to ensure the AEX feedstock was essentially the same but containing significantly less product- and process-related impurities. In two separate experiments, virus breakthrough was monitored as a function of MAb load density. The purified MAb U2 feedstock has higher virus retention than when compared to the feedstream that is less pure (Figure 2), which indicates that virus breakthrough is not MAb-specific but due to the presence of impurities. Virus breakthrough correlates with CHOP (CHO protein) load density as demonstrated by evaluating three MAbs of varying purity (Figure 3) and comparing the amount of CHOP loaded per milliliter of AEX resin when virus begins to break through. Virus begins to break through at 1 mg CHOP/mL resin. This correlation has been reproduced with additional MAbs (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Variable LRV can occur when AEX is positioned post-affinity chromatography.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Effect of feedstream purity on X-MuLV LRV using MAb U2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

XMuLV breakthrough correlates with CHOP load density on AEX resin using three MAbs.

Virus Removal by AEX Chromatography in Flowthrough (F/T) Mode: Studies To Support Submission of Marketing Authorization (Richard Chen, Imclone)

ImClone Systems, a wholly owned subsidiary of Eli Lilly and Company, performed a series of AEX virus challenge and auxiliary studies to support the submission of marketing authorization for a MAb process. Challenge studies were performed to ensure adequate clearance data for a full panel of four viruses—XMuLV, MMV, PrV, and Bovine viral diarrhea virus (BVDV). The model virus BVDV was included in the virus panel based on the use of an animal derived raw material for the NS0 cell culture process. A combination of worst-case manufacturing conditions were simulated for the challenge studies and included other process controls such as aging of the AEX chromatography resin to end of service life and cleaning effectiveness to mitigate virus carryover.

An AEX model qualification study was completed before the virus challenge studies to verify that the model was representative of the commercial process. Two types of aged resins were evaluated. Clean-in-place (CIP)-aged resin columns (90 cycles) were evaluated for the full virus panel. A single product-aged resin column (90 cycles) was evaluated for only XMuLV, the most relevant specific model virus for the process. CIP/virus carryover studies were also executed for XMuLV (new and product-aged columns). Each CIP/virus carryover (CO) run was performed by executing a blank run (i.e., no virus spike) on the same column as the preceding XMuLV challenge run. A summary of all studies (and study linkages) performed to support marketing authorization is provided in Figure 4. Resulting log reduction factors (LRFs) are provided in Table II.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Overview of all AEX studies to support marketing authorization.

Significant and consistent virus reduction was obtained for all four model viruses. In addition, there was no detectable virus for any of the XMuLV carryover blank runs, thereby verifying a robust AEX CIP process. Proper planning and scheduling was a key factor for the successful execution of all AEX viral clearance studies due to study linkages and long duration times for some precursor studies (e.g., resin re-use study).

AEX Chromatography: Case Study in Unexpectedly Anomalous Viral Clearance (David Roush, Peter Machielsen, Miranda van Beuningen, Merck, Sharp and Dohme)

A case study was presented with an AEX operated in F/T mode as a second chromatography step following protein A chromatography. Although the process was operated within proposed generic operating conditions (buffer, pH, conductivity, loading, and number of reuses) presented in the literature (1) and with a well characterized resin (6), the LRV was less than 1.4 for all four viruses examined (X-MuLV, MVM, PrV, Reo-3). This observation was in stark contrast to the expectation of LRV > 4 based on literature precedent (1, 7⇓–9).

Four potential root causes/hypotheses were proposed as follows: (1) virus retention and associated LRV were sensitive to competitive adsorption of impurities (resin-specific?); (2) differences in electrostatic retention mechanism between MAb and virus were too small (e.g., ΔpI of MAb and virus, partitioning of virus and MAb were based on charge patches?); (3) the capacity factor, k', for the MAb was too low (not predicted based on pI of MAb); (4) weak retention of both product and viruses contributed to overall low LRV.

Experiments were planned and executed to obtain the necessary mechanistic data to address potential root causes, specifically, evaluation of the breakthrough of impurities and virus, sensitivity analysis to evaluate the impact of operating conditions on breakthrough, and the overall cumulative effect (impact of impurity and loading). The experimental studies evaluated the impact of process conditions including loading, pH, and conductivity on LRV. One additional variable examined was the impact of the resin on the LRV with the same feedstock.

Experimental design involved collection and analysis of fractions across the elution pool to determine relative partitioning of virus and impurities. Process residuals that were tracked included aggregates HMW (high molecular weight species), host cell protein (HCP), and DNA. Virus titers for PrV and MVM were also tracked. Examination of the reconstructed chromatographs developed from the fraction analysis illustrates co-elution of the two viruses with the process residuals (see Figure 5), supporting the hypothesis of competitive adsorption of virus and impurities.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Reconstructed chromatograph illustrating product, impurity, and virus elution in AEX F/T on Q-Sepharose FF.

Potential solutions examined included change in operating conditions (reduce pH, conductivity, and loading), evaluation of a higher capacity resin, or reduction in impurity load in the feed. Reduction in the loading (50 to 30 g/L), conductivity (6 to 4 mS/cm), or a combination of both did not have a significant impact on LRV observed for MVM or PrV. The lack of impact of loading and conductivity was expected and is consistent with the literature precedent, as the operating conditions were well within the “generic” operating condition space. The combination of reduced loading (50 to 30 g/L) and reduction in pH (7.5 to 7.2) had a modest impact of MVM LRV (0.5 to 1.4) and a significant impact on LRV for PrV (<1 to >5.3).

Another set of experiments independently assessed the impact of resin and impurity load on LRV for XMuLV and MVM. The same feedstock was employed for both resins but a 4× higher loading was evaluated for an alternate resin (resin B). Resin B had a significantly higher LRV for both XMuLV and MVM, and the LRV that were consistent with previous published data (LRV > 4). The combination of employing resin B, further optimization of the operating/buffer conditions, and an associated reduction in HCP (range of 2.3× to 3.5×) and reduction in DNA (46×) further increased the clearance of XMuLV (LRV > 6.7) and MVM (LRV > 7.8) to the high end of the range reported in the literature.

Conclusions from this case study are that competitive adsorption of viruses and impurities can exist under conditions of weak MAb retention, and breakthrough of DNA and HCP correlate with breakthrough of virus (MVM and PrV) in AEX chromatography when operated in F/T mode. The competitive adsorption challenge could be resolved by utilizing a combination of a different AEX ligand chemistry and a higher capacity (higher ligand density) resin.

Using Phage To Understand the Impact of Process Parameters on Virus Clearance for a Platform AEX Chromatography Step (Timothy Iskra, Chris Gallo, and Ranga Godavarti, Pfizer Global Biologics; and Kurt Brorson and Scott Lute, Center for Drug Evaluation and Research, U.S. Food and Drug Administration)

AEX chromatography operated under a weak partitioning mode has proven to be a robust process for the removal of a wide range of impurities and contaminants including virus. To gather further data to support the development and use of a modular viral clearance package, initial clearance studies were conducted using bacteriophage as a surrogate for virus. Pseudomonas phage 7 (PPV), and φX174 bacteriophage were used as models to identify process parameters that would likely have the greatest influence on virus clearance for the platform AEX step. Using four different MAbs, a multivariate design-of-experiments (DOE) study was conducted to investigate several process parameters including load challenge (both product and impurities), load pH, load conductivity, contact time, and bed height. These molecules were evaluated under both weak partitioning and F/T conditions to expand the range of operation.

Results from these studies showed clear removal of PP7 (pI = 4.3–4.9) over a wide range of operating conditions including a load pH of 7.0–8.7, load chloride of 10–75 mM, bed heights from 10 to 20 cm, and a load challenge up to 250 g/L (Figure 6). A range of impurities were also challenged based on the molecules chosen to include high levels of DNA, HCP, and aggregate. Although PP7 was effectively removed, φX174, which has a pI of 6.7–7.0, showed that the concentration of salt greatly affected clearance with little to no removal at or above 45 mM (Figure 7). These results indicate that φX174 is not an appropriate surrogate to model viral clearance over the AEX chromatography step because robust clearance for both a model retrovirus and parvovirus has been historically achieved over a much greater range of salt concentration.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

(A) PP7 removal using four different MAbs preparations operated under AEX weak partitioning (WP) conditions and AEX flowthrough (F/T) conditions. Line indicates LRV of 4 demonstrating a robust step. (B) Plot showing load pH and load chloride levels for four MAbs evaluated and the respective WP conditions and F/T conditions. WP conditions are indicated by red circles while F/T conditions are indicated by the blue circle.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Impact of salt concentration on removal of φX174 over the AEX chromatography step. As shown, there is a complete loss of φX174 removal within a 20 mM concentration range of NaCl.

Data from this initial clearance study with bacteriophage will help define an appropriate operating range for evaluation of clearance with a model retrovirus and parvovirus, which will ultimately lead to the development of a modular viral clearance package for AEX chromatography operated in the weak partitioning mode.

Session 1.2 Conclusions

Data presented for four MAb/virus systems suggest that process residuals (CHOP and DNA) at high enough levels in AEX feed are associated with virus breakthrough and loss of LRV for multiple viruses (XMuLV, MMV, and PrV).

A prospective, side-by-side comparison of purified MAb and AEX feed (affinity-purified) demonstrated a drop in LRV with increase in resin loading (∼60 g MAb/L resin), as presented in Figure 3. These data suggest that XMuLV breakthrough correlates with increased CHOP loading. However, the potential impact or contribution of DNA involvement in the virus breakthrough was not independently evaluated.

Hence, the data presented at the 2013 Viral Clearance Symposia suggest a strong correlation with host cell residual breakthrough and virus breakthrough over a range of MAb and processing conditions. These results suggest that host cell residual breakthrough (CHOP, DNA) could be used as a surrogate for virus breakthrough, associated with reduced LRV. These observations are also consistent with previous research on host cell residual breakthrough reported for AEX membrane absorbers and loss of LRV (4).

A case study extended the application of the “generic operating conditions” for AEX to four viruses (XMuLV, MVM, PrV, and BVDV) where effective clearance or LRV ∼4 was achieved with naïve and used resins. This research extends the combination of worst-case conditions and resin reuse (aged and CIP-aged) of up to 90 cycles from the initial value of 50 cycles proposed for generic conditions (1).

A case study evaluating the use of bacteriophage as a surrogate for viral clearance in AEX chromatography (both F/T and weak partitioning modes) demonstrated feasibility of using phage (PP7) although the pI of the PP7 is relatively low (4.3–4.9). However, some significant challenges and limitations were encountered with the use of φX174. One important note is that the pI for φX174 (6.7–7.0) is closer to the pI range (5.4–6.2) for viruses (1) typically used in viral clearance studies (SV40, XMuLV, MMV) than the pI for PP7 (4.3–4.9). Since the pI of φX174 is higher than typical viruses and closer to the pI of MAbs used in viral clearance studies, this may result in reduced or weaker retention on AEX, potential co-elution with the MAb, and hence reduced LRV. One potential interpretation of the research with φX174 is that use of a surrogate for virus partitioning is complex in that matching the pI of the virus with the phage is insufficient and that there is a potential impact of charge distribution of the virus on LRV in addition to the pI. Hence, additional research on the use of phage to resolve these technical challenges is warranted.

Background

Affinity-based purification of MAbs via protein A is widely employed in the industry. Although extensive characterization on the impact of cleaning and reuse on protein A performance have been completed (10, 11) and techniques [for example, quantitative PCR (qPCR)] are available to evaluate inactivation from partitioning (12), the specific mechanisms governing virus partitioning which result in variable LRV of the step are still unclear (13). Hence, the protein A step remains a well-established unit operation that has some knowledge gaps.

Owing to the variable clearance observed among viruses and products for protein A (1, 8, 13), additional focus on better mechanistic understanding of virus retention on protein A is warranted.

Overview

Improvements in the mechanism of virus partitioning and removal in the protein A step could allow for development of a more integrated viral clearance/viral inactivation process including the potential to combine the protein A/low-pH inactivation step. Presentations at the 2013 Viral Clearance Symposia focused on the ability to retain the orthogonality of the mechanisms for viral removal and inactivation, for example, use of a detergent wash in protein A for removal from low pH inactivation.

Mechanism of XMuLV Clearance by Protein A Chromatography (Julia Bach, Shivanthi Chinniah, and Lisa Connell-Crowley, Amgen)

Protein A chromatography is a relatively poor clearance step for XMuLV. Typical XMuLV clearance by protein A chromatography is 1–4 LRV, although XMuLV removal is highly reproducible for a given protein A step. The mechanism of XMuLV removal by protein A was studied by examining the partitioning of XMuLV on protein A resin in equilibration buffer alone or in the presence of product and/or impurities. Table III shows that when XMuLV is loaded onto the column in buffer alone or in MAb-free harvested cell culture fluid (HCCF with MAb removed), where the majority of XMuLV flows through the column, resulting in an elution pool with an LRV above 4. However, when XMuLV is loaded onto the column in the presence of product, either as purified drug substance or in HCCF, a portion of the virus co-elutes with the product, which results in lower elution LRVs. These data suggest some of the XMuLV interacts with MAb bound to the resin, resulting in poor to moderate removal of XMuLV by the step.

In an attempt to improve XMuLV clearance by disrupting the interaction between the virus and product, various column washes such as high-salt, detergent, and urea/Isopropanol (IPA) were examined for their impact on XMuLV LRV in the elution pool (Table IV). While salt A had a small, if any, effect, washing the column with salt B, Triton X-100, or 3M urea/20% IPA significantly improved XMuLV removal from 1.5 logs up to 4 logs. However, virus was still detectible in the elution pool by qPCR with the product regardless of the wash.

Since XMuLV appeared to be difficult to completely remove from the column, an additional experiment was performed to look at inactivating the virus that remained on the column to further improve the clearance of the step. Three of the washes used in the previous experiment were shown to inactivate XMuLV in solution and, as expected, only trace amounts of infectious virus were detected in the elution pool after these washes were applied to the column (Table V). In cases where a 30 minute static hold was used during the inactivating wash phase, >6 LRV were achieved with no XMuLV detected in the elution pool. These data indicate that a wash step can significantly improve XMuLV clearance by protein A chromatography by either improving removal or by incorporating both removal and inactivation into a single step.

Inactivation of Viruses Using Novel Protein A Wash Buffers (Glen Bolton, Keith Selvitelli, Ionela Iliescu, Doug Cecchini, Biogen Idec)

A low-pH viral inactivation step is typically performed in the eluate pool following the Protein A capture step during the manufacturing of MAbs and Fc-fusion proteins. However, exposure to acidic conditions has the potential to alter protein quality. This can be worse at large scale where it can be time-consuming and difficult to manually acidify, mix, transfer, and then neutralize a large pool. The acidification and subsequent neutralization can increase pool conductivity, which can affect the subsequent ion exchange chromatography step. When performing simulated moving bed or continuous multicolumn chromatography, it is challenging to perform the low-pH inactivation step on the many elution pools that are generated. To avoid these difficulties, novel protein A wash buffers capable of inactivating viruses while antibodies were bound to chromatographic resins were developed.

By equilibrating the column in high-salt buffer (2 M ammonium sulfate) after loading, the interactions between antibodies and protein A ligands were increased enough to prevent elution at pH 3. The ammonium sulfate was also found to cause binding of an antibody to a mixed-mode AEX resin (Capto Adhere) at a pH value (3.5) that caused elution in a conventional AEX resin. This indicated that retention was likely due to enhanced hydrophobic interactions.

The potential of the 2 M ammonium sulfate pH 3 buffer, a 1 M Arginine buffer, and a buffer containing the detergent (LDAO) to inactivate viruses when used as protein A wash buffers with a 1 h contact time were studied. The high-salt and detergent-containing column wash buffers provided about 5 logs of removal determined using PCR, and complete combined removal and inactivation (>6 log) determined by measuring infectivity. The arginine provided complete removal of XMuLV, as determined using PCR.

The novel protein A washes could provide more rapid, automated viral inactivation steps with lower pool conductivities. In free solution, the 2 M ammonium sulfate pH 3 buffer did not completely inactivate XMuLV virus, probably due to viral aggregation, indicating that this buffer may be more effective when used during protein A chromatography.

Session 1.3 Conclusions

Research presented at the 2013 Viral Clearance Symposium shed some additional insights into the mechanism of virus retention on protein A chromatography. For XMuLV, when spiked into the protein A feed, the majority of virus flows through the column but a fraction (4 to 5 log) was observed to interact with the ligand and to co-elute with MAbs and/or impurities. Similar results were observed with purified MAbs, suggesting one potential mechanism of retention of XMuLV on protein A is via a MAb/virus interaction. Addition of detergents allowed for virus inactivation and/or dissociation of the virus from the MAbs.

In the spirit of developing an integrated and more efficient viral clearance/inactivation, the concept of exploiting the MAb/virus interactions in protein A, which may have contributed to variable LRV historically, to combine the protein A and low pH inactivation step into one unit operation was presented. This combined approach could minimize potential product quality changes associated with low-pH conditions used in viral inactivation but may also result in overall lower LRV for the integrated step. Additional mechanistic experiments to map the variability in protein A LRV to more MAb/virus systems with various species (salt washes, detergents) would be extremely beneficial. The results from these experiments would improve the viral clearance capabilities for the protein A step via more predictable performance/reduced variability.