Viral Clearance Using Traditional, Well-Understood Unit Operations (Session I): Virus-Retentive Filtration ========================================================================================================== * Dayue Chen ## Background Virus-retentive filtration is routinely incorporated into downstream purification processes as a dedicated virus removal unit operation to provide robust and reliable virus clearance capacity. In recent years, use of viral filters designed for parvovirus removal (one of the smallest mammalian viruses, ∼20 nm in diameter) has become an industry standard and regulatory expectation. It is generally agreed that virus-retentive filtration technology removes viruses primarily by size exclusion (1, 2). Therefore, virus filters designed to retain parvoviruses are expected to provide reliable and effective removal of larger viruses such as retroviruses (∼100 nm). The fact that no retrovirus breakthrough of parvovirus filters (1) has been reported so far in the literature is consistent with the size exclusion mechanism. In comparison to other dedicated virus clearance unit operations such as low pH or heat treatment, virus-retentive filtration is much gentler, thus minimizing potential adverse impacts on product quality. In this session, diversified topics, ranging from generic validation with a single virus to root cause investigation of unexpected results, were discussed. ## Generic Validation Given that the mechanism of virus removal by virus-retentive filtration is size-based exclusion, it is natural to rationalize that clearance studies with larger model viruses are unnecessary when robust clearance of parvovirus has already been demonstrated. In fact, a potential key area of improvement described in the 2009 Viral Clearance Symposium white paper (1) was to assess the feasibility of using parvovirus (e.g., mouse minute virus, MMV) as a single model virus to establish virus clearance across parvovirus-retentive filters. Several firms provided data on this particular topic. While approaches taken by different firms varied, collective studies from the industry as a whole indicate that product and patient safety will not be compromised with this type of generic validation strategy. ## Characterization of the Virus-Retentive Filtration Unit Operation Careful characterization is essential for understanding the robustness of the virus-retentive filtration unit operation and setting appropriate operational ranges to ensure product safety. Participants from several companies shared their data and experience on this topic. Janssen Pharmaceutical provided data indicating that multiple parameters could affect the effectiveness of virus removal by the virus-retentive filtration unit operation and ranked several parameters based on statistical analysis. The study by Eli Lilly and Company indicated that depressurization prior to buffer flushing could result in significant parvovirus breakthrough. Genentech shared their study of using cation exchange chromatography (CEX) guard columns to improve process capacity. ## Generic Validation with a Single Model Virus (Eva Gefroh, Amgen) Historically, Planova 20N results from 24 monoclonal antibody (mAb) processes indicate that log reduction values (LRVs) achieved with MMV and murine leukemia virus (MuLV) are comparable, as shown in Figure 1. In an overwhelming majority of the cases (29 out of 30), no MMV was detected in the filtrate. These results present a compelling evidence for using only MMV, the smallest of all commonly used model virus in filter validation, to claim LRV for all viruses ≥20 nm in size. ![Figure 1](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F1.medium.gif) [Figure 1](http://journal.pda.org/content/68/1/38/F1) Figure 1 Amgen's cumulative experience with xMuLV and MMV clearance in Planova 20N validation studies. Modular clearance has been employed for process steps with a demonstrated consistent, robust clearance (1). Based on Amgen's experience with parvovirus-retentive filters, it is believed that process conditions appropriate for a specific filter can produce consistent, robust clearance. Amgen performed a screening design of experiment (DoE) with the Viresolve® Pro (EMD Millipore) filter, evaluating the impact of pressure, pH, sodium chloride concentration, and protein concentration in a wide operating space. The mAb feedstock material was pre-filtered by either the Viresolve® Pre-filter or the Viresolve® Pro+ Shield, and MMV was used as the model virus. The study was based on a resolution III screening design with five independent conditions tested in duplicate, for a total of 10 runs. A throughput target of 500 L/m2 was set for all conditions. This required the use of the inline spiking strategy for a couple of the runs to eliminate the pre-filter decoupling effect; details of the inline spiking technique are described in a separate paper (3). As shown in Table I, consistent and robust LRVs were achieved under all conditions evaluated, including a range of flow decay that was observed due to process conditions. To further test the robustness of the Viresolve® Pro filter, the potential impact of static holding was also assessed. Specifically, after the completion of the mAb + MMV loading from the screening DoE study, the filters were flushed with the corresponding buffer, held statically for 20 h, and the subsequent flush post-hold was assayed for virus. Results are shown in Table II as a comparison of the LRV before hold and LRV of the product + flush after a 20 h hold. The results show that the static hold had minimal impact on virus clearance, further demonstrating the robustness of the virus filter. View this table: [Table I](http://journal.pda.org/content/68/1/38/T1) Table I Summary of Amgen DOE Screening Study of MMV removal by Viresolve Pro View this table: [Table II](http://journal.pda.org/content/68/1/38/T2) Table II Amgen's DOE Screening Study Evaluating the Effect of Static Hold on MMV Clearance by Viresolve Pro Filters Taken together, these data demonstrate that the virus filters designed for parvovirus removal can achieve consistent, robust parvovirus clearance. In addition, it is scientifically justifiable to use MMV as the worst-case model virus for demonstrating robust, effective clearance of larger viruses for parvovirus filters. Furthermore, when robust operating spaces can be defined for a parvovirus filter that shows effective clearance over a wide range of conditions, the concept of generic validation for this unit operation should be considered. ## Virus-Retentive Filtration, Genentech Experience (Rachel Specht, Aimee Durbin; Genentech, a Member of the Roche Group) Genentech has extensive experience with both large and small virus-retentive filters. Breakthrough of MuLV has not been observed with Viresolve® NFR, Viresolve® NFP, or Viresolve® Pro filters, based on Genentech's experience. In addition, no process parameters have been identified that affect retrovirus retention for these filters, making it suitable for modular application when the filtration process is operated within ranges previously tested. On the other hand, parvovirus breakthrough is not an uncommon phenomenon (1, 4, 5). It has been previously shown that flux decay often accompanies a decrease in virus retention for the Viresolve® NFP filter (6). The correlation between flux decay and parvovirus breakthrough has not been observed for the Viresolve® Pro filter, as shown in Figure 2. When 16 mAbs with wide ranges of flux, volume throughput, pH, conductivity, and protein concentration were examined, no significant parvovirus breakthrough of the Viresolve® Pro filter was observed based on the limit of quantification (LOQ) using quantitative polymerase chain reaction (Q-PCR). Complete or near-complete MMV removal was achieved within broad ranges of permeability decay (19 to 82%) by the Viresolve® Pro filter. ![Figure 2](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F2.medium.gif) [Figure 2](http://journal.pda.org/content/68/1/38/F2) Figure 2 Genentech's experience with MMV removal by Viresolve Pro filter after flux decay. Claimable LRVs obtained by virus-retentive filtration can be influenced by multiple factors such as virus titer and loss of virus during pre-filtration prior to virus-retentive filtration. Table III shows the impact of virus stock titer from four individual MMV stocks on the overall claimable LRV achieved by the Viresolve® Pro filter. The claimable LRV is limited by the load titer that is determined by the stock titer and the pre-filtration loss. Complete or near-complete MMV removal is observed for all eight mAbs. For all MMV stocks examined there is no decrease in volume throughput caused by the virus spike when the virus stock is sonicated before use (data not shown). The results for the four different MMV stocks with varying infectivity titers highlight the need for a consistent, high-titer virus stock that would result in predictable LRVs for the Viresolve® Pro filter. View this table: [Table III](http://journal.pda.org/content/68/1/38/T3) Table III MMV LRVs Obtained in a Genentech Study of Viresolve Pro Filtration of Different MMV Stock Preparations. There appears to be a general trend between stock infectivity titer (far left column) and achievable LRV (far right column). ## Defining Robust Operating Ranges for Virus-Retentive Filtration in a mAb Platform Process for Reproducible Viral Clearance (Hong Shen, Dominick Vacante, and Pedro Alfonso; Active Pharmaceutical Ingredient - Large Molecule, Janssen Pharmaceuticals, Johnson & Johnson) The most common virus-retentive filtration modes used are nominal (dead-end) and tangential flow filtration. The filtration mode, membrane composition, virus spike, mAb, filtration buffer, and virus-retentive filtration parameters used may influence the virus-retentive filtration operation. Janssen's platform utilizes nominal size dead end filter to remove viruses larger than 20 nm. In order to study Janssen's virus-retentive filtration platform, a representative mAb in-process intermediate was spiked with MMV and analyzed for key performance filtration parameters, with regard to their ranking in importance in affecting virus clearance as measured by LRV. The analysis resulted in a polynomial model fit against combined data for pressure (9 to 14 psi), mass load (50 to 1250 g/m2), volume load (14 to 310 L/m2), buffer flush volume (0 to 10 L/m2), immunoglobulin G (IgG) load concentration (1 to 16 g/L), and virus spike ratio (0.5 to 1.0%). The model fit is supported by comparing LRVactual versus LRVpredicted, showing an R square value of 0.92, *P*-value of less than 0.0001, and square root of mean squared error (RMSE) of 0.2154 (Figure 3). The polynomial model fit was used to understand process parameter interactions and ranking. As shown in Figure 4a, the interaction between protein mass load (mAb) with spike ratio, total volume load with spike ratio, buffer flush volume with pressure, and mAb concentration with pressure showed statistical significance with *P* ≤ 0.002 and were the highest ranking in importance. Those interaction terms are demonstrated as nonparallel lines in Figure 4b. For example, in the plot of total volume load with spike ratio, located at the end of the third row, the effect of total volume load is small at the low values of spike ratio, but it diverges widely for the high values of spike ratio (Figure 4b). This indicates that no virus breakthrough occurs at the low value of total volume load (14 L/m2), and therefore LRV increases with ascending spike ratio. In contrast, virus breakthrough occurs at the high value of total volume load (310 L/m2), so LRV decreases significantly with an increasing spike ratio. In this fitted model, the interactions tend to mask the effect of main effects. Main factors with statistical significance that followed in ranking were shown to be the following: Total volume load > Pressure > Buffer flush volume > mAb mass load. Load protein (mAb) concentration and spike ratio alone did not show any statistical significance (Figure 4a). Spike ratio was the lowest-ranked single parameter. No further comment was provided. ![Figure 3](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F3.medium.gif) [Figure 3](http://journal.pda.org/content/68/1/38/F3) Figure 3 Plot of filtration LRV actual values obtained by Janssen versus LRV predicted by polynomial fit model. ![Figure 4](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F4.medium.gif) [Figure 4](http://journal.pda.org/content/68/1/38/F4) Figure 4 Result from Janssen filtration design of experiment filtration study. Shown are (a) sorted parameter estimates, (b) interaction profiles; see text for details. Janssen's filtration platform database, consisting of over 100 data points, shows a window for virus-retentive filtration LRV claims of >4 LRV under a wide mAb mass load, salt, and mAb concentration as shown in Figure 5. The green box within each plot is the robust region, with LRV clearance greater than approximately 4.0, and this robust region can be outlined by ranges of pressure 9–14 psi, mAb load 200–2500 g/m2, mAb load concentration 7–17 mg/mL, buffer pH 6.7–7.1, and buffer salt 0–150 mM. However, it should be pointed out that the pH and salt conditions are largely overlapped in the current database, so additional data distributed outside of the triangle area is needed to complete the robust region in the right contour plot in Figure 5. ![Figure 5](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F5.medium.gif) [Figure 5](http://journal.pda.org/content/68/1/38/F5) Figure 5 Contour plots for LRV from Janssen filter DoE study. Left: load mAb concentration versus mAb mass load. Right: buffer pH versus salt (mM). In conclusion, analysis of Janssen filtration platform using a representative mAb showed combined and single factors with statistical significance that affect LRV. Janssen's platform filtration process is robust as outlined and capable of removing 4 log10. ## Evaluation of a 20 nm Hollow-Fiber Filter for Parvovirus Clearance (Adith Venkiteshwaran, J. Nolting, and Dayue Chen; Eli Lilly and Company) Breakthrough of parvovirus has been commonly observed during post-filtration buffer flushing. In order to effectively control or mitigate this potential risk, a study was carried out to evaluate whether the observed breakthrough can be influenced by operating pressure (in a constant pressure operation) and depressurization prior to the buffer flushing. All experiments were performed with a hollow-fiber virus filter module of 0.001 m2 surface area (Planova BioEX) and utilized two mAbs (mAb-1 and mAb-2) spiked with 0.1% porcine parvovirus (PPV) (vol/vol). Laboratory-scale virus-retentive filtration experiments were carried out at 15, 30, or 45 psi. The filtrate and flush pools were collected separately. To monitor the kinetics of potential breakthrough, three *grab samples* (i.e., instantaneous sampling during a run) were collected for each individual experiment. All samples were titrated by a median tissue culture infective dose (TCID50) assay using PK-13 indicator cells. As shown in Figure 6, PPV breakthrough in the filtrate pool was only detected in experiments carried out at the lowest pressure evaluated (15 psi). Also, significant PPV (3–4 log10 TCID50/mL) was observed in the buffer flushing samples in all cases. It was not surprising to detect PPV in the flush when a consistent breakthrough of virus was detected throughout filtration (15 psi; Figure 6, panel A), but it was unexpected in the flush for the 30 and 45 psi cases. The overall LRV in the filtrate decreased by 1.5 to 2 log10 after adjusting for differences in volume of the filtrate and buffer flush pools. ![Figure 6](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F6.medium.gif) [Figure 6](http://journal.pda.org/content/68/1/38/F6) Figure 6 PPV titers from Lilly mAb1 filter grab samples at (A) 15 psi, (B) 30 psi, and (C) 45 psi, with depressurization. Breakthrough in filtrate pool samples was only observed in runs performed at 15 psi. Based on observed kinetics (grab samples), the presence of PPV in flush samples from 15 psi runs was anticipated but was unexpected for the 30 psi and 45 psi runs. A similar observation was made for mAb-2 (data not shown). Flow was stopped between the end of filtration and the start of buffer flushing in all these experiments by depressurization. To assess whether depressurization played any role in the PPV breakthrough during buffer flushing, the experiments were repeated with mAb-1 in the exact same manner without depressurization. As shown in Figure 7, PPV breakthrough was observed in all the samples from the 15 psi experiments. On the other hand, no PPV was detected in any of the samples including buffer flushing filtrates at 30 and 45 psi. ![Figure 7](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F7.medium.gif) [Figure 7](http://journal.pda.org/content/68/1/38/F7) Figure 7 PPV titers for Lilly mAb1 filter grab samples at (a) 15 psi, (b) 30 psi, and (c) 45 psi, without depressurization. Detectable PPV in filtrate and flush pool samples was only observed in runs performed at 15 psi. Note that unlike the pressurized runs (see Figure 16), PPV in flush samples was not detected at 30 psi and 45 psi. In summary, the Planova BioEX hollow-fiber virus-retentive filtration module provided effective PPV clearance at the 30 and 45 psi operating pressures as recommended by the manufacturer. However, when the filter was flushed with buffer following the mAb filtration with depressurization, significant PPV breakthrough was observed that resulted in a 1.5 to 2 log10 decrease in overall LRV. Consistent PPV breakthrough during mAb filtration was observed at 15 psi (outside of the manufacturer's recommendations). This study indicates that careful evaluation of virus breakthrough with respect to operating pressure range and depressurization is essential in order to understand the operating space and to set appropriate control for the virus-retentive filtration unit operation. ## Parvovirus Filter Protection via CEX Guard Columns (J. Bill Jr.; Genentech, a Member of the Roche Group) Parvovirus filters are an effective, size-based, virus removal step used in many purification processes for biologicals such as mAbs derived from mammalian cell culture. However, due to the propensity to foul resulting in lower-than-expected capacities, the cost associated with virus-retentive filtration can be expensive. Previously, guard media such as CEX membranes have been shown to increase capacity over somewhat ineffective microfilters with 0.2 μm diameter pores when used prior to the virus filter (7). Other CEX media, such as resins, have not been evaluated for this application, even though CEX resin chromatography operations are a known approach to remove impurities (host cellular proteins DNA) and aggregates (1). To assess whether CEX resins could be used as a guard column to a parvovirus filter, two resins with high recommended linear velocities were evaluated while varying bed height and flow rate. The guard columns were located upstream of a Millipore Viresolve® Pro (VPro). A mAb feedstock stream at pH 5.5, 6 mS/cm, and 5 g/L was filtered at constant flux and the resulting VPro permeability (liter/m2/h /psi) was graphed versus VPro capacity (kg/m2). The data show that the guard columns enable the VPro to reach higher capacities than those observed while using 0.2 μm microfilters under all conditions evaluated (Figure 8). Additionally, shorter bed heights result in reduced VPro permeability for Resin 1 (Figure 8a). In contrast, bed heights from 1.0 to 7.5 cm have no significant impact on the VPro permeability for Resin 2, as shown in Figure 8b. Figures 8c and 8d show that while holding bed height constant and varying flow rate, permeability trends for both resins are more consistent, resulting in VPro capacities of 10–15 kg/m2 for Resin 1 and up to 50 kg/m2 for Resin 2. ![Figure 8](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F8.medium.gif) [Figure 8](http://journal.pda.org/content/68/1/38/F8) Figure 8 Effect of bed height and flow rate on VPro capacity and flux decay in Genentech study evaluating CEX prefiltration (Resin 1 and Resin 2). (a) Resin 1, 1.2 mL/min; (b) Resin 2, 1.8–2.1 mL/min range; (c) Resin 1, 4.8 cm bed height; (d) Resin 2, 4.8 cm bed height. These results suggest that varying bed height at constant flow rate, which simultaneously affects column volume and residence time, has the potential to adversely influence parvovirus-retentive filter performance, as seen for Resin 1. The relatively consistent performance observed with constant bed height and varying flow rate, which only affects residence time, suggests the that adverse impact is primarily due to changes in column volume. In comparison to Resin 1, Resin 2 shows little effect from column volume or residence time. Consequently, Resin 2 is a better choice for VPro protection in a manufacturing setting. To draw comparison to previous work demonstrating increased filter capacities using CEX guard membranes, guard columns containing Resin 2 were tested alongside the guard membrane Millipore Shield using the previously described mAb feedstock stream. The data in Figure 9 show that VPro was able to reach capacities of 5.4 kg/m2, 6.9 kg/m2, and 10.6 kg/m2 using the 1 cm guard column, Millipore Shield, and 2 cm guard column, respectively. The capacity data were used to calculate mAb processed (kg) assuming an equivalent area of VPro for each guard media. Guard column prices were based solely on the resin price (no packing, labor, or capital associated costs), while the Millipore Shield and VPro prices were based on their respective marketed prices. All data was then normalized against the Millipore Shield filtration data for a more simplified comparison. ![Figure 9](http://journal.pda.org/https://journal.pda.org/content/pdajpst/68/1/38/F9.medium.gif) [Figure 9](http://journal.pda.org/content/68/1/38/F9) Figure 9 Process economics of resin 2 guard columns at 1 and 2 cm bed heights (0.34 and 0.68 mL, respectively) versus CEX membrane (Millipore Shield) at 1.6 mL/min in Genentech pre-filtration study. As also shown in Figure 9, the 1 cm bed height guard column is cheaper (0.8×) which offsets the lower mAb filtered (0.8×) despite a lower capacity, yielding a slightly higher operating cost (1.1×) when compared to the Millipore Shield. The 2 cm bed height guard column would be able to filter more mAb (1.5×) and was slightly cheaper (0.8×), yielding a significantly lower operating cost (0.6×). This demonstrates that a guard column can be economically competitive, or cheaper than, a CEX membrane. In summary, CEX resins can be used to effectively protect a parvovirus filter, and may be a more economic guard media for reducing the overall cost of goods. The resin used, bed height and/or volume, and residence time through the column may all impact performance and should be thoroughly studied. Practical implementation concerns can be addressed through thoughtful selection of the resin and design of the guard column to be used. Pre-packed, single-use columns can be leveraged to eliminate labor intensive column packing operations, cleaning validation, or re-use studies. Properly selected column dimensions can provide sufficient residence time while also minimizing flow resistance through the column. ## Systematic Approach for Characterization of the Virus-Retentive Filtration Unit Operation: Additional Knowledge or Extra Frustration? (D. Chen, A. Venkiteshwaran, and D. Strauss; Eli Lilly and Company) Despite significant advancement in recent years, virus-retentive filtration remains one of the most challenging unit operations to define a data-based design space based on laboratory scale spike/recovery studies. In an attempt to better understand virus-retentive filtration in Eli Lilly's downstream purification process, DoE methodology was employed to systematically characterize the unit operation. A full factorial DoE study consisting of 12 individual experiments was performed to evaluate the potential effects of three selected parameters (pressure, product concentration, and throughput volume) for two mAb molecules, using PPV as the model virus. To monitor potential breakthrough, three grab samples (∼1.0 mL each) were collected at the start, the middle, and the end of the filtration process. Buffer flushing permeate was combined with the mAb permeate pool. As shown in Table 4, while 11 out of 12 experiments provided complete or near-complete PPV clearance, one experiment had only achieved 2.6 log10 of PPV removal. The fact that no PPV was detected in any of the grab samples from Experiment 8 (Table V) strongly suggested that PPV “breakthrough” took place during buffer flushing. This was confirmed by repeating Experiment 8 and collecting the buffer flush separately. Significant PPV breakthrough was indeed observed during buffer flushing (not shown). It is worth noting that in Lilly's experience such severe parvovirus breakthrough has not been observed for this given brand of filter. View this table: [Table IV](http://journal.pda.org/content/68/1/38/T4) Table IV Initial Viral Clearance Results from Lilly mAb-1 DOE Study, Exhibiting Sporadic Breakthrough of PPV. Pressure, capacity, and mAb concentration were evaluated by DOE methodology in a full factorial manner. M = midpoint of the range; L = low end of the range; and H = high end of the range. View this table: [Table V](http://journal.pda.org/content/68/1/38/T5) Table V PPV Titer of Individual Grab Samples from the Experiment with Significant PPV Breakthrough Results from our mAb-2 DoE study can be divided into two groups based on the lot of laboratory-scale filters used. As shown in Table VI, individual experiments with lot A filter gave consistent and robust PPV clearance that was consistent with previous process knowledge. On the other hand, individual experiments with lot B filters showed varying PPV clearance ranging from 3.75 log10 to ≥6.67 log10. Statistical analysis indicated a significant difference between the two filter lots. View this table: [Table VI](http://journal.pda.org/content/68/1/38/T6) Table VI Viral Clearance Results from a Lilly mAb-2 DOE Study Comparing Two Filter Lots. Experiments with filter lot A removed PPV better than those performed with lot B. Pressure, capacity, and mAb concentration were evaluated by DOE methodology in a full factorial manner. M = midpoint of the range; L = low end of the range; and H = high end of the range. Lilly's initial plan was to compare the performance of the two lots of virus-retentive filters in a head-to-head fashion with identical conditions. However, this plan had to be abandoned because lot B filters were no longer available internally or from the supplier. Upon request, the supplier manufactured additional laboratory-scale devices using the same lot of filter membrane as the devices from lot B. The performance of these reconstructed lot B filters was compared to those of lot A under the exact same conditions. As shown in Table VII, robust and consistent LRVs were achieved using both lot A and reconstructed lot B filters. However, the results from experiments using the reconstructed lot B filters were strikingly different from those obtained with the original lot B filters. Contrast to the original lot B filers, the reconstructed lot B filters achieved as consistent and robust PPV clearance as lot A filters. Consistent with the better performance of the reconstructed lot B filters, the significant PPV breakthrough detected in the DoE experiments with the original lot B filters was not observed in the experiments with the reconstructed lot B filters (data not shown). View this table: [Table VII](http://journal.pda.org/content/68/1/38/T7) Table VII Comparison of Viral Clearance Results from Different Filter Lots. Repeat experiments using vendor-claimed analogs (longer being manufactured; see text for details) of filter lot B did not reproduce the cross-lot variability in PPV clearance previously observed (Table VI). All comparison experiments were carried out in duplicate to provide assurance of repeatability. M = midpoint of the range; L = low end of the range; and H = high end of the range. * Indicating reconstructed lot B filters; *#* derived from the DOE study (Table VI) using the original lot B filters. Despite of Lilly's significant efforts with complete support from our supplier, the investigation was not able to identify the root cause for the performance inconsistency and LRV fluctuation of the original lot B filters in the DoE study, primarily due to the unavailability of the lot. However, by using the reconstructed lot B filters, we have provided compelling circumstantial evidence suggesting that the poor and inconsistent performance of the original lot B filters was likely caused by the faulty laboratory-scale filters for that particular lot (original lot B). ## Summary In recent years, implementation of virus filters designed for parvovirus removal in the downstream process has become an industry standard, and regulatory expectation, for biologics manufactured in mammalian cell cultures. This is accompanied by the availability of a new generation of parvovirus filters such as Millipore's Viresolve® Pro and Asahi's Planova BioEX. These new filters have improved significantly in throughput as measured by liters per square meter per hour (L/m2/h) as well as consistency and robustness of viral clearance. The correlation between flux decay and parvovirus breakthrough observed with Millipore NFP filter (6) has not been seen with the new filters. Because virus-retentive filters remove viruses by a size-based exclusion mechanism (2) and because no retrovirus breakthrough has ever been observed with parvovirus filters (1), it is logical to streamline laboratory-scale virus clearance studies when parvovirus filters are used. Data presented by Amgen and Genentech at the conference further strengthen the case that a generic claim for larger model viruses, such as MuLV, seems reasonable if parvovirus-retentive filtration is employed in the purification train. As to exactly what number to use for claiming generic retrovirus clearance, individual firms have to provide sufficient data to justify the LRV claimed. One possibility is to claim the LRV achieved with parvovirus. Another possibility is to cite in-house historical MuLV LRVs. Also discussed, though data were not directly presented at the symposium, was the possibility of using LRVs achieved with surrogate viruses such as phage (e.g., see Reference 4). Parvovirus breakthrough remains a common phenomenon, even with the new parvovirus-retentive filters currently on the market. However, the breakthrough detected is relatively insignificant and not correlated with flux decay (filter fouling). The mechanism of the observed breakthrough has not been well studied and may vary from brand to brand. Genentech reported that MMV breakthrough pattern could not be linked to changes of flux, volume throughput, pH, conductivity, or mAb concentrations for Millipore Viresolve® Pro filters based on data from 16 mAbs. Janssen's study showed that multiple processing parameters could influence parvovirus breakthrough, with the total volume load being the most significant parameter. Eli Lilly and Company provided evidence suggesting that parvovirus detected following buffer flushing was linked to pressure release, and therefore could be easily mitigated by avoiding depressurization. Filter capacity can be influenced by product aggregates as well as virus preparations used in spike/recovery studies (8⇓–10). Data presented by Genentech indicated that CEX can potentially significantly improve the capacity of the throughput capacity of the Viresolve® Pro filter, thus reducing cost. Development work is essential to select the most effective and robust CEX conditions, including resin type and column volume, for achieving the maximum filter throughput capacity. Finally, the adoption of DoE methodology and quality by design (QbD) concepts requires significantly more laboratory experiments to better understand the viral filtration unit operation. As a result, the likelihood of encountering unexpected LRVs or “outlier results” also increases, as reported by Eli Lilly and Company. All outlier results must be thoroughly investigated to determine the root cause so that the product safety and quality system integrity are not compromised. As demonstrated by Eli Lilly and Company's case, it is essential to work with filter supplier closely in order to carry out the investigation to the best of one's capability. A thorough investigation would not only provide a scientific basis to justify the appropriate response and determine whether the results are truly outliers, but it also allows filter suppliers to improve their quality control strategies for their products. * © PDA, Inc. 2014 ## References 1. 1. Miesegaes G., Bailey M., Willkommen H., Chen Q., Roush D., Blümel J., Brorson K. Proceedings of the 2009 Viral Clearance Symposium. Dev. Biol. (Basel) 2010, 133, 3–101. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=21516942&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 2. 2. 1. Flickinger M. Miesegaes G., Lute S., Aranha H., Brorson K. Virus Retentive Filtration. In Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation, and Cell Technology, Flickinger M. , Ed.; Wiley Interscience: Hoboken, NJ, 2010. 3. 3. Lutz H., Chang W., Blandl T., Ramsey G., Parella J., Fisher J., Gefroh E. Qualification of a novel inline spiking method for virus filter validation. Biotechnol. Prog. 2011, 27 (1), 121–128. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=20878721&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 4. 4. Lute S., Bailey M., Combs J., Sukumar M., Brorson K. Phage passage after extended processing in small-virus-retentive filters. Biotechnol. Appl. Biochem. 2007, 47 (3), 141–151. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=17286554&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 5. 5. Miesegaes G., Lute S., Brorson K. Analysis of viral clearance unit operations from monoclonal antibody regulatory submissions. Biotechnol. Bioeng. 2010, 106 (2), 238–246. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=20073086&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 6. 6. Bolton G., Cabatingan M., Rubino M., Lute S., Brorson K., Bailey M. Normal-flow virus filtration: detection and assessment of the endpoint in bio-processing. Biotechnol. Appl. Biochem. 2005, 42 (2), 133–142. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=15901236&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 7. 7. Brown A., Bechtel C., Bill J., Liu H., Liu J., McDonald D., Pai S., Radhamohan A., Renslow R., Thayer B., Yohe S., Dowd C. Increasing parvovirus filter throughput of monoclonal antibodies using ion exchange membrane adsorptive pre-filtration. Biotechnol. Bioeng. 2010, 106 (4), 627–637. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=20229510&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom) 8. 8. Cabatingan M. Impact of virus stock quality on virus filter validation: a case study. BioProcess Int. 2005, 3 (Suppl 10), 39–43. 9. 9. Miesegaes G., Lute S., Dement-Brown J., Kaushal S., Strauss D., Chen D., Brorson K. A survey of quality attributes of virus spike preparations used in clearance studies. PDA J. Pharm. Sci. Technol. 2012, 66 (5), 420–433. [Abstract/FREE Full Text](http://journal.pda.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoicGRhanBzdCI7czo1OiJyZXNpZCI7czo4OiI2Ni81LzQyMCI7czo0OiJhdG9tIjtzOjIxOiIvcGRhanBzdC82OC8xLzM4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 10. 10. Slocum A., Burnham M., Genest P., Venkiteshwaran A., Chen D., Hughes J. Impact of virus preparation quality on parvovirus filter performance. Biotechnol. Bioeng. 2013, 110 (1), 229–239. [PubMed](http://journal.pda.org/lookup/external-ref?access_num=22766979&link_type=MED&atom=%2Fpdajpst%2F68%2F1%2F38.atom)