How to run a filterability test and virus challenge of Pegasus™ SV4 virus removal membrane filter discs

How to run a filterability trial

For reliable results, a filterability trial should be conducted using a calibrated balance (accuracy ≤ 0.1 g) to collect and measure the filtrate mass over time. A sufficiently accurate estimation of density for water and simple buffers is 1 g/L, but you should determine your product feed density if you believe it to be significantly different from water. Keep the temperature at the same level as the full-scale process where possible, to give the correct product viscosity.

If desired, you can use a buffer conditioning step to reduce the risk of aggregate formation at the water-product interface of the prewetted membrane. A buffer flush of ≥ 3 mL is sufficient to condition the membrane if water is drained from the upstream of the disc assembly before processing the buffer. If buffer flux needs to be determined, measure the buffer flow rate over 10 min. Follow the procedures and instructions for testing of 47 mm discs.

We strongly recommend that filterability trials are run to full processing time for the most accurate estimation of performance. If there are any time or product volume constraints, then we recommend running for the longest processing time possible, and forward predicting the results using the V_max model to estimate throughput.

Forward prediction of throughput using V_max analysis

V_max is the estimated value of the maximum capacity of a membrane. In simpler terms: the throughput that would be reached when the membrane is completely plugged, if there were no restrictions to time and feed quantity, and the membrane fouls in line with the standard pore constriction model.

We calculate V_max using a plot of time over throughput (At/V) against time (t) and is the inverse of the gradient, as shown by the linear form of the standard blocking law, Equation 1:

Constant pressure V_max linear equation (Eq 1)

A = filtration area (m²), t = time (h), V = volume (L)

V_max = estimated maximum throughput capacity (L/m²)

J₀ = initial flux (L/m²/h).

Determine the gradient from the linear portion of the graph only, as indicated in Figure 2. Calculate the initial flux from the same linear portion of the graph; evaluated as the y-intercept of that linear data. This initial flux is not always as accurate as direct flux measurement, but gives a suitable forward prediction in many cases. We can then rearrange Equation 1 to estimate the throughput achieved at a given time as described in Equation 2:

Forward prediction of throughput at constant pressure (Eq 2)

Fig 2. Data analysis for forward prediction.

Figure 2 shows typical data collected during a 47 mm disc filterability run, showing three phases of data collection. It should be noted that:

• Start-up effects cause inaccurate and variable data due to the low flux decay relative to measurement accuracy and start time accuracy. The slope can be under or over- predicted and the time for this phase varies between tests.
• The linear portion of the graph used to determine V_max and J₀ for forward prediction.
• End effects, only seen if the feed sample is filtered to completion and flow reduces to zero due to the feed running out.

Take caution when forward predicting, and adhere to the following conditions to minimize estimation errors:

• Forward predicted throughput no more than twice the measured throughput
• Forward predicted throughput < 90% of the calculated V_max
• Coefficient of determination (R² value) > 0.95

Accuracy in estimated throughput improves the closer the raw data collection time is to the estimation time. Although fouling is more complex than a simple constriction mechanism, this model is the most appropriate of all the traditional membrane fouling mechanisms for small forward predictions of limited data sets.

Using V_max to forward-predict throughput assumes that the gradient measured from the At/V vs t plot remains constant up to the estimation time and complete blockage of the membrane. Therefore, take caution when quoting the V_max value itself as a maximum capacity it can be a long and potentially very inaccurate extrapolation for high-capacity membranes like Pegasus™ SV4.

Generally, V_max values for our virus removal filters are very high and exceed the throughput that can be reached in typical processing times. Most of the time, membrane performance (batch area requirement) is either independent or weakly dependent on V_max and batch area requirement is governed by the processing time and membrane initial flux. For high V_max values, performance comparisons using V_max are not recommended, whereas for cases when the membrane is plugged, V_max can potentially be quoted with caution as highlighted above.

Typical filterability results

For filterability testing, we used proprietary monoclonal antibody solutions (mAbs) and processed a commercially available hIgG solution to represent plasma-derived product applications.

Figure 3A shows a typical flux profile for the Pegasus™ SV4 membrane challenged with mAb solution. Flux remains constant throughout the experiment for both operating pressures. Increasing the operating pressure for Pegasus™ SV4 membranes to 3.1 bar (45.0 psi, 0.31 MPa) yields higher flux without any negative impact on flux decay. We therefore recommend using an operating pressure of 3.1 bar (45.0 psi, 0.31 MPa) to achieve maximum flux performance. Results are typical for multiple mAb solutions tested, up to 25 g/L.

Figure 3(B) demonstrates the robust nature of the Pegasus™ SV4 membrane tested in plasma-derived protein solutions, again showing that a steady flux can be achieved at high operating pressures. The flux decay rates per unit mass are comparable to customer testing conducted with high purity intravenous immunoglobulin (IVIG) solutions.

Due to complexities including process impurity levels, buffer conditions and donor profiles, results from plasma sources vary significantly from product to product and therefore direct comparisons must be made using your own specific feed solutions.

Fig 3. Increasing performance at higher operating pressure—typical flux profiles for Pegasus™ SV4 membrane filterability tests at 2.1 bar (30 psi/0.21 MPa) or 3.1 bar (45 psi/0.31MPa) with (A) 10 g/L (1%) hIgG, (B) up to 25 g/L (2.5%) mAb.

Scale up

Table 3. Pegasus™ SV4 filter relative effective filtration areas for scaling calculations

Constant flow operation

We recommend that small-scale filterability tests and virus validation studies are carried out at constant pressure. Difficulties in maintaining a constant flow and accurately measuring an increasing test pressure typically generate more experimental noise than a constant pressure test and the results will be less reliable. Therefore, constant pressure testing is always preferable.

The flow decay for Pegasus™ SV4 membrane is slow and steady and there is no difference seen between constant flow and constant pressure testing. Permeability (L/m²/h per unit pressure) decays relative to the product throughput and is independent of the pressure applied. Therefore, using constant flow (or stepwise increases in the flow rate) to eliminate fouling due to high initial flow, is not necessary for Pegasus™ SV4 virus filtration. What may appear to be a lower flux decline will actually be an equivalent permeability decline.

If required, the key to successful constant flow operation is a pump that is capable of supplying the required flow rates accurately up to the maximum test pressure, without pump slippage or pulsing of flow/pressure.

Protein transmission

Protein transmission studies can be conducted alongside filterability testing by collecting samples of feed solution before and after filtration and subjecting them to protein assays.

Where the target protein is the major protein species, either a generic protein assay or a target protein-specific assay can be applied. Other assays may be used on the filtrate to assess conformation, biological or enzymatic activity, as appropriate.

Typical protein transmission is > 95%, although the exact level will depend on the product concentration, quality, stability and process throughput.

Calculating flux

When aliquots are taken the flux (L/m²/h) for that aliquot is simply the total unadjusted throughput (L/m²) divided by the time (h). If continuous data is collected using a balance, then the flux can be charted throughout the experiment. Many different options for calculating flux from continuous data exist with varying complexity. One solution we recommend is that the flux at a given data point should be calculated by the slope of the throughput and time data up to 5 min either side of the data point. Calculating the instantaneous flux between every time point collected, potentially leads to significant variation in the calculated flux due to the discrete nature of the filtrate drops, especially when collecting data over small time intervals.

Validation of virus filtration processes requires special attention, as both the filter manufacturer and end user serve vital roles. As a filter manufacturer, we have responsibility for ensuring that each filter will perform to the same specification whereas you, as an end user, must demonstrate that the selected filter satisfies the needs of your process.

Important factors to consider when designing product-specific viral filter retention validation studies

Good design of the product specific validation study is critical to ensure success of the study. Usually, the retention study follows a scaled down version of the full-scale process. Some of the important factors that need to be addressed in the study design are:

• The choice of spike viruses (models)
• Target reduction factor
• Spike virus titer
• Test feedstock comparable to process feedstock with respect to, for example, concentration, temperature, chemistry, etc.
• Equivalence of scale-down filter to process scale filter
• Same volume to filter area ratio for the test and process
• Inclusion of proper study
• Inclusion of measures aimed at removing virus aggregates potentially present in the spiked challenge solution (e.g., spike prefiltration).

How to run a virus spike challenge test

Similarly to filterability studies, we strongly recommend running virus spike challenges in constant pressure operation. Constant pressure testing is always preferable, even where a pump is to be used at process scale, with the pressure limits set by the expected full-scale pump performance. Constant pressure is also always required to run post-use installation checks to verify the correct operation of the challenge and release the samples for viral assay.

General protocol recommendations

Minimizing the amount of nonviral contaminants added to the product in spike studies is best practice to keep maximum equivalency between viral validation and production-scale feed streams. Therefore, excessive spiking, which also increases virus preparation-derived contaminants, is not optimal. We recommend that virus spikes should be designed on the basis of required input titer, rather than a particular spike percentage. Our recommended approach is to use a spike level that achieves a 10⁶ pfu/mL input titer (or another appropriate target titer based on your requirements).

Sometimes, high spike percentages (> 1%) can be necessary, for example, due to low stock titers. In these specific cases the robustness to fouling of the Pegasus™ SV4 filters allows the use of these spike percentages without additional flux decay impacting on the throughput that can be validated. Spike level should always be minimized to maintain equivalency and control any additional contaminants, but spikes of up to 5% can generate acceptable throughputs where required.

Filterability and protein transmission trial results guide the target throughput for virus filtration. For initial virus spike challenges (or bacteriophage studies) the filtrate should be collected in a minimum of two aliquots. We recommend a maximum aliquot volume of 111 mL (100 L/m²). Once retention data is established, aliquot volumes can be increased based on assessment of the data with respect to target retention.

Collect aliquots in individual graduated sterile containers, and record the time taken to collect each aliquot to calculate the flux.

If you require a product recovery buffer flush sample, we recommend priming the upstream volume with buffer and flushing through 3 mL of buffer, or another appropriate amount as determined by protein transmission studies.

Virus clearance is measured by the log titer reduction (LTR) or log reduction value (LRV), as shown in Equation 3, which is the base 10 logarithm of the ratio of the total virus input and total virus measured in all filtrate aliquots. For an individual aliquot or grab sample this simplifies to the ratio of feed concentration (C_feed) to filtrate concentration (C_filtrate).

Log titer reduction (LTR) or log reduction value (LRV) (Eq 3)

Typical virus spike challenge results

Table 4. Example virus spike challenge results for Pegasus™ SV4 filters

Table 4 shows details of the typical virus removal performance we expect from the Pegasus™ SV4 membrane in a variety of protein feed solutions. Figure 4 shows the typical performance of the Pegasus™ SV4 membrane in a 1 g/L BSA solution as per the PDA guidelines (1) for bacteriophage model parvovirus (PP7) and porcine parvovirus (PPV). Live virus testing was carried out at an independent virus validation test laboratory. Filtration was carried out at 3.1 bar (45.0 psi, 0.31 MPa) to achieve the maximum flux performance and demonstrate the viral clearance capability of the Pegasus™ SV4 membrane under these conditions.

Fig 4. Live virus (PPV, n = 4) and model virus (PP7, n = 9) retention performance of Pegasus™ SV4 virus filter membrane in 1 g/L BSA at 3.1 bar (45 psi, 0.31 MPa).

Table 5. Parameters potentially affecting microbial retention by filtration

Consider these parameters when running filterability optimization studies and designing viral clearance validation tests for virus filters. Specific recommendations for Pegasus™ SV4 virus filters are detailed in the following sections.

Operating differential pressure

As shown in Figure 3, increasing operating differential pressure increases the flux of Pegasus™ SV4 membrane and this is maintained across the course of the test. Testing with polyclonal human IgG solutions with different fouling levels at a variety of different pressures demonstrated no significant change in the level of fouling of the Pegasus™ SV4 filter membrane (as measured by V_max) from 2.1 bard (30 psid, 0.21 MPad) to 3.1 bard (45 psid, 0.31 MPad) operating differential pressure.

Typical bacteriophage clearance by the Pegasus™ SV4 filter membrane in a 1 g/L BSA solution (as per the PDA recommendations) is > 4 logarithm and consistent from 2.1 bard (30 psid, 0.21 MPad) to 3.1 bard (45 psid, 0.31 MPad) operating differential pressures. This demonstrates that with the Pegasus™ SV4 filter membrane, the optimum filterability performance seen at higher pressures does not impact retention performance.

Throughput

Under many process conditions, other virus filters characterized by high initial flow rates display rapid decay in flow and become less economical over time when compared to a fouling resistant constant flow filter such as the Pegasus™ SV4 filter. For this reason, the most economical approach for virus filtration is allowing for longer processing times using a fouling resistant filter and therefore achieve higher throughputs with a minimized cost per batch.

Run viral filtration validation testing to at least the expected maximum process throughput (volume to filter area ratio), which corresponds to the expected maximum process time. Conduct filterability studies to the maximum throughput as well, although initial scouting studies can use smaller volumes and forward predict performance. This is important due to factors related to process throughput and time such as product stability over the processing time and changes in performance at higher loading levels during extended processing.

Temperature and viscosity

Higher processing temperatures can reduce product viscosities and thereby increase filtration flux. Lower temperatures tend to increase viscosities and reduce filtration flux rates.

pH and ionic strength

Both ionic strength and pH can affect processing parameters like filtration flux rates and total throughput, as well as properties of the spiked viruses in the carrier fluid. Therefore, maintain careful control of pH during all virus filter testing.

Optimal conditions can vary for different products, so there are no specific recommendations for pH and ionic strength when using Pegasus™ SV4 filter membrane. Other buffer components, like stabilizers and excipients, etc., can also impact the overall filterability performance. In general, try to avoid extremes of pH (< 4, > 8) and high ionic strength (> 1 M) unless there is evidence of product stability at these conditions.

Product aggregation

Product aggregation can be caused by a variety of factors such as extremes of temperature, ionic strength and pH, and when pH ≈ pI (isolectric point). Both percentage aggregate content and aggregate size distribution can impact virus filter performance. Process steps, including virus inactivation and freeze-thawing, can also introduce aggregation. Some products can also aggregate over time due to intrinsic instability.

An important benefit of the Pegasus™ SV4 filter membrane is its high resistance to fouling for a range of aggregates, resulting in outstanding throughput capacity in both dilute and complex, concentrated biological fluids. These features ensure that our filters give maximum virus filtration economy and efficiency.

Precautions taken to maintain product stability during product development are typically sufficient to ensure that the Pegasus™ SV4 filter can process the aggregate burden in product feed streams with low flux decay. We recommend using prefiltration to improve the overall process performance in cases where significant flux decay is seen (see Prefiltration section below for details).

Protein concentration

Pegasus™ SV4 filters can achieve stable flows over a wide range of process conditions, including different protein concentrations, due to their robustness of flux and resistance to fouling.

For all protein solutions, an optimum concentration exists where a given mass can be processed with the minimum amount of filter area. It is a balance between three effects:

1. Reduced flow at higher concentrations due to increased viscosity.
2. A decrease in capacity at higher concentrations.
3. Reduction in process volume at higher concentrations.

Operating either at, or close to this optimum not only minimizes costs, but also means variations in batch concentration will have lower impact on performance. This is especially true for robust, fouling-resistant virus filters like the Pegasus™ SV4 filter, which typically has a relatively wide and flat optimum design space.

Optimum protein concentrations for Pegasus™ SV4 filters are typically > 30 g/L and performance is stable around these optima. Although typical variations in process concentrations are unlikely to impact on the Pegasus™ SV4 filter performance, it is beneficial to consider it in robustness studies. Generally, we don't recommend diluting or selecting a process position for virus filtration with a lower concentration. However, for certain extremely high fouling feeds or products with unusual viscosity trends, this can be necessary and is more likely to be beneficial where undiluted concentrations exceed 50 g/L. Where concentrations are low (< 20 g/L) Pegasus™ SV4 filter performance will still be strong, however where there are process positioning options it is likely that performance will improve by operating at higher concentrations.

Prefiltration

Prefiltration requirements vary from feed-to-feed based on the presence of various sizes of aggregates or contaminants. Pegasus™ SV4 filters can perform without any prefiltration beyond upstream sterilizing filtration (0.2 or 0.1 µm membrane) that may already be built into the purification process. Many mAb and plasma protein tests have demonstrated this, highlighting the robustness to flow decay of the Pegasus™ SV4 membrane. If an existing sterilizing grade pre-filter is not in place, we recommend the Fluorodyne™ II DJL filters (0.1 µm rated). If required, we have several other sterilizing-grade filter options.

For processes with high levels of particulate contaminants, protection of Pegasus™ SV4 filters by membrane prefiltration can be required. If flow decays are substantially faster than the typical performance seen in Figure 3, consider an additional prefilter. If flux decay is < 20% the process is unlikely to benefit from an additional prefilter.

If flux decay is > 50%, a prefilter is likely to make the process more economical.

Summary of recommended design space using Pegasus™ SV4 filters

Please note that the following design space specifications are a guideline for optimal performance of Pegasus™ SV4 filters. Apply prior knowledge and understanding of the particular feed to be tested, and we recommend conducting filterability studies before virus spiking to confirm performance and reproducibility.

Table 6. Design space recommendations when using Pegasus™ SV4 filters

1. A consensus rating method for small virus-retentive filters. I. Method development. PDA of Pharm. Sci. Tech. 2008;62:5:318–333

CY53816