Studying resin variability using a Process Characterization Kit

Process robustness and success requires a thorough understanding of critical process parameters (CPP), critical material attributes (CMA), and sometimes interactions between these. Resin attributes can impact critical quality attributes (CQAs) as well as process performance depending on the process design and the separation at hand.

Ligand density is an important resin attribute that may display interaction with process parameters, such as conductivity and pH. In this study, we illustrate a process development workflow to assess the potential impact of ligand density variability on product quality and process performance using a Process Characterization Kit. The results show that the kit is a useful tool for gaining deeper process understanding.

Introduction

Resin variability and its potential impact on the process outcome can be studied at various stages in process development. Most commonly, it is addressed in the process characterization study after a suggested process design has been decided upon.

Historically, assessment of the impact of resin variability has been difficult. One approach is to evaluate three random lots of the resin in question. The limitations of such an approach is the lack of detailed control over the resin properties for each lot. Therefore, such a study will, in most cases, be of limited use and not generate much understanding of how resin attributes interplay with process parameters.

However, combining factors in a study using a design of experiment (DoE) approach facilitates estimation of interaction effects and relative effects of variability. In order to use a DoE approach, both the resin and process parameters should be varied in a controlled fashion in the same study. This is now possible using Process Characterization Kits.

A tool for gaining deeper process understanding

These kits consist of three bottles of the same resin with three different ligand densities (high, average, and low). The kit enables thorough characterization of the potential impact of this important resin parameter on product quality and process performance. If an impact is identified, a control strategy can be developed to mitigate this, resulting in a robust manufacturing process.

A proposed way of incorporating the Process Characterization Kit during process development is discussed in this article.

Incorporating the kit during process development

Figure 1 describes a general process development workflow. The steps highlighted in green will be the focus of this article. However, a brief description of the first steps will also be given to set the context.

Fig 1. Illustration of a general process development workflow where a Process Characterization Kit is used in the process characterization step to investigate process robustness.

Process description: deciding the basics of the process. The goal of the process is defined and suggested ways to reach the goal are described. Considerations covered are, for example, chromatographic mode, what options for the separation are available in terms of buffer composition, loading, flow rate and so on.

Identifying potential factors: listing factors that could affect the outcome of the process. Typically, a Fishbone diagram is used.

Preliminary risk assessment: selecting the factors, most likely to affect the outcome, for further study. Some examples are where the nature or magnitude of the effect is unknown or if potential interplay between parameters is suspected.

Factor screening: screening of the identified factors. This work defines the process in more detail and, if possible, narrows down the number of factors that should be studied. If several separation options exist, these should be explored, and the most promising way forward decided.

Process design: identifying factor settings and acceptable factor intervals, which would have the potential to reach the separation target. When then process details have been decided, the detailed effect of important process parameters must be understood. This is typically done by using a DoE approach.

Risk assessment: revisiting the risk assessment to decide which factors to include in the process characterization study.

Process characterization: studying the robustness of the process.

Study of a simplified mAb separation

In this article, a simplified separation will be discussed from setting the process design to the process characterization study using a Process Characterization Kit. The separation involves the removal of Fab from a monoclonal antibody (mAb) sample using Capto S ImpAct cation exchange chromatography (CIEX) resin.

The main purpose of this article is to illustrate how a Process Characterization Kit can be incorporated in the process development workflow. It is therefore important to realize that in a real situation the process development process will be more complex and more elaborate studies will be needed.

Setting the process design

In this study, the process involved separation of a mAb from an unwanted Fab fragment. It was concluded early-on that a cation-exchange step using Capto S ImpAct could potentially achieve the desired goal; to achieve complete removal of Fab while ensuring high yields of the mAb.

Factor screening

Two potential options for carrying out the cation exchange step were identified: to use salt or pH elution. Two experiments were therefore run as a prestudy. One was run with a salt gradient (not shown), the other with a pH gradient (Fig 2). The results showed that pH elution was the preferred alternative, as it seemed possible to wash out Fab with minimal mAb loss using a pH change step.

Fig 2. Gradient elution with a pH gradient ranging from 5.5 to 7.5. Load was performed at pH 5.5.

Preliminary risk assessment conclusions

The preliminary risk assessment concluded that if pH change was chosen as elution strategy, two factors, sample load and wash pH, were probably important. These two factors should be studied in more detail in order to define the process. The ionic capacity of the Capto S ImpAct resin was also identified as a factor that could potentially have an effect on the process outcome. However, it was concluded that load and pH were the most important factors to study at this stage.

Process design study using a DoE approach

The study was performed using a full factorial central composite face design to find a potentially robust area with high yield of mAb and good clearance of Fab (Fig 3).

Fig 3. Illustration of the process design study used to investigate the effect of the identified main factors.

Based on the prestudy, the Fab/mAb sample was loaded at pH 5.2. A frontal analysis was performed to find the breakthrough capacity of the resin for the mAb at this pH. The binding capacity at 10% breakthrough (Q_B10) was determined to be 76 g mAb/L resin at the selected conditions.

The factor settings for load in the subsequent DoE study were 60% to 80% of Q_B10, that is 46 to 61 g mAb/L resin. The wash pH was between 5.2 and 6.0 based on the prestudy results. To verify that these factor settings were suitable, worst-case conditions for Fab clearance and mAb yield were first evaluated.

At high load and high pH, the mAb yield was low (20%) while the Fab clearance was complete. The center point conditions were also analyzed (load 53.5 g/L and pH 5.6). These conditions gave high yield and good Fab clearance.

No loss of Fab was observed during load at either of the settings used. Based on these observations, the DoE setting was narrowed to pH 5.4 to 5.8.

Outcome of the process design study

The study was performed according to Table 1, which also shows the outcome of the experiments. The center point was replicated to estimate experimental reproducibility.

Table 1. Factor settings and outcome of the process design study (center point replicates are highlighted in bold text).

Fab removal could not be modeled because total clearance was achieved at all conditions studied. Figure 4 shows that there are three observations with low yield (experiments 3, 4, and 8). A wash with high pH was used in all three experiments. This indicates that a lower pH for the wash is required to obtain high mAb yields.

Fig 4. Replicate plot on the outcome from the process design study.

A good overall model for yield was obtained with high predictability as indicated by a Q² of 0.885 (Fig 5). The model showed that pH had the largest effect on mAb yield. It also showed that the amount of mAb loaded also had an effect, but to a lesser extent than pH. The response contour plot in Figure 6 supports that conclusion.

Fig 5. Effect on mAb yield together with summary of fit for the model. The coefficient plot indicates that pH is the most important factor for yield. N = number of experiments, R² = goodness of fit, RSD = residual standard deviation, Q² = goodness of prediction

Fig 6. Response contour plot showing highest mAb yield in the lowest pH and load ranges.

It was estimated that a yield above 95% can be achieved if the pH in the intermediate wash is kept below pH 5.6 and the load does not exceed 54 g/L resin. However, the response contour plot will only give an indication of the outcome. The model error is, for example, not accounted for in the plot.

Setting the potentially robust area

Once a model for the purification step was obtained, a Monte Carlo simulation was performed to find more detailed information as a basis for setting the process design. All significant process parameters were varied with different settings ten thousand times. The outcome could be studied in silico by using the obtained model. In this way, potentially robust conditions could be set for the process, which would be verified in the subsequent process characterization study.

When a Monte Carlo simulation with an aim of 95% yield and with failure risk of 1% was performed, it was concluded that the necessary pH interval (± 0.04 pH unit) was too narrow for a robust process (Fig 7). Therefore, a lower yield of 90% had to be accepted as a worst-case scenario (Fig 8).

Fig 7. Results from the Monte Carlo simulation. The area within the dashed lines indicates a robust area where the yield of the process is predicted to be above 95% with a 1% risk of failure

Fig 8. Result from the Monte Carlo simulation. The area within the dashed lines indicates a robust area where the yield of the process is predicted to be above 90% with a 1% risk of failure.

The potentially robust area identified to secure at least 90% yield was a pH between 5.4 and 5.6 for the wash, and a load between 46 and 53 g mAb/L resin. These settings were used for the subsequent process characterization study.

The acceptable load range is larger when the minimum acceptable yield limit is reduced to 90%. Loads lower than 46 g mAb/L resin might also be acceptable, but this was not tested in this study.

Process characterization

Risk assessment

The revisited risk assessment suggested that variability in the ligand density (in this case ionic capacity) of the CIEX resin should be considered in addition to the critical process parameters, as discussed below. The mAb yield is related to the binding strength of the protein. Variability in resin ionic capacity may affect the strength with which proteins interact with the resin and could therefore influence the yield.

The mAb yield will also be affected by pH, as shown in the process design study, because the protein charge and thus binding strength is affected by changes in pH. As the suggested process design involved a relatively narrow acceptable pH interval to ensure 90% yield, variability in ionic capacity in combination with load and pH was considered risk factors for lower yields in this separation.

In order to examine the robustness of the proposed process design to variability in ionic capacity, the process characterization study included the Process Characterization Kit Capto S ImpAct along with the critical process parameters load and pH. By combining all potential risk factors in the same DoE study, the effects of each individual factor as well as interactions between these factors could be examined and potentially mitigated.

Set-up of the process characterization study using a DoE approach

A full factorial design was performed with factors pH, sample load, and ionic capacity (Fig 9). The center point was replicated to estimate experimental reproducibility. Responses were mAb yield and Fab clearance.

Fig 9. Illustration of the Process Characterization study design.

Outcome of the process characterization study

Table 2 shows the outcome of the process characterization study.

Table 2. Factor settings and outcome of the process characterization study (center point replicates are highlighted in bold text).

Fab clearance was complete in all performed experiments. Figure 10 indicates some variation in mAb yield however. The plot also shows that experiments 4 and 5 deviated more than the others. The conditions used for experiments 4 and 5 were the extremes, that is high load and high pH and low load and low pH, respectively. These two experiments were further examined and are discussed in the Investigating the extreme conditions section.

Fig 10. Replicate plot from the process characterization study indicating some variation in yield.

Figure 11A shows the coefficient plot with all replicates included. The results indicate that no factors are statistically significant and that the model shows no predictability. The unit operation can thus be considered robust even with variation in resin ionic capacity as one of the factors.

There is, however, a noticeable deviation for replicate 11 in Figure 10 and this deviation could hide some unexamined effect. To exclude that possibility, this replicate was removed from the model (Fig 11B). The result was the same as in the first evaluation, which indicates that there is no predictability for the model and that the process is robust.

Fig 11. Coefficient plots from the robustness study (A) including all three replicates; (B) excluding the deviating replicate (experiment No. 11). N = number of experiments, R² = goodness of fit, RSD = residual standard deviation, Q² = goodness of prediction.

Investigating the extreme conditions to confirm robustness

Table 3 shows that the mAb yield observed for experiment number 4 and 5 in the robustness study was similar to the yield predicted by the process design model with the same settings. This further supports the outcome of the Process Characterization study—that variability in the resin ionic capacity did not influence the outcome of the purification because the extreme values could be well predicted with a model without ionic capacity included.

Table 3. Observed yield in the process characterization study vs predicted mAb yield obtained by the process design study model where ionic capacity was not included

Discussion

Process design

The process design study was performed on a single Capto S ImpAct resin lot and showed complete Fab removal under all tested conditions. The yield was, however, affected by pH and to some degree also mAb load.

The potentially robust area ensuring more than 95% mAb yield relied on a pH interval that was too narrow according to the Monte Carlo simulation. The conclusion was that this effect could not be mitigated by changing the mAb load. Therefore, mAb yield had to be sacrificed in some parts of the robust area and a minimum yield of 90% was accepted.

Including resin ligand density variation in the process characterization study

In this purification example, the step yield was sensitive to changes in pH with a narrow operating window of ± 0.1 pH units. In ion exchange chromatography, the ligand density (ionic capacity) of the resin can also affect the yield.

Consequently, the Process Characterization Kit was included in the process characterization study of the unit operation together with the critical process parameters pH and load. In this example, the outcome of the process characterization study showed no impact of any of the tested factors. The unit operation could therefore be considered robust.

If resin variability would have had an impact on the process outcome

If a significant effect on product quality or process performance caused by variability in ligand density is observed, a mitigation strategy should be explored. In such cases, thorough understanding of the interplay between process parameters and ligand density is highly valuable because it potentially allows shifting the process parameter settings to minimize the effect.

For example, if high ligand density gives lower yield, but the effect is more pronounced at lower pH, the allowed pH range can be shifted upwards provided that the impact on CQAs is acceptable. The process understanding is critical to make an informed decision, and this can be achieved by using the Process Characterization Kit.

If shifting process parameter target settings does not give the desired outcome, an adaptive process can be considered. This entails changing the process parameter settings depending on the resin lot properties. This is however a more complicated route and a change in process parameter settings is a preferred mitigating action.

Conclusions

In this study, a process development workflow including a Process Characterization Kit was suggested to evaluate the potential resin variability impact on a purification process.

A relevant variability in ligand density can be studied in a DoE setting by using the Process Characterization kit in the process characterization study. In the study discussed in this article, the ligand density variation showed no significant impact on process robustness when tested together with process parameters in the proposed robust area. Hence, this process can be considered robust also with respect to variability in resin ligand density.

If variability in ligand density had caused a significant effect, the deeper knowledge gained by studying this parameter in a DoE setting would have resulted in a better chance to find a suitable mitigation strategy, for example by shifting process parameter settings.

The Process Characterization Kit can thus be used to increase process understanding and minimize unforeseen deviations during manufacturing.

See what resins are available as Process Characterization Kits

Related products

• Process Characterization Kit

Related content

• Raw material variability: the need for deeper process understanding
• Risk assessment: studying chromatography resin variability
• 5 control strategy options to prevent resin variability impact

TR29455168