Maintaining the supply chain of single-source raw materials is of utmost importance for a biopharmaceutical company’s manufacturing operations. As often happens, a supplier will notify its customer of process changes that might affect the quality or properties of supplied materials. Occasionally, a supplier might notify the customer of substitutions in its own supply chain or other changes in the source of its own raw materials. Customers must conduct appropriate testing using the “new” raw material(s) to ensure acceptable comparability, even if the change is only in sourcing and not the process itself.
Here we illustrate one such example in which a supply chain notification (SCN) described a source change in one of our supplier’s cellulose-based ion-exchange chromatography media raw materials (cotton linter). Upon receipt of the SCN, chromatography tests were executed in the laboratory using a representative small-scale model of the affected manufacturing process using new and prechange media to test their comparability. Later, the new resin was used in packing studies with radial columns at large scale, and packing differences were observed when the results were compared with historical data. Since then, and with the supplier’s support, significant efforts have been made to further understand the implications of this change in chromatography media and prevent negative impact on performance of the biopharmaceutical purification process.
PRODUCT FOCUS: BIOPHARMACEUTICALS
PROCESS FOCUS: DOWNSTREAM PROCESSING
WHO SHOULD READ: MANUFACTURING AND PROCESS DEVELOPMENT
KEYWORDS: CHROMATOGRAPHY, RESIN, VARIABILITY, COLUMN PACKING
Materials and Methods
Comparability Studies and Pressure Flow Profiles at Small Scale: To evaluate the impact of this change, small-scale (34-mL column bed volume) studies compared the chromatographic performance of new resin lots provided by the supplier against the prechange resin. During these experiments, we compared chromatographic profiles and column performance parameters of the media with two different linter sources. We also generated pressure-flow profiles at small scale to evaluate potential effects on the packing process of the ion-exchange chromatography column resulting from the change (1). For the small-scale experiments, we used XK axial-flow chromatography columns from GE Healthcare (www.gelifesciences.com).
For our comparability studies, we tested three different lots of resin with the new linter and compared the results against a control resin manufactured with the prechange linter. We ran each packed column with a corresponding control on a single ÄKTAexplorer small-scale chromatography system from GE Healthcare to reduce system-to-system variability.
We obtained pressure-flow profiles from columns using two lots of the new and one lot of the prechange linter resin. To monitor pressure differences, we tested six column packs from the new-linter resin and six from the prechange resin using the small-scale chromatography system. To obtain the specific pressure profile for each resin type, we recorded pressure readings at flow rates of 23, 57, 90, 124, 158, 192, and 226 cm/h for each packed column using equilibration buffer. Those flow rates covered the range of roughly zero to the maximum pressure rating for the columns. We subtracted the system pressure contributions by performing the same pressure-flow rate readings with a column filled only with equilibration buffer.
Packing Studies with Radial Columns at Large Scale: To determine whether changes in resin and scale could affect chromatographic pressure and flow dynamics, we performed the same ion-exchange chromatography step at large scale (20-L column bed volume) using a radial-flow column. In such columns, the mobile phase flows in a radial path from the outer column area into the center (Figure 1). Because our small-scale pressure-flow studies used axial-flow columns, it was important to test the packing of the new-linter resin at a larger scale to ensure that hardware differences did not produce dissimilar results from the small-scale studies (2). During execution of the large-scale test run, we packed the column with three new-linter resin lots. After the resin was properly conditioned, the column was packed at a set flow rate to a final target pressure (Figure 2). At this scale, the amount of resin packed is ≥11.3 kg.
Once the column was packed, the resin bed was conditioned by passing two upward and two downward flushes through the column before the pack was tested. The flushes allowed resin beads to disperse or set evenly across the column and create a uniform bed. We tested the column pack using a sodium chloride (NaCl) tracer solution spike in an equilibration buffer background solution and evaluated the peak asymmetry and packing ratio against the expected range. The column packing ratio is defined as the elution peak volume divided by the column volume.
Small-Scale Purification Performance: Using the new-linter resin, column performance was comparable to that of the resin made with linter from the prechange supplier. Figure 3 overlays one of the three sets of ion-exchange chromatographic profiles resulting from the purification performance study. Small differences in the heights and shapes of the elution profile peaks are typical of this unit operation, mainly due to differences in protein concentration of the loading material (3).
We sampled product pools from six chromatography runs (three test runs and three controls) and sent them for analysis, then compared the results for multiple column performance parameters such as recovery and purity against an expected range. All results were within the preestablished ranges and comparable with the control runs. A paired t test was performed to establish whether small differences between the new and prechange linter resins were statistically significant (p-value <0.05). Pair matching between each test run and its corresponding parallel control material showed no statistically significant difference between results from the new-linter and control resins (Table 1).
Table 1: Paired t test for column performance parameters of new and previous linter resins; even though the results for all runs were within preestablished acceptance criteria and comparable to the controls, the analysis was performed to determine whether small differences between the resins were statistically significant (p-value <0.05).
Small-Scale Pressure-Flow Profiles: We evaluated the pressure-flow profiles obtained at small scale to look for effects on the column packing process from the new resin linter. Results from these studies showed no significant differences among pressure-flow profiles generated at small scale (Figure 4). A qualitative observation was made that the overall pressure profiles from all new-resin runs were consistently below the pressure profile of the prechange runs. Although we saw this with all runs, the difference was still within our chromatographic system’s detection error of p-value of 0.76) among columns packed with either resin type (Figure 5).
Because we executed our small-scale pressure-flow studies using axial flow columns, the small pressure differences observed cannot be directly extrapolated to a radial-flow column in which fluid dynamics are different (4). If the radial column is packed to a final target pressure, any potential variation in the pressure-flow profile of the packed resin could alter the final resin amount loaded. So we used the small-scale pressure-flow data only as an initial assessment of the new-linter resin’s pressure-flow behavior. Our next step was to evaluate its potential impact on the column packing process at large scale.
Radial-Flow Column Packing Studies at Large Scale: Three large-scale column packs used a radial-flow column to assess whether the change in resin linter had any effect on the column packing process. We used three different new-linter resin lots provided by the supplier to conduct these packing exercises. And we evaluated the amount of resin packed and both the column packing ratio and asymmetry upon completion of those studies, then compared the expected range and historical values±1 standard deviation (SD). Figures 6, 7, and 8 show the results.
Results for column packs 1 and 2 met acceptance criteria (Figures 68). When compared against the historical average±1 SD, however, the amount of resin packed in the first two packs was found to be beyond –1 SD of the historical average (Figure 6). Given these results, we incorporated changes into the third packing exercise to determine whether additional resin could be packed to values closer to those seen historically at the manufacturing facility. Specifically, we increased the packing target pressure 1.8-fold to allow for additional resin. During the third execution, only a small amount of additional resin could be packed into the column, even when using a higher pressure target. The resin amount we calculated for the third pack was higher than for the first two packs, but a good portion of the resin remained within the packing inlet manifold. So the third column pack failed the peak asymmetry acceptance criteria (Figure 8).
Results from these packing exercises prompted us to question the possibility that the particle size distribution of the new-linter resin might be different from that of the prechange resin. We took samples from both resin lots and analyzed their particle size distribution using an LS 13 320 laser-diffraction particle size analyzer from Beckman Coulter (www.beckmancoulter.com). Results showed that the average particle size range for the new-linter resin was 126.9–141.7µm compared with a range of 127.2–130.1µm for the prechange resin (5). Based on the particle size analyzer’s measuring tolerance, we observed no appreciable differences.
A subsequent communication from the media supplier reported the results of its internal validation exercise using the resin with the new linter source. Data indicated that the resin made with the new linter source met all supplier specifications and acceptance criteria.
Upon receipt of that notification, we decided to execute a series of small-scale experiments to ensure that changes in the raw material did not affect our chromatography process. We tested the new resin for purification performance alongside the prechange version as a control and found no statistically significant difference between the two. Pressure-flow studies reveal no statistically significant differences between the pressure-flow profiles of the two resins.
Because we used an axial-flow column for our small-scale experiments, and hardware differences brought about by scale changes were possible, we performed large-scale packing studies using a radial column. Results from our large-scale packing studies revealed differences in packing behavior when a column is packed at a set flow rate, to a final target pressure, and the amount of resin packed must be equal to or greater than a target value. For the new-linter resin, values for packed resin were beyond –1 SD of the historical average even after increasing the specified packing pressure to allow for more resin in the column. This suggested the possibility that the average particle size of the resin with the new linter could be greater than that of the prechange version, but subsequent particle size analysis showed both resins to have very similar particle size ranges. Another possibility is that the new-linter resin is more rigid, making it less compressible.
Since we received the SCN sent by our chromatography media supplier, a significant amount of time and resources have been allocated to identify differences caused by that single change in the supplier’s resin linter source. A comprehensive investigation has been started with support of the media supplier to understand the full extent of the impact to our chromatography process. That investigation is still ongoing, as is additional experimentation on our part. This report should reinforce the need to ensure that even apparently small changes in raw materials by suppliers can have noticeable effects on processes.