Chromatography Optimization Strategy

ABSTRACT

The optimization of a large-scale chromatography operation involved a strategy combining bench and pilot scale packing stability studies, and evaluation of hardware limitations at bench and large scales.¹ The effects of resin amount on column stability, hydrodynamic properties of compressible media, process resolution, and targeted recovery were evaluated at the bench scale. The process was scaled up for hardware and process performance evaluation. A robust packing procedure was developed at pilot scale to optimize system suitability, reduce the time required for packing, and improve process performance. The packing optimization provided an increased bed stability, which offered the basis to extend the number of chromatography cycles per packed column. The lifetime extension reduced resin costs per lot. The faster packing method and reduced frequency for packing reduced labor requirements for this step at large scale.

This case study relates to protein purification through ion exchange chromatography with a start of collection based on ultraviolet (UV) absorbance and end of collection based on reaching a volumetric endpoint. This approach mitigates the potential impact of impurities that elute on the back side of the main peak, but led to higher yield losses for runs, in which peak resolution was broader because of nonoptimized packing conditions.

(LONZA, LTD., BASEL, SWITZERLAND)

Inconsistent performances of the chromatography step for the large-scale process led to high variability in process yields and purity. Therefore, basic process development techniques are required to assess and resolve the cause of variability and thereby improve and optimize the chromatography process.

RESIN QUANTIFICATION METHOD

The quantity of resin required to achieve targeted compression on a packed bed is a factor that significantly affects the chromatography elution profile. Achieving targeted compression is critical to avoid headspace formation and overcompression in stainless-steel columns, in which this cannot be visually observed. Previously, the resin concentration on slurry was determined by the gravity settling (GS) method, which is a simple way to determine resin concentration as a volumetric ratio. The method calls for pouring a determined amount of homogenous slurry into graduated cylinders and allowing the resin to settle by gravity over an extended period, which is dependant on the storage solution. There are, however, certain drawbacks associated with this approach such as a long settling time before reading. The calibration resolution of the graduated cylinder and wall support effects provided misleading readings of bed volume. The resin concentration varied because of intricacies of the technique related to visual interpretation of the cylinder calibration resolution and an unstable resin settling using the previous storage solution.

Figure 1. Experimental bed height and predicted height at different aspect ratios

FLOW CONSOLIDATION METHOD

An alternative to the GS method is the flow consolidation (FS) method, which takes into consideration the hydrodynamic properties of compressible media, showing a linear relationship between height and flow (Figure 1). The linearity between bed height and flow can be expressed in terms easily read on a chromatography column tube (Equation 1):

in which L is the bed height as a function of flow velocity (u) and L₀ represents the initial settled bed height at a flow velocity of 0 cm/h. B can be obtained as the slope constant and L₀, the curve intercept from a linear regression analysis (Figure 2). The linearity of bed height to flow appears at various aspect ratios (AR) defined as the ratio between initial bed height (L₀) and column diameter (D). By allowing the resin to consolidate by flow, this approach provides the advantage of reducing wall support effects that lead to variability in the resin quantification and allow for a homogeneous settling with results obtained much faster than the GS method.

Figure 2. Flow consolidation method

For a constant cross-sectional area (AC) of the column, the volumetric concentration of the resin can be determined from Equation 2 by associating the intercept L₀ from linear regression analysis with the initial amount of slurry poured in the column as follows:

in which LS is the slurry height poured into the column.

This method requires using a complex and expensive setup to quantify resin, which is not ideal for a manufacturing environment because of sensitivity.

CENTRIFUGATION METHOD

The quantification of solids by weight through centrifugation was also developed for this application.² After a resin sample is loaded on the top section, centrifugal force pushes the liquid through the filter into a receiver. Centrifugal speed and time are controlled to minimize variability. The amount of resin is then determined based on net mass retained in the centrifuge concentrator. The mass is then normalized against the total slurry weight poured into the concentrator defining the weight ratio M_R (Equation 3):

in which Mass_RETAINED and Mass_SLURRY are the net masses of the resin (top reservoir) after centrifugation and the total slurry (top and bottom reservoirs) respectively. Although there are concentrators with volumetric marks for volume determination, this approach provided less accuracy than using an analytical scale for measuring resin amount.

The disadvantage of this method is that a volumetric relationship must be established to adapt the model to a manufacturing environment, in which packing procedures call for volumetric amounts. In addition, the developed correlation is specific to the type of resin and its interactions with the storage media and its physical properties.

Therefore, a relationship between volumetric and mass ratio was established to optimize the advantages of methods described earlier. The relationship was generated through a plot of volumetric ratio (GS and FC) versus mass retained (Figure 3). The data showed a region of linear relationship between volumetric and mass retained ratios. Given that the GS method was historically unstable, the centrifuge method was related to the FC method. A plot of mass retained versus volumetric ratios demonstrate a linear relationship within the tested range.

Figure 3. Volumetric versus mass ratio

Using the centrifuge method for resin determination significantly reduced the packing procedure cycle time from greater than 70 hours down to 60 minutes. This method was important in the process characterization strategy because it paved the way to rapidly complete numerous packs at bench and large scales and thereby reduced the characterization timeframe and raw material consumption.

COMPRESSIBLE MEDIA PRESSURE-FLOW CHARACTERISTICS

A small-scale study was established to understand the hydrodynamic properties to forecast effects of resin amount before consuming expensive raw materials and efforts at large-scale. This provided the advantage to predict the column performance at given conditions of resin amount, fluid properties, and flow rates at various aspect ratios. Pressure-flow studies were executed at small scale. The model was tested with resin slurry at applicable process temperatures. The experimental plan was based on the work performed by Jonathan Stickel and Alexandros Fotopoulos.³ Pressure and bed height for various aspect ratios were collected on flow-rate increments until a maximum flow rate was achieved. Bench-scale experiments on a 100-mm diameter column reached maximum compressions of approximately 20.4% average at various aspect ratios.

The results from small-scale experiments provided a linear expression on the plot of the critical velocity (u_CRI) versus aspect ratio with slope and intercept values of 349.97 cm²/h and 3718.88 cm²/h respectively (Equation 4):

in which u_CRI is the critical flow velocity as a function of AR given in cm/h, L₀ is the gravity settled bed height given in cm, and D is the column inner diameter given in cm. This expression is capable to predict maximum flow rates as a function of column aspect ratio, which is dictated by the desired compression on a column (Figure 4).

Figure 4. Critical velocity versus aspect ratio

The relation of aspect ratio and flow rate was embedded into the Blake-Kozeny equation below (Equation 5) to predict pressure drop as a function of flow rate.

Finally, the correlation was developed from a curve fit analysis of the expression with the experimental data. The reliability of the model developed at a 100-mm scale was challenged by comparing the model with large-scale column experimental data (Figure 5) with promising results.

Figure 5. Large-scale experimental data versus model

Process parameters could be predicted at given conditions providing insights into the stability of the packed bed and settings to achieve optimum performance. Significant influence of the interstitial bed porosity and particle size on pressure drop (Figure 6) shows that a nonhomogeneous packed column will yield a larger pressure drop than expected from a homogeneous pack in which the interstitial porosity or void fraction is smaller. Therefore, a typical pressure drop may be an indicator of column bed deterioration. Increased interstitial porosity suggests an unstable pack because of partial motility of the beads in the column. Decreased bead size or particle diameter for unstable packs increases the pressure drop substantially at any given flow rate.

Figure 6. Pressure drop of packed beds of different interstitial bed porosities

The bench-scale results confirmed headspace formation at process flow rate with previous resin compression parameter thus marking a relationship of resin amount to the pack instability and performance variability seen on large scale. To optimize the chromatography operation, it was considered to pack at higher compression to increase resin stability at process flow rates. The use of more resin rather than lowering the bed height to compress further the resin or decreasing flow rates to minimize head space was avoided to prevent resolution loss and additional changes to regulatory filing.

The pressure-flow model predicted that incrementing compression to 1.18 (as recommended by manufacturer) will require a minimum flow velocity of 191.6 cm/h to achieve a bed height of 12 cm. For large-scale columns using flow-packing techniques, in which the resin is compressed by increasing the mobile phase velocity, this flow is not enough because of the resin expansion experienced once the flow is stopped to lower the top plate adaptor to the desired height of 12 cm, and therefore, a higher flow rate is required.⁴

It was determined from the pressure-flow simulation data that, for an optimum packing flow (85% of u_CRI), the required flow rate to flow-pack the resin to 1.18 compression is close to 226 cm/h. For the existing 1,000-mm column hardware, this represented a limitation because the expected pressure drop of the resin plus the pressure drop of the hardware exceeded the maximum allowable working pressure of the column. Axial compression was introduced to the packing procedure to compensate for system limitations of the flow-pack technique at higher compression.

LARGE-SCALE PACKING STUDIES

Packing studies were performed at large scale (1,000 mm) to understand the influence of column hardware, flow cell geometry effects, and variable resin amount on the pack stability. The studies consisted of a series of experiments that include dye and cleaning studies, and evaluation of various packing approaches.⁵

Dye studies using phenol red revealed equipment limitations of flow distribution under the top slurry nozzle through inspection of bed cross-sectional cuts at various angles, indicating that the flow velocity under the slurry nozzle is lower than the average velocity of the sample bands (~2.5 CV) or dye injection. A low compression pack revealed low stability underneath mobile phase inlet across bed diameter. By increasing the resin amount and thus the compression, the sample band distribution was more homogeneous throughout the bed under the top slurry nozzle.

A cleaning study evaluated the column cleaning using phenol red as a visual indicator for cleaning aptness. The packed column was buffer exchanged from packing buffer into a 0.1 g/L phenol red in packing buffer for 2 CV duration. The equilibration followed a five-hour static hold to allow diffusion toward stagnant zones at the edge of the packed bed. Once hold was completed, the bed was washed with packing buffer at a 3 CV duration followed by a 1.5 static hold and storage. These set of conditions were selected as a stress-case scenario when compared to current hold parameters. Inspection of cross-sectional cuts of the packed bed revealed no traces of phenol red. This study provided cleaning capability of the column as retention volume basis (it is not an evaluation of impurity or carryover clearance that could bind into the resin during the chromatography step).

COLUMN QUALIFICATION METHODOLOGY

The large-scale column pack qualification pulse injection testing used flow rates different from the chromatography elution step flow rate. In addition, the previous pulse injection and elution phase compositions had ionic interactions with the media. The ionic interactions led to qualification results that deviated from ideal asymmetry for the resulting peak. The pulse injection and elution flow rates were changed to match that of the process and a different testing media consisting of a 0.1 M sodium chloride solution/1.0 M sodium chloride pulse injection. This combination reduced the ionic interaction effects on the elution profile. By minimizing these effects HETP and peak asymmetry values measured by this method gave a greater sensitivity to the quality of flow distribution. This procedure was used to test all column packs in this study.

Figure 7. Low compression bed stability

High (1.18) and low (1.09) compression packs were exposed to stress cycles induced by high flow rates on a single pack. Figure 7 shows that the column packed at the lower 1.09 compression displayed a tendency toward increasing peak asymmetry values with increasing chromatography cycles. Figure 8 shows superior stability over the column lifetime at higher compression as indicated by the repeatability of the elution profile with repeated cycles.

Figure 8. High compression bed stability

By comparing the elution profiles for each compression, the higher compression pack suggests increased bed stability toward cycle stresses.

PACKING PROCEDURE EVALUATION

The pressure flow model suggested that flow pack techniques were not sufficient to reach the bed height of 12 cm at the desired compression because of high pressure drop and resulting hardware limitations. The packing procedure was simplified and optimized based on predicted hydrodynamic properties and use of readily available hardware. Several packing modifications were tested to keep the pack simple, reliable, and reproducible with the current hardware limitations and availability (i.e., flow meter, pressure gauge, pump, and column).

The final procedure includes resin quantification through centrifugation. Exact resin amount is transferred into a primed column chamber at a rate of 61 cm/h monitoring flow rate at the column top mobile phase port. The buffer from the column is recycled back into the resin slurry-mixing tank to assure that all the resin is transferred. After the transfer is completed, the resin is consolidated with a downward flow at 61 cm/h followed by lowering the top head-plate to 12 cm with the top port open and bottom closed. The top head-plate descent rate was controlled indirectly by targeting a predetermined top mobile port flow rate. This methodology allowed control over the top head-plate descent rate without expensive hardware upgrades to the large-scale column. Gradual increments are made to hydraulic pressure on the column as packed bed compression increases. This mechanical compression is then followed by a downward high flow conditioning to uniformly stabilize the effect of this compression. Unnecessary steps that added to the time required to pack the column and increased the variability of the packs were removed. Buffer consumption was reduced to a minimum through recycling and priming. During characterization, the procedure was repeated various times and packs were qualified with modified method. Table 1 shows the elution volume, height equivalent to a theoretical plate (HETP), and peak asymmetry values for the six experimental packs.

Table 1. Qualification of low and high compression packs

By definition, low HETP values are associated with increased separation or peak resolution. Reduction of HETP indicator from the new qualification method suggested a better performance on highly compressed packs. Lower HETP results were observed with the use of final packing equipment once implemented on the large-scale process.

Ultimately, the process-development data generated from pressure-flow and packing studies fed information for the design of a packing skid, defining flow rates, pump sizing, pipe ID, instrumentation, and procedural logic.

PILOT-SCALE AND LARGE-SCALE CHROMATOGRAPHY RUNS

The purification process was scaled to fit pilot 450 mm and 1,000-mm column keeping column-loading ratio consistent and equal linear velocities. The sole idea of performing an actual run at large scale as part of the process development strategy was to account for geometrical differences found on flow-cell hardware between each scale that could affect performance or scalability of results from bench scale. Table 2 shows the pertinent column information.

Table 2. Column information for chromatography runs

Each column pack was qualified with new pulse injection method and judged for acceptability based on the results of the packing studies. All the resin slurries were buffer exchanged into packing buffer. The resin quantification method relied on a column-based method to determine the percentage of resin. The pilot-plant chromatograms (Figure 9) reflected a good resolution but also a shift in retention volume.⁶

Figure 9. Pilot plant run at high compression versus commercial run at low compression

Large-scale results matched the experimental data from laboratory and pilot runs for this project with improved resolution at the beginning of the elution profile (Figure 10).

Figure 10. Overlay of previous 1.09 compression and current 1.18 compression elution profiles

The bench-scale approach to evaluate end-of-collection was applied to pilot and large scales to evaluate the impact of the shift in retention. A new elution collection set up was developed to handle the retention shift and generate data necessary for a filing change. For each run with the proposed elution collection set up, the elute pool was well within historical results and met all acceptance criteria. In fact, product purity improved.

Reproducibility between purity and recovery results between bench and large scales demonstrated that the process was scalable and reproducible at both scales. This paved the way to qualify a bench-scale method to assist manufacturing events.

The overall chromatography process showed that it is receptive to the amount of resin or compression factor. A higher compression sharpened the prepeak elution profiles. Specifying the compression factor as a process characterization variable is critical in maintaining a scaled-down model.

LARGE-SCALE BLANK ELUTION RUN

The large-scale run was followed with a blank elution run to evaluate cleaning capability at a higher compression. The pool was analyzed with molecule concentration of below maximum limits. The results indicated that the regeneration is sufficient to mitigate concerns of product carry-over. Additional characterization efforts based on packing stability promoted the increase in pack re-uses from 10 to 30 cycles. Significant improvement was seen on yield recovery and the performance was more stabilized. The 30-cycle lifetime extension has been achieved at scale.

CONCLUSION

A scalable and consistent performance was demonstrated at large scale by an increase in both chromatographic step recovery and product purity. Asymmetry and HETP results were improved in both average results and consistency. The stability of the new pack procedure allowed for a lifetime extension for the chromatography step. The work culminated in the successful completion of three validation lots in the facility. The filings were submitted and approved and the facility now runs with this new large process.

Overall, the project had an impact of drug substance yield improvement by 7%, reduced recovery variability by 50%, reduced impurity levels by 12%, and zero nonconformance events post implementation. The resin determination by centrifuge and changes to packing procedure dramatically reduced cycle time.

AKNOWLEDGEMENTS

The authors are grateful to the cross-functional team of engineers and scientists for their dedication and support throughout all stages of the project.

Javier O. Tapia is a process engineer, Carlos Escobar is a principal engineer, and James Weidner is a director, all at bulk process development, Amgen Manufacturing, Ltd., Juncos, PR, 787.916.6871, javiert@amgen.com

REFERENCES

1. Escobar C, Keener N. Chromatography unit operation optimization plan, process development study. PD-008-06, Amgen; 2006.

2. Rivera O. Determination resin percent for resin slurry by centrifuge method. Technical report 070072TR. Amgen; 2007.

3. Stickel J, Fotopoulos A. Pressure-flow relationships for packed beds of compressible chromatography media at laboratory and production scale. Biotechnol Prog. 2001;17:744-751.

4. Tapia J. Pressure flow correlation for resin. Technical report 070002TR. Amgen; 2007.

5. Keener N. Column packing study. Technical report 060409TR. Amgen; 2007.

6. Britton D. Pilot plant campaign summary. Technical report 060408TR. Amgen; 2007.