ON-LINE PROCESS CONTROL: Automating the Control of Process-Scale Purification Columns Using On-Line Liquid Chromatography

ABSTRACT

This article discusses how on-line high-performance liquid chromatography (HPLC) can measure product purity in the column eluent stream in near–real time. These data can then enable the automation and control of a purification column operation, thus reducing product variability, shortening process cycle time, and increasing yield. An example application demonstrates how on-line HPLC is used as a process analytical technology to ensure the process can accommodate variability in the separation while ensuring the product meets its critical quality attributes.

Pharmaceutical products, particularly those derived from biotechnology based processes, frequently use process chromatography in downstream operations to isolate and purify the product of interest. The high selectivity provided by chromatography is especially important when dealing with the complex streams encountered in biotech processes in which the product of interest often is a minor process component at the initiation of downstream isolation and purification operations. Knowing when the product of interest at the targeted purity elutes from a process-scale purification column can be difficult to determine because these separations rarely provide baseline separation of the product of interest from other components in the process stream.

Determining when product elution occurs generally involves collecting the column eluent stream into small-volume time slices or "fractions" that are analyzed off-line in the laboratory by means of higher resolution and frequently orthogonal chromatographic techniques such as reversed phase chromatography. This process of eluent fractionation and off-line chromatographic analysis is performed during process development to develop sufficient process understanding so that developers can predict when the product will elute from the process scale chromatography column. The eluent fractionation and off-line analysis process often continues even after the product moves into manufacturing.

Fractionating the column eluent has several disadvantages, including:

• Labor intensive

• Opportunities for errors to occur

• Increased risk for product contamination and exposure of operators and facilities to process components

• Increased process cycle time caused by time needed to obtain off-line assays

• Suitable storage required for "work in process" (the column fractions)

• Degradation of product caused by these delays

• Reduced production capacity resulting from long cycle times

• Difficult to automate processing because of manual handling of fractions.

Because of the many negative issues associated with column fractionation, companies strive to eliminate this operation. Options to eliminate fractionation include collecting product based on column eluent volume; collecting product based on column eluent volume and optical density; and measuring product purity by on-line HPLC.

Collecting product based on column eluent volume.

In this approach, the eluent volume where the product of sufficient purity is expected to elute is determined from historical data; the start-collection and stop-collection setpoints are set based on these empirical data. An in-line flow meter measures the eluent volume to control the collection of the product of interest as it elutes from the process chromatography column.

The difficulty with this approach is that it is totally dependent on the reproducibility of the column elution profile which is affected by many variables. These variables include column loading, the purity of the starting material, the affinity of various components in the process stream for the stationary phase, and the column's operating parameters such as the generation of the gradient used to elute the column. Because of variability in these inputs, it can be challenging to maintain reproducible purity levels of the recovered product when using this approach, even with feedback control of the column's critical operating parameters (including gradient generation). This variability generally leads to conservative collection setpoints that reduce product yield.

Collecting product based on column eluent volume and optical density.

In this approach, an in-line ultraviolet (UV) sensor set to monitor absorbance at 280 nm measures the amount of product present in the eluent stream by relating the absorbance at 280 nm (A280) to product concentration. Because this measurement is not product-specific (i.e., any peptide in the stream will generate an A280 absorbance signal) the output of the optical density (OD) sensor output is examined in combination with the in-line flow meter to improve resolution. That is, the product is only collected when the OD at 280 nm exceeds the OD setpoint and falls within the eluent volume setpoint window.

This approach imparts greater selectivity than using the eluent volume approach alone, but is still dependent on the reproducibility of the column's elution profile, because the measurement is not specific for the product of interest. Although the variability is lower than with the approach of collecting product based on column eluent volume, the lack of a high level of selectivity still leads to conservative collection setpoints that reduce product yields.

Measuring product purity by on-line HPLC.

The critical quality attribute of the product is its purity. Meeting a specified purity value is the criterion that determines if the process chromatography step has been successful and if the in-process material is suitable for forward processing. By transferring the specificity of HPLC to an on-line analyzer, it is possible to directly measure the critical quality attribute in near–real time, thus allowing the process decision—when to start and stop collection of the product eluting from the process chromatography column—to be based on the critical quality attribute (product purity) rather than being based on a surrogate measurement that is affected by the process variability.

Using On-Line HPLC to Automate Process Chromatography Operations

Figure 1 is a diagram of how product purity data generated by the on-line HPLC are transmitted to the distributed control system (DCS), where the data are compared to a product purity setpoint to determine when to start and stop the collection of the product based on the critical quality attribute (i.e., product purity). The diagram also shows the typical critical operating parameters for the unit operation such as conductivity (CT), flow rate (FT), pressure drop (PT), and eluent pH (pH) that are also transmitted to the DCS where feedback control of these parameters is performed to minimize process variability. A parameter that is also controlled, but not shown in this diagram, is the eluent feed to the column if the column is eluted by application of a mobile-phase gradient.

Figure 1. Control of process-scale chromatography column using on-line HPLC to determine product purity

In this operation, the on-line HPLC sends the product purity value, derived from the on-line HPLC data, to the DCS. The product purity value generated by the on-line HPLC is compared to the product purity setpoint in the DCS. If the product purity is lower than the product purity setpoint, the DCS sends a signal to the three-way valve in the column eluent stream to divert the eluent stream to waste. If the product purity is greater than the product purity setpoint, the DCS sends a signal to the three-way valve in the column eluent stream to divert the eluent stream to the product collection tank. By basing the mainstream pooling decision on a direct measurement of product purity, variability in the mainstream pool purity is minimized. Thus, by varying the timing of output collection according to a near–real time measurement of product purity determined by the on-line HPLC analyzer, the process can accommodate variability in the process input and process operations. The dependency of product purity on the operation of the process chromatography column is significantly reduced when compared to collecting the product pool based on elution volume or a combination of elution volume and an A280 value.

By using the on-line liquid chromatography approach to provide near–real time information on product purity, a more efficient and robust process is achieved which provides more consistent product purity. This is accomplished in several ways:

• Eliminating fraction collection significantly reduces opportunities for errors and contamination

• Eliminating off-line analysis of the fractions reduces cycle time

• Yield may be increased by having an on-line, mainstream collection measurement that is more specific to the product, because this allows a larger amount of the elution peak to be collected when compared to measurements such as elution volume or A280 that are not product-specific.

• The process endpoint—determined by achieving a particular product purity value—is flexible and can adjust to variability in the process input and operations.

• Increased levels of process automation (e.g., automated sequencing) can be implemented, thus leading to more precise operations than typically can be achieved with manual operations.

Using on-line HPLC to Enable Unique Operational Approaches

Most biotech processes involve multiple column chromatography steps to achieve adequate purification of the biosynthetic origin product. By using on-line HPLC, Eli Lilly and Company was able to successfully combine two process-scale ion exchange chromatography steps into a single, automated process.¹ This process is explained in Figures 2 through 7.

Figure 2 shows the two process chromatography columns combined using one three-way and two four-way process valves to enable the columns to be operated either in parallel (i.e., as two independent columns) or in series (i.e., as one continuous column) and to allow the eluent stream to be diverted to the waste stream or product collection tank. In this operation, an on-line HPLC analyzer automatically samples and determines the purity of the product in the column eluent stream. The DCS uses these data to determine the appropriate position of the process valves (to operate the process columns either in parallel or series) and to divert the eluent stream to waste or to the product collection tank.

Figure 2. Two process-scale columns combined into a single operation. The positioning of the on-line HPLC downstream of the first four-way valve allows the one on-line HPLC analyzer to monitor the eluent stream from either process column.

Figure 3 shows the two process scale columns operated in parallel mode. The on-line HPLC automatically analyzes the Column 1 eluent stream and sends the data to the DCS system where they are compared to the product purity setpoint. If only frontside impurities are detected in the eluent stream (i.e., no product is present), the stream is automatically diverted to waste and the two process columns continue to operate in the parallel operation mode.

Figure 3. Columns operated in parallel mode. Frontside impurities from column 1 are automatically diverted to waste.

Figure 4 shows the two process scale columns operated in series mode. The on-line HPLC automatically analyzes the Column 1 eluent stream and sends the data to the DCS system where they are compared to the product purity setpoint. At this point in the process, the product of interest is detected in the Column 1 eluent stream, so the process valves are automatically positioned to operate the two process columns in series, thus allowing a "heart cut" from Column 1 to be automatically charged to Column 2 for further purification.

Figure 4. Columns operated in series mode with the heart cut from column 1 are being automatically charged to column 2.

In the next step, shown in Figure 5, backside impurities are detected eluting from Column 1. As a result, the DCS automatically adjusts the process valves so that the two process columns operate in parallel, allowing the backside impurities from Column 1 to be eluted to waste. Repositioning the process valves also allows the same on-line HPLC that was monitoring the eluent from Column 1 to begin monitoring the eluent stream from Column 2. The frontside impurities are detected eluting from Column 2 and are automatically diverted to waste under the control of the DCS.

Figure 5. Columns operated in parallel mode. Backside impurities are eluted from Column 1 while frontside impurities are eluted from Column 2 and automatically diverted to waste.

When the on-line HPLC detects high purity product eluting from Column 2 (Figure 6), the DCS repositions the valves to allow the product to be automatically collected in the product collection tank. With the process valves in this position, the columns operate independently (i.e., in parallel mode). This allows Column 1 to undergo regeneration while Column 2 is being eluted. Performing these two operations simultaneously further reduces the overall process cycle time.

Figure 6. Columns operated in parallel mode. Column 1 is undergoing regeneration while the high purity product is automatically collected from Column 2.

In the final step of the process, the columns continue to be operated in parallel mode (Figure 7). The backside impurities are eluted from Column 2 and are automatically diverted to waste. At this point, the regeneration of Column 1 is complete and the column is ready to receive the next production lot. Column 2 can be regenerated while Column 1 is charged with the next lot to be purified.

Figure 7. Columns operated in parallel mode. The regeneration of Column 1 is completed and ready to receive the next production lot. The backside impurities are automatically eluted to waste from Column 2 to make it ready it for regeneration.

Thus, by using on-line HPLC to provide a near–real time measurement of product purity, two batch purification steps have been combined into a single operation. This combination reduces the overall cycle time because the elution of Column 1 occurs simultaneously with the charging of Column 2. All fraction handling is eliminated along with its associated opportunities for errors, thus removing a significant potential for process failure.

The on-line measurement of product purity provides the information necessary to automatically sequence the column operation as well. In addition, since the product never leaves the process piping and all valves are equipped with position switches, the control system has continuous confirmation of where the product is flowing.

Eliminating the off-line HPLC analysis of fractions in the central laboratory reduced the overall cycle time of the process by two days. The combination of automated process sequencing and eliminating off-line analysis allowed the throughput of this particular process step to be increased tenfold.¹ These process changes also reduced process variability and increased process yield.

Conclusions

Using on-line HPLC to monitor the eluent from process-scale chromatography columns allows a critical quality attribute (product purity) to be measured on-line in near–real time. The product purity data generated from an on-line HPLC analyzer can be used to automate the collection of product pools. Using the measurement of the critical quality attribute to adjust the mainstream pool collection points allows the process to vary as needed to compensate for variability in the process scale separation. This flexible process endpoint allows this variability to be accommodated by varying the timing of mainstream pool collection rather than have it show up as increased variability in the mainstream pool purity.

In addition to reducing product variability, using on-line HPLC also increases product yield; enables the use of increased levels of automation; reduces opportunities for errors; and reduces overall cycle time.

Obtaining product purity data in near–real time can enable unique approaches to product purification to be implemented. For example, two process-purification column steps can be combined and their sequencing can be fully automated. When this combination of column operations was carried out at Eli Lilly and Company process throughput was significantly increased, primarily as a result of cycle time reduction, allowing the production area to increase its throughput for the process step tenfold.

Acknowledgements

The author thanks Eli Lilly and Company for allowing the use of its process example to demonstrate the value of on-line HPLC in process scale chromatography operations.

The author would like to particularly acknowledge the invaluable contributions of several scientists, engineers, and technicians who were involved in the development, implementation, and support of this pioneering work at Eli Lilly and Company: Charles Stevenson, Danny Johnson, David Crozier, Richard Moss, Michael Hilgert, Robert Wilken, Jim Owens, Jerry Shrake, and Mearl Gibson. In addition, without the challenging questions of the late Dr. Leroy Baker, the development and use of on-line HPLC probably would not have been initiated at Eli Lilly and Company in 1980.

RICK E. COOLEY is the manager of process analytics center of excellence at Dionex Corporation, 765.349.6002, rick.cooley@dionex.com

References

1. Cooley RE. Utilizing PAT to monitor and control bulk biotech processes Presentation at University of Michigan Pharmaceutical Engineering Seminar, 2003 Mar 4. Available from: www.fda.gov/cder/OPS/cooley/index.htm