Send to printer »

Feature Articles : Sep 1, 2010 (Vol. 30, No. 15)

Process Monitoring Realm Expanding

Traditional Sensor Approaches Evolve to Include Measurements that Enable PAT
  • Angelo DePalma, Ph.D.

Bioprocess monitoring is moving rapidly beyond process sensors for dissolved oxygen (DO), pH, glucose, and carbon dioxide. Development groups and vendors are now coupling those critical measurements to more complex quality attributes, and ultimately to feedback control, as part of process analytic technology (PAT) initiatives. Ultimately, these techniques will lead to a top-down quality by design (QbD) approach to manufacturing.

“The idea of PAT is to have more and more data measured online,” says Gernot John, Ph.D., of PreSens Precision Sensing. “In the past the process was a black box, where at the end you analyze the product to assure the quality of the process.”

Online measurement, Dr. John explains, enables not only process control but ongoing process improvement. This differs from fixed processes that were “not touched” during production. “We are not there yet, but at some point we will be.”

PreSens manufactures chemical-optical sensors for oxygen, carbon dioxide, and pH. The first two devices are suitable for both stainless steel and single-use equipment; the pH sensor is used overwhelmingly in disposable equipment because “the classical pH sensor is more suited for multiple usage" for fixed-tank reactors. PreSens also supplies sensing-data transmitters and complete monitoring systems that include two or more sensors or probes.

Riding a Workhorse Method

Process monitoring using high-performance liquid chromatography, arguably the workhorse analytical method in biology, is an intriguing idea that is just now catching on. Traditional HPLC has for years been deemed too slow for in-line or online measurements, but wider adoption of fast LC techniques like uHPLC (ultra-HPLC) and monolithic LC columns is changing the equation.

The utility of fast LC monitoring depends on the application, says Helmut Schulenberg-Schell, Ph.D., worldwide marketing manager for LC at Agilent Technologies. For amino acids and small peptides, uHPLC’s one-minute cycle time can be quite efficient for monitoring runs in near-real-time. “But for proteins or larger molecules, runs are significantly longer and fast feedback more difficult to achieve.”

Nevertheless quite a few large pharmaceutical companies, according to Dr. Schulenberg-Schell, are working toward the goal of real-time uHPLC-based monitoring. Moving uHPLC equipment into process suites is easy, and the equipment is robust.

The challenge, he says, is to combine knowledge from several disciplines, for example the process engineer running a bioreactor, sampling, chromatography, and data analysis. Automation can be a tremendous help but analyzing chromatograms, for purposes of real-time control, will probably have to wait until the first shift crew arrives.

HPLC is more likely found in a PAT-like role during process development. Shelly Pizarro at Genentech has published a technique for confirming the validity of dissolved oxygen on a protein refolding reaction using reverse-phase HPLC. The overriding quality issue was to achieve sufficient levels of active protein while avoiding oxidative degradation.

Working with a 3 L bioreactor, Pizarro correlated folding and degradation levels to DO concentrations as measured directly by in-line sensing. She found that DO profiles and refolding correlated and were comparable across scales, thus validating her PAT approach.

Traditional process monitoring is based on life-size sensors, transmitters, and probes familiar to most bioprocessors. Most of these simply don’t fit the form factors of process-development equipment, which creates a disconnect between PAT-worthy large-scale processes and the smaller vessels in which process ideas are evaluated.

In fact Govind Rao, Ph.D., of the Center for Advanced Sensor Technology at the University of Maryland Baltimore County, argues that lack of adequate sensors has hamstrung process development.

Early process development is characterized by extensive experimentation on process parameters that is generally carried out in many small, un-instrumented “scouting devices” such as flasks, small bioreactors, and microtiter plates. “Ideally, one would want to employ sensing and monitoring of quality-related parameters at the very earliest stages,” Dr. Rao explains.

The cost of “instrumenting” small-scale devices has been prohibitive, however, and the underlying sensing technology inadequate. “Current technology does not permit sufficient control of process scouting devices to adequately replicate large-scale processes.” Part of the problem is integration, and the need for conventional sensors to come into contact with process fluids.

The introduction of disposable sensor devices is beginning to change this picture, Dr. Rao says. In particular, he mentions low-cost, disposable, semi-invasive optical sensor patches that are easily calibrated and provide through-space readout.

These sensors reside within the process but transmit signals through the vessel wall. Several such devices, for example for measuring pH, dissolved oxygen, pCO2 have been developed in Dr. Rao’s lab. These sensors have been commercialized by Fluorometrix and sub-licensed to Sartorius Stedim Biotech. Presens, Finesse Solutions, and other manufacturers offer similar sensors. In addition to being used with process scouting devices, these sensors are finding use in conventional and disposable bioreactors ranging from a few milliliters to several hundred liters.

Issues still remain with microsensors. Foremost is their proprietary status. End-users are uncomfortable with being locked in to a single supplier, while the lack of complete information on sensor composition causes problems with toxicology departments and potentially with regulatory agencies.

To circumvent these issues, Dr. Rao’s lab has developed next-generation technology where the sensors are placed outside the bioprocess vessel with a suitable barrier. Sartorius Stedim Biotech has acquired the exclusive license to this technology, which can allow for entire bioprocess trains to be integrated and monitored using noninvasive sensors.

Ultimately, Dr. Rao feels that a broad technology sharing consortium approach is essential for widespread adoption of the sensor technology and for its incorporation into disposables. He likens the development of the USB standard for computers as an example that the bioprocess instrumentation community could follow.

Such an approach could revolutionize the field with plug-and-play sensors that would be available from several vendors and yet offer standard outputs for connectivity to control systems. This evolution will greatly facilitate PAT approaches that will drive QbD.

Tips for Improving Bioprocess Monitoring—Consider:

1 Measuring as many of your critical bioprocess monitoring variables as possible online.
2 Adopting fast LC techniques that utilize tools such as ultra-HPLC and monolithic LC columns for process monitoring.
3 Employing sensing and monitoring of quality-related bioprocess parameters at the very earliest stages.
4 Using disposable sensor devices and disposable buffer handling bags wherever possible.
5 Carrying out in-line buffer dilution as it may reduce the need for storage tanks.
6 Relying on total organic carbon (TOC) analysis as a method for verifying the effectiveness of clean-in-place procedures.

Cleaning and cleaning validation, while not technically manufacturing steps, nevertheless factor into processing time. Hyde Engineering  has introduced a technique, based on total organic carbon (TOC) analysis, for verifying the effectiveness of clean-in-place (CIP) operations. TOC should drop as cleaning proceeds.

The technique collects rinse water from the CIP return lines and delivers it to a high-throughput GE 900 Turbo TOC analyzer. A pump pulls rinse samples for systems under vacuum, while the return pump suffices for CIP systems under positive pressure.

The GE instrument takes a data point every 4.5 seconds, but the technique is not quite that fast. Due to residence times within the instrument and sample lines, and instrument equilibration, there is a two- to five-minute delay between sampling and readout.

“This is not quite an in-line sensor, it’s more online,” comments Keith Bader, director of technical and quality services at Hyde.

Timeliness is approximately the same as that of an at-line analyzer where samples are pulled manually. The advantage lies in the automation, which greatly reduces the likelihood of sample contamination.

Bader believes the the TOC technique, which has been tested at a large biopharmaceutical manufacturer in a microbial fermentation, will eventually be used as a release test, or verification of CIP effectiveness. “Users can determine within five minutes of completing a cleaning cycle if the equipment is clean,” he says, based on a method that is orthogonal to traditional criteria such as conductivity.

Another essential but nonproduction step, buffer preparation, has come under increasing scrutiny due to the capital expense for holding/mixing tanks, related cleaning and cleaning validation, and the serious demands these activities place on floor space. Disposable buffer handling bags can eliminate the first two concerns, at least up to the bag’s working volume, but concerns over where to put these tanks—whether full or empty—remain.

In-line buffer dilution can reduce the need for storage tanks but precision metering of buffer concentrates has traditionally been a problem. Michael Li, Ph.D., process sciences manager at Asahi Kasei Bioprocess, has devised a solution based on real-time conductivity measurements that is fully PAT worthy.

Because it is real-time and controls the flow of both concentrate and water, the technique does not require precise composition of the buffer concentrate to achieve exact final buffer concentrations. “Once you know the proper conductivity of the end solution, the feedback control assures consistency every time,” Dr. Li says.

Building on the Scientific Foundation

With increasing numbers of bioprocessors taking PAT and QbD seriously, the future appears bright for bioprocess monitoring and advanced technologies for measuring quality attributes in real- or near-real time.

One hurdle to adoption of new process-monitoring technologies is the risk aversion of many biopharmaceutical companies, particularly with respect to highly regulated operations like cleaning and fermentation. But the FDA has made an effort, says Bader, to evaluate new techniques and technologies scientifically.

Regulators and companies alike, he says, are looking for favorable, substantiated precedents, preferably in already approved manufacturing processes. “Once they see a few case studies, see that people are doing it, and the technique has gone through regulatory approval, everyone is much more open.”

Sidebar: Applying PAT to Vaccine Manufacturing

The complexity and heterogeneity of vaccines makes them much more difficult to characterize than therapeutic proteins, and even more so in real time. For example, mAbs of molecular weights up to about 150 kDa range are easily analyzed for sequence, isoforms, and glycosylation.

“By contrast, virus-like particles can be more than 1,000-fold greater in size than proteins and hold much more complex structures,” notes Gerard Powell, Ph.D., analytical development scientist at Eden Biodesign. Vaccines are also held to a higher safety standard than other biologicals, which makes safety an even higher priority, both during manufacture and after release.

The question, given the complexity of vaccine formulations and relevant analytics, is which techniques lend themselves to in-process monitoring, and how many such orthogonal methods will provide manufacturers and regulators with a reasonable level of safety assurance?

Electrophoresis, immunoblotting, PCR, amino acid sequencing, quaternary structure analysis, and electron microscopy are all quite slow. And given the heterogeneity of vaccines, real-time platform analytics are unlikely to emerge.

“But where it is impossible to identify every attribute of the whole vaccine, well-chosen assays can ensure an overall quality picture is achieved,” Dr. Powell says. For example process engineers can monitor total protein and quantify sub-visible particles; purity may be established using ion-exchange HPLC using novel monolithic columns and size-exclusion chromatography with multi-angle laser light scattering analysis; biological activity is assessed through serological assaying and, in the case of adenoviruses, measuring infectious titer.

However, classical analytics such as HPLC/ uHPLC and spectroscopy, while useful, lack the resolving power and sensitivity to detect small changes that may result in production of an effective vaccine, or even a safe product, in real time.

Yet, says Dr. Powell, applying PAT to vaccine manufacturing and process development is essential “if we are to fully understand, and ultimately be able to control, the manufacturing process.”

First, vaccine-makers must establish critical quality attributes whose real-time measurements may be used to control the process. Applicable techniques may include on-line liquid chromatography approaches such as size-exclusion chromatography coupled with multi-angle light scattering detection, and ion-exchange chromatography using monolithic columns. These techniques, he says, can estimate quality during process development as well as production in much the same way that analytical protein A chromatography allows with mAbs.

“These on-line PAT techniques allow much more flexibility in rapid decision making during vaccine process development, which is critical to building optimal platform processes for vaccine manufacturing.”