Send to printer »

Feature Articles : Jun 1, 2008 (Vol. 28, No. 11)

PAT Initiative Helps Move HPLC Forward

Achieving the Speed, Reliability, and Robustness Firms Demand
  • Angelo DePalma, Ph.D.

High-performance liquid chromatography (HPLC) has provided analytical support for biopharmaceutical research, development, and manufacturing for as long as biotech has existed. Only recently, though, have HPLC, its methods and associated instrumentation achieved the speed, reliability, and robustness sufficient to analyze process streams in real time or near-real time.

If process analytic technology (PAT) is not at the tip of your tongue, it should be. This FDA-inspired initiative for real-time, in-process analytics, has been slowly gaining traction thanks to steadfast efforts by instrument vendors and best-in-class biomanufacturers. HPLC will be part of PAT moving forward.

Another important factor in the emergence of process HPLC is quality by design (QbD), a concept, like PAT, borrowed from nondrug process and manufacturing industries. Under QbD, quality is designed-in as the process moves forward, rather than tested-in afterward by quality groups.

From Lab to Manufacturing Floor

The complex nature of biomolecules in their raw form and the realities of the manufacturing floor have both contributed to holding process HPLC back. At-line or in-line analytics must be rugged, close to plug and play, and utilize methods suitable for production settings. Seemingly, every technical paper published on bioprocess PAT notes the difficulties in transferring otherwise robust analytical platforms and methods from the ideal conditions of back-room laboratories to the center stage of biomanufacturing. “Production environments are different from analytical labs,” notes John Waraska, marketing manager at ESA Laboratories(www.esalabs.co.uk).

Problems arise from every angle. Conventional columns and pressures are too slow for some processors’ tastes, while methods that serve central laboratories or benchtop analysis are unsuitable for real-time analysis. Some detection methods, such as ultraviolet, may not provide the detail required for analyzing complex process streams; others, like mass detectors, are still too complex for general use at-line or in-line. “You don’t want a mass spec on the production floor at this point in time. They’re still too skill-intensive,” Waraska adds.

The idea of a universal detector therefore holds great appeal. The closest thing on the market is ESA’s Corona CAD® (charged aerosol detector), an HPLC detector that uses evaporative techniques and light scattering to quantify almost any analyte, including proteins, small molecules, nutrients, and counter-ions—every relevant component of a fermentation sample. During method development, the CAD may be configured to split the sample stream and send half to the CAD and half to a mass detector to validate the in-process CAD method.

Recently, scientists from Discovery Laboratories (www.discoverylabs.com) presented data on the analysis of complex mixtures of lipids and peptides using CAD. Discovery Labs’ Surfaxin® pulmonary surfactant, which is currently under FDA review, consists of a peptide, KL4, two phospholipids, and palmitic acid. Where currently marketed surfactants are animal-derived—and therefore complex both medically and compositionally—Surfaxin is synthetic.

Senior director of analytical services at Discovery Labs, Michelle DeCrosta, Ph.D., says that the synthetic approach leads to tighter control over product quality. “It’s more simplistic, the impurity profiles are predictable, and it’s reproducible.”

Simplicity notwithstanding, analyzing Surfaxin conventionally requires that one detector quantify the four active ingredients that belong to three distinct chemical classes. Ultraviolet detection is not quantitative for lipids, while fluorescence requires conjugation to a fluorophore and may require calculating response corrections. Consequently, at one time Discovery Laboratories used six separate HPLC methods to quantify the actives and key impurities in Surfaxin.

ESA’s CAD detector reduces analysis to a single injection and elution in about 35 minutes. Not ultrarapid by any stretch, but “not bad by biotech standards,” Dr. DeCrosta points out. The CAD analysis is sensitive, with a limit of detection at less than half a microgram per milliliter plus good linearity for actives and impurities. “CAD allows us to increase specificity, to where we can see all actives and impurities.”

Because it relies on evaporation of sample droplets, CAD works independently of UV wavelength and does not require chromophores. According to Discovery Labs, CAD substantially reduces sample-preparation complexity and overall cycle time. The company now employs the CAD method for stability and release testing.

Comparing Methods

Amgen’s process-development group recently performed a feasibility study of HPLC to help guide fraction-pooling decisions during preparative liquid chromatography.

Researcher Anurag Rathore, Ph.D., analyzed fractions from a hydroxyapatite column, comparing results by three methods: UV absorbance at 280 nm, an online DX-800 process analyzer from Dionex, and an offline reference system consisting of a TSK-GEL G3000 SWXL column from Tosoh Bioscience in an Agilent (www.agilent.com) 1100 LC system. The objective was to identify fractions containing undesirable amounts of protein dimers and other high molecular weight species.

The three methods have their pluses and minuses. UV is simple and robust, but cannot differentiate among chemically distinct species absorbing at 280 nm. The offline reference analysis is accurate but time- and resource-inefficient. Online analysis was effective in the test separation, but its complexity raises issues for other applications.

Dr. Rathore was able to tweak the online analysis method by increasing the HPLC flow rate while simultaneously slowing down elution from the preparative column. Analysis might be speeded up even more by switching to ultrahigh performance LC.

One might ask at this point if real-time analysis is ever necessary in biotech. “That depends on the time scale of the unit operation,” Dr. Rathore says.

For a cell culture, which may last weeks, the window for real-time analysis is long, on the order of hours, while for others, such as fraction pooling for preparative chromatography the real-time requirements, are much more apparent.

In addition, the fact that chromatography always involves a retention time means that HPLC can never provide true real-time analysis as ultraviolet- or other spectrometric-detection methods can. “Bioprocessors rarely need to know what’s going on every second,” observes Ira Krull, Ph.D., professor at Northeastern University. Depending on the process, even rather slow HPLC can serve process engineers adequately.

One Answer: Column Technologies

Craig Dobbs, senior pharmaceutical operations manager at Waters (www.waters.com), disagrees with the knock on real-time HPLC. He explains that, while spectroscopic detection is instantaneous, its associated data workup is not. When all factors are considered, he argues, a fast LC technique like Waters’ UltraPerformance LC (UPLC) may be time-equivalent to techniques like near-infrared, UV, and Raman, providing what amounts to real-time analysis.

UPLC is a column or stationary-phase solution that achieves high speed without sacrificing resolution. Waters’ Patrol UPLC Process Analyzer uses sub-2 micron particles operating in the 12,000–15,000 psi pressure range while conventional HPLC runs at up to about 6,000 psi.

Waters promotes UPLC as the first real advance in LC in 35 years.

“You can run any conventional HPLC method on 35-year-old systems,” Dobbs says. “Instruments have become easier to use and more reliable and rugged, but the basic system hasn’t changed since 1972.”

Data from UPLC experiments is impressive. For real-time analysis, UPLC does not demand a trade-off among resolution, sensitivity, and speed. UPLC routinely separates complex solutions in less than 30 seconds. “It makes real-time LC a reality,” notes Dobbs.

He explains that while UV, Raman, and fluorescence operate in real time, deconvoluting the spectra takes several minutes. In that time most operators can perform one or more UPLC runs.

Patrol UPLC’s standard online configuration enables fully automatic sampling from two independent process streams or sources. The system uses high-precision syringe drives for all sample and reagent metering to ensure accuracy throughout all phases of operation.

Patrol is also available in an at-line configuration, which enables an operator to manually introduce a barcoded sample. The system reads the barcode; checks with the laboratory information management (LIMS) or distributed control system (DCS) software to confirm that the vial contains an appropriate sample for analysis; loads the appropriate analytical method, which includes column and buffer selection; begins the UPLC analysis; and, at the end, sends the data to the DCS or LIMS.

Waters is currently collaborating with several pharmaceutical companies to quantify Patrol UPLC’s impact on manufacturing, with the goal of implementing it for real-time PAT.

“Process HPLC differs from conventional analytical HPLC in the complexity of its samples, and often in its need for speed without sacrificing resolution or accuracy,” says Ales Strancar, Ph.D., CEO of BIA Separations (www.biaseparations.com).

Process-level analysis is complicated by impurities, side products, additives, buffers, protein folding, and aggregation—most of which are gone by the time quality groups get hold of the product.

BIA specializes in polymeric monolithic columns that satisfy process HPLC’s need for speed, versatility, high resolution, and ease of scale-up. Separations that take an hour with porous-particle stationary phases run in two minutes on a monolith. That is one reason these columns have become a hot topic at HPLC conferences.

One exciting bioprocess application of monolithic columns is the analysis of vaccines, particularly virus-based products. Viruses and virus fragments tend to plug porous stationary phases, but not monoliths. A group in England is using BIA columns to separate viral particles from capsids and partially degraded viruses. According to Dr. Strancar these methods are suitable for process control and stability testing.

The first vaccine purification using BIA’s CIM® monolithic columns took place in 2006. Other adopters include Novartis and Octapharma.

Boehringer Ingelheim (www.boehringer-ingelheim.com) was reportedly the first company to use monolithic columns for the industrial cGMP purification of plasmid DNA, noting a 15-fold increase in productivity.