Anda di halaman 1dari 9

ARTICLE Process Analytical Technology (PAT) for Biopharmaceutical Products: Part I.

Concepts and Applications


E.K. Read,1 J.T. Park,1 R.B. Shah,1 B.S. Riley,1 K.A. Brorson,1 A.S. Rathore2
2

Ofce of Pharmaceutical Sciences, CDER, FDA, Silver Spring, Maryland, USA Process Development, Amgen Inc., Department of Chemical Engineering, Indian Institute of Technology, Hauz Khas, New Delhi 110016, India; telephone: 805-774-8986; fax: 91-1126581120; e-mail: asrathore@biotechcmz.com

Received 3 July 2009; revision received 26 August 2009; accepted 31 August 2009 Published online 3 September 2009 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bit.22528

ABSTRACT: Process analytical technology (PAT) has been gaining momentum in the biotech community due to the potential for continuous real-time quality assurance resulting in improved operational control and compliance. In this two part series, we address PAT as it applies to processes that produce biotech therapeutic products. In the rst part, we address evolution of the underlying concepts and applications in biopharmaceutical manufacturing. We also present a literature review of applications in the areas of upstream and downstream processing to illustrate how implementation of PAT can help realize advanced approaches to ensuring product quality in real time. In the second part, we will explore similar applications in the areas of drug product manufacturing, rapid microbiology, and chemometrics as well as evolution of PAT in biotech processing. Biotechnol. Bioeng. 2010;105: 276284. Published 2009 Wiley Periodicals, Inc.y KEYWORDS: process analytical technology (PAT); cell culture; harvest; chromatography; UF/DF; virus ltration

Introduction
In September 2004, the US Food and Drug Administration published the guidance document: PATA Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance (FDA, 2004). This guidance described a regulatory framework to encourage the voluntary development and implementation of innovative pharmaceutical development, manufacturing, and quality assurance. PAT was subsequently described by ICH as an opportunity to develop more exible regulatory approaches (ICH, 2008b).
Correspondence to: A.S. Rathore

The essence of PAT involves process understanding combined with monitoring of raw material or in-process product attributes in real-time to control the process by adjusting process parameters, with the goal of ensuring nal product quality. The decision to transition from one process step to the next or an adjustment of a parameter is made at dened decision points (points in a process at which transition decisions are made) and based on clear decision criteria (real-time information that triggers a transition). Overall, this strategy transitions product quality assurance from after the fact xed controls (i.e., batch testing of end product) to adaptive controls (i.e., monitored and controlled during the manufacturing process) with a goal of more efciently achieving the desired material attributes of the end product. The introduction of the Quality by Design (QbD) initiative (ICH, 2008a; Rathore, 2009; Rathore and Winkle 2009) has reshaped the regulatory arena since the initial release of the PAT guidance. QbD can be viewed as an umbrella encompassing several sometimes overlapping concepts that include creation of a manufacturing knowledge base, risk-management principles, process design spaces and PAT. Some of these concepts, like riskmanagement, have at this point been adopted to a greater extent by industry than PAT (Ahmed et al., 2008). Like PAT, many of the concepts under the QbD umbrella had a genesis that predated ICH Q8. For example, process design spaces are similar to the modular/bracketed validation concept initially proposed by CBER in the Monoclonal Antibody Points to Consider document of 1997 (Brorson et al., 2003; FDA, 1997), with some differences (e.g., the process design space concept stresses multivariate analysis). The PAT approach was originally embraced by the small molecule world, where active pharmaceutical ingredients (APIs) are produced by series of chemical reactions and where powders are blended and/or granulated prior to

276

Published 2009 Wiley Periodicals, Inc. yThis article is a US Government work and, as such, is in the public domain in the United States of America.
Biotechnology and Bioengineering, Vol. 105, No. 2, February 1, 2010

tableting. Many processes have been described in the scientic literature where monitoring and controlling the chemical end products or blended powders is achieved by technologies such as Near Infrared (NIR) (Scott and Wilcock, 2006). PAT has been embraced by some traditional pharmaceutical rms in discrete areas such as ensuring powder blend uniformity and control. Specic applications are under review and starting to achieve regulatory approval. The fact that a step transition towards universal PAT implementation has not occurred is not a result of conceptual aws with or the inapplicability of PAT, but rather economic and business factors driving manufacturing decision-making including but not limited to implementation costs, legacy systems, existing strategies, sensor issues, and regulatory concerns (Scott and Wilcock, 2006). The biotech world presents a unique set of challenges: (1) proteins are produced by highly complex cell-based biological systems which are highly sensitive to external conditions such as temperature, pH and nutrients; (2) Inprocess intermediates can be complex mixtures where the desired protein may be a small fraction of the bulk liquid containing host cell proteins and other process related impurities; (3) Worrisome impurities (e.g., viruses) would be a concern even when present at levels undetectable by even the most sensitive in-line/on-line/at-line technologies; (4) raw materials are complex and often variable; and (5) the functional/clinical signicance of product variants, small changes in molecular structure in the product or prole of impurities is often incompletely characterizable or understood. In contrast, some signicant challenges for small molecule drugs may not apply to biotech; for example, blending of aqueous protein solutions is usually never as problematic as small-molecule drug powder blending where PAT approaches can substantially improve operations and control. While at the present time, PAT has not been implemented in every aspect of bioprocess validation (e.g., removal of low-level impurities and virus is still evaluated in spike/ removal studies), the examples described below are feasible and can easily add value to biotech manufacturing. PAT should be viewed as a QbD enabler as well as an approach to reduce after-the-fact testing and to facilitate real-time release. It may not make business sense to apply PAT for every unit operation in the process and the pros and cons should be weighed to make this decision (Rathore et al., 2006). As sensor technology improves over time, expansion of the list of feasible applications is likely.

CQA, or (3) process control based on real-time measurement of parameters that conrm that a unit operation/piece of equipment continues to be t for purpose (e.g., lter back pressure, column integrity, feedback from a vibrational monitor on a pump). These categories are based on whether the sensor measures a product attribute (i.e., an indirectly controlled material output) or a parameter which has been shown to directly affect the product CQA. Whether these #2 and #3 examples are truly PAT is a matter of perspective concerning whether one should include control of inputs into the PAT paradigm. The #2 and #3 examples, as well as techniques that enhance process understanding without control may be viewed steps on the path to the more narrow #1 denition for these highly complicated biologic systems. In this paper, we present case studies of all these types, but will focus on the rst set of applications. The goal of PAT is to enhance understanding and control the manufacturing process. Hence a focus needs to be placed on relevant multi-factorial relationships among material, manufacturing process, environmental variables, and their effects on quality. This enhanced focus provides a basis for identifying and understanding relationships among various critical formulation and process factors and for developing effective risk mitigation strategies (e.g., product specications, process controls, training). The multi factorial studies are required to identify variables that interact to impact product quality versus those that do not interact. There may be cases where such studies show that univariate controls sufce for a given application. Simple control systems is a desirable outcome, and the best systems are those where the process itself assures quality without intervention. Successful application of PAT requires an appropriate combination of some, or all, of PAT tools (Multivariate tools for design, data acquisition and analysis; Process analyzers; Process control tools; Continuous improvement and knowledge management tools) may be applicable to a single-unit operation, or to an entire manufacturing process and its quality assurance. The control strategy and the decision on what to analyze are based on process understanding that is generated during the design and routine operation of the process/ unit operation. Together, three steps, design, analyze and control, ensure that product CQAs are met consistently (Fig. 1).

PAT Example Case Study Denition of PAT


After the release of the PAT guidance by FDA, the wider biotech community has viewed PAT approaches in three categories. (1) Process control based on real-time, direct measurement of a product (or raw material) critical quality attributes (CQAs), (2) process control based on real-time measurement of parameters that directly correlate with a In the following example, we show that PAT is more than just analysis; it also requires process understanding and process control. This case study involves depth ltration monitored by UV absorbance at 280 and 410 nm, adapted from an observation by Yigzaw et al. (2006) that host cell protein complexes adsorb 410 nm light. Depth lters are commonly employed during the post cell culture clarication stage to remove cell fragments and protein complexes

Read et al.: PAT for Biopharmaceutical Products Biotechnology and Bioengineering

277

Figure 1.

Illustration of three steps for successful execution of PAT: design, analyze and control. [Color gure can be seen in the online version of this article, available at www.interscience.wiley.com.]

undesirable host cell proteins (HCP) complexes. HCP complexes can breakthrough the depth lter, when present at high levels in the process uid and/or when the depth lter is overloaded with high volumetric throughputs. This is an undesirable outcome because HCP complexes can foul or clog downstream columns. A410 has been described as potential monitoring tool for HCP complexes in depth lter and protein A unit operations (Yigzaw et al., 2006). Monitoring at A280 would determine the start of peak collection. When A410 signal in the ltrate starts to increase, the process is terminated to prevent HCP from co-purifying with the product. Thus, elevated A410 can effectively monitor a relevant attribute, and hence facilitate real-time process control, and either terminate processing or switch to a new lter. Hence A410 is an appropriate process analyzer for this application and the control scheme utilized in Figure 2C follows PAT.

that can ultimately co-purify with product or clog capture columns (as illustrated in Fig. 2A). Figure 2B illustrates typical monitoring by A280. Monitoring ow rate, transmembrane pressure and protein content by A280 is commonly employed in protein production to conrm process consistency and detect excursions. Because UV absorbance at 280 nm detects all proteins, it does not differentiate between the protein product and protein related impurities (i.e., host cell protein complexes) and is not useful for this aspect of process control. In this case, A280 cannot detect when these complexes saturate the depth lter and start breaking through. Hence, UV absorbance at 280 nm can not be used to measure a relevant product attribute, and hence control the process by dening the end point. This is not a PAT application. Figure 2C illustrates an example of what could be a PAT application. In this case, the depth lter is monitored for typical parameters (ow rate, transmembrane pressure and protein content by A280), but also the efuent is also monitored and controlled by A410, specically detecting

PAT Applications for Cell Culture Unit Operations


Cell culture represents a challenge for the application of PAT, but also has the potential to yield the most dramatic improvements in drug substance quality and manufacturing costs by measurement and control of CQAs. Despite recent progress, gaps remain in the cell culture knowledge space where important variables and interactions between variables are complex and poorly understood because the biological processes are inherently complex making operational control challenging. The complexity and heterogeneity of biopharmaceutical proteins, as well as the instrumentation employed to assess relevant quality attributes contribute to difculty in applying PAT. This is not to say that the eld is devoid of real-time measurement tools, indeed, in-line sensors that measure important cell culture parameters like temperature, pH and dissolved oxygen have enabled process control for decades but these do not facilitate direct control of product quality attributes. The eld seems poised to leap forward with variety of high tech instruments, some of which are listed in Table I. Discussion of the challenges and benets of controlling cell culture for therapeutic protein production will be generically divided into the following unit operations: media design and preparation, seed train and inoculation, production phase, and harvest.

Media Design and Preparation The formulation of a basal media is the rst stage of cell culture where PAT approaches can be envisioned to improve drug substance quantity and quality. Media components are generally well controlled process inputs, even so, variability in raw material composition can lead to process perturbations (Kirdar et al., 2008). The cell culture process outputs are quality attributes and nal product yield and quality

Potential PAT application for depth ltration. A: product and HCP complex ow during depth ltration that is detected by a sensor (S) and ow valve controlled (V) controlled by a PAT actuator (- - -). B: monitoring process by A280. C: monitoring process by A410.

Figure 2.

278

Biotechnology and Bioengineering, Vol. 105, No. 2, February 1, 2010

Joeris et al. (2002), Camisard et al. (2002) Kacmar and Srienc (2005), Sitton et al. (2006), Sitton and Srienc (2008) Kreij et al. (2005), Clemente et al. (2008) Veale et al. (2007), Tridaphillou et al. (2007), Holm-Nielsen et al. (2008)

Detect metabolic end products Sensing metabolic process failure Detect media components and metabolic end products Detect media components and metabolic end products Measure homogenate components Detect metabolic end products Detect media components and metabolic end products

Assess product concentration and afnity Assess product concentration and structure Assess product concentration and structure

Application

attributes. Although not a quality attribute, product yield is important to manufacturers for process economics and to regulatory authorities as a gauge of process consistency and overall control of the facility. One approach involves a Design of Experiments (DOE) studies of media formulation that can reveal previously unappreciated interactions between ingredients and the cell culture. DOE also avoids quasi-optimums frequently derived from single variable experiments (Mandenius and Brundlin, 2008). Such media optimization experiments can be expensive at the manufacturing scale. Historical data obtained from previous production runs can be retrospectively analyzed by a variety of chemometric algorithms such as MVDA (see Part II of this review series) to reveal variability within raw material sources (Kirdar et al., 2008) and to guide advanced parameter control (Johnson et al., 2007; Kirdar et al., 2008). To gain this understanding, sensor enabled, disposable scale-down bioreactors provide a cost-effective alternative during the scouting phase of bioprocess design (Rao et al., 2008). Finally, commercial scale data from protein expression platforms can streamline the gathering and application of process understanding, which can be applied cross-products with biochemical similarities (e.g., monoclonal antibodies).

References

Jacquemarte et al. (2008) Larson et al. (2002) Klyushnichenko (2004)

P, H P, H P, H

S, P, H

S, P S, P

M, S, P, H P

Stage

Characterize cell population Characterize cell population

Sensing biological contaminants Detect media components

P S, P P P M P P

Teixeira et al. (2009) Ulber et al. (2001) Noui et al. (2002) Mukhopadhyay et al. (2005) Larson et al. (2002)

Carvell and Dowd (2006)

Determine biomass

Seed Train and Inoculation While the medium is prepared, the cell culture that will seed the bioreactor must be thawed from frozen parent stock and expanded to sufcient biomass through successive cycles of dilution and growth. Occasionally, a small fraction of the production phase is reserved as seed for the next production run. Before seeding of the nal production phase, the inoculum is subjected to microbial analyses, some of which are amenable to PAT control (see Part II of this review series). The quality attributes of the seed culture that contribute most to undesirable production phase variability in product yield and quality are batch-to-batch differences in biomass concentration and specic biomass growth rate, thus PAT strategies that target these cell culture attributes would prove useful. For example, Jenzsch et al. (2006) induced a seed organism to exert feedback control via control of biomass growth rate. Control of growth rate dampens process and subsequent product variability. Variability in product quality can also be introduced by culture perturbations during the seed train by subtle process differences; the culture perturbations are often discernable only by specic analyses such as ow cytometry. For example, during a partial factorial experiment, small amounts of residual trypsin were found to reduce expression of a progenitor potential cell surface marker. This marker was considered a CQA for therapeutic stem cell treatment (Thomas et al., 2009). Such variability in expressed cell surface markers can be detected during seed processing and hence controlled.

Process control and monitoring of product CQAs Surface plasmon resonance Refractive index change High performance liquid chromatography Physicochemical properties Capillary electrophoresis Physicochemical properties Process control and monitoring of parameters mathematically correlated to CQAs Dielectric spectroscopy Capacitance of membrane bound biovolume In situ microscopy Optical imaging of cell population Flow cytometry Light scattering and uorescence

Table I.

Sensor

Metal oxide eld effect transistor Infrared spectroscopy

In situ 2D uorometry Raman spectroscopy UV spectroscopy Mass spectroscopy High performance liquid chromatography

Stages of most likely utility. M, media preparation; S, seed train; P, production; H, harvest.

PAT enabling sensors for cell culture unit operations.

Volatile organic compounds Fluorophore emissions

Fluorophore emissions Shifted wavelength scattering Absorbance at various wavelengths Mass Physicochemical properties

measurement principle

Read et al.: PAT for Biopharmaceutical Products Biotechnology and Bioengineering

279

Production Phase For the purposes of this review, we dene the production phase of cell culture as the growth of biomass and accumulation of expressed protein product starting from seeding of the full scale bioreactor to the decision point that triggers harvest. Production phase can be run in batch mode, lasting generally less than 15 days or perfusion mode, where the bioreactor can be run for months. From a practical standpoint, all of the PAT applications described in this section can also be applied to optimization and control of the seed train which is typically performed in bioreactors, albeit smaller ones. Currently, the production phase culture paradigm maximizes product yield and quality through rigid standardization strategy where the process and subsequent off-line batch release testing are tightly controlled around prospectively dened set points. Parametric measurement devices for cell culture have existed for decades in the form of temperature, pH and dissolved oxygen in-line probes. In such cases, recorded data directs the bioreactor to maintain a desired parameter set-point by the addition of compensatory ingredients (i.e., heat, base, oxygen). The cell culture process also generates a variety of other real-time data from cell culture (not product) attributes (Nyberg et al., 2008) that could in theory be exploited for control if sufcient process understanding exists. For example, some researchers have successfully applied articial neural networks to oxygen uptake rate data to estimate total biomass, a cell culture attribute that can impact the ultimate product (Jenzsch et al., 2006). Another recent application involved development of in-line probes for dielectric spectroscopic measurement of viable cell volume for bioreactor control (Carvell and Dowd, 2006). Finally, in situ microscopy probes that have been developed for measurement of cell density and individual cell volume generate real-time data suitable for control (Camisard et al., 2002; Joeris et al., 2002). A summary of instruments being developed as on-line sensors facilitating operational control of the culture process based upon detection of metabolites can be found in Table I. The data generated by most of these instruments do not directly measure CQAs and require advanced mathematical manipulations thereby limiting the process understanding derived from such analyses. While empirical models can be built to support mechanistic understanding, since they are not based on a rst principles understanding, the correlations found with chemometric models are limited in their utility and in most cases cannot be extrapolated. A potential real-time on-line measurement of total concentration and a product CQA, binding rates, has been achieved by Surface Plasmon Resonance (Jacquemarte et al., 2008). This technology has not yet developed enough to be used for control in a PAT application as it is currently limited to sensing excursions from a previously modeled culture. Another potential PAT tool is Metal Oxide Field Effect Transistor detection of culture contamination from head space gasses (Kreij et al., 2005). Early detection of

culture contaminants would confer considerable benets to bioprocessing by allowing early termination of contaminated cultures. Process control would need to be implemented for both of these applications to be considered PAT. Finally, automated sampling devices allowed high frequency bioreactor analysis by delivering culture broth samples directly to capillary electrophoresis and HPLC for determination of amino acid and protein concentrations within the media (Larson et al., 2002), again cell culture attributes that can inuence product attributes. In summary, on-line applications need to be linked to a rational real-time control strategy to further the future of PAT in bioprocessing. This hinges on additional development of a process understanding, that is, a knowledge space (Lepore and Spavins, 2008), that encompasses instrument detection of culture process response to a variety of perturbations and corrective measures. The primary decisions of the production phase remaining to be subjected to PAT are: when to feed, what to feed and how much to feed to ensure process control and ultimately product quality; these will be discussed in the case study below.

Harvest Phase The rst and most critical decision to be made after a successful cell culture process is when to start the harvest. As in other phases, the ultimate goal of PAT is to base the harvest decision on the real-time measurement of the product quality attributes. The current practice is to time harvest based on operating ranges set in a standard operating procedure. However, with real-time analyzers that can monitor the production phase, the decision to harvest can be based on cell culture attributes such as biomass or direct measurements of actual product yield and quality. Before continuing with the harvest process, microbiological methods are usually implemented to ensure the product has not been contaminated (see Part II of this review series). The next stage of harvest is clarication, where separation of the uid phase product from the cell material takes place. This is usually achieved by centrifugation, tangentialow ltration or expanded bed absorption. Typically, the efuents depleted of cell material by these means are further claried by passage through (charged) depth lters that remove undesirable residual debris and increase clarity by removing aggregates and colloids. In cases of some bacterial expressed products, a solubilization and refolding step may be necessary. In one case, control of clarication was obtained through real-time measurement of throughput clearness and adjustments in feed:occulant ratio towards previously determined clarity set-points (Koggersbol, 2005). DoE strategies have also been applied to acoustic cell retention and expanded bed absorption (EBA) clarication unit operations. In these cases, the goal is to maximize product recovery (Mandenius and Brundlin, 2008). In principle, all of the ltration-based unit operations during

280

Biotechnology and Bioengineering, Vol. 105, No. 2, February 1, 2010

the harvest phase can be controlled by applying the PAT principles and tools, such as those described later in this article (see PAT Applications for Downstream Unit Operations Section).

Case Study: Automated Flow Cytometry for Scale-Up and Fed-Batch Control of Production Phase Cell Culture Unit Operation
Design Frequently, biopharmaceutical culture processes are run in a fed-batch mode where additional nutrients are added to a batch culture to increase the nal concentration of viable cells or to prolong the production phase. Usually feeding to restore depleted nutrients occurs in a specic, pre-dened tight time range specied in a rigorously supervised standard operating procedure. However, such a rigid strategy may introduce variability by not adjusting to the heterogeneity within the cell population and that may exhibit subtle differences in growth kinetics from batch-tobatch. From a PAT perspective, culture feeding should be triggered in response to extant conditions. In mammalian cell culture, stationary phase is presaged by the onset of cell death during exponential phase but off-line measures of cell titer and viability occur too infrequently for precise timing of the feeding and would require manual operator control (Sitton et al., 2006). In-line probes for turbidity and capacitance provide unstructured data incapable of quantifying an absolute concentration of dead cells. However, ow cytometry can distinguish nonviable cells within a population (Schwarz and Osborne, 1995) and has been brought online by integration with an automated cell preparation system which processes a bioreactor sample (Abu-Absi et al., 2003). This allows cell culture adjustments to maintain cell viability and subsequent yield of high quality protein.

simultaneous cultures (Sitton and Srienc, 2008). Exhausted cell cultures are likely to produce lower quantities of correctly processed proteins. To maintain control over viability (and protein quality), non-viable cells (NVC) was monitored to determine the point when cell concentration (NVC) exceeded four times the average NVC prior to the onset of cell death. This triggered a bioreactor feed to boost viability, improving protein yield and quality. The second strategy involved measuring total cell density in a second bioreactor to determine cessation of growth, 30 h after this point a media feed was implemented. This forestalled viability decrease (maximizing culture density, but also improving protein yield and quality) for an additional 15 h before culture death resumed. A comparison of the two control strategies suggests that each is best suited for a different purpose. If the goal of the process is to maximize the area under the cell concentration curve as a function of time, as might be desired for increased protein expression, then the stationary phase-directed feeding based upon total cell concentration is more suitable. However, if the goal is to obtain the highest viable cell concentration in the shortest amount of time, as is often the case during seed train scaleup, then the NVC triggered batch feed is preferable.

PAT Applications for Downstream Unit Operations Performance of downstream purication processes is affected by the microbial or cell culture step and the condition of the harvested material. The downstream purication processes are designed to achieve the certain level of product quality and desired formulation buffer condition. Also, it should minimize bioburden and remove process-related (e.g., host cell proteins, DNAs, fermentation media, etc.) and product-related (e.g., aggregates and fragments) impurities. Typical downstream unit operation steps used in biopharmaceutical industry with an appropriate control strategy for the PAT applications are listed in Table II.

Analyze The automated system was programmed to deliver a bioreactor sample every 25 min to detect the increases in the absolute concentration of nonviable cells that heralds the onset of stationary phase or cell cycle physiological events that occurs on a shorter time scale (Kacmar and Srienc, 2005). Light absorbance due to FSC and SSC measurements of individual cells are used to calculate the concentrations of viable and nonviable cells that were plotted as a function of time. These data were also used to calculate the total cell concentration and the relative fraction of each subpopulation. UltraltrationDialtration (UFDF) Currently, dialtration steps are often operated using xed volumes,; these are pre-specied based on process development studies. Filtration steps can be controlled through the use of transmembrane pressure and/or surrogates (e.g., uid ow rate). A PAT-based control strategy has been used to end a UFDF step when the actual dialtration process is complete (i.e., salt and buffer concentrations are equivalent between retentate and buffer feed). These can be easily monitored in the retentate by pH or conductivity probes (Rathore et al., 2006). The authors found that the implementation of this PAT-based control strategy for a DF step is technically and economically feasible and would offer several benets, when using an expensive dialtration buffer or when the duration of ltration step has to be reduced due to product-stability concerns.

Control Sitton and Srienc implemented two control strategies triggered by real-time ow cytometry data upon identical

Read et al.: PAT for Biopharmaceutical Products Biotechnology and Bioengineering

281

Table II.

Typical downstream unit operation steps and potential control strategies. Parameters to be controlled Process end points Membrane integrity Peak cutting Gradient elution Column equilibration Column cleaning and regeneration Repacking Flow rate End points Reagent concentration Process end points Monitoring or control strategy pH, conductivity, product concentration, formulation buffer concentration Transmembrane pressure, ux feed/permeate/retentate In-line analytical methods, on-line detectors, as UV, RI System control pH, concentration Mock eluate HETP System control pH, concentration, conductivity Addition control Product quality, reaction time

Unit operation Dialtration/ultraltration Chromatography

In-line mixing Refolding Steps, Conjugation reactions (e.g., PEGylation, antibody-drug conjugation)

Chromatography Currently chromatography steps in biotechnology processes often use pooling criteria that are based on UV absorbance. The key advantage of the pooling by UV absorbance is simplicity; A280 probes are inexpensive, robust, readily available and suitable for intended use. Although pooling by absorbance is simple, absorbance methods are not able to differentiate between the product and other protein impurities, which have similar absorbance proles. Also, they do not distinguish between routine protein charge variants (e.g., C-terminal lysine) and glycoforms (e.g., G0 vs. G1 vs. G2, etc). Therefore, product impurities levels (both process- and product-related) may vary from lot to lot due to variations in the feed conditions and chromatography operating conditions, and typical columns do not separate glycoforms and charge isoforms. Several researchers (Doyle and Netwon, 2002; Rathore et al., 2006, 2007) have used a PAT-based control strategy to assure quality of column pooled product. With on-line analytical liquid chromatography, such as HPLC, continuous monitoring of the chromatography step for aggregate peaks or other undesirable efuent components can be achieved, thus ensuring that product variations are detected mitigating the risk of lot failure. An on-line HPLC system (Dionex DX-800 process analyzer) has been used for examining the real-time pooling of a process chromatography column (Rathore et al., 2007). The system was programmed to collect sample from a side stream directly from the process column to monitor for aggregates. Rathore et al. found that on-line HPLC offers a feasible approach for analysis that can facilitate real-time decisions for column pooling based on product quality attributes (i.e., stop collecting when the aggregate peak starts). Conductivity may also be used as an indicator of when to start and stop peak collection in gradient elution of process chromatography, based on process knowledge (i.e., when undesirable components elute). Similar to the case of dialtration step discussed earlier, column equilibration or cleaning/regeneration can be controlled through the use of surrogates (e.g., pH, UV absorbance, and/or conductivity) and the use of transition analysis to monitor continued

tness for the purpose of the column (e.g., HETP). The key element for PAT applications for process chromatography is to be able to gather information for the process and elicit a timely response to facilitate control.

In-Line Mixing Large-scale pharmaceutical manufacturing operations widely utilize in-line mixing systems to blend large volumes of pre-mixed reagent solutions. Vendors have developed a PAT system with programmable logic controller (PLC) that can be used to monitor the composition of the blend during mixing and prevent release of the blend to the process until its composition accurately matches the set point. This type of PLC-controlled PAT system for in-line mixing enables creation of point-of-use blending for accurate and reproducible mixing of the reagent solutions. For a process chromatography unit operation, a small decrease in variability of the mobile phase by a PAT controlled in-line mixing system or tank blending system can often result in a large improvement in reproducibility, which, in turn, dramatically affects product quality.

PEGylation of Proteins Proteins vary in their stability, solubility, and susceptibility to proteolysis depending on their ionic charge, size, and structure. The intermolecular packing and surface chemistry of proteins determine many of these properties, and may promote protein unfolding, degradation or aggregation, which ultimately leads to loss of their biological activities. PEGylation, (i.e., to link an active protein to a polyethylene glycol (PEG)), is one of the most successful methods for stabilizing proteins and increasing their solubility (Chapman, 2002). In a typical PEGylation reaction, an activated monofunctional PEG is reacted with one or more accessible lysine residues or the N-terminal amino groups of protein. As proteins and PEG are complexed, a mixture of PEGylated protein isomers is, in most case, obtained during PEGylation reaction. Additionally, separation of the

282

Biotechnology and Bioengineering, Vol. 105, No. 2, February 1, 2010

mixture of PEGylated protein isomers is extremely difcult, because the physiochemical characteristics of the species are very similar. It is our opinion that a PAT-based control strategy using an on-line analytical HPLC system to monitor and control the PEGylation reaction is a valuable approach to produce a homogenous/active product.

ux-decay is set to ensure adequate viral clearance at process scale. Although ux-decay is being monitored and not the actual CQA of virus load, this control strategy can be viewed as PAT and used to ensure adequate viral clearance because ux-decay is: (1) tightly correlated with viral clearance and (2) serves as the control strategy to dene the process endpoint. A typical control strategy is outlined in Figure 3.

Case Study: Virus retentive Filter Unit Operation


For direct-ow small virus retentive lters, extended throughputs and overloading may result in a decline of virus titer reduction (Lute et al., 2007). For certain lter types (i.e., Viresolve NFP), the instantaneous clearance capacity is tightly associated with instantaneous ow rate relative to the initial product ow rate (Bolton et al., 2005). For this lter type, this is hypothesized to be due to partial plugging by particulates of the ner end of the membrane pore size distribution. According to this theory, the product ow is diverted through larger pores leading to lower viral titer reductions. It should be noted that this theory may not apply to other small virus retentive lter types. Like all lters, virus retentive lters eventually foul as protein and other particles build up in and on the membrane surface. Membrane fouling leads to a decrease in ux/ow rate under constant pressure mode, which can be monitored during processing. Because of the tight correlation described above, the extent of fouling is considered a critical process parameter for Viresolve NFP lters. The actual relationship between ux decay and viral clearance should be veried for individual process uids in validation studies on a process/product specic basis. Generally, a process endpoint denition of 5060%

Summary
In rst part of this series, we have focused on presenting the concept of PAT and case studies in upstream and downstream processing. The second part focuses on case studies in the areas of dosage form manufacturing, rapid microbiology and chemometrics, as well as discussing how PAT is likely to evolve in biotech processing.
The authors thank Chris Watts, Carla Lankford, Keith Webber, and Ali Afnan for careful review of this manuscript. This work was supported in part by a CDER/FDA Critical Path Grant, project number 1500. Views expressed in the article are those of the authors and not necessarily policy of the Food and Drug Administration or the US Government.

References
Abu-Absi NR, Zamamiri A, Kacmar, J, Balogh SJ, Srienc F. 2003. Automated ow cytometry for acquisition of time-dependent population data. Cytometry A 51(2):8796. Ahmed R, Baseman H, Ferreira J, Genova T, Harclerode W, Hartman J, Kim S, Londeree N, Long M, Miele W, Ramjit T, Raschiatore M, Tomonto C. 2008. PDA survey of quality risk management practices in the pharmaceutical, devices, & biotechnology industries. PDA J Pharm Sci Technol 62:121. Bolton G, Cabatingan M, Rubino M, Lute S, Brorson K, Bailey M. 2005. Normal-ow virus ltration: Detection and assessment of the endpoint in bio-processing. Biotechnol Appl Biochem 42:133142. Brorson K, Krejci S, Lee K, Hamilton E, Stein K, Xu Y. 2003. Bracketed generic inactivation of rodent retroviruses by low pH treatment for monoclonal antibodies and recombinant proteins. Biotechnol Bioeng 82(3):321329. Camisard V, Brienne J-P, Baussart H, Hammann J, Suhr H. 2002. Inline characterization of cell concentration and cell volume in agitated bioreactors using in situ microscopy: Application to volume variation induced by osmotic stress. Biotechnol Bioeng 78(1):7380. Carvell JP, Dowd JE. 2006. On-line measurements and control of viable cell density in cell culture manufacturing processes using radio-frequency impedance. Cytotechnology 50:3548. Chapman AP. 2002. PEGylated antibodies and antibody fragments for improved therapy: A review. Adv Drug Del Rev 54:531545. Clemente JJ, Monteiro SMS, Carrondo MJT, Cuhna AE. 2008. Predicting sporulation events in a bioreactor using and electronic nose. Biotechnol Bioeng 101(3):545552. Doyle MP, Netwon BJ. 2002. Chromatography with On-line HPLC and Ion Chromatography For Process Control. CAST: Jan/Feb 2002, pp. 47. FDA. 1997. Points to consider in the manufacture and testing of monoclonal antibody products for human use. Rockville, MD: U.S. Department of Health and Human Services. Food and Drug Administration, Center for Biologics Evaluation and Research. FDA. 2004. PATA framework for innovative pharmaceutical development, manufacturing, and quality assurance. Rockville, MD: U.S. Department of Health and Human Services. Food and Drug Administration Center for Biologics Evaluation and Research.

Figure 3. Flux decay and viral clearance as a function of volumetric throughput during Viresolve NFP ltration under constant pressure mode. Solid line is instantaneous decay relative to initial ux (the actual real-time measured CPP); dashed line is instantaneous viral clearance (not monitored, but the CQA of concern). A typical process endpoint denition would be 50% ux decay (sometimes referred to as V50).

Read et al.: PAT for Biopharmaceutical Products Biotechnology and Bioengineering

283

Holm-Nielsen JB, Lomborg CJ, Oleskowicz-Popiel P, Esbensen KH. 2008. On-line near infrared monitoring of glycerol-boosted anaerobic digestion processes: Evaluation of process analytical technologies. Biotechnol Bioeng 99(2):302313. ICH. Q9: Quality Risk Management. 2005. Geneva, Switzerland, International Conference on Harmonisation of Technical Requirements for the Registration of Pharmaceuticals for Human Use. ICH. Q10: Pharmaceutical Quality System. 2008a. Geneva, Switzerland, International Conference on Harmonisation of Technical Requirements for the Registration of Pharmaceuticals for Human Use. ICH. Q8(R1): Pharmaceutical, Development. 2008b. Geneva, Switzerland, International Conference on Harmonisation of Technical Requirements for the Registration of Pharmaceuticals for Human Use. Jacquemarte R, Chavane N, Durocher Y, Hoemann C, De Crescenzo G, Jolicouer M. 2008. At-Line Monitoring of Bioreactor Protein Production by Surface Plasmon Resonance. Biotechnol Bioeng 100(1):184 188. Jenzsch M, Gnoth S, Beck M, Kleinschmidt M, Simutis R, Lubbert A. 2006. Open-loop control of the biomass concentration within the growth phase of recombinant protein production processes. J Biotechnol 127: 8494. Joeris K, Frerichs J-G, Konstantinov K, Scheper T. 2002. In-situ microscopy: Online process monitoring of mammalian cell cultures. Cytotechnology 38:129134. Johnson R, Yu O, Kirdar AO, Annamalai A, Ahuja S, Ram K, Rathore AS. 2007. Applications of multivariate data analysis in biotech processing. Biopharm Int 20:130144. Kacmar J, Srienc F. 2005. Dynamics of single cell property distributions in Chinese hamster ovary cell cultures monitored and controlled with automated ow cytometry. J Biotechnol 120:410420. Kirdar AO, Green KD, Rathore AS. 2008. Application of multivariate data analysis for identication and successful resolution of a root cause for a bioprocessing application. Biotechnol Prog 24:720726. Klyushnichenko V. 2004. Capillary electrophoresis in the analysis and monitoring of biotechnological processes. Methods Mol Biol 276: 77120. Koggersbol A. 2005. The role of Process Control in Process Analytical Technology. Presented at: Advances in Process Analytics and Control Technology. Kreij K, Mandenius CF, Clemente JJ, Cunha AE, Monteiro SM, Carrondo MJ, Hesse F, Los Molinas MM, Wagner R, Merten OW, Geny-Fiamma C, Leger W, Wiesinger-Mayr H, Muller D, Katinger H, Martensson P, Bachinger T, Mitrovics J. 2005. On-line detection of microbial contaminations in animal cell reactor cultures using an electronic nose device. Cytotechnology 48:4158. Larson TM, Gawlitzek M, Evans H, Albers U, Cacia J. 2002. Chemometric evaluation of on-line high-pressure liquid chromatography in mammalian cell cultures: Analysis of amino acids and glucose. Biotechnol Bioeng 77:553563. Lepore J, Spavins J. 2008. PQLI design space. J Pharm Innv 3:7987. Lute S, Bailey M, Combs J, Sukumar M, Brorson K. 2007. Phage passage after extended processing in small-virus-retentive lters. Biotechnol Appl Biochem 47:141151.

Mandenius C-F, Brundlin A. 2008. Bioprocess optimization using designof-experiments methodology. Biotechnol Prog 24:11911203. Mukhopadhyay TK, Halliwell D, ODwyer C, Shamlou PA, Levy MS, Allison N, Gorringe A, Reddin KM. 2005. Rapid characterization of outer-membrane proteins in Neisseria lactamica by SELDI-TOF-MS (surface-enhanced laser desorption ionization-time-of-ight MS) for use in a meningococcal vaccine. Biotechnol Appl Biochem 41:175182. Noui L, Hill J, Keay PJ, Wang RY, Smith T, Yeung K, Habib G, Hoare M. 2002. Development of a high resolution UV spectrophotometer for atline monitoring of bioprocesses. Chem Eng Process 41:107114. Nyberg G, Green K, Hashimura Y, Rathore AS. 2008. Modeling of biopharmaceutical processes. Part 1: Microbial and mammalian unit operations. Biopharm Int 21:5665. Rao G, Moreira A, Brorson K. 2008. Disposable bioprocessing: The future has arrived. Biotechnol Bioeng 102(2):348356. Rathore AS. 2009. Roadmap for implementation of quality by design (QbD) for biotechnology products. Trends Biotechnol 27:546553. Rathore AS, Winkle H. 2009. Quality by design for biopharmaceuticals. Nat Biotechnol 27:2634. Rathore AS, Sharma A, Chilin D. 2006. Applying process analytical technology to biotech unit operations. Biopharm Int 19:4857. Rathore AS, Branning R, Cecchini D. 2007. Design space for biotech products. Biopharm Int 20:3640. Schwarz LM, Osborne BA. 1995. Cell death. Methods Cell Biol 46:xvxviii. Scott B, Wilcock A. 2006. Process analytical technology in the pharmaceutical industry: A toolkit for continuous improvement. PDA J Pharm Sci Technol 60(1):1753. Sitton G, Srienc F. 2008. Mammalian cell culture scale-up and fed-batch control using automated ow cytometry. J Biotechnol 135:174 180. Sitton G, Hansgate A, Srienc F. 2006. Transient gene expression in CHO cells monitored with automated ow cytometry. Cytotechnology 52:1324. Teixeira AP, Portugal CA, Carinhas N, Dias JM, Crespo JP, Alves PM, Carrondo MJ, Oliveira R. 2009. In situ 2D uorometry and chemometric monitoring of mammalian cell cultures. Biotechnol Bioeng 102: 10981106. Thomas RJ, Anderson D, Chandra A, Smith NM, Young LE, Williams D, Denning C. 2009. Automated, scalable culture of human embryonic stem cells in feeder-free conditions. Biotechnol Bioeng 102:1636 1644. Tridaphillou S, Martin E, Montague G, Norden A, Jeffkins P, Stimpson S. 2007. Fermentation process tracking through enhanced spectral calibration modeling. Biotechnol Bioeng 97(3):554567. Ulber R, Protsch C, Solle D, Hitzmann B, Willke B, Faurie R, Scheper T. 2001. Use of bioanalytical systems for the improvement of industrial tryptophan production. Chem Eng Technol 24:1517. Veale EL, Irudayaraj J, Demirci A. 2007. An on-line approach to monitor ethanol fermentation using FTIR spectroscopy. Biotechnol Prog 23(2): 494500. Yigzaw Y, Piper R, Tran M, Shukla AA. 2006. Exploitation of the adsorptive properties of depth lters for host cell protein removal during monoclonal antibody purication. Biotechnol Prog 22:288296.

284

Biotechnology and Bioengineering, Vol. 105, No. 2, February 1, 2010

Anda mungkin juga menyukai