Current Status of Analytical Techniques for Characterization of Protein Stability

Abstract
A major goal of pharmaceutical development is to characterize pathways of chemical and physical instability and develop strategies to minimize them. The potential for the presence of multiple degradation products in protein-based pharmaceuticals highlights done proofreading before a need for analytical methods capable of reliably and accurately identifying and measuring those variants. Our review aims to provide a comprehensive overview of the current state of analytical methods for the detection and characterization of protein aggregation, particulate formation, and chemical degradation.

Introduction
Biotherapeutics are large macromolecules derived from biological sources using recombinant DNA technology. Therapeutic proteins, especially monoclonal antibodies (mAbs) and antibody-related products, such as Fc-fusion proteins, antibody fragments, and antibody-drug conjugates (ADCs) have grown to become a dominant product class within the biotechnology market1. At the current approval rate of five to seven new products per year, approximately 70 monoclonal antibody products will be on market by 2020, with combined world-wide sales approaching $125 billion1.
Most globular proteins are marginally stable macromolecules2, 3. Intrinsic factors such as presence of aggregation prone regions and disulfide bonds, and extrinsic factors, such as temperature, solution conditions and composition (excipient types and levels, peroxides, metal impurities, ionic strength, extractables and leachables), processing steps (fermentation, purification, shaking, freeze-thaw, shear, interfaces, formulation/filling) can potentially impact protein stability4. Protein instability has been increasingly recognized as a problem limiting the efficacy and shelf-life of protein therapeutics.
Protein degradation pathways are complex5, 6. Protein aggregation is the most common, complex and varied physical instability pathway7. Aggregation typically involves irreversible association of partially unfolded monomeric units of proteins to form higher order structures, such as dimers and trimers8. These protein aggregates can be either soluble or insoluble. Soluble protein aggregates can be characterized by techniques such as size exclusion chromatography, analytical ultracentrifugation (SEC), field flow fractionation (FFF), light scattering, differential scanning calorimetry (DSC), fluorescence dye binding (using dyes such Thioflavin T or Congo Red), and NMR5, 9. The most recent advances in this area, including the use of methods orthogonal to SEC, are discussed below.
On the other hand, insoluble aggregates can be categorized as subvisible particles (SVPs) and visible particles (VPs). Techniques such as light obscuration, flow imaging, coulter counter, fluorescence microscopy, and visual inspection are routinely used to characterize SVPs and VPs10, 11. The term “aggregation” will be used here to refer to soluble aggregates whereas “particulates” will be used in reference to comprise of subvisible particles (100 nm – 100 µm) and macroscopic insoluble aggregates appearing as visible particles (> 100 µm).
Chemical modifications or degradation involves the covalent modifications of proteins. Residue-specific modifications such as oxidation, deamidation, isomerization, disulfide shuffling, pyroglutamic acid formation and post-translational modifications such as glycosylation (i.e. galactosylation, sialylation, fucosylation), oxidation, lipidation, sulfation, deamidation come under the category of chemical degradation11, 12. However, this review will only discuss analytical methods used to characterize oxidation and deamidation products.
This review article presents an overview of the current state of analytical techniques employed for characterizing protein aggregation, particulate formation, and chemical degradation. In particular, new and emerging analytical characterization tools and recent improvements to existing techniques will be highlighted. As the literature is already rich with general articles on protein characterization techniques10, 11, 13-15, this review will focus primarily on advances published in the last three to four years.

Analytical Tools for Characterizing Particulates
Figure 1 provides an overview of analytical techniques that can be potentially used for analysis of aggregates, SVPs, and VPs. Appearance of particulates and observable foreign particles have long been recognized as a critical quality attribute (CQA) for injectable protein formulations16, 17. Visual inspection is the industry standard for detecting VPs in parenteral formulations and is routinely performed as part of drug product release7, 18-20. Nephelometry and turbidimetry are both light scattering-based methods that are used in drug product development to measure opalescence and detect relative changes in particulate levels21. In the subvisible particle (SVP) region, particles greater than 10 µm and 25 µm have historically received attention because of the pharmacopeial requirements for parenteral products22, 23. Currently, the compendial methods use light obscuration (LO) and microscopy to measure SVPs in parenteral products22, 23. The HIAC system is the industry standard for LO and has been used to quantify particles in the size range of 2 µm and larger13, 24. Quiroz and colleagues implemented a modified light obscuration sensor for particle characterization and demonstrated that this modification resulted in improved detection of particles smaller than 2 µm and reduction of particle counting artifacts25.

Flow imaging analysis (also called flow microscopy or dynamic flow imaging) method utilizes light guided detection of particles as the fluid passes through a flow-cell. Digital visualization of particles enables detailed analysis of size, shape, transparency, and compactness of particles suspended in solution and provides information about size and particle count in the 1-100 µm range. This can enable differentiation between protein and non-proteinaceous particles such as silicone oil, air bubbles, excipient degradation products, particles shed from pumps or primary packaging material13, 14, 26-31. Numerous articles discussing the application of flow imaging microscopy to characterizing SVPs in parenteral products are available32-38. A comparative evaluation of the two commonly used flow imaging systems (MFI and FlowCAM) has been performed by several groups39-42. In general, most light-based particulate measurement techniques have lower precision in the subvisible range primarily due to low sampling volumes and the corresponding large extrapolation factors25, 43, 44. Developing optimal imaging methods31 and application of precise software filters for data analysis is therefore critical for accurately counting the number of particles and differentiating different types of particles. In the last couple years, several novel software filters such as aspect ratio-based filter45, variable threshold algorithms (determination of boundaries of imaged SVPs)46, four image morphology algorithm (discrimination between proteinaceous particles and silicone oil)47, and improved local pixel intensity variance (LPIV) algorithms (detecting translucent particles)48 have been developed and have resulted in a dramatic reduction of particle characterization artifacts. In their work, Ripple and Hu demonstrated that it is possible to correct the relative biases and get good quantitative agreement between MFI, FlowCAM, and LO instruments by introducing bias corrections49.
More recently, focus has moved to submicron SVPs. An FDA guidance document recommends that assessment should be made of the range and levels of submicron SVPs (0.1 – 1.0 µm) present in therapeutic products50. For this reason, nanoparticle tracking analysis (NTA) was developed for the characterization of particles in the nanometer size range (range: 40 nm – 1 µm) and provides quantitative information on the particle size and semiquantitative information on particle concentration51. In NTA, samples are illuminated by a laser, particle movement is logged via light scattering by a CCD camera, while software tracks the particles moving under Brownian motion. In contrast to DLS, where bulk intensity changes are measured, NTA tracks particles individually, which, in theory, enables distinction of particle subpopulation51. Recent publications have evaluated NTA for the measurement of protein aggregates and particulates52-58. These articles have reported on factors that could potentially improve NTA data quality and reproducibility53, 57 and have provided recommendations on instrument settings and sample preparations to properly characterize protein samples54. Interference from proteins and surfactants on particle size distribution during NTA has also been reported58. It should also be noted that NTA evaluates only a small fraction of the total sample population, requiring extensive extrapolations. For these reasons, and the sensitivity to data collection parameters, one must be careful in the interpretation of NTA data58.
Another promising technology for particle analysis that is still maturing is Resonant Mass Measurement (RMM). In RMM, particles are drawn through a MicroElectroMechanical System (MEMS) microfluidic channel which resonates mechanically. The presence of a particle in the sensor changes the total mass and shifts the sensor resonant frequency. From these mass measurements, the corresponding size distributions are derived by translating the mass of each particle to equivalent circular diameter using the density of the particles measured. Size-range for RMM depends on the density of the analyzed particles and is approximately 300 nm to 5 µm for proteinaceous particles and 500 nm to 5 µm for silicone oil droplets. RMM is also capable of providing information on sample concentration, viscosity, polydispersity, density and volume and distinguishing between negatively buoyant proteinaceous particles and positively buoyant silicone oil droplets. Several articles comparing the performance of RMM relative to other particle sizing techniques such as DLS, MFI, LO have reported that RMM is an effective complementary technique capable of providing accurate particle size measurements59, 60, determining particle density61, and differentiating between protein particles and silicone oil41. In fact, RMM has been successfully applied to discriminate silicone oil from protein particles in several commercial prefilled syringe (PFS) products62-66.
Flow cytometry has also been used to distinguish silicone oil droplets from fluorescently-labeled protein particulates67. Flow cytometry system equipped with light and scattering detectors has been successfully used to characterize SVPs68. In a recent report, Nishi et al. evaluated the capability of the flow cytometry (FACS) method to detect and count SVPs and benchmarked against conventional techniques such as LO and MFI69. They reported that FACS can be used to monitor SVPs down to 500 nm and that particle counts from FACS were proportional to LO and MFI. Preparative FACS has also been successfully used to fractionate microscopic proteinaceous particles based on their size70.
Taylor dispersion analysis (TDA),which determines the hydrodynamic size of particles based on their diffusion coefficient, is another novel method for characterizing protein particles14. In TDA, the diffusion coefficient is based on band broadening of the flow profile which is detected by UV signal. Findings by Hawe et al. suggested that TDA allowed accurate determination of hydrodynamic radius over a wide concentration range (0.5 to 50 mg/mL), with little interference from excipients71. Hulse and colleagues have developed quick and low-volume TDA method that is capable of measuring hydrodynamic radius with a standard deviation of less than 5%72, 73. Summary of all recently published articles discussing application of TDA for measuring diffusion constants and/or hydrodynamic radius in protein formulations is provided here74-78.
In the last three to four years, several novel techniques have been successfully implemented to characterize protein aggregates and particles. Quantitative Laser Diffraction (qLD), which is an extension of laser diffraction method, is a recently developed technique that has been successfully applied for simultaneously assessing the concentration distributions of protein particles with diameters in the 0.2 – 10 µm range79. Water protein NMR (wNMR) is another tool that has been tested to characterize protein aggregates and particles80. The authors demonstrated that wNMR outperformed SEC, DLS, and MFI in that it is most sensitive to increases in both soluble and insoluble aggregates, including SVPs. In a recent article, Patil et al. have discussed the application of Diffusion Ordered SectroscoY (DOSY) NMR to measure diffusion coefficients in formulated drug product and suggested that this technique could be used to infer particle size distribution81.
The number of analytical techniques for the quantification and characterization of protein particles has continuously increased during the last few years. Numerous characteristics such as size, shape, chemical composition, and structure can be determined based on different measurement principles. However, no single method is capable of providing information on all desired parameters for the entire size range. This necessitates a combination of several methods based on different measurement principles for a comprehensive characterization.

Detection and Quantitation of Soluble Aggregates
By far and away, the most common method for detection and quantitation of soluble aggregates of therapeutic proteins has been SEC. In fact, an SEC method is usually found in nearly every regulatory filing for a protein therapeutic. However, the potential shortcomings of SEC, which have been examined in some detail82, 83, has led regulatory agencies to request the use of orthogonal methods for determining the exact extent of aggregation in drug products84, 85. Specific aspects of SEC methodology have been identified as being problematic82, ranging from the composition of the mobile phase86 to the size and type of vial used in SEC87. Despite these limitations, one can still obtain reliable estimates of soluble aggregate levels using SEC88.
Of all of the orthogonal methods available to researchers, two have received the most attention: analytical ultracentrifugation (AUC)9, 89 and asymmetrical flow field-flow fractionation (AF4)90. In both of these cases, advances in methodology and data analysis continue to appear, thereby providing more reliable estimates of soluble aggregate contents. For example, the importance of rotor calibration, maintenance, and alignment has long been recognized9, 91, 92. A recent article revisits this subject, showing that improper alignment can be highly detrimental for quantitation of aggregates using sedimentation velocity (SV)-AUC [93]. Improvements in data analysis software can also be beneficial94. As these techniques become more well refined, their ability to function as orthogonal methods becomes clearer95. In fact, articles have appeared showing that AF4 can be reliably used to monitor antibody aggregation96 and even validated as a stability-indicating assay97.
Mass spectrometry (MS) has emerged as a tool for detecting oligomeric states of proteins. Most often, MS is performed under strongly denaturing conditions which are well suited to the characterization of covalent protein structure and is readily able to detect modifications that result in changes in mass. However, the mass to charge ratio detected by MS also appears to be sensitive to the association state of the protein. More recently it has been demonstrated the analysis under nondenaturing conditions is possible allowing the successful characterization of noncovalent aggregated species98-100. While these approaches work well for detecting and quantifying oligomers (dimer, trimers, etc.), they lack sensitivity for larger higher molecular weight species. The lack of success in detecting higher oligomers is however not due to a fundamental limitation and with further refinement in experimental techniques and advances in instrumentation it is reasonable to expect that the characterization of larger aggregates by MS will be possible.
Other techniques have been reported and may present advantages for characterization of soluble aggregates. These include gas-phase electrophoretic mass measurement analysis (GEMMA) or differential mass analysis (DMA)101-103. In this technique, dilute protein solutions are aerosolized and the encapsulated protein species is analyzed using a mass analysis system. It provides direct mass analysis of aggregates, but requires low protein concentrations and volatile buffers. Still, it has been touted as an orthogonal method to SEC102. Other electrophoretic methods have recently been promoted for their ability to monitor aggregation, as with native gel electrophoresis104, which was used to monitor antibody aggregation.
Surface plasmon resonance (SPR), often used to measure antigen-antibody binding, has been adapted to monitor aggregation of proteins and peptides105-108. This includes a study comparing SPR with methods like DLS and AF4 for following the aggregation of antibody and antibody fragments105. Studies like these show methods that are used for one purpose and can be redirected to provide information regarding aggregation and association of protein molecules.

Chemical Instability
There are several main pathways by which the primary structure of therapeutic antibodies can be changed during manufacturing and storage. Among them are oxidation and hydrolytic reactions (including deamidation and isomerization) are the most common109-113. The extent of chemical degradation, which can impact the efficacy and immunogenicity of therapeutic proteins114, generally depends on the formulation and storage conditions, as well as the intrinsic properties of the protein115. Among the techniques that have traditionally been used to characterize these kinds of chemical modifications are ion-exchange chromatography, electrophoretic methods, mass spectrometry, and spectroscopic methods116. Most of these methods analyze the intact protein. However, for larger proteins, such as monoclonal antibodies (mAbs), there have been advances in the use of these techniques in combination with preparative methods enabling one to analyze smaller domains. These types of developments will be a primary focus of this section.
Oxidation. Oxidative processes are a particular concern for vulnerable amino acid residues with aromatic or sulfur-containing side chains115, 117, 118. When oxidized, they can have a significant impact on protein function114. At the screening level, a fluorescence method has been demonstrated that detects oxidation at the whole protein level based on derivatization of Tyr and Phe residues119. However, traditionally, mass spectrometry has been the method of choice for monitoring oxidation of proteins120, but site specificity is often lacking when analyzing intact proteins.
Where greater depth of information is needed, specificity and sensitivity must be balanced with throughput in order to distinguish oxidation products that may be indicative of particular reaction pathways120. Peptide mapping is an established approach that typically involves proteolytic digestion of the protein of interest, separation of the fragments by reversed phase HPLC, and analysis by electrospray mass spectrometry. Proteolytic digestion is often time-intensive121 and also potentially prone to artifacts122-124, thereby limiting throughput and accuracy. The addition of organic solvents has been explored as a means of improving the rate of digestion125, 126. On-line digestion may be another alternative approach to minimize these issues127.
For monoclonal antibodies, the discovery of the protease, IdeS (immunoglobulin-G degrading enzyme from S. pyogenes), provides a relatively fast and robust way to cleave human IgG1 antibodies with good specificity under aqueous condition, yielding Fab, Fc, and Fc/2 fragments128. This enzyme also is known as FABricator in some publications. With a well-defined mixture of fragments, the researcher is better equipped not only to examine oxidation in a site-specific manner, but also to analyze charge heterogeneity caused by hydrolytic reactions, as discussed below. This approach has been demonstrated for the characterization of domain-level protein hot spots with respect to oxidation susceptibility and impact on protein function129-131. The throughput advantages of the IdeS protease have also been extended with the incorporation of UHPLC separation methods132, further increasing the potential for practical and site-specific monitoring of oxidation.
Finally, hydrogen-deuterium exchange mass spectrometry (HDX-MS) has, in recent years, found wider use for investigating the dynamics of oxidation and other degradation reactions133. Although this method is experimentally demanding, a number of studies have employed HDX-MS to provide insights into oxidation in the complementarity-determining regions (CDRs) of mAbs134, the influence of Fc oxidation on mAb conformation135, and the mechanism of His photo-oxidation136. Native mass spectrometry has also been used to investigate Met oxidation in IgG1 mAbs137, 138.
Deamidation. Another common post-translational modification is deamidation, wherein Asn (or less often, Gln) residues undergo hydrolysis to form Asp-based species. The rate of this reaction is influenced by a number of factors, both intrinsic and extrinsic139-141. The change in local charge associated with deamidation can disrupt protein conformation and function. Chromatographic methods, including ion exchange and reversed phase chromatography, represent two common techniques for separating these different charged species. Use of these techniques in analysis of deamidation (and other charge variants) is widespread140. In material-limited situations, it may be possible to use the stationary phase in small-scale capillary separations142.
When RP and IEX chromatography are combined with other techniques, they enable subunit analysis of protein deamidation132, 143, 144. By yielding site-specific information about chemical reaction sites on the protein, it also becomes possible to identify critical quality attributes to be monitored in development and manufacturing145.
Capillary electrophoresis represents a second group of methods that has been applied to charged-based analysis of proteins and peptides146-148. However, with these techniques, separation performance is hindered by protein adsorption to the capillary walls. Coatings to address this issue were the focus of a recent review149 and the concept has been demonstrated for the separation and characterization of antibody charge variants150. Another limitation arises when CE is operated with only UV absorbance detection. Specifically, the small dimensions of most capillaries limit the optical path length. For this reason, alternative detection methods, such as laser-induced fluorescence (LIF) and MS, have been highlighted as more sensitive alternatives151. Increasing the optical path length by increasing column cross-section is generally not feasible for CE, which is important when larger-scale separation is desired, for example, in the isolation and downstream analysis of a particular charge variant. To this end, free-flow IEF152 and immobilized pH gradient (IPG)-IEF153, 154 have been applied as preparative methods in the characterization of post-translational modifications, including deamidation, in antibodies.
By contrast, there may be other situations where it is advantageous to shrink the dimensions of CE systems even further as with microchip CE systems. Developments in this area have been reviewed155, 156, and in one case, a microchip IEF system with UV detection was reported to show a several-fold throughput improvement over conventional two-step cIEF157. The resolving power of CE is often combined with the sensitivity of MS for site-specific analysis of charge heterogeneity. In recent years, the use of a sheathless capillary interface has enabled lower flow rates at the capillary outlet, improving ionization efficiency and thus allowing greater sensitivity.156, 158 This strategy has been used for the analysis of peptides159 and mAbs160-162.
Finally, efforts to characterize charge-based protein species with higher throughput have leveraged the advantages of plate-based assays, and a bioluminescent assay for detecting deamidation (as well as ubiquitination) was recently reported with turnaround times on the order of hours163, 164. The method was validated against an established HPLC protocol and demonstrated for two control proteins.

Comparability and Higher Order Structure (HOS)
The use of various analytical methods to establish comparability, especially in the context of higher order (secondary, tertiary, quaternary) structure (HOS), has been examined in some detail of late165-168. Primarily, these studies use spectroscopic methods, such as infrared spectroscopy169 or circular dichroism (CD) spectroscopy170, but there has been a number of recent articles employing hydrogen-deuterium exchange mass spectrometry (HX-MS)171-175. Often, HX-MS is used in conjunction with other techniques, such as in recent studies that employed SEC and ion mobility spectroscopy (IMS)-mass spectrometry176, 177.
Once a set of analytical data is obtained, whether from separation techniques or spectroscopy, it is essential to employ quantitative metrics for assessing similarity or comparability. These may involve the calculation of a single comparator, as with area of overlap for spectra169, 170, or one may use chemometric approaches178. With these metrics in hand, one can use them to make decisions about the similarity of two drug products or lots179-181. In this field, the evaluation and processing of the analytical data are critical to obtaining usable information regarding comparability. The nature of the analytical method becomes less important in this context.

Interrelationship of Chemical and Physical Instability
Although it is convenient to segregate chemical degradation from physical instability processes, the reality is these can be highly interrelated. In the case of deamidation at Asn residues, there are a number of examples where deamidation can facilitate or accelerate aggregation182-184. At the same time, the literature contains case studies where deamidation actually slows the aggregation process99, 185. In these latter cases, it may be that the increased negative charge that is accumulated via deamidation (conversion of an Asn residue to various Asp species) provides an increased electrostatic repulsion that inhibits aggregate growth. In one system, the deamidation can either increase or decrease aggregation rate, depending on the actual deamidation site186.
Likewise, there are examples of oxidation affecting aggregation rates. For Aß peptide, Met oxidation appears to increase fibrillation rates187. Similarly, MAP kinase 4 also displays increased aggregation upon oxidation188. The same is seen for SOD189. Carbonylation, another manifestation of oxidation, has also been shown to promote aggregation190, 191. Not only can chemical degradation impact aggregation behavior, there are examples of aggregates undergoing chemical modification after the aggregates are formed192, 193.
One final facet of these studies should be noted. Many of them used SDS-PAGE to monitor the aggregation process188, 190, 191. More often, aggregation kinetics are monitored using SEC185, 192, 194 or extrinsic fluorescence183, 184, 186, 187. Certainly, SDS-PAGE is a well established method for detecting aggregates, although there can be issues with quantitation due to the limited linear range of most stains. In addition, the throughput is relatively low for SDS-PAGE. Still, the ability to distinguish covalent vs. non-covalent aggregates makes it a method still in widespread use today.

Acknowledgements
The authors thank Monica Adams, Mark Bolgar, and Madhushree Gokhale, all of Bristol-Myers Squibb Co., for helpful comments and suggestions.



Subscribe to our FREE newsletter

We will not sell or give your information to a third party. See our Privacy Policy