Article pubs.acs.org/ac
Deep-Ultraviolet Resonance Raman (DUVRR) Spectroscopy of Therapeutic Monoclonal Antibodies Subjected to Thermal Stress Justin Bueno,* Dianna Long, John F. Kauffman, and Sergey Arzhantsev Division of Pharmaceutical Analysis, Office of Testing and Research, Center for Drug Evaluation and Research, Food and Drug Administration, 645 S. Newstead Avenue, St Louis, Missouri 63110, United States S Supporting Information *
ABSTRACT: The structural assessment of Rituximab, an IgG1 mAb, was investigated with deep-ultraviolet resonance Raman (DUVRR) spectroscopy. DUVRR spectroscopy was used to monitor the changes to the secondary structure of Rituximab under thermal stress. DUVRR spectra showed obvious changes from 22 to 72 °C. Specifically, changes in the amide I vibrational mode were assigned to an increase in unordered structure (random coil). Structural changes in samples heated to 72 °C were related to loss in drug potency via a complement dependent cytotoxicity (CDC) bioassay. The DUVRR spectroscopic method shows promise as a tool for the quality assessment of mAb drug products and would represent an improvement over current methodology in terms of analysis time and sample preparation. To determine the scope of the method, protein pharmaceuticals of different molecular weights (ranging from 4 to 143 kDa) and secondary structure (β-sheet, α-helix and unordered structure) were analyzed. The model illustrated the method’s sensitivity for the analysis of protein drug products of different secondary structure. Results show promise for DUVRR spectroscopy as a rapid screening tool of a variety of formulated protein pharmaceuticals.
M
Circular dichroism (CD), Raman and Fourier transform infrared (FTIR) spectroscopies are also well established techniques for the determination of protein secondary structures.7,8 Recently, Raman, FTIR and CD have been investigated to monitor aggregation of mAb therapies caused by thermal stress. Thiagarajan et al. applied Raman spectroscopy and Raman optical activity (ROA) to the analysis of a thermally stressed IgG4 mAb.9 The use of a nonresonant excitation (532 nm) resulted in a relatively high concentration requirement and spectral accumulation times of 48 h. Additionally, spectra had large contributions from amino acid side chains, excipients and fluorescence. Spectral subtraction and multivariate analysis were required to illustrate the changes in mAb secondary (Raman) and tertiary (ROA) structures for samples subjected to heat stress over a 1 month period. Joshi et al. found that CD data was in agreement with SEC when monitoring mAb aggregation caused by thermal stress.10 Andersen et al. correlated aggregation of the therapeutic mAb Rituximab, to the unfolding of specific mAb subdomains at different temperatures using CD and SEC.11 FTIR microscopy was used to illustrate differences in spectra collected from a native mAb therapy in solution and aggregated mAb particles after exposure to 80 °C for 20 min.12 Thermal stability of therapeutic mAbs was also monitored with attenuated total reflectance (ATR)-FTIR.13 mAbs in solution at 1 mg/mL provided very weak ATR-FTIR signal, but as
onoclonal antibody (mAb) pharmaceuticals are a vital class of patient therapies marketed to treat an increasing range of disorders. In 2012, sales of mAb therapies grossed over $24B in the United States, illustrating both the large market and inherent effectiveness of these drug products.1 Because of the complexity of protein therapeutics, numerous quality attributes need to be considered to ensure product quality. Compared to other quality attributes, characterization of formulated mAb secondary structure is relatively underdeveloped. Although thermal unfolding can lead to aggregation causing an increase in immunogenicity,2,3 the relationship between changes to protein secondary structure under thermal stress and pharmaceutical function has previously been neglected. The techniques for high resolution (atomic level) determination of protein secondary structure, nuclear magic resonance (NMR) spectroscopy and X-ray crystallography, are not suitable for rapid quality assessment of mAb drug products because they are time-consuming and labor intensive. Specifically, NMR of high molecular weight proteins (such as mAbs) requires complicated isotopic labeling. Recent advances have been made in 2D NMR measurements of mAbs, although this approach is still relatively time-consuming.4 X-ray diffraction requires protein crystals, which are unobtainable when working with a formulated product. Techniques, such as size exclusion chromatography (SEC), to monitor protein aggregation resulting from thermal stress are well established.5,6 However, the relationship between thermal unfolding and changes to mAb function require further research. This article not subject to U.S. Copyright. Published 2015 by the American Chemical Society
Received: April 28, 2015 Accepted: July 1, 2015 Published: July 1, 2015 7880
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
Article
and size). Next, Rituximab, the active pharmaceutical ingredient (an IgG1 therapeutic mAb from the drug product Rituxan), was subjected to thermal stress and the resulting changes in the DUVRR spectra were monitored. Rituxan is marketed to treat non-Hodgkin’s lymphoma, leukemia, rheumatoid arthritis and granulomatosis with polyangiltis, and grossed >$3B in the United States in 2012.1 In its native state, Rituximab is composed of mostly β-sheet conformations.22 Unfolding of Rituximab was observed in DUVRR spectra as the formulated product was heated from 22 to 72 °C. Changes in the DUVRR spectra were assigned to an increase in unordered structure (random coil) determined by literature assignments18 and qualitative comparison to an unordered structure protein model. The structural assessment determined by DUVRR spectroscopy was correlated to loss of mAb function via a complement dependent cytotoxicity (CDC) bioassay. The fully development method represents a novel and rapid approach to determination of mAb product quality via secondary structure characterization.
precipitation occurred with increasing temperature, the protein signal increased due to an increase in the local concentration around the ATR crystal. However, pitfalls remain for these analyses when targeting drug products. The application of CD for analysis of mAb drug products is limited due to the numerous buffers and excipients that are not transparent in the far-UV region (170−250 nm). Signals from these components often obscure protein signal in CD spectra.14 CD also suffers from relatively weak specificity, as illustrated by the presence of one negative band at approximately 218 nm (for β-sheet protein). FTIR suffers from interferences from water and ATRFTIR requires mAb precipitation to increase the local concentration around the crystal (in horizontal mode). Recent research utilizing CD and FTIR on therapeutic mAbs has focused on measuring precipitates resulting from aggregation. Consequently, there has been very little research on the changes to the secondary structure of mAbs under thermal stress and its impact on drug function. Deep-ultraviolet resonance Raman (DUVRR) spectroscopy is uniquely capable of structural assessment for therapeutic mAb drug products. Compared to nonresonant Raman, the use of a UV excitation can dramatically increase the sensitivity of the method for protein analysis. Structural assessment of formulated mAb pharmaceuticals via nonresonant Raman is problematic due to interferences from excipients and protein fluorescence that can mask signal from the amide backbone. To overcome these drawbacks, additional processing and multivariate analysis procedures are required to extract information from the spectra.9 The enhanced sensitivity when a UV excitation is utilized removes the need for these processing steps. The enhanced sensitivity originates from the fourthpower dependence of scattering efficiency as the frequency of light increases (ν4), which greatly enhances Raman signal when a UV excitation is used as compared to visible light. In addition, a resonance enhancement of ∼106−108 is expected when the excitation is resonant with the π → π* transition of the protein amide backbone. These enhancements allow for analysis of mAbs in solution at a concentration of approximately 1 mg/ mL. Furthermore, fluorescence occurs at longer wavelengths, well removed from the Stokes scattered radiation frequencies, and nonresonant Raman contributions to the signal from excipients are extremely weak compared to the resonant protein signal. The technique is sensitive to the Φ/Ψ angles of the peptide bonds that are unique for different secondary structures. Previous research has illustrated the effectiveness of deep-ultraviolet resonance Raman (DUVRR) spectroscopy for determining protein secondary structure.15−19 Additionally, information about protein tertiary structure can be obtained from changes to the aromatic amino acid contributions in DUVRR spectra.20 DUVRR spectroscopy has been applied to the quality assessment of protein pharmaceuticals on a limited basis,21 and has not been applied to the analysis of mAb drug products. Here, three different diluted drug products (including the mAb Rituximab) were discerned with DUVRR spectroscopy based upon secondary structure. The work described here investigates the use of DUVRR spectroscopy as a quality assessment tool of protein pharmaceuticals. Three diluted drug products of varying secondary structure and molecular weights were subjected to DUVRR spectroscopic analysis. The qualitative models that were developed indicated that the method is applicable on a range of protein pharmaceuticals independent of excipients, and the physical properties of the protein (i.e., secondary structure
■
MATERIALS AND METHODS Rituxan Samples. Three lots of Rituxan (Genentech, San Francisco, CA and Biogen Idec, Cambridge, MA) were purchased from Bradly Drugs (Bethesda, MD). The active ingredient in Rituxan is Rituximab, an IgG1κ anti-CD20 mAb. Concentrations of the drug product include 10 mg/mL (0.07 mM) Rituximab, 150 mM NaCl, 25 mM citrate buffer and 0.5 mM of the excipient polysorbate 80. For DUVRR analysis, Rituxan was diluted to 1 mg/mL Rituximab in 25 mM citrate buffer (Sigma-Aldrich, St. Louis, MO) and 150 mM NaCl (Acros Organics, NJ, USA). For analysis of samples at room temperature, dilution was the only sample preparation required. The concentration of the excipient polysorbate 80 was not adjusted after dilution; however, the ratio of protein to polysorbate 80 concentration before and after dilution remained constant. To avoid interference from aggregates during thermal stress investigations, Rituximab samples at pH 1.9 (adjusted by HCl) were utilized. The drug product Epogen was diluted to a concentration of albumin of 1 mg/mL in 25 mM citrate buffer and 150 mM NaCl. The drug product Protamine Sulfate was diluted in 100 mM sodium phosphate (Fisher Scientific, Fair Lawn, NJ) buffer. DUVRR Spectroscopy. The DUVRR instrumental setup has been described in detail previously.21 The 205 nm Raman excitation was produced through fourth harmonic generation (FHG) of the fundamental laser (820 nm). The laser power on the sample ranged from 1.5 to 2.0 mW. The CCD camera was controlled by WinSpec/32 2.5.23.0 software (Princeton Instruments, Trenton, NJ). Spectra of protein were collected with two 600 s accumulations, over a spectral range of 656− 2566 cm−1 and the use of the WinSpec temporal cosmic ray removal algorithm. The spectrograph was controlled with SynerJY V3.1.5.11 software (Horiba Jobin Yvon, Kyoto, Japan). Calibration of the spectrograph was performed daily based upon a five point linear fit of the Raman bands from the analysis of neat cyclohexane (Millipore, Billerica, MA). Sample temperature was controlled by a Flash 300 temperature controlled cuvette holder and regulated through T-App 1.32 software (Quantum Northwest, Liberty Lake, WA). Temperature points (±0.02 °C precision) of 22, 30, 40, 50, 60 and 72 °C were utilized during sample acquisition. For each lot of Rituxan, three samples were analyzed from 22 to 72 °C. 7881
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
Article
(10 mg/mL) are possible, it was found that the optimum protein concentration for high quality Raman spectra was 1 mg/mL. The three drug products were diluted to 1 mg/mL protein concentrations in the buffers used in the drug products; no other sample preparation was performed. Figure 1 illustrates
DUVRR analysis was performed under stirring (to prevent photodegradation of the sample), with 2 mL of sample in quartz cuvettes of 1 cm path length. DUVRR spectra were imported into MATLAB R2013a (MathWorks, Natick, MA) with the PLS_Toolbox 7.3.1 (Eignvector Research, Wenatchee, WA) for data treatment. An internally developed MATLAB toolbox was used to import the spectra and calibrate the Raman scale (x-axis). The TrendTool feature within the PLS_Toolbox was used to identify the peak locations/band centers of the Raman spectra. TrendTool sensitivity was set to identify all peaks in the Raman spectra. Raman spectra were truncated (spectra in Figure 3 have a spectral range of 1157−1706 cm−1), smoothed (Savitzky−Golay, five point filter width, zero order polynomial), baseline corrected (weighted least-squares, second order and weight mode) and normalized (total area). CDC Bioassay. Rituxan was diluted to 1 mg/mL protein concentration and heated at 72 °C for 30 min. Centrifuging was used to remove aggregates. Protein concentration after the removal of aggregates was determined by UV absorbance at 280 nm (ε = 229 220 M−1 cm−1), via an Agilent 8453 UV−Vis spectrometer. Rituxan was diluted to a concentration range of 0.0038−250 μg/mL. The CDC assay was performed in a 96well format (Corning Inc., Corning, NY). Each plate contained triplicates of a heated sample and an unheated sample of Rituximab. A high concentration IgG1 isotype control (Infliximab) and a maximum cell death control of 10% SDS were also included in triplicate. A no-drug control well and a background control well, containing only assay medium, were included for each row. Ramos cells (Genentech, American type culture collection number: CRL-1596) were grown in RPMI-1640 medium with glutamine (Hyclone Laboratories Inc., South Logan, Utah) containing 10% heat inactivated fetal bovine serum (Hyclone), 50-50 Pen-Strep (Gibco, Washington, DC), and 1× 2mercaptoethanol (Gibco). Before use, Ramos cells were pelleted and resuspended at 2.0 × 106 cells/mL in assay medium containing RPMI-1640 with glutamine, 0.1% bovine serum albumin (Sigma-Aldrich), and 50-50 Pen-Strep. Ramos cells were added to each well at 1.0 × 105 cells/well and incubated at 37 °C, 5% CO2 for 15 min to allow antibody to bind to cells. Pooled complement human serum (Innovative Research, Novi, MI) was diluted and added at 10% of the final well volume. Plates were incubated at 37 °C, 5% CO2 for 60 min to allow for complement-antibody binding and subsequent cell death. To monitor cell viability, 20 μL CellTiter-Blue resazurin dye (Promega, Madison, WI) was added to each well and plates were incubated at 37 °C, 5% CO2 for 12−16 h, allowing living cells to convert the dye to a fluorescent end product. Plates were read using a BioTek Cytation3 plate reader set for fluorescence (560 nm excitation/590 nm emission). Percent viability was calculated for each data point using the no-drug control as a reference for 100% viability. Curves were determined using a four-parameter sigmoid fit and analyzed using an internally developed Excel plugin.
Figure 1. DUVRR spectra of protein pharmaceuticals. Each drug product represents a model for different protein pharmaceutical secondary structure. Peak locations are indicated for amide III, II and I vibrational modes. Spectra were collected at room temperature. Drug products were diluted to 1 mg/mL concentration of protein.
the resulting DUVRR spectra from three different diluted drug products, Protamine Sulfate, Epogen and Rituxan. The spectra were preprocessed as described in the DUVRR Spectroscopy section. Protamine is a relatively low molecular weight protein (4 kDa) composed mostly of Arg residues with an unordered conformation. Human serum albumin (HSA) is an inactive ingredient (carrier protein)23 in the drug product Epogen (Epoetin), and was utilized as an α-helical model for this work. To aid bloodstream circulation of Epoetin, a high fraction of Epoetin should be bound to HSA.23 As such, a considerably larger concentration of HSA as compared to Epoetin is required in the drug product, resulting in the DUVRR spectra resembling HSA. Rituximab is an IgG1 mAb composed mostly of β-sheet conformation with a molecular weight of ∼143 kDa. It is important to note that the excipient polysorbate 80 is present at a concentration of an order of magnitude larger than Rituximab. The resonance enhancement of the protein at this excitation results in the Raman spectrum dominated by signal from the protein. The protein secondary structures determined by DUVRR were consistent with CD spectra of these proteins (Figure S-1, Supporting Information). Figure 1 illustrates the sensitivity of DUVRR to the secondary structure of protein pharmaceuticals. For instance, the frequency of the amide I vibrational mode (primarily CO stretching)19 for the β-sheet protein Rituximab is higher (1667 cm−1) as compared to the α-helical (albumin at 1659 cm−1) and unordered (protamine at 1645 cm−1) proteins. As expected from the literature, the amide I vibrational mode for protamine (unordered structure) is observed to be considerably more broad as compared to the narrow amide I bands characteristic of β-sheet and α-helical proteins.18,24 Furthermore, the literature assignments of the amide I Raman bands report lower frequencies for unordered proteins as compared to β-sheet conformation, a trend
■
RESULTS AND DISCUSSION DUVRR Spectroscopy of Protein Pharmaceuticals. Three protein drug products were subjected to DUVRR analysis. Each drug product represented a model for different protein secondary structure (unordered structure, α-helix and β-sheet). For analysis, diluted drug products were utilized. Although DUVRR spectra of proteins at high concentrations 7882
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
Article
observed here.18,19 Similarly, literature assignments show an increase in the frequency of the amide III vibrational mode (C−N stretching and N−H in-plane bending) from β-sheet (1240 cm−1) to unordered (1246 cm−1) to α-helix (1262 cm−1), a trend also reproduced here.18,19 Amide II shows relatively less variation between secondary structures.19 The Raman peak at ∼1600 cm−1 is absent in the DUVRR spectrum of protamine sulfate (green trace) due to it originating from the side chains of the aromatic amino acids Tyr/Phe and, protamine is composed of Arg residues. Figure 1 illustrates the ability of DUVRR to probe diluted protein drug products with different excipient concentrations, protein molecular weights and secondary structures. As such, a fully developed DUVRR-based quality assessment method is expected to be applicable to a wide variety of protein pharmaceuticals. Thermal Stress of Rituximab. To monitor the thermal unfolding of Rituximab, the drug product Rituxan was diluted to 1 mg/mL concentration of mAb, and subjected to temperatures between 22 and 72 °C for 30 min. DUVRR spectra were collected during heating for each temperature interval. It was determined that aggregation of Rituximab at elevated temperatures interfered with collection of Raman data as a result of aggregate precipitation. Under these conditions estimations could not be made about changes to Rituximab secondary structure in the presence of aggregated protein, due to an increase in solution turbidity and/or significant loss in concentration of soluble protein. In order to maintain consistent protein concentration and collect useful DUVRR spectra at elevated temperatures, the pH of the samples were reduced below the isoelectric point of Rituximab. At pH 1.9 and 72 °C, the vibrational modes originating from the amide backbone of Rituximab could be characterized (Figure S-2, Supporting Information). Alternatively, centrifugation of the heat stressed samples could be used to remove aggregates at elevated temperatures and to allow for the characterization of the protein amide vibrational modes (Figure S-3, Supporting Information). However, an approximate 50% loss in protein concentration due to aggregation could be expected when this approach is utilized (determined by A280 nm, Table S-1, Supporting Information). As previously noted, methods to monitor/characterize mAb aggregation are well established. In order to characterize changes to mAb secondary structure under thermal stress and constant concentrations by DUVRR, the Rituximab samples were adjusted to pH 1.9 prior to analysis, as noted above. Figure 2 illustrates minor changes in DUVRR spectra between pH 1.9 and 7.5 Rituximab samples at 22 °C.
To characterize interbatch heterogeneity of Rituxan, three lots were investigated under thermal stress. Individual samples were heated from 22 to 72 °C for a total of 30 min at each temperature point. DUVRR spectra were collected during sample heating. For each Rituxan lot, triplicate analysis was performed, resulting in nine spectra for each of the six temperature points. The 54 spectral data set was preprocessed as described in the DUVRR Spectroscopy section. Next, the nine spectra for each temperature point were averaged. The resulting mean spectra are plotted in Figure 3. Obvious trends
Figure 3. DUVRR spectra of 1 mg/mL Rituximab at pH 1.9 collected from 22 to 72 °C. Arrows indicate trends with temperature at vibrational modes related to the amide backbone of the protein.
can be observed in the mean spectra from different temperature points. As temperature increases the amide I vibrational mode experiences enhancement, broadening and a shift in peak location. Figure 4 is a magnified view of the amide I vibrational
Figure 4. Magnification of amide I vibrational mode from spectra plotted in Figure 3. Purple and yellow traces represent spectra collected at 60 and 72 °C, respectively. A shift of about 10 cm−1 at the amide I band can be observed for the spectra collected at higher temperatures. These changes indicate the unfolding of Rituximab at higher temperatures.
mode and illustrates a shift from 1670 cm−1 for Rituximab spectra collected at low temperatures, to 1660 cm−1 for high temperature spectra. The Raman spectra collected at 60 and 72 °C (purple and yellow trace, respectively) show considerable broadening as compared to low temperature spectra. A shift to lower wavenumbers and broadening of the amide I peak is characteristic of an increased content in unordered conformation.18,19 Signal enhancement for the Phe/Tyr (1600 cm−1) Raman peaks and loss in signal at amide II, III and Cα-H vibrational modes were also observed. The loss in signal for the other vibrational modes was not a result of loss in Rituximab
Figure 2. DUVRR spectra of 1 mg/mL Rituximab collected at 22 °C, pH 1.9 (blue trace) and pH 7.5 (green trace). 7883
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
Article
concentration. At pH 1.9, the concentration of Rituximab in solution was constant at all temperature points (Table S-1, Supporting Information). Therefore, any changes in the DUVRR spectra should be resulting from changes to Rituximab secondary structure. The DUVRR spectroscopic approach is designed as a rapid structural assessment and quality estimation method of mAb pharmaceuticals. It should be considered that in a real world scenario, the quality assessment of the mAb product would occur at ambient temperatures. This is due to the fact that the storage conditions prior to the quality assessment would be unknown. Therefore, it is important to determine if the unfolding process is reversible. If temperature reduction (i.e., from 72 °C to room temperature) caused the protein to return to its native secondary structure, then DUVRR spectroscopic analysis would not be able to identify the improper storage (although it would not necessarily exclude loss in product quality). This point is relevant because a scenario could occur where Rituximab is improperly stored at >70 °C for a period of time but returned to proper conditions before quality assessment could be performed. To investigate this scenario, native Rituximab at 1 mg/mL was heated at 72 °C for 30 min. Next, it was cooled to 22 °C for 30 min. To solvate fully the sample, the pH was reduced to 1.9. Figure 5 compares the
Figure 6. Comparing amide vibrational mode peak locations for native Rituxan (green trace), heat stressed Rituxan (red trace) and an unordered structure protein model (protamine, blue trace). Protamine was measured at neutral pH; Rituxan samples were measured at pH 1.9.
cm−1 closely resemble the bands at 1245/1254 cm−1 for protamine as compared to the amide III shape of native Rituximab. Figures 3 and 4 illustrate that significant changes do not occur below 60 °C. CD was performed as an orthogonal method on pH 1.9 Rituximab samples. CD analysis confirmed changes to the secondary structure of Rituximab above 60 °C (Figure S-4, Supporting Information). In parallel, an experiment was performed in which Rituximab was subjected to heat stress and cooled. However, to remove aggregates, centrifugation was utilized (samples were of neutral pH). Figure S-3, Supporting Information compares the control with the heat stressed and centrifuged sample. The heat stressed sample displays an enhancement in the amide I peak (amide I/amide II ratio of 0.97 as compared to 0.80 for the control) and a shift to lower frequency. The parallel experiment on Rituximab samples of neutral pH, reproduced the results observed for the low pH samples. These results indicate that thermally induced denaturation of Rituximab is not reversible, allowing for quality assessment to be performed at a later time and at ambient temperature. The dilution of the drug product should also be considered as a factor of protein unfolding. However, as investigated by Andersen et al., the concentration dependence on the thermal unfolding of Rituximab is “surprisingly weak.11” Specifically, the aggregation kinetic reaction rates (at elevated temperatures) have a “weak concentration dependence.” Both results presented here and information from the literature support the conclusion that the changes in DUVRR spectra are due to thermal unfolding of Rituximab and not an effect of pH or protein concentration. The preliminary results illustrate the ability of DUVRR spectroscopy to monitor denaturation of Rituximab under thermal stress. Markers in the DUVRR spectra were identified that could indicate unfolding of Rituximab; peak location of the amide I mode (shift to lower frequency, ∼1660 cm−1) and ratio of amide I/amide II (ratio ≥1). The shape of the amide I and III Raman bands and the peak location of the Cα-H vibrational mode also showed promise as markers for thermal unfolding. Prior research on Rituximab determined the specific melting points for each subdomain of the mAb via differential scanning calorimetry (DSC).11 At 72 °C, only one subdomain (CH2, the second constant subdomain on the heavy chain, located in the Fc region) experiences a phase transition. The fab region and the CH3 subdomain of the Fc region unfold at higher
Figure 5. Comparing heat stressed (red trace) and control (green trace) Rituximab samples. Both samples were collected at room temperature and pH 1.9. The spectra were not preprocessed (raw), with the exception of normalization at the amide II vibrational mode.
DUVRR spectra of the heat stressed sample after pH reduction (red trace) and a pH 1.9 Rituxan sample that was not exposed to heat stress (control, green trace). The spectra are raw with the exception of normalization at the amide II vibrational mode, which is relatively insensitive to changes in secondary structure.19 A drastic enhancement and broadening in the amide I vibrational mode can be observed in the heat stressed sample. The ratio of amide I/amide II peak intensity was calculated to be 0.77 for the control and 1.2 for the heat stressed sample. As observed before, the amide I peak location for the heat stressed sample experiences a 10 cm−1 shift to lower frequency. Figure 6 compares DUVRR spectra of native Rituximab at 1 mg/mL (green trace), heat stressed Rituxan at 1 mg/mL and pH 1.9 (red trace), and native protamine (blue trace), which is a model for unordered secondary structure. A qualitative comparison of the spectra indicates that the heat stressed Rituxan sample trend toward the peak locations of protamine. Specifically, the broadening of the amide I vibrational mode of the heat stressed sample more closely resembles protamine as compared to the native Rituximab spectrum. Additionally, the peak locations of the amide I and Cα-H vibrational modes for the heat stressed sample shift closer to the unordered structure model and the peak shape of the amide III bands at 1237/1251 7884
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
Article
temperatures.11 A cartoon illustration of the IgG mAb subdomains is provided in Figure S-5, Supporting Information. It is impressive that DUVRR spectroscopy is capable of observing changes to the secondary structure of the mAb based upon the unfolding of one subdomain. These minor changes in the mAb structure have the potential to impact protein function. Thermally unfolded Rituximab was subjected to a CDC bioassay to determine if the structural assessment performed by DUVRR spectroscopy could be related to mAb function. CDC Bioassay. To determine the potency of heat stressed Rituximab, a complement dependent cytotoxicity (CDC) bioassay was performed. CDC is a main effector function of Rituximab for killing cancerous or overexpressed cells. CDC requires the generation of an antibody-complement complex through the binding of the complement molecule at the Fc domain of the mAb. The assay measures the ability of the mAb to generate this complex as a function of cell viability over a range of mAb concentration. The fluorescent response is directly proportional to the amount of viable cells. Figure 7 is
formulated mAbs is a physicochemical property that is relatively under investigated. This work illustrates a new method for probing the secondary structure of mAb drug products via DUVRR spectroscopy. DUVRR spectroscopy holds several advantages over other techniques for the characterization of mAb secondary structure. Although the DUVRR instrument is not commercially available, results presented here and literature data illustrate that the technique is promising. It was found that the method was applicable to different types of protein pharmaceuticals, independent of protein molecular weight and secondary structure. Similar to other quality assessment methods (i.e., NMR, SEC-HPLC and electrophoretic methods) DUVRR analysis of mAbs should be considered as destructive (analyzed samples should not be administered to patients). Other drawbacks of the approach include the requirement to remove protein aggregates for DUVRR analysis to be performed, achieved here via minor sample preparation. Centrifugation or pH reduction of samples was determined to be effective sample preparation methods to remove protein aggregates. Dilution of the drug product to 1 mg/mL protein concentration was found to provide higher quality Raman spectra as compared to analysis of the original drug product. Although dilution of the drug product is an additional sample preparation technique, it is a common procedure used by other mAb analysis methods (i.e., CD, mass spectrometry and NMR). DUVRR spectroscopy was able to monitor unfolding of the therapeutic mAb Rituximab under thermal stress. Spectroscopic markers were discovered that could be used to identify unfolding of the mAb from its native β-sheet to a partially unordered (random coil) conformation. Changes in the DUVRR spectra of unfolded Rituximab were correlated to loss in mAb Fc function and impacting the potency of the mAb as determined by a complement dependent cytotoxicity (CDC) bioassay. Additionally, it was found that the thermal unfolding of Rituximab was not reversible. Thus, in a real world scenario, cooling would not reverse the loss of product quality. DUVRR spectroscopy shows promise as a method to probe mAb secondary structure and expedite the identification of pharmaceutical efficacy loss due to mAb degradation. This approach could replace current methods used to investigate the relationship of mAb structure and function, all of which are time-consuming and demand considerable sample preparation. DUVRR spectroscopy is applicable across a wide range of protein pharmaceuticals and, as shown here, is able to monitor unfolding caused by thermal stress. Although obvious changes were observed in DUVRR spectra above 60 °C, a trend of degradation was also observed from 22 to 60 °C. As such, a direct application of this method would be to determine the exact temperature at which the mAb drug product was improperly stored. This approach is relevant because visible aggregates may not occur until elevated temperatures yet degradation could still occur at lower temperatures. Work presented here illustrates the ability of DUVRR to monitor degradation by thermal stress; however, it is expected that the method will also be sensitive to changes in secondary structure caused by other sources of stress. Future directions will include DUVRR studies that utilize other forms of stress to denature proteins, and the application of multivariate analyses to extract additional information from the spectra.
Figure 7. Complement dependent cytotoxicity (CDC) bioassay. Comparing Rituxan control (blue trace) to heat stressed sample (red trace). The plot represents calculated % cell viability versus concentration of Rituximab.
the resulting CDC plot measuring the cell response in calculated % viability (measured by fluorescence: 560 nm excitation/590 nm emission) as compared the concentration of Rituximab, fit with a four parameter sigmoidal curve. Samples exposed to thermal stress at 72 °C for 30 min were compared to a Rituximab control sample that was not heated. Samples were centrifuged before the assay was performed to remove aggregates. As observed, to reach the same effectiveness the heat stressed sample required double the concentration (shifted to higher concentration on x-axis) as compared to the control, indicating loss of CDC effector function when Rituximab is heated to 72 °C. Previous research has indicated that Rituximab has a thermal unfolding point at 71 °C, associated with the Fc domain of the antibody.11 This unfolding point explains the loss in function for the Fc domain to complex with the complement molecule, resulting in loss of mAb function.
■
CONCLUSION Quality assessment of monoclonal antibody (mAb) drug products requires investigating numerous attributes including physicochemical properties, biological activity, immunogenicity, purity and quantity. Secondary structure assessment of 7885
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
Analytical Chemistry
■
Article
(21) Arzhantsev, S.; Vilker, V.; Kauffman, J. Appl. Spectrosc. 2012, 66, 1262−1268. (22) Du, J.; Wang, H.; Zhong, C.; Peng, B.; Zhang, M.; Li, B.; Huo, S.; Guo, Y.; Ding, J. J. Biol. Chem. 2007, 282, 15073−15080. (23) Fanali, G.; di Masi, A.; Trezza, V.; Marino, M.; Fasano, M.; Ascenzi, P. Mol. Aspects Med. 2012, 33, 209−290. (24) Xu, M.; Shashilov, V.; Lednev, I. K. J. Am. Chem. Soc. 2007, 129, 11002−11003.
ASSOCIATED CONTENT
S Supporting Information *
Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org/. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b01606.
■
AUTHOR INFORMATION
Corresponding Author
*J. Bueno. E-mail:
[email protected]. Notes
This publication reflects the views of the authors and should not be construed to represent FDA’s views or policies. The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Funding for this work was provided by the CDER Critical Path Program.
■
REFERENCES
(1) Aggarwal, S. R. Nature 2014, 201, 33−39. (2) Schellekens, H. Nat. Rev. Drug Discovery 2002, 1, 457−462. (3) Rosenberg, A. S. AAPS J. 2006, 8, E501−E507. (4) Arbogast, L. W.; Brinson, R. G.; Marino, J. P. Anal. Chem. 2015, 87, 3556−3561. (5) Vázquez-Rey, M.; Lang, D. A. Biotechnol. Bioeng. 2011, 108, 1494−1508. (6) Therapeutic Monoclonal Antibodies: from Bench to Cinic; An, Z., Ed.; John Wiley & Sons: Hoboken, NJ, 2009. (7) Barton, C.; Spencer, D.; Levitskaya, S.; Feng, J.; Harris, R.; Schenerman, M. A. In State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Vol. 1. Monoclonal Antibody Therapeutics: Structure, Function, and Regulatory Space; American Chemical Society: Washington, DC, 2014; pp 69−98. (8) Brorson, K.; Kendrick, B. In State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Vol. 1. Monoclonal Antibody Therapeutics: Structure, Function, and Regulatory Space; American Chemical Society: Washington, DC, 2014; pp 99− 116. (9) Thiagarajan, G.; Widjaja, E.; Heo, J. H.; Cheung, J. K.; Wabuyele, B.; Mou, X.; Shameem, M. J. Raman Spectrosc. 2015, 46, 531. (10) Joshi, V.; Shivach, T.; Yadav, N.; Rathore, A. S. Anal. Chem. 2014, 86, 11606−11613. (11) Andersen, C. B.; Manno, M.; Rischel, C.; Thórólfsson, M.; Martorana, V. Protein Sci. 2010, 19, 279−290. (12) Telikepalli, S. N.; Kumru, O. S.; Kalonia, C.; Esfandiary, R.; Joshi, S. B.; Middaugh, C. R.; Volkin, D. B. J. Pharm. Sci. 2014, 103, 796−809. (13) Boulet-Audet, M.; Byrne, B.; Kazarian, S. G. Anal. Chem. 2014, 86, 9786−9793. (14) Li, C. H.; Nguyen, X.; Narhi, L.; Chemmalil, L.; Towers, E.; Muzammil, S.; Gabrielson, J.; Jiang, Y. J. Pharm. Sci. 2011, 100, 4642− 4654. (15) Oladepo, S. A.; Xiong, K.; Hong, Z.; Asher, S. A.; Handen, J.; Lednev, I. K. Chem. Rev. 2012, 112, 2604−2628. (16) Chi, Z.; Chen, X.; Holtz, J. S.; Asher, S. A. Biochemistry 1998, 37, 2854−2864. (17) Copeland, R. A.; Spiro, T. G. Biochemistry 1987, 26, 2134−2139. (18) Shashilov, V. A.; Sikirzhytski, V.; Popova, L. A.; Lednev, I. K. Methods 2010, 52, 23−27. (19) Huang, C.-Y.; Balakrishnan, G.; Spiro, T. G. J. Raman Spectrosc. 2006, 37, 277−282. (20) Jordan, T.; Eads, J. C.; Spiro, T. G. Protein Sci. 1995, 4, 716− 728. 7886
DOI: 10.1021/acs.analchem.5b01606 Anal. Chem. 2015, 87, 7880−7886
