A detailed protocol for the purification and subsequent analysis of a monoclonal antibody from harvested cell culture fluid (HCCF) of automated microbioreactors has been described. Use of analytics to determine critical quality attributes (CQAs) and maximizing limited sample volume to extract vital information is also presented.
Monoclonal antibodies (mAbs) are one of the most popular and well-characterized biological products manufactured today. Most commonly produced using Chinese hamster ovary (CHO) cells, culture and process conditions must be optimized to maximize antibody titers and achieve target quality profiles. Typically, this optimization uses automated microscale bioreactors (15 mL) to screen multiple process conditions in parallel. Optimization criteria include culture performance and the critical quality attributes (CQAs) of the monoclonal antibody (mAb) product, which may impact its efficacy and safety. Culture performance metrics include cell growth and nutrient consumption, while the CQAs include the mAb’s N-glycosylation and aggregation profiles, charge variants, and molecular weight. This detailed protocol describes how to purify and subsequently analyze HCCF samples produced by an automated microbioreactor system to gain valuable performance metrics and outputs. First, an automated protein A fast protein liquid chromatography (FPLC) method is used to purify the mAb from harvested cell culture samples. Once concentrated, the glycan profiles are analyzed by mass spectrometry using a specific platform (refer to the Table of Materials). Antibody molecular weights and aggregation profiles are determined using size exclusion chromatography-multiple angle light scattering (SEC-MALS), while charge variants are analyzed using microchip capillary zone electrophoresis (mCZE). In addition to the culture performance metrics captured during the bioreactor process (i.e., culture viability, cell counts, and common metabolites including glutamine, glucose, lactate, and ammonia), spent media is analyzed to identify limiting nutrients to improve the feeding strategies and overall process design. Therefore, a detailed protocol for the absolute quantification of amino acids by liquid chromatography-mass spectrometry (LC-MS) of spent media is also described. The methods used in this protocol take advantage of high-throughput platforms that are compatible for large numbers of small-volume samples.
Protein therapeutics are being used to treat a growing variety of medical conditions including tissue transplant complications, autoimmune disorders, and cancers1. Since 2004, the United States Food and Drug Administration (USFDA) has documented an increasing proportion of biologic license applications (BLAs) of all approvals regulated by the Center for Drug Evaluation and Research (CDER), with BLAs accounting for over 25% in 2014 and 20152.
Considering this expanding market, biopharmaceutical manufacturers are challenged with quickly delivering more product with consistent quality. Efforts to augment product yield have focused on CHO cell engineering and production line screening, though the most significant improvements are due to advances in media/feed strategy optimization and cell culture environmental controls1,3,4,5 during the manufacturing process.
Since mAbs are produced in a biological system, there can be inherent protein variability. Antibody composition can be altered post-translationally, such as glycosylation or impacted by degradation or enzymatic reactions. These structural variations may provoke dangerous immune reactions or alter antibody binding, which in turn can reduce or eliminate the intended therapeutic function5. Thus, critical quality attributes (CQAs) of monoclonal antibodies – N-glycan profile, charge variant distribution, and the percentage of antibody in monomeric form – are regularly monitored and controlled as part of a Quality by Design (QbD) approach during manufacturing processes1,6. In a regulated production environment, therapeutic proteins must meet acceptance criteria to be licensed as an approved commercial drug product7. The methods presented herein would typically be part of the quality characterization process for an antibody7,8, and any protein scientist will be familiar with their usage.
In prior work9, the application and operation of microbioreactors for high throughput screening of cell culture conditions in upstream bioprocessing has been described. The purified product obtained from the varying media conditions is subjected to N-glycan analysis using LC-MS. Glycosylation patterns of therapeutic proteins can be detected and characterized using LC-MS techniques10,11, and the presence of various glycan species has been linked to bioprocess parameters such as feed strategy, pH, and temperature12. The effect of the varying media conditions on product quality, indicated by the percentage of the resulting IgG in monomeric form, is also evaluated with Size Exclusion Chromatography- Multi-Angle Light Scattering (SEC-MALS)13,14,15. The charge variant profile represents a number of modifications16 that could impact the function of a product. Microcapillary zone electrophoresis (mCZE) is a technique that offers a considerably faster analysis time compared to traditional cation exchange (CEX) chromatography and capillary isoelectric focusing (cIEF) methods used for charge variant analysis17,18. Spent bioreactor media was analyzed to track amino acid consumption during protein production as it relates to changes in the antibody's identifying attributes19,20,21,22,23.
Protein analytics allow us to identify critical process parameters (CPPs) based on the relationships between process inputs and changes in CQAs. During bioprocess development, identifying and measuring CPPs fundamentally demonstrates process control and ensures that the product has not changed, which is essential in highly regulated manufacturing environments. In this paper, analytical techniques to measure some of the biochemical characteristics of the protein most pertinent to product CQAs (N-glycan profile, charge variants, and size homogeneity) are presented.
1. Purification of antibody
NOTE: The equilibration buffer for the in-house antibody is 25 mM Tris, 100 mM NaCl, pH 7.5. The elution buffer used is 0.1 M acetic acid. The buffers and resin (Protein A) are dependent on the specific antibody purified. Column volume is equivalent to the bed height of the resin. The amount of mobile phase used is determined in terms of column volume.
2. Concentration of purified antibody
NOTE: The Tris-acetate buffer is 0.1 M acetic acid neutralized with 1 M Tris Base to a pH of ~5.5.
3. Analysis of N-glycans using mass spectroscopy
4. Analysis of antibody aggregation using SEC-MALS
5. Charge variant analysis
6. Amino acid analysis
The harvested cell culture fluid from the automated microscale bioreactor is purified using fast protein liquid chromatography (FPLC), as seen in Figure 1 and the purified proteins' critical quality attributes (CQAs) were characterized by various downstream analytical methods. This is a key benefit of the automated microbioreactor system; differences in CQAs can be rapidly assessed across a wide range of conditions. N-glycan data from CHO-produced mAbs that are processed by mass spectrometry should appear like the chromatograms shown in Figure 2. The figure depicts a comparison between two chromatograms showing that the mannose 5 peak (M5) from one sample is considerably lower. If only a noisy baseline is observed instead of peaks, this may mean that the chromatography setup is faulty or that the procedure is not successful. Using controls, troubleshooting can be simplified. First, assess the FLR peaks from the dextran ladder; these peaks indicate that the chromatographic system is working correctly. Next, compare the experimentally obtained peaks with those obtained from a processed intact mAb standard. If peaks from the standard are visible, but no sample peaks are identified, then the mAb samples were not processed correctly. This may be due to SDS or nucleophile presence in the buffer interfering with N-glycan labeling and purification.
SEC-MALS can be used to assess two more CQAs: the aggregation profile and the molecular weight of the antibody. A representative SEC-MALS chromatogram is comparable to the one shown in Figure 3. The molecular mass distribution and the absolute molecular weight were determined using the required software with an extinction coefficient of 1.37 mL*(mg*cm)-1 and a dn/dc of 0.185 mL/g. As peak calling and setting the baseline in the software is performed manually, results may vary slightly from user to user. The absolute molecular weight of monomeric IgG1 from Figure 3 is 1.504 x 105 Da ± 0.38% (blue) and the higher order complex is 7.799 x 105 Da ± 3.0% (red). The polydispersity of the aggregates is much greater than that of the monomer, as indicated by the red molar mass distribution of Peak 1 (Figure 3). The small quantity of sample and importance of aggregation as a CQA make this technique a highly valuable complementary analytical tool to the automated microbioreactor system.
The result of mCZE is an electropherogram, such as in Figure 4, which shows the charge variant profile for a monoclonal antibody. The profile is a unique signature for the protein being investigated and is highly sensitive to the operating pH. Also visible is a free-dye peak to the left of the charge variant profile. When establishing an operating pH, there is some discretion to the operator to balance resolution and signal; in addition, the operator must ensure good separation from the free-dye peak which migrates at ~30 s. The sample can be desalted after labelling to remove this peak, though this leads to a significant loss in signal. Once an operating pH is established, the sample profiles can be compared. While generally consistent, changes in labeling efficiency or differences in excipients can lead to minor differences in the migration of a sample and the charge variant profile making electropherograms hard to directly compare. Instead, the method of comparison is usually based on the percentages of basic, main, and acidic species. In this case, relative differences as small as 1-2% can be identified using mCZE.
Amino acid consumption can be monitored to determine if depletion is causing changes in CQAs. Chromatogram readouts from the mass spectrometer can be used to evaluate the successful creation of a calibration curve for the absolute quantification of amino acids in crude bioreactor media samples. Figure 5 depicts two total ion chromatograms (TIC) and one extracted ion chromatogram (XIC) as representative results during this process. In Figure 5A, the TIC shown depicts the background signal from the buffer system as only a water blank was injected. Figure 5B depicts a representative TIC of the amino acid standard where, when compared to the water blank, small peaks that correspond to the individual amino acid species can be observed (such as lysine at 7.96 minutes). To integrate the peak and facilitate the quantification of peak area (and therefore the concentration), the XIC is used where only the signal from a defined "chromatogram mass window" is displayed. Depending on the sensitivity of the instrument and the quality of the chromatographic separation, the optimal mass window will have to be determined by the user. In this example (Figure 5C), the XIC of lysine (m/z = 147.1144) with a mass window of 10 ppm is shown where lysine in the amino acid standard elutes off the column at 8.03 minutes.
Figure 1. Representative chromatogram of the purification scheme using the Fast Protein Liquid Chromatography (FPLC) technique. Purification method phases corresponding to volume (mL) are labeled along the x-axis. UV absorbance at 280 nm (mAU y-axis, solid line) is monitored throughout the purification cycle. Non-specifically bound impurities are displaced by increasing conductivity (mS/cm y-axis, dashed line) during the High Salt Wash. Antibody is eluted from Protein A column with the introduction of elution buffer (Conc B, dotted line) when the pH decreases to 4 (not shown). Please click here to view a larger version of this figure.
Figure 2. A representative fluorescence chromatogram obtained from tagged glycans that are mass verified. The x-axis is retention time (minutes) while the y-axis is signal intensity. The peak at 14.94 min represents the Mannose 5 (M5) glycan, where a large difference between the M5 signal strength can be observed between the two samples that are overlaid. Please click here to view a larger version of this figure.
Figure 3. Molecular weight distribution of IgG1 monoclonal antibody. Chromatogram of an intact IgG1 monoclonal antibody separated by size exclusion chromatography in 1x PBS (pH7.4). Absorbance is monitored at 280 nm (black; left axis) and light scattering and refractive index detectors were used to calculate the absolute molecular weight of each peak (red and blue; right axis). High Molecular Weight species are indicated with the peak labeled "HMW". Please click here to view a larger version of this figure.
Figure 4. Charge variant profile of a IgG1 monoclonal antibody. This electropherogram is generated on a mCZE platform. A free-dye peak migrates at ~30 s and is well separated from the IgG1. For quantification, peaks were split into basic, main, and acidic species using instrument data analysis software. The red line outlines the integrated peak areas. Please click here to view a larger version of this figure.
Figure 5. Representative results of the ion chromatograms for mass spectrometry-based amino acid analysis of crude bioreactor media. The x-axis is time (minutes) while the y-axis is signal intensity (A) A water blank serves as the negative control and reveals the background signal observed over the course of the liquid chromatography gradient (B) The 225 pmol/µL amino acid standard is used here as a positive control, as the individual peaks observed in this total ion chromatogram represent the different amino acids of the standard mix being resolved chromatographically (C) A representative extracted ion chromatogram for m/z 147.1144, which is lysine. The 7.96 min peak in B corresponds to the 8.03 peak in C of lysine. Please click here to view a larger version of this figure.
HCCF contains debris and large particles that can clog and destroy costly instrumentation, thus culture clarification is needed before further downstream processing. Centrifugation is generally the first approach to separate cells and other insoluble particles from proteins followed by filtration. This filtered HCCF is then subjected to Fast Protein Liquid Chromatography (FPLC) for purification. Purification of HCCF from automated microbioreactors to obtain the product is an important step in downstream processing. Here, a benchtop FPLC system with a protein A column is used to obtain monoclonal antibodies from the HCCF. Analytics for upstream processes can provide useful insight into cell behavior and guide bioprocess design, helping to obtain a consistent and reliable quality product. Analytics also allows us to link Critical Quality Attributes (CQAs) to upstream and downstream processes. Presented here are four assays that are commonly used in the characterization of monoclonal antibodies. These techniques are robust, reliable, and readily deployable for process and product analysis from a variety of upstream sources which are only partially purified and may still contain residual levels of DNA and HCP.
When cleaning up samples for analytics, an important balance must be struck between creating a sample that is sufficiently clean for analysis while preserving the variability present in the bioreactor. The two most common contaminants impacting product are DNA and HCP, which can be checked by measuring the ratio absorbance at 260/280 nm and through SDS-PAGE or µCE-SDS. The assays presented here are not sensitive to low levels of DNA content. The purity of the product is >95% pure, as determined by µCE-SDS.
Charge variant analysis with a microcapillary electrophoresis system provides a high-throughput method to identify charge variants, with chips and reagents that are relatively easy to implement. The nature of the technique and the chemistry of the labeling reagent are both sensitive to excipients and other primary amines, thus requiring a desalting step for most sample matrixes. From experience, low levels of DNA co-migrate with the free-dye from the labeling reaction and do not impact the quality of results. While variability of basic, main, and acidic peak quantification is typically <1%, higher levels of DNA and other contaminants can increase the variability of the assay. It is extremely important to be consistent with the protein labeling and ensure the prompt use of DMF after removal from the bottle and being mixed with the dye. Lysine and/or histidine standards are recommended as labeling controls. Over time and depending on sample quality, chips can foul or lose the coating on the microfluidics channels, leading to greater noise, the presence of ghost peaks, and greater sample-to-sample variation. To identify this occurrence, blanks and a system suitability standard (i.e. NISTmAb) were concurrently analyzed with the samples at regular intervals. When chip issues arise, the chips can be washed with the storage solution or replaced.
The methods used for glycan analysis of therapeutic glycoproteins primarily involves liquid chromatography (LC) and/or mass spectrometry (MS), with lectin microarray analysis gaining popularity as a third option25. The method described in this paper uses both LC and MS, which has benefits and disadvantages. Mass spectrometric methods have the advantage of mass verification of the analyzed glycans, which is not possible with LC-based methods using a fluorescent detection output or lectin microarrays. This method uses LC and fluorescence detection to assign glycan identities using retention time comparison to a dextran ladder standard. Fluorescence monitoring allows for increased sensitivity and quantification due to the ease of its detection, where MS alone might not be able to quantify low abundance species due to the poor ionization efficiency of oligosaccharides. The mass information from MS is used to confirm glycan identities, but the processing software does not use mass information as the primary assignment criteria. Hence, without reproducible chromatography and easily resolvable peaks, this method can suffer regarding glycan assignments. Fortunately, the mass information can help with glycan assignments even in situations when the chromatography is subpar, such as shifts in retention time that hinder reproducible glycan assignments. If this method is used without MS, the chromatography must be at the highest level since mass information cannot be used to correct for residence time drift.
The amino acid analysis method described here utilizes LC-MS for rapid quantitation of underivatized amino acids in crude cell culture media. Alternative amino acid analysis methods require amino acid derivatization agents to enable UV detection26. The LC-MS method offers important advantages over the LC-UV method: it allows for identification based on both retention time and ion mass as opposed to the LC-UV method, which is limited by a lack of mass characterization. Furthermore, the LC-MS method offers time and reproducibility advantages, as the LC-UV method requires a time-consuming derivatization reaction, which may impart sample variability27. However, the injection of crude cell culture media in the LC-MS method can cause detrimental effects on MS signal due to ion skimmer fouling. A calibration ladder is injected frequently as a system suitability check, and sample order is randomized to prevent bias in the data.
The cell culture process for antibody production using microbioreactors is previously described9. In this study, detailed protocols for monoclonal antibody characterization methods that maximize data acquired from limited sample volumes are well-defined. Limited amounts of harvested cell culture fluid can sometimes restrict the product information acquired and selection of the right analytical procedures to obtain product quality data is essential. Analytics are important to link together upstream process parameters to the changes in product quality. Here, a guideline is provided for users to characterize mAbs when working with microbioreactors.
The authors have nothing to disclose.
The authors would like to thank Scott Lute for the analytical support he provided. Partial internal funding and support for this work is provided by the CDER Critical Path Program (CA #1-13). This project is supported in part by an appointment to the Internship/Research Participation Program at the Office of Biotechnology Products, U.S. Food and Drug Administration, administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and FDA.
CHO DG44 Cell Line | Invitrogen | A1100001 | |
Akta Avant 25 | General Electric Life Sciences | 28930842 | |
Pro Sep vA Ultra Chromatography Resin | Millipore Sigma | 115115830 | Purification Stationary Phase |
Omnifit 10cm Column | Diba Fluid Intelligence | 006EZ-06-10-AA | Housing for Stationary Phase |
Tris Base | Fisher Scientific | BP154-1 | |
Superloop 10 mL | GE Healthcare | 18-1113-81 | |
µDawn Multi Angle Light Scattering Detector | Wyatt | WUDAWN-01 | |
0.22 µm Millex GV Filter Unit PVDF Membrane | Merck Millipore | SLGV033RB | |
10X Phosphate Buffered Saline | Corning | 46-013-CM | |
12 mL Syringe | Covidien | 8881512878 | |
1290 Infinity Binary Pump | Agilent Technologies | G4220A | |
1290 Infinity DAD | Agilent Technologies | G4212A | |
1290 Infinity Sampler | Agilent Technologies | G4226A | |
1290 Infinity Thermostat | Agilent Technologies | G1330B | |
1290 Infinity Thermostatted Column Compartment | Agilent Technologies | G1316C | |
15 mL Falcon tube | Corning Inc. | 352097 | |
150 uL Glass Inserts with Polymer Feet | Agilent Technologies | 5183-2088 | |
50 mL Falcon tube | Corning Inc. | 352070 | |
96-Well Plate | Bio-Rad | 127737 | |
Acetic Acid | Sigma-Aldrich | 695072 | |
Acetonitrile | Fisher Chemical | BPA996-4 | |
ACQUITY I-Class UPLC BSM | Waters Corporation | 18601504612 | |
ACQUITY I-Class UPLC Sample Manager | Waters Corporation | 186015000 | |
ACQUITY UPLC FLR Detector | Waters Corporation | 176015029 | |
Amicon Ultra-4 100 kDa centrifugal filters | Merck Millipore | UFC810096 | |
Amino Acid Standard, 1 nmol/µL | Agilent Technologies | 5061-3330 | |
Amino Acid Supplement | Agilent Technologies | 5062-2478 | |
Ammonium Formate Solution – Glycan Analysis | Waters Corporation | 186007081 | |
Blue Screw Caps with Septa | Agilent Technologies | 5182-0717 | |
CD OptiCHO AGT Medium | Thermo Fisher Scientific | A1122205 | |
Centrifuge Tubes | Eppendorf | 22363352 | |
Charge Variant Chip | Perkin Elmer | 760435 | |
Charge Variant Reagent Kit | Perkin Elmer | CLS760670 | |
Chromatography Water (MS Grade) | Fisher Chemical | W6-4 | |
Dimethylformamide | Thermo Scientific | 20673 | |
Extraction Plate Manifold for Oasis 96-Well Plates | Waters Corporation | 186001831 | |
Formic Acid | Fisher Chemical | A117-50 | |
GlycoWorks RapiFlour-MS N-Glycan Starter Kit – 24 Sample | Waters Corporation | 176003712 | |
GXII Buffer Tubes | E&K Scientific | 697075- NC | |
GXII Detection Window Cleaning Cloth | VWR | 21912-046 | |
GXII HT Touch | Perkin Elmer | CLS138160 | |
GXII Ladder Tubes | Genemate | C-3258-1 | |
GXII Lint-Free Swab | ITW Texwipe | TX758B | |
Hydrochloric Acid | Fisher Scientific | A144-500 | |
Intact mAb Mass Check Standard | Waters Corporation | 186006552 | |
Intrada Amino Acid Column 150 x 2 mm | Imtakt | WAA25 | |
NanoDrop One Microvolume UV-Vis Spectrophotometer | Thermo Fisher Scientific | 840274100 | |
Optilab UT-rEX Differential Refractive Index Detector | Wyatt | WTREX-11 | |
Perchloric acid | Aldrich Chemistry | 311421 | |
Pipet Tips with Microcapillary for Loading Gels | Labcon | 1034-960-008 | |
Polypropylene 96-Well Microplate, F-bottom, Chimney-style, Black | Greiner Bio-One | 655209 | |
RapiFlour-MS Dextran Calibration Ladder | Waters Corporation | 186007982 | |
Screw Top Clear Vial 2mL | Agilent Technologies | 5182-0715 | |
Sodium Chloride | Fisher Scientific | S271-1 | |
Sodium Iodide | Sigma Aldrich | 383112 | |
TSKgel UP-SW3000 4.6mm ID x 30 cm L | Tosoh Biosciences | 003449 | |
UNIFI Scientific Information System | Waters Corporation | 667005138 | |
Vacuum Manifold Shims | Waters Corporation | 186007986 | |
Vacuum Pump | Waters Corporation | 725000604 | |
Xevo G2 Q-ToF | Waters Corporation | 186005597 | |
Zeba Spin Desalting Column, 0.5 mL | Thermo Scientific | 89883 |