Article pubs.acs.org/ac
Automated High-Throughput Permethylation for Glycosylation Analysis of Biologics Using MALDI-TOF-MS Archana Shubhakar,*,†,§ Radoslaw P. Kozak,† Karli R. Reiding,‡ Louise Royle,† Daniel I. R. Spencer,† Daryl L. Fernandes,† and Manfred Wuhrer‡,§ †
Ludger Ltd., Culham Science Centre, Abingdon, Oxfordshire, United Kingdom Leiden University Medical Center, Center for Proteomics and Metabolomics, Leiden, The Netherlands § Division of BioAnalytical Chemistry, VU University Amsterdam, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands ‡
S Supporting Information *
ABSTRACT: Monitoring glycoprotein therapeutics for changes in glycosylation throughout the drug’s life cycle is vital, as glycans significantly modulate the stability, biological activity, serum half-life, safety, and immunogenicity. Biopharma companies are increasingly adopting Quality by Design (QbD) frameworks for measuring, optimizing, and controlling drug glycosylation. Permethylation of glycans prior to analysis by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) is a valuable tool for glycan characterization and for screening of large numbers of samples in QbD drug realization. However, the existing protocols for manual permethylation and liquid−liquid extraction (LLE) steps are labor intensive and are thus not practical for high-throughput (HT) studies. Here we present a glycan permethylation protocol, based on 96-well microplates, that has been developed into a kit suitable for HT work. The workflow is largely automated using a liquid handling robot and includes N-glycan release, enrichment of N-glycans, permethylation, and LLE. The kit has been validated according to industry analytical performance guidelines and applied to characterize biopharmaceutical samples, including IgG4 monoclonal antibodies (mAbs) and recombinant human erythropoietin (rhEPO). The HT permethylation enabled glycan characterization and relative quantitation with minimal side reactions: the MALDI-TOF-MS profiles obtained were in good agreement with hydrophilic liquid interaction chromatography (HILIC) and ultrahigh performance liquid chromatography (UHPLC) data. Automated permethylation and extraction of 96 glycan samples was achieved in less than 5 h and automated data acquisition on MALDI-TOF-MS took on average less than 1 min per sample. This automated and HT glycan preparation and permethylation showed to be convenient, fast, and reliable and can be applied for drug glycan profiling and clinical glycan biomarker studies.
F
Organization (HUPO) and Human Disease Glycomics/ Proteome Initiative (HGPI)14 endorsed MS-based analysis for glycomics studies. The study also demonstrated that MALDITOF-MS of permethylated glycans is highly effective for glycan analysis during development and optimization of biotherapeutics, as well as for identifying disease biomarkers. This method can provide an accurate and reliable overview of the glycosylation profile in a short time span. Permethylation of released N- and O-glycans is routinely performed prior to MALDI-TOF-MS analysis,15,16 because it provides a number of advantages such as (i) improvement and enhancement of ionization efficiency of glycans when compared to nonderivatized oligosaccharides, (ii) stabilization of the labile sialic acid moieties, (iii) the detection of both neutral and acidic glycans in positive ion mode, (iv) easier determination of branching and glycosidic linkage positions, (v) the resulting
or most therapeutic glycoproteins, the glycosylation patterns correlate strongly with the clinical safety and efficacy profiles.1 In biological tissues and biological fluids2,3 (e.g., blood, saliva), these patterns can also correlate with the state of health or disease of the individual.4 Given this, there is an increasing interest in detailed characterization of glycosylation changes, for example, in Quality by Design (QbD) approach throughout biopharmaceutical development of monoclonal antibodies (mAbs).5,6 Changes in glycosylation can be complex and subtle, so a large number of samples (ranging from hundreds to thousands) often need to be analyzed to obtain conclusive results. Monitoring glycosylation is often challenging due to the complexity and heterogeneity of glycans.7 This is impacted by long turnaround times, high labor intensity, low concentrations of sample, lack of automation, and the high cost for sample analysis.8 The European Medicines Agency’s (EMA) guidelines for biosimilar mAbs recommends the use of several orthogonal techniques9−11 to provide complementary data to support identification and quantitation of glycoforms.12,13 A multi-institutional assessment of glycomics methodologies coordinated by the Human Proteome © XXXX American Chemical Society
Received: April 27, 2016 Accepted: August 1, 2016
A
DOI: 10.1021/acs.analchem.6b01639 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
using an AutoFlex Speed instrument from Bruker Daltonics (Bermen, Germany). Samples. Human IgG glycoprotein, fetuin glycoprotein, Nglycan standards A3G3S3, A2G2S2, and A2G2S1, and 13C permethylated human-IgG standard were obtained from Ludger; intact mAb IgG1 mass check standard was purchased from Waters (U.K.). University College London provided IgG4 monoclonal antibody (mAb) expressed in Chinese hamster ovary (CHO) cell line. The Centre for Genetic Engineering and Biotechnology (CIGB; Havana, Cuba) provided recombinant human erythropoietin (rhEPO) expressed by Chinese Hamster Ovary (CHO) cell lines.
increased glycan hydrophobicity enabling C18 chromatography analysis, and (vi) faster profiling and analysis times, in comparison to, for example, UHPLC analysis.17−19 Relative and absolute quantitation can also be achieved through permethylation by introducing isotope-labeled internal standards.20−22 However, when analyzing large numbers of samples, the conventional manual in-solution permethylation technique would be labor-intensive with long turnaround times. Various research groups have improved the permethylation technique to achieve a simplified and enhanced sample preparation workflow. Examples include the widely used solid phase micro spin columns technique published by Kang et al.,23 whereas Jeong et al. reported 96-well plate HT format24 and Gao et al. developed the open tubular capillary reactor technique.25 Most of these methods described are still largely manual, which is not desirable for analysis of large cohorts and appear limited in securing the long-term stability of the method (e.g., for the characterization of biopharmaceuticals). Our aim was to develop a fully validated comprehensive system of commercially available kits, which are easy to use, robust, and suitable for large numbers of sample to perform quality control for biopharmaceuticals. This prompted us to develop a robotics compatible product: a microplate-based permethylation kit, which is automated, HT, and reliable. In this paper, we present the HT and largely automated workflow for (i) N-glycan release, (ii) hydrophilic liquid interaction chromatography (HILIC)-solid phase extraction (SPE) of glycans, (iii) permethylation of released glycans, (iv) liquid− liquid extraction (LLE), (v) data acquisition, and (vi) semiautomated data analysis. The HT permethylation kit can be used to process up to 96 samples in parallel using either a manual method or an automated method using a liquid handling robot. In this article we demonstrate how the automated HT method was successfully applied to the analysis of N-glycans from glycan standards, human polyclonal IgG, biopharmaceutical IgG monoclonal antibody (mAb) and recombinant human erythropoietin (rhEPO) samples, and the analysis of O-glycans from rhEPO. We also present data comparing the automated HT permethylation method to an automated UHPLC data obtained after fluorescent labeling (which is referred to as the gold standard method for glycan analysis).26
■
METHOD DESCRIPTION Sample Preparation Prior to Permethylation. NGlycans were released from glycoproteins and enriched using a HILIC SPE filter plate using the liquid handling robot. The O-glycans from rhEPO and fetuin glycoproteins were released as reported previously.27 For the detailed method description of IgG4 mAb affinity purification, N- and O-glycan release, HILIC-SPE glycan enrichment, and manual in-solution permethylation, see SI, experimental protocols, section 1.1. Automated HT Permethylation (96-Well Microplate Format). The sample preparation for permethylation has been automated on the liquid handling robot with the exception of few steps such as off-deck plate sealing and incubation steps. The liquid handling workstation used in this study has 8 independent pipetting channels and an integrated vacuum manifold system. The programs for the automated methods were created using Hamilton Microlab VENUS 2 base package 4.3 software. A 96-well PCR plate containing the enriched and dried N-glycan samples and a LT-PERMET-96 permethylation plate were both placed on the robot deck. Two 150 μL aliquots of dimethyl sulfoxide (DMSO) were dispensed into the 300 μL PCR sample plate; the contents were mixed by pipetting action and transferred to the 96-well format permethylation plate. The permethylation plate was sealed with a 96-well silicone plate lid and incubated at RT for 15 min on a plate shaker off the robot deck. Following the incubation step, the seal was removed after brief centrifugation and the permethylation plate was placed back on the deck where 55 μL of methyl iodide (MeI) was added to each reaction well. Finally, the plate was removed from the robot deck, sealed again with 96-well plate lid and placed on a shaker incubator for 60 min at RT. This step takes approximately 2.5 h to perform, including the incubation steps. Automated Liquid−Liquid Extraction (LLE). After incubating the samples for 1 h, the LT-PERMET-96 permethylation plate was placed back onto the robot deck to perform the LLE steps. A total of 400 μL of dichloromethane (DCM) was applied to each sample well after prewetting the tips to support the dispensing of volatile liquid, followed by 245 μL of water with resistivity of 18.2 MΩ. The resulting twophase solution was mixed thoroughly by pipetting action. The samples were then transferred to a 2 mL deep well collection plate and an additional 750 μL of water was added to the deep well plate followed by mixing by pipetting in order to perform LLE. The program on the robot was set up to perform LLE four times in a loop setting, where 750 μL water was added to the deep well collection plate and mixed by pipetting action during each cycle, and after 2.5 min wait step each time, 1000 μL of the top aqueous layer was discarded to extract salts and render the pH of the aqueous layer neutral prior to mass spectrometry. In the final extraction step, with liquid level
■
EXPERIMENTAL SECTION Ethanol, formic acid, acetonitrile (ACN), methanol, sodium hydroxide (NaOH), and 9:1 2,5-dihydroxybenzoic acid and 2hydroxy-5-methoxybenzoic acid (super-DHB) were obtained from Sigma (Dorset, U.K.). Reagents for affinity purification, in-solution manual permethylation, N- and O-glycan release, fluorescent labeling, and post-labeling cleanup are mentioned in the Supporting Information (SI), materials section. PCR plates, foil pierce seals, plate sealer, polypropylene collection plates, and silicone plate lids were purchased from 4titude (Surrey, U.K.). HT permethylation kit (LT-PERMET-96) and LudgerClean prepermethylation cleanup plate (LC-PERMET-96, a HILIC-SPE plate housing polypropylene membrane used for glycan enrichment) were obtained from Ludger (Oxfordshire, U.K.). Samples were dried down in a Thermo Savant centrifugal evaporator from Thermo (Hampshire, U.K.). All automated steps in the analytical workflow described were performed using a Hamilton MICROLAB STARlet Liquid Handling Workstation, from Hamilton Robotics Inc. (Bonaduz, Switzerland) MALDI-TOF-MS data acquisition was performed B
DOI: 10.1021/acs.analchem.6b01639 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Scheme 1. Automated and HT Streamlined Workflow for Glycoprotein Sample Preparation and Permethylation for Analysis Using MALDI-TOF-MS
detection setting activated, 650 μL of top aqueous layer is discarded twice to waste, leaving the hydrophobic permethylated glycans present in the lower organic DCM layer ready for centrifugal evaporation. This step of LLE of 96 samples could be performed in 2.5 h using the liquid handling robot. Sample Drying in Centrifugal Evaporator. After the automated LLE was completed, the deep well collection plate was placed in a centrifugal evaporator and spun for 2 min without any vacuum setting switched on to allow the DCM to settle. After 2 min, the vacuum pump was turned on and the contents were dried down completely. MALDI-TOF-MS. The dried permethylated samples were resuspended in 10 μL of 70% methanol. Matrix solution (0.5 μL of 5 mg/mL super-DHB with 1 mM NaOH in 50% ACN) was spotted manually on a Bruker, MTP Anchor-Chip 384 well MALDI-target plate followed by an equal volume of permethylated sample, and allowed to dry. Data acquisition was performed three times for each sample from three different spots on the MALDI-target plate, to obtain average relative intensities for glycan peaks. However, for the intraday and interday variation experiments the samples were spotted once on the MALDI target. Data acquisition of one sample spot took less than 1 min on average. For MALDI-TOF-MS data acquisition and sample processing details, see SI, experimental protocols, section 1.1. For data analysis, we used an automated data extraction tool employing a previously published Python script.28,29 Sample Preparation for Automated and HT Fluorescent Labeling. N-Glycans were released from glycoproteins, filtered through a protein binding membrane (PBM) cleanup plate and fluorescently labeled (e.g., 2-aminobenzamide (2-AB) or procainamide dye). All the sample preparation and processing steps were performed on the liquid handling robot
using the previously published method.30 See SI, experimental protocols, section 1.2, for sample preparation for automated and HT fluorescent labeling. Glycan Representation. Glycan structures were visualized by cartoons built with GlycoWorkBench, version 2.1.31 Structures for glycans are depicted following the Consortium for Functional Glycomics (CFG) notation: N-acetylglucosamine (N; blue square), fucose (F; red triangle), mannose (H; green circle), galactose (H; yellow circle), N-acetylneuraminic acid (S; purple diamond), and N-glycolylneuraminic acid (Sg; light blue diamond).
■
RESULTS AND DISCUSSION Automated HT Glycan Analysis Workflow on a Liquid Handling Robot. The increasing demand for detailed characterization of glycosylation during biopharmaceutical development, as well as in glycan biomarker discovery, has necessitated HT and automated technological advancements in glycomics.10,32 Glycan analysis workflows would benefit from expedited sample processing, higher productivity, less sampleto-sample variation, and increased speed and efficiency. MALDI-TOF-MS cannot separate isobaric structures and the coefficients of variation (CVs) for quantification are usually higher for MALDI-TOF-MS measurements compared to UHPLC methods. However, the high sensitivity, ability to perform structural elucidation via fragmentation studies, the ability to analyze samples fast and the HT approach makes MALDI-TOF-MS a valuable tool for glycomics. We, therefore, have developed a largely automated, HT glycan characterization workflow for analysis by MALDI-TOF-MS, as shown in Scheme 1. The Hamilton Microlab VENUS 2 software provides a graphical user interface that was used to setup and write scripts C
DOI: 10.1021/acs.analchem.6b01639 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 1. Comparison of PNGase F released and purified human IgG N-glycans analyzed with orthogonal methods. N-glycans were analyzed by (A) HILIC UHPLC after 2-AB labeling and (B) MALDI-TOF-MS after permethylation. Sample processing and preparation for both methods was performed using the liquid handling robot. Permethylated N-glycans were registered as [M + Na]+.
from 13 major N-glycans were integrated and then normalized to the sum of intensities. The averaged relative intensities (RIs), standard deviations (SDs), and CVs were calculated from triplicate sample analysis using UHPLC and MALDI-TOF-MS, as shown in SI, Table S-1. For MALDI-TOF-MS signals with RIs above 5%, the CVs were
