An automated robotic platform for rapid profiling oligosaccharide analysis of monoclonal antibodies directly from cell culture

Analytical Biochemistry 442 (2013) 10–18

Contents lists available at ScienceDirect

Analytical Biochemistry journal homepage: www.elsevier.com/locate/yabio

An automated robotic platform for rapid profiling oligosaccharide analysis of monoclonal antibodies directly from cell culture Margaret Doherty a,⇑, Jonathan Bones a, Niaobh McLoughlin a, Jayne E. Telford a, Bryan Harmon b, Michael R. DeFelippis b, Pauline M. Rudd a a b

National Institute of Bioprocessing Research and Training (NIBRT), Mount Merrion, Blackrock, Dublin 4, Ireland Bioproduct Research and Development, Lilly Research Laboratories, Eli Lilly, Indianapolis, IN 46285, USA

a r t i c l e

i n f o

Article history: Received 23 May 2013 Received in revised form 26 June 2013 Accepted 1 July 2013 Available online 16 July 2013 Keywords: Monoclonal antibodies Oligosaccharides Automation Glycomics Ultra-performance liquid chromatography Cell culture media

a b s t r a c t Oligosaccharides attached to Asn297 in each of the CH2 domains of monoclonal antibodies play an important role in antibody effector functions by modulating the affinity of interaction with Fc receptors displayed on cells of the innate immune system. Rapid, detailed, and quantitative N-glycan analysis is required at all stages of bioprocess development to ensure the safety and efficacy of the therapeutic. The high sample numbers generated during quality by design (QbD) and process analytical technology (PAT) create a demand for high-performance, high-throughput analytical technologies for comprehensive oligosaccharide analysis. We have developed an automated 96-well plate-based sample preparation platform for high-throughput N-glycan analysis using a liquid handling robotic system. Complete process automation includes monoclonal antibody (mAb) purification directly from bioreactor media, glycan release, fluorescent labeling, purification, and subsequent ultra-performance liquid chromatography (UPLC) analysis. The entire sample preparation and commencement of analysis is achieved within a 5h timeframe. The automated sample preparation platform can easily be interfaced with other downstream analytical technologies, including mass spectrometry (MS) and capillary electrophoresis (CE), for rapid characterization of oligosaccharides present on therapeutic antibodies. Ó 2013 Elsevier Inc. All rights reserved.

Monoclonal antibodies (mAbs)1 represent a large proportion of approved and in-pipeline biotherapeutics for use in oncology, treatment of autoimmune diseases, and the prevention of xenograft rejection [1]. The majority of approved therapeutic mAbs are of the immunoglobulin G1 (IgG1) isotype. IgG1 is composed of four polypeptide chains, two heavy and two lights chains, covalently linked by disulfide bonds [2]. A single highly conserved N-linked glycosylation site occurs at Asn297 in each CH2 domain in the Fc region; however, glycosylation may also occur in the antigen binding region of the IgG molecule [3]. The CH2 glycosylation can play an important role in the monoclonal antibody’s effector functions by modulating the affinity of interaction with Fc receptors or C1q complement displayed on cells of the innate immune system [4,5]. However, the dis-

⇑ Corresponding author. Fax: +353 1 215 8116. E-mail address: [email protected] (M. Doherty). Abbreviations used: mAb, monoclonal antibody; IgG, immunoglobulin G; CGE–LIF, capillary gel electrophoresis with laser-induced fluorescence detection; PNGaseF, peptide N-glycosidase F; APTS, 8-aminopyrene-1,3,6-trisulfonic acid; ESI, electrospray ionization; HILIC, hydrophobic interaction liquid chromatography; SPE, solid phase extraction; UPLC, ultra-performance liquid chromatography; HTS, high-throughput screening; CE, capillary electrophoresis; MS, mass spectrometry; HPLC, high-performance liquid chromatography; 2-AB, 2-aminobenzamide; PBS, phosphate-buffered saline solution; GU, glucose unit. 1

0003-2697/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ab.2013.07.005

tributions of oligosaccharide structures present on mAbs can vary significantly depending on the bioprocess parameters used, including the cell line used for production [6], the amount of dissolved oxygen [7,8], the nutrients available [9], and the manufacturing mode (i.e., batch-fed or perfusion-based bioreactors) [10,11]. Regulatory authorities expect thorough characterization of the glycosylation profile and understanding of its relationship to the bioprocessing parameters. In certain cases, routine testing may be required to demonstrate that glycan structures are maintained within specific ranges in order to confirm batch-to-batch consistency and ensure product safety and efficacy [12]. The rapidly expanding mAb market has led to an increased demand for the development and implementation of high-throughput analytical technologies for the characterization of the glycosylation at all stages of bioprocess development and production. This demand is set to increase further, especially with the emergence of biosimilar products [13–15]. Current limitations in glycosylation analysis include the hands-on time required for sample preparation and analysis. There has been a committed effort to address this shortfall, and an integrated microfluidic chip for glycan profiling of mAbs with combined sample preparation and sample analysis time has been developed with a timeframe of just 10 min [16]. Another strategy proposed by Ruhaak and coworkers

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

performs N-glycan analysis by capillary gel electrophoresis with laser-induced fluorescence detection (CGE–LIF) employing a multicapillary format using a DNA sequencer following N-glycan release by peptide N-glycosidase F (PNGaseF) [17]. Optimization of the derivatization reaction was accomplished by varying the concentrations of the APTS (8-aminopyrene-1,3,6-trisulfonic acid) labeling reagent as well as the type and amount of reducing agent. Attention also has been focused on increasing the speed of the derivatization reaction, and citric acid has proved to be a more powerful catalyst than the conventionally used acetic acid [18]. More recently, an automated method was developed based on glycopeptide analysis consisting of two steps. First IgG was purified out of the supernatant, followed by a tryptic digest. The glycopeptides were subsequently purified from the tryptic digest and analyzed [19]. Although viable methods for particular applications, these methods also have several drawbacks; the preparation of the APTS-labeled N-glycans prior to multiplexed CGE–LIF requires two overnight incubations, and there is also the requirement for synthetic standards for each glycan analyzed [17]. Using the mAb chip, the method requires optimization for each sample analyzed, and the potential for cross-contamination could be a concern because the same enzyme is used repeatedly and could have major implications when performing analysis in biopharmaceutical processes. In positive mode electrospray ionization (ESI), this microfluidic chip method uses the glucosylamine intermediate glycan to achieve improved sensitivity [16]. However, these intermediates are relatively unstable even under mild acidic conditions. Furthermore, both anomeric forms of the free reducing end glycan are present. To obtain accurate quantitation, it is preferable to use a single glycan standard rather than a potentially unstable glucosylamine intermediate. In addition, glycans may exist as both protonated and sodiated adducts using positive mode ESI, further complicating the characterization process. Previously, Royle and coworkers [20] described a 96-well plate analytical platform that included sample immobilization, enzymatic N-glycan release, fluorescent labeling, and quantitative hydrophobic interaction liquid chromatography (HILIC)–fluorescence-based profiling. In the current article, we describe further development, modification, and automation of the entire procedure using a robotic platform that is capable of sample analysis directly from a bioreactor. mAbs were purified and captured from cell culture media on protein A resin in a 96-well plate format. After washing, the glycans were removed enzymatically via N-glycanase while the mAbs were still immobilized on the protein A. The released glycans were fluorescently labeled, and excess underivatized material was removed via solid phase extraction (SPE) using a synthetic polyamide stationary phase packed in 96-well plates. Furthermore, method transfer to ultra-performance liquid chromatography (UPLC) using a 1.7-lm HILIC phase allowed for a 12-fold reduction in analysis time, from 180 min to just 15 min. Using the automated platform, the time taken to achieve this overall process is approximately 5 h as opposed to a number of days for classical analysis. The automated platform is directly suited to the biopharmaceutical industry, where high-throughput methods are required for development activities such as highthroughput screening (HTS) of clones for cell line selection and design of experiment (DOE) studies supporting quality by design (QbD) approaches to understanding the impact of process parameters on product quality. A particular advantage of this automated method is its amenability to multiplexing with other analytical devices such as direct coupling to UPLC, capillary electrophoresis (CE), and mass spectrometry (MS). This high-throughput strategy bridges the interface among bioprocessing operations, sample preparation, and analysis, allowing for rapid data generation for the processes under optimization.

11

Materials and methods Chemicals Reagent water used throughout this study was obtained from a MilliQ Gradient A10 Elix system (Millipore, Bedford, MA, USA) and was 18.2 MX or greater with a total organic carbon (TOC) content less than 5 parts per billion (ppb). All solvents used were high-performance liquid chromatography (HPLC) gradient grade and received from Sigma–Aldrich (Dublin, Ireland). Chemicals were of the highest possible grade and were obtained from Aldrich. Protein A plates were purchased from Pierce (Rockford, IL, USA). Clarified Chinese hamster ovary (CHO) cell culture supernatant was provided by Eli Lilly (Indianapolis, IN, USA). PNGaseF was purchased from Prozyme (Hayward, CA, USA). 2-Aminobenzamide (2-AB) Nglycan labeling was performed using the 2-AB LudgerTag labeling kit (Ludger, Oxfordshire, UK). Liquid handling parameter programming All methods and scripts were programmed using Vector, the Hamilton MicroLab STAR software (Reno, NV, USA). A graphical user interface facilitated the setup of both the glycan release and sample preparation method and the deck layout of the platform. All reagents were loaded into the appropriate reservoirs, and interactive functionality prompted the user to conduct visual checks on the platform before progressing with the liquid handling. All errors were detected by the automatic error response device, allowing the user to respond appropriately. Liquid handling steps were performed employing the 8-channel head, whereas sample transfer was accomplished by operating the 96-channel head. Disposable tips were used at all times. Test methods were run in simulation mode before initiating the actual run to maximize productivity and minimize any potential errors. Automated glycan preparation on a robotic platform Protein A plates (Pierce) were equilibrated with 200 ll of phosphate-buffered saline solution (PBS) and vacuumed to waste. Briefly, 200 ll of clarified bioreactor cell culture supernatants was applied to the protein A plate and incubated at room temperature for 10 min to enable IgG1 capture. The plate was washed with 200 ll of PBS twice to remove unbound contaminants and vacuumed to waste. N-Glycans were released from the IgG while immobilized to the protein A resin with the addition of 50 ll of 0.25 U/ml PNGaseF (Prozyme, San Leandro, CA, USA) in 20 mM NaHCO3 (pH 7.2). The samples were moved to the integrated 37 °C incubation via the iSWAP robotic plate handler using the CORE gripper tools and incubated for 60 min. Glycans were eluted with 5  100 ll of water and concentrated in a centrifugal evaporator (Thermo, Basingstoke, Hampshire, UK). The eluted glycans were labeled with 2-AB using the LudgerTag 2-AB kit. Glycan labeling and cleanup Labeling was performed via reductive amination using the LudgerTag 2-AB labeling reagent kit prepared according to the manufacturer’s instructions. In all instances, 5 ll of labeling reagent was added to each dried glycan sample in a deep well plate on the robotic platform, and the plate was subsequently transferred to the integrated incubator and incubated at 65 °C for 2 h. All samples were prepared in triplicate for each experiment. Postlabeling sample cleanup was accomplished by conditioning a 96-well SPE plate packed with 25 mg of the synthetic polyamide polymer DPA-6S material (Sigma–Aldrich, Poole, Dorset, UK) with

12

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

acetonitrile using the 8-tip pipettor arm on the liquid handling platform. The SPE plate was conditioned by washing three times with 1 ml of 95% (v/v) acetonitrile, followed by three washes with 20% (v/v) acetonitrile and finally three washes with 95% (v/v) acetonitrile. Labeled glycan samples were diluted via the addition of 95 ll of MilliQ water and 900 ll of acetonitrile. Solutions were added to the 96-well SPE plate under mild vacuum. Excess fluorescent label was removed by five successive washes of 1 ml of 95% (v/v) acetonitrile. Glycans were eluted into a deep well collection block using five individual applications of 100 ll of MilliQ water. All eluants were combined and concentrated to dryness in a centrifugal evaporator. Exoglycosidase digestions were performed according to the method detailed by Royle and coworkers [21]. A combination of exoglycosidase digestions and comparison of glucose unit (GU) values with GlycoBase was used in structural assignment [22]. N-Glycan nomenclature and symbolic representations used in this article were described previously by Harvey and coworkers [23].

HILIC–fluorescence and UPLC–fluorescence 2-AB-labeled N-glycans were analyzed on a Tosoh TSKgel Amide 80 250  4.6-mm i.d., 5-lm particle-packed analytical column on a Waters Alliance 2695 Separations Module with a Waters 2475 Multi Wavelength Fluorescence Detector under the control of Empower chromatography workstation software (Waters, Milford, MA, USA). A linear gradient of 50 mM ammonium formate buffer (pH 4.4) (mobile phase A) and acetonitrile (mobile phase B) was used for glycan separation. A linear gradient of 80 to 42% acetonitrile at a flow rate of 0.4 ml/min over 152 min was employed, with a total analytical runtime of 180 min with reequilibration. The detection wavelengths were kex = 330 nm and kem = 420 nm. This method was transferred to UPLC using a Waters Acquity UPLC instrument under the control of Empower 2 software (Waters). Separations were performed on a Waters BEH glycan column (2.1  150 mm, 1.7-lm particles) at a temperature of 40 °C. An injection volume of 20 ll was used throughout, and samples were prepared in 80% (v/v) acetonitrile and maintained at 5 °C prior to injection. A linear gradient of 65 to 58% acetonitrile over 12 min at a flow rate of 0.37 ml/min was used, with a total analytical runtime of 15 min with reequilibration. For both HPLC and UPLC, the dextran ladder was fitted with a fifth-order polynomial distribution curve that was used to allocate GU values following retention time-based normalization.

Statistical analysis Results are expressed as means ± standard errors, and data were compared using the unpaired Student’s t test. All statistics were calculated using GraphPad Prism version 6.0. Groups of data were considered to be significantly different if P < 0.05.

Results and discussion Automating glycan preparation strategy on liquid handling platform The acceleration of the prolific market for therapeutic glycoproteins has necessitated technological developments in the glycomics domain. The ability to profile glycans in a high-throughput manner facilitates the bioprocessing industry in screening large numbers of cell lines and rapidly extracting the most promising conditions. In general, N-linked glycans are covalently attached to glycoproteins, extending away from the polypeptide surface and playing key roles in many biological functions. However, in IgG, the N-glycans are sequestered within the interstitial space between the CH2 domains in the Fc region. They play a role in effector functions of the immunoglobulin as well as the structural stability [5]. Here, a Hamilton Microlab STAR liquid handling platform (Fig. 1) was adapted to automate a high-throughput glycan preparation method for the analysis of monoclonal antibodies directly from cell culture media. Immunoglobulins were immobilized on protein A prior to deglycosylation via PNGaseF. N-Glycans were subsequently labeled with 2-AB and analyzed by UPLC with fluorescence detection. Deck modification included two custom-built integrated incubators for both the deglycosylation and labeling incubation steps. A modified insert for the vacuum manifold was also required for the deep well collection plate to ensure that there was no cross-contamination between wells. The Hamilton STAR platform offers patented tip coupling technology, interchangeable disposable tips, monitored air displacement pipetting, independently spreadable pipetting channels, and enhanced throughput options [24]. The disposable tips used for this liquid handling platform have integrated conductivity probes that allow for liquid level detection. These functions, when combined, greatly enhance the precision of the method, eliminating human error. All liquid transfer steps were automatically performed by the robotic liquid handling system, resulting in reduced sample preparation time, simultaneously reducing the possibility of human error as time-consuming and labor-intensive manual pipetting steps were

Fig.1. Layout of the deck custom built to facilitate our automated method.

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

eliminated. This platform also included an integrated vacuum manifold, a plate shaker, and two custom integrated incubators that allowed for unattended capabilities. Scripts for the method of the developed glycan workflow were written using the Hamilton software Vector (Reno, NV, USA). (These scripts can be made available by contacting the corresponding author.) The newly modified deck layout was programmed into the software prior to writing the scripts for the glycan method. All consumables used were preprogrammed, specifying liquid classes and the dimensions of all plates used. Acceleration of current processes for automation Current sample preparation workflows were examined to assess the possibility of automation. Within these procedures, a number of time-consuming method steps were identified for modification in the liquid handling robotic platform. The majority of therapeutic

13

mAbs lack Fab glycosylation [25]. In this study, Fc glycosylation was the focus because it represents the most common form in biotherapeutics and none of the mAbs had Fab glycosylation present. Immobilization of IgG on protein A prior to glycan cleavage The time required for the N-glycanase digestion was subsequently investigated. To expedite the process of purification of IgG from bioreactor cell culture supernatants, protein A was used in a 96-well plate high-throughput format. Protein A was chosen to exploit the strong affinity of protein A to bind to the Fc portion of IgG molecules and efficiently purify the immunoglobulin from the cell culture media. IgG from clarified medium was captured on protein A resin in each of the wells of the filter plate. Glycans were successfully released from the immunoglobulin using PNGaseF while immobilized on protein A. This was compared directly with purifying the IgG from the protein A first and subsequently

Fig.2. (A) Cleavage of N-glycans after purification of IgG from protein A. (B) Removal of glycans while IgG immobilized to protein A. (C) Protein A binding to IgG. Protein A helix is depicted in purple and coil in blue. The Fc component of IgG1 is illustrated in yellow. PBD ID: 1FC2 [26,37].

14

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

cleaving the glycans (Fig. 2) and generated perfectly superimposable chromatograms. The relative percentage of the core fucosylated biantennary structure, F(6)A2, present after IgG was purified via protein A was 64.8 ± 0.7% compared with 61.7 ± 0.6% when glycans were released while the antibody was still immobilized to protein A. The relative percentage area of Man 5 was 4.8 ± 0.4% in the released sample compared with 4.4 ± 0.4% in the immobilized sample. Using a t test, there was also no significant difference in the total percentage areas of galactosylated structures. F(6)A2[6]G1, F(6)A2[3]G1, and FA2G2 were 27.3 ± 1.1% in the released samples versus 30 ± 0.2% in the immobilized samples. These results indicate that the glycans were successfully released from the immobilized IgG, eliminating the need to release the IgG prior to cleavage of the glycans present on the molecule. It was initially thought that immobilizing the substrate on protein A and removing the glycans may have been a challenge because the accessibility of PNGaseF to the N-glycans may have been compromised. However, PNGaseF was successful in deglycosylation even though the N-glycans located within the interstitial space between the CH2 domains are in close proximity to the binding site of protein A (see Fig. 2C) [26]. Immobilization of enzymes has recently gained attention in integrated glycan preparation workflows. Palm and Novotny [27] demonstrated that by incorporating N-glycanase into a monolithic

column, deglycosylation of small and medium-size glycoproteins can occur within minutes. More recently, this strategy was developed further by using a monolithic reactor in a capillary format with immobilized PNGaseF, which effectively deglycosylated even large proteins [28]. Another approach reported by Bynum and coworkers involved immobilization of PNGaseF in a microfluidic chip format, where 98% of the antibody was deglycosylated within 6 s [16]. The advantages of immobilizing the substrate instead of the enzyme are that there is no concern regarding enzyme stability or performance with time and no potential carryover or contamination issues, which is especially relevant in a biopharmaceutical setting. This procedure also has the benefit of simplifying the overall process, reducing the number of steps that need to be taken, and accelerates the overall strategy. Optimization: Eliminating reduction and alkylation steps Conventionally, samples are reduced and alkylated prior to deglycosylation. However, not all proteins require a reduction and alkylation strategy for deglycosylation, as results here demonstrate, and this must be investigated for each IgG analyzed. Direct comparison of samples with and without reduction and alkylation was examined as a means for accelerating the overall process. Results

A

B

D

A2 F(6)A2 M5 F(6)A2[6]G(4)1 F(6)A2[3]G(4)1 F(6)A2G2

C

Fig.3. (A, B) Investigation of reducing and alkylating the sample (A) versus eliminating the reduction and alkylation step (B). (C) Comparison of the relative percentage areas of N-glycans employing each method. (D) N-Glycan nomenclature.

15

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

indicated quantitatively and qualitatively comparable glycosylation profiles in the presence (Fig. 3A) and absence (Fig. 3B) of reducing and alkylating agents. Subsequent statistical analysis via a t test indicated no significant difference in the relative percentage areas of glycans present. The relative percentage of core fucosylated biantennary structure F(6)A2 present in the reduced samples was 42.6 ± 0.8% compared with the removal of this step, 43.1 ± 1.2%. The average percentage of mannosylated structures present in the reduced samples was 7.3 ± 1.5% compared with 7.1 ± 0.3% in the nonreduced samples. The biantennary galactosylated structures F(6)A2[6]G1, F(6)A2[3]G1, and FA2G2 were 32.6 ± 1.4, 7.4 ± 0.3, and 8.5 ± 0.5%, respectively, in the reduced samples versus 32 ± 0.3, 7.2 ± 0.4, and 7.7 ± 0.2%, respectively, in the nonreduced samples. The reduction and alkylation steps were eliminated from all further experiments on the robotic platform to expedite the overall process because there was no significant difference in the relative percentage areas for the studied immunoglobulins.

Optimization of deglycosylation time Classically, N-glycans are cleaved during an overnight incubation at 37 °C in the presence of PNGase to ensure complete deglycosylation of any glycoprotein (Fig. 4). As displayed in Fig. 4, results indicate that for IgG this deglycosylation time can be reduced to 60 min in a high-throughput format without loss of yield. This incubation time can be further decreased by increasing the amount of PNGaseF; however, this leads to increased expense and an overall loss in yield. The recovery of the glycans from the longer incubation time was comparable to the recovery of glycans from the decreased incubation time. As shown in Table 1, there were no significant differences in the relative percentage areas of the glycans between each of the two incubation times investigated. The percentage of core fucosylation corresponded to 66.1 ± 0.7% of the total peak area after 16 h of incubation in the presence of PNGaseF compared with 61.5 ± 0.4% of the total peak area after just 60 min of incubation. This may be due to slight losses in the sample preparation, or the unsubstantial increase in galactosylated structures may account for the decrease in core fucosylation in these samples because the areas are all relative. The combined relative percentage area of galactosy-

lated structures F(6)A2[6]G1, F(6)A2[3]G1, and FA2G2 was 27.2 ± 0.8% with the longer incubation time compared with 30.1 ± 0.2% with the significantly shorter incubation time. The relative percentage area of Man5 was 3.4 ± 0.9% after 16 h of incubation and 4.9 ± 0.2% after 60 min in the presence of the N-glycanase. This incubation was facilitated by the modification of the deck with a custom-made integrated incubator on the robotic platform, allowing for complete automated integration. This is consistent with other studies investigating the acceleration of deglycosylation [16,28,29]. Bynum and coworkers [16] described using immobilized PNGaseF to deglycosylate mAbs in 6 s. With a total turnaround time of 10 min per sample, this results in 6 samples being processed per hour. However, using the method described in this manuscript, N-glycans from 96 samples can be efficiently removed within just 60 min, enabling higher sample throughput. Svec and coworkers used immobilized PNGaseF to deglycosylate IgG in 5.5 min [28]; however, this is not an automated strategy. To facilitate high-throughput glycan analysis, the deglycosylation strategy needs to be amenable to automation, and the method developed and described here addresses that need. All further experiments used a 60-min incubation time for deglycosylation of the molecule. To verify the efficiency of the deglycosylation step, we also performed multiple PNGaseF incubations on the same immobilized sample. Analysis of each of the resulting releases indicated that the enzymatic liberation of N-glycans was complete after a single PNGaseF incubation, with no detectable glycans present at the expected retention times in the chromatograms of each subsequent enzymatic incubation (data not shown). Table 1 Comparison of relative percentage areas when deglycosylation time is reduced. Peak

Glycan structure

16-h incubation % area

1-h incubation % area

1 2 3 4 5 6

A2 F(6)A2 M5 F(6)A2[6]G(4)1 F(6)A2[3]G(4)1 FA2G2

3.0 ± 0.6 66.1 ± 0.7 3.4 ± 0.9 18.9 ± 0.2 4.8 ± 0.1 3.5 ± 0.5

3.3 ± 0.2 61.5 ± 0.4 4.9 ± 0.2 22.1 ± 0.1 3.8 ± 0.05 4.2 ± 0.07

Fig.4. Investigation of reducing the deglycosylation time while the substrate is immobilized to protein A.

16

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

Removal of excess underivatized material Reducing the 2-AB incubation time was investigated; however, it was revealed that this required a 2-h incubation step to avoid selective labeling of the oligosaccharides (data not shown). Several products were evaluated to find an efficient means of postderivatization cleanup retaining the minimum amount of contaminating free fluorescent dye. Microcrystalline cellulose has been used successfully in HILIC mode for the cleanup of 2-anthranilic acid (2-AA) derivatives of N-glycans [30]. Another more recent approach has been the use of cotton wool for the preparation of HILIC SPE in a microtip format to remove salts, nonglycosylated peptides, and detergents [31]. Classically, Whatmann 3MM chromatography paper has been used to remove excess 2-AB via hydrophilic interaction [20,32]. The sample is adsorbed onto the cellulose paper and dried, and excess 2-AB is separated from the sample using an acetonitrile mobile phase, as outlined by Royle and coworkers [20]. In an attempt to use this method for automated high-throughput purposes, 96-well plates packed with cellulose paper were custom made. However, results indicated that the sample was not sufficiently adsorbed prior to dye removal, resulting in sample loss (data not shown). This could be due to the fact that drying was inadequate in the 96-well plate format. Fiber glass plates were also

investigated and were successful in removing the excess dye; however, the glycans were not fully recovered (data not shown). This may be due to the sugars binding strongly to the borosilicate in the presence of a high percentage of solvent. The SPE plate packed with 25 mg of synthetic polyamide DPA-6S stationary phase outperformed all other tested formats. Hence, it was considered to be the method of choice for the effective removal of excess underivatized material in all subsequent experiments. A direct comparison between these two different SPE 96-well plates is displayed in Fig. 5A, which illustrates how remaining dye can mask the detection of glycans comigrating with the large overloaded dye front. A more efficient purification by the DPA-6S 96-well plate is depicted with a dotted line in Fig. 5A, where excess derivatization reagent is efficiently removed and does not interfere with peak annotation and quantitation of the glycans present. This plate contained polyamide DPA-6S resin, which when used in the hydrophilic interaction mode substantially reduced the amount of excess label and did not interfere with glycan detection. To ascertain whether all N-glycans were initially retained and then successfully recovered from the resin, the flow-through was collected and analyzed by HPLC. Negligible amounts of N-glycans were present in the background noise in the resulting chromatogram corresponding to the collected SPE flow-through fraction. In

Fig.5. (A) Comparison of two SPE phases for the removal of excess fluorescent label. (B) Determination of optimal elution volume for DPA-6S 96-well plate.

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

17

Fig.6. (A) Comparison of time taken for N-glycan analysis on HPLC versus UPLC. (B) Overlays of eight independent releases of cell culture media analyzed via UPLC.

Fig. 5B, the optimal elution volume was determined by collecting each wash and analyzing the glycans present. A series of five successive washes of 100 ll each was determined to be the optimal procedure for successful recovery of the glycans present. The advantage of this particular SPE cleanup method is its versatility because it can also be readily applied for the purification of underivatized glycans for complementary analytical techniques such as MS analysis and APTS-labeled glycans for CE–LIF analysis [33]. Reduction of HILIC separation time UPLC is fast becoming the separation method of choice due to the importance in increasing throughput with a more rapid and efficient analysis, allowing faster detection of problems in production and decreasing analytical costs in the biopharmaceutical industry [34–36]. To significantly reduce analysis time and maintain separation efficiency the separation methods were transferred from HPLC to UPLC. The chromatographic separation was achieved on a Waters Acquity UPLC BEH glycan analytical column (150  2.1 mm, 1.7-lm amido functionalized BEH hybrid silica particles). The time taken for the separation of the glycan pool was reduced from 180 min (Fig. 6A) to 15 min, thereby increasing throughput while maintaining separation efficiency. Statistical t

test analysis revealed no significant differences in the relative percentage areas of each of the glycan structures when analyzed on HPLC compared with UPLC. The peak eluting at a retention time of 79 min via the HPLC separation (Fig. 6A) corresponds to the afucosylated biantennary structure A2; this peak elutes in less than 4 min using the UPLC separation (Fig. 6A). The dominant structure, the core fucosylated biantennary structure F(6)A2, has a relative percentage area of 65 ± 0.8% and eluted at approximately 83 min when analyzed by HPLC; using UPLC, F(6)A2 corresponds to 61.4 ± 1.6% of the total glycan pool and has a retention time of approximately 4.2 min. The fucosylated mono- and digalactosylated glycans F(6)A2[6]G1, F(6)A2[3]G1, and FA2G2 corresponded to 19 ± 0.3, 4.7 ± 0.1, and 3.4 ± 0.4%, respectively, when the glycan pool was separated by HPLC. When the same mAb was separated on ULPC, F(6)A2G[6]1, F(6)A2G[3]1, and FA2G2 accounted for 21 ± 1.5, 5.8 ± 0.5, and 3.6 ± 0.3%, respectively, of the total glycan pool. These results demonstrate a 12-fold decrease in the time taken to analyze one sample via UPLC while maintaining separation efficiency. Glycans from eight IgG samples isolated on protein A from cell culture media were released and analyzed via UPLC (Fig. 6B). The standard deviations of the peak areas were compared and were on average less than 1% for each of the integrated peaks. Furthermore, the releases yielded superimposable chromatograms,

18

Automated oligosaccharide profiling / M. Doherty et al. / Anal. Biochem. 442 (2013) 10–18

and quantitatively there was no significant difference in the relative percentage areas of the glycoforms present. Conclusion We have described the successful automation of a highthroughput strategy for glycan preparation and analysis suitable for the biopharmaceutical industry. The automated workflow allows for direct capture of the mAb molecule of interest to facilitate glycan cleavage in a more rapid manner. The time taken for deglycosylation to occur was successfully reduced from 16 h to just 60 min. Once the glycans were released, they were labeled with a fluorophore and excess label was removed prior to analysis by UPLC. The glycan analysis was successfully decreased from 180 min via HPLC to just 15 min (including reequilibration) via UPLC while maintaining efficient separation and resolution. The total time taken for this glycan workflow is accomplished within a 5h timeframe, making it a viable option for an automated glycan preparation and analysis approach. This automated strategy has numerous advantages over manual interaction. First, the low volumes required result in lower cost, and the screening capabilities allow faster cell line and clone selection and cell culture process optimization in the biopharmaceutical industry, thereby reducing overall expenditure. The opportunity for parallel screening using a larger liquid handling platform would further accelerate the automated process development of recombinant therapeutics. The UPLC gradient method described here combined with the successful reduction in time taken for glycan preparation outlined in this article and its automation on a robotic liquid handling platform makes this a viable tool for the biopharmaceutical industry.

[12] [13] [14] [15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24] [25]

References [1] R.O. Duthaler, B. Ernst, R. Fischer, A.G. Katopodis, W. Kinzy, W. Marterer, R. Oehrlein, M.B. Streiff, G. Thoma, In vivo neutralization of naturally existing antibodies against linear a(1,3)-galactosidic carbohydrate epitopes by multivalent antigen presentation: a solution for the first hurdle of pig-tohuman xenotransplantation, Chimia (Aarau) 64 (2010) 23–28. [2] R.R. Porter, The structure of antibodies: the basic pattern of the principal class of molecules that neutralize antigens (foreign substances in the body) is four cross-linked chains: this pattern is modified so that antibodies can fit different antigens, Sci. Am. 217 (1967) 81–87. passim. [3] J.J. van Bueren, T. Rispens, S. Verploegen, T. van der Palen-Merkus, S. Stapel, L.J. Workman, H. James, P.H. van Berkel, J.G. van de Winkel, T.A. Platts-Mills, P.W. Parren, Anti-galactose-a-1,3-galactose IgE from allergic patients does not bind a-galactosylated glycans on intact therapeutic antibody Fc domains, Nat. Biotechnol. 29 (2011) 574–576. [4] R. Jefferis, Antibody therapeutics: isotype and glycoform selection, Expert Opin. Biol. Ther. 7 (2007) 1401–1413. [5] R. Jefferis, Recombinant antibody therapeutics: the impact of glycosylation on mechanisms of action, Trends Pharmacol. Sci. 30 (2009) 356–362. [6] C.H. Chung, B. Mirakhur, E. Chan, Q.T. Le, J. Berlin, M. Morse, B.A. Murphy, S.M. Satinover, J. Hosen, D. Mauro, R.J. Slebos, Q. Zhou, D. Gold, T. Hatley, D.J. Hicklin, T.A. Platts-Mills, Cetuximab-induced anaphylaxis and IgE specific for galactose-a-1,3-galactose, N. Engl. J. Med. 358 (2008) 1109–1117. [7] V. Restelli, M.D. Wang, N. Huzel, M. Ethier, H. Perreault, M. Butler, The effect of dissolved oxygen on the production and the glycosylation profile of recombinant human erythropoietin produced from CHO cells, Biotechnol. Bioeng. 94 (2006) 481–494. [8] E. Trummer, K. Fauland, S. Seidinger, K. Schriebl, C. Lattenmayer, R. Kunert, K. Vorauer-Uhl, R. Weik, N. Borth, H. Katinger, D. Muller, Process parameter shifting: I. Effect of DOT, pH, and temperature on the performance of Epo-Fc expressing CHO cells cultivated in controlled batch bioreactors, Biotechnol. Bioeng. 94 (2006) 1033–1044. [9] N.S. Wong, L. Wati, P.M. Nissom, H.T. Feng, M.M. Lee, M.G. Yap, An investigation of intracellular glycosylation activities in CHO cells: effects of nucleotide sugar precursor feeding, Biotechnol. Bioeng. 107 (2010) 321–336. [10] J. Rodriguez, M. Spearman, T. Tharmalingam, K. Sunley, C. Lodewyks, N. Huzel, M. Butler, High productivity of human recombinant b-interferon from a lowtemperature perfusion culture, J. Biotechnol. 150 (2010) 509–518. [11] E. Trummer, K. Fauland, S. Seidinger, K. Schriebl, C. Lattenmayer, R. Kunert, K. Vorauer-Uhl, R. Weik, N. Borth, H. Katinger, D. Muller, Process parameter shifting: II. Biphasic cultivation—A tool for enhancing the volumetric

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34] [35]

[36]

[37]

productivity of batch processes using Epo-Fc expressing CHO cells, Biotechnol. Bioeng. 94 (2006) 1045–1052. G. Walsh, R. Jefferis, Post-translational modifications in the context of therapeutic proteins, Nat. Biotechnol. 24 (2006) 1241–1252. B.R. Hirsch, G.H. Lyman, Biosimilars: are they ready for primetime in the United States?, J Natl. Compr. Cancer Network 9 (2011) 934–943. H.I. Miller, Why an abbreviated FDA pathway for biosimilars is overhyped, Nat. Biotechnol. 29 (2011) 794–795. G. Dranitsaris, E. Amir, K. Dorward, Biosimilars of biological drug therapies: regulatory, clinical, and commercial considerations, Drugs 71 (2011) 1527– 1536. M.A. Bynum, H. Yin, K. Felts, Y.M. Lee, C.R. Monell, K. Killeen, Characterization of IgG N-glycans employing a microfluidic chip that integrates glycan cleavage, sample purification, LC separation, and MS detection, Anal. Chem. 81 (2009) 8818–8825. L.R. Ruhaak, R. Hennig, C. Huhn, M. Borowiak, R.J. Dolhain, A.M. Deelder, E. Rapp, M. Wuhrer, Optimized workflow for preparation of APTS-labeled Nglycans allowing high-throughput analysis of human plasma glycomes using 48-channel multiplexed CGE–LIF, J. Proteome Res. 9 (2010) 6655–6664. Z. Szabo, A. Guttman, T. Rejtar, B.L. Karger, Improved sample preparation method for glycan analysis of glycoproteins by CE–LIF and CE–MS, Electrophoresis 31 (2010) 1389–1395. D. Reusch, M. Haberger, M.H. Selman, P. Bulau, A.M. Deelder, M. Wuhrer, N. Engler, High-throughput work flow for IgG Fc-glycosylation analysis of biotechnological samples, Anal. Biochem. 432 (2013) 82–89. L. Royle, M.P. Campbell, C.M. Radcliffe, D.M. White, D.J. Harvey, J.L. Abrahams, Y.-G. Kim, G.W. Henry, N.A. Shadick, M.E. Weinblatt, D.M. Lee, P.M. Rudd, R.A. Dwek, HPLC-based analysis of serum N-glycans on a 96-well plate platform with dedicated database software, Anal. Biochem. 376 (2008) 1–12. L. Royle, C.M. Radcliffe, R.A. Dwek, P.M. Rudd, Detailed structural analysis of Nglycans released from glycoproteins in SDS–PAGE gel bands using HPLC combined with exoglycosidase array digestions, Methods Mol. Biol. 347 (2006) 125–143. M.P. Campbell, L. Royle, C.M. Radcliffe, R.A. Dwek, P.M. Rudd, GlycoBase and autoGU: tools for HPLC-based glycan analysis, Bioinformatics 24 (2008) 1214– 1216. D.J. Harvey, A.H. Merry, L. Royle, M.P. Campbell, R.A. Dwek, P.M. Rudd, Proposal for a standard system for drawing structural diagrams of N- and O-linked carbohydrates and related compounds, Proteomics 9 (2009) 3796–3801. M.E. Dilorenzo, C. Timoney, R.A. Felder, Technological advancements in liquid handling robotics, J. Lab. Autom. 6 (2001) 36–40. R. Jefferis, Glycosylation of recombinant antibody therapeutics, Biotechnol. Prog. 21 (2005) 11–16. B. Huang, F.F. Liu, X.Y. Dong, Y. Sun, Molecular mechanism of the affinity interactions between protein A and human immunoglobulin G1 revealed by molecular simulations, J. Phys. Chem. B 115 (2011) 4168–4176. A.K. Palm, M.V. Novotny, A monolithic PNGase F enzyme microreactor enabling glycan mass mapping of glycoproteins by mass spectrometry, Rapid Commun. Mass Spectrom. 19 (2005) 1730–1738. J. Krenkova, N.A. Lacher, F. Svec, Multidimensional system enabling deglycosylation of proteins using a capillary reactor with peptide-Nglycosidase F immobilized on a porous polymer monolith and hydrophilic interaction liquid chromatography–mass spectrometry of glycans, J. Chromatogr. A 1216 (2009) 3252–3259. W.N. Sandoval, V. Pham, E.S. Ingle, P.S. Liu, J.R. Lill, Applications of microwaveassisted proteomics in biotechnology, Comb. Chem. High Throughput Screening 10 (2007) 751–765. L.R. Ruhaak, C. Huhn, W.J. Waterreus, A.R. de Boer, C. Neususs, C.H. Hokke, A.M. Deelder, M. Wuhrer, Hydrophilic interaction chromatography-based highthroughput sample preparation method for N-glycan analysis from total human plasma glycoproteins, Anal. Chem. 80 (2008) 6119–6126. M.H. Selman, M. Hemayatkar, A.M. Deelder, M. Wuhrer, Cotton HILIC SPE microtips for microscale purification and enrichment of glycans and glycopeptides, Anal. Chem. 83 (2011) 2492–2499. J.C. Bigge, T.P. Patel, J.A. Bruce, P.N. Goulding, S.M. Charles, R.B. Parekh, Nonselective and efficient fluorescent labeling of glycans using 2-amino benzamide and anthranilic acid, Anal. Biochem. 230 (1995) 229–238. M. Olajos, P. Hajos, G.K. Bonn, A. Guttman, Sample preparation for the analysis of complex carbohydrates by multicapillary gel electrophoresis with lightemitting diode induced fluorescence detection, Anal. Chem. 80 (2008) 4241– 4246. L. Novakova, L. Matysova, P. Solich, Advantages of application of UPLC in pharmaceutical analysis, Talanta 68 (2006) 908–918. J. Bones, S. Mittermayr, N. O’Donoghue, A. Guttman, P.M. Rudd, Ultra performance liquid chromatographic profiling of serum N-glycans for fast and efficient identification of cancer associated alterations in glycosylation, Anal. Chem. 82 (2010) 10208–10215. S. Mittermayr, J. Bones, M. Doherty, A. Guttman, P.M. Rudd, Multiplexed analytical glycomics: rapid and confident IgG N-glycan structural elucidation, J. Proteome Res. 10 (2011) 3820–3829. J. Deisenhofer, Crystallographic refinement and atomic models of a human Fc fragment and its complex with fragment B of protein A from Staphylococcus aureus at 2.9- and 2.8-Å resolution, Biochemistry 20 (1981) 2361–2370.