Characterization of IgG N-Glycans Employing a Microfluidic Chip that

Oct 6, 2009 - ... Darryl W. Brousmiche , Zhengmao Hua , Stephan M. Koza , Paula Magnelli , Ellen Guthrie , Christopher H. Taron , and Kenneth J. Fount...
0 downloads 0 Views 3MB Size
Download PDF
Anal. Chem. 2009, 81, 8818–8825

Characterization of IgG N-Glycans Employing a Microfluidic Chip that Integrates Glycan Cleavage, Sample Purification, LC Separation, and MS Detection Maggie A. Bynum,*,† Hongfeng Yin,† Katherine Felts,‡ Yvonne M. Lee,‡ Craig R. Monell,‡ and Kevin Killeen† Agilent Laboratories, Agilent Technologies, Santa Clara, California 95051, and Agilent Technologies, 11011 North Torrey Pines Road, La Jolla, California 92037 A novel polymeric microfluidic device with an on-chip enzyme reactor has been developed for the characterization of recombinant glycoproteins. The enzyme reactor chip packed with PNGase F-modified solid support material was combined with a microfluidic glycan cleanup chip and a commercially available HPLC-chip to perform glycoprotein deglycosylation, protein removal, glycan capture, glycan LC separation, and nanoelectrospray into a time-of-flight mass spectrometry (TOF-MS) system. With this integrated chip, the combined sample preparation and sample analysis time was reduced from multiple hours to less than 10 min. A once tedious and timeconsuming glycan analysis workflow is now integrated into an HPLC-chip device. Glycan profiling analysis has been achieved with as little as 100 ng of monoclonal antibody. Furthermore, a single chip was shown to retain activity and perform equivalently for over 250 replicate glycan profiles from a recombinant antibody. Glycosylation is of great importance for many recombinant protein drugs, with over one-third of recombinant protein drugs being glycoproteins. Changes in the glycosylation profile can lead to dramatic differences in glycoprotein pharmaceutical efficacy, pharmacokinetics, immunogenicity, folding, and stability.1-3 Thus, the characterization and routine analysis of protein-linked glycans is critical in the development and commercialization of glycoprotein pharmaceuticals, such as monoclonal antibodies. For monoclonal antibody therapeutics, glycosylation has been shown to alter biological efficacy.4 There are three typical steps in glycan profiling analysis: (1) release of glycans from glycoproteins, (2) separation of glycans, and (3) detection, identification, and quantitation of released glycans. * To whom correspondence should be addressed. E-mail: [email protected]. Phone: (408)553-3152. † Agilent Laboratories, Agilent Technologies, Santa Clara, California. ‡ Agilent Technologies, La Jolla, California. (1) Walsh, G.; Jefferis, J. Nat. Biotechnol. 2006, 24, 1241–1252. (2) Shields, R.; Lai, J.; Keck, R.; O’Connell, L.; Hong, K.; Meng, Y. G.; Weikert, S.; Presta, L. J. Biol. Chem. 2002, 30, 26733–26740. (3) Davies, J.; Jiang, L. Y.; Pan, L. Z.; LaBarre, M.; Anderson, D.; Reff, M. Biotechnol. Bioeng. 2001, 4, 288–294. (4) Jefferis, R. Expert Opin. Biol. Ther. 2007, 7, 1401–1413.

8818

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

The initial step in preparation of N-glycans for glycan profiling analysis commonly involves enzymatic deglycosylation of glycoprotein using PNGase F ((peptide-N4-(N-acetyl-β-glucosaminyl)asparagine amidase F; EC 3.5.1.52.)5 PNGase F is an amidase that cleaves the asparagine-sugar amide bond between asparagine and innermost N-acetyl-β-D-glucosamine residue of Nglycans, yielding intact glycans and deglycosylated protein (with the asparagine residue converted to aspartic acid). Initially, glycans are released as β-glycosylamines, with ammonia conjugated to the reducing end.6 The stability of β-glycosylamine is pH dependent and under mild acidic conditions β-glycosylamines spontaneously hydrolyze, liberating free ammonia and a free reducing end glycan. PNGase F acts with wide specificity, releasing all major N-linked glycans found on mammalian proteins. Traditional methods for deglycosylation with PNGase F add significantly to the glycan profiling analysis time as solution phase enzymatic glycan release routinely involves long incubation times. Recent approaches by multiple groups to increase the deglycosylation speed have included microwave assisted deglycosylation7 and deglycosylation using PNGase F immobilized on a monolithic substrate.8,9 Once deglycosylated, the glycans must typically be separated from the deglycosylated protein. Separation can be achieved by a variety of chromatographic methods, such as reverse phase, anion exchange chromatography, or normal phase chromatography.10 Nonchromatographic methods for separating glycans from proteins include ultrafiltration and precipitation. Detection and identification of glycans has been achieved with chromatographic and electrophoretic methods. Normal phase HPLC has also been demonstrated for the detection of glycans with the fluorophores anthranilic acid (2-AA) or 2-aminobenzamide (5) Maley, F.; Trimble, R. B.; Tarentino, A. L.; Plummer, T. H. Anal. Biochem. 1989, 180, 195–204. (6) Rasmussen, J. R.; Davis, J.; Risley, J. M.; Van Etten, R. L. J. Am. Chem. Soc. 1992, 114, 1126. (7) Sandoval, W. N.; Arellano, F.; Arnott, D.; Raab, H.; Vandlen, R.; Lill, J. R. Int. J. Mass Spectrom. 2007, 259, 116–123. (8) Palm, A. K.; Novotny, M. V. Rapid Commun. Mass Spectrom. 2005, 19, 1730–1738. (9) Krenkova, J.; Lacher, N. A.; Svec, F. J. Chromatogr. 2009, 1216, 3252– 3259. (10) Weitzhandler, M.; Hardy, M.; Co, M. S.; Avdalovic, N. J. Pharm. Sci. 1994, 83, 1670–1675. 10.1021/ac901326u CCC: $40.75  2009 American Chemical Society Published on Web 10/06/2009

(2-AB).11,12 Efficient reverse phase separation of glycans is possible when the glycans are derivatized with hydrophobic chromophores or hydrophobic fluorophores.13,14 Most fluorophores are coupled to the glycans via a reductive amination reaction which requires the free reducing end glycan. Commonly, these protocols include an acidification step to completely convert all β-glycosylamines to free reducing end glycans. Such labeling processes are laborious and introduce additional manipulation of the sample and time to the experiment. Capillary electrophoresis is another powerful glycan separation technique. Both laser induced fluorescence detection (LIF)15-17 and electrospray MS have been used to detect APTS (8-aminopyrene-1,3,6-trisulfonic acid) labeled glycans.18 Mass spectrometry is an efficient tool for accurate mass-based detection of glycans. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been widely used for detection of labeled and unlabeled glycans and provides a relatively fast measurement.19,20 However, MALDI-TOF is unable to provide resolution of isobaric glycans, such as G1 isomers. Further, MALDI-TOF analysis of glycans requires a cation-exchange step to desalt the glycans after deglycosylation. Electrospray MS has been used to detect and identify 2-AB labeled glycans.13 However, electrospray mass spectrometry’s primary advantage is label-free glycan detection and identification. For instance, graphitized carbon chromatography and hydrophilic interaction chromatography21 are used to separate glycans without derivatization and have been successfully coupled to nanoelectrospray MS.22 In this article, we describe an integrated microfluidic chip that incorporates all of the following functionalities: (1) glycan release with immobilized PNGase F, (2) removal of the deglycosylated protein, (3) capture of the released glycan, (4) glycan separation based on graphitized carbon chromatography, and (5) nanoelectrospray into a TOF MS for glycan detection and quantitation (see Supporting Information). This approach has numerous advantages over current methods. Our method includes a simplified sample preparation, a 10 min turnaround time from antibody injection to glycan data, and an automated protocol without any manual interference. In comparison, the traditional methods listed above require 1-3 days to deglycosylate, purify, or label and detect the glycans. The integrated chip significantly reduces the long (11) Anumula, K. R.; Dhume, S. T. Glycobiology 1998, 8, 685–694. (12) Guile, G. R.; Rudd, P. M.; Wing, D. R.; Prime, S. B.; Dwek, R. A. Anal. Biochem. 1996, 240, 210–226. (13) Prater, B. D.; Connelly, H. M.; Qin, Q.; Cockrill, S. L. Anal. Biochem. 2009, 385, 69–79. (14) Chen, X.; Flynn, G. C. Anal. Biochem. 2007, 370, 147–161. (15) Ma, S.; Nashabeh, W. Anal. Chem. 1999, 71, 5185–5192. (16) Raju, T. S.; Briggs, J. B.; Borge, S. M.; Jones, A. J. S. Glycobiology 2000, 10, 477–486. (17) Mechref, Y.; Muzikar, J.; Novotny, M. V. Electrophoresis 2005, 10, 2034– 2046. (18) Gennaro, L. A.; Salas-Solano, O. Anal. Chem. 2008, 80, 3838–3845. (19) Papac, D. I.; Brigg, J. B.; Chin, E. T.; Jones, A. J. S. Glycobiology 1998, 8, 445–454. (20) Keck, R. G.; Briggs, J. B.; Jones, A. J. S. In Therapeutic Proteins: Methods and Protocols, Methods in Molecular Biology, Vol. 308; Humana Press Inc.: Totowa, NJ, 2005; pp 381-396. (21) Alpert, A. J.; Shukla, M.; Shukla, A. K.; Zieske, L. R.; Yuen, S. W.; Ferguson, M. A. J.; Mehlert, A.; Pauly, M.; Orlando, R. J. Chromatogr., A 1994, 676, 191–202. (22) Ninonuevo, M.; An, H.; Yin, H.; Killeen, K.; Grimm, R.; Ward, R.; Germna, B.; Lebrilla, C. Electrophoresis 2005, 26, 3641–3649.

incubation time needed for solution phase deglycosylation with the PNGase F enzyme reaction. Furthermore, the need for acidification to convert the β-glycosylamines to the deamidated form is eliminated. Finally, the integrated chip also renders unnecessary the lengthy labeling and cleanup steps that are needed when analyzing glycans by UV or fluorescence. Further, the cation-exchange step required for MALDI analysis is also obviated. This integrated chip is compatible with existing nanoflow HPLC-chip/MS TOF systems. This integrated chip exhibits a significant increase in workflow efficiency and can be used for a broad range of applications. It has the potential to highly expedite the analyses of a wide range of glycans, including those present on current biotherapeutics. For example, routine analysis of glycans from monoclonal antibodies in biopharmaceutical development is a particularly well suited application for this integrated chip. In addition, this in-line technique can serve as a valuable measurement approach for process analytical technology (PAT). The use of this integrated chip technology offers the opportunity to monitor the effect of production parameter changes on cellular glycosylation patterns by sampling directly from the production bioreactor. We believe that this approach provides significant advantages in productivity, sensitivity, speed, and efficiency. We hope that this technology will help to accelerate the pace of discovery in the field of biotherapeutics and glycobiology. EXPERIMENTAL SECTION Reagents. The recombinant PNGase F enzyme was purchased from Prozyme Inc. (San Leandro, CA). Silica beads (5 µm) and 5 µm C8 beads (both with 300 Å pore size) were provided by Agilent Technologies (Santa Clara, CA). Common biochemical and organic reagents and ammonium acetate, formic acid, acetonitrile, and (3-glycidyloxipropyl)trimethoxysilane (GOPTS) were purchased from Sigma Chemical Co. (St. Louis, MO). Two antibody samples were obtained from Genentech for evaluation of the microfluidic enzyme bioreactor chip (referred to as “Ab1” and “Ab2” in this article.) “Ab3”, produced in a mouse NSO cell line, was provided by Pfizer. Enzyme Immobilization. PNGase F was covalently coupled to silica beads after silanization with GOPTS, following the published method.23 A total of 10 µg of PNGase F was buffer exchanged into 100 mM phosphate buffer (pH 8) to a final concentration of 2 µg/µL. A total of 2 mg of GOPTS functionalized silica beads were added to the enzyme solution. The slurry was incubated at room temperature for 24 h. The beads were centrifuged and washed three times with 1 mL of 200 mM TrisHCl, pH 8.0. The beads were incubated in Tris-HCl for 2 h and then washed three times with PBS. Chip Fabrication. In addition to the commercially available HPLC-chip, two more polymer microfluidic chips were constructed using the laser ablation and lamination technologies in polyimides as described previously.24 These two chips were used for in-line deglycosylation and removal of antibody, respectively, while the separation of glycans was achieved with the HPLC-chip packed with graphitized carbon. The PNGase F-immobilized beads were (23) Piehler, J.; Brecht, A.; Valiokas, R.; Liedberg, B.; Gauglitz, G. Biosens. Bioelectron. 2000, 15, 473–481. (24) Yin, H.; Killeen, K.; Brennen, R.; Sobek, D.; Werlich, M.; van de Goor, T. Anal. Chem. 2005, 77, 527–533.

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

8819

slurry packed into the 180 nL channel located in the first polyimide chip. The C8 beads were packed into a 180 nL channel in the second polyimide chip and used to remove deglycosylated protein from free glycan. The HPLC-chip has a sample enrichment column (160 nL), a separation column (43 mm) filled with porous graphitized carbon, and an integrated MS tip. All three chips were stacked, aligned, and inserted into the chip frame holder. Alternatively, for the analysis of deglycosylated antibody, the PNGase enzyme reactor chip was paired with a HPLC-chip that was packed with C8 reverse phase media. The antibodies were diluted in 100 mM ammonium acetate buffer, pH 7.6, to a concentration of 500 ng per µL, and 0.2 µL was injected. The total amount of antibody injected was 100 ng. LC/MS System Experimental Method. The HPLC-chip MS system was previously described in detail.24 An Agilent 1200 nanoPump was used to generate gradient nanoflow at 400 nL/ min, using typical mobile phases of 0.1% formic acid in MS-grade water (solvent A) and 0.1% formic acid in acetonitrile (solvent B). The gradient for fast deglycosylation and glycan analysis was 2-32% B in 1.5 min, 32-75% B in 30 s with a hold at 75% B for another 30 s, followed by a 75-2% B drop over the last 30 s. A longer gradient was used to separate glycan isomers. The gradient for isomer separation was 2-22% B in 10 min and 22-75% B in 2 s with a hold at 75% B for another 3 min, followed by 75-2% B over the last 1 min. An Agilent 1200 CapPump was used for both loading and deglycosylation with 100 mM ammonium acetate, pH 7.6. An Agilent 1200 ChipCube was used as the chip interface to the mass spectrometer. An Agilent 6224 TOF MS was used to record mass spectra in positive ion mode. For glycan analysis, the nanoelectrospray voltage was set to 1750 V, with a drying gas of 8 L/min nitrogen at 325 °C. The fragmentor voltage was 120 V, and the skimmer voltage was 65 V. Data was acquired at 2 GHz, in extended dynamic range mode. For the deglycosylated antibody analysis, the fragmentor voltage was raised to 375 V. Data Analysis. Data analysis and evaluation was performed using the Agilent MassHunter software version B.02. The molecular feature extraction (MFE) tool in MassHunter identifies all charge states and adducts for each compound and combines these entities into one compound (molecular feature). Following this extraction, the results were plotted as an extracted compound chromatogram (ECC) and exported as an analysis report in Microsoft Excel format. A custom Excel macro was developed that automatically computes the ratios for each glycan as a percentage of the total glycans. RESULTS AND DISCUSSION Microfluidic Chip Design. The multilayered polymeric microfluidic chip architecture was first described in 2004 as a twodimensional HPLC separation device.25 This design enabled the integration of two independent sample separation functions on a single microfluidic device, where each sample separation function is on a different layer of the device. To integrate multiple steps of the glycan profiling analysis workflow, we have drawn on the multilayer chip concept and expanded it, using an alternative (25) Yin, H.; Killeen, K. HUPO poster presentation, 2004. (26) Ghiun, M.; Bonneil, E.; Pomies, C.; Marcantonio, M; Yin, H.; Killeen, K.; Thibault, P. Miniaturization and Mass Spectrometry; PSC Publishing: Cambridge, U.K., 2009; Chapter 8.

8820

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

Figure 1. Diagram of the three chips used for enzymatic deglycosylation, protein (deglycosylated antibody) removal, glycan concentration, and glycan separation. From top to bottom are the PNGase F enzyme chip, C8 chip, and HPLC-chip. Inside the ChipCube, the three chips are sandwiched between the stator on the top side and the rotor on the bottom side. Sample and solvent are automatically introduced through the capillary flow pump and the nanoflow pump.

approach in which three separate chips were created and combined. The first chip has a reactor chamber packed with PNGase F immobilized on silica beads. The second chip is packed with 5 µm reverse phase C8 beads for capture of the deglycosylated antibody. The third chip, a commercially available HPLCchip, has both sample enrichment column and an LC separation column packed with porous graphitized carbon (PGC) material. These three chips were stacked, aligned, and sealed with each other inside an Agilent HPLC ChipCube instrument previously described.26 Figure 1 is an expanded view of the three combined chips. From top to bottom are the PNGase F enzyme chip (green), C8 chip (purple), and HPLC-chip (yellow.) Inside the ChipCube, the three chips were sandwiched between the stator on the top side and the rotor on the bottom side. This design results in the intact glycoprotein sample flowing into the enzyme reactor from a transfer capillary that is connected to the stator. Both the deglycosylated protein and the released glycans then travel together into the C8 bead-packed channel where the proteins are trapped. In the first rotor valve position (Figure 2A), the glycans are then trapped on the enrichment column in the HPLC-chip. In the second rotor valve position (Figure 2B), the gradient flow delivered by the nanopump elutes the glycans from the enrichment column followed by the LC separation column before being sprayed through the on-chip nanoESI tip. While it is possible to integrate these functions onto a single multilayered chip using the architectures described in ref 25, we found that the multiple chip approach described here is better suited for method development, where different prototype designs of the individual chip layers can be mixed and varied depending on the application needs. On-Chip Deglycosylation with Immobilized PNGase F Enzyme Reactor. To study the analytical characteristics of the microfluidic chip, Ab1 and Ab2 were evaluated. The enzyme reactor chamber has a total volume of 180 µL. At a flow rate of 1 µL/min, the residence time (the total time the antibody is exposed to the PNGase F enzyme) was 6 s. When the deglycosylated antibody was tested with ESI MS, approximately 98% of the antibody was deglycosylated within this residence time. When the flow rate was increased to 2 µL/min, (reducing the residence time

Figure 2. HPLC-chip and ChipCube rotor diagram showing the flow path during sample preparation and analysis. In the valve configuration shown in Figure 2A, the chip is in sample preparation and deglycosylation mode. The intact antibody sample travels through the PNGase F enzyme reactor from a transfer capillary, where the glycans are cleaved from the antibody. Both the deglycosylated antibody and the free glycans then travel together into the C8 bead-packed channel where the antibodies are retained. The free glycans travel further, via a rotor groove (shown in red), to the enrichment column on the HPLC-chip. Figure 2B illustrates the valve configuration and flow path in the LC/MS analysis configuration. In this valve configuration, an Agilent Nano Pump delivers gradient nanoflow to elute glycans from the enrichment column and separate the glycans with the separation column before the electrospray source and the mass spectrometer.

to 3 s), the amount deglycosylated antibody dropped to approximately 65%. The close coupling of the MS detector to the newly released glycans enables the direct detection the β-glycosylamine intermediates before the conversion to the free-reducing end glycans. To our knowledge, this is the first time β-glycosylamine intermediates have been separated by LC. All of the traditional labeling methods based on reductive amination require the presence of a free reducing terminus of the glycans, which necessitate incubating the released glycans with acid to accelerate the hydrolysis and to achieve complete conversion from the β-glycosylamine to the free reducing end glycan required for labeling. Our approach obviates the need for this acid incubation step. Table 1 illustrates the glycan structures and the corresponding names used in this article, as well as the monoisotopic masses of the free reducing end glycan forms. Glycan cartoons are constructed using standard Consortium for Functional Glycomics (CFG) nomenclature for monosaccharides.27 Note that these glycans in their β-glycosylamine forms will have a molecular mass that is 0.984 amu smaller than those noted in Table 1. Figure 3 shows the extracted compound chromatogram (ECC) separation of the dominant glycan peaks. The β-glycosylamines elute between 2.2 and 2.4 min, and the free reducing end glycans elute between 2.5 and 2.7 min. In the β-glycosylamine form, each glycan compound exists as a single peak over the course of this 1.5 min separation gradient. However, the data show that in the free reducing end glycan form, each glycan is split into a pair of peaks. This splitting arises from the anomeric carbon at the reducing end.28 In a conventional in-solution deglycosylation experiment, a reduction step is used to convert the split peaks into a single peak. The inset in Figure 3 shows the mass spectra of the doubly charged β-glycosylamine-form and the free reducing end glycan form of G0. The difference in m/z of the doubly charged monoisotopic ions is 0.492, or 0.984 amu for the mass difference, which corresponds to the difference between -OH and -NH2. When the extracted ion chromatogram (EIC) for the β-glycosylamine G0 is (27) Consortium for Functional Glycomics (CFG). http://www.functionalglycomics.org/static/consortium/Nomenclature.shtml. (28) Lindhorst, T. K. Essentials of Carbohydrate Chemistry and Biochemistry; WileyVCH GmbH and Co.: Weinheim, Germany, 2007; p 19.

Table 1. Glycan Structures and Names Used in This Articlea

a The masses for the glycans in the free reducing end glycan form are also noted. Glycan cartoons are constructed using the standard Consortium for Functional Glycomics (CFG) nomenclature for monosaccharides.

plotted using a narrow window around m/z 731.78, only the first peak at 2.25 min is observed. However, when the EIC is plotted for the free reducing end form at m/z 732.27, both the monoisotopic peak of the doubly charged free reducing end form G0 and the C13 isotope peak of the G0 β-glycosylamine form are observed (data not shown.) For this reason, we decided to use ECC instead of EIC to distinguish the β-glycosylamines from the free reducing end glycans in this article. Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

8821

Figure 3. Extracted compound chromatogram (ECC) of glycans from Ab1 with on-chip deglycosylation, antibody removal, glycan capture, and glycan separation. The β-glycosylamine intermediates eluted between 2.2 and 2.4 min, while the free reducing end glycans eluted between 2.5 and 2.7 min. In addition to the peak annotation shown by glycan structures, they are also color coded: red trace ) G0, green trace ) G1, blue trace ) G2. The inset in Figure 3 shows the mass spectra for the doubly charged β-glycosylamine form (top) and the free reducing end glycan form (bottom.) The mass difference is 0.984 amu, which is the mass difference between -OH and -NH2.

Figure 4. Glycan profile plotted as ECC for Ab2. In addition to the fucosylated glycans seen in Ab1, Ab2 contains lower abundance nonfucosylated glycans that are separated from the other glycans and elute between 2.1 and 2.2 min.

Glycan Separation by the Graphitized Carbon HPLC-Chip. The separation conditions for the PGC column were optimized to reduce total analysis time while maintaining the analytical performance of the chips. Figure 4 is the total glycan profile of Ab2. From Figure 4, it can be seen that the nonfucosylated glycans elute between 2.1 and 2.2 min. Under these separation conditions, the two G1 isomers are not separated. 8822

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

Figure 5. ECC of antibody 3 made in a mouse NSO cell line. β-Glycosylamine G1 isomers (green trace) as well as G2 isomers (blue trace) were separated using a 10 min LC gradient. In addition to the more commonly observed glycans G0, G1, and G2, other glycans were observed and identified in this sample, such as G0GlcNAc-F, G0-F, Man-6, G1-GlcNAc-F, G0-GlcNAc, and G1-GlcNAc.

With shallower LC mobile phase gradient conditions, we were able to separate isomeric forms of the glycans. Figure 5 shows the separation of G1 isomers from Ab3 over the course of a 6 min separation. In addition, the data show three peaks with the equivalent mass of the G2 glycan, which we believe to be isomers of G2 with the structure 8 and 9 shown in Table 1. This antibody was expressed in an NSO mouse cell line, which is known to produce G2 isomers.29 Conversion from β-Glycosylamines to Free Reducing End Glycans. We examined the utility of measuring the β-glycosylamine forms and compared the relative ratios measured to the more traditionally measured free reducing end glycan form. We accomplished this by injecting Ab1 over the PNGase F column on-chip and maintaining the gradient at 2% B for extended lengths of time from 0 to 120 min. As a result, the glycans remained on the PGC trapping column and hydrolyzed to the free reducing end form in the presence of water supplemented with 0.1% formic acid. Figure 6 shows the ECC and depicts the decrease in abundance of the β-glycosylamine form and corresponding conversion to the free reducing end form over time. Figure 6A shows the ECC with no wait time on the trapping column, as the chip would normally be used. β-Glycosylamines elute between 2.2 and 2.4 min, while the free reducing end glycans elute between 2.5 and 2.7 min. For the time course experiment, the glycans remained on the trapping column for 30, 60, and 120 min, as shown in parts B, C, and D of Figure 6, respectively. Figure 6B shows the increase in the free reducing end glycans after 30 min; 75% of the glycans hydrolyzed to the free reducing (29) Stadlmann, J.; Pabst, M.; Kolarich, D.; Kunert, R.; Altmann, F. Proteomics 2008, 8, 2858–2871.

Figure 6. Time course experiments showing the kinetics of hydrolysis of the β-glycosylamine form to the free reducing end-glycan form. After the free glycans are captured on the enrichment column, the ChipCube rotor rotates to allow the nanopump to deliver the LC gradient through the enrichment column. By varyiation of the gradient delay time, the captured glycans were hydrolyzed, to different degrees, to free reducing end glycans. The gradient delay times were A, 0 min; B, 30 min; C, 60 min; and D, 120 min. The β-glycosylamines eluted between 2.2 and 2.4 min and decreased in intensity over time, while the free reducing end form eluted between 2.5 and 2.7 min increased over time. Within 120 min (Figure 6D), the β-glycosylamines are completely hydrolyzed to the free reducing end form. The distribution of G0, G1, and G2 is maintained.

end form. Figure 6C shows the glycans after 60 min; 90% of the glycans are hydrolyzed to the free reducing end form. After 120 min, the β-glycosylamines are completely hydrolyzed to the free reducing end form, as shown in Figure 6D. The data also show that by measuring the glycans in their β-glycosylamine state, the sensitivity is about double to that of measuring the glycans in their more typically measured free reducing end form. The three most abundant glycans, G0, G1, G2 were identified in the free reducing end form shown in Figure 6D. Notably, in the β-glycosylamine form, lower abundance glycans could be measured, such as G1 - GN, G0 - GN, G2 + GN, each at less than 1% of the total. (Lower abundance glycans are not shown in Figure 6.) The glycan distribution was calculated and maintained for each time point during the conversion of the β-glycosylamines and free reducing end glycans. The relative percent of the top four glycans (G0, G1, G2, and G0 - GlcNAc) were 62%, 31%, 3%, and 2%, respectively. Analysis of Deglycosylated mAb. In a second chip configuration, Ab1 was deglycosylated in the enzyme reactor and the deglycosylated antibody was analyzed. For this application, the device configuration consisted of two integrated chips, instead of three as in the device described above for measuring cleaved glycans. The first chip contained the enzyme reactor, while the second chip contained the LC/MS chip packed with C8 beads in the trapping column and in the LC column. This chip also had an integrated electrospray tip. By varying of the residence time of the antibody in the deglycosylation reactor, the necessary residence time for complete

deglycosylation was determined. First, the intact antibody was analyzed (without deglycosylation) on the C8 LC/MS chip. Figure 7A shows this deconvoluted antibody spectrum. Three main peaks are observed. The difference between two consecutive peaks is 163.9 and 161.8 amu, respectively, which is in agreement with the characteristic mass difference of the terminal galactose on G0, G1, and G2 glycans of 162.05 amu. Because this antibody has two glycosylation sites, one on each heavy chain, these peaks are suggestive of the presence of the combination of G0, G1, and G2 glycans on the antibody. The peak masses measured are 149 089.6, 149 253.5, and 149 415.3 amu and most likely represent antibody with G0/G0, G0/G1, and G1/G1 or G0/G2, respectively. Next, the antibody was deglycosylated with the enzyme-reactor chip with a residence time of 3 s and then again with a residence time of 6 s. Figure 7B shows the partially deglycosylated antibody following a 3 s residence time in the enzyme chip. About 65% of the antibody is deglycosylated, as indicated by the peak at 146 202.2 amu, which is the mass for the deglycosylated antibody. The measured mass difference between the deglycosylated antibody and the antibody with two G0 glycans is 2887.4 amu, which agrees with the expected mass difference of 2889.1 amu. Incompletely deglycosylated antibody with one glycan remaining can be detected with these methods, as shown in Figure 7B and represented by the peaks at 147 646.9 and 147 807.9 amu. The difference between these two peaks is 161.0 amu and represents the terminal galactose addition. The mass difference between the deglycosylated antibody and the antibody with one G0 glycan is measured to be 1443.7 amu, which is in agreement with the Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

8823

Figure 7. Analysis of deglycosylated antibody with on-chip deglycosylation. Figure 7A shows the deconvoluted antibody spectrum of Ab1. Three main peaks are observed. These peaks are suggestive of the antibodies with the combination of G0, G1, and G2 glycans. The peak masses measured are 149 089.6, 149 253.4, and 149 415.3 amu and most likely represent antibody with G0/G0, G0/G1, and G1/G1 or G0/G2, respectively. Figure 7B shows the partially deglycosylated antibody following a 3 s residence time. Antibodies with one glycan are observed as peaks with mass of 147 645.9 and 147 806.4 amu. Figure 7C shows the deglycosylated antibody peak after a 6 s residence time in the PNGase F enzyme reactor chip.

expected mass difference of 1444.5 amu. Lastly, the antibody was deglycosylated in the enzyme-reactor chip with a residence time of 6 s, as shown in Figure 7C, where nearly all of the antibody is deglycosylated. As a result, the glycan chip can be used to analyze deglycosylated antibody mass without the need for lengthy deglycosylation steps. Lifetime of the Integrated Chip. The enzyme reactor contained in the microfluidics chip has remained active and fully functional for at least 250 injections with no loss of activity. The chip was stored at 4 °C between experiments. During storage, the enzyme chip was allowed to dry, and when reconstituted with buffer and retested, there was no apparent loss of enzyme activity, as indicated by comparable chip performance (data not shown.) CONCLUSIONS A novel microfluidics chip is described that incorporates all of the steps of glycan profiling analysis of monoclonal antibodies: glycan cleavage, cleanup, capture, separation, and detection. This microfluidic chip approach enables a simplified sample prepara8824

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

tion, rapid deglycosylation in 6 s, a 10 min turnaround time from antibody injection to glycan data, and an automated protocol without the need for manual intervention. In comparison, the conventional methods for characterizing N-glycans from antibodies are time-consuming and laborious, involving long incubation times, with multiple cleanup, labeling, and transfer steps, typically requiring 1-3 days. The rapid MS detection following deglycosylation allows for the measurement of the β-glycosylamine intermediates, which we have shown to be analogous to measuring the free reducing end forms. We demonstrate several advantages to measuring this β-glycosylamine intermediate form. First, each β-glycosylamine structure exists as one peak, thereby facilitating analysis, as compared to the split peaks that occur in the free reducing end form due to the anomeric nature of a nonreduced free reducing end glycan. Second, the isomers of G1 and G2 can be readily separated in less than 10 min in the β-glycosylamine intermediate form. In addition to analyzing the released glycans, this microfluidics device with integrated enzyme cleavage has also been

shown to be suitable for analysis of the deglycosylated antibody. We believe that this novel integrated microfluidics device will facilitate the fast and accurate routine characterization of glycans, glycoproteins, and antibodies in biotherapeutics development and research. ACKNOWLEDGMENT We thank Rod Keck and Dr. Tomasz Baginski from Genentech for providing antibody samples used in these experiments, as well as for their helpful discussions of the field of glycobiology and therapeutic antibody development and characterization. We thank Dr. Nathan Lacher of Pfizer for his donation of the antibody produced in the NSO cell line. We would also like to thank Dr.

Karla Robotti and Dr. Ahmed Faizy for their contribution in the silica bead chemistry, Dr. Pat Perkins for his in depth knowledge in oligosaccharide chemistry, Debbie Ritchey for her support in chip manufacturing, Dr. Reid Brennen for the mechanical drawings, and Dr. George Yefchak for developing the Excel macro. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review June 18, 2009. Accepted September 10, 2009. AC901326U

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

8825