Integrated Microfluidic LC-MS

Glycoproteins comprise an important class of biological compounds. The hallmark feature of a glycoprotein is the presence of N-linked glycans, which exhibit a tremendous amount of structural complexity and diversity. It is this structural complexity that makes N-linked glycans difficult to characterize. Current characterization workflows are tedious, time-consuming and as such, are unsuitable for real-time bioprocessing applications. A microfluidic-based workflow that integrates and automates glycan cleavage, purification, and chromatographic separation onto a microfluidic liquid chromatography–mass spectrometry (LC–MS) chip for subsequent MS detection is a valuable approach to resolve these shortcomings in current methods.

Glycans' importance to bioprocessing

Protein glycosylation is an important class of post-translational modification often necessary for correct protein folding and full biological function. More than one-third of recombinant protein drugs are glycoproteins, and antibodies are the largest group of recombinantly produced glycoproteins. Glycans are covalently linked carbohydrate moieties of glycoconjugates such as glycoproteins. N-linked glycans (N-glycans), the focus of this article, constitute one of the major classes of glycans found on mammalian glycoproteins. N-glycans are most commonly attached to the nitrogen of asparagine present in the consensus amino-acid sequence AsnXxxSer/Thr (where Xxx can be any amino acid except proline), as a result of the multistep enzymatic process of glycosylation.

Table I

Recombinant monoclonal antibodies constitute a major class of anticancer therapeutics, and their N-glycans have earned much attention in recent years. Antibodies include evolutionarily conserved sites of N-glycosylation, which contain heterogeneous glycan structures of biantennary class. Glycan heterogeneity is a natural outcome of complex enzymatic synthesis pathways where carbohydrate residues may be attached to each other in many different ways, thereby resulting in various linkages and positional isomers. Table 1 provides a list of some of the N-glycans commonly found on natural and recombinant monoclonal antibodies along with their structure, name, and monoisotopic mass.

Glycans are a source of heterogeneity that can affect the biological attributes of recombinant antibody therapeutics. In antibodies possessing effector functions, N-glycans have been shown to be involved in the removal of cancer cells via antibody-dependent cell-mediated cytotoxicity (ADCC.) In vivo, after the Fab region of the antibody binds to the antigen, natural killer cells bind to the Fc region. This triggers the release of cytotoxic agents that destroy the antigen. The lack of a core fucose in the Fc region of the antibody leads to increased Fc-receptor binding and is correlated with up to 50 times higher ADCC activity (1).

Glycan heterogeneity varies with species, expression system, and cell-culture conditions. Chinese hamster ovary (CHO) cells, a major expression host for many industrially manufactured human therapeutic antibodies, produce "human-like" N-linked glycans. On the other hand, the "same" antibody produced in a mouse NS0 cell line can have very different, sometimes undesirable glycan structures. NS0 cells contain α1,3-galactosyltransferase (α1,3GT), an enzyme that transfers α-Gal residues to N-acetyllactosamine (LacNAc) termini of glycans. Antibody therapeutics with the Gal-α-Gal linkages (α-Gal epitopes) have been shown to cause immune reactions (2). In Table 1, the glycan structure labeled G2 isomer shows the immunogenic Gal-α-Gal structure, while the structure labeled G2 is the nonimmunogenic form.

Because glycoprotein heterogeneity can arise from incomplete synthesis, glycans can provide an important measure of production process consistency. Enzymes involved in glycan synthesis can be sensitive to subtle changes in conditions, and manufacturing processes rarely achieve 100% efficiency. Because variants may not be as safe and effective as the desired product, they must be characterized and quantified. N-linked glycan characterization is becoming more routinely performed as an in-process test during clone selection and screening of cell-culture conditions for recombinant glycoproteins derived from mammalian cell lines. And, it is expected that regulatory agencies will more frequently request glycan characterization prior to product release (3).

Because glycosylation of recombinant monoclonal antibodies can influence their biological activity, efficacy and immunogenicity, there is a growing interest in antibody design based on glycan engineering, and a growing need in drug discovery, drug development, and bioprocessing for a fast and reliable separation and detection technique to characterize glycan microheterogeneity.

Traditional N-glycan characterization methods

Various methods are used to characterize N-glycans. Figure 1 summarizes those most commonly used. Most require multiple steps and a day or more to complete. All typically follow the same basic process as follows:

• Release of glycans from glycoproteins

• Separation of glycans from the deglycosylated protein

• Analysis (separation, detection, identification, and quantification) of glycans with or without derivatization.

Enzymatic digestion with PNGase F ((peptide-N4-(N-acetyl-β-D-glucosaminy-I)) asparagine amidase F; EC 3.5.1.52) is commonly used to release glycans from glycoproteins. PNGase F is an amidase that cleaves the amide bond between the glycosylation-site asparagine and the innermost N-acetyl-β-D-glucosamine residue of N-glycans. It has a wide specificity, releasing all major classes of N-glycans found on mammalian glycoproteins. As shown in Figure 1, this method of deglycosylation involves long incubation times of 3 to 20 h, a significant component of the overall analysis time.

After releasing the glycans, the next step is to separate and recover them from the deglycosylated protein. This step is commonly done using reverse-phase or graphitized carbon-based solid-phase extraction, ultrafiltration, and precipitation. Centrifugal evaporation is typically needed to concentrate and dry the purified glycans. Depending on the method used, this step can take several hours to complete (see Figure 1).

Various methods are used to separate and detect glycans. Although high pH anion-exchange chromatography (HPAEC), with pulsed amperometric detection (PAD), has been popular because it does not require derivatization (4), it is being replaced by newer methods such as high-performance liquid chromatography (HPLC) and capillary electrophoresis (CE) with fluorescence detection. To improve the separation and enhance detection sensitivity, these methods involve derivatizing free glycans with fluorophore (via reductive amination.) After separation, the glycans' retention time or migration time is measured, and peaks are correlated with standards. Relative quantitation is performed based on the fluorescence of individual peaks (5). For CE analysis, glycans are usually labeled with fluorescent tag APTS (8-aminopyrene-1,3,6-trisulfonic acid.) (6). For HPLC analysis, glycans are typically labeled with a fluorescent label such as 2-aminopyridine (2-AP), anthranilic acid (2-AA), or 2-aminobenzamide (2-AB), and separation is performed using either hydrophilic interaction chromatography (HILIC) or reverse-phase chromatography. Depending on the method used, labeling requires about 1–3 h to complete. Postlabeling cleanup further increases the time requirements needed to complete the analysis.

Because sample labeling and cleanup is laborious and time consuming, direct label-free detection and identification by MS is a popular technique. MS also has the advantage of being able to provide exact mass data and identify very complex glycan profiles. Matrix-assisted laser desorption time-of-flight (MALDI–TOF) MS is one of the MS techniques often used. However, because MALDI–TOF-MS is not coupled to chromatographic separation, isomeric information is not obtained. Sample preparation for MALDI–TOF-MS involves deglycosylation, purification, acidification, and cation exchange to remove salts after deglycosylation, and takes about 6 h (see Figure 1) (7).

Figure 1: Characterization workflows vary widely in terms of the techniques used and the time required to perform them. Even the fastest workflow is not fast enough to be practical for real-time process monitoring. LC-LIF is laser-induced fluorescence. LC is liquid chromatography. CE–MS is capillary electrophoresis–mass spectrometry. HPLC is high-performance liquid chromatography. MALDI is matrix-assisted laser desorption. (ALL FIGURES BUT FIGURE 3 ARE COURTESY AGILENT TECHNOLOGIES.)

Hyphenated techniques have been developed to draw upon the combined attributes that labeling, chromatographic separation, and MS provide: improved detection sensitivity, isomer identification, and accurate mass information. For example, CE–MS methods enable faster separations as well as separation of isomeric species (8). HILIC and reverse-phase HPLC methods can be used in combination with electrospray MS detection. These methods also provide useful isomeric information (9, 10). Though these hyphenated methods provide powerful data, they often require 2–3 days to complete, thereby making them unsuitable for real-time applications.

Because porous graphitized carbon (PGC) chromatography is amenable to the separation of unlabeled glycans, it, in combination with nanoelectrospray MS, is emerging as the preferred hyphenated technique to separate and identify glycans without derivitization (11, 12). Unlike reverse-phase HPLC, glycans are readily retained on PGC without the need for a labeling step. However with this method, the glycans must be reduced to simplify their chromatographic profile. Without this lengthy reduction step, each glycan yields two peaks due to the anomeric carbon at the reducing end.

Integrated microfluidic LC–MS workflow

Current glycan characterization methods are proven, but are time-consuming and labor-intensive. As discussed and shown in Figure 1, these methods use lengthy solution-phase enzyme cleavage of glycans, fluorescent labeling for CE and LC analysis, and sample cleanup for MALDI, LC–MS, or CE–MS detection. A reliable, rapid method, simplified by automated sample preparation and separation, combined with the specificity of MS detection would be very useful in all stages of glycoprotein development and production.

Figure 2: Glycan profiling workflow using an integrated microfluidic chip. The time required for glycan cleavage, sample preparation, and glycan analysis is reduced to 10 min. In comparison, current methods require at least 6 h and involve many sample handling, cleanup, and labeling steps. LC-LIF is laser-induced fluorescence. MALDI–MS is matrix-assisted laser desorption–mass spectrometry. LC–MS is liquid chromatography–mass spectrometry. mAb is monoclonal antibody. HPLC is high-performance liquid chromatography. TOF MS detector is time of flight–mass spectrometry.

Recent developments in microfluidics technology enable a new approach that integrates and automates established glycan analysis sample preparation and characterization steps onto a single microfluidic chip coupled to a TOF-MS (see Figure 2) (13). These steps are as follows:

• N-glycan release with immobilized PNGase F

• Removal of deglycosylated protein

• Capture of released glycans

• Separation by PGC chromatography

• Nanoelectrospray ionization for subsequent detection by TOF-MS.

As shown in Figure 3, the integrated microfluidic chip is composed of three biocompatible polyimide microfluidic chips constructed using patented laser ablation and lamination technologies developed by Agilent Laboratories, the central research laboratory for Agilent Technologies (Santa Clara, CA) (14). Laser ablation of polyimide film is used to fabricate the chip's surface structures, and lamination of multiple layers together is used to create multifunction chips.

Figure 3: Diagram of the three-layer integrated microfluidic liquid chromatography–mass spectrometry (LC–MS) chip. The three chips are fixed between a stator and a rotor on an Agilent HPLC Chip Cube MS. The chip components are constructed using patented polyimide laser ablation and lamination technology developed by Agilent Technologies Agilent Laboratories. HPLC is high-performance liquid chromatography. (FIGURE 3 IS COURTESY REID BRENNEN, AGILENT TECHOLOGIES AGILENT LABORATORIES)

The first chip, on top, has a reactor chamber packed with PNGase F-immobilized silica beads. The second chip, in the middle, is packed with 5µm reverse-phase C8 beads to capture deglycosylated protein. The third chip, at the bottom, contains a sample enrichment column, an LC separation column packed with PGC and a nanospray tip for nanoelectrospray ionization. The chips are stacked, aligned, and sealed to each other, and then fixed between a stator on the top and a rotor on the bottom of an Agilent HPLC-Chip Cube MS interface.

Figure 4 shows the sample flow during chip operation. In the first-rotor valve position (i.e., the sample-preparation configuration), a transfer capillary attached to the stator delivers intact antibody sample into the enzyme reactor, and the glycans are cleaved from the antibody. The deglycosylated antibody and the released glycans flow together into the C8 bead-packed retention column where the protein is trapped. The glycans are trapped and concentrated on the PGC enrichment column. In the second-rotor valve position (i.e., the sample-analysis configuration), the aqueous to organic gradient flow delivered by a nanoflow pump (Agilent 1200 Series, Agilent Technologies) elutes the glycans from the enrichment column and separates them on the LC column. Finally the glycans are sprayed through the nanoelectrospray tip into the TOF-MS.

Figure 4: Microfluidic chip operation and flow path during sample preparation and analysis. In the valve configuration and flow path shown in Figure 4 (a) (i.e., the sample-preparation configuration), the intact glycoprotein sample travels to the peptide-N4-(N-acetyl-β-D-glucosaminy-I) asparagine amidase F (PNGase F) enzyme reactor where it is deglycosylated. Both the deglycosylated protein and the free glycans travel together into the C8 bead-packed retention column where the proteins are retained. Free glycans continue, via a rotor groove shown in red, to the enrichment column. Figure 4 (b) illustrates the valve configuration and flow path for sample analysis. In the sample-analysis configuration, the nanopump delivers an aqueous-to-organic gradient to elute the glycans from the enrichment column and to separate them on the porous graphitized chromatography column. PGC is porous graphitized carbon. MS is mass spectrometry.

The microfluidic LC–MS chip workflow completes all analysis steps, from antibody injection to LC–MS results, in 10 min and is completely automated once sample is introduced.

Microfluidic LC/MS workflow results

CHO cell-derived monoclonal antibody A1 and Ab2 and mouse NS0 cell-derived Ab3 were analyzed to evaluate the integrated microfluidic LC–MS chip workflow. Figure 5 shows the extracted compound chromatogram (ECC) with separation of the dominant glycan peaks. The β-glycosylamines eluted between 2.2 and 2.4 min, and the free-reducing-end glycans eluted later, between 2.5 and 2.7 min. In the -glycosylamine form, each glycan is represented by a single peak in the 1.5-min. separation. In contrast, the free-reducing-end form is represented by pairs of peaks. This pair is due to the anomeric carbon at the reducing end and is characteristic of PGC chromatography. In a conventional deglycosylation experiment, a reduction step would have been used to simplify the chromatogram.

Figure 5: Extracted compound chromatogram of the glycans released from antibody sample Ab1. The β -glycosylamine intermediates were eluted between 2.2 and 2.4 min and the free-reducing-end glycans were eluted between 2.5 and 2.7 min. The three predominant glycan (G) peaks are annotated with the glycan cartoon structures and color coded: red = G0, green = G1, and blue = G2. The inset shows the mass spectra for the doubly charged -glycosylamine form (top) and the free-reducing-end glycan form (bottom). The mass difference is 0.984 atomic mass units (amu), which is the mass difference between –OH and –NH2.

The inset in Figure 5 shows the mass spectra of the doubly charged -glycosylamine form and the free-reducing end for the glycan G0. The difference in mass–to–charge ratio (m/z) of the doubly charged monoisotopic ions is shown to be 0.492, which corresponds to the 0.984 atomic mass units (amu) mass difference between the –OH and –NH₂ forms. Because The β-glycosylamines exist as single peaks, a reduction step is not needed. An important advantage of coupling TOF-MS to the microfluidic chip is that it enables detection of -glycosylamine intermediates prior to conversion to free-reducing-end glycans. The acidification step used with traditional HPLC approaches involving labeling or derivatizing the free glycans by reductive animation is not needed.

Figure 6 shows the glycan profile of Ab2. In addition to the fucosylated glycans found in A1 (see Figure 5), Ab2 contains lower abundance nonfucosylated glycans that were separated from the other glycans, eluting earlier between 2.1 and 2.2 min.

Figure 6: Extracted compound chromatogram for the glycan profile of Ab2. In addition to the fucosylated glycans seen in Ab1 (see Figure 5), Ab2 contains lower abundance nonfucosylated glycans that were separated from the other glycans (eluting earlier, between 2.1 and 2.2 min).

As shown for Ab3 in Figure 7, by using a longer 10-min, LC gradient, it is possible to separate isomeric glycan forms. The data show three peaks with masses equal to the G2 glycan, which are likely to be G2 isomers with the structures labeled G2 and G2 isomer shown in Table I. Antibody Ab3 was expressed in a NS0 mouse cell line that is known to produce G2 isomers.

Figure 7: Extracted compound chromatogram of Ab3 made in a mouse NSO cell line. The -glycosylamine G1 (green trace) and G2 isomers (blue trace) were separated using a longer liquid chromatographic gradient. In addition to the more commonly observed G0, G1, and G2 glycans, many others were found and identified.

Mass spectral data were generated by an Agilent 6224 Accurate-Mass TOF LC–MS. Data analysis was performed using Agilent MassHunter Workstation software. The molecular feature extraction (MFE) algorithm in MassHunter identifies all charge states and adducts for each compound and combines these into one compound or molecular feature. Following extraction, the mass spectral results are plotted as an ECC and exported into Microsoft Excel. An Excel macro was created to automatically compute the ratios for each glycan as a percentage of the total.

β-glycosylamines versus free-reducing-end glycans

Because direct coupling of the microfluidic LC–MS chip to the TOF-MS enables detection of -glycosylamine intermediates, it is important to understand the kinetics of hydrolysis from the -glycosylamine to free-reducing end forms. By varying the glycan residence time on the PGC trapping column, it is possible to monitor the consistency of the distribution of G0, G1, and G2 during hydrolysis.

In the time-course experiment summarized in Figure 8, the glycans remained on the trapping column for 0, 30, 60, and 120 min. The ECC's depict the decrease in abundance of the -glycosylamine forms and the corresponding conversion (increase) to the free-reducing-end forms over time. With no wait time on the trapping column, as the chip would be used in a real-time process-monitoring application, The β-glycosylamines were eluted between 2.2 and 2.4 min, and the free-reducing-end glycans were eluted between 2.5 and 2.7 min. The ECC showed that, with no wait time, the abundance counts of The β-glycosylamines were more than double those of the free-reducing-end glycans. With the microfluidic LC–MS chip approach, measuring β-glycosylamines, rather than the free-reducing-end forms, doubles the sensitivity of analysis. After 120 min, The β-glycosylamines were completely hydrolyzed to free-reducing-end forms. The relative percent of the three most abundant glycans was maintained throughout the experiment.

Figure 8: Kinetics of hydrolysis of β-glycosylamines to free-reducing-end forms. By varying the gradient delay time, the time the glycans were trapped on the porous graphitized chromatographic column, the glycans were allowed to hydrolyze for 0, 30, 60, and 120 min. The β-glycosylamines eluted between 2.2 and 2.4 min, and their peak intensity decreased over time. The free-reducing-end glycans eluted between 2.5 and 2.7 min, and their peak intensity increased over time. By 120 min, The β-glycosylamines were completely hydrolyzed to the free-reducing-end forms. The calculated distribution of the glycans G0, G1, and G2 was maintained.

Analysis of intact and deglycosylated antibodies

Using a different chip configuration, it is possible to analyze deglycosylated antibody. For this analysis, the microfluidic chip was constructed of two layers rather than three: the PNGase enzyme reactor chamber and the LC–MS C8 packed chip layer. Because laser ablation of polyimide film is used, it is possible to create chips with a wide variety of features and functionality.

By varying the residence time in the reactor chamber, it is possible to determine the time required to completely deglycosylate the antibody. Figure 9 shows the results from the analysis of intact and degylcosylated antibody A1. Figure 9 (a) shows deconvoluted mass spectrum of the intact antibody without deglycosylation (no time in the reactor chamber). Three main peaks were found. The mass difference between consecutive peaks is 163.9 and 161.8 amu, which corresponds to the characteristic mass difference, 162.05, of the terminal galactose on the G0, G1, and G2 glycans. Because A1 has two glycosylation sites, the peaks suggest a combination of G0, G1, and G2 glycans attached. Specifically, the masses indicate that A1 has the following glycan attachment configurations: G0 and G0, G0 and G1, G1 and G1, or G0 and G2.

Figure 9: Characterization of intact and deglycosylated antibody Ab1. Figure 9 (a) shows the deconvoluted mass spectrum of the intact antibody Ab1 prior to deglycosylation. The three peaks suggest a combination of G0, G1, and G2 glycan modifications. The deconvoluted mass spectrum in Figure 9 (b) shows the partially deglycosylated antibody following a 3-s deglycosylation in the reactor chamber. Figure 9 (c) shows the single-peak mass spectrum of the completely deglycosylated antibody following a 6-s residence time in the reactor chamber. mAb is monoclonal antibody.

Figure 9 (b) shows the partially deglycosylated antibody following a 3-s deglycosylation in the reactor chamber. Partially and completely deglycosylated A1 was measured. The peaks represent the partially deglycosylated forms as follows: 147,645.93 amu is from the A1 G0 form; 147,806.94 amu is the A1 G1 form; 146,201.21 amu is the deglycosylated form. The peak measuring 149,089.41 amu is intact A1 with two G0 glycans. The high mass accuracy of the TOF-MS enables detection of glycans differing by only one sugar monomer on very high molecular weight antibodies. Figure 9 (c) shows the deglycosylated antibody peak after 6-s residence time in the reactor chamber. Nearly all of the antibody was deglycosylated. The data indicated that this second chip configuration can be used to rapidly analyze deglycosylated antibody without the need for time-consuming sample preparation, deglycosylation, and separation steps.

Conclusion

Until today, N-Linked glycan characterization workflows were tedious and time-consuming, making them unsuitable for real-time bioprocessing applications. A microfluidic LC–MS chip approach is described that integrates and automates all analysis steps: glycan cleavage, capture and purification, chromatographic separation, and MS detection. This approach reduces the previously lengthy incubation time for deglycosylation with PNGase F from many hours to 6 s. It eliminates reaction steps, including acidification, labeling, or derivatizing the free glycans used with traditional LC and MS approaches. Sample cleanup and labeling steps common to LIF detection, and acidification and desalting steps common to MALDI-TOF-MS are not needed. Rapid TOF-MS detection following deglycosylation makes it possible to measure -glycosylamine intermediates directly. Compared with measuring free reducing end glycans, this approach provided the advantages of simpler chromatographic peaks, isomer detection, and enhanced sensitivity.

In total, the integrated microfluidic LC–MS chip reduces analysis time from antibody injection to LC–MS results to 10 min and is completely automated once sample is introduced. It was also shown to separate isomers quickly and efficiently. The chip has the advantage of being able to characterize very complex glycan profiles. Low-level, unknown and nonfucosylated glycans were easily identified. Using a different LC–MS chip configuration, it is possible to analyze the deglycosylated antibody. The authors believe that this microfludic LC–MS chip-based workflow will facilitate fast and accurate glycan characterization in all stages of glycoprotein development and processing.

Maggie Bynum* is a scientist at Agilent Technologies' Agilent Laboratories, 5301 Stevens Creek Blvd, Santa Clara, CA 95051, tel. 408.553.3152, fax 408.553.2161, maggie_bynum@agilent.com. Tomasz Baginski is a scientist, and Rodney Keck is a senior scientist and senior group leader at Genentech. Kevin Killeen is a director at Agilent Technologies' Agilent Laboratories.

*To whom all correspondence should be addressed.

References

1. R.L. Shields et al., J. Biol. Chem. 277 (30), 26733–26740 (2002).

2. C.H. Chung et al., New Eng. J. Med. 358 (11), 1109–1117 ( 2008).

3. B.S. Kendrick et al., BioPharm Intl. 22 (8), 32–44 (2009).

4. P. Hermentin et al., Anal. Biochem, 203 (2), 281–289 (1992).

5. K.R. Anumula, Anal. Biochem. 283 (1), 17–26 (2000).

6. A. Guttman et al., Anal. Biochem. 233 (2), 234–242 (1996).

7. D.I. Papac et al., Glycobiology 8 (5), 445–454 (1998).

8. L.A. Gennaro and O. Salas-Solano, Anal. Chem.80 (10), 3838–3845 (2008).

9. G.R. Guile et al., Anal. Biochem. 240 (2), 210–226 (1996).

10. M. Wuhrer et al., Mass Spec. Rev. 28 (2), 192–206 (2009).

11. M.R. Ninonuevo et. al., Agric. Food Chem. 54 (20), 7471–7480 (2006).

12. N. Tao et al, J. Dairy. Sci.92 (7), 2991–3001 (2009).

13. M.A. Bynum et al., Anal. Chem. 81 (21), 8818–8825 (2009).

14. H. Yin et al., Anal. Chem, 77 (2), 527–533 (2005).

15. Consortium for Functional Glycomics, www.functionalglycomics.org/fg/site_guide/about.shtml, accessed Feb. 18, 2010.