Quantification of the α2-6 sialic acid linkage in branched N-glycan structures with capillary nanogel electrophoresis

Abstract

Sialylation and sialic acid linkage in N-glycans are markers of disease, but are analytically challenging to quantify. A capillary electrophoresis method is reported that integrates a unique combination of enzymes and lectins to modify sialylated N-glycans in real time in the capillary so that N-glycan structures containing α2–6 linked sialic acid are easily separated, detected, and quantified. In this study, N-glycans were sequentially cleaved by enzymes at the head of the separation capillary so that the presence of α2–6 linked sialic acids corresponded to a shift in the analyte migration time in a manner that enabled interpretation of the N-glycan structure. Following injection, only afucosylated N-glycan structures were passed through enzyme zones that contained α2–3 sialidase, followed by β1–3,4 galactosidase, which cleaved any terminal α2–3 linked sialic acid and underlying galactose yielding a terminal N-acetyl glucosamine. With this treatment complete, a third zone of α2–3,6,8 sialidase converted the remaining α2–6-linked sialic acid to terminal galactose. With these enzyme processing steps, the α2–6 linked sialic acid residues on an N-glycan correlated directly to the number of terminal galactose residues that remained. The number of terminal galactose residues could be interpreted as a stepwise decrease in the migration time. Complex N-glycans from alpha-1-acid glycoprotein were analyzed using this approach, revealing that a limited number of α2–6 linked sialic acids were present, with biantennary, triantennary, and tetraantennary N-glycans of alpha-1-acid glycoprotein generally containing 0 or 1 α2–6 linked sialic acid.

Graphical Abstract

INTRODUCTION

Asparagine-linked carbohydrates, or N-glycans, are a type of post-translational protein modification that impacts physiological function. N-linked glycans have a common core structure¹ which can be modified such that the non-reducing terminus is capped with sialic acid that is α2–3 or α2–6 linked to galactose residues. Sialic acid linkage is a direct measure of altered synthesis, metabolism, and regulation because the processes that affect sialic acid linkage are controlled by glycosyltransferases, glycosidases, and a variety of feedback mechanisms.^2–4 Increases in the amount of α2–6 linked sialic acids in serum proteins is reported in cancer.^3,5 For example, the sialic acid linkages of alpha-1-acid glycoprotein (AGP) in serum are associated with liver, pancreatic, and kidney cancer.^6,7

Analyses of N-glycan isomers with different sialic acid linkages are challenging. Sialylation can be heterogeneous and sialic acid residues are labile. Currently the measurement of sialic acid linkage is achieved utilizing chemical derivatization^8–10 or tandem MS.¹¹ Moreover, the linkage specific enzyme α2–3 sialidase is used to further verify the linkage composition.^8,12 Because the distribution of N-glycan structures is complex, glycan samples must be fractionated prior to analyses. Promising strategies for fractionating complex N-glycan samples include HILIC,^13,14 ion mobility spectrometry,^12,15 and microfluidic or capillary electrophoresis.^8,10,12

Capillary electrophoresis has high separation efficiency and high peak capacity, making it ideally suited for complex N-glycan samples.¹⁶ Because capillary electrophoresis separations are based on differences in charge-to-size ratio, they are affected by the highly charged nature of sialylated N-glycans. Thus, sialic acids are best resolved using reverse polarity and suppressed electroosmotic flow. Under these conditions, the anionic glycans are driven to the detector via electrophoretic transport. If a high viscosity media is also included in the background electrolyte, additional separation enhancement is observed as subtle differences in shape (i.e. sialic acid linkage isomers) are intensified by an increase in the frictional drag.

Capillary nanogel electrophoresis has been used to simultaneously separate and identify glycan structures.^17–19 In this approach a stationary zone of exoglycosidase enzyme was included near the injection end of the separation capillary. The enzyme in the capillary cleaved specific terminal N-glycan monomers. For neutral residues such as galactose or N-acetylglucosamine, this resulted in a change in the size of the glycans, but not the charge, when monomers were cleaved. This created a migration time shift of product from that of the substrate allowing identification and quantification. This approach was effective for analyzing mixtures of asialylated N-glycans,^18–20 or a single sialylated N-glycan for sialic acid linkage.¹⁸ Sequencing complex mixtures containing sialic acids with capillary electrophoresis and exoglycosidases is more difficult because removing the sialic acid changes both charge and size. For complex N-glycans containing both α2–3 vs α2–6 linked sialylation, the products do not have predictable charge-to-size ratio confounding interpretation of the electropherogram without the use of expensive glycan standards.

In this study, enzymes were patterned in an order that simplified the mass shift interpretation, enabling the analysis of complex mixtures of α2–3 and α2–6 linked sialylation. The approach was used for N-glycans with different branching and consumed only nanoliter volumes of enzyme, sample and phospholipid. The N-glycans injected into the capillary were passed through three enzyme zones. The first enzyme zone contained α2–3 sialidase, which specifically cleaved α2–3 linked sialic acids. The modified N-glycan then migrated through the second zone that contained β1–3,4 galactosidase, which cleaved terminal galactose. Employing α2–3,6,8 sialidase then removed the α2–6 linked sialic acids. With this process of in-capillary enzyme modification and separation, shifts in migration time reflected the α2–6 linked sialic acid composition and the amount of α2–6 linked sialic acid correlated directly to the number of galactose residues on the N-glycan. This enzyme-based approach provided quantitative information about the amount of α2–6 linked sialic acid for biantennary, triantennary, and tetraantennary N-glycans. The multiple enzyme approach was validated with a sialylated N-glycan standard and used together with lectins to quantify α2–6 linked sialic acid in AGP which is a heavily sialylated positive acute phase protein which serves as a marker of disease.

MATERIALS AND METHODS

Chemicals and reagents

The sialylated triantennary complex N-glycan (GKC-335300), Human AGP N-Linked Glycan Library (GKLB-001), α2–3 sialidase (GK80021 Sialidase S,) and the β1–3,4 galactosidase (GKX-5013) were from Agilent (Santa Clara, CA, formerly Prozyme Hayward, CA). Acetic acid was from Fisher Scientific (Pittsburgh, PA). Sodium phosphate, 8-aminopyrene-1,3,6-trisulfonic acid, triethylamine, AGP, acetonitrile, sodium hydroxide, methanol, sodium acetate, α2–3,6,8 sialidase (N2876–6UN) and sodium cyanoborohydride (dissolved in tetrahydrofuran) were from Sigma-Aldrich (St. Louis, MO). Aleuria aurantia lectin (AAL), Erythrina Cristagalli lectin (ECL), and Maackia Amurensis lectin (MAL) were from Vector Labs (Burlingame, CA). PNGase F was purchased from New England Biolabs (Ipswich, MA). Deionized water was from an Elga Purelab ultra water system (Lowell, MA). The phospholipids 1,2-dimyristoyl-sn-glycero-3phosphocholine (DMPC) and 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC), were from Avanti Polar Lipids (Alabaster, AL). Nanogels were prepared by weighing phospholipids and combining them such that [DMPC]/[DHPC] = 2.5. The phospholipids were dissolved in an aqueous solution of 50 mM sodium acetate buffered to pH of 5.0 to achieve a concentration of 10, 20, or 40% weight/volume. The process to reconstitute phospholipids has been described previously.^19–21 Each preparation was then aliquoted and stored at −20 °C.

Preparation and derivatization of standards

The glycans were labeled with a fluorescent chromophore as previously described.^{19, 20,22} The sample 1 glycans were cleaved from AGP as described previously.²⁰ Glycan samples were purified using DPA-6s columns and stored at −20 °C as described previously.²² For further analysis glycan samples were diluted at least 50-fold in 1mM sodium acetate buffered to pH 5.

All enzymes were reconstituted so that the appropriate concentration of enzyme was in 20% phospholipid buffered at pH 5 with 50 mM sodium acetate. The separation and enzyme zones were buffered at pH 5, which was compatible with enzyme conversion^23–25 and was near the isoelectric point of the enzymes. At this pH the enzymes will be essentially stationary in the capillary in the absence of electroosmotic flow. The α2–3,6,8 sialidase powder was reconstituted to a concentration of 250 mUnits/μL with 50 mM sodium acetate buffered to pH 5. The appropriate volume of master stock (1 μL) was diluted with 20% nanogel to a final enzyme concentration of 5 mUnits/μL. For the α2–3 sialidase, 1 μL of enzyme was reconstituted in 20% nanogel in sodium acetate phospholipid buffered at pH 5 to a concentration of 2 mUnits/μL. The appropriate volume of master stock was diluted with 20% nanogel to make concentrations between 5 μUnits/μL and 750 μUnits/μL. For analysis using β1–3,4 galactosidase, this enzyme was desalted and concentrated to 20 mUnits/μL using a 10 kDa molecular weight cut-off filter (UFC01025 EMD Millipore, Burlington, MA). The 20 mUnits/μL preparation was mixed with 40% nanogel to achieve a solution with final concentrations of 10 mUnits/μL β1–3,4 galactosidase in 20% nanogel buffered to pH 5 with sodium acetate.

Lectins (i.e. AAL, ECL, MAL) were desalted prior to use. The salts were removed using a 10 kDa molecular weight cut-off filter. The lectins were rinsed with a solution 50 mM sodium phosphate buffered to pH 6, and each lectin was then collected from the molecular weight cut-off filter. Each lectin sample was then mixed 1:1 with a 40% nanogel preparation that contained 50 mM sodium acetate buffered to pH 5. The final pH was approximately 5.5. The final concentration of the lectin preparation was 250 μM AAL, 150 μM ECL, or 100 μM MAL in 20% nanogel.

Capillary electrophoresis

Analyses were performed using a SCIEX MDQ Plus (Sciex, Redwood City, CA) configured by the manufacturer with laser induced fluorescence detection and appropriate filters (3 mW solid state laser with λ_ex = 488 nm, λ_em = 520 nm). Silica capillaries with a 25 μm internal diameter and 360 μm outer diameter fused silica capillary (Polymicro Technologies, Phoenix, AZ) were used for the separations. Each day capillaries were prepared as previously described.²² The electrophoresis was performed using reversed polarity to separate the negatively-charged APTS-labeled glycans (anodic reservoir near detector, cathodic reservoir near the site of injection). The background electrolyte was sodium acetate buffered to pH 5. Ambient thermal control of the room and instrument was maintained using a portable air conditioner as was described previously.²⁶ Prior to each run the capillary was held at 15 °C and the 40% nanogel was introduced in the capillary.

For enzyme patterning, the zones were introduced in the opposite order they are in contact with the glycan (see Figure 1). Patterning in the capillary was an automated process performed by the capillary electrophoresis instrument in multiple steps. First, the capillary was filled with 40% nanogel. Second, the α2–3,6,8 sialidase enzyme zone was introduced at 20.7 kPa (3 psi) for 14 s followed by injection of background electrolyte at 3.4 kPa (0.5 psi) for 5 s. The purpose of the 5 s introduction of background electrolyte was to clean the outer surface of the capillary. A 10% lipid spacer at 20.7 kPa (3 psi) 20 s was then introduced to prevent contamination. Third, the β1–3,4 galactosidase was introduced at 20.7 kPa (3 psi) for 42 s followed by injection of background electrolyte at 3.4 kPa (0.5 psi) for 5 s, and then a 10% lipid spacer at 20.7 kPa (3 psi) 20 s. Fourth, the α2–3 sialidase was then introduced at 20.7 kPa (3 psi) for 14 s followed by injection of background electrolyte at 3.4 kPa (0.5 psi) for 5 s, and then a 10% lipid spacer at 20.7 kPa (3 psi) 20 s. Finally, the AAL zone was patterned into the capillary at 20.7 kPa (3 psi) for 28 s followed injection of 10% nanogel at 13.8 kPa (2 psi) for 23 s and then a background electrolyte pre-plug of 13.8 kPa (2 psi) for 20 s. After the capillary was filled, the temperature of the separation was increased to 23 °C for the sample injection, incubation, and separation. Sample injections were at 8 kV 5 s followed by a background electrolyte post-plug of 3.4 kPa (0.5 psi) for 5s. After a run was complete, a 276 kPa (40 psi) background electrolyte flush for 15 min was applied in the reverse direction to push out any remaining protein toward the site of injection. Data collection and analyses were performed using 32 Karat Software version 10.2. Reducing the velocity while passing through the enzyme zone maximized enzyme conversion.¹⁷ Therefore, a pre-separation at 4kV for 9 minutes in the beginning of the run was performed to drive substrate through the enzyme zones.

Figure 1.

(A) The presence of α2–6 linked sialic acid is related to galactose residues following processing with serial enzymes. Saccharides are represented with nomenclature defined by the Consortium for Functional Glycomics. (B) Electropherogram showing migration time shifts based on the number of terminal galactose residues present. (C) Verification of the α2–6 linked sialic acid distribution following treatment with α2–3 sialidase only. Areas labeled in the traces are normalized to the internal standard (IS). Y-axes for trace C is normalized to the internal standard in trace B. Runs were performed in triplicate. See Tables S-1 and S-2 in the supporting information for areas of replicate runs. Separations performed at 23 °C in a 25 μm i.d. capillary, with an effective length of 50 cm and E=450V/cm (reverse polarity). Pre-separation at 4kV for 9 min to drive substrate past enzyme plug. All enzymes are suspended in 20% nanogel with 50 mM sodium acetate buffered to pH 5.

Statistical analysis

Significance testing was used to compare peak areas for optimization of residue cleavage from N-glycans for different enzymatic methods. The samples were considered to be normally distributed, assuming only random error existed in the measurement, and independent. Levene’s test was used to remove bias from the statistical analysis when choosing between a pooled t-test and a Welch test. In every case, at the 0.01 significance level, the results of Levene’s Test did not support a significant difference in sample (group) variance, indicating a 2-sided pooled t-test be conducted for all comparisons

RESULTS AND DISCUSSION

Serial enzyme processing to determine α2–6 linked sialic acid

The amount of α2–6 linked sialic acid was quantified by modifying the N-glycan so that the number of α2–6 linked sialic acids was related to the number of galactose residues remaining after sequential in-capillary digestion. As depicted in Figure 1, the separation capillary was patterned so that the APTS labeled N-glycan passed through three successive zones which were α2–3 sialidase, β1–3,4 galactosidase, and α2–3,6,8 sialidase. Migration through the first two zones resulted in cleavage of any α2–3 linked sialic acid (zone 1) and the penultimate galactose (zone 2) attached to the cleaved α2–3 linked sialic acid (Figure 1A first arrow). Migration through the third zone cleaved the α2–6 linked sialic acid (Figure 1A second arrow). As a result, the galactose residues remaining on the products were those that were linked to the α2–6 linked sialic acids. A unique advantage of this method is that it can be applied to fully or partially sialylated N-glycans because the lack of a sialic acid does not impact the quantification of the α2–6 linked sialic acid residues. It was established in the literature²⁷ that α2–3 linked sialic acids are more readily cleaved than α2–6 linked sialic acids because of structural differences. As α2–3 linked sialic acid is more susceptible to loss associated with sample handling, derivatization, or storage, quantification of α2–6 linked sialic acid is a viable strategy for linkage analyses of physiological samples.

Quantifying α2–6 linked sialic acid in a triantennary N-glycan standard

A sialylated N-glycan standard was enzymatically processed in-capillary to demonstrate the feasibility of enzyme-based quantification of α2–6 linked sialic acid. The top trace of Figure 1B shows the stepwise shift in migration that occurred with each additional loss of a galactose residue. The structures were identified by migration order to determine the number of α2–6 linked sialic acid residues, which was measured as area. The α2–6 linked sialic acid distribution on the triantennary N-glycans summarized in Figure 1B was similar to that reported by the vendor,²⁸ which for the triantennary structures containing 0, 1, 2, and 3 α2–6 linked sialic acids was reported to be 4%, 51%, 36%, and 5%, with 3% of the composition ambiguous. As shown in Figure 1B, split peaks were observed for the mono- (i.e. 23.1, 23.4 min) and di-sialylated (i.e. 24.7, 25.1 min) structures. This peak splitting was due to the asymmetry of the structure emanating from the α1–3 and α1–6 mannose at the N-glycan core.²⁹

Verification of serial enzyme processing

The fidelity of three zone enzyme processing was verified by analyzing the migration order and areas obtained for the triantennary N-glycan standard using only the first zone of α2–3 sialidase (Figure 1C). If the peak assignment is correct, then when enzyme conversion is as expected the areas of peaks associated with structures of the same sialic linkage composition will be the same. The order of migration of the peaks obtained with a single α2–3 sialidase treatment was reversed when compared to the order obtained with the 3 enzyme processing. This was due to the difference in mobility associated with the sialic acid charge. Another notable difference was that the peaks obtained with the single enzyme treatment were not symmetrically spaced because the sialic acid residue was charged. The peak splitting of the mono-galactosylated and di-galactosylated structures observed in trace B and in other reports^22,30 for asialylated peaks was not observed in the partially sialylated glycans. This was attributed to the conformation of sialylated residues forming structures with similar hydrodynamic radius. In fully sialylated glycans the α2–6 linked sialic acids were reported to adopt a more compact structure with the penultimate galactose residues.²⁷ In another report fully sialylated N-glycans were reported to adopt structures that allowed the sialic acid to even fold against the N-acetylglucosamine residues at the reducing terminus of the structure.²⁹ With the peak identities associated with each structure assigned, the areas obtained with the 3 enzyme versus the single enzyme treatment were compared. No significant difference was observed in the areas obtained between the single enzyme in Figure 1C (replicates in Table S-2A in the Supporting Information) and the three-enzyme analysis in Figure 1B (replicates in Table S-1A in the Supporting Information) at a 0.01 significance level, when compared using a two-sided pooled T test.

Working range of specific sialidase

Although α2–3 sialidase was specific for α2–3 linked sialic acid, at sufficiently high concentration it will cleave α2–6 linked sialic acid at a slower velocity. The concentration of 250 μUnits/μL α2–3 sialidase used for the analyses in Figure 1 was verified to ensure that the amount of α2–3 sialidase was sufficient to fully convert α2–3 linked sialic acids, but did not cleave α2–6 sialic acid. The same concentration of triantennary standard that was used to quantify the α2–6 linked sialic acid in Figure 1B was used to evaluate the working concentration range of the specific sialidase. A different instrument was used for this study. Although the response of these instruments differs the ratio of the peak area of the N-glycan to that of the internal standard is similar (see Table S-3C in the Supporting Information). In-capillary digestions were performed at α2–3 sialidase concentrations of 250 μUnits/μL (Figure 2B) and 750 μUnits/μL (Figure 2A) and the distribution of desialylated peaks compared. The peaks produced by these in-line enzyme reactions had reproducible migration times and areas, with the relative standard deviation less than 1% and less than 10%, respectively. The areas of the product peaks obtained at these two concentrations were not significantly different at a 0.01 significance level when compared using a two-sided pooled T test. When lectin with high specificity to α2–3 sialic acids (MAL) was included in the capillary (Figure 2C), no difference was observed in the trace, which further confirmed that the α2–3 sialic acids were completely cleaved from the sample. A lower concentration of α2–3 sialidase (5 μUnits/μL) generated incomplete cleavage of α2–3 linked sialic acid (Figure 2D).

Figure 2.

The concentration range of α2–3 sialidase enzyme suitable for complete conversion of α2–3 linked sialic acid, but not α2–6 linked sialic acid is established for triantennary standard. Electropherograms obtained at 750 μUnits/μL(A) and 250 μUnits/μL(B) enzyme are not significantly different. When MAL lectin specific for α2–3linkedsialic acid is included with 250 μUnits/μLenzyme (C) no change is observed confirming complete removal of α2–3 linked sialic acid. At concentrations lower than this working range (D) partial conversion is achieved (5 μUnits/μL). Separations condition are similar to Figure 1

For this study it was unnecessary to establish a working range for the β1–3,4 galactosidase or the α2–3,6,8 sialidase because stepwise sequencing was only dependent upon these enzymes being above a minimum threshold for complete digestion. Concentrations of 10 mUnits/μL and 5 mUnits/μL, were used for β1–3,4 galactosidase or the α2–3,6,8 sialidase, respectively. These concentrations were selected as working at higher concentrations increased the potential to foul the capillary surface, which required more substantial flushes in-between the separations to maintain reproducible peak migration times and areas. These experiments confirmed that the enzyme processing strategy effectively quantified the abundance of sialic acid linkages for a single N-glycan standard; however, in order to apply the approach to sialylated glycoproteins it was evaluated with mixtures of N-glycans.

Measuring the α2–6 sialic acid content simultaneously in a mixture of N-glycans

To analyze glycan mixtures, sample migration windows and branching distribution were established from reference traces shown in Figure 3A for AGP, which contains bi-, tri-, and tetraantennary N-glycans. The N-glycans were cleaved from AGP and labeled with APTS prior to in-capillary desialylation and separation. In-capillary enzyme processing cleaved the N-glycans to the terminal N-acetyl glucosamine (Figure 3A blue trace) or galactose residues (Figure 3A black trace). A lectin specific for fucose (AAL) was used to remove fucose containing structures. This produced three peaks (Figure 3A) for which the migration times delineated the range of the stepwise shift for each branched structure. The area of each branched structure obtained from triplicate runs (see Table 1 and Table S-4A in Supporting Information) was used to assist in the analyses and were normalized to a mannose trisaccharide internal standard. The peak overlap of the fully galactosylated biantennary and the agalactosylated tetraantennary structures was addressed in α2–6 sialic acid determinations that followed by including a lectin with specificity to galactose residues (ECL) in processing. This removed the biantennary structure which allowed for quantification of the tetraantennary structure (Figure 3B blue trace).

Figure 3.

(A) reference traces establish the area associated with biantennary, triantennary, and tetraantennary glycans. (B) The α2–6 linked sialic acid area of each branched structure is established through serial enzyme processing in the absence of ECL (black trace). Presence of ECL (blue trace) identifies agalactosylated glycans in co-migrating peaks. Separation conditions are as specified in Figure 1. The asterisk denotes contaminant peaks.

Table 1.

Peak area of α2–6 linked sialic from AGP glycans

Branch						Total
	25(10)	89(7)	0	—	—	114(12)
	104(5)	336(8)	0	0	—	440(9)
	94(7)	265(3)	0	0	0	359(8)

Areas are reported in units of 10³ RFU. The areas and percent relative standard deviations (RSD) of 2, 3, and 4 branch N-glycans from normalized alignment traces are: 107 (4), 443 (1), and 374 (2), respectively. Normalization is based on the highest quantified area for the internal standard within the alignment traces, sequencing traces, and sequencing traces with ECL (i.e. 35,090 RFU). Solid lines in the table indicate that the structure is not possible. Separation conditions are as stated Figure 1.

Working range of specific sialidase for an N-glycan mixture

Although the AGP was a commercial standard, the amount of N-glycan present in the sample was reported as total N-glycan mass, but the concentration of each branched structure was not specified by the manufacturer. The bands of different N-glycans in the AGP sample, which passed through the enzyme zones at different velocities and times, were adjusted to maintain the enzyme working range established for the triantennary standard. The AGP sample was diluted so that the total peak area of triantennary N-glycan of the AGP sample matched the peak area of the triantennary standard used in Figure 1. While the areas of the tetraantennary and triantennary AGP N-glycans were similar, the area of biantennary N-glycan was 4-fold lower. As a result, the applicability of the working range for the AGP sample was validated.

The working range of α2–3 sialidase was established as described for the triantennary standard. Three replicates of in-capillary digestions were performed at α2–3 sialidase concentrations of 750 μUnits/μL and 250 μUnits/μL (Figure S-1 and Table S-5A in the Supporting Information) and the distribution of desialylated peaks compared. The peaks produced by these in-line enzyme reactions had reproducible migration times and areas, with the relative standard deviation less than 1% and less than 10%, respectively. The areas of the product peaks obtained at these two concentrations were not significantly different at a 0.01 significance level when compared using a two-sided pooled T test. When lectin with high specificity to α2–3 sialic acids (MAL) was included in the capillary no statistical difference in areas, which further confirmed that the α2–3 sialic acids were completely cleaved from the sample. A lower concentration of α2–3 sialidase, 50 μUnits/μL, showed incomplete cleavage of α2–3 linked sialic acid. Two different instruments were used for the sialic acid analysis and the range study. However, the ratio of N-glycan to internal standard was the same because the same concentrations were used (see Table S-5C in the Supporting Information).

Enzyme processing

Serial enzyme processing was then applied to the AGP N-glycans to identify the α2–6 linked sialic acid distribution. The migration windows for the bi-, tri- and tetraantennary peaks with different numbers of galactose residues are emphasized with arrows in Figure 3A. The enzyme processing was performed in the absence (Figure 3B black trace) and presence (Figure 3B blue trace) of lectin with affinity to galactose bearing structures (i.e. ECL) to distinguish any comigrating peaks. From these runs areas were determined by assigning the biantennary structures to peaks until the summed area for biantennary peaks equaled the corresponding area established from the reference traces in Figure 3A. This process was repeated for the triantennary and then for the tetraantennary structures. Peak splitting observed with the triantennary standard (Figure 3B) due to differences in the α1–3 and α1–6 arm mannose linkages in the core were also observed in the AGP mixture (Figure 3B, black trace) for the mono-galactosylated structures that are biantennary (19.6 min, 19.9 min), triantennary (20.8 min, 21.1 min), and tetraantennary (22.1 min, 22.4 min). The agalactosylated biantennary peak (18.6 min) is broad because some of the biantennary N-glycans were partially desialylated and differences in mobility of the fully and partially sialylated structure induced this broadening.

Comparison of the α2–6 sialic acid content from different sources of AGP

A second sample, which was a commercially available source of free N-glycans cleaved from AGP, was analyzed with the same method to determine α2–6 sialic acid composition. The second AGP sample was diluted so that the total peak area of triantennary N-glycan of the AGP sample matched the peak area of the triantennary standard used in Figure 1. The areas of the biantennary and tetraantennary N-glycan peaks were 1.4-fold and 2.3-fold lower than that of the triantennary N-glycan. The working range of α2–3 sialidase was confirmed and the area of each branched structure was normalized to the internal standard (see Table S-6A and Figure S-2 in the Supporting Information). Two different instruments were used for the sialic acid analysis and the range study. However, the ratio of N-glycan to internal standard was the same because the same concentrations were used (see Table S-7C in the Supporting Information).

The second sample contained more unreacted APTS than sample 1. Additionally, the total glycan area of the second sample was approximately 2.4 times higher than that of sample 1, shown in Figure 3. As a result, more contaminant peaks, which are labeled with an asterisk (see Figures S-3 and S-4 in the Supporting Information), were evident in the second sample. The contaminant peaks were distinguishable from the N-glycan peaks by comparing the signal obtained from serial enzyme processing performed in the absence and presence of ECL (see Figure S-4 in the Supporting Information). Although different branching ratios were observed for sample 1 as compared to sample 2 for the biantennary (13% vs 21%), triantennary (51% vs 47%), and tetraantennary (36% vs 31%) structure, these branching distributions were within the ranges reported in the literature.^6,31,32 Interestingly, the predominant form of α2–6 sialic acid in each branched structure contained only a single α2–6 sialic acid. Small amounts of tetraantennary N-glycan containing two to four α2–6 sialic acids were observed (see Table 2). This was expected as α2–3 sialic acids adopt an extended linear conformation; whereas, α2–6 sialic acids are oriented toward the mannose core forming a more compact glycan structure.²⁷ Although quantification of α2–6 linked sialylation in AGP had not previously been reported, AGP has been reported to contain monogalactosylated,⁸ as well as di-, and tri-galactosylated tetraantennary structures.³³

Table 2.

Peak area of α2–6 linked sialic from AGP glycans

Branch						Total
	56(1)	339(3)	0	—	—	395(3)
	226(3)	649(6)	0	0	—	876(7)
	216(5)	339(6)	19(10)	52(7)	19(10)	645(6)

Areas are reported in units of 10³ RFU. The areas and percent relative standard deviation (RSD) of 2, 3, and 4 branch N-glycans from alignment trace normalized to internal standard are: 380 (2), 890 (4), and 630 (4), respectively. Normalization is based on the highest quantified area for the internal standard within the alignment traces, sequencing traces, and sequencing traces with ECL (i.e. 81,675 RFU). Solid lines in the table indicate that the structure not possible. Separation conditions are as stated Figure 1.

CONCLUSIONS AND FUTURE DIRECTIONS

Quantification of the amount of α2–6 sialic acid in N-glycan structures was achieved using an enzyme based method of glycan modification. With the use of nanogel, enzymatic conversion of N-glycans was achieved in real-time and consumed only nanoliter volumes of enzymes or lectins for each analysis. Capillary electrophoresis analyses of mixtures of sialylated N-glycans are complex because of the charge on the sialic acids. Additionally, N-glycans can be desialylated when exposed to heat and/or to different solutions used in conventional sample processing. Moreover, differences in the stability of α2–6 vs α2–3 linked sialic acids are found in the literature. By incorporating the enzymes as outlined in this report, a predictable mobility shift based on a change in the charge-to-size ratio of the N-glycan was achieved. With this method, the α2–6 linked sialic acid linkage of N-glycans was determined for a pure standard as well as for mixtures of biantennary, triantennary, and tetraantennary structures derived from AGP. Moreover, the exact site of exact site of α2–6 linked sialylation in biantennary and potentially triantennary N-glycans can be inferred based on the difference in hydrodynamic radius for the α1–3 mannose, which is more extended,²⁹ creating a larger hydrodynamic radius and a slower electrophoretic migration in the electric field observed in capillary electrophoresis separations of N-glycans.^34–36 Determination of sialic acid linkage is critical to validating ongoing research on the use of α2–6 linked sialic acid as a biomarker of disease. Future work centers on the application of this technology to glycoproteins as well as glycan standards to authenticate the degree of sialylation and sialic acid linkage.