High-Sensitivity Glycan Profiling of Blood-Derived Immunoglobulin G, Plasma, and Extracellular Vesicle Isolates with Capillary Zone Electrophoresis-Mass Spectrometry

Abstract

We developed a highly sensitive method for profiling of N-glycans released from proteins based on capillary zone electrophoresis coupled to electrospray ionization mass spectrometry (CZE-ESI-MS) and applied the technique to glycan analysis of plasma and blood-derived isolates. The combination of dopant-enriched nitrogen (DEN)-gas introduced into the nanoelectrospray microenvironment with optimized ionization, desolvation, and CZE-MS conditions improved the detection sensitivity up to ~100-fold, as directly compared to the conventional mode of instrument operation through peak intensity measurements. Analyses without supplemental pressure increased the resolution ~7-fold in the separation of closely related and isobaric glycans. The developed method was evaluated for qualitative and quantitative glycan profiling of three types of blood isolates: plasma, total serum immunoglobulin G (IgG), and total plasma extracellular vesicles (EVs). The comparative glycan analysis of IgG and EV isolates and total plasma was conducted for the first time and resulted in detection of >200, >400, and >500 N-glycans for injected sample amounts equivalent to <500 nL of blood. Structural CZE-MS² analysis resulted in the identification of highly diverse glycans, assignment of α-2,6-linked sialic acids, and differentiation of positional isomers. Unmatched depth of N-glycan profiling was achieved compared to previously reported methods for the analysis of minute amounts of similar complexity blood isolates.

Graphical Abstract

INTRODUCTION

Glycosylation of proteins is one of the most common post-translational modifications (PTMs) and is responsible for controlling many biological phenomena, including cell–cell and cell–matrix interactions, protein folding, receptor binding, and protein clearance.^1,2 The majority of blood plasma proteins are N- and/or O-glycosylated. Changes in glycosylation (e.g., decrease or increase of fucosylation, bisection, galactosylation, or sialylation) may be associated with a number of pathologies, including hereditary, immune, cardiovascular, inflammatory, and oncological diseases.^3–5 Altered glycosylation has been observed in the glycocalyx of cancer cells. It was reported that increased core fucosylation was associated with lung and breast cancers, and that high-grade tumors were correlated with polysialic acid expression.^6,7 Higher levels of 5-N-glycolyl-neuraminic acid, which is not endogenously synthesized in humans but consumed from dietary sources and displayed on the cell glycocalyx, were reported in ovarian, prostate, and colon cancers.^8,9 These findings inspired a significant interest in the discovery of novel biomarkers based on changes in glycosylation patterns in biofluids such as serum and plasma for early detection of various types of cancer.^8,10 Recent advances in the development of analytical techniques and instrumentation, especially in mass spectrometry (MS), made reliable identification of multiple blood-derived glycans practical using glycomic strategies. Altered N-glycan profiles in serum and plasma of patients diagnosed with breast, prostate, ovarian, liver, or lung cancer have recently been reported.^11,12 Minimally invasive collection of blood samples (liquid biopsy) appears as an attractive alternative to surgical biopsies of solid tissues for diagnostic and prognostic purposes.

Advanced glycoprofiling techniques can be applied to the characterization of whole serum, plasma, or other blood-derived specimens, e.g., immunoglobulin G (IgG) and extracellular vesicle (EV) isolates, to address multiple urgent biomedical needs. The high abundance of IgG in plasma and serum and its prominent role in the immune system make its glycome an ideal system to examine for changes in cancer³ and other pathologies. Several recent studies reported altered IgG glycosylation patterns in patients with gastric¹³ and ovarian^14,15 cancers.

EVs are submicron-size bilayer phospholipid membrane-enclosed blobs released from cells into physiological fluids.^16–18 Significant heterogeneity of EV subpopulations, which include exosomes, microparticles, apoptotic bodies, apolipoprotein complexes,¹⁷ and exomers,¹⁹ is expected based on their diversity of biogenesis, size, cellular/tissue origin, protein composition, mRNA and microRNA content, and biological function.^17,20 With increasing evidence of their important roles in cell–cell communication, representation of pathological alteration of molecular profiles in affected tissues, and relevance in the transmission of pathogenic and signaling molecules in diseases, EVs have been exploited as attractive sources for biomarkers for clinical diagnostics and vehicles for therapy delivery.^{10,17,21–26} Characterization of EV proteins, and especially their PTMs, is challenging because of the low abundance of EVs in highly complex biological fluids. EV proteins and their glycosylation moieties among other PTMs are expected to represent tissues of origin and biological states of the EV-producing cells. Comprehensive characterization of variations in EV proteins and their glycosylation could contribute to the development of novel diagnostic, prognostic, and therapy delivery strategies, biomarker-based surveillance for disease progression, and treatment efficacy.

N-glycans are typically analyzed by high-performance liquid chromatography (HPLC) using different modes, including reversed-phase, hydrophilic interaction, and anion exchange. Capillary electrophoresis (CE) is an alternative powerful separation technique for glycan analysis. With its high resolving power and high selectivity, CE can resolve both linkage and positional isomers. Capillary zone electrophoresis (CZE) interfaced with MS enables unequivocal structural characterization through both MS and tandem MS.^27,28

The analysis of released oligosaccharides from biological sources commonly requires a glycan derivatization step to facilitate the separation and enhance the detection of glycans. A wide variety of derivatization strategies with a large number of different labels have been developed.^29–31 CE with laser-induced fluorescence (LIF) detection of glycans is now routinely achieved using 8-aminopyrene-1,3,6-trisulfonic acid (APTS). CE-LIF of APTS-labeled glycans was utilized to detect and quantify N-glycans derived from various glycoproteins³² and permitted unmatched isomeric separation.³³ However, the identification of glycans via this strategy greatly depends on the availability of standards, and only a handful of applications were demonstrated using effective but rather tedious experimental workflows involving exoglycosidases.^34,35 Such limitations can be alleviated by interfacing CE with MS.

Significantly different ionization efficiencies of neutral and acidic glycans, along with the generally low ionization efficiency of nonlabeled glycans, make MS characterization of glycans challenging. Derivatization with labeling reagents such as APTS is often used to enhance ionization.³⁶ In a recent study, Jacobson et al. developed a CZE-MS method to analyze N-glycans that were released from human serum and labeled with APTS.³⁷ However, the majority of structural identities of the 77 detected glycan-like features could not be confirmed by MS² fragmentation because of their low abundance.

Recently, Kammeijer et al. evaluated the use of sheathless CE-ESI-MS combined with dopant-enriched nitrogen (DEN)-gas introduced into the space between the electrospray emitter and an inlet into a mass spectrometer for glycan and glycopeptide analysis.^38,39 Similar to the previously shown improved desolvation and ionization efficiency with the overall enhanced sensitivity in LC applications,⁴⁰ improved sensitivity levels for evaluated glycans and glycopeptides were observed compared to the conventional sheathless CE-ESI-MS.

In the present study, we developed a high-sensitivity CZE-MS method for profiling of N-glycans at fmol level released from complex biological samples using a sheathless CZE-MS interface. The method was developed and optimized with dextran ladder standards (maltooligosaccharides). The glycans were labeled with APTS, and several strategies (with or without DEN-gas) to improve the profiling sensitivity were evaluated. DEN-gas combined with optimized ESI conditions and MS parameters resulted in a significant sensitivity improvement (up to 100-fold), as directly compared to the conventional mode of instrument operation through peak intensity measurements. Then, the technique was applied to in-depth profiling of N-glycans released from human blood-derived total plasma, IgG, and EV isolates injected at the levels corresponding to sub-microliter volumes of blood. Negative ESI MS resulted in highly informative MS² spectra, which enabled unequivocal structural identification and assignment of α-2,6-linked sialic acids. Analysis of the three examined blood-derived sample types resulted in unique qualitative and quantitative N-glycan profile characteristic to each specimen type. In comparison to the methods reported for N-glycan profiling of similar complexity blood isolates available in limited amounts, the number of identified glycans was increased ~5-fold.^{37,39,41–44} The developed technique can provide important structural and quantitative information about glycome alterations in biological phenomena and pathologies using limited sample amounts (i.e., liquid micro-biopsies).

EXPERIMENTAL SECTION

Sample Preparation.

APTS-labeled dextran ladder standards and all of the N-glycans analyzed in this study were prepared following the fast glycan kit protocol developed by Sciex (Brea, CA)³³ (see the Supporting Information). Human serum IgG (purity ≥95%, based on nonreduced sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and verified by nanoLC-MS/MS of tryptic digests) was purchased from Sigma-Aldrich (St. Louis, MO). The total EV isolates were prepared from 1 mL of platelet-free anticoagulated with ethylenediaminetetraacetic acid (EDTA) human blood plasma using ultracentrifugation and differential centrifugation, as described in previously published work.⁴⁵ The purity of the EV isolates in respect to the amount of co-isolated free plasma proteins was ≥90–95%, based on a combination of EV counting (using tunable resistive pulse sensing (TRPS) and nano-flow cytometry), and total protein concentration measurements (using UV-spectrophotometry and fluorescence), which was in accordance with other reports⁴⁶ (see the Supporting Information). N-glycans were then released from the total EV isolates following the procedure described in the Supporting Information.

CZE-MS Methods.

CZE-MS experiments were conducted using a CESI 8000 instrument (Sciex, Brea, CA). For all experiments, bare fused silica (BFS) OptiMS capillaries (91 cm × 30 μm i.d. × 150 μm o.d.) were used. No specific adsorption-related phenomena resulting in peak tailing were observed for released N-glycans using these uncoated capillaries. Sample injections were performed at 5 psi for 40 or 60 s (corresponding to 34 and 51 nL injection volumes, respectively) depending on the sample. Unless stated otherwise, the experiments were carried out with a background electrolyte (BGE) of 10 mM (ionic strength) ammonium acetate pH 4.5 with 10% isopropanol. For glycan profiling of IgG, EV, and plasma isolates, CZE-MS analyses were performed using DEN-gas combined with ion transfer tube (ITT) at 150 °C and in-source collision-induced dissociation (ISCID) at 70 eV, and the CZE supplemental pressure (SP) was switched off for 18 min before applying 5 psi during the rest of the analysis. For the repeatability of the CZE method and other experimental details, see the Supporting Information.

MS Instrumentation.

QE Exactive and QE Exactive Plus Orbitrap MS (Thermo Fisher Scientific, Bremen, Germany) were used. All analyses were carried out in negative ESI mode (see the Supporting Information).

Data Analysis.

For data acquisition and processing, Xcalibur (v. 2.8) software was used. CZE-MS data were processed with GlycReSoft (v. 3.10) software (Boston University, Boston, MA). Analyses of CZE-MS² data were performed with SimGlycan (v. 5.91) software (Premier Biosoft, Palo Alto, CA). The generated results were based on the processing of three replicate analyses. The glycan composition identification results were mainly based on CZE-MS data processing using GlycReSoft. As additional verification of the plausible glycan identifications made using GlycReSoft, several supplementary levels of manual data examination were applied, including (1) CZE-MS migration patterns, (2) charge state and isotopic distributions characteristic to glycan ions, (3) detection of neutral losses, and (4) manual examination of CZE-MS² data (see the Supporting Information).

Additional experimental details in accordance with the MIRAGE guidelines⁴⁷ are provided in the Supporting Information. The generated raw data were deposited in GlycoPOST, a dedicated repository for MS-based glycomics⁴⁸ (https://glycopost.glycosmos.org).

RESULTS AND DISCUSSION

Optimization of Detection Sensitivity and Separation Performance in the CZE-MS of N-Glycans.

The potential of APTS has been demonstrated for glycoprofiling of complex biological and clinical samples, including human serum.^37,49–53 To increase the sensitivity and throughput of glycoprofiling, Guttman et al. used carboxyl-activated magnetic beads for the enrichment of first nonlabeled N-glycans released from sample proteins and then APTS-labeled glycans upon completion of the derivatization step.^33,54 A commercially available kit for APTS labeling and enrichment/cleanup using carboxylic magnetic beads was selected for the development of our CZE-MS method for profiling of N-glycans. In our initial experiments, we were able to reliably detect up to 24 glucose units (GU) in the dextran ladder standard injected at a concentration level of ~20 μg/mL (34 nL injected). Several analytical strategies were evaluated to improve the detection sensitivity and separation performance of the CZE-MS method.

Detection Sensitivity Improvement via Control of the Nanoelectrospray Microenvironment and the Use of Dopant-Enriched Nitrogen.

We first evaluated the effects of the microenvironment around the nESI interface and the MS source by either supplying the purified air or the dopant-enriched nitrogen (DEN)-gas into the space where the nESI emitter and the inlet into the mass spectrometer are located. We compared the background noise intensity levels and the spray stability for the following conditions: (1) conventional mode of instrument operation in the atmosphere of the lab environment; (2) the nESI microenvironment filled with the purified air supplied using an Active Background Ion Reduction Device (ABIRD); and (3) the nitrogen gas enriched with either acetonitrile (DEN-gas, ACN) or isopropanol (DEN-gas, IPA) (Figure S1). The purified air conditions resulted in the most significant decrease (~6-fold) of the background ion signal and the highest stability of nESI according to the fluctuations of the TIC. DEN-gas also led to a decrease in the background ion signal, but the decrease was either ~2-fold (IPA) or ~3-fold (ACN) less pronounced compared to the purified air approach (see the Supporting Information).

The three evaluated modes (conventional mode, purified air, and DEN-gas) were then applied to the analysis of the dextran ladder standards. As shown in Figures 1 and S2, the purified air mode led to a 1.2–1.5-fold increase in peak heights and peak areas for oligoglycans with a low degree of polymerization (DP), i.e., ≤13 glucose units. For larger oligosaccharides (DP ≥ 18), a more significant increase in the signal was observed, ranging for DP18–23 from 6- to 10-fold and 7- to 13-fold for peak heights and peak areas, respectively. Using the DEN-gas with IPA (Figures 1 and S2), peak heights and peak areas for glycans with the lowest DPs (≤13) were increased 2–4-fold. For larger glycans (DP18–23), the increase in MS signal ranged from 17- to 20-fold and from 21- to 40-fold for peak heights and peak areas, respectively. The DEN-gas experiments using ACN as a dopant, in our hands, led to the detection of strong contaminant background ions, resulting in ion suppression that was detrimental for the overall profiling sensitivity in the analysis of the dextran ladder standards. Among the three tested modes, the DEN-gas setup yielded the best detection sensitivity as evident from the number of detected glycans (see the Supporting Information) and gains in signal intensity levels as well as peak areas for detected glycans. We noticed that all ratios of abundances for triply/doubly charged ions were higher with DEN-gas, compared to the conventional mode and purified air conditions, based on peak height/area measurements, indicating a shift toward higher charge states when using the DEN-gas (see the Supporting Information and Figure S3).

Figure 1.

Comparison of MS signal intensity levels for glycan peaks in CZE-MS analyses (n = 3) of APTS-labeled dextran ladder standards at the evaluated conditions. (A, B) Comparison of different nESI microenvironment control modes: (1) a conventional mode of instrument operation; (2) a purified air (ABIRD); and (3) a nitrogen gas enriched with IPA. (C, D) Optimization of the MS ion transfer tube (ITT) temperature, without applying any additional control of the nESI microenvironment.

Optimization of the Ion Transfer Tube (ITT) Temperature.

To study the impact of the temperature of the ITT (aka heated capillary) at the inlet into the mass spectrometer, we examined the temperature range from 110 to 300 °C, using the conventional mode of instrument operation. As shown in Figures 1 and S4, the highest signal, as assessed through peak intensity and area measurements, was obtained for 150 °C for all detected DPs, which we interpret as evidence of the improved ionization and desolvation efficiencies. Compared to the experiments carried out at the ITT temperature of 110 °C, selecting 150 °C, peak heights and peak areas were increased 1.5–2-fold for all of the detected oligosaccharides.

In-Source CID.

While the ammonium acetate-based BGE used in our experiments resulted in impressive separation performance, it led to the formation of high-abundance ammonium adduct ions (mass shift of 17.03 Da), with a deleterious effect on the detection sensitivity. The MS signal corresponding to adduct ions was proportionally increased when using DEN-gas in correlation with the enhancement of the signal of glycan ions without adducts. To minimize the formation of ammonium adducts, collisional activation experiments were performed with ISCID set at 50, 70, and 100 eV. The use of 100 eV removed clusters most efficiently (~70%), but substantial evidence of in-source decay, was observed. Seventy electron volt was selected since this CID energy level resulted in a significant decrease (~50%) in the signal intensity of ammonium adducts without detectable in-source decay. At this ISCID level, the detection sensitivity was increased over 3-fold for DPs ≤ 18 and over 10-fold for DPs ≥ 18 (see the Supporting Information).

Combination of DEN-Gas, Increased ITT Temperature, and ISCID.

In the next set of experiments, we evaluated the above-described parameters in combination. The dextran ladder standard was analyzed using the following conditions shown in Figure 2: (1) conventional mode of instrument operation (with ITT temperature set at 110 °C); (2) ITT temperature at 150 °C; (3) ITT temperature at 150 °C with ISCID at 70 eV; and (4) ITT temperature at 150 °C with ISCID at 70 eV in combination with DEN-gas using IPA. As shown in Figure 2, an ITT temperature set at 150 °C resulted in an increase (over 2-fold) in the peak intensities but with no major effect on the adduct formation. On the contrary, ISCID was highly efficient in not only increasing the peak heights but also decreasing the formation of adducts. Finally, the combination of DEN-gas with the other two parameters provided the highest S/N ratios and best sensitivity levels, with almost complete suppression of the ammonium adduct ion species (decrease of >90% in the signal intensity of the predominant ammonium adduct, making the rest of the adductions undetectable).

Figure 2.

Comparison of extracted ion electropherograms (EIEs, panels (A)–(K)) and MS spectra corresponding to DP7 (C–L) in CZE-MS analyses (n = 3) of APTS-labeled dextran ladder standards DP4–DP29 at the following evaluated conditions: (A–C) ITT at 110 °C; (D–F) ITT at 150 °C; (G–I) ITT at 150 °C and ISCID at 70 eV; and (J–L) ITT at 150 °C, ISCID at 70 eV, and DEN-gas with IPA.

Our additional validation experiments confirmed that DEN-gas alone or ISCID alone did not allow achieving the intensity levels obtained from the combination of both parameters applied at the same ITT temperature. We therefore selected a CZE-MS method based on the combination of ITT temperature at 150 °C, ISCID at 70 eV, and DEN-gas with IPA. As directly compared to the conventional mode of instrument operation, this optimized method resulted in an increase of the peak heights ranging from 25- to 45-fold for smaller oligosaccharides (DP ≤ 17) and from 70- to 100-fold for larger oligosaccharides (DP ≥ 18), which allowed detecting, above the LOD (at S/N > 3), oligosaccharides as large as DP46, DP58, and DP62 depending on the selected mass scan range (m/z 500–2000, 500–2500, and 1000–3000, respectively). The largest DPs detected above the LOQ (at S/N > 10) were, respectively, DP41, DP52, and DP55. Using these optimized conditions, additional low-abundance isobaric forms could be detected for several DPs, including DP18, DP23, and DP25, based on the experimental masses of the oligosaccharides that were detected with a mass accuracy ≤2 ppm and the CZE-MS migration patterns of the oligosaccharidic ions.

Optimized Control of Supplemental Pressure.

The initial CZE-MS experiments were carried out applying continuous supplemental pressure (SP) of 5 psi at the inlet of the separation capillary. These conditions resulted in the detection of maltooligosaccharides up to 24 GUs, but glycans were poorly resolved. The first strategy to increase the resolution between the peaks of maltooligosaccharides was to introduce a step without applying SP at the beginning of the CZE-MS analysis. The very low EOF mobility (2.02 × 10⁻⁸ m²/V/s) generated by the BGE consisting of 10 mM ammonium acetate pH 4.5 and 10% isopropanol, lower than the electrophoretic mobilities (μ_ep) of the oligosaccharides, made this step possible. As the longitudinal diffusion increased when the SP was delayed, which resulted in the broadening of the later migrating peaks with a decrease in their signal intensity levels⁵⁵ (see the Supporting Information), the time without SP was carefully optimized to sharpen and compress later migrating peaks without providing noticeable disturbance to the earlier migrating peaks. The application of the SP (5 psi) 18 min after the application of the separation voltage resulted in a significant extension of the separation window from 3.5 min (when the pressure of 5 psi was continuously applied) to 8.5 min, and the resolution increased ~3-fold (e.g., from 1.1 to 3.4 for the DP2–DP3 pair) while preserving the spray stability and signal intensity. The separation performed without SP over the whole duration of the run resulted in a further extension of the separation window to 21 min, and a ~5-fold increased resolution. Applying the parameters that enable the best sensitivity (i.e., DEN-gas with ITT temperature at 150 °C combined with ISCID at 70 eV), the separation window was extended to over 54 min when the SP was not applied at all, and large oligosaccharides (up to DP46) were detected (Figure S5). However, a ~10-fold decrease in signal intensity was observed. μ_ep values of dextran ladder oligosaccharides were measured in these experiments conducted without applying SP, and μ_ep values ranged from 4.57 × 10⁻⁸ to 2.90 × 10⁻⁸ m²/V/s for DP1–DP46, respectively (Figure S6).

CZE-MS ANALYSIS OF HUMAN SERUM IGG

The optimized CZE-MS method was applied to the analysis of N-glycans released from the total isolate of human serum IgG. In each CZE-MS analysis, we injected N-glycans released from ~25 ng of total serum IgG, which corresponds to ~0.5 ng of IgG glycans, and ~3 nL of serum (see the Supporting Information). Signal intensity levels demonstrated a similar trend of changes in comparison to the dextran ladder experiments (see the Supporting Information). As shown in Figures 3A and S7, the combination of DEN-gas, ITT at 150 °C, and ISCID at 70 eV resulted in an increase of the peak heights and peak areas up to 70-fold as directly compared to the conventional mode of instrument operation. The level of in-source decay of glycan structures was either undetectable or minimal (<5%), which did not noticeably impact the qualitative and quantitative characterization of fucosylated and sialylated glycans. The number of N-glycans detected in serum IgG isolates increased in correlation with the signal intensity improvements. We detected 60 ± 5, 110 ± 5, 170 ± 13, and 210 ± 22 nonredundant N-glycan compositions (i.e., the compositional sum of monosaccharides) using the conventional mode, the mode with the purified air, the DEN-gas mode, and the optimized CZE-MS method, respectively. The dramatic changes in the appearance of ion species patterns are evident from the comparison of the ion density maps shown in Figure 3C,D. These ion density maps also illustrate: a. the characteristic CZE-MS migration patterns of glycan ions caused by discrete changes in net charge and hydrodynamic volume and b. the m/z shifts related to each monosaccharide mass increment, which were used for validation of identified glycan compositions (see the Supporting Information).

Figure 3.

Effect of experimental conditions on CZE-MS-based glycan profiling. (A, B) CZE-MS analyses (n = 3) of human serum IgG at five different conditions: (1) conventional mode of instrument operation (ITT at 110 °C); (2) purified air (ABIRD) (ITT at 110 °C); (3) nitrogen gas enriched with IPA (ITT at 110 °C); (4) nitrogen gas enriched with IPA (ITT at 150 °C); and (5) nitrogen gas enriched with IPA (ITT at 150 °C) and ISCID at 70 eV. (C, D) Ion density maps (intensity level: 3 × 10⁵) of the CZE-MS analyses of human serum IgG before (C) and after (D) method optimization.

To improve the separation of different IgG glycans, we included an initial step without SP for 18 min, followed by the application of the SP between 1.5 and 5 psi. This strategy enabled the separation of multiple glycans, which could not be separated applying continuous SP. For instance, the three structurally related glycans G0F, G1F, and G2F were baseline separated with resolution ≥1.7 using SP of 1.5 or 2 psi after the no-pressure step (Figure S9). We noticed that when we slowed down the separation and extended the separation window, the profiling depth and sensitivity in the analysis of IgG isolates were increased, most probably due to the decreased ionization suppression, decreased extent of spectral chimeracy, as well as a better match between the separation and the duty cycle and dynamic range of MS data acquisition. As expected, the analyses performed without SP during the entire time of the run led to the best resolution between the peaks of closely related glycan species. Thus, the four glycans G0F, G1F, G2F, and G2FS1 were baseline separated at mean resolution ≥3.5 (while maintaining the separation efficiency level at ~2.5 × 10⁵ theoretical plates/m), and several additional isobaric forms could be detected for G0F and G1F, making possible the separation of positional isomers for G1F even if they were not baseline separated, as shown in several reported studies^41,43 (Figure 4). Although CZE in these experiments without SP expectedly led to a 10-fold decrease in peak intensities compared to the analyses performed with SP of 5 psi, such conditions appear promising for the thorough characterization of glycans derived from biological samples and biopharmaceutical proteins. Finally, we found that the right balance between the sensitivity, throughput, and separation performance was achieved with the SP of 5 psi applied 18 min after the beginning of the CZE run.

Figure 4.

Effect of experimental conditions on the resolution in CZE-MS separation of selected APTS-labeled glycans. EIEs of G2FS1, G0F, G1F, and G2F IgG glycans analyzed at the following conditions: (A) conventional mode of instrument operation (ITT at 110 °C) and supplemental pressure of 5 psi, (B, C) nitrogen gas enriched with IPA (ITT at 150 °C) and ISCID at 70 eV with (B) or without (C) supplemental pressure of 5 psi. For (A) and (B), the supplemental pressure was applied 18 min after the beginning of the CZE-MS run.

Processing of single-stage MS (MS¹) data acquired in CZE-MS experiments resulted in the identification of 210 nonredundant N-glycan compositions in the serum IgG isolates. Approximately half of the detected N-glycans were fucosylated: ~31% were monofucosylated, ~15% difucosylated, ~5% trifucosylated, and ~1% tetrafucosylated (Figure 5C). Approximately 32% of the N-glycans were shown to be mono- or disialylated with 5-N-acetyl-neuraminic acid (Neu5Ac). A few low-abundance trisialylated glycans were also identified (≤1%) (Figure 5F). As labile monosaccharides (sialic acid and fucose) may be lost during the ESI process, we correlated the migration patterns, which were strongly dependent on the hydrodynamic volume and charge of the glycans, to the structural features of the detected glycans. Different glycan compositions/structures were separated by CZE and not detected by MS as co-migrating species, as could be indicative of ESI- and in-source-induced decay. Also, our optimized CZE-MS conditions resulted in detection of a higher number of highly sialylated and highly fucosylated glycans in the serum IgG isolates in comparison to other reported glycan profiling studies of IgG.^41,43,44 These experimental results and observations support the hypothesis that glycan decomposition during the CZE-ESI process was negligible. In addition, ESI-induced fragmentation is expected to be limited compared to MALDI-MS, and the loss of labile monosaccharides should not significantly impact the results of the glycan profiling.⁵⁶

Figure 5.

Differential N-glycan profiling of blood-derived samples. (A, B) Results of Euclidean distance-based hierarchical clustering of N-glycan quantitative profiles detected in human serum IgG, human plasma total EV isolate, and total human plasma. (A) Circular heatmap showing clustered groups of all detected N-glycans. (B) Heatmap cluster showing N-glycans detected in all three sample types. Red, yellow, and light blue colors correspond to high, medium, and low relative abundances based on the N-glycan signal intensities. N-glycans that are not detected in the samples are highlighted in dark blue. (C–H) Distribution of different types of N-glycans based on the number of identifications in the above-mentioned biological samples. (C–E) Fucosylated N-glycans. (F–H) Neu5Ac-containing N-glycans.

The processing of CZE-MS² data enabled reliable and accurate characterization of 109 N-glycan structures (i.e., exact monosaccharide arrangements and interconnecting glycosidic linkages) in the IgG isolates (Table S2). The MS² spectra were dominated by Y-type glycosidic and Y/Y-type glycosidic/glycosidic fragments. Additionally, as expected for negative ESI mode, efficient cross-ring fragmentation, resulting in A- and X-type ions, was observed, making MS² spectra structurally informative. The MS² spectra that did not lead to reliable characterization of glycan structural features were examined manually to verify if the MS² spectral patterns were characteristic of glycan fragmentation. This manual verification helped confirm the MS¹-based glycan composition identification results scored above the selected threshold (see the Supporting Information).

The detailed structural characterization of monofucosylated glycans G2FS1, G0F, G1F, and G2F, migrating between 26 and 28 min, showed several fragments characteristic of the core fucosylation (see the Supporting Information and Figure S10). Two of the most abundant N-glycans detected in the IgG isolates were FA2G2S2 (Mr_th 2368.84 Da) and its bisecting form FA2BG2S2 (Mr_th 2571.92 Da), migrating between 25.3 and 25.5 min. For these glycans, the core fucosylation was indicated with the fragments Y₁ (m/z 403.05), Y₂ (m/z 504.59), Y₁/Y₁ (m/z 330.02), and Y₁/Y₂ (m/z 431.56) (Figure 6). The detection of the glycosidic fragment ${\text{B}}_1^{1-}$ (m/z 290.09) and the glycosidic/glycosidic fragments ${\text{Y}}_{\text{6}}/{\text{Y}}_6^{3-}$ (m/z 741.53 and 809.22, respectively) made possible localization of the binding sites of the sialic acid residues at the termini of the antennas. The presence of a bisecting group in FA2BG2S2 was confirmed by detection of the ions ${\text{Y}}_{\text{3}}{\text{/Y}}_{\text{3}}/{\text{Y}}_6^{2-}$ (m/z 849.21) and ${\text{Y}}_{\text{3}}{\text{/Y}}_{\text{6}}/{\text{Y}}_6^{3-}$ (m/z 741.53) derived from multiple internal fragmentation cleavages. Also, the tandem mass spectra exhibited the singly charged diagnostic ion ^0,4A₂-CO₂ at m/z 306.12 that revealed the presence of a α-2,6 Neu5Ac linkage.⁵⁷ The nonfucosylated analog A2G2S2 (Mr_th 2222.78 Da) migrated at ~25.1 min, and the bisecting form A2BG2S2 (Mr_th 2425.86 Da) was detected 0.1 min later. CZE-MS² analyses also enabled the structural characterization of tri- and tetrafucosylated glycans (Table S2 and Supporting Information).

Figure 6.

Examples of characteristic tandem mass spectra of APTS-labeled FA2G2S2 (A) and FA2BG2S2 (B) detected in human serum IgG and human plasma EV isolates. The MS²-based structural characterization was performed by selecting the [M − 2H]²⁻ molecular ion at m/z 1403.89 and the [M − 3H]³⁻ molecular ion at m/z 1003.29, respectively, as precursor ions. Fragment ions are annotated based on the Domon and Costello nomenclature. Blue square, GlcNAc; red triangle, Fuc; green circle, Man; yellow circle, Gal; purple diamond; Neu5Ac; and orange star, APTS. Symbol Z relates to cross-ring fragmentation. Only the most intense/relevant fragments are annotated in the shown spectra.

High-mannose glycans consisting of up to 10 mannose residues were also detected and structurally characterized in the IgG isolates (see the Supporting Information). In addition, a few hybrid-type glycans, composed of both lactosamine and oligomannosyl branches on the trimannosyl core, were determined using CZE-MS² (Table S2). Finally, in the IgG isolates, ~83% of the structures characterized by tandem MS were complex-type glycans, ~10% high-mannose-type glycans, and ~7% hybrid-type glycans. Selected polylactosamines could be detected as well (~4%).

The number of N-glycans identified in the serum IgG isolates in the reported here study is higher than in other IgG glycan profiling reports (mostly focused on mAb characterization), which described less than 50 different IgG N-glycans.^41,43,44 However, considering four types of human IgG molecules, sequence variations, allotypes, variability in PTMs, isomerization, Fab-arm exchange, cells of origin, age, physiological conditions, enormous dynamic range, and other variables greatly contribute to the huge diversity of structural subpopulations and proteoforms of IgG molecules in human blood.⁵⁸ We think that further investigations are needed to develop a comprehensive catalog of blood IgG glycans, and that our knowledge of how substantial the heterogeneity of IgG glycoforms still remains in its infancy.

CZE-MS ANALYSIS OF EXTRACELLULAR VESICLES ISOLATED FROM HUMAN BLOOD

To the best of our knowledge, only a handful of studies were dedicated to the analysis of EV glycomes,^24,26,59,60 and there were no reports focused on glycan profiling of EVs isolated from human plasma. Our developed and optimized CZE-MS method was applied to the analysis of N-glycans released from human plasma-derived EVs. CZE-MS analyses of injected amounts equivalent to ~340 nL of plasma allowed for detection of 447 nonredundant N-glycan compositions in these EV isolates (these identification results were based on different levels of MS¹ data examination, as explained above).Approximately 72% of the detected EV-derived N-glycans were fucosylated, and, interestingly, high degrees of fucosylation (up to six fucose residues) were observed. Approximately 25% of the EV glycans were monofucosylated, ~20% difucosylated, ~12% trifucosylated, ~8% tetrafucosylated, ~4% pentafucosy-lated, and ~3% sextafucosylated (Figure 5D).

A large number of sialylated species (up to five sialic acids) were also detected (~70%). Sialic acid residues detected in the EV-derived glycans included Neu5Ac, the most common mammalian sialic acid, and 5-N-glycolyl-neuraminic acid (Neu5Gc). Neu5Gc is not endogenously synthesized in humans but is metabolically incorporated into human tissues from dietary sources and detected at higher levels in some human cancers.^7–9 Less than 10% of the detected EV-derived N-glycans were shown to contain Neu5Gc. For the sake of comparison with the sialylated species detected in the IgG isolates, the EV glycans containing only Neu5Ac were quantified: 29% were monosialylated, 18% disialylated, 8% trisialylated, 4% tetrasialylated, and 1% pentasialylated (Figure 5G). Our results show that tetra- and pentasialylated oligosaccharides were only detected in a nonlabeled state in the same APTS-labeled samples, suggesting that these highly negatively charged species are problematic for labeling with APTS. One explanation could be the electrostatic repulsion between the negatively charged moieties of sialylated glycans and the APTS reagent and possibly an additional steric hindrance related to the large size of these glycans. The CZE-MS² experiments confirmed the significant presence of sialylated glycans in the total EV isolates. Most of the sialylated glycans were shown to contain Neu5Ac residues. Neu5Gc-containing glycans were detected as well but only in a few instances (~5%).

The most abundant glycan detected and fully structurally characterized by MS/MS in the total EV isolates was A2G2S2 (Mr_th 2222.78 Da) (see the Supporting Information and Figure S11). In contrast, in IgG, the abundance of A2G2S2 was relatively low compared to the most abundant species detected in the IgG isolates (~6-fold lower). The fucosylated form of A2G2S2 with a core-linked fucose, FA2G2S2 (Mr_th 2368.84 Da), was detected in high abundance, and the bisecting form FA2BG2S2 (Mr_th 2571.92 Da) migrated 0.1 min later. See the discussion and Figure 6 for the fragmentation patterns of these glycans in the previous section. The second most abundant glycan detected in the EV isolates, migrating at ~23.3 min, was the tri-antennary trisialylated oligosaccharide A3G3S3 (Mr_th 2879.01 Da) (see the Supporting Information for the fragmentation pattern of A3G3S3). The presence of α-2,6 sialic acid linkages was supported by the detection of the diagnostic ion ${}^{0,4}{\text{A}}_2-{\text{CO}}_2^{1-}$ at m/z 306.12 and ${}^{0,4}{\text{X}}_{\text{5}}{\text{/C}}_5^{2-}$ (m/z 1059.87) and ${}^{0,4}{\text{X}}_{\text{5}}{\text{/C}}_6^{2-}$ (m/z 1161.40) fragment ions. Several monofucosylated forms of A3G3S3 (Mr_th 3025.07 Da) migrated shortly after their unfucosylated counterpart. Remarkably, two positional isomers, one with a core fucose, FA3G3S3, and the other one with an outer arm fucose, A3F1G3S3, could be separated. FA3G3S3 exhibited a ~2% lower μ_ep compared to the other isomer. Other sialylated tri-antennary glycans were detected at lower abundance. For instance, the disialylated glycan A3G3S2 (Mr_th 2587.91 Da) was detected as a set of three positional isomers, which were baseline separated by CZE at resolution ≥2.2 (Figure S12). We also detected highly sialylated glycans containing ≥4 sialic acid residues in their nonlabeled state, such as the tetra-antennary tetrasialylated glycan A4G4S4 (Mr_th 3535.24 Da) (see the Supporting Information and Figure S13).

Importantly, changes in the plasma levels of most of the N-glycans described above have been reported as putative blood-derived cancer biomarkers.¹² For instance, an increased level of A2G2S2 has been noticed in the serum of ovarian cancer patients, and a decrease of A3G3S3 abundance has been reported in the serum of patients with liver cancer.

CZE-MS² data also confirmed the detection of highly fucosylated species, containing three to five fucose residues, and provided information regarding their position in the glycan structure (see the Supporting Information). High-mannose (6–11 mannose residues) and hybrid-type glycans were also detected in the EV isolates. Finally, CZE-MS² experiments enabled the structural characterization of 105 N-glycans in the EV isolates from plasma (Table S3). Approximately 84% were complex-type glycans, ~11% high-mannose-type glycans, and ~5% hybrid-type glycans. Several polylactosamines could be detected as well (~11%).

Comparative Profiling of Whole Plasma and IgG and EV Isolates.

To demonstrate that N-glycans detected in the EV isolates from blood plasma were highly specific to EVs and not related to nonvesicle impurities (i.e., high-abundance plasma proteins) co-isolated from plasma matrix, we conducted a differential analysis of glycans released from all three described blood-derived samples (i.e., platelet-free anticoagulated with EDTA human blood plasma, EV isolates from the same plasma, and total IgG isolates from blood serum). To the best of our knowledge, such comparison of the quantitative N-glycan profiles of plasma and both types of isolates (IgG and EVs) was conducted for the first time. As shown in Figure 5A,B, the quantitative and qualitative profiles of N-glycans detected in the three types of specimens were significantly different for approximately identical amounts of glycans injected (as estimated by the total signal intensity). For the CZE-MS-based glycan profiling of total plasma, an injected sample amount equivalent to ~14 nL of plasma (i.e., ~40 nL of blood) was used. This resulted in the identification of over 500 nonredundant N-glycans. Highly sialylated glycans (≥4 sialic acid residues) could not be detected in the IgG isolate but were readily detected in the EV isolate with a broad range of intensity levels (Table S1). These highly sialylated glycans could also be detected in the plasma samples but at significantly different abundance levels for most of them. Over 50 glycans, including highly abundant ones, were detected only in the EV isolates and were not detected in both IgG and plasma samples (Figure S14). Most of these EV-specific glycans were highly sialylated. On the other hand, glycans that were either not detected (possibly due to their low abundance and under-representation in EV isolates) or detected at very low levels in the EV isolates (e.g., G0F, G1F, and G2F) could be detected at high or medium abundance in IgG and/or plasma samples. Highly fucosylated glycans (≥5 fucose residues) were detected only in EV and plasma samples. However, these highly fucosylated glycans were detected mostly at very low abundance levels in plasma but at relatively high abundance in the EV isolate. Moreover, the sextafucosylated glycans detected in EV isolates were specific to EVs and were different from the ones detected in plasma. Over 100 glycans were only detected in plasma, most of them being fucosylated with or without one or two sialic acids.

Several structurally related glycans, including the set of glycans A2G2S2, FA2G2S2, and FA2BG2S2, or A4G4S4 and its fucosylated analog FA4G4S4, were clustered together because of similar abundance levels across the assessed sample types (Figure 5B). Other clusters reflected similar glycan structures and ratios of abundances between different samples. For example, G0F and G2F, detected at very high abundance in the IgG isolates but at very low abundance in the EV isolates, were clustered together because the ratios between abundances of these glycan species were similar in both samples.

Based on the results of EV proteomic profiling experiments reported by our and other groups,¹⁷ similarly to IgG isolates and plasma samples, the EV isolates also contained IgG. The presence of IgG in EV isolates is expected due to the possible intercellular incorporation of IgG into EVs and the “sponge effect” in EVs, when highly abundant plasma proteins, including IgG, bind to or infiltrate into EVs specifically or nonspecifically. We expect a minor extent (<5%) of co-isolation of serum or plasma glycosylated and nonglycosylated proteins in IgG and EV preparations, respectively. However, despite this expected phenomenon, qualitative and quantitative N-glycan profiles of all three types of blood-derived samples were substantially different.

CONCLUSIONS

In this study, we developed and optimized a CZE-MS method for profiling of N-glycans and applied the developed approach to characterization of N-glycans released from human blood-derived IgG and EV isolates. The combination of DEN-gas with optimized MS conditions gave rise to a sensitivity improvement that reached 100-fold for some glycan species. The optimized control of the supplemental CZE pressure significantly increased the CZE resolution and allowed us to separate closely related and isobaric glycan species. For the first time, we report the results of qualitative and quantitative differential analysis of N-glycan profiles in blood plasma, EV isolates, and serum IgG isolates. Despite possible cross-contamination of these types of blood-derived specimens, highly distinct and characteristic N-glycan profiles were elucidated. Injected amounts equivalent to ~3 nL of serum and ~340 nL of plasma (i.e., ~1 μL of blood) resulted in the identification of >200 and 400 N-glycans in IgG and EV isolates, respectively. Approximately half of them were structurally characterized by MS², among which positional as well as linkage isomers could be differentiated. Several glycans detected in the EV isolates have been previously reported as putative cancer biomarkers. The CZE-MS-based glycan profiling of total plasma, using an injected sample amount of ~14 nL, demonstrated the uniqueness of plasma glycome with the detection of >500 N-glycans, with highly diverse structures. The N-glycan profiling results were based on different levels of data examination (software-assisted and manual), and the glycan compositions and structures identified in the three types of blood-derived isolates reflect the actual differences of the examined N-glycomes. The detection of a significant number of highly sialylated and highly fucosylated glycans in the three types of analyzed human blood isolates supported the observation that the gentle conditions of the developed CZE-MS method did not result in substantial losses of labile monosaccharide residues. The reported here results demonstrate that the developed CZE-MS method is a highly sensitive approach for the qualitative and quantitative glycan characterization of complex biological and clinical specimens available in scarce amounts. Compared to the techniques previously reported for N-glycan profiling of similar complexity blood isolates at similar amounts, the developed method resulted in an unmatched depth of characterization (enhancement of ~5-fold in the number of identified N-glycans). In addition, to the best of our knowledge, MS-based N-glycan profiling of human plasma-derived EVs has not been reported yet. The developed technique showed potential for deep, sensitive, and highly informative N-glycan profiling of EVs, exosomes, and other trace-level constituents isolated from nL to low μL volumes of physiological fluids.