Analytical and Bioanalytical Chemistry

, Volume 409, Issue 2, pp 579–588

Isotope-targeted glycoproteomics (IsoTaG) analysis of sialylated N- and O-glycopeptides on an Orbitrap Fusion Tribrid using azido and alkynyl sugars

  • Christina M. Woo
  • Alejandra Felix
  • Lichao Zhang
  • Joshua E. Elias
  • Carolyn R. Bertozzi
Research Paper

DOI: 10.1007/s00216-016-9934-9

Cite this article as:
Woo, C.M., Felix, A., Zhang, L. et al. Anal Bioanal Chem (2017) 409: 579. doi:10.1007/s00216-016-9934-9
Part of the following topical collections:
  1. Glycomics, Glycoproteomics and Allied Topics


Protein glycosylation is a post-translational modification (PTM) responsible for many aspects of proteomic diversity and biological regulation. Assignment of intact glycan structures to specific protein attachment sites is a critical step towards elucidating the function encoded in the glycome. Previously, we developed isotope-targeted glycoproteomics (IsoTaG) as a mass-independent mass spectrometry method to characterize azide-labeled intact glycopeptides from complex proteomes. Here, we extend the IsoTaG approach with the use of alkynyl sugars as metabolic labels and employ new probes in analysis of the sialylated glycoproteome from PC-3 cells. Using an Orbitrap Fusion Tribrid mass spectrometer, we identified 699 intact glycopeptides from 192 glycoproteins. These intact glycopeptides represent a total of eight sialylated glycan structures across 126 N- and 576 O-glycopeptides. IsoTaG is therefore an effective platform for identification of intact glycopeptides labeled by alkynyl or azido sugars and will facilitate further studies of the glycoproteome.


Glycoproteomics Chemical proteomics LC-MS/MS Metabolic labeling Sialic acid 


Glycosylation is a heterogeneous post-translational modification (PTM) that decorates proteins from the extracellular matrix to the intracellular nucleus in mammalian cells. The glycoproteome influences diverse biological processes, such as immunological regulation [1] or cancer progression [2, 3], via modulation of innate biophysical properties [4]. Recent attention has been focused on sialylated glycans due to their prevalence on many cancers and their proposed roles in metastasis [5, 6] and immune evasion [7, 8]. Likewise, fucosylated glycans have been associated with cancer and may play similar roles [9, 10]. Thus, these glycan classes are important to understand in molecular details as they may hold keys to new therapeutic and diagnostic strategies.

Until recently, cancer glycosylation was largely characterized at the level of global cell surface abundance using lectins, antibodies, or related detection reagents [11, 12]. A more detailed structural analysis of cancer glycoproteomes has been challenging due to the complex and heterogeneous nature of glycosylation [13, 14]. Nonetheless, progress has been accelerating. Structural profiling of glycoproteins from complex mixtures has been tackled using various enrichment strategies and mass spectrometry (MS) to obtain information about glycan attachment site and glycan heterogeneity. Glycoprotein enrichment by chemical oxidation followed by hydrazide capture [15], lectin affinity [16], chemical labeling [17], and metabolic labeling [18] approaches have been reported, among other techniques [14, 19, 20]. In most studies, glycans were separated from their peptide scaffolds for direct observation of the glycosite. This can be accomplished by treatment with PNGase F or Endo H for N-glycosite characterization [21, 22], oxidative proteolysis [23], beta elimination for O-glycans [24], or acid hydrolysis of terminal sialic acid [25]. As well, a genetic engineering approach to identify O-glycosites has been developed [26, 27].

Several MS methods show promise for intact glycopeptide analysis [28, 29], including correlation of the released glycans with deglycosylated peptides to assign the intact glycopeptide [30]. However, these methods require that glycans and their peptide substrates are characterized in separate analyses [31, 32, 33]. Consequently, when these approaches are applied to complex protein mixtures, glycan–protein associations can only be made indirectly and are therefore less accurate than what might achieved by direct observation. But current methods for direct observation, while increasing confidence of glycan–protein associations, are burdened with challenging assignments due to potential insufficient tandem MS coverage [34, 35]. A directed method to characterize intact glycopeptides is necessary as the assignment of heterogeneous glycoproteins within a complex mixture is computationally challenging.

An orthogonal approach to enrich and confidently assign intact glycopeptides involves use of enrichment with cleavable probes together with mass-independent MS. Recent work from our lab has applied this approach to a platform termed isotope-targeted glycoproteomics (IsoTaG) to tag and characterize a range of glycopeptides, including those bearing sialylated glycans [36]. The process begins with metabolic labeling of cultured cells with azide-functionalized monosaccharide substrates. For example, peracetylated N-azidoacetylmannosamine (Ac4ManNAz) is metabolized to the corresponding azido sialic acid (SiaNAz) and then labels sialylated glycans. Likewise, peracetylated N-azidoacetylgalactosamine (Ac4GalNAz) metabolically labels glycans bearing GalNAc or GlcNAc, which includes numerous mammalian glycan structures. After metabolic labeling, the glycoproteins are affinity enriched with cleavable alkynyl biotin probe 1, which we attach to the target substrates by copper-catalyzed azide–alkyne cycloaddition (CuAAC, also called “click chemistry,” Fig. 1a). Simultaneously, glycopeptides are isotopically recoded by two bromine atoms embedded within the probe for targeted assignment of the glycopeptide by MS. Isotopic recoding enables mass-independent MS where glycopeptides are immediately recognized by full-scan MS to guide subsequent targeted assignment.
Fig. 1

Chemical probe design and application in IsoTaG. a Complementary probe and glycan pairs. Silane probe 1 reacts selectively with azidosugar (e.g., Ac4ManNAz) glycoproteins. Silane probe 2 reacts selectively with alkynylsugar (e.g., Ac4ManNAl) glycoproteins. b Application of IsoTaG to sialylated glycans expressed on the cell surface and secreted

The IsoTaG approach relies on the efficiency of metabolic labeling by azido sugars to install a handle for tagging with probe 1. Metabolic labeling enables selection of glycoproteins as they are biosynthesized and has been performed in cell culture [18], live animals [37, 38], and human tissues [39]. Labeling efficiency, however, is highly dependent on the activity of the biosynthetic enzymes. In some cases, metabolic labeling with alkyne-bearing sugar analogs is advantageous. For example, metabolic labeling with peracetylated N-(4-pentynoyl)mannosamine (Ac4ManNAl) was found to proceed at higher efficiency than with Ac4ManNAz in mice [38]. Alkynyl analogs of fucose [e.g., peracetylated 6-alkynyl fucose (Ac4FucAl)] are incorporated into mammalian glycans with less toxicity as compared to azido analogs [40]. Furthermore, alkynyl GlcNAc analogs for specific labeling of O-GlcNAc vs other GlcNAc-containing glycans have been developed [41]. To exploit these alkynyl sugars in the context of IsoTaG will require the development of new probes and methods.

Here, we describe the design of an IsoTaG-compatible azide probe (compound 2) that enables analysis of intact alkyne-labeled glycopeptides. Glycoproteins labeled with alkynylsugars in the secretome of PC-3 cells were structurally characterized by IsoTaG on an Orbitrap Fusion Tribrid, alongside azide-labeled glycoproteins for comparison. Thirty-seven alkyne-labeled and 695 azide-labeled sialoglycopeptides, respectively, were found from the secretome of PC-3 cells. In sum, 699 intact glycopeptides whose structures included eight discrete sialylated glycans were identified from 192 glycoproteins across 126 N- and 576 O-glycopeptides. These results establish IsoTaG as an effective methodology to characterize intact metabolically labeled glycopeptides across N- and O-glycans.

Materials and methods

Chemical materials

Commercial solvents and reagents were used as received with the following exceptions. Dichloromethane was purified according to the method of Pangborn and co-workers [42]. Triethylamine was distilled from calcium hydride under an atmosphere of nitrogen immediately before use. RapiGest was prepared according to the method of Lee and co-workers [43]. 3-[4-({Bis[(1-tert-butyl-1H-1,2,3-triazol-4-yl)methyl]amino}methyl)-1H-1,2,3-triazol-1-yl]propanol (BTTP) was prepared according to the method of Wu and co-workers [44]. Tetraacetylated N-(4-pentynoyl)mannosamine (Ac4ManNAl) was prepared according to the method of Wong and co-workers [40]. Peracetylated 6-alkynyl fucose (Ac4FucAl) was obtained from Thermo Fisher. Tetraacetylated N-azidoacetyl mannosamine (Ac4ManNAz) was prepared according to the method of Bertozzi and co-workers [45]. EDTA-free protease inhibitor cocktail was obtained from Roche Diagnostics (Version 11). Streptavidin–agarose beads were obtained from Thermo Scientific and washed with PBS prior to use. The silane probe 1 was prepared according to the procedure of Bertozzi and co-workers [36]. Bovine serum albumin (BSA) was obtained from Sigma. Biotin-PEG3-azide was obtained from Sigma.

Synthetic procedures

The silane probe 2 was prepared as shown in Electronic Supplementary Material (ESM) Figure S1. Please see ESM for full synthetic procedures.

Cell culture and metabolic labeling

PC-3 cells were obtained from the American Type Culture Collection (ATCC) and maintained at 37 °C and 5 % CO2 in a water-saturated incubator. PC-3 cells were metabolically labeled between passages 17–22. Cell densities were counted using a hemacytometer and seeded at 2 × 105 cells/mL at the start of metabolic labeling experiments. PC-3 cells were maintained in RPMI-1640 supplemented with 10 % FBS and 1 % penicillin/streptomycin.

Azido and alkynyl sugars (Ac4ManNAz, Ac4FucAl, or Ac4ManNAl) were prepared as 500 mM stock solutions in dimethylsulfoxide (DMSO). Tissue culture dishes (150 mm) were seeded with 100 μM of Ac4ManNAl, Ac4ManNAz, Ac4FucAl, or vehicle control containing DMSO (3.0 μL). Six dishes per condition were prepared. A suspension of cells at a density of 2 × 105 cells/mL in complete media (RPMI supplemented with 10 % FBS and 1 % penicillin/streptomycin) was added to the dish (15 mL per dish) and the dishes were incubated for 48 h at 37 °C. The media were aspirated and the adherent cells were washed with PBS (1 × 10 mL). Washed dishes were resuspended in RPMI containing 100 μM glycan metabolite without FBS additive (15 mL), and the cells were incubated an additional 48 h at 37 °C in a humidified 5 % CO2 incubator.

The conditioned media (100 mL) was harvested and cleared by centrifugation (150×g, 3 min). Clarified media was spin concentrated (Amicon, 15 mL 10 kDa spin filter) to 1 mL. The concentrated residue was washed with 1 % Triton X-100/PBS (3 × 15 mL) and transferred to an Eppendorf microcentrifuge tube as the conditioned media fraction. The conditioned media fraction was adjusted to a final concentration of 1 % RapiGest/PBS with a 10 % RapiGest/PBS stock solution. Protein concentrations from the conditioned media fractions were measured by bicinchonic acid assay (Pierce).

Chemical enrichment

Conditioned media from PC-3 cells treated with Ac4ManNAl, Ac4ManNAz, Ac4FucAl, or DMSO vehicle were aliquoted to 2.0 mg fractions (667 μL). To Ac4ManNAz-labeled conditioned media, click chemistry reagents (40.0 μL, 200 μM 1, 300 μM CuSO4, 100 mM BTTP, 2.50 mM sodium ascorbate, mixed immediately before addition to lysates) were added and the reaction was incubated for 3 h at 24 °C. To Ac4ManNAl-labeled, Ac4FucAl-labeled, or DMSO vehicle conditioned media, click chemistry reagents (40.0 μL, 200 μM 2, 300 μM CuSO4, 100 mM BTTP, 2.50 mM sodium ascorbate, mixed immediately before addition to lysates) were added and the reaction was incubated for 3 h at 24 °C. Methanol (1 mL) was added to quench the reaction, and proteins were precipitated for 1 h at −80 °C. Precipitated proteins were pelleted by centrifugation (16,100 × g, 10 min, 4 °C) and the supernatant was discarded. Pelleted proteins were air-dried for 10 min at 24 °C. Dried protein pellets were resuspended in 400 μL of 1 % RapiGest/PBS and solubilized by probe sonication (Misonix, 1.0 min, 4 °C). Streptavidin–agarose resin [200 μL, washed with PBS (3 × 1 mL)] was added, and the resulting mixture was incubated for 12 h at 24 °C with rotation. The beads were pelleted by centrifugation (3000 × g, 3 min) and the supernatant containing uncaptured proteins was separated. The beads were washed with 1 % RapiGest/PBS (1 mL), 6 M urea (2 × 1 mL), and PBS (5 × 1 mL), and the beads were pelleted by centrifugation (3000 × g, 3 min) between washes.

Washed beads were resuspended in 5 mM DTT/PBS (200 μL) and incubated for 30 min at 24 °C with rotation. Ten millimolars of iodoacetamide (4.0 μL, 500 mM stock solution in PBS) was added to the reduced proteins, and the resultant solution allowed to react for 30 min at 24 °C with rotation, in the dark. Beads were pelleted by centrifugation (3000 × g, 3 min) and resuspended in 0.5 M urea/PBS (200 μL). Trypsin (1.5 μg) was added to the resuspended beads, and digestion proceeded for 12 h at 37 °C. Beads were pelleted by centrifugation (3000 × g, 3 min), and the supernatant digest was collected. The beads were washed with PBS (1 × 200 μL) and H2O (2 × 200 μL). Washes were combined with the supernatant digest to form the trypsin digest of nonconjugated peptides. Probe 1 or 2 was cleaved with two treatments of 2 % formic acid/H2O (200 μL) for 30 min at 24 °C with rotation and the eluent was collected. The beads were washed with 50 % acetonitrile–water + 1 % formic acid (2 × 400 μL), and the washes were combined with the eluent to form the cleavage fraction. The trypsin digest and cleavage fraction were concentrated using a vacuum centrifuge (i.e., a speedvac, 40 °C) to 50 μL. Samples were desalted with a ZipTip P10 and stored at −20 °C until analysis.

Western blotting

Aliquots collected during the enrichment procedure (10.0 μL) were reduced and separated by standard SDS-PAGE (Bio-Rad, Criterion system), electroblotted onto nitrocellulose, blocked in 5 % BSA in Tris-buffered saline with Tween (10 mM Tris pH 8.0, 150 mM NaCl, 0.1 % Tween-20), and analyzed by standard enhanced chemiluminescence immunoblotting methods (Pierce). The staining agent was streptavidin–HRP (Pierce, 1:100,000).

Liquid chromatography–tandem mass spectrometry (LC-MS/MS)

Peptides were analyzed by online capillary nanoLC-MS/MS. Samples were separated on an in-house-made 20-cm reversed phase column [100 μm inner diameter, packed with ReproSil-Pur C18-AQ 3.0 μm resin (Dr. Maisch, GmbH)] equipped with a laser-pulled nanoelectrospray emitter tip. Peptides were eluted at a flow rate of 400 nL/min using a two-step linear gradient including 3–25 % buffer B in 70 min and 25–40 % B in 20 min (buffer A: 0.2 % formic acid in water; buffer B: 0.2 % formic acid in acetonitrile) in a Dionex UltiMate 3000 HPLC system (Thermo Fisher Scientific). Peptides were then analyzed using an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). Data acquisition was executed in data dependent mode (top speed with duty cycle of 3 s) with full MS scans acquired in the Orbitrap mass analyzer with a resolution of 120,000 and m/z scan range of 400–1500. Precursor ions with charge state 2–7 and intensity threshold above 5000 counts were selected for fragmentation using higher-energy collisional dissociation (HCD) with quadrupole isolation, isolation window of 4 m/z and collision energy of 30 %. The HCD fragments were analyzed in the Orbitrap mass analyzer with a resolution of 15,000. Dynamic exclusion was enabled with a repeat count of 1 and exclusion duration of 30 s. The automatic gain control (AGC) target was set to 400,000 and 50,000 for full FTMS scans and FTMS2 scans, respectively. The maximum injection time was set to 50 and 250 s for full FTMS scans and FTMS2 scans. Data from both scans were acquired in profile mode. For additional HCD-triggered MS2 scans, targeted loss trigger and targeted product trigger were performed in separate runs. For targeted loss trigger, the masses of both neutral loss species and their corresponding M + H forms were added to the inclusion mass list. For product ion trigger, the m/z values of the M + H products were added to the inclusion mass list. The inclusion mass list was scanned for throughout data collection. Electron transfer dissociation (ETD) reactions were performed for triggered MS2 following previous HCD scans. Triggered precursor ions with charge state 3–7 were subject to ETD with quadrupole isolation, isolation window of 4 m/z, reaction time of 40 ms, reagent target of 200,000, and maximum reagent injection time of 200 ms. ETD supplemental activation (EThcD) was enabled with collision energy of 25 %. The detection of ETD fragment ions was performed with the same parameters as HCD scans.

Data analysis

The raw data were processed using Proteome Discoverer 1.4 software (Thermo Fisher Scientific) and searched against the human-specific SwissProt database downloaded on April 29, 2016. Indexed databases for tryptic digests were created allowing for up to three missed cleavages, one fixed modification (carboxyamidomethylcysteine, +57.021 Da), and variable modifications (methionine oxidation, +15.995 Da; and others as described below). Precursor ion mass tolerance for spectra acquired using the Orbitrap were set to 10 ppm. The fragment ion mass tolerance was set to 20 ppm. The SEQUEST HT search engine was used to assign nonconjugated peptides from on-bead tryptic digests. Normalized protein assignments at a 1 % false discovery rate (FDR) were considered enriched if the fold change was greater than two and the associated p value was ≤0.05 (t test) in labeled samples than in the DMSO control (ESM Figure S2).

Samples with intact glycopeptides were searched with the Byonic search algorithm v2.8.2 as a node in Proteome Discoverer 1.4. Byonic is a software package that allows definition of the number of occurrences for each modification, which prevents a combinatorial explosion when searching for multiple glycan structures simultaneously [46]. The Byonic score is defined by the absolute quality of the peptide–spectrum match over a range of 0 to 1000, where 300 is a good score and 400 a very good score. Example glycan structure input files are provided in the ESM for Ac4ManNAl, Ac4ManNAz, and Ac4FucAl glycans, respectively. Following a search against the SwissProt human proteome, a search against the pool of enriched glycoproteins was performed. Glycopeptide assignments were aggregated and a minimum of two separate assignments at 5 % FDR to the same underlying peptide was used as a cutoff for further analysis. The DMSO control sample was searched separately against each of the glycan structure input files and no intact glycopeptide assignments were reported. Assignments of all spectra in Ac4FucAl-labeled samples were validated by manual inspection for the precursor isotope pattern and expected glycan fragments.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [47] partner repository with the dataset identifier PXD004302.

Results and discussion

The azide biotin probe 2 was designed to closely mimic the IsoTaG-compatible alkyne probe 1 (Fig. 1a). Preparation of probe 2 was achieved in three steps (ESM Figure S1). Probe 2 preserves the dibromide motif for isotope recoding with a triplet signature (1:2:1 over M, M + 2, M + 4) and the cleavable silane linker [48], which is cleaved in acidic conditions (2 % formic acid) tolerated by glycoconjugates. The dibromide motif is detectable computationally using a pattern-recognition algorithm developed in house, termed IsoSTAMP [49].

To evaluate the performance of the azide biotin probe 2, PC-3 cells were metabolically labeled with 100 μM Ac4ManNAl for 2 days in FBS-free media. Ac4ManNAl is intracellularly deacetylated, converted to the corresponding alkynyl sialic acid (SiaNAl), and installed on sialylated glycoproteins [38, 40]. The conditioned media were collected and reacted with azide probe 2 by CuAAC. For comparison, PC-3 cells were labeled with 100 μM Ac4ManNAz to produce the corresponding azido sialic acid (SiaNAz), and the conditioned media were reacted with alkyne biotin probe 1 [45, 50]. Western blot analysis showed similar tagging and enrichment of conditioned media from cells treated with Ac4ManNAl or Ac4ManNAz, with little off-target tagging from cells treated with the DMSO vehicle (Fig. 2a and ESM Figure S3).
Fig. 2

Global analysis of glycoproteins and glycopeptides identified by IsoTaG analysis of conditioned media from metabolically labeled PC-3 cells. a Anti-biotin Western blot analysis of tagging and enrichment efficiency with DMSO control, Ac4ManNAl, and Ac4ManNAz samples. Total protein levels shown alongside (ponceau). Lanes: (1) protein input following CuAAC reaction with probe 1 or 2, (2) protein supernatant following overnight incubation with streptavidin–agarose beads, (3) aliquot of streptavidin–agarose beads (10 μL) after enrichment. b Venn diagram of glycoproteins identified from nonconjugated on-bead tryptic peptides from Ac4ManNAz and Ac4ManNAl samples. c Subcellular localization of proteins identified by direct intact glycopeptide assignments

Tagged glycoproteins were affinity enriched with streptavidin–agarose beads and glycoproteins were proteolyzed on-bead with trypsin. Nonconjugated peptides from trypsin proteolysis were assigned by SEQUEST HT and analyzed for enrichment (Fig. 2b). High-confidence peptide assignments (1 % FDR) were aggregated into protein groups. Proteins were considered enriched if a fold change of greater than two and associated p value ≤0.05 was found in metabolically labeled samples as compared to the DMSO control. A pool of 1015 proteins were enriched from a total 1539 identified proteins (66 %, ESM Figure S2 and Table S1). Of these glycoproteins, 198 were found in both metabolically labeled samples representing a proteomic overlap of greater than 50 % between Ac4ManNAz and Ac4ManNAl samples, with respect to the Ac4ManNAl sample. The greatest number of enriched proteins was identified from Ac4ManNAz conditioned media (835 proteins). Differences in proteomic overlap may reflect the ability of azido and alkynyl sugars to label distinct glycoprotein subsets, as well as the relative permissiveness of PC-3 cells for Ac4ManNAz.

Following the release of nonconjugated peptides for glycoprotein identification, beads were treated with 2 % formic acid to cleave the silane linker and recover intact glycopeptides for mass-independent MS analysis. The released glycopeptides were analyzed by reversed-phase nanoflow liquid chromatography coupled to an ETD-enabled Orbitrap Fusion Tribrid mass spectrometer. The Fusion Tribrid enables acquisition of HCD and ETD with high mass accuracy to improve fragment assignment by database searching. Acquisition strategies such as HCD product-dependent ETD (HCD-pd-ETD), where ETD is triggered in real-time from observed glycan oxonium ions in HCD spectra, have been shown to optimize MS time for glycopeptide analysis [29]. Thus, an HCD-pd-ETD method with supplemental activation was employed for glycopeptide analysis. Two technical replicates were obtained with ETD triggered from the neutral/charged loss ion or the product ion, respectively.

Dibromide incorporation into the probes 1 and 2 creates an isotopically recoded mass envelope (Fig. 1b). We use this isotopic signature as an orthogonal, mass-independent handle for confident glycopeptide identification. The isotopic signature thus introduces a metric to evaluate the efficiency of the product-dependent ETD trigger and validate glycopeptide false positives and false negatives during database assignment. The isotopic pattern additionally allows rapid assessment of the existence of intact glycopeptides. We applied the pattern searching algorithm IsoStamp to the raw data and found that without any further analysis, Ac4ManNAz conditioned media possessed approximately 40 times more species carrying the isotopic pattern as compared to the full-scan MS from Ac4ManNAl conditioned media. Spectra were searched against the Swiss-Prot human proteome and the database of enriched glycoproteins described above using Byonic v2.8 [46]. Without any further manipulation, up to 928 spectra were assigned as intact glycopeptides in a single MS analysis. A product ion or neutral/charged loss trigger for ETD were equally advantageous for glycopeptide assignment. Glycopeptide assignments with two or more spectral counts at 5 % FDR were considered for further analysis. Manual validation for the dibromide precursor of 100 of these glycopeptide assignments in the Ac4ManNAz sample found a 98 % true positive rate for the isotopic pattern. Spectra that were assigned as a glycopeptide by database searching, but did not possess an isotopically recoded precursor were thus excluded. As an indication of the confidence that site level analysis enables in glycoproteome studies, no intact glycopeptides were assigned to the DMSO control.

A total of 699 intact glycopeptides were assigned from 192 glycoproteins in PC-3 conditioned media. Intact glycopeptides represented 126 N-glycans and 576 O-glycans. Subcellular localization of the proteins from which they derived was predominantly at the plasma membrane (35 %) and secretome (42 %), as expected for the analysis of conditioned media (Fig. 2c). Proteins were additionally annotated to the endoplasmic reticulum/Golgi apparatus (11 %) and other (12 %, e.g., lysosome, vesicles, cytoplasmic) locations within the cell. The molecular function annotation of these glycoproteins includes binding to ions (e.g., calcium, metal), protein binding interactions, and glycan binding (Table 1). The identified glycosites derive from approximately 20 % of the glycoproteins observed by nonconjugated peptide analysis (ESM Table S2).
Table 1

Top 15 molecular function annotations for identified glycoproteins

Molecular function

Protein count

Calcium ion binding


Heparin binding


Metal ion binding


Identical protein binding


Structural molecule activity


Protein homodimerization activity


Integrin binding


Poly (A) RNA binding


Zinc ion binding


Cytokine activity


Enzyme binding


Protease binding


Serine-type endopeptidase inhibitor activity


ATP binding


DNA binding


Many unassigned spectra were found to derive from isotopically recoded precursors, an indicator that additional assignments may be made on closer analysis. Spectra from the highest yielding Ac4ManNAz analysis revealed that of 2692 MS2 spectra derived from patterned species, 1542 spectral assignments were made. Spectral unassignment may be caused by low spectrum quality, modification types that were not included in the database search (e.g., acetylation, formylation, alternate glycan structures, amino acid variants) [36], and the remaining challenges in automated assignment of glycan, peptide, and particularly glycopeptide fragments. Further analysis using the isotopically recoded precursor may guide assignment of these spectra.

The observed sialylated glycopeptides contained eight discrete sialylated glycan structures (Fig. 3a). During database searching, N- and O-glycan structures were searched simultaneously. Sialylated glycans included three O-glycans (S1–S3), moderately complex N-glycans (S4, S6, S8), and their fucosylated counterparts (S5, S7). In particular, the sialyl Tn O-glycan epitope (S1) and core 1 sialylated glycans (S2, S3) were assigned at a greater frequency than any of the N-glycans. The higher assignment rate of O-glycans may be related to their relative abundance or the relative depth of peptide fragmentation observed with O- vs N-glycopeptides. Glycopeptide assignments were two orders of magnitude higher for conditioned media from azido sugar-labeled PC-3 compared to alkynyl sugar-labeled PC-3 cells (Table S2). This discrepancy is supported by the greater number of glycoproteins enriched from azido sugar-labeled media (ESM Table S1). Due to incorporation differences across metabolites, any quantitative application of IsoTaG would require comparison only when the same metabolite and cell line was used.
Fig. 3

Intact sialylated glycan structures and frequency of occurrence. a Intact sialylated glycans identified (S1–S8). S1–S3 are O-glycans. S4–S8 are N-glycans. b. Frequency of sialyl glycans observed by spectral counting

Fucosylated glycopeptide assignments were additionally sought by treatment of PC-3 cells with 100 μM Ac4FucAl [40]. Enrichment and MS analysis of Ac4FucAl-labeled conditioned media led to the identification of 145 intact glycopeptides with a minimum of two spectral counts and no corresponding assignments found in the DMSO control. However, the assignments corresponding to FucAl-labeled glycopeptides carried low scores (max score = 82, Byonic), and the absence of isotopic recoding due to the dibromide in the full-scan MS precursor confirmed these assignments as false positives. Independent analysis by our computational algorithm of the MS1 dataset revealed no isotopically recoded precursors from FucAl-labeled glycopeptides. The incorrect assignment of fucosylated glycopeptides may derive from the fact that fucose, unlike sialic acid, does not generate oxonium ions during tandem MS and therefore must be inferred indirectly through neutral or charged losses. The low abundance of FucAl-labeled glycopeptides may reflect the low permissivity of the fucose salvage pathway enzymes in PC-3 cells, low rates of fucosylation due to enzymatic suppression [10], or steric inaccessibility of FucAl-labeled glycans to the chemical probe. We investigated the potential for steric inhibition by comparison of tagging efficiency by probe 2 and a biotin-PEG3-azide. By Western blot analysis, we found no anti-biotin signal from conditioned media of Ac4FucAl-labeled PC-3 cells using either probe (ESM Figure S4).

Over 80 glycopeptides were identified that exist as multiple glycoforms, a selection of which are highlighted in Table 2. Up to four unique glycoforms were assigned to a single glycopeptide (entries 9, 15). Several glycoproteins displayed multiple glycosites with high heterogeneity (entries 9, 12, 15). In some cases, both N- and O-glycosylations were found on the same glycopeptide in separate instances (entries 2, 9). The full GO molecular function annotation and subcellular localization are presented in ESM Tables S3 and S4. Thus, IsoTaG can identify both N- and O-glycans in an unbiased manner by mass-independent MS in a single analysis.
Table 2

Selected examples of intact glycopeptides with more than three glycoforms identified


Row labels


Protein name (accession)



S1, S2, S3

Latent-transforming growth factor beta-binding protein 1 (Q14766)



S2, S5, S7

Dickkopf-related protein 1 (O94907)



S1, S2, S3

Dystroglycan (Q14118)



S4, S5, S6, S7

Procollagen-lysine,2-oxoglutarate 5-dioxygenase 1 (Q02809)



S1, S2, S3

Nucleobindin-1 (Q02818)



S1, S2, S3

Prostate-associated microseminoprotein (Q1L6U9)



S2, S5, S7

Clusterin (P10909)



S1, S2, S3

Dickkopf-related protein 1 (O94907)



S3, S4, S5, S7

Metalloproteinase inhibitor 1 (P01033)



S5, S6, S7

Metalloproteinase inhibitor 1 (P01033)



S1, S2, S3

Latent-transforming growth factor beta-binding protein 1 (Q14766)



S1, S2, S3

Latent-transforming growth factor beta-binding protein 1 (Q14766)



S1, S2, S3

Latent-transforming growth factor beta-binding protein 1 (Q14766)



S6, S7, S8

Collagen alpha-2 (VI) chain (P12110)



S4, S5, S6, S8

Procollagen-lysine,2-oxoglutarate 5-dioxygenase 1 (Q02809)



S1, S2, S3

Cathepsin D (P07339)

Glycosite(s) is indicated by lower case letter. IsoTaG enriched glycopeptides carrying SiaNAz were analyzed on an Orbitrap Fusion Tribrid and assigned by database searching (Byonic)

An intact glycoproteomics method enables correlation of the glycan structure to the glycosite. We compared the spectral assignment of YSQAVPAVTEGPIPEVLK carrying glycan S2 from cathepsin D derived from SiaNAl or SiaNAz-labeled PC-3 cells (Fig. 4a, b, respectively). Both spectra carry isotopically recoded precursors and product ions from the glycan. Closer assessment of the spectra revealed differential localization of the sialic acid residue between the terminal or reducing position of the glycan. Measurement of the relative abundance of oxonium ions revealed differential ratios. The 2,6-sialic acid linkage produced oxonium ions in the ratio of 1:0.3:1 for HexNAc, HexNAcHex, and dibromide-tagged sialic acid (Fig. 4a, ESM Table S5), while the corresponding 2,3-sialic acid linkage produced oxonium ions in the ratio of 1:1.2:1.4, respectively (Fig. 4b, ESM Table S6). These ratios may reflect specific glycan structures more broadly [28]. IsoTaG delivers the potential to evaluate these intact glycopeptide fragmentation ratios, which may eventually reveal the heterogeneity inherent to glycosylation both within the glycan structure and the peptide modification site.
Fig. 4

Example spectral assignment of the glycopeptide YSQAVPAVTEGPIPEVLK from cathepsin D carrying SiaNAl (a) or SiaNAz (b) incorporation to glycan S2. Assignment of glycan and peptide fragments enables localization of sialic acid in the glycan structure. The relative abundance of highlighted peaks is reported in ESM Tables S1 and S2


IsoTaG enables the study of intact glycopeptides from azido and alkynyl sugar-labeled samples with high confidence. To evaluate samples labeled by alkynyl sugars (e.g., Ac4ManNAl) by IsoTaG, we developed azide biotin probe 2 for chemical enrichment and isotopic recording. Isotopic recoding imparts an immediate and orthogonal validation of glycopeptide assignments from complex mixtures by full-scan MS. Integration of IsoTaG with an Orbitrap Fusion Tribrid led to the assignment of 699 intact glycopeptides from 192 glycoproteins from the PC-3 cell line. These glycopeptides represent both N- and O-glycans (126, 578, respectively) that are modified with eight sialylated glycan structures. These results expand applications of IsoTaG to alkyne-functionalized metabolic labels and will speed the characterization of intact glycoproteins for biomarker discovery and functional studies.


Financial support from the US National Institutes of Health (CA200423, C.R.B.), Jane Coffin Childs Memorial Fund (C.M.W.), Burroughs Wellcome Fund Career Awards at the Scientific Interface (C.M.W.), Stanford Undergraduate Advising and Research Student Grant (A.F.), the W.M. Keck Foundation Medical Research Program (J.E.E.), and the Bill and Melinda Gates Foundation (J.E.E.) are gratefully acknowledged.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

216_2016_9934_MOESM1_ESM.pdf (3.3 mb)
ESM 1(PDF 3393 kb)
216_2016_9934_MOESM2_ESM.xlsx (571 kb)
ESM 2(XLSX 571 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2016

Authors and Affiliations

  • Christina M. Woo
    • 1
  • Alejandra Felix
    • 1
  • Lichao Zhang
    • 2
  • Joshua E. Elias
    • 2
  • Carolyn R. Bertozzi
    • 1
    • 3
  1. 1.Department of ChemistryStanford UniversityStanfordUSA
  2. 2.Chemical and Systems BiologyStanford UniversityStanfordUSA
  3. 3.Howard Hughes Medical InstituteStanford UniversityStanfordUSA

Personalised recommendations