Methods

Materials. A uniform mixture of six standard glycoproteins was used as the starting material, including IgG (56834, catalog numeber), IgA (I4036), IgM (I8260), Alpha-1-acid glycoprotein (G9885), Alpha-2-macroglobulin (M6159) and Haptoglobin (H3536) (Sigma). Trypsinization and HILIC enrichment were performed as previously reported22.

LC-MS/MS conditions. The standard glycoprotein mixture was analyzed by nanospray LC-MS/MS on an Orbitrap Fusion Tribrid (Thermo Scientific) coupled to an EASY-nano-LC System (Thermo Scientific) without the trap column. For one LC-MS run, 10 ug mixture was used (before HILIC enrichment). In our experience, for glycoprotein mixtures, about 1 ~ 5% sample remained after HILIC enrichment. Therefore, the loading was about 0.1 ~ 0.5 ug. The mixture was loaded onto a C18 spray tip 15 cm × 75 μ m i.d. column and was separated at a flow rate of 350 nL/min by using a gradient of 3% to 22% solvent B (100% acetonitrile with 0.1% formic acid) in 42 min, followed by an increase to 30% B in 6 min, and then to 90% B in 6 min and held for another 6 min. Solvent A was 0.1% formic acid in water.

Two separate LC-MS/MS runs were performed: the first one collected HCD-MS/MS and CID-MS/MS spectra, and the other one collected HCD-MS/MS and MS3 spectra.

HCD-pd-CID-MS/MS. The parameters used for MS data acquisition of HCD-MS/MS and CID-MS/MS spectra were: (1) MS: top speed mode, cycle time = 3 sec; scan range (m/z) = 400–2,000; resolution = 60,000;AGC target = 200,000; maximum injection time = 5 ms; MS1 precursor selection range = 700–2,000; included charge state = 2–6; dynamic exclusion after n times, n = 1; dynamic exclusion duration = 15 sec; precursor priority = most intense; (2) HCD-MS/MS: isolation mode = quadrupole; isolation window = 2; collision energy = 40%; detector type = orbitrap; resolution = 15,000; AGC target = 100,000; maximum injection time = 35 ms; microscan = 1; product ion table: at least n product ions detected, n = 1; product ion threshold = 30%; product ion table = 138.055 Da (the triggering ion of HCD-pd-CID is the 138.055 Da ion with at least 30% relative intensity);(3) CID-MS/MS: isolation mode = ion trap; isolation window = 2; collision energy = 35%; detector type = orbitrap; resolution = 15,000; AGC target = 100,000; maximum injection time = 35 ms; microscan = 1.

HCD-pd-MS3. HCD-pd-MS3 was performed in another MS run. The parameters used for data acquisition of MS and HCD-MS/MS were the same as the parameters described above. And the parameters used for MS3 acquisition were: exclusion mass list = 100–700; top N = 3 (three most intense ions in the HCD-MS/MS spectrum were subjected to MS3 analysis respectively); isolation mode = ion trap; isolation window = 2; collision energy = 35%;detector type = ion trap; AGC target = 500,000; maximum injection time = 500 ms; microscan = 1.

Results and discussion

Workflow of the identification of glycopeptides. Raw data were converted to Mascot Generic Format (MGF) by pParse (version 2.0.6)23. After removal of glyco-oxonium ions (see the section “Glyco-oxonium ions” and Table S-1 in the Supporting Information), HCD- and CID-MS/MS spectra were deisotoped and then analyzed by pGlyco against a glycan database that combined two previously reported glycan databases6,24,25: 1) three largest plausible N-glycans, corresponding to high mannose-, hybrid-, and complex-type glycan structures in human serum were compiled into one glycan database, containing 2,012 non-isomorphic glycan structures; 2) the GlycomeDB database (http://www.glycome-db.org) recorded 34,457 glycans, and only N-linked glycans were extracted. After combining these two glycan database and removing redundancies, we obtained 2,860 N-glycan structures. For the analysis of the HCD/CID-MS/MS spectrum pair, the mass tolerance for precursors and fragments were set as ±10 ppm and ±20 ppm respectively.