Distinguishing N -acetylneuraminic acid linkage isomers on glycopeptides by ion mobility-mass spectrometry †

Diﬀerentiating the structure of isobaric glycopeptides represents a major challenge for mass spectrometry-based characterisation techniques. Here we show that the regiochemistry of the most common N -acetylneuraminic acid linkages of N -glycans can be identified in a site-specific manner from individual glycopeptides using ion mobility-mass spectrometry analysis of diagnostic fragment ions.

Protein glycosylation as post-translational modification tremendously influences cellular events such as cell-cell interactions and receptor recognition. 1,2Glycosylation is highly dynamic, cell-type specific and depends on a variety of additional factors such as the developmental status of the cell. 3,46][7][8][9] Therefore, knowledge of both individual glycan structure and the location on a given protein is important in revealing these complex structure-function relationships. 6iquid chromatography-mass spectrometry (LC-MS) is a very sensitive technique that is widely used for studying site-specific protein glycosylation. 5,7,9,10Differentiation of minor changes in glycan structure, such as terminal N-acetylneuraminic acid (NeuAc) linkages, directly from glycopeptides is very challenging due to limits in LC separation and the isobaric nature of the fragments observed by MS. 7,10 A promising technique capable of providing additional structural information is ion mobility spectrometry (IM) coupled to mass spectrometry (IM-MS). 11In IM, analyte ions travel through a cell filled with an inert neutral gas aided by a weak electric field and undergo a series of low-energy collisions with the gas.Compact ions undergo fewer collisions with the drift gas than more extended ions and therefore traverse the IM cell faster.As a result, in IM-MS molecular ions or their fragments are not only separated according to their mass and charge, but also according to their size and shape, which enables the separation of isomers.The obtained drift times can be converted into rotationally-averaged collision cross sections (CCSs), 12,13 which are absolute biophysical properties that can be used for structural classification by reference to database values. 14M-MS studies on isolated glycan isomers show great promise, [15][16][17][18][19] however, very few reports focus on isomeric glycopeptides. 18,20Here we report a universally applicable and rapid approach capable of differentiating a2,3 and a2,6 NeuAc linkages in N-glycopeptides without any additional sample preparation steps.Using IM-MS we evaluated a small set of well-defined, synthetic glycopeptides carrying N-glycans containing either a2,3 or a2,6 linked NeuAc residues.To illustrate the robustness of the method, we tested complex mixtures using two forms of a-1 proteinase inhibitor (A1PI) produced in different cell types.
Homogeneous glycopeptides were generated by chemical and chemo-enzymatic synthesis for systematic IM-MS method development.An asparagine (Asn) building block carrying biantennary, mono-or disialylated N-glycans was obtained from egg yolk using a combination of extraction and proteolytic digestion steps (for details see ESI †). 21,22This glycan-Asn building block was Fmoc protected, the NeuAc residues benzylated and the molecule subsequently used to synthesize N-glycopeptides by solid-phase peptide synthesis (SPPS).
The first set of glycopeptides was designed based on the naturally occurring tryptic peptide sequence from human butyrylcholinesterase ( 505 YGNPNETQNNSTSWPVFK 522 , UniProt entry P06276). 23This peptide contains three possible glycosylation sites (boldface) defined by the consensus sequence N-X-T/S/C (X a P), two of which are glycosylated in serum. 23To reduce complexity, the sequence was simplified to YGNVNETQNNSFK and an a2,6 disialylated, biantennary glycan was selectively incorporated by SPPS at one glycosylation site, either near the N-(GP1) or near the C-terminus (GP2).IM-MS experiments were performed to determine whether the method can separate the isobaric glycopeptides.As protonated ions, both isomers, regardless of the charge state, could not be separated and showed identical drift times (Fig. S4, ESI †).However, when quadruply deprotonated ions ([M À 4H] 4À = 928) were measured, GP1 and GP2 had noticeably different drift times of 5.80 and 5.33 ms (Fig. 1).Both isomers, when examined as mixtures, were nearly baseline separated illustrating that IM-MS can in principle be used to differentiate isobaric glycopeptides that merely differ in their glycosylation site.
Next, we examined two glycopeptides with a single glycosylation site but different glycan structures.Specifically, the attached complex-type glycans differed in the linkage of the terminal NeuAc residue, that is either a2,3or a2,6-linked to galactose.Subtle differences in NeuAc regiochemistry are of biological and biopharmaceutical importance 8,24 and are challenging to characterise using established glycoproteomics techniques. 25,26he investigated peptides are designed based on the human protein C fragment 284 EVFVHPNYSK 293 (UniProt entry P04070) that contains one glycosylation site (boldface).The peptide was synthesized by SPPS using an Asn building block containing an a2,6 monosialylated, biantennary N-glycan.Subsequently, a fraction of the resulting glycopeptide (GP3) was desialylated using trifluoroacetic acid followed by enzymatic re-sialylation using recombinant b-galactoside a2,3-sialyltransferase 3 (see ESI †).Finally, a glycopeptide (GP4) exclusively carrying a2,3 linked NeuAc residues was obtained.
Only marginal, non-significant drift time differences between GP3 and GP4 were observed for 3+ molecular ions of the intact glycopeptides such that they cannot be separated in mixtures (Fig. 2b).Given the minor structural differences in the glycan moiety compared to the overall size of the molecule this result was not surprising.
When collision-induced dissociation (CID) is applied to positively charged glycopeptide ions, glycosidic cleavages are the preferred fragmentation pathway resulting in a multitude of oligosaccharide-only fragment ions (B-and Y-type fragments in Fig. 2).¶ 27 Tandem MS analysis of GP3 and GP4 yielded almost identical CID fragment spectra and did not provide diagnostic information for any glycan structural features (Fig. 2a).However, from these oligosaccharide-only fragment ions the m/z 657 B 3 type fragment is of particular interest because this oxonium ion corresponds to a trisaccharide consisting of Gal, GlcNAc and NeuAc residues in N-glycopeptides.To elucidate whether regiochemical differences of the NeuAc linkage in GP3 and GP4 leads to drift time differences of the resulting fragments, CID experiments were followed by IM-MS analysis.The extracted arrival time distributions (ATDs) of these m/z 657 B 3 ions were vastly different for GP3 and GP4 (Fig. 2b), with the a2,6 fragment exhibiting a considerably shorter drift time when compared to the a2,3 equivalent.When both glycopeptides were mixed, the isomeric NeuAc-containing fragments showed baseline separation.In addition, the collision cross sections in nitrogen drift gas ( TW CCS N2 ) of 236 Å 2 for the a2,6 linked NeuAc and 246 Å 2 for the a2,3 linked NeuAc fragments differed about 4%, well above the 1.5% error of the method. 13These values are highly diagnostic to the regiochemistry of the underlying NeuAc linkage and can be used to gain site-specific information on important glycan structural features directly from individual glycopeptides in a single experiment.
To evaluate whether similar diagnostic fragments could be obtained from complex mixtures within a glycoproteomics workflow, forms of the glycoprotein A1PI were tested (Fig. 3a).N-Glycan NeuAc linkages on A1PI differ depending on the biological source of the protein.Human plasma A1PI contains mostly a2,6 linked NeuAc residues 25 whereas the same protein recombinantly expressed in Chinese hamster ovary (CHO) cells only contains a2,3 sialylated glycans. 28A1PI samples from both sources were purified using SDS-gel electrophoresis followed by in-gel tryptic digestion and glycopeptide enrichment using hydrophilic interaction chromatography (HILIC) to remove unglycosylated peptides as these can suppress efficient glycopeptide detection (for details see ESI †). 22Subsequently, the purified glycopeptides were analysed off-line by IM-MS (Fig. S5 and S6, ESI †).Known glycopeptide precursor ions 25 were m/z-selected for tandem IM-MS experiments and characteristic B 3 fragments were observed after CID (Fig. 3b).Importantly, the drift times of the obtained B 3 fragments of 7.03 ms (from CHO precursor m/z 1273 and 1401) and 6.44 ms (from human plasma precursor m/z 1321) were essentially identical to those observed for the synthetic glycopeptides GP4 (6.96 ms) and GP3 (6.38 ms), respectively.In addition, the TW CCS N2 of the oxonium fragment ions (Fig. 3b) were consistent with the synthetic reference glycopeptide data and can be used to differentiate a2,3 from a2,6 sialylated N-glycans from all sialylated glycopeptides obtained from CHO-derived and human A1PI (Fig. 3b).The universal applicability of this approach is demonstrated, as size and sequence of the glycopeptide precursors did not affect the drift time of the resulting B 3 fragment.This finding underscores the diagnostic nature of the a2,3 or a2,6 fragments and justifies this approach as a reliable fragment-based method to obtain N-glycan structure information directly from glycopeptides.
In conclusion, we show that IM-MS can significantly improve the identification of isomeric glycopeptides and fits seamlessly within existing glycoproteomics workflows.Peptides with two distinct glycosylation sites can be differentiated directly as intact molecular ions using IM-MS.The regiochemistry of the prevalent a2,3 and a2,6 NeuAc linkages in N-glycosylated peptides can be distinguished and allows for the unambiguous identification in a site-specific manner on basis of the CCSs of diagnostic B 3 -type fragments that are cleaved directly from mass-selected glycopeptide precursors.The approach is fast, does not require derivatisation and is universally applicable regardless of the nature of the investigated glycoprotein.Our data highlight the immense potential of IM-MS to be implemented into existing glycoproteomic workflows.

Fig. 1 Fig. 3 Fig. 2
Fig.1IM-MS separation of the isobaric glycopeptides GP1 and GP2.Two isomeric glycopeptides that share the same sequence and attached glycan, but differ in the site of glycan attachment can be distinguished based on their drift time (top and middle) and separated in mixtures (bottom) when analysed as quadruply deprotonated ions.§