Analysis of Viral Spike Protein N-Glycosylation Using Ultraviolet Photodissociation Mass Spectrometry

Abstract

Characterization of protein glycosylation by tandem mass spectrometry remains challenging owing to the vast diversity of oligosaccharides bound to proteins, the variation in monosaccharide linkage patterns, and the lability of the linkage between the glycan and protein. Here, we have adapted an HCD-triggered-ultraviolet photodissociation (UVPD) approach for the simultaneous localization of glycosites and full characterization of both glycan compositions and intersaccharide linkages, the latter provided by extensive cross-ring cleavages enabled by UVPD. The method is applied to study glycan compositions based on analysis of glycopeptides from proteolytic digestion of recombinant human coronaviruse spike proteins from SARS-CoV-2 and HKU1. UVPD reveals unique intersaccharide linkage information and is leveraged to localize N-linked glycoforms with confidence.

Graphical Abstract

INTRODUCTION

Although the importance of protein glycosylation in numerous disease pathways is undisputed, glycosylation remains a challenging post-translational modification to localize, characterize, and quantitate.¹ As one very timely example, the viral spike glycoprotein of SARS-CoV-2 is the antigen in leading COVID-19 vaccines, and extensive initiatives have been launched to decipher the full glycan topography to guide vaccine development.² Protein glycosylation can be grouped into two general classes based on the covalent linkage connecting the glycan to the protein: N-linked and O-linked, in which glycans are attached to asparagine or serine/threonine side chains, respectively. N-linked glycans follow a genetically coded sequence (N-X-S/T) that facilitates their localization to appropriate asparagine side chains, yet characterization of their composition continues to be a challenging analytical task.³ O-linked glycans are not explicitly encoded, but some mucin-type consensus motifs may be predicted based on the amino acid sequence,⁴ and glycosaminoglycans are often found bound to serine residues at Ser–Gly repeats;⁵ nonetheless, their localization and characterization remain challenging. In addition to glycan macroheterogeneity, the microheterogeneity of glycan modifications at specific residues is another daunting aspect of glycoproteomic analysis. Tandem mass spectrometry (MS/MS) has been an integral tool in the glycoproteomic toolkit, primarily using various MS/MS methods for the examination of glycopeptides produced upon proteolytic digestion of glycoproteins.^6-9

Collisional-activated dissociation (CAD) and electron transfer dissociation (ETD) are the two activation methods most often used in glycopeptide analysis by MS/MS.^10-13 These methods are often integrated in a complementary manner to enhance the ability to pinpoint glycosylation sites (via ETD) and provide glycan structural insights (via CAD) due to the facile fragmentation of the attached glycan. Although these methods are able to provide monosaccharide compositional information, they provide minimal intersaccharide linkage information in the way of cross-ring glycosidic cleavages. Targeted methods have been implemented to take advantage of the abundant glycan-specific oxocarbenium ions produced upon CAD to trigger ETD of the same glycopeptide, the latter providing site-specific localization of glycans and peptide sequence information.^11,14-17 In addition, hybrid MS/MS methods such as electron-transfer higher-energy collisional dissociation (EThcD),¹⁸ which combines CAD and ETD, have improved the fragmentation efficiency of ETD and resulted in greater abundances of meaningful glycosidic fragment ions.¹⁹ Another innovative strategy employs multiple-stepped collisional energies during CAD to modulate the degree of both glycan and peptide fragmentation, as demonstrated for analysis of N-glycopeptides.^20,21 Extensive optimization and comparison of CAD and ETD methods for glycopeptides suggest that CAD methods outperform ETD approaches for N-linked glycoproteomics.¹⁷

Despite the advances in enrichment techniques,^22-24 MS/MS strategies,^9,25,26 and search engines^27,28 for glycoproteomics, the need for methods that perform well for proteins containing heterogeneous mixtures of glycans (such as viral proteins) and for profiling glycoproteins in complex mixtures and for unraveling intersaccharide linkages remains unmet.²⁹ As such, key aims in the quest to map glycosylation are to characterize the peptide sequence, localize the sites of modifications, and determine both glycan composition and intersaccharide linkages. We have explored the development and application of ultraviolet photodissociation (UVPD) for O-glycoproteomics.^30-32 UVPD allows access to fragmentation pathways with significantly higher activation energies, enabling the formation of a wide array of diagnostic ions and preserving labile bonds that facilitate PTM mapping.^33,34 UVPD also results in glycosidic cross-ring cleavages of glycans^30,31,35-38 or oligosaccharides,^32,33 resulting in A/X fragments that are important for the characterization of intersaccharide linkages in complex glycosylated molecules.

Collisional-activated and electron-based MS/MS methods provide compositional glycan information of glycopeptides, but determination of intersaccharide linkages, especially at the non-reducing termini of branching saccharide structures or at bisected glycans, requires cross-ring fragmentation. The specific patterns of intersaccharide linkages influence the topologies of glycans and thus impact the recognition and docking of ligands based on the spatial orientation of the receptor and ligand.^39,40 The glycan topologies also play a role in aggregation and may even modulate stabilization of protein conformation.^39-41 The ability to construct complete maps of intersaccharide linkages is a key step in elucidating glycan topologies, and this aim served as a key motivating impetus for the present work utilizing UVPD.

Here, we apply an HCD-triggered-UVPD MS/MS method to characterize the SARS-CoV-2 alpha variant and HCoV-HKU1 spike glycoproteins. We focus on viral spike proteins as representative glycoproteins due to their role as vaccine antigens in the battle against coronaviruses (CoVs) and their incorporation of both abundant N-glycosylation and less-prevalent O-glycosylation.⁴² A significant structural feature of the SARS-CoV-2 S protein is the 22 genetically coded N-glycans that play an essential role in modulating the spike protein receptor-binding domain’s conformation and thus recognition of angiotensin-converting enzyme 2 (ACE2) that facilitates entry into the host cell.^43-45 The extensive glycosylation of viral proteins also plays a vital role in proper protein folding and evading host immune system responses by shielding exposed viral protein epitopes, an ongoing challenge as new SARS-CoV-2 variants emerge.^46,47 The latest SARS-CoV-2 glycan profiling studies indicate that up to 20 of the 22 N-linked glycosylation sites are composed of complex-type glycoforms, with reports of up to 9 occupied O-linked sites containing mucin-type glycoforms.^48-57 Beyond glycan occupancy, glycan linkage information directly reflects the glycan landscape topography of protein glycosylation which affects antigen binding sites and thus antibody binding profiles.^58-62 Although conventional methods can be leveraged to provide general glycan profiles of target proteins, UVPD can be harnessed to not only provide compositional glycoform information but also structural characteristics of the glycans themselves. As described in the present study, we demonstrate the use of HCD-triggered UVPD to identify the N-glycosylation sites of SARS-CoV-2 and HCoV-HKU1 spike glycoproteins and, importantly, position UVPD as a fragmentation modality capable of characterizing complex branched glycans with intersaccharide linkage details.

EXPERIMENTAL SECTION

Sample Preparation.

The three glycopeptide standards were previously synthesized and characterized, with different glycoforms attached to the Asn of the same peptide sequence EEQFNSTFR.⁶³ The glycoforms are Gal(β1-4)GlcNAc(β1-2)Man(α1-6)[GlcNAc(β1-2)Man(α1-3)]Man(β1-4)GlcNAc-(β1-4)GlcNAc(β1) for GP1, Gal(β1-4)GlcNAc(β1-2)Man-(α1-6)[GlcNAc(β1-2)Man(α1-3)]Man(β1-4)GlcNAc(β1-4)-[Fuc(α1-6)]GlcNAc(β1) for GP2, and Neu5Ac(α2,6)Gal(β1-4)GlcNAc(β1-2)Man(α1-6)[GlcNAc(β1-2)Man(α1-3)]Man-(β1-4)GlcNAc(β1-4)[Fuc(α1-6)]GlcNAc(β1) for GP3.

The recombinant stabilized form of the spike (S) glycoprotein from the severe acute respiratory syndrome-related coronavirus 2 (SARS-CoV-2) alpha variant, Wuhan-Hu-1 (GenPept: YP_009724390), was obtained through BEI resources. The recombinant form of the spike glycoprotein contains residues 5-1202 of the SARS-CoV-2 spike glycoprotein with substitutions at the furin S1/S2 cleavage site RRAR to SGAG at residues 673 to 676, and stabilizing point mutations noted at K986P and V987P. The glycoprotein was produced in human embryonic kidney HEK293 cells, purified by immobilized metal affinity chromatography, and dialyzed into 50 mM Tris (pH 8.0), 150 mM NaCl, and 0.25% l-histidine.

Plasmid-encoding residues 1-1277 of the HCoV-HKU1 spike with a mutated S1/S2 cleavage site, proline substitutions at residues 1067 and 1068, a C-terminal T4-fibritin trimerization motif, an 8× HisTag, and a TwinStrepTag (HCoV-HKU1 S-2P) were transiently transfected into Free-Style293F cells with polyethyleneimine. Three hours posttransfection, these cells were treated with kifunensine resulting in the generation of primary oligomannose glycans, preventing glycan maturation through the ER/Golgi complex beyond N-linked glycosylation. After a 6 day expression period, transfected supernatants were harvested, and protein was purified using StrepTactin resin (IBA). The affinity-purified spike was then further purified by size-exclusion chromatography using a Superose6 10/300 column (Cytiva).

20 μg of each spike protein was denatured for 10 min at 100 °C in 100 mM Tris–HCl (pH 8.0) and 20 mM tris(2-carboxyethyl)phosphine (TCEP). Each protein was subsequently reduced and alkylated by adding 35 mM iodoacetamide (Sigma-Aldrich, St. Louis, MO) and incubated for 30 min in the dark, followed by a final addition of 10 mM TCEP to halt alkylation (Pierce, Rockford, IL). Next, 5% sodium dodecyl sulfate (w/v) was added to each aliquot (Sigma-Aldrich, St. Louis, MO), to which trypsin and chymotrypsin were added in a 1:25 enzyme: protein (w/w) ratio (Pierce, Rockford, IL). The proteolytic reaction was conducted within an S-Trap micro spin column and then cleaned using a (9:1) methanol: 100 mM Tris–HCl (pH 8.0) ratio (Protifi, Farmingdale, NY). Following digestion, peptides were eluted from the spin column using 50% acetonitrile and 0.2% formic acid and dried under medium heat to concentrate the sample. Samples were reconstituted in 0.1% formic acid and 2% acetonitrile in water for LC–MS analysis.

LC–MS Analysis.

Each aliquot was separated on a Dionex RSLC 3000 nano-LC system (Thermo Fisher Scientific, San Jose, CA, USA). Approximately, 500 ng of the digested glycoprotein was injected into an in-house packed 3 cm C-18 trapping column (3 μm, 300 Å pore size, 75 μm ID). Peptides were then eluted onto a 20 cm fritted C-18 (1.8 μm, 300 Å pore size, 75 μm ID, packed in-house) analytical column (New Objective, Woburn, MA, USA). Separation was performed using mobile phase A consisting of 0.1% formic acid in water and mobile phase B consisting of 0.1% formic acid in acetonitrile. A 145 min linear gradient was applied from 8 to 35% in mobile phase B at a flow rate of 300 nL/min. The nanoLC system was interfaced to a Thermo Scientific Instruments Lumos Orbitrap mass spectrometer (San Jose, CA, USA) equipped with a 193 nm excimer laser for UVPD as described previously, and each analysis was conducted in triplicate.³⁶ MS¹ spectra were collected in a top-speed data-dependent fashion with a dynamic exclusion of the precursor for 20 s after one repeated activation event. All MS² spectra were acquired in the Orbitrap mass spectrometer (30,000 resolution at 200 m/z) with an isolation width of 1.6 Da. HCD mass spectra were collected using 30 NCE, and UVPD was performed in the high-pressure linear ion trap using six 5 ns pulses (3.0 mJ per pulse) from a 193 nm, 500 Hz excimer laser (Coherent Excistar, Santa Clara, CA). EThcD was performed in the high-pressure linear ion trap with 50 ms reaction time for ETD (2 × 10⁵ reagent AGC) with 15% supplemental collisional energy.

Database Searching.

Database searching and glycopeptide quantitation were accomplished using Protein Metrics Inc. software by combining Byonic and Byologic proteomics packages. Glycan linkage analysis was accomplished using SimGlycan v. 5.94, a software by PREMIER Biosoft. Byonic (Protein Metrics, Cupertino, CA, USA) proteomic database search was used for large-scale identification and site localization of N-linked glycopeptides with a precursor and a fragment mass tolerance of 10 ppm. The main scoring metrics used to gauge the confidence of identified glycopeptides are the Delta Mod Score and the PEP 2D score, as previously described.³² All glycopeptide fragmentation data were evaluated manually for each identified glycopeptide meeting scoring metrics described in the Supporting Information. Each MS replicate data set was searched using the Protein Metrics N-glycan library and quantitated based on extracted ion chromatographic (EIC) areas for each glycopeptide precursor using Byologic software. The relative quantification of glycan composition was represented as the mean of three technical replicates ± standard error of the mean. Glycan linkage analysis of model glycopeptides was conducted by matching confidently identified glycopeptides and manually searching the Byonic PSM scan number within SimGlycan 5.94 (PREMIER Biosoft, San Francisco, CA, USA). Standard glycopeptides were validated by deconvolution using the Xtract algorithm (Thermo) and analysis using Prosight Lite using a fragment tolerance of 10 ppm. The “protein characterization score” (PCS) is used in the analysis of these infused glycopeptides to determine the quality of fragment matching.⁶⁴ Glycan branches are listed as alpha (α) or beta (β) following the Domon and Costello nomenclature,⁶⁵ and intersaccharide linkages are deliberately listed as a or b linkages to designate alpha or beta linkage positions, respectively.⁶⁶

RESULTS AND DISCUSSION

In this work, we extend the use of 193 nm UVPD to an N-glycoproteomic workflow to provide both comprehensive glycopeptide characterization, including determination of intersaccharide linkage positions and relative quantification of glycan subclasses from intact glycopeptides (Figure 1). We demonstrate the capacity of UVPD to provide glycan localization, both glycan and peptide sequencing of N-linked branched glycopeptides, and both intersaccharide linkages and branching pattern analysis of antennary glycans derived from tryptic/chymotryptic digestion of viral spike proteins. To develop this method, we used two viral spike proteins, the SARS-CoV-2 spike protein (SARS-2-S), which has been the subject of a flurry of recent work owing to the tremendous impact of the COVID-19 pandemic, and the HKU1 spike protein (HKU1-S), a common beta-coronavirus known to cause respiratory illness in humans (sequences shown in Figure S1).^67,68 UVPD fragmentation patterns enable precise localization of the glycans and informative monosaccharide-level composition of the complex branched glycan structures. UVPD is able to dissect branching patterns of antennary N-glycans which are integral in the analysis of antigenic characterization and development of mature glycoproteins involved in biotherapeutics.⁶⁹ In addition, relative quantitation of individual glycoforms was performed by monitoring extracted ion chromatograms of glycopeptides (i.e., intact masses of precursor glycopeptides validated by UVPD).

Figure 1.

HCD-triggered-UVPD approach for N-glycoproteomic analysis providing glycopeptide characterization and relative quantitation.

To develop guidelines for the fragmentation of complex glycopeptides by 193 nm UVPD, we used a standard glycopeptide (GP1)⁶³ (EEQFNSTFR) and glycan structure, the latter composed of Gal(β1-4)GlcNAc(β1-2)Man(α1-6)[GlcNAc(β1-2)Man(α1-3)]Man(β1-4)GlcNAc(β1-4)GlcNAc(β1) attached via an amide β-linkage to the Asp (N) side chain of the peptide (Figure S2A). Upon HCD of protonated GP1 (3+ charge state), the fragment ions of greatest abundance ions are oxonium ions and glycosidic B/Y-type cleavages, allowing full glycan sequencing and determination of the branching pattern but not providing insights into the specific intersaccharide linkages (Figure S3 and Table S1). Owing to the labile nature of the N-glycan, collisional activation primarily cleaves the N-glycan from the peptide, a well-known outcome of CAD methods in the context of glycopeptide analysis. The few diagnostic b/y sequence ions that are produced afford some peptide sequence information but do not retain the glycan, thus impeding confident localization.

For the benchmark peptide GP1, UVPD results in the complete characterization of the peptide sequence, and many glycan-type ions, see Figure 2 and the summary in Table S2). UVPD imparts high energy deposition, inducing backbone cleavages of the peptide without disrupting the intact glycan and promoting glycoside cleavages that map the composition and sequence of the glycan. For GP1, nearly the entire series of peptide backbone fragment ions are observed, many retaining the key 1460.53 Da mass shift attributed to the glycan (e.g., a₆, a₇, y₆, y₇, y₈, etc.) in addition to ones that unequivocally bracket the site of modification to Asn5 (e.g., a₄, a₅, y₄, y₅, etc.,). In general, many b and y ions retain the intact glycan upon UVPD, whereas conventional collisional activation (like HCD) more often produces b/y ions that have lost the labile modification. The expansion in Figure 2C shows one of the diagnostic cross-ring cleavage ions produced by UVPD; it is the ^2,4X₂ fragment ion that identifies the linkage pattern at the antennae containing mannose, localizing Gal to the a1-6 branch. Although ions generated by UVPD often have lower relative abundances, they still exhibit excellent S/N ratios, match theoretical isotope fits, and have high mass accuracies. The numerous glycosidic cleavages upon UVPD that leads to B/Y/Z fragment ions allow the glycan to be sequenced and its composition identified (Figure 2D). Moreover, UVPD causes cross-ring fragmentation that provides information on intersaccharide linkage information via A/X-type ions.^36,70

Figure 2.

(A) Glycan mapping of the GP1 glycopeptide (m/z 873.69, 3+) by 193 nm UVPD (six pulses, 3.0 mJ per pulse). UVPD induces iterative fragmentation throughout the peptide backbone and glycan chain for peptide and glycan sequencing. (B) Expanded region of the mass spectrum from m/z 600 to 1300 showing many glycan-type product ions. (C) Expansion of the ^2,4X₂ ion produced from a cross-ring cleavage displays a close match to the theoretical isotope pattern in red and identifies the linkage pattern at the antennae containing mannose, localizing Gal to the a1-6 branch. (D) Glycan sequence information and cross-ring cleavages. The array of glycan fragments is listed in Table S2 in the Supporting Information.

The GP1 glycopeptide was also analyzed using an EThcD method for comparison. As displayed in Figure S4 and Table S3, EThcD allowed localization of the glycan based on peptide fragment ions that bracketed the glycan site and resulted in several B/Y-type glycosidic ions that aid glycan sequencing. However, charge reduction of the precursor ion was the most dominant process, few of the longer peptide fragment ions retained the glycan, and there was a small gap in the peptide sequence coverage. Nonetheless, as shown in Figure S4B, EThcD reveals the complete glycan sequence and confirmation of monosaccharide composition but does not offer complete intersaccharide linkage information.

After systematically optimizing of UVPD parameters, we confirmed that UVPD provides intact structural information for both the peptide and glycan moiety of three glycopeptides (GP1, GP2, and GP3) (Figure S2). Specifically, the number of laser pulses and the laser power were varied to evaluate their impact on sequence coverage and retention of each glycan based on the average number of matching peptide fragment ions containing the intact glycan moiety (Figure S5A). A negligible difference in peptide fragmentation is noted beyond five pulses of 2.5 mJ. However, considering glycan-specific fragments (Figure S5B), a decrease in identifiable glycan fragments was noted beyond six pulses of 3.0 mJ. Therefore, an effort was made to optimize both the number of explained glycan fragments and peptide fragments without inducing excessive fragmentation, precluding glycan characterization. The parameters that reliably yielded the most informative spectra with optimum sequence coverage and glycan retention consisted of six 5 ns laser pulses of 3.0 mJ. Full sequence coverage of the peptide backbone, precise glycan localization, and characterization of the glycan composition and linkages were accomplished using UVPD to analyze GP1, GP2, and GP3 (Figures S6 and S7 and Tables S4 and S5) based on the production of a combination of reducing-end glycopeptide ions containing the intact peptide and non-reducing end glycan-specific fragment ions.

The array of low-mass oxocarbenium ions (m/z 127.039, 138.055, 145.049, 168.066, 186.076, and 204.087) observed in Figure S3 generated upon HCD are characteristic of hexose-(Hex) and N-acetylhexosamine- [HexNAc, such as N-acetylglucosamine (GlcNAc)] containing glycopeptides^15,16,71 and can be used as reporter ion filters to trigger UVPD of the same precursor ion that was selected for HCD, a method termed HCD-triggered UVPD.³² Detection of any oxocarbenium ion within the top 20 product ions in an initial HCD scan is thus used to trigger UVPD of the same precursor ion in the subsequent scan, allowing full characterization of glycopeptides. Combining the oxonium reporter ions from HCD with the extensive sequencing/glycan characterization information from UVPD affords a powerful MS/MS strategy that facilitates selective targeting of glycopeptides. The selective targeting aspect is particularly crucial for complex mixtures in which the glycopeptides typically have low abundances relative to numerous non-glycosylated counterparts. HCD-triggered UVPD was applied to both map and characterize glycans of glycopeptides produced upon digestion of two human coronavirus spike proteins. The first protein examined was the SARS-2-S protein, a key glycoprotein directly involved in viral entry through recognition of the ACE2 receptor on the cell.^72,73 The second protein was the HKU1-S glycoprotein for which no protein receptor has been identified; however, a neutralizing antibody was discovered that binds its receptor-binding domain (RBD), effectively blocking HKU1 infection of cells.^74-76 HCoVs are composed of four structural proteins, including the spike, envelope, membrane, and nucleocapsid proteins. Reportedly, HCoVs are not as heavily shielded by glycans as other pathogenetic virion surfaces, such as HIV-1.⁶⁷ However, glycosylation of the spike protein of HCoVs plays a role in virulence, such as the 22 predicted N-glycosylation sites on SARS-2-S protein which modulate the conformation of the spike protein and thus mediate cellular recognition.⁴³

A reversed-phase LC–MS/MS approach was implemented to separate and analyze the tryptic/chymotryptic digests of these two heavily glycosylated spike proteins (Figure S8). The HCD-triggered UVPD approach generated thousands of spectra that were searched against a combined SARS-CoV-2 and HKU1 proteome reference database with the addition of each recombinant protein sequence appended to each respective FASTA file. As before, HCD was leveraged to create abundant oxocarbenium ions corresponding to Hex-containing ions (m/z 127.039, 145.049, 163.060, and 325.112) or HexNAc-containing ions (m/z 138.055, 168.066, 186.076, 204.087, and 366.138). The identification of specific oxonium ions was used to trigger UVPD, thus yielding confident peptide sequence information and conclusive localization of glycosylation sites for every glycopeptide in the digests. The product ion-triggered methods used here filter peptides in the LC runs and only subject glycan-containing glycopeptides to MS/MS, thus dedicating the duty cycle to glycopeptide characterization and not the analysis of non-glycosylated peptides.

For the SARS-2-S protein overall, 99% protein sequence coverage was observed using the HCD-triggered UVPD strategy with 106 unique glycoforms corresponding to 21 different N-glycosylated sites across each of the three replicates. The peptide-level sequence coverages obtained for SARS-2-S averaged 79% for HCD, 79% for HCD-triggered EThcD, and 80% for HCD-triggered UVPD. The annotated UVPD mass spectrum of one representative glycopeptide, TQSLLIVNN(+1437.5128)ATNVVIK containing a HexNAc(3)Hex(5) identified as a (GlcNAc)3(Man)5, and glycan map are shown in Figure S9 (Table S6). Fragmentation throughout the peptide backbone pinpoints the location of the hybrid-class glycan at residue Asn113 of the glycoprotein, and production of numerous A/X, B/Y, and C/Z ions provides characterization of the triantennary N-linked glycan. Figure S10 displays the UVPD mass spectrum of a peptide (4+ charge state) containing a highly branched HexNAc(6)Hex(7)Fuc(1) glycan on peptide LHVTYVPAQEKN(+2516.9145)F. Without cross-ring cleavages at the tri-mannose core (^0,4A₄, ^1,5A₃), one might assume that this is a bi-antennary linear glycan, but cleavage at each mannose indicates it is a more complex highly branched glycan. The UVPD mass spectrum of another peptide (3+ charge state) containing a highly branched glycan HexNAc(6)Hex(7)Neu(1) on EGVFVSN(+2661.9520)-GTHWF is shown in Figure S11. Once again, the glycan is highly branched (complex-type) with a bound neuraminic acid. Due to the near-symmetric nature of this glycan, the highly branched structure can be confirmed through the pattern of cross-ring cleavages at the tri-mannose core (^2,4X₂, ^1,5X₃), but the precise location of the bound sialic acid cannot be pinpointed.

The number of glycosylated sites identified and distribution of glycan compositions obtained by the HCD-triggered-UVPD approach were compared to stepped HCD and HCD-triggered-EThcD for SARS-2-S (Figure 3) by counting peptide spectral matches (PSMs) and by integrating the area of specific identified glycopeptides from extracted ion chromatograms generated by Protein Metrics Byologic software from each LCMS run. The HCD-triggered UVPD approach identified 21 N-glycosylated sites compared to 17 sites for stepped HCD and 15 sites for HCD-triggered EThcD. Overall, the stepped-HCD approach resulted in 5639 ± 57 PSMs, HCD-triggered UVPD returned 4368 ± 48 PSMs, and the HCD-triggered-EThcD approach yielded 2128 ± 42 PSMs. Despite the markedly higher number of PSMs provided by the stepped-HCD method, the HCD-triggered-UVPD approach afforded greater structural detail of the assessed glycopeptide due to the specificity and speed of the oxonium-triggered method. Furthermore, among all oxonium ion-triggered data-dependent MS2 methods, UVPD displays a significant reduction in cycle time when compared to EThcD and is comparable to spHCD (Figure S12). The glycan class was categorized according to the monosaccharide composition observed upon MS/MS and comparison against an accurate intact glycan mass using the Protein Metrics N-glycan library within a 10 ppm error. The compositions of the N-linked glycans of each glycoform for each site are categorized based on the three major classes of oligosaccharides: complex, hybrid, or oligomannose (see Scheme S1). Based on MS/MS analysis of each glycopeptide observed, breakdowns of the specific glycan subclass compositions are shown in Figures S13-S15 and Tables S9-S11 for each of the three MS/MS methods. The capability of HCD-triggered UVPD for characterization of hybrid-type glycans exceeded that of the other two MS/MS methods, especially at glycosylation sites N343, N616, and N657.

Figure 3.

Distribution of glycan compositions at each N-linked amino acid site of SARS-2-S using (A) HCD-triggered UVPD, (B) stepped HCD, and (C) HCD-triggered EThcD. Glycans are categorized as complex, hybrid, or oligomannose.

By focusing on the glycopeptides containing three specific glycosylation sites (N343, N616, and N657), the performance of UVPD can be compared to that of the other MS/MS methods (Tables S9-S11). Glycans on site N343 were found on peptide GEVFNATR (2+ charge state), following the N-X-S/T motif. For this peptide, UVPD and HCD vastly outperformed EThcD likely due to the preferential charge-reduction pathways and preferential fragmentation of the glycan moiety with few key site-localizing peptide sequence ions present. Preferential fragmentation of the glycan thus generated low-scoring PSMs which precluded glycan localization or peptide fragmentation. Four glycans were found on site N343 using EThcD, in which 60% of the EIC area of the N343-modified peptides was composed of oligomannose glycopeptides and 40% from peptides containing complex glycans. The spHCD method identified 17 glycans associated with N343, in which 11% of the EIC area was from oligomannose glycopeptides and 89% was from peptides containing complex glycans. Finally, UVPD identified 28 glycans for the N343 site, consisting of 13% oligomannose-type glycopeptides, 4% hybrids, and 83% were glycopeptides containing complex-type glycans. N616 glycans were found on peptide QDVNCTEVPVAIHADQLTPTWR (3+ or 4+ charge state, with glycans on the fourth residue). This higher-charge state peptide should be well-suited for ETD; however, the slow acquisition speed through the elution of this glycopeptide impeded rapid acquisition of spectra. For site N616, only nine glycans were identified for this glycopeptide using EThcD, and all were oligomannose glycans. For the spHCD method, a total of 22 glycans were found, with 7% containing oligomannose glycans and 93% containing complex glycans. UVPD identified 19 glycans at site N616, with 7% containing hybrid glycans and 93% complex-type glycans. Finally, the glycoforms identified on N657 were identified based on analysis of peptide AGCLIGAEHVNNSYECDIPIGAGICASY (3+ charge state, glycans at position 4). No glycans were identified using EThcD, and spHCD generated PSMs with poor fragmentation throughout the peptide sequence, and thus, no glycopeptide met PSM score cutoffs for reporting. Eight glycans on site N657 were identified by UVPD, all corresponding to hybrid-type glycans.

Comparison of the global glycan composition shows agreement with recombinantly expressed SARS-2-S glycosylation patterns previously reported.^49-53 The N-glycosylation pattern at Asn234 has been reported to inform on the origin of the spike protein categorizing the protein as recombinantly expressed or native.⁵² The glycosylation pattern observed in the present analysis agrees with initially reported recombinantly expressed signatures and closely matches the glycosylation patterns observed in the harvested virus from Calu-3 lung epithelial cells,⁶¹ with a predominant complex-type glycosylation and oligomannose character observed at Asn61, Asn234, Asn709, Asn717, and Asn801.

In comparison to N-glycosylation analysis, characterization of O-glycosylation is an even greater challenge that requires a focused approach. N-glycans vastly outnumber O-glycans on viral spike proteins, suppressing ionization of O-glycopeptides and generally requiring the use of O-glycopeptide enrichment strategies or the addition of endoglycosidases such as PNGaseF that can remove the N-glycans. Nonetheless, the HCD-triggered-UVPD approach filters stochastic precursors based on identification of unique Hex- or HexNAc-oxonium ions. The detection of these glycan-specific oxonium ions, ones created by both N- and O-linked glycopeptides, triggered acquisition of UVPD mass spectra and led to identification of O-glycosylation sites on SARS-2-S: Ser677 and The323. A bound HexNAc moiety was identified on Thr323 (Figure S16A), and two core 2 glycoforms were observed on Ser677 (Figure S16B,C), the serine adjacent to the mutated furin S1/S2 cleavage site, where RRAR was mutated to SGAG.⁷⁷ Although this recombinant SARS-2-S is not a native spike protein; the identified O-glycosylated positions align with proposed O-glycosylated sites.^42,51,78 In terms of the O-glycan structure, the fragment ions generated by UVPD infer that these glycans (Figure S16B,C) are branched based on the evidence of Y₁Y₂ glycan fragments and the complementary reducing-end glycosidic C₂ cleavage products retaining two glycosidic oxygens (Table S12). A more elaborate approach incorporating enrichment would be needed to enhance the abundances of O-glycopeptides and facilitate comprehensive analysis of cross-ring cleavages and determination of linkages.

For the HKU1-S protein overall, 97% protein sequence coverage was obtained using the HCD-triggered UVPD strategy with six unique glycoforms corresponding to 24 different N-glycosylated sites across each of the three replicates. The average peptide-level sequence coverages generated by UVPD were comparable to those from EThcD and stepped HCD, all providing approximately 80% peptide sequence coverage across all detected HKU1-S glycopeptides (77% for stepped HCD, 76% for HCD-triggered EThcD, and 77% for HCD-triggered UVPD). The annotated UVPD mass spectrum of one representative glycopeptide, SLPLVN-(+1882.6447)VTINNFNPSSW containing a HexNAc(2)-Hex(9), and the corresponding glycan map are shown in Figure S17 (and detailed in Table S13). A significant number of cross-ring cleavages occur upon UVPD, allowing characterization of all inter-saccharide linkages, and the glycan-containing sequence ions permit localization of the M9 oligomannose glycan to Asn924.

The performance of HCD-triggered UVPD was compared to that of stepped HCD and HCD-triggered EThcD based on the glycosites identified and distributions of glycan compositions (Figure 4). Although HCD-triggered UVPD identified 24 N-glycosylated sites, 20 sites were identified by HCD-triggered EThcD, and 22 sites were found using stepped HCD. The glycan subclasses determined by each method are summarized in Figures S18-S20 and Tables S14-S16. The predominant glycan of the HKU1-S protein was oligomannose with lesser amounts of complex-type glycans located primarily toward the C-terminal end.

Figure 4.

Distribution of glycan compositions at each N-linked amino acid site of HKU1-S using (A) HCD-triggered UVPD, (B) stepped HCD, and (C) HCD-triggered EThcD. Glycans are categorized as complex, hybrid, or oligomannose.

CONCLUSIONS

The ability to map and characterize glycans is of vital importance as they have been shown to confer functional viral escape from neutralizing antibodies.^79-81 Although viral mutation affecting N-linked glycans could readily be traced by virtue of the genetically encoded N-linked sequence, mutations affecting glycan composition cannot be readily tracked. We have developed UVPD-MS for the simultaneous localization of glycan sites and characterization of glycan compositions based on analysis of glycopeptides from tryptic/chymotryptic digestion of recombinant viral spike proteins. The rich fragmentation patterns produced by UVPD reveal complete glycan linkage information, important to topographical glycan analysis and are leveraged to allow confident site-specific localization of each complex N-linked glycoform regardless of the glycan subclass. The HCD-triggered-UVPD approach serves as a way to filter glycopeptides based on the presence of oxonium ions in survey HCD mass spectra and then performing UVPD only on precursors that generated glycan-specific oxonium ions. Because data-dependent MS/MS analysis is stochastic, lower-abundance glycopeptides may not be adequately sampled, a shortcoming that could be resolved in future strategies. With routine glycopeptide sample processing, the UVPD-MS approach is agnostic to the origin of the sample. Thus, an opportunity exists for in-depth glycosylation analysis of proteins in complex biological samples, cell lines, tissue, or biological fluid. Finally, although the glycoforms of the spike glycoproteins studied here do not reflect the glycosylation composition exhibited in active virions,⁸² it is a first step toward designing an MS/MS workflow that offers in-depth analysis of complex glycoproteoforms.