Glycopeptide characterization of Sf9‐derived SARS‐CoV‐2 spike protein recombinant vaccine candidates expedited by the use of glycopeptide libraries

Abstract

Rationale

We report the N‐glycosylation pattern of Sf9 insect cell‐derived recombinant spike proteins being developed as candidate vaccine antigens for SARS‐CoV‐2 (COVID‐19) (Sanofi). The method has been optimised to produce peptides with single, isolated glycosylation sites using multiple protease digests. The development and use of glycopeptide libraries from previous developmental phases allowed for faster analysis than processing datasets from individual batches from first principles.

Methods

Purified spike proteins were reduced, alkylated, and digested with proteolytic enzymes. Three different protease digests were utilised to generate peptides with isolated glycosylation sites. The glycopeptides were then analysed using a Waters Q‐TOF while using a data‐dependent acquisition mass spectrometry experiment. Glycopeptide mapping data processing and glycan classification were performed using Genedata Expressionist via a specialised workflow that used libraries of previously detected glycopeptides to greatly reduce processing time.

Results

Two different spike proteins from six manufacturers were analysed. There was a strong similarity at each site across batches and manufacturers. The majority of the glycans present were of the truncated class, although at sites N61, N234, and N717/714 high mannose structures were dominant and at N1173/1170 aglycosylation was dominant for both variant proteins. A comparison was performed on a commercially available spike protein and our results were found to be similar to those of earlier reports.

Conclusions

Our data clearly show that the overall glycosylation pattern of both spike protein variants was highly similar from batch to batch, and between materials produced at different manufacturing facilities. The use of our glycopeptide libraries greatly expedited the generation of site‐specific glycan occupancy data for a large glycoprotein. We compared our method with previously obtained data from a commercially available insect cell‐derived spike protein and the results were comparable to published findings.

1. INTRODUCTION

The addition of N‐glycans on asparagine is one of the most common and critical post‐translational modifications that a protein can undergo. The glycosylation of viral surface proteins can affect the stability, protein folding, and functional efficiency of the virus as well as shielding of the epitopes from immune recognition. ¹ , ² , ³ In order to understand the impact of glycans on the quality of biologically active proteins such as vaccine antigens, it is vital to carry out glycoproteomics or, more specifically, glycopeptide mapping, which is the site‐specific assignment and quantitation of glycans. For regulatory purposes, site occupancy is typically reported as a percent relative abundance of the aglycan and glycoform class at each site. ³

For glycosylation to occur at an asparagine residue, it must follow the sequence motif: asparagine‐X‐serine, threonine, or cysteine where X is any amino acid with the exception of proline. N‐glycosylation occurs in the endoplasmic reticulum and glycans are further processed in the Golgi apparatus of eukaryotic cells. A large glycan is added to the asparagine and subsequently broken down by glycosidases, which removes the glycoses. The remaining oligomannose or high mannose structures can persist or be further processed into truncated, hybrid, or complex structures. The addition of other saccharides by transferase enzymes, i.e. fucose, provides considerable structural heterogeneity and functionality. N‐glycan structures remain attached to the protein during processing. ⁴ , ⁵ , ⁶ , ⁷ N‐linked glycans have a core structure of GlcNAc₂Man (GlcNAc = N‐acetylglucosamine and Man = mannose). ⁸ Typically, several related but different glycans are found at a glycosylation site rather than a single glycan, ⁹ resulting in complex chromatogram/mass spectrum patterns increasing the difficulty of a glycopeptide mapping experiment.

Baculovirus–insect cell systems (BICS) have been used to produce recombinant proteins since the 1980s. ¹⁰ Spodoptera frugiperda (Sf) or fall armyworm and Trichoplusia ni or cabbage looper are the most commonly used hosts for BICS. ¹¹ Insect cells tend to produce short paucimannose structures (Man_1–3GlcNac₂Fuc_0–1). ¹² , ¹³ Insect cells are capable of producing α‐1,3 fucosylation as well as α‐1,6 fucosylation, whereas mammalian cells only produce α‐1,6 fucosylation, and therefore insect cells can produce both mono‐ and di‐fucosylated glycans. The extent of di‐fucosylation is determined by the nature of the host cell, the cell culture conditions, and the nature of the expressed glycoprotein. ¹⁴ , ¹⁵ , ¹⁶ Di‐fucosylation at the GlcNAc bound to asparagine is a potential problem for vaccines as it has been shown to elicit an allergenic response in mammals. ¹⁰ , ¹⁷ Sf9 cells cannot produce sialylated structures due to low levels of sialic acid. ¹³ The species used for viral production strongly affects what glycans can be produced as its genetics dictate what glycan processing enzymes are available. The specific cell line, media, and other external factors can also influence the glycans present on a protein. A proprietary Sf9 cell line, expressSF+®, was used to produce the spike protein for COVID‐19 candidate vaccines (Sanofi).

Glycopeptide mapping of the spike protein can provide insight into biological functions such as receptor binding, cell entry, fusion, and replication. ¹⁸ The SARS‐CoV‐2 virus uses the spike protein on the viral surface to bind to angiotensin‐converting enzyme 2 present on the surface of host cells, initiating entry into the host cell causing infection. The spike protein is a homotrimer with each monomer being composed of two subunits, S1 and S2. In total, the monomer has 22 N‐linked glycosylation sites within the two subunits. The glycoforms at each site as well as the site occupancy can be monitored to demonstrate consistency of the manufactured vaccine antigen between batches and across different manufacturing sites. ⁵

Mass spectrometry (MS) has been shown to be a viable technique to analyse glycopeptides. ¹ , ² , ⁴ , ⁹ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Glycoproteomic analyses of the SARS‐CoV‐2 spike protein have been performed using MS by several laboratories. ¹ , ² , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ These studies differ in terms of the cell line that was used to produce the spike protein analysed. In general, either a human cell line (i.e. HEK293) or an insect cell line (i.e. Sf9) was used. This is expected to cause variability in the glycopeptide mapping results; for instance, BICS can produce mono‐ or di‐fucosylated glycans whereas mammalian cells can only produce α‐1,6 mono‐fucosylated glycans but can produce larger, complex sialylated glycans that insect cells cannot. ¹⁹

Typically, glycans are reported in four different categories: high‐mannose, hybrid, complex, or truncated. High‐mannose glycans contain five or more mannose saccharides attached to two core GlcNAc; hybrid glycans contain only one GlcNAc antenna (sometimes further extended with Gal) on one of the core mannose branches and sometimes containing additional mannose on the other core mannose branch; complex glycans must have at least one GlcNAc attached to both core mannose branches which can be further extended with Gal; and truncated glycans include paucimannose (Man_1–3GlcNAc₂Fuc_0–1) and other smaller glycans consisting of one or two GlcNAc structures, but four or fewer Man in total. ⁵ All glycans that contained any type of fucosylation were classified according to the structure as if the fucosylation was not present, i.e. Man₃GlcNAc₂Fuc₂ = truncated, Man₄GlcNAc₃Fuc₁ = hybrid, or Man₃GlcNAc₄Fuc₁ = complex. Aglycosylation, where no glycans are attached to the sequon, is also reported when performing glycopeptide mapping.

Briefly, the reports of spike protein synthesized in HEK293 cell lines and Sf9 cell lines are reviewed. First, an overview of spike proteins produced in HEK293 cells is presented. Watanabe et al reported 8 sites that were primarily high mannose with the 14 other sites containing complex‐type glycans. ²⁰ The data from this publication were then analysed with a different algorithm by Klein and Zaia. ²⁴ The results from their analysis were consistent with the results of Watanabe et al. ²⁰ Wang et al analysed the S1 subunit (which only contains the first 685 amino acids and 13 of the 22 glycosylation sites) of the protein from HEK293 and BICS. ²¹ In the HEK293 spike protein, 11 of the 13 S1 subunit sites were dominated by complex glycans. The site at N616 was glycosylated 100% with high mannose. ²¹ Wang et al reported similar results, ¹ where of the 13 sites in the S1 subunit, 11 of them were dominated by complex glycans. The N657 site, however, was mostly aglycosylated with only about 15% of the site containing complex glycans. ¹ Zhang et al also reported that the S1 subunit contained mostly complex glycans. ²² Sanda et al analysed the full‐length protein and reported that most of the glycopeptide sites were fully occupied by complex glycans. However, two of the sites did not follow that pattern. At site N603, it was observed that this site was aglycosylated and at site N234, the glycans were mostly high mannose. ²⁵ , ²⁶

Second, a review of the Sf9‐produced spike protein glycosylation is presented. Sun et al reported 20 N‐linked glycosylation sites on the spike protein produced by Sf9 cells; the remaining two predicted sites, N17 and N149, were found to be fully aglycosylated. The glycan composition was predominantly paucimannose or high mannose, except for sites N74, N343, N709, N717, and N1173 which contained complex and hybrid glycans. Two sites (N603 and N657) were only glycosylated at 43% and 74%, respectively. ²³ Zhang et al reported that the glycans from the spike protein produced in insect cells contained paucimannose and fucose‐type oligosaccharides. ²¹ Wang et al analysed the S1 subunit of the protein. Eight of the 13 sites were found to contain mostly high‐mannose glycans. Sites N149, N165, N282, N616, and N657 were not dominated by high mannose. These sites, except for N149, were found to contain more complex glycans than any other type of glycan. Site N149 had hybrid glycans as the most abundant. ²¹ Wang et al analysed the full protein and reported slightly different results overall. Truncated glycans were found to be the most common form of glycans at all sites except for sites N61, N234, and N717. These sites contained mostly high‐mannose glycans. Less than 5% of the glycans were found to be complex on any of the sites. ¹ From analysing these studies, the environment and/or cell line that is used to produce the spike protein has a dramatic impact on the glycans attached to the protein. Most of the previous work was performed on peptides that were a mixture of single and two glycosylation sites.

Herein, we report the results of the glycosylation of two spike proteins that were produced in expressSF+® cells (D614 and B.1.351 variants). The results show the batch‐to‐batch comparability as well as manufacturer‐to‐manufacturer comparability for each variant. The results using this methodology on a commercially available insect cell‐derived spike protein were compared to what has been previously published. The spike protein was digested using three different conditions with a total of five different proteases: trypsin/Lys‐C, rAsp‐N, chymotrypsin, and α‐lytic protease. Our digestion method produces peptides containing a single glycosylation site for all 22 sites within the spike proteins. The glycoproteomic analysis was performed using a Waters Acquity I class UPLC coupled to a Synapt G2‐Si mass spectrometer using reversed‐phase ultrahigh‐performance liquid chromatography (UHPLC) and data‐dependent acquisition (DDA) with collision‐induced dissociation MS. Data from previous batches and method development phases were used to generate a comprehensive list, referred to as a library, of previously detected glycopeptides within Genedata Expressionist processing software. A library of glycopeptides was generated for each variant as the protein sequence is different. These library searches facilitated speedier analysis of batches and replicates than glycopeptide mapping from first principles.

2. METHODS

2.1. Materials

All reagents were purchased from Honeywell (Charlotte, NC) unless otherwise stated. For the reduction and alkylation steps dithiothreitol and iodoacetamide were purchased from ThermoFisher (Waltham, MA). RapiGest™ SF (Waters, Milford, MA) was used to resolubilise the protein after drying. Proteases trypsin/Lys‐C, rAsp‐N, and chymotrypsin were purchased from Promega (Madison, WI). α‐Lytic protease was purchased from New England Biolabs (Ipswich, MA). Formic acid, Optima LC/MS grade, was purchased from ThermoFisher (Waltham, MA). A recombinant SARS‐CoV‐2 spike protein expressed in baculovirus insect cells (S1 + S2 ECD, cat. no. 40589‐V08B1) was purchased from Sino Biological (Wayne, PA).

2.2. Digestion of spike proteins

Two different spike protein variants were analysed, D614 and B.1.351, expressed in expressSF+® cells (Sanofi). A commercially available D614 variant spike protein from Sino Biological was analysed as well. The digestion procedure was based on previously published work. ¹ , ¹⁹ , ²⁷ For each batch, six replicates were prepared, to allow duplicates of the three different digestion conditions. An amount of 20 μg of the protein was denatured by adding 2 μL of 500mM dithiothreitol solution and incubated at 60°C for 60 min. Samples were then alkylated by the addition of 4 μL of 500mM iodoacetamide solution and incubated at room temperature in the dark for 30 min. Following alkylation, 800 μL of acetone was added to each sample and placed on ice for 45 min. The samples were centrifuged at 18 000g and 4°C for 30 min. The majority of the supernatant was removed, leaving ca 50 μL. An amount of 500 μL of acetone–water (80:20) was added to the samples which were centrifuged at 18 000g and 4°C for 15 min. Approximately 500 μL of supernatant was removed and 500 μL of acetone was added followed by centrifugation at 18 000g and 4°C for 15 min. The supernatant was removed leaving behind ca 20 μL of solution. The samples were then dried in a vacuum centrifuge.

After drying, 50 μL of 0.1% RapiGest™ SF solution was added to the samples. The samples were briefly vortexed mixed, ultrasonicated for 5 min, and briefly vortexed mixed again to aid in dissolution. For the rAsp‐N/chymotrypsin digested samples, a 1:5 (w/w) ratio (protease to protein) of rAsp‐N was used and samples were incubated at 50°C for 1 h. After 1 h, a 1:5 (w/w) ratio of chymotrypsin was added, and the samples incubated at 37°C with agitation at 600 rpm overnight. For the trypsin/Lys‐C digested samples, a 1:10 (w/w) ratio of trypsin/Lys‐C was used, and the samples were incubated at 37°C with agitation at 600 rpm overnight. For the α‐lytic protease digested samples, a 1:20 (w/w) ratio of α‐lytic was used and the samples were incubated at 37°C with agitation at 600 rpm overnight.

Following the overnight incubation, the samples were acidified with 10 μL of a 10% (v/v) trifluoroacetic acid aqueous solution to quench the proteases and hydrolyse the RapiGest™ SF. The acidified samples were incubated at 37°C for 45 min and centrifuged for 10 min at 14 000g and 4°C. The supernatant was dried in a vacuum centrifuge. An amount of 25 μL of a solution of water–acetonitrile–trifluoroacetic acid–formic acid (97/3/0.05/0.2) was added and the samples were briefly vortexed mixed, ultrasonicated for 5 min, and briefly vortexed mixed again to aid dissolution. The samples were transferred to HPLC vials and used for UHPLC/MS analysis.

2.3. Mass spectrometry analysis

Liquid chromatography coupled with electrospray ionization MS/MS analysis was performed using a Waters (Milford, MA) Acquity I class UPLC connected to a Synapt G2‐Si mass spectrometer. The protein digest was separated using a Waters (Milford, MA) Acquity Premier Peptides CSH C18 (150 × 2.1 mm 1.7 μm) column. The mobile phase was 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B) using the following gradient: 3% B (0–2 min); 3–30% B (2–78 min); 30–60% B (78–80 min); 60% B (80–84 min); 60–3% B (84–85 min); 3% B (85–93 min). The flow rate was 0.300 mL/min, and 10 μL of the sample, stored at 6°C in the autosampler, was injected. The column was heated to 40°C.

All the MS data acquisition was conducted in centroid mode using DDA. The mass spectrometer collected data from retention times of 1.92 to 86 min in positive ion resolution mode (resolution ca 18 000 FWHM at 556 m/z units). The MS survey scan range was 310–2000 Da with a 0.1 s scan rate. MS/MS was activated when the intensity of an ion rose above 40 000 arbitrary units. The MS/MS scan range was 100–2000 m/z units with a maximum number of three MS/MS experiments per single MS scan. The scan rate for MS/MS was 0.4 s; only one scan per identified ion was performed per chromatographic peak. The collision energy was ramped from 10 to 40 V (low mass: 310 m/z units) up to 20 to 90 V (high mass: 2000 m/z units) depending on the mass of the ion.

2.4. Data analysis

The MS data were processed using Genedata Expressionist v. 15.0.7 (Genedata AG, Basel, Switzerland). Because each digestion method gives a different peptide cohort leading to a different chromatographic profile, each protease datafile (trypsin/Lys‐C, rAsp‐N/chymotrypsin, and α‐lytic protease) is processed individually in a separate workflow stream and the results are merged at the end of the workflow. Replicates within each protease condition are processed at the same time. Data processing was divided into three different steps and the corresponding workflows are shown in supporting information 1. The default settings for the workflows were used except where stated in supporting information 2.

The first data processing step (‘preprocessing’) was performed as follows. The raw data were first loaded into the workflow and an intensity threshold applied to remove low‐abundance noise. Chemical noise subtraction was applied to smooth the chromatogram signal and remove chemical noise by using a quantile baseline subtraction function with 70% quantile and 501 scans. After baseline subtraction, intensity thresholding was performed by clipping the data at an intensity of 20 (arbitrary value in Genedata) to remove any low‐intensity noise remaining from the noise subtraction process. Chromatographic peaks were detected, and isotope clusters were detected based on an averaging mass isotope clustering algorithm. Singleton peaks were then removed from the data. Only the 45 000 most abundant isotope clusters were kept; all other features were removed. Next, charge and adduct grouping was applied, whereby the isotope clusters were grouped together if they elute closely: within 0.12 min in the chromatogram, have a tolerance within 15 ppm m/z units, and have either (1) the same monoisotopic mass (but different charge state) or (2) correspond to a mass shift for an adduct of a lower mass feature. The MS/MS consolidation was applied to MS/MS spectra corresponding to the same MS1 isotope cluster group. MS/MS peaks were identified and extracted from the new averaged spectra and then deisotoped. The processed datafiles were then saved in Expressionist's ‘snapshot’ format for later analysis.

When ready for evaluation, preprocessed snapshot files are loaded into the second process: the library search. A glycopeptide library of unique, isolated N‐glycan‐containing peptides (Table 1) was created over the course of the protease digest method development. This glycopeptide library contains the peptide sequence specific to a given protease digest with previously observed glycans for that sequon (including the agly peptide), its calculated monoisotopic mass, and a retention time. This library is used to quickly identify known glycopeptides of interest based on previously observed monoisotopic mass and retention time. The library search based on previous glycopeptide mapping results saved considerable time on such a complex glycoprotein. The glycopeptide library development was performed using developmental phase material. The development data for each peptide sequon with all the suggested glycans were reviewed. A glycan was selected based on its consolidated MS/MS ion score produced by the glycopeptide‐aware peptide‐spectrum‐matching algorithm built into Expressionist. The consolidated MS/MS ion score is a combination of peptidic and glycan fragments detected in the MS/MS spectrum of an ion that mass matches to a potential glycan‐containing peptide. The MS/MS spectra were manually inspected for both peptide and glycan fragments including oxonium ions, glycan fragments attached to the intact peptide, and other larger glycan fragments. Once a series of glycopeptide structures were identified for a particular peptide sequon (which generally elute close to each other in a discernible elution pattern), other related glycoforms of the same peptide were identified even if they did not include a robust MS/MS score or pattern. Assigning hits that do not contain robust MS/MS scores was performed by precursor ion mass search tolerance (<10 ppm) and analysing the elution pattern of the tentative glycoform collection with the Expressionist ion map (plot of retention time [min] versus m/z units versus intensity) (Figure 1). Higher scores will be based on an MS/MS spectrum that contains more fragments of both the glycan and the associated peptide. The di‐fucosylated core would be identified by a number of different mechanisms: (1) if it produced an MS/MS spectrum its presence would be confirmed by fragments that show the position of the core fucosylation; (2) a mass match at the MS level plus retention time. Di‐fucosylated glycans are typically smaller glycans, i.e. M3F2 or M4F2, and no evidence of fucosylation on anything other than core sites was detected. Since the protein sequence was the same (differing only for the two variants), a glycopeptide library was developed for each variant and used for future batches that compared high‐resolution MS1 monoisotopic mass and retention time. This library was used for all subsequent analyses. Isotope cluster groups were searched against library items based on accurate monoisotopic m/z units (±15 ppm) and retention time window of ±0.3 min. A peptide mapping view activity was used to manually review and reject source fragments. The files were then saved as a ‘snapshot’ with their annotated features.

TABLE 1.

Peptides used for glycopeptide mapping for spike protein variants: (A) D614 variant; (B) B.1.351 variant (enzymatic digestion used is located at the top of each column; RT, retention time)

(A) D614 peptides
	α‐Lytic		Chymotrypsin/rAsp‐N		Trypsin/Lys‐C
Site	Peptide	RT	Peptide	RT	Peptide	RT
17	QCVNLT	30.8			QCVNLTTR	20.7
61			FSNVTWF	45.5
74	WFHAIHVSGTNGT	80.1	HAIHVSGTNGTKRF	5.3
122					TQSLLIVNNATNVVIK	43.6
149			YHKNNKSWMESEF	17.9
165					VYSSANNCTFEYVSQPFLMDLEGK	57
234			DLPIGINITRF	51.7	DLPQGFSALEPLVDLPIGINITR	77
282					YNENGTITDAVDCALDPLSETK	49.3
331	RFPNIT	19.4	RVQPTESIVRFPNITNLCPFGEVF	62
343	NLCPFGEVFNAT	49	NATRF	4.3
603			GGVSVITPGTNTSNQVAL	43.4
603			GGVSVITPGTNTSNQVALYQ	47.3
616	VLYQDVNCT	22.2	DVNCTEVPVAIHA	30.3
616	LYQDVNCT	15.4
657			QTRAGCLIGAEHVNNSYEC	21.6
709	YSNNSIA	9.2
709	YSNNSIAIPT	26.2
717	NFTIS	20.6
801			NFSQILP	37.4	DFGGFNFSQILPDPSKPSK	55.3
1074	QEKNFT	3			NFTTAPAICHDGK	13.2
1098					EGVFVSNGTHWFVTQR	36.7
1134			DVVIGIVNNTVY	52.8
1158			DKYFKNHTSPDV	9.8
1173			DISGINASVVNIQKEI	46.8
1194			DRLNEVAKNLNESLI	40.3	NLNESLIDLQELGK	43.8

(B) B.1.351 peptides
	α‐Lytic		Chymotrypsin/rAsp‐N		Trypsin/Lys‐C
Site	Peptide	RT	Peptide	RT	Peptide	RT
17	QCVNFTT	36.4			QCVNFTTR	23.2
61			FSNVTWF	43.3
74	GTNGTKRFANPVLPFNDGVYFA	47.5	HAIHVSGTNGTKRF	5.3
122					TQSLLIVNNATNVVIK	39.7
149			YHKNNKSWMESEF	16
165					VYSSANNCTFEYVSQPFLMDLEGK	57
234			DLPIGINITRF	52.2	GLPQGFSALEPLVDLPIGINITR	71.6
279					YNENGTITDAVDCALDPLSETK	49.2
328	RFPNIT	17	RVQPTESIVRFPNITNLCPFGEVF	61.1
340	NLCPFGEVFNAT	49.2	NATRF	3.5
600			GGVSVITPGTNTSNQVAL	43.6
613	LYQGVNCT	14	QGVNCTEVPVAIHA	27.6
654			QTRAGCLIGAEHVNNSYEC	20.3
706	YSNNSIAIPT	26.2
714	NFTIS	16.6
798			NFSQILP	37.6	TPPIKDFGGFNFSQILPDPSKPSK	50.6
798					DFGGFNFSQILPDPSKPSK	53.7
1071	QEKNFT	3			NFTTAPAICHDGK	13.3
1095					EGVFVSNGTHWFVTQR	34.6
1131	GNCDVVIGIVNNTVYDPLQPELDS	38.5	DVVIGIVNNTVY	47.6
1155			DKYFKNHTSPDV	9.8
1170			DISGINASVVNIQKEI	47.3
1191			DRLNEVAKNLNESLI	37.5	NLNESLIDLQELGK	44

FIGURE 1.

Example of complete glycan mapping for a specific spike antigen peptide, DLPIGINITRF. For (A) and (B), the x‐axis is the m/z units range, and the y‐axis is the intensity (on the left) and retention time in minutes (on the right). (A) Demonstration of complexity and grouping of identified features with the accepted glycans highlighted in red. (B) Accepted glycans for the given peptide highlighted in red with their corresponding glycan structures circled. Green circles, high‐mannose features with names above; yellow circles, hybrid glycans with names below; blue circles, truncated glycans names above; purple circle, aglycosylated; dark red circles, other charge states found from accepted features. (C) Expressionist output of the MS/MS spectra of the Man8 feature ranging from 100 to 3000 m/z units. All fragments have their charge removed and are plotted as neutral entities to ensure clarity and avoid confusion between singly and multiply charged ions. Therefore, the mass range is larger than the scanned range. (D) Expressionist output of the MS/MS spectra showing fragmentation found from the Man8 feature at the higher m/z units (1500–3000 m/z units) [Color figure can be viewed at wileyonlinelibrary.com]

During the final process, the datafiles from the library search process are loaded and then a glycopeptide mapping is performed over the library search results to catch any new glycoforms of previously identified glycopeptides (new peak detection). This was achieved by using the protein sequence, the modifications of glutamine conversion to pyroglutamic acid, deamidation, and oxidation, and a glycan library for S. frugiperda gleaned from GlyCosmos with a few additional probable sequences (supporting information 3). ²⁸ The glycopeptide mapping was viewed and the results were manually checked and accepted or rejected while looking at the retention times and the ion map for each peptide of interest and glycans on a given peptide (supporting information 4). The data were carefully analysed to avoid the acceptance of source fragments. Figure 1 shows a completed example of the glycan mapping for a given peptide along with the classification for the accepted features. The MS/MS spectra output from Expressionist shows a strong fragmentation pattern for the Man8 structure on the peptide DLPIGINITRF (Figures 1C and 1D). These fragments are all singly protonated with the glycopeptide structure shown above the corresponding m/z units. This glycopeptide search was greatly aided by the primary library search because only glycoforms eluting close to previously identified glycopeptides needed to be considered and most of the features had been annotated during the library search. In most cases, no new features were detected. After the glycopeptide search results were accepted, the results were saved in a final, reportable ‘snapshot’ file.

For reporting, the glycopeptide library file contains categorical annotations in the ‘comment’ column to indicate glycan class (high mannose, hybrid, complex, truncated, or agly) and the number of fucoses on the glycan. Up to this point, different digest conditions have been treated independently as they produce different peptides, but for reporting purposes the results are then combined in the final reporting workflow (supporting information 5). A custom Python‐based plugin (‘Glycan Report’) was used to calculate the percent relative abundance of glycan classes and number of fucoses at each sequon across all accepted glycopeptides in all three digest conditions. The Excel report generated by the plugin lists for each observed and accepted feature: the peptide sequence, glycan structure, the location of the glycan on the protein, a comment that states the glycan class, the intensity, the relative percentage of each glycan on a specific site, the percentage of each glycan class on a specific site, and the percentage of di‐fucosylated glycans on a specific site. The plugin also creates two Excel charts: the first chart displays the percentage of each type of glycan on each site for the batch and the second displays the percentage of di‐fucosylated glycan on each site (Figure 2).

FIGURE 2.

Glycopeptide mapping profile for Sino spike protein. (A) Overall glycans; (B) di‐fucosylation [Color figure can be viewed at wileyonlinelibrary.com]

3. RESULTS

To compare our method to previously reported results, ¹ , ²¹ , ²² , ²³ a purified spike protein (D614 variant) from Sino Biological that was expressed in insect cells was analysed. Wang et al reported that truncated glycans were the major glycans at most of the N‐glycosylation sites. However, at sites N17, N61, N234, and N717, this was not the case. In fact, at site N17 no glycans were reported, i.e. 0% occupancy or fully aglycosylated. For sites N61, N234, and N717, high mannose was the major glycan group. ¹ Zhang et al reported similar results. They found that the major glycan group for the whole protein was paucimannose and fucose‐type oligosaccharides. ²² Wang et al analysed the S1 region only of the spike protein that was expressed in insect cells from Sino Biological. Since only the S1 region was analysed, 12 sites were analysed for glycopeptides (up to site N603). It was reported that the major glycan groups were also paucimannose and fucose‐type oligosaccharides, where at site N17, truncated glycans were found at nearly 90% occupancy. ²¹ Sun et al expressed their own spike protein in Sf9 cells. They reported that the glycan moieties were equally distributed between the four glycan groups with glycans at 21 out of the 22 possible sites. Site N17 was observed to be aglycosylated. ²³

For these previously reported studies, reviewed here, there are some key differences in the way the experiments were carried out. Wang et al reported on peptides that contained more than one glycosylation site. ¹ Zhang et al analysed the intact protein rather than peptides. ²² Wang et al analysed only the S1 region of the spike protein. ²¹ Sun et al analysed the glycans using a released glycan method rather than a glycopeptide method as well as expressing their own protein. ²³

We analysed the same spike protein from Sino Biological as Wang et al and Zhang et al. ¹ , ²² Figure 2 shows the glycan mapping results for each site as well as the di‐fucosylation at each site. No glycans were found at site N17 in our analysis. Truncated glycans were the most dominant group found on the protein and, not surprisingly, truncated glycans were the major glycan group at all sites with the exception of N61, N234, and N717. At those three sites, high mannose was the most abundant glycans. Hybrid glycans made up less than 30% for all sites with complex glycans making up less than 1% at all sites. These results are in agreement with those of previous reports. ¹ , ²² Di‐fucosylation was also analysed for each site due to its potential to cause an allergenic response in mammals. Every site was found to have some di‐fucosylated glycans. Wang et al also reported high abundance of di‐fucosylation. At 15 of the 22 sites, a di‐fucosylated glycan was reported as one of the top two glycans. Wang et al reported the highest aglycosylation at sites N657 and N1194. ¹ However, our results show that sites N603 and N1173 had the highest aglycosylation levels of any of the sites. The differences between our results and those of Wang et al could be due to their use of peptides containing two glycosylation sites for some sites. For peptides with two glycosylation sites, the glycans would be split evenly between the sites, which could lead to the discrepancies in the results. Overall, there is a strong correlation between what has been previously reported in the literature and our results.

3.1. D614 variant batch‐to‐batch comparison from different manufacturers

There were four different manufacturers used to produce the D614 spike protein vaccine candidate antigen, referred to as Manufacturers A through D. Three batches were compared from Manufacturer A, four batches from Manufacturer B, six batches from Manufacturer C, and three batches from Manufacturer D. For the batches from Manufacturer A, truncated glycans were the most abundant for the majority of sites except at N61, N234, N717, and N1173. At N61, N234, and N717, high‐mannose glycans were found to be the majority of the glycans while at N1173 it was observed to be mostly aglycosylated with an average of 71.41 ± 1.11% (Figure 3). The largest difference between the batches was observed at site N603, namely the amount of hybrid glycan detected. The first two batches were very similar, but the third batch had a substantially higher percentage of hybrid structures. With that being said, a number of glycoforms that were found in the other batches were not identified in this batch, i.e. Man6, Man5, Man4F1, Man4F1N, Man3F2, and Man2F1. The third batch had the lowest intensity compared to the other two batches and this could explain why there is such a difference at this site. Di‐fucosylation levels in these batches were similar. Site N1158 had the highest di‐fucosylation level with an average of 2.59 ± 0.19%. All of the other sites contained less than 1% except for site N74 that had an average of 1.24 ± 0.18% (Table 2). Overall, there was good batch‐to‐batch consistency.

FIGURE 3.

Glycopeptide mapping profiles of D614 variant spike protein produced by manufacturer A: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 2.

N‐glycosylation relative site occupancy for spike D614 variant antigen batches produced at different manufacturing facilities (relative mean % occupancy ± SD)

Site	Manufacturer	Complex	High mannose	Hybrid	Truncated	Aglycosylated	Di‐fucosylated
17	A	0.02 ± 0.02	2.24 ± 0.47	1.16 ± 0.05	90.67 ± 1.00	5.92 ± 0.81	0.15 ± 0.01
	B	0.00 ± 0.01	1.44 ± 1.11	0.49 ± 0.09	92.10 ± 0.91	5.97 ± 1.11	0.36 ± 0.04
	C	0.00 ± 0.00	0.56 ± 0.52	0.31 ± 0.27	93.32 ± 4.51	5.80 ± 4.54	0.44 ± 0.27
	D	0.01 ± 0.02	0.95 ± 0.51	0.58 ± 0.24	94.22 ± 1.67	4.24 ± 0.90	0.16 ± 0.07
61	A	0.12 ± 0.08	87.81 ± 0.51	0.92 ± 0.04	11.15 ± 0.41	0.00 ± 0.00	0.00 ± 0.00
	B	0.09 ± 0.09	85.22 ± 2.02	0.62 ± 0.21	14.07 ± 1.78	0.00 ± 0.00	0.00 ± 0.00
	C	0.03 ± 0.04	84.67 ± 2.68	1.24 ± 0.23	14.07 ± 2.80	0.00 ± 0.00	0.00 ± 0.00
	D	0.175 ± 0.02	86.79 ± 3.75	1.09 ± 0.26	11.95 ± 4.02	0.00 ± 0.00	0.00 ± 0.00
74	A	0.00 ± 0.00	0.00 ± 0.00	0.28 ± 0.05	91.93 ± 0.69	7.79 ± 0.66	1.24 ± 0.18
	B	0.00 ± 0.00	0.00 ± 0.00	0.00 ± 0.00	94.11 ± 0.83	5.89 ± 0.83	0.79 ± 0.23
	C	0.00 ± 0.00	0.00 ± 0.00	0.47 ± 0.49	86.17 ± 9.13	13.36 ± 9.16	4.01 ± 2.06
	D	0.00 ± 0.00	0.00 ± 0.00	0.08 ± 0.11	96.70 ± 2.33	3.22 ± 2.28	0.67 ± 0.27
122	A	0.91 ± 0.70	20.20 ± 1.90	2.02 ± 0.16	76.81 ± 2.43	0.07 ± 0.03	0.00 ± 0.00
	B	0.36 ± 0.11	16.25 ± 2.60	1.9 ± 0.69	81.48 ± 2.53	0.01 ± 0.01	0.00 ± 0.00
	C	0.45 ± 0.28	15.87 ± 2.16	2.33 ± 0.91	81.32 ± 2.49	0.02 ± 0.03	0.00 ± 0.00
	D	0.74 ± 0.32	13.86 ± 1.47	1.91 ± 1.03	83.42 ± 2.69	0.07 ± 0.02	0.00 ± 0.00
149	A	0.00 ± 0.00	2.17 ± 0.86	0.33 ± 0.17	92.49 ± 1.17	5.02 ± 0.29	0.12 ± 0.05
	B	0.00 ± 0.00	2.18 ± 1.55	0.17 ± 0.11	92.59 ± 1.17	5.05 ± 1.38	0.15 ± 0.04
	C	0.00 ± 0.00	0.73 ± 0.43	0.87 ± 0.43	93.68 ± 1.12	4.72 ± 1.67	0.18 ± 0.04
	D	0.00 ± 0.00	1.39 ± 0.41	0.22 ± 0.06	92.03 ± 0.85	6.36 ± 1.29	0.13 ± 0.04
165	A	0.00 ± 0.00	2.828 ± 1.08	0.51 ± 0.57	96.65 ± 1.06	0.02 ± 0.03	0.10 ± 0.06
	B	0.01 ± 0.03	1.38 ± 0.86	1.04 ± 0.6	97.57 ± 0.32	0.00 ± 0.00	0.10 ± 0.01
	C	0.02 ± 0.02	0.10 ± 0.14	1.83 ± 1.03	98.06 ± 1.05	0.00 ± 0.00	0.17 ± 0.12
	D	0.00 ± 0.00	1.24 ± 0.40	1.47 ± 0.13	97.29 ± 0.52	0.00 ± 0.00	0.07 ± 010
234	A	0.39 ± 0.09	86.33 ± 1.43	7.58 ± 0.85	5.44 ± 0.79	0.27 ± 0.21	0.00 ± 0.00
	B	0.29 ± 0.19	83.99 ± 1.46	9.81 ± 1.46	5.63 ± 0.47	0.28 ± 0.16	0.00 ± 0.00
	C	0.05 ± 0.04	90.04 ± 1.60	6.62 ± 1.05	3.23 ± 0.87	0.07 ± 0.06	0.00 ± 0.00
	D	0.36 ± 0.13	86.25 ± 3.63	7.29 ± 1.17	5.91 ± 2.31	0.18 ± 0.14	0.00 ± 0.00
282	A	0.21 ± 0.14	0.29 ± 0.09	2.85 ± 0.71	95.34 ± 0.64	1.32 ± 0.17	0.15 ± 0.02
	B	0.18 ± 0.10	0.10 ± 0.13	4.71 ± 0.87	94.48 ± 0.95	0.53 ± 0.35	0.12 ± 0.06
	C	0.54 ± 0.18	0.08 ± 0.05	6.52 ± 0.93	91.60 ± 1.07	1.26 ± 0.52	0.51 ± 0.47
	D	0.46 ± 0.12	0.08 ± 0.01	3.54 ± 0.15	94.99 ± 0.10	0.93 ± 0.17	0.11 ± 0.01
331	A	0.03 ± 0.02	8.30 ± 2.46	0.24 ± 0.08	90.74 ± 2.48	0.69 ± 0.10	0.00 ± 0.00
	B	0.03 ± 0.03	6.02 ± 2.14	0.15 ± 0.07	93.62 ± 2.10	0.18 ± 0.14	0.01 ± 0.01
	C	0.01 ± 0.02	4.33 ± 1.51	0.47 ± 0.31	93.09 ± 4.01	2.09 ± 2.69	0.00 ± 0.00
	D	0.03 ± 0.01	4.41 ± 0.67	0.16 ± 0.05	95.09 ± 0.68	0.31 ± 0.07	0.00 ± 0.00
343	A	0.08 ± 0.03	23.25 ± 2.67	0.69 ± 0.04	75.87 ± 2.61	0.10 ± 0.01	0.02 ± 0.01
	B	0.06 ± 0.01	17.41 ± 2.74	0.06 ± 0.06	82.47 ± 2.68	0.01 ± 0.01	0.00 ± 0.00
	C	0.15 ± 0.09	15.95 ± 4.49	1.68 ± 1.13	82.04 ± 4.51	0.17 ± 0.25	0.02 ± 0.02
	D	0.11 ± 0.02	17.71 ± 3.84	0.44 ± 0.46	81.74 ± 4.30	0.00 ± 0.00	0.00 ± 0.00
603	A	0.32 ± 0.09	0.04 ± 0.05	30.91 ± 11.79	61.58 ± 12.33	7.16 ± 0.68	0.07 ± 0.06
	B	0.20 ± 0.22	0.00 ± 0.00	5.80 ± 1.04	58.90 ± 3.60	35.10 ± 3.47	0.16 ± 0.28
	C	0.65 ± 0.41	0.00 ± 0.00	16.64 ± 2.37	73.96 ± 6.62	8.75 ± 5.23	0.00 ± 0.00
	D	0.28 ± 0.06	0.00 ± 0.00	32.58 ± 6.98	57.94 ± 7.92	9.20 ± 1.04	0.00 ± 0.00
616	A	1.27 ± 0.23	1.45 ± 0.45	4.86 ± 0.66	92.38 ± 0.66	0.04 ± 0.02	0.18 ± 0.03
	B	1.54 ± 0.24	0.88 ± 0.53	4.01 ± 0.67	93.55 ± 0.48	0.01 ± 0.02	0.09 ± 0.01
	C	1.52 ± 0.25	0.25 ± 0.08	9.55 ± 1.53	88.00 ± 2.36	0.68 ± 1.40	0.41 ± 0.12
	D	1.23 ± 0.32	0.52 ± 0.06	4.33 ± 0.23	93.90 ± 0.03	0.02 ± 0.00	0.12 ± 0.02
657	A	0.00 ± 0.00	0.00 ± 0.00	0.72 ± 0.07	67.94 ± 1.47	31.34 ± 1.39	0.24 ± 0.02
	B	0.00 ± 0.00	0.00 ± 0.00	0.53 ± 0.03	65.97 ± 3.12	33.50 ± 3.12	0.17 ± 0.03
	C	0.00 ± 0.00	0.00 ± 0.00	1.15 ± 0.58	63.00 ± 25.25	35.85 ± 25.69	0.65 ± 0.18
	D	0.00 ± 0.00	0.00 ± 0.00	0.40 ± 0.10	63.12 ± 3.61	36.48 ± 3.71	0.14 ± 0.05
709	A	1.81 ± 0.22	7.05 ± 2.32	5.60 ± 0.54	85.53 ± 2.56	0.00 ± 0.00	0.02 ± 0.02
	B	1.02 ± 0.22	9.07 ± 2.64	7.37 ± 1.14	82.55 ± 2.59	0.00 ± 0.00	0.00 ± 0.00
	C	1.03 ± 0.21	3.91 ± 2.40	6.86 ± 1.87	8,820 ± 3.97	0.00 ± 0.00	0.05 ± 0.07
	D	1.10 ± 0.29	5.75 ± 3.01	5.32 ± 2.45	87.83 ± 5.50	0.00 ± 0.00	0.00 ± 0.00
717	A	1.26 ± 0.17	73.44 ± 0.38	6.32 ± 0.32	18.90 ± 0.86	0.08 ± 0.06	0.00 ± 0.00
	B	1.54 ± 0.31	66.13 ± 1.94	5.26 ± 0.46	27.06 ± 1.85	0.00 ± 0.00	0.01 ± 0.02
	C	0.75 ± 0.27	81.37 ± 2.67	6.30 ± 1.29	9.15 ± 3.19	2.43 ± 1.40	0.00 ± 0.00
	D	1.29 ± 0.33	67.79 ± 2.12	5.84 ± 0.90	25.08 ± 3.21	0.00 ± 0.00	0.00 ± 0.00
801	A	0.17 ± 0.05	22.01 ± 3.41	1.36 ± 0.03	76.46 ± 3.37	0.01 ± 0.00	0.00 ± 0.00
	B	0.18 ± 0.04	15.10 ± 3.48	5.14 ± 2.40	79.57 ± 1.12	0.01 ± 0.01	0.00 ± 0.00
	C	0.14 ± 0.07	22.28 ± 10.23	2.43 ± 0.44	74.85 ± 10.83	0.30 ± 0.59	0.00 ± 0.00
	D	0.16 ± 0.03	13.62 ± 2.60	1.61 ± 0.34	84.61 ± 2.97	0.00 ± 0.00	0.00 ± 0.00
1074	A	0.98 ± 0.14	17.99 ± 1.12	4.03 ± 2.00	75.91 ± 1.88	1.09 ± 0.22	0.01 ± 0.00
	B	1.38 ± 0.43	23.70 ± 3.86	3.63 ± 210	70.93 ± 3.02	0.36 ± 0.36	0.02 ± 0.01
	C	0.97 ± 0.28	18.24 ± 2.08	10.23 ± 1.35	64.99 ± 7.10	5.57 ± 6.43	0.04 ± 0.02
	D	1.23 ± 0.19	20.23 ± 3.89	5.77 ± 0.37	72.41 ± 4.30	0.35 ± 0.25	0.01 ± 0.01
1098	A	0.87 ± 0.10	2.25 ± 0.74	8.40 ± 1.07	88.27 ± 1.61	0.21 ± 0.02	0.00 ± 0.00
	B	0.93 ± 0.16	1.28 ± 0.55	7.76 ± 1.04	89.93 ± 1.04	0.09 ± 0.09	0.00 ± 0.00
	C	0.79 ± 0.14	1.58 ± 0.45	13.03 ± 2.59	84.37 ± 2.74	0.23 ± 0.08	0.00 ± 0.00
	D	0.95 ± 0.34	1.03 ± 0.31	8.11 ± 1.82	89.71 ± 2.44	0.20 ± 0.03	0.00 ± 0.00
1134	A	0.67 ± 0.12	1.41 ± 0.67	8.59 ± 1.22	87.46 ± 2.47	1.86 ± 0.75	0.47 ± 0.09
	B	1.11 ± 0.46	1.60 ± 0.95	9.42 ± 0.89	87.26 ± 1.64	0.61 ± 0.58	0.73 ± 0.23
	C	0.94 ± 0.21	0.37 ± 0.30	13.79 ± 1.65	82.27 ± 1.63	2.63 ± 0.56	0.75 ± 0.55
	D	0.87 ± 0.14	0.79 ± 0.15	8.82 ± 0.31	87.71 ± 0.45	1.81 ± 0.34	0.42 ± 0.14
1158	A	0.99 ± 0.50	0.00 ± 0.00	3.31 ± 1.87	95.70 ± 2.37	0.00 ± 0.00	2.59 ± 0.19
	B	0.77 ± 0.51	0.00 ± 0.00	4.46 ± 3.46	94.77 ± 3.43	0.00 ± 0.00	2.45 ± 0.26
	C	0.68 ± 0.52	0.00 ± 0.00	4.61 ± 2.63	94.70 ± 2.92	0.00 ± 0.00	9.85 ± 14.01
	D	0.91 ± 0.28	0.00 ± 0.00	4.11 ± 1.32	94.98 ± 1.21	0.00 ± 0.00	2.23 ± 0.32
1173	A	1.02 ± 0.49	0.00 ± 0.00	0.34 ± 0.28	27.23 ± 1.64	71.41 ± 1.11	0.29 ± 0.13
	B	0.71 ± 0.47	0.00 ± 0.00	0.81 ± 0.30	31.83 ± 4.22	66.64 ± 3.89	0.13 ± 0.08
	C	0.36 ± 0.27	0.00 ± 0.00	1.00 ± 0.23	26.11 ± 7.47	72.54 ± 7.75	0.45 ± 0.09
	D	0.68 ± 0.47	0.00 ± 0.00	0.46 ± 0.21	26.53 ± 4.93	72.33 ± 5.31	0.15 ± 0.11
1194	A	2.91 ± 2.11	1.36 ± 0.15	8.18 ± 2.16	69.61 ± 4.13	17.93 ± 3.97	0.35 ± 0.04
	B	1.49 ± 0.32	1.43 ± 0.50	10.89 ± 2.05	44.00 ± 4.78	42.19 ± 5.22	0.16 ± 0.03
	C	1.81 ± 0.58	0.86 ± 0.16	19.34 ± 1.94	61.01 ± 2.23	16.98 ± 3.55	0.51 ± 0.07
	D	1.25 ± 0.62	1.00 ± 0.06	12.63 ± 1.67	56.77 ± 5.52	28.34 ± 7.44	0.29 ± 0.04

Four batches from Manufacturer B were compared (Figure 4). Truncated glycans were observed to be the most abundant overall for each batch. At sites N61, N234, N717, and N1173, truncated glycans were not found to be the most abundant, though. At N61, N234, and N717, high‐mannose glycans made up the highest percentage. At N1173, the peptide was observed to be mostly aglycosylated with an average of 66.64 ± 3.89%. There was little variation observed between these four batches. The di‐fucosylation was limited in these batches with N1158 having the highest average percentage of 2.45 ± 0.26%. All of the other sites were found to have less than 1% di‐fucosylation (Table 2). The four Manufacturer B batches were thus consistent with one another.

FIGURE 4.

Glycopeptide mapping profiles of D614 variant spike protein produced by manufacturer B: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

Six batches from Manufacturer C were compared (Figure 5). All of the sites had truncated glycans as the most abundant glycans except for sites N61, N234, N717, and N1173. At sites N61, N234, and N717, high mannose was the most abundant glycan type. At N1173, aglycosylation was observed to have the highest percentage. Looking at Figure 5, it is clear that the first batch had many differences compared to the other batches at sites N17, N74, N603, N657, N1074, and N1173 due to a manufacturing issue. It is interesting to note that the main difference observed at those sites was an increase in the percentage of aglycosylation. The amount of di‐fucosylation was higher in Manufacturer C batches compared to previous batches from Manufacturers A and B. The last five batches had an average of 3.64 ± 2.08% at site N1158. The large standard deviation is due to di‐fucosylation not being detected in one specific batch (Table 2). Comparing the last five batches from this manufacturer, it is clear to see a strong similarity from batch‐to‐batch.

FIGURE 5.

Glycopeptide mapping profiles of D614 variant spike protein produced by manufacturer C: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

Three batches from Manufacturer D were compared (Figure 6). Truncated glycans were the most abundant type at the majority of sites except for sites N61, N234, N717, and N1173. High‐mannose glycans made up the largest percentage at sites N61, N234, and N717. Aglycosylation made up the majority of site N1173's composition (average of 72.33 ± 5.31%). Only site N1158 had over 1% of di‐fucosylation glycans with an average of 2.23 ± 0.32%. In general, all batches were highly similar. The largest difference was observed at site N603 with the truncated/hybrid glycans. These glycans had a standard deviation of 7.92% and 6.98%, respectively. This variance was due to the first batch having more hybrid glycans detected than the other two batches which in turn led to a decreased amount of truncated glycans (Table 2). These three batches displayed consistency with one another overall.

FIGURE 6.

Glycopeptide mapping profiles of D614 variant spike protein produced by manufacturer D: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

3.2. D614 manufacturer‐to‐manufacturer comparison

In addition to the batch‐to‐batch comparison for each of the four D614 manufacturers, an overall comparison of the similarities and the differences between the manufacturers was performed (Table 2). For all of the manufacturers of the D614 spike protein, the most abundant glycan type overall was truncated. Complex structures were the least abundant glycan (and were undetectable in some cases) for all of the manufacturers. Truncated glycans made up the majority of the glycans observed at all of the sites for each manufacturer with the exceptions of N61, N234, N717, and N1173. At sites N61, N234, and N717, high‐mannose structures were most abundant. All of the manufacturers' spike protein batches were found to be mostly aglycosylated at site N1173, with abundance ranging from 66.64% to 72.54%, while site N657 was observed to be more than 30% aglycosylated. For the di‐fucosylation, N1158 had the highest percentages for all of the manufactures with N74 being the next highest.

The observed differences between the manufacturers were in terms of the percentages of hybrid and aglycosylation, and these were minor. Manufacturer A was observed to have the most aglycosylation at sites N657, N1173, and N1194 with 31.34%, 71.41%, and 17.93%, respectively. At site N603, 30.91% of the glycans found were hybrid structures. The rest of the sites for Manufacturer A were below 10% for hybrid abundance. Manufacturer B had the highest percentage of aglycosylation at sites N603, N657, N1173, and N1194: 35.10%, 33.50%, 66.64%, and 42.19%, respectively. These results correspond with Manufacturer A for sites N657 and N1173, but not at sites N603 and N1194. For the hybrid structures, only N1194 had a hybrid percentage above 10% with an average of 10.89%. For Manufacturer C, the first batch that was tested was different from the other five batches, so for comparison with the other manufacturers the data from that batch were removed here for the following calculations. Manufacturer C batches had the highest percentage of aglycosylation at sites N657, N1173, and N1194. The percentages of aglycosylation for the sites were 23.70%, 69.14%, and 16.22%, respectively. Manufacturer C showed five different sites that had over 10% hybrid structures: N603, N1074, N1098, N1134, and N1198. The highest percentage was 19.96% at N1198. Manufacturer D had three sites with more than 10% aglycosylation: N657, N1173, and N1194. The percentages of aglycosylation for these sites were 36.48%, 72.33%, and 28.34%, respectively. Sites N603 and N1194 had hybrid glycans that had percentages over 10%. N603 was calculated to have 32.58% and N1194 was calculated to have 12.63%.

Overall, the N‐glycosylation profiles are highly comparable across spike protein antigen batches produced by all of the manufacturers. The sites that differ the most between the manufacturers are N603 for the hybrid/truncated/aglycosylation ratios, N717 for the truncated/high‐mannose ratios, and N1194 for the hybrid/truncated/aglycosylation ratios.

3.3. B.1.351 variant batch‐to‐batch comparison from different manufacturers

This variant has, among other sequence changes, a sequence deletion of three amino acids at 229–231, which changes the AA numbering by 3 compared to the D614 variant from site N282/279 (D614/B.1.351). B.1.351 spike proteins from two different manufacturers (referred to here as Manufacturers E and F) were analysed. A total of six batches were analysed from Manufacturer E and three batches were analysed from Manufacturer F. For Manufacturer E batches, truncated glycans constituted the majority of the glycosylation sites (Figure 7). However, at sites N61, N234, N714, and N1170, truncated structures were not the most abundant glycan type; for sites N61, N234, and N714, high‐mannose glycans made up the highest percentage. Site N1170 was observed to be mostly aglycosylated. The biggest variation between the batches was observed at site N654. The last batch that was analysed contained only one glycan which was a truncated structure; for this batch no aglycosylation was observed at this site. All of the other batches analysed had aglycosylation of over 30%. Another site that had substantial variation was site N1191. The fifth batch was calculated to be 18.35% aglycosylated, while for the other batches this ranged from 33.38% to 43.62%. There were seven sites that recorded over 10% of hybrid structures. These sites were N600, N613, N706, N1071, N1095, N1131, and N1191. Site N600 had the highest percentage of 26.73 ± 4.39%. Just like the D614 variant samples, the highest di‐fucosylation occurred at site N1155. There was an average of 3.94 ± 0.64% (Table 3). The batch‐to‐batch comparison for Manufacturer E demonstrates the strong similarity for this manufacturer for the production of this spike protein variant.

FIGURE 7.

Glycopeptide mapping profiles of B.1.351 variant spike protein produced by manufacturer E: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 3.

N‐glycosylation relative site occupancy for spike B.1.351 variant antigen batches produced at different manufacturing facilities (relative mean % occupancy ± SD)

Site	Manufacturer	Complex	High mannose	Hybrid	Truncated	Aglycosylated	Di‐fucosylated
17	E	0.00 ± 0.00	0.05 ± 0.05	0.43 ± 0.21	93.07 ± 1.69	6.44 ± 1.58	0.56 ± 0.20
	F	0.00 ± 0.00	1.40 ± 0.43	1.56 ± 0.42	91.38 ± 1.80	5.66 ± 1.11	0.75 ± 0.46
61	E	0.00 ± 0.00	86.38 ± 2.58	0.10 ± 0.06	13.52 ± 2.59	0.00 ± 0.00	0.00 ± 0.00
	F	0.00 ± 0.00	92.13 ± 0.53	0.06 ± 0.03	7.82 ± 0.52	0.00 ± 0.00	0.00 ± 0.00
74	E	0.00 ± 0.00	0.00 ± 0.00	0.22 ± 0.12	96.55 ± 2.54	3.24 ± 2.64	0.60 ± 0.20
	F	0.00 ± 0.00	0.00 ± 0.00	0.38 ± 0.06	92.04 ± 0.77	7.58 ± 0.75	0.23 ± 0.17
122	E	0.27 ± 0.28	4.96 ± 1.61	1.44 ± 0.38	93.28 ± 2.18	0.05 ± 0.08	0.00 ± 0.00
	F	0.10 ± 0.02	7.78 ± 0.79	3.27 ± 0.36	88.62 ± 0.42	0.23 ± 0.01	0.00 ± 0.00
149	E	0.00 ± 0.00	0.00 ± 0.00	0.31 ± 0.16	86.74 ± 4.60	12.94 ± 4.56	0.20 ± 0.04
	F	0.00 ± 0.00	0.09 ± 0.12	3.04 ± 0.40	89.87 ± 1.24	7.01 ± 1.16	0.29 ± 0.21
165	E	0.00 ± 0.00	5.20 ± 1.32	0.34 ± 0.61	94.46 ± 1.55	0.00 ± 0.00	0.08 ± 0.04
	F	0.00 ± 0.00	7.15 ± 3.22	0.23 ± 0.32	92.62 ± 3.07	0.00 ± 0.00	0.00 ± 0.00
234	E	0.00 ± 0.00	96.26 ± 1.04	2.10 ± 0.65	1.35 ± 0.62	0.29 ± 0.07	0.00 ± 0.00
	F	0.00 ± 0.00	98.10 ± 0.33	1.09 ± 0.26	0.59 ± 0.09	0.23 ± 0.04	0.00 ± 0.00
279	E	0.21 ± 0.09	0.30 ± 0.09	3.72 ± 0.43	95.18 ± 0.50	0.59 ± 0.25	0.06 ± 0.01
	F	0.07 ± 0.11	1.04 ± 0.28	3.80 ± 0.21	95.03 ± 0.35	0.06 ± 0.04	0.04 ± 0.01
328	E	0.01 ± 0.02	2.42 ± 0.86	0.37 ± 0.17	96.96 ± 1.02	0.24 ± 0.14	0.00 ± 0.01
	F	0.07 ± 0.01	7.00 ± 0.84	1.10 ± 0.13	91.49 ± 0.78	0.34 ± 0.14	0.02 ± 0.01
340	E	0.01 ± 0.02	16.62 ± 1.97	0.41 ± 0.43	82.96 ± 2.16	0.00 ± 0.00	0.00 ± 0.00
	F	0.00 ± 0.00	25.26 ± 2.17	1.06 ± 0.72	73.65 ± 1.59	0.03 ± 0.02	0.00 ± 0.00
600	E	0.33 ± 0.43	0.00 ± 0.00	26.73 ± 4.39	62.72 ± 6.41	10.22 ± 3.76	0.06 ± 0.13
	F	0.00 ± 0.00	0.00 ± 0.00	29.88 ± 1.58	59.99 ± 0.85	10.12 ± 1.17	0.00 ± 0.00
613	E	1.69 ± 0.35	0.69 ± 0.21	12.97 ± 1.81	84.66 ± 2.11	0.00 ± 0.00	0.32 ± 0.30
	F	1.98 ± 0.32	2.90 ± 0.28	12.92 ± 1.45	82.20 ± 1.54	0.00 ± 0.00	0.49 ± 0.29
654	E	0.00 ± 0.00	0.00 ± 0.00	0.49 ± 0.30	70.17 ± 13.49	29.34 ± 13.29	0.00 ± 0.00
	F	0.00 ± 0.00	0.00 ± 0.00	1.39 ± 0.23	75.23 ± 3.24	23.38 ± 3.47	0.00 ± 0.00
706	E	0.28 ± 0.30	11.06 ± 2.77	11.96 ± 4.44	76.71 ± 4.33	0.00 ± 0.00	0.00 ± 0.00
	F	0.49 ± 0.08	25.50 ± 1.73	11.84 ± 1.18	62.17 ± 1.05	0.00 ± 0.00	0.00 ± 0.00
714	E	1.60 ± 0.4	71.17 ± 3.88	9.29 ± 0.96	17.83 ± 4.02	0.11 ± 0.17	0.00 ± 0.00
	F	1.84 ± 0.22	80.75 ± 2.27	6.59 ± 2.59	10.81 ± 0.42	0.00 ± 0.00	0.02 ± 0.02
798	E	0.27 ± 0.18	18.60 ± 2.99	4.73 ± 0.70	76.39 ± 3.79	0.01 ± 0.01	0.00 ± 0.00
	F	0.18 ± 0.07	26.24 ± 3.13	4.51 ± 0.54	69.06 ± 3.45	0.00 ± 0.00	0.00 ± 0.00
1071	E	0.84 ± 0.17	28.07 ± 5.97	11.68 ± 1.71	59.41 ± 5.26	0.00 ± 0.00	0.00 ± 0.01
	F	0.97 ± 0.07	38.65 ± 0.63	13.57 ± 0.95	45.76 ± 2.77	1.06 ± 1.50	0.00 ± 0.00
1095	E	1.07 ± 0.25	0.84 ± 0.21	16.69 ± 1.95	81.40 ± 2.19	0.00 ± 0.00	0.00 ± 0.00
	F	1.41 ± 0.10	3.58 ± 0.13	18.03 ± 0.21	76.97 ± 0.27	0.00 ± 0.00	0.00 ± 0.00
1131	E	1.00 ± 0.13	0.19 ± 0.10	17.00 ± 1.45	81.50 ± 2.03	0.31 ± 0.62	0.62 ± 0.20
	F	0.42 ± 0.05	1.07 ± 0.19	16.14 ± 0.47	81.65 ± 0.68	0.72 ± 0.73	0.42 ± 0.15
1155	E	1.40 ± 0.31	0.00 ± 0.00	6.83 ± 4.56	91.76 ± 4.77	0.00 ± 0.00	3.94 ± 0.64
	F	1.83 ± 0.22	0.00 ± 0.00	6.14 ± 1.01	92.03 ± 1.15	0.00 ± 0.00	5.00 ± 0.95
1170	E	0.07 ± 0.06	0.00 ± 0.00	0.63 ± 0.47	22.42 ± 3.56	76.89 ± 3.65	0.44 ± 0.04
	F	0.06 ± 0.01	0.00 ± 0.00	0.98 ± 0.09	21.90 ± 1.86	77.05 ± 1.96	0.42 ± 0.03
1191	E	1.38 ± 0.47	0.38 ± 0.43	17.35 ± 3.23	46.83 ± 6.73	34.05 ± 7.69	0.37 ± 0.12
	F	1.34 ± 0.60	0.86 ± 0.40	17.67 ± 2.41	59.91 ± 3.64	20.21 ± 2.17	0.22 ± 0.08

Three batches from Manufacturer F were analysed and compared. The most abundant glycan type for these batches was also truncated (Figure 8). The only sites where truncated glycans were not the most abundant were N61, N234, N714, and N1170. High‐mannose glycans were predominant at sites N61, N234, and N714, and at site N1170 the peptide was observed to be mostly aglycosylated. At seven sites (N600, N613, N706, N1071, N1095, N1131, and N1191), hybrid structures made up more than 10% of the glycans. Variability between these three batches was minimal. The standard deviation for each glycan type was below 4% for all sites. N1155 had the highest di‐fucosylation percentage with an average of 5.00 ± 0.95% (Table 3). Overall, the three Manufacturer F batches were highly consistent.

FIGURE 8.

Glycopeptide mapping profiles of B.1.351 variant spike protein produced by manufacturer F: batch‐to‐batch comparison [Color figure can be viewed at wileyonlinelibrary.com]

3.4. B.1.351 manufacturer comparison

Spike proteins produced by the two B.1.351 manufacturers (E and F) had highly similar glycosylation profiles (Table 3). Batches from both manufacturers showed truncated glycans to be the most abundant at all sites except for sites N61, N234, N714, and N1170. Sites N61, N234, and N714 were observed to have high mannose as the most abundant glycan group, while at N1170 it was mostly aglycosylated. Complex glycans made up less than 2% of the glycans over all sites for both Manufacturers E and F spike protein batches. For the di‐fucosylated glycans, site N1155 contained the largest percentage for both Manufacturers E and F at 3.94 ± 0.64% and 5.00 ± 0.95%, respectively. The main differences between the two manufacturers were observed at sites N706 and N1191. For N706, Manufacturer F was observed to have 14% more high‐mannose structures and 14% less truncated structures compared to Manufacturer E. At site N1191, Manufacturer E was calculated to have over 13% more aglycosylation and 13% less truncated glycans. All of the other differences were close to 10% or less between the two sites.

4. CONCLUSIONS

Using glycopeptide mapping by LC/MS, we conducted a thorough characterization of the N‐glycans profiles in insect cell‐derived recombinant D614 and B.1.351 spike protein antigens for the SARS‐CoV‐2 vaccines (Sanofi). The use of five different enzymes allowed the analysis of one glycosylation site per peptide, simplifying the data analysis. Data produced from the LC/MS analysis of the digests were processed using Expressionist software. Glycopeptide libraries were developed and used to significantly shorten the analysis time. We compared our method to published results for a commercially available spike protein and reported similar results to those obtained previously. Our data for the spike proteins show that the N‐glycosylation profiles are highly similar from batch‐to‐batch, as well as between manufacturers, for both variants. The di‐fucosylation percentages reported herein for the spike proteins (Sanofi) were at much lower level than that of the commercially available spike protein. The high degree of similarity of N‐glycosylation profiles over multiple batches of the two spike protein variants is evidence of consistency of the manufacturing process at different manufacturing facilities. More broadly, our results show that glycopeptide mapping by LC/MS is a powerful analytical tool for assessing batch‐to‐batch comparability and manufacturing process consistency, for N‐glycosylated recombinant proteins produced using eukaryotic expression systems.

CONFLICT OF INTEREST

David Bush, Aude Tartiere, and Nick DeGraan‐Weber are employees of Genedata and may hold shares or stock options in the company. George Perkins was an employee of Sanofi and may hold shares or stock options in the company. Roland Miller is an employee of Sanofi and may hold shares or stock options in the company.

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1002/rcm.9452.

Supporting information

Data S1. Supporting Material 1: Genedata Expressionist Workflows: Preprocessing (left), Library Search (top right), Final process (bottom right).

Supporting Material 2: Modifications to the Genedata Expressionist Workflow That Differ from the Default Settings.

Supporting Material 3: Glycan Library

Supporting Material 4: Visualization of the Selection of Glycans While Looking at Retention Time and Ion Map.

Supporting Material 5: Combination Workflow for the Different Digests.

Miller RM, Perkins GL, Bush D, Tartiere A, DeGraan‐Weber N. Glycopeptide characterization of Sf9‐derived SARS‐CoV‐2 spike protein recombinant vaccine candidates expedited by the use of glycopeptide libraries. Rapid Commun Mass Spectrom. 2023;37(5):e9452. doi: 10.1002/rcm.9452

DATA AVAILABILITY STATEMENT

Research Data are not shared.