N-Glycosylation of Seasonal Influenza Vaccine Hemagglutinins: Implication for Potency Testing and Immune Processing

In the present study, the glycosylation patterns of the 2014-2015 influenza vaccine season standard antigens A/California/07/2009 H1N1, A/Texas/50/2012 H3N2, and B/Massachusetts/02/2012 were revealed, and the sensitivity of the single radial immunodiffusion (SRID) potency test to glycosylation was tested. Differences in hemagglutinin glycosylation site composition and heterogeneity seen in antigens produced in different cell substrates suggest differences in processing and downstream immune responses. The SRID potency test used for vaccine release is not sensitive to differences in glycosylation under standard use conditions. This work reveals important differences in vaccine antigens and may point out areas where improvements may be made concerning vaccine antigen preparation, immune processing, and testing.

ABSTRACT

Prior to each annual flu season, health authorities recommend three or four virus strains for inclusion in the annual influenza vaccine: a type A:H1N1 virus, a type A:H3N2 virus, and one or two type B viruses. Antigenic differences between strains are found in the glycosylation patterns of the major influenza virus antigen, hemagglutinin (HA). Here we examine the glycosylation patterns of seven reference antigens containing HA used in influenza vaccine potency testing. These reagents are supplied by the Center for Biologics Evaluation and Research (CBER) or the National Institute for Biological Standards and Control (NIBSC) for use in vaccine testing. Those produced in hen egg, Madin-Darby canine kidney (MDCK), and insect (Sf9) expression systems were examined. They are closely related or identical to antigens used in commercial vaccines. The reference antigens studied were used in the 2014–2015 influenza season and included A/California/07/2009 H1N1, A/Texas/50/2012 H3N2, and B/Massachusetts/02/2012. Released glycan and HA-specific glycopeptide glycosylation patterns were examined. We also examined the sensitivity of the single radial immunodiffusion (SRID) potency test to differences in HA antigen glycosylation. Based on deglycosylation studies applied using standard assay procedures, the SRID assay was not sensitive to any HA antigen glycosylation status from any cell system. Mapping of glycosites with their occupying glycan to functional regions, including antigenic sites, lectin interaction regions, and fusion domains, was performed and has implications for immune processing, immune responses, and antigenic shielding. Differences in glycosylation patterns, as dictated by the cell system used for expression, may impact these functions.

IMPORTANCE In the present study, the glycosylation patterns of the 2014-2015 influenza vaccine season standard antigens A/California/07/2009 H1N1, A/Texas/50/2012 H3N2, and B/Massachusetts/02/2012 were revealed, and the sensitivity of the single radial immunodiffusion (SRID) potency test to glycosylation was tested. Differences in hemagglutinin glycosylation site composition and heterogeneity seen in antigens produced in different cell substrates suggest differences in processing and downstream immune responses. The SRID potency test used for vaccine release is not sensitive to differences in glycosylation under standard use conditions. This work reveals important differences in vaccine antigens and may point out areas where improvements may be made concerning vaccine antigen preparation, immune processing, and testing.

INTRODUCTION

Influenza vaccines contain antigens from three or four viruses: trivalent vaccines contain A/H1N1 and A/H3N2 antigens and either an influenza B/Victoria-lineage or B/Yamagata-lineage virus antigen, while quadrivalent vaccines contain A/H1N1 and A/H3N2 antigens as well as both B-lineage antigens. Most influenza vaccines are produced from viruses cultured in embryonated eggs, but vaccines prepared from viruses grown in mammalian cells (Flucelvax) and from recombinant hemagglutinin (rHA) expressed in insect cells (Flublok) are also available.

HA is the primary antigen in the seasonal influenza vaccine. HA is a glycosylated protein on the virus surface and plays a key role during viral infection. Glycosylation of this protein plays a key role in host-pathogen interactions. N-linked glycosylation occurs at specific Asn residues within the sequon Asn-Xaa-Ser/Thr (where Xaa is any amino acid except for Pro) (1). HA N-linked glycosylation may affect virulence by modulating virus receptor binding (2), masking antigenic sites (3), and stimulating the host immune response (4). Influenza virus undergoes sequence changes that are selected under immune pressure or as host cell adaptations occur. The resultant HA amino acid changes may result in a gain or loss of glycosylation sites that alter antigenic regions (5). Therefore, from season to season, antigenic changes may be related not only to the protein sequence but also to the glycosylation patterns of the vaccine strains.

Glycosylation differs depending on the host cell used. It is not clear whether different glycosylation patterns of HAs originating from avian, mammalian, and insect cell substrates impact vaccine performance. For instance, two cell lines that produce different glycans and/or glycosite occupancies at key antigenic sites may generate nonequivalent antibody responses to circulating virus. Zost et al. (6) demonstrated that human and ferret antibodies elicited by egg-grown virus used in the 2016-2017 vaccine season, which had lost an HA glycosylation site at antigenic site B, poorly neutralized circulating H3N2 viruses. In contrast, antibodies against the rHA expressed from insect cells, retaining the glycosylation site and used in the same season, were effective, suggesting that antibodies against egg-derived strains were directed to an antigenic site that was not available in the circulating virus due to differences in glycosylation. Clearly, changes in glycosylation status of the HA antigen can affect vaccine efficacy.

How such differences impact the measured potency are not known. The single radial immunodiffusion (SRID) assay is the accepted potency assay for influenza vaccines (7, 8). The assay quantifies HA, the major vaccine antigen, by measuring the diameter of precipitant rings that form in agarose gels containing HA-specific antisera. A reference antigen, either the same antigen or one highly similar to that being tested and with known HA content, is used as the standard. Antisera used in the assay are raised in sheep against egg-derived HAs. It is not clear if the antibodies in the polyclonal antisera can discriminate between antigens that have differences in glycosylation patterns. Here we report our findings.

We used a mass spectrometry-based analytical platform to perform glycosylation characterization of HAs from reference antigens of the 2014-2015 vaccine season. The reference antigens analyzed here are those available through the Center for Biologics Evaluation and Research (CBER) and the National Institute for Biological Standards and Control (NIBSC) and include H1N1 A/California/07/2009 X-179A, H3N2 A/Texas/50/2012 X-223A, and B/Massachusetts/02/2012 wild type (WT). Reference sera, provided by the same institutions, were raised in sheep immunized with bromelain-cleaved and purified HA produced in hen eggs. A/H1N1 reference antigens derived from egg, MDCK, and Sf9 cells were analyzed. Egg- and MDCK-derived viruses were analyzed for both the A/H3N2 and type B viruses. The SRID assay was performed according to standard protocols used by regulatory agencies and manufacturers. We performed the SRID assay on fully glycosylated and deglycosylated HA to test the assay’s sensitivity to glycosylation. Here we report the glycosylation patterns present on these antigens, variances between those derived using different cell substrates, and the locations of glycans on HA and assess the impact of glycosylation on SRID potency measurements. The glycosylation patterns of HAs have an impact on HA function as well as immune processing and immune responses, and these aspects are discussed.

RESULTS

Details of our analytical strategy were published previously (9, 10). A detailed description of glycan cartoon representations used herein, including definitions of high-mannose, complex, hybrid, intermediate, and paucimannose subtypes, is provided in Fig. 1 and its legend.

FIG 1.

N-glycan subtype examples. N-glycan compositions were divided into the following five groups in this study: high-mannose, complex, hybrid, intermediate, and paucimannose types. Representative glycan cartoons are presented in Fig. 2. High-mannose glycans have the composition Man5-9GlcNAc2 (Fig. 2, m/z 1,579.90, 1,783.99, 1,988.08, and 2,192.14). Complex glycans consist of a trimannosyl core with additional residues replaced by one or more monosaccharides other than Man and typically extended by the addition of GlcNAc, GalNAc, Gal, and Fuc (Fig. 2, m/z 2,244.16, 2,418.17, and 2,693.14). Hybrid N-glycans here also consist of a trimannosyl core; one branch retains some or all of its Man residues, and only one arm is elongated, typically with GlcNAc, GalNAc, Gal, and Fuc (Fig. 2, m/z 2,029.11), in the same way as that for complex N-glycans. Intermediate N-glycans are similar to hybrid glycans, since only one arm is elongated, but it is limited to the addition of a GlcNAc. Here we define intermediate N-glycans as Man3-5GlcNAc3 structures with or without a core Fuc residue (Fig. 2, m/z 1,417.09 and 1,591.21). The distinction between intermediate and hybrid glycans is important for comparisons made between cell platforms. The latter are characteristic of insect cell glycosylation, being more abundant in insect cells than in eggs or MDCK cells. The paucimannose N-glycans consist of a trimannosyl core with no substituents on either terminal mannose residue. The core GlcNAc residue can be replaced by Fuc (Fig. 2, m/z 1,171.91 and 1,346.02) in some insect cell lines (10, 13), and an additional Fuc core substitution is possible, although this is usually not seen in proteins derived from the Sf9 cell lines used here. Green circles are Man, blue squares are GlcNAc, red triangles are Fuc, and yellow circles are Gal.

Characterization of glycosylation of influenza virus A/H1N1 HA derived from egg, MDCK, and Sf9 cells.

The N-glycans of reference antigens were released using peptide-N-glycosidase A (PNGase A) and permethylated (11). Permethylated glycan profiling by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) is highly reproducible and can be used to provide semiquantitative oligosaccharide comparisons (12). Figure 2 shows example MALDI-TOF mass spectra for H1N1 reference antigen-released glycans derived from egg, MDCK, and Sf9 cells. Major peaks are indicated. Permethylated glycan compositions detected for each antigen standard virus or recombinant HA are shown in the histograms presented in Fig. 3A (H1N1 egg, MDCK, and Sf9 systems), Fig. 5A (H3N2 egg and MDCK systems), and Fig. 6A (B-type egg and MDCK systems).

FIG 2.

N-glycan profiles of H1N1 reference antigens from eggs and MDCK and Sf9 cells. Glycans were released from proteins, permethylated, and analyzed by MALDI-TOF MS. Highly abundant peaks are annotated with possible compositions.

FIG 3.

(A) Relative abundances of permethylated N-glycans from H1N1 reference antigens. Measurement was performed in triplicate by MALDI-TOF MS. The composition of each glycan is displayed beneath the histogram. The relative abundance of each glycan is displayed as a percentage of the total abundance, with standard deviations displayed as error bars. (B) Individual subgroup glycans were summed, and the percentage of each subgroup is displayed in a pie diagram. Abbreviations: H, hexose; N, N-acetylglucosamine; dHex, deoxyhexose; NeuAc, sialic acid. High-mannose glycans are noted by the number of mannoses they have, i.e., Man9.

FIG 5.

(A) Relative abundances of permethylated N-glycans from H3N2 reference antigens. Measurement was performed in triplicate by MALDI-TOF MS. The composition of each glycan is displayed underneath the histogram, abbreviated as described in the legend to Fig. 3. The relative abundance of each glycan is displayed as a percentage of the total abundance, with standard deviations displayed as error bars. (B) Individual subgroup glycans were summed, and the percentage of each subgroup is displayed in a pie diagram.

FIG 6.

(A) Relative abundances of permethylated N-glycans from influenza B virus reference antigens. Measurement was performed in triplicate by MALDI-TOF MS. The composition of each glycan is displayed underneath the histogram, abbreviated as described in the legend to Fig. 3. The relative abundance of each glycan is displayed as a percentage of the total abundance, with standard deviations displayed as error bars. (B) Individual subgroup glycans were summed, and the percentage of each subgroup is displayed in a pie diagram.

The relative abundance of each glycoform was calculated as a fraction of the total abundance, with standard deviations derived from three analytical replicates, as shown in the histograms. The glycoforms were divided into subtype groups as described above, and the relative abundances of compositions within each of the subclasses are summed and displayed in pie charts. There were 18 glycoforms detected for the egg-derived H1N1 virus, as shown in Fig. 3A. High-mannose glycans were the majority, accounting for 62.8% of the total abundance of glycans. The abundance of complex glycans was 29.8%, and that of hybrid glycans was 7.4%. Fifteen glycoforms were detected for the MDCK cell-derived virus HA. The abundance of high-mannose glycans was 84.1%, that of complex glycans was 14.9%, and that of the hybrid was 1% of the total amount. Fifteen glycoforms were detected for the Sf9 cell-expressed recombinant HA. The glycans from insect cells are relatively small compared to those from egg or MDCK cells. Paucimannose glycans were the majority, accounting for 54.7% of the total glycans. The abundance of high-mannose glycans was 28.5%, and that of intermediate glycans was 16.8%. Glycans released from egg- and MDCK-derived antigens included not only glycans from HA but also those from neuraminidase (NA) and other glycoproteins present in these inactivated virus antigen preparations. The insect cell-derived sample was purified recombinant HA, and the profile is therefore essentially from HA. The overall glycosylation patterns of all three preparations were weighted toward high-mannose glycans, including paucimannosidic forms on the insect cell-derived antigen. It should be noted that PAGE gel shift assays were performed for all standard antigen preparations, and shifts resulting from enzymatic deglycosylation in expected viral glycosylation antigens were seen, with no shifts seen in any bands that would correlate with host cell contaminant proteins (13). Therefore, released glycans detected by permethylation analysis likely originate from viral proteins, primarily HA and NA. However, we cannot rule out that some host cell protein contamination may have contributed to some glycoform intensities.

Example nanoscale liquid chromatography-mass spectrometry (NanoLC-MS^E) spectra performed on H1N1 HA glycopeptides containing N276 are shown in Fig. 4. The green peak in each spectrum is the intact glycopeptide ion signal, which is annotated with “[M+H]^+” and the molecular weight. The peptide sequence was determined by a series of b and y ions, which are indicated by blue and red lines on the left side of the spectra. The oxonium ions at m/z 204.09, 366.14, 528.19, and 690.25 indicate the presence of HexNAc, Hex-HexNAc, Hex-Hex-HexNAc, and Hex₃-HexNAc, respectively. They can be used as diagnostic markers to assess whether the predicted glycoform composition is correct. The fragment ions on the right half of the spectra are the glycosidic bond cleavage products where the peptide remains intact. They can be used to predict the sequence and composition of the glycan moiety as shown in Fig. 4. The ion at m/z 2,789.28 is the peptide plus the innermost GlcNAc residue. Formation of this ion is often favorable, leading to a high relative intensity compared to those for other fragment ions, and it can be used to verify peptide identity. Glycan sequences for some fragment ions are annotated in the figure for clarity. All observed glycoforms and their relative abundances on each glycosylation site of the H1N1 HAs are summarized in Table 1.

FIG 4.

MS/MS spectra of glycopeptide NAGSIIISDTPVHDCN₂₇₆TTCQTPK of H1N1 HAs. The colors in the figure indicate the following: red, y ions; blue, b ions; green, y/b ions after neutral loss; pink, glycopeptides after neutral loss assigned by BiopharmaLynx; and gray, ions unassigned by BiopharmaLynx (some were assigned manually, as indicated). Monosaccharide symbols are as follows: blue squares, GlcNAc; green circles, Man; red triangles, Fuc; and yellow circles, Gal. Peptide fragments with neutral glycosyl loss are indicated. Oxonium ions at m/z 204.09, 366.14, 528.19, and 690.25 are indicated.

TABLE 1.

N-glycans detected from H1N1 reference antigens and their relative abundances

Glycosylation site	N-glycan detected for indicated system
	Egg		MDCK cells		Sf9 cells
	Composition	Relative abundance (%)	Composition	Relative abundance (%)	Composition	Relative abundance (%)
N(11)STD	Hex5HexNAc2	24	Hex6HexNAc2	65	Hex5HexNAc2	31
	Hex5HexNAc3	9	Fuc1Hex6HexNAc4	16	Fuc1Hex3HexNAc2	63
	Hex5HexNAc5	22	Fuc1Hex7HexNAc5	19	Fuc1Hex3HexNAc3	6
	Fuc1Hex5HexNAc5	16
	Fuc2Hex5HexNAc5	29
N(23)VTV	Hex5HexNAc2	2	Hex6HexNAc3	21	Hex2HexNAc2	10
	Hex5HexNAc4	16	Fuc1Hex5HexNAc4	20	Hex3HexNAc2	44
	Hex5HexNAc5	12	Fuc1Hex6HexNAc4	23	Hex7HexNAc2	6
	Hex6HexNAc3	13	Fuc2Hex5HexNAc5	36	Fuc1Hex3HexNAc2	40
	Hex7HexNAc2	3
	Fuc1Hex5HexNAc4	54
N(87)GTC	Hex5HexNAc2	26	Hex5HexNAc5	69	Hex3HexNAc2	17
	Hex5HexNAc3	16	Fuc1Hex5HexNAc4	31	Hex3HexNAc3	17
	Hex5HexNAc5	25			Hex5HexNAc2	9
	Hex6HexNAc2	20			Hex6HexNAc2	57
	Hex6HexNAc3	13
N(276)TTC	Hex5HexNAc2	23	Hex5HexNAc2	2	Hex5HexNAc2	3
	Hex5HexNAc3	3	Fuc1Hex4HexNAc4	2	Hex7HexNAc2	1
	Hex6HexNAc3	13	Fuc1Hex5HexNAc4	6	Hex8HexNAc2	3
	Hex6HexNAc4	3	Fuc1Hex6HexNAc4	5	Hex9HexNAc2	1
	Hex7HexNAc2	5	Fuc1Hex6HexNAc5	14	Fuc1Hex3HexNAc2	92
	Fuc1Hex4HexNAc3	6	Fuc2Hex5HexNAc5	53
	Fuc1Hex4HexNAc4	3	Fuc2Hex6HexNAc5	18
	Fuc1Hex5HexNAc4	26
	Fuc1Hex6HexNAc5	18
N(287)TSL	Hex5HexNAc2	17	Hex5HexNAc2	12	Hex2HexNAc2	0.4
	Hex6HexNAc2	20	Hex6HexNAc2	12	Hex3HexNAc2	15
	Hex6HexNAc3	1	Hex7HexNAc2	26	Hex3HexNAc3	8
	Hex7HexNAc2	43	Hex8HexNAc2	50	Hex7HexNAc2	26
	Hex8HexNAc2	19			Hex8HexNAc2	50
					Fuc1Hex3HexNAc2	0.6
N(481)GTY	Hex5HexNAc4	60	Fuc1Hex5HexNAc4	12	Hex3HexNAc2	6
	Fuc1Hex5HexNAc4	40	Fuc1Hex5HexNAc5	17	Fuc1Hex3HexNAc2	52
			Fuc1Hex6HexNAc5	7	Fuc1Hex3HexNAc3	42
			Fuc2Hex5HexNAc5	24
			Fuc2Hex6HexNAc5	12
			Fuc2Hex6HexNAc6	28

Assuming ∼100% glycosite occupancy, overall HA glycosylation was 34%, 28%, and 88% high mannose in egg-, MDCK cell-, and Sf9 cell-derived forms, respectively, with the balance essentially composed of complex and hybrid glycoforms. Glycoform intensities are provided in Table 1 (see Fig. 7A for relative ratios). In egg-grown H1N1 HA, biantennary glycans (without a fucosyl group) were the majority glycoforms on glycosites N11, N23, and N276. High-mannose, hybrid, and biantennary glycans were identified at site N481. High-mannose glycans were the dominant glycoforms at site N287. Site N481 had only biantennary glycans in the egg-grown H1N1 HA, whereas that of MDCK cells was larger, with up to tetra-antennary forms detected. High-mannose glycan is the dominant species at site N11, with some fucosylated complex glycoforms observed. Larger amounts of complex fucosylated glycans were observed at sites N23, N87, and N276. Site N287 is occupied by high-mannose glycans only. For the HA expressed in Sf9 cells, paucimannose glycans are the dominant species at sites N11, N23, N276, and N481; high-mannose and intermediate glycoforms were also detected. Sites N87 and N287 are modified by high-mannose glycans, with several paucimannose and intermediate forms detected. Table 1 provides the relative abundances and heterogeneity at each glycosylation site (see Fig. 7A for the locations of glycosylation sites and dominant forms at each site).

FIG 7.

Glycosylation and antigenic site mapping for hemagglutinins. (A) Most abundant glycans at each site in A/California/07/2009 H1N1 HA produced in eggs and MDCK and Sf9 cells. Shown are surface area representations of two sides of an HA monomer from PDB structure 3UBQ, which is 98% identical to California H1N1 HA. The RBS is indicated with a light blue circle. N-glycosylation sites are colored cyan. Antigenic sites are colored as follows: Sa, red; Sb, orange; H1C, yellow; Ca1, green; Ca2, pink; and Cb, blue. Numbers under glycoforms from the egg and MDCK cell lines are ratios of forms with 2 HexNAcs (high mannose) to those with 3 HexNAcs (complex and hybrid). Under the Sf9 cell line glycoforms, the ratios are percentages of sugars with 2 HexNAcs and 5 or more hexoses (high mannose), 2HexNAcs and fewer than 5 hexoses (paucimannose), and 3 or more HexNAcs (intermediate). (B) Most abundant glycans at each site in A/Texas/50/2012 H3N2 HA produced in eggs or MDCK cells. Shown are surface area representations of two sides of an HA monomer from PDB structure 4WE8, which is 98% identical to Texas H3N2 HA. The RBS is indicated with a purple circle. N-glycosylation sites are colored cyan. Antigenic sites are colored as follows: A, red; B, orange; C, yellow; D, green; and E, blue. Numbers under glycoforms are ratios of forms with 2 HexNAcs to those with 3 or more HexNAcs. (C) Most abundant glycans at each site in B/Massachusetts/02/2012 HA produced in eggs or MDCK cells. Shown are surface area representations of two sides of an HA monomer from PDB structure 4M40, which is 96% identical to B/Massachusetts/02/2012 HA. The RBS is indicated with a light blue circle. N-glycosylation sites are colored cyan. Antigenic sites are colored as follows: BA, red; BB1, orange; BB2, pink; BC, yellow; BD, green; and BE, blue. Numbers under glycoforms are ratios of forms with 2 HexNAcs to those with 3 or more HexNAcs. Figures were made with Pymol. Relevant antigenic site mapping reports used in this work are cited in the text.

Characterization of HA glycosylation of influenza virus A/H3N2 strains derived from egg and MDCK cells.

Fig. 5A shows the relative abundances of the N-glycans from the H3N2 reference antigens grown in egg and MDCK cells in this study. There were 24 glycoforms detected in the egg-derived virus-released glycans. Approximately 73% of the N-glycans were high-mannose, 14% were complex, and 13% were hybrid glycans. Twenty-three glycoforms were detected in the MDCK cell-derived virus sample. The abundance of high-mannose glycans was 70%, that of complex glycans was 20%, and that of hybrid glycans was 10% of the total amount. The complex glycans were highly branched. Fucose (Fuc) was detected on both hybrid and complex glycans, with compositions indicating that it was present in both core and antenna substitutions. Trace amounts of glycans with sulfate substitution were detected. Overall glycosylation profiles of the reference antigens show high-mannose glycan dominance with similar amounts of complex and hybrid glycans. As for H1N1 reference antigens, it should be noted that these profiles represent glycosylation primarily from HA and NA.

See Fig. 7B for ratios of glycans detected (high-mannose glycans to complex and hybrid glycans) at each glycosite in H3N2 standard antigens. Assuming nearly 100% glycosite occupancy, overall HA glycosylation was 60% and 55% high mannose in egg- and MDCK cell-derived forms, respectively, with the balance essentially composed of complex and hybrid glycoforms. Table 2 shows the glycoform intensities (see Fig. 7B for relative ratios). H3N2 HA (A/Texas/50/2012) is a highly glycosylated protein, with 12 potential glycosylation sites. The HA monomer chain consists of globular head and stem regions. For the H3N2 HA in this study, six potential glycosylation sites are located on the globular head (N63, N122, N133, N144, N165, and N246), and six are on the stem region (N8, N22, N38, N45, N285, and N483). Table 2 summarizes the glycoforms detected on each site and their relative abundances. The glycan types on each site are similar between the egg-derived HA and the MDCK cell-derived HA, with the former showing more heterogeneity. Three glycosylation sites, N165, N246, and N285, contain only high-mannose glycans. Sites N38, N45, and N133 have primarily high-mannose glycans, with some hybrid glycan forms. Sites N122 and N483 contain primarily complex glycans. The complex glycans are highly branched, with many containing Fuc at the core and some at the antennae. The range of glycoforms detected was especially diverse at site N144 of egg-derived HA, with about two-thirds complex glycans and one-third high-mannose glycans. The glycans on site N144 of MDCK cell-derived HA were exclusively complex glycans which were bi- and triantennary with Fuc residues. The glycans on egg-derived HA site N144 were 37% high mannose. Sites N8 and N22 are on the same tryptic peptide, and no proteolytic strategy was available to cleave this dual-site glycopeptide. Although the composition of the two glycans could be determined, the location of each glycan was not resolved. High-mannose and hybrid glycans are the major glycoforms on these two sites for both proteins. The egg-derived HA and MDCK-derived HA have about 28% and 37% biantennary complex glycans, respectively, on these two sites. Sites N38, N45, and N63 are on the same tryptic peptide. To differentiate these, we combined trypsin with Glu-C to cleave the peptide into three short peptides. We detected the two peptides with sites N38 and N45 but did not identify any glycopeptide with site N63. See Fig. 7B for the most abundant glycoforms detected at each site for each H3N2 HA alongside the three-dimensional (3-D) structure. The percentages of abundance of high-mannose and complex-plus-hybrid forms are shown below the glycan cartoon representations. The amount of high-mannose glycans is shown to the left of the colon, and the complex and hybrid total is shown to the right. For example, egg-derived HA glycosylation site N133 is 91% high mannose and 9% complex plus hybrid. It should be noted that the most abundant glycan does not always coincide with the most abundant subclass seen at a given site. Significant differences were seen between cell substrate HA forms. Table 2 shows relative abundances and heterogeneity at each glycosylation site. See Fig. 7B for the locations of glycosylation sites and dominant glycoforms at each site.

TABLE 2.

N-glycans detected from H3N2 reference antigens and their relative abundances

Glycosylation site	N-glycan detected for indicated system
	Egg		MDCK cells
	Composition	Relative abundance (%)	Composition	Relative abundance (%)
N(8)STA/N(22)GTI	Hex3HexNAc2/Hex3HexNAc3	2	Hex3HexNAc2/Hex3HexNAc3	2
	Hex4HexNAc2/Hex3HexNAc3	7	Hex4HexNAc2/Hex3HexNAc3	17
	Hex5HexNAc2/Hex3HexNAc3	14	Hex5HexNAc2/Hex3HexNAc3	19
	Hex6HexNAc2/Hex3HexNAc3	21	Hex6HexNAc2/Hex3HexNAc3	25
	Hex4HexNAc2/Hex4HexNAc2	17	Hex5HexNAc2/dHex1Hex3HexNAc2	14
	Hex5HexNAc2/Hex5HexNAc2	11	Hex6HexNAc2/dHex1Hex3HexNAc2	23
	Fuc1Hex4HexNAc4/Hex3HexNAc4	28
N(38)ATE	Hex4HexNAc3	17	Hex4HexNAc3	100
	Hex5HexNAc2	35
	Hex5HexNAc3	11
	Hex6HexNAc2	24
	Hex7HexNAc2	13
N(45)SSI	Hex5HexNAc2	20	Hex6HexNAc2	65
	Hex6HexNAc2	21	Hex6HexNAc3	22
	Hex6HexNAc3	36	Hex7HexNAc2	13
	Hex7HexNAc2	23
N(122)ESF	Hex5HexNAc4	15	Hex6HexNAc5	14
	Hex5HexNAc5	11	Hex7HexNAc6	12
	Hex7HexNAc3	23	Fuc1Hex6HexNAc5	24
	Hex7HexNAc4	25	Fuc1Hex7HexNAc5	15
	Hex7HexNAc6	26	Fuc1Hex7HexNAc6	20
			Fuc1Hex8HexNAc6	15
N(133)GTS	Hex5HexNAc2	16	Hex5HexNAc2	46
	Hex6HexNAc2	15	Hex6HexNAc4	5
	Hex6HexNAc3	7	Hex7HexNAc2	39
	Hex6HexNAc4	2	Hex8HexNAc2	10
	Hex7HexNAc2	53
	Hex8HexNAc2	7
N(144)NSF	Hex3HexNAc2	1	Fuc1Hex5HexNAc4	9
	Hex4HexNAc2	1	Fuc1Hex5HexNAc5	18
	Hex5HexNAc2	10	Fuc1Hex6HexNAc5	7
	Hex5HexNAc3	1	Fuc2Hex5HexNAc5	36
	Hex5HexNAc4	1	Fuc2Hex6HexNAc5	30
	Hex5HexNAc5	2
	Hex6HexNAc2	4
	Hex6HexNAc3	8
	Hex6HexNAc4	1
	Hex8HexNAc2	13
	Hex9HexNAc2	8
	Fuc1Hex3HexNAc4	1
	Fuc1Hex4HexNAc3	1
	Fuc1Hex4HexNAc4	1
	Fuc1Hex5HexNAc3	1
	Fuc1Hex5HexNAc4	4
	Fuc1Hex5HexNAc5	41
	Fuc1Hex6HexNAc4	1
N(165)VTM	Hex5HexNAc2	1	Hex6HexNAc2	10
	Hex6HexNAc2	6	Hex7HexNAc2	21
	Hex7HexNAc2	19	Hex8HexNAc2	50
	Hex8HexNAc2	71	Hex9HexNAc2	19
	Hex9HexNAc2	3
N(246)STG	Hex7HexNAc2	3	Hex8HexNAc2	12
	Hex8HexNAc2	12	Hex9HexNAc2	88
	Hex9HexNAc2	85
N(285)GSI	Hex6HexNAc2	8	Hex6HexNAc2	8
	Hex7HexNAc2	43	Hex7HexNAc2	28
	Hex8HexNAc2	20	Hex8HexNAc2	29
	Hex9HexNAc2	29	Hex9HexNAc2	35
N(483)GTY	Hex5HexNAc2	2	Fuc1Hex5HexNAc4	9
	Hex5HexNAc4	3	Fuc1Hex5HexNAc5	3
	Hex6HexNAc3	4	Fuc1Hex6HexNAc4	2
	Fuc1Hex4HexNAc3	7	Fuc1Hex6HexNAc5	13
	Fuc1Hex4HexNAc4	2	Fuc1Hex6HexNAc6	17
	Fuc1Hex5HexNAc4	36	Fuc1Hex7HexNAc5	5
	Fuc1Hex5HexNAc5	12	Fuc1Hex7HexNAc6	1
	Fuc1Hex6HexNAc3	4	Fuc2Hex5HexNAc5	5
	Fuc1Hex6HexNAc5	7	Fuc2Hex6HexNAc5	18
	Fuc2Hex5HexNAc4	19	Fuc2Hex6HexNAc6	27
	Fuc2Hex5HexNAc5	4

Characterization of HA glycosylation of influenza B virus strains derived from egg and MDCK cells.

Figure 6A shows the relative abundances of the N-glycans from the influenza B virus reference antigen grown in egg and MDCK cells. There were 13 glycoforms detected from the egg-derived virus. High-mannose glycans were the majority, accounting for 85% of the total glycan abundance. The complex, paucimannose, and hybrid glycans comprised 10%, 4%, and 1% of the total, respectively. Twenty-five glycoforms were detected from the MDCK cell-derived virus HA. The abundances of high-mannose, complex, hybrid, and paucimannose glycans were 67%, 22%, 8%, and 3% of the total abundance detected (see Fig. 7C). The glycoforms of MDCK cell-derived antigen were more diverse: the complex glycans were highly branched, with Fuc substitutions at the core and antennae. These profiles represent glycans released primarily from HA and NA as well as trace contaminants and provide an overall representation of the glycosylation status of the reference antigen preparations.

Based on LC-MS^E analysis, overall HA glycosylation was 66% and 53% high mannose in egg- and MDCK cell-derived forms, respectively, with the balance essentially composed of complex and hybrid glycoforms. Table 3 shows glycoform intensities (see Fig. 7C for relative ratios). There are 10 potential glycosylation sites on HA (B/Massachusetts/02/2012). Seven glycosylation sites are outside the globular head, among which two are on the transmembrane domain and one is on the cytoplasmic tail. All seven of the ectodomain sites are predicted to be glycosylated. Indeed, all the predicted occupied sites are glycosylated. Table 3 summarizes the glycoforms detected and their relative abundances on each site. The glycan types on each site are similar between the egg-derived HA and the MDCK cell-derived HA, with latter showing more heterogeneity. Three glycosylation sites, N144, N166, and N331, contain only high-mannose glycans, with some paucimannose observed. Sites N24 and N302 have various types of glycans, including high-mannose, hybrid, and complex glycans. The complex glycans of these sites on egg-derived HA are bi- and triantennary, and some contain fucose. The glycans on these sites from MDCK cell-derived HA were highly branched, with many containing fucose at the core and some at the antennae. The glycoforms detected on site N58 were complex bi- and triantennary glycans with fucose residues. Only high-mannose glycans were detected on site N490 from egg-derived HA, whereas that from MDCK cells had about 58% high-mannose, 31% complex biantennary, and 11% hybrid glycans at this site. Table 3 shows relative abundances and heterogeneity at each glycosylation site. See Fig. 7C shows the locations of glycosylation sites and dominant forms at each site.

TABLE 3.

N-glycans detected from influenza B virus reference antigens and their relative abundances

Glycosylation site	N-glycan detected for indicated system
	Egg		MDCK cells
	Composition	Relative abundance (%)	Composition	Relative abundance (%)
N(24)VTG	Hex4HexNAc3	2	Hex6HexNAc5	15
	Hex5HexNAc4	50	Hex9HexNAc2	12
	Hex5HexNAc5	21	Fuc1Hex6HexNAc5	70
	Hex8HexNAc2	8	Fuc1Hex7HexNAc6	2
	Fuc1Hex5HexNAc4	19	Fuc2Hex6HexNAc6	1
N(58)CTD	Fuc1Hex5HexNAc4	65	Fuc1Hex4HexNAc4	2
	Fuc1Hex6HexNAc5	35	Fuc1Hex5HexNAc4	4
			Fuc1Hex5HexNAc5	23
			Fuc1Hex6HexNAc4	2
			Fuc1Hex6HexNAc5	8
			Fuc2Hex5HexNAc5	33
			Fuc2Hex6HexNAc5	28
N(144)ATS	Hex5HexNAc2	49	Hex5HexNAc2	32
	Hex7HexNAc2	47	Hex6HexNAc2	68
	Hex8HexNAc2	4
N(166)ATN	Hex5HexNAc2	12	Hex3HexNAc2	9
	Hex6HexNAc2	10	Hex4HexNAc2	25
	Hex7HexNAc2	38	Hex5HexNAc2	31
	Hex8HexNAc2	40	Hex6HexNAc2	16
			Hex7HexNAc2	19
N(302)KSK	Hex5HexNAc4	9	Hex4HexNAc4	3
	Hex6HexNAc2	45	Hex4HexNAc2	2
	Hex7HexNAc2	6	Hex5HexNAc3	1
	Hex7HexNAc6	37	Hex5HexNAc6	1
	Fuc1Hex5HexNAc4	3	Hex6HexNAc3	2
			Fuc1Hex4HexNAc4	10
			Fuc1Hex5HexNAc4	6
			Fuc1Hex5HexNAc5	23
			Fuc1Hex6HexNAc5	10
			Fuc2Hex5HexNAc5	32
			Fuc2Hex6HexNAc5	10
N(331)GTK	Hex5HexNAc2	14	Hex4HexNAc2	44
	Hex6HexNAc2	17	Hex6HexNAc2	56
	Hex7HexNAc2	53
	Hex8HexNAc2	16
N(490)QTC	Hex5HexNAc2	20	Hex4HexNAc2	11
	Hex6HexNAc2	25	Hex5HexNAc2	17
	Hex7HexNAc2	55	Hex5HexNAc4	21
			Hex6HexNAc2	21
			Hex6HexNAc3	11
			Hex6HexNAc4	10
			Hex7HexNAc2	9

The glycan compositions detected for the HAs studied using either MALDI-TOF MS or NanoLC-MS^E glycopeptide analyses are in good agreement in terms of class distributions and compositions. Complete agreement cannot be expected due to differences in techniques and the physicochemical properties of the analytes. Additionally, MALDI-TOF MS was performed on total antigen, where egg- and MDCK cell-derived samples were whole inactivated virus, whereas NanoLC-MS^E exclusively targets HA. Therefore, MALDI-TOF MS provides information related to overall virus glycosylation for egg and MDCK cells. All N-glycans detected on the HAs and overall viruses are consistent with the known N-glycan processing pathways in each cell type (14, 15).

Mapping glycosylation sites to the HA three-dimensional surface.

Figure 7 shows the 3-D structures of the HA monomers: A/Texas/50/2012 (H3N2) (Fig. 7A), A/California/07/2009 (H1N1) (Fig. 7B), and B/Massachusetts/02/2012 (Fig. 7C). The H1N1 HA sequence was aligned with A/California/04/2009 HA, which is 98% identical to A/California HA. The X-ray crystal structure of A/California/04/2009 HA was reported previously (Protein Data Bank [PDB] number 3UBQ) (16). The six H1N1 HA antigenic sites are indicated as follows: Sa, Sb, Ca1, Ca2, and Cb are on the head, and H1C is on the stem (Fig. 7B). Site Sa, in red, includes residues 128, 129, 158, 160, and 162 to 167. Site Sb comprises residues 192, 193, and 196 of a region of α-helix, 198, 156, and 159, and these residues are along the upper edge of the receptor binding site (blue circle). Site Ca1, in green, contains residues 169, 173, 207, and 240. Site Ca2, in pink, includes two loops that contain residues 140, 143, and 145 and residues 224 and 225. Site Cb, in blue, defines a region near the bottom of the globular head that includes residues 78 to 83 and 122. Site H1C, in yellow, includes residues 86, 272 to 277, 279 to 282, and 286 (17–19). The glycosylation site N87 (in cyan) is not located near any antigenic site on the head. The glycosylation site N276 is near antigenic site H1C. The most abundant glycoforms detected at each site for each H1N1 HA are shown alongside the 3-D structure in Fig. 7A. All other glycosylation sites are outside major antigenic sites.

The H3N2 HA sequence was aligned with A/Victoria/361/2011 HA, which is 98% identical to the A/Texas HA. The X-ray crystal structure of A/Victoria/361/2011 was reported previously (PDB number 4WE8) (20). There are five antigenic sites on the H3N2 HA surface: A, B, D, and E are on the globular head region, and C is on the stem region. These antigenic sites on the protein surface are indicated with various colors (Fig. 7A). Site A, shown in red, is a protruding loop from amino acids 140 to 146 and surrounding residues 122, 126, and 133 to 139. Site B, shown in orange, comprises the external residues 187 to 196 of an α-helix and adjacent residues 155 to 160 and 163 to 165, which are along the upper edge of a pocket of conserved residues of the host receptor binding site. Site C, shown in yellow, is a bulge in the tertiary structure at the disulfide bond between Cys52 and Cys277. Site D, shown in green, is in the interface region between subunits in the HA trimer and includes residues 174, 182, 201 to 220, 226, and 242 to 248. Site E, shown in blue, is from residues 63 to 83 (21–23). The receptor binding site (RBS) is indicated with a blue circle. The glycosylation sites (in cyan) are all located in or around the antigenic sites: N133 and N144 are within site A, N122 is at the edge of sites A and E, N165 is at the edge of sites B and D, N45 and N285 are at site C, and N246 is within site D (21, 23, 24).

The type B HA sequence was aligned with B/Yamanashi/166/1998 HA, which is 98% identical to the B/Massachusetts HA. The X-ray crystal structure of B/Yamanashi/166/1998 HA was reported previously (PDB number 4M40) (25). There are six antigenic sites on the B/Massachusetts HA surface: BA, BB1, BB2, BD, and BE are on the head, and BC is on the stem (Fig. 7C). Site BA, in red, contains residues 135, 141, and 145 to 149. Site BB1, in orange, includes residues 188, 190, and 194 to 200. Site BB2, in pink, contains residues 128, 130, 158 to 160, and 162 to 165. Site BC, in yellow, contains residues 47, 48, 80, 81, 116, 276, and 281. Site BD, in green, includes residues 171, 172, 179 to 181, and 217 to 219. Site BE, in blue, contains residues 56, 58, 62, 67 to 69, and 79 (18, 26). The glycosylation site N144 is in antigenic site BA, N166 is in antigenic site BB2, and N58 is in antigenic site BE. The most abundant glycoforms detected at each site for each type B HA are shown alongside the 3-D structure in Fig. 7C.

SRID analysis of intact and deglycosylated reference antigens.

The SRID assay is routinely used by manufacturers and regulatory agencies to measure the potency of commercial influenza vaccines. It is used to quantitate the amount of HA in vaccine lots at drug product release and stability testing. A predetermined amount of polyclonal antiserum is plated on an agarose gel punched with wells. The antigen is applied in a dilution series into the wells. Antigen diffuses into the gel and interacts with the antiserum, resulting in precipitation of antigen-antibody complexes when a critical concentration of antigen is reached. This is visualized as precipitin rings. The ring diameters are measured, and the amount of antigen is calculated based on a reference standard with a known HA concentration that is highly similar to or is the same antigen as that being tested. We tested the sensitivity of this assay to glycosylation for antigens in cell substrate-matched samples differing only by incubation with or without use of an endoglycosidase (Endo) F1, F2, and F3 cocktail. The SRID assay used antisera produced from sheep immunized with egg-derived HA. The sheep were immunized with bromelain-cleaved HA. The protocol used is essentially that used by regulatory agencies and manufacturers.

The H1N1 A/California/07/2009 (X-179A) egg-derived, A/California/07/2009 (H1N1) rHA insect cell-derived, and California-like A/Brisbane/10/2010 MDCK cell-derived antigens were analyzed. The endoglycosidase F1, F2, and F3 enzymes used are less sensitive to protein conformation than PNGase F. Since retention of protein antigenic structure with removal of glycosylation only was the goal, we chose to use this enzyme combination. Enzyme specificities are as follows: for Endo F1, high-mannose and hybrid N-glycans; for Endo F2, complex N-glycans; and for Endo F3, highly branched complex N-glycans. Results were monitored by SDS-PAGE. Deglycosylation approached 70% to 75%, based on gel shifts and band intensities. Incubations were performed with and without the enzyme cocktail at 37°C for 1 h.

The values and slopes for H1N1 antigens that had been deglycosylated and held under the same conditions in the absence of enzyme were highly similar and essentially superimposable, as shown in Fig. 8A to C. There was no significant difference. Egg- and MDCK cell-derived antigens lost some radial diameter due to incubation conditions (∼10%), based on comparisons to control antigen not subjected to incubation conditions. These results indicate that polyclonal sheep sera raised against H1N1 antigen are not sensitive to differences in glycosylation. The H3N2 antigen, with or without enzyme, behaved similarly, with nearly overlapping slopes and values (Fig. 8D and E). The egg-derived HAs lost about 12% of radial diameter due to incubation conditions, whereas the HA of the MDCK cell-cultured virus preparation was not significantly affected. Again, the amount of HA measured was not affected by removal of ∼70 to 75% of glycans. Type B antigen, with or without enzyme, also behaved similarly (Fig. 8F and G). Egg-derived antigen lost approximately 12% radial diameter, whereas MDCK cell-derived antigen was essentially stable under incubation conditions. Again, the SRID assay was not sensitive to removal of glycans, showing that antibodies generated in response to egg-grown HA bind to similar degrees to glycosylated and deglycosylated HAs. Overall, loss of glycosylation did not impact SRID measurement. These data demonstrate that the standard SRID assay used in potency testing is not sensitive to differences in glycosylation.

FIG 8.

SRID analysis of reference antigens. Reference antigens were treated with and without an Endo F1, F2, and F3 endoglycosidase cocktail. Resultant antigens were compared to untreated control antigen. Blue, control antigen; orange, incubation without enzyme; gray, incubation with enzyme cocktail. The reference antigen is shown at the top of each graph. Dilutions are shown on the x axis. Immunoprecipitation ring diameters are shown on the y axis.

DISCUSSION

In this work, we report the glycosylation patterns of influenza virus reference antigens produced in egg, MDCK, and Sf9 cell substrates. The antigens studied include A/California/07/2009 (H1N1) antigen produced in eggs and in MDCK and Sf9 cells, A/Texas/50/2012 (H3N2) antigen produced in eggs and in MDCK cells, and B/Massachusetts/02/2012 antigen produced in eggs and in MDCK cells. The samples derived from egg and MDCK cells were inactivated whole virus, while that produced in Sf9 cells was purified recombinant HA antigen. Released glycan and glycopeptide analyses were performed for all. Representative glycans and subtype ratios were mapped to the three-dimensional surfaces of the HA monomers. Many glycosylation sites were located close to established antigenic sites. We also report on the impact that differential glycosylation has on the SRID potency test used in influenza vaccine manufacture. The results have implications for immune system processing, immune responses, and antigenic masking. Our findings concerning glycosylation patterns are discussed in relation to these implications.

Strikingly, all released glycan profiles consisted predominantly of mannosyl glycans. Complex glycans were most abundant in the MDCK cell-derived H3N2 (Fig. 5) and type B (Fig. 6) viruses. H1N1 egg-derived virus contained twice the complex glycans as those from MDCK cells (Fig. 3 and 4). As expected, the Sf9 cell-derived H1N1 glycans were predominantly paucimannose (27). The distributions of glycan subtypes are dependent upon both the virus and the cell substrate. The predominance of mannosyl glycans may have implications for immune processing and immune surveillance. Mannose-binding lectins, such as DC-SIGN and lung surfactant SP-D, are known to facilitate influenza virus antigen uptake, and the implications of this are discussed here.

The overall H1N1 HA glycosylation was 34%, 28%, and 88% mannosyl glycan in egg-, MDCK cell-, and Sf9 cell-derived forms, respectively (Fig. 7A and Table 1). The H1N1-derived HA head region, containing only glycosite N87, was not located in an antigenic site and was differentially glycosylated according to cell substrate. Therefore, glycan shielding of the site should not be significant. In this region, the egg-derived HA glycans were an equal mixture of high-mannose and complex/hybrid glycan types, whereas MDCK cell-derived HA glycans were all complex, and Sf9 cell-derived HA glycans were predominantly mannosyl, with a minority of intermediate glycans. This observation may be significant, since this region interacts with lung surfactant SP-D, which binds mannosyl glycans (28). Infection with virus containing the complex glycan subtype at this position would likely be a poor substrate for SP-D (29). Referring to egg- and MDCK cell-derived HAs, and moving away from the head region, site N276 had predominantly complex glycans for egg- and MDCK cell-derived HAs and paucimannose for Sf9 cell-produced rHA. Further down the stem, site N287 had predominantly high-mannose glycans. Sites N11, N23, and N481, near the bottom of the stem, were occupied by increasingly complex/hybrid-type glycans. These sites are important in membrane fusion, as shown previously in studies where stem glycosite ablation negatively affected fusion (30). However, molecular ablation of the glycosylation sites of at least N11 and N23 may not be the best approach, since these sites are required for protein folding and editing control (30, 31). A better approach may be to use HAs derived from systems where glycosylation can be manipulated in these regions. This study defines glycosylation patterns in this region in three systems, which can allow informed targeting of region-specific glycosylation patterns. The antigenic sites of the H1N1 HA head region were not significantly shielded by glycans, as no glycosylation site was present within the predicted antigenic regions.

The H3N2 HAs were highly glycosylated, with 11 glycosites occupied. Overall, the egg- and MDCK cell-derived HAs were 60% and 55% high mannose, respectively, with the balance present as complex and hybrid glycoforms (Fig. 7B and Table 2). The distributions of glycoforms at all glycosites were highly similar in these two HAs, although fucosylation was more abundant in the MDCK cell-derived form. Glycans at head region sites N133, N165, and N246 were essentially all high mannose. These three sites have been reported by us and others to facilitate SP-D activity during infection (9, 32–36). Head region sites N122 and N144, located in antigenic site A, were decorated with complex glycans. The most highly abundant forms at these sites tended to be highly branched, possibly providing a high degree of shielding. Stem sites N285 and N45 were mannosyl dominant. Site N38 differed between the two substrates, as the egg-derived antigen contained predominantly high mannose and the MDCK cell-derived antigen contained essentially all complex glycans. However, the glycosite is not located in a major antigenic site, so any possible impact of glycosylation differences may be minimal, although interactions with host lectins may differ. At the base of the stem, N8 and N22 had mixed glycoforms and N483 had essentially complex glycans, with a majority of highly branched forms. Glycosylation at these sites is required for proper protein folding and may be required for membrane fusion (37). Overall, the whole virus A/Texas/50/2012 H3N2 reference antigen (inactivated whole virus) was highly mannosylated. The egg- and MDCK cell-derived HAs were also highly mannosylated, in contrast to A/California/07/2009 H1N1 HA. The head region is decorated with both high-mannose and large, complex glycans, making it likely to be well shielded. All antigenic sites, except for site B, contained glycosites within them or in close proximity. Based on the abundance of high-mannose glycans on the head region, this HA is likely a good receptor for surfactant SP-D in cases of natural infection, if indeed these cell substrates mimic the natural glycosylation state. It may also be a target of DC-SIGN, which can modulate the immune response, as discussed later.

The B/Massachusetts/02/2012 HAs carried 66% and 53% high-mannose glycans for egg- and MDCK cell-derived viruses, respectively (Fig. 7C and Table 3). Seven glycosites were occupied. Head sites N144 and N166 were occupied by high-mannose glycans and N58 by complex glycans. Moving toward the stem, N302 glycosylation differed between the two cell substrates. Egg-derived glycoforms were half mannose and half complex/hybrid, while MDCK cell-derived glycans were almost exclusively complex glycans. Site N331 contained exclusively high-mannose glycans in both HAs. Sites N24 and N490 differed significantly between the two HAs. While N24 was predominantly complex and hybrid glycans in both, MDCK cell-derived glycans tended to be larger and more highly branched at this site. However, this glycosite is not within a major antigenic site. N490 glycans, at the foot of the stem, contained much more complex glycans in MDCK cell-derived HAs. As for other HAs, membrane fusion and protein folding may be affected by stem glycosylation, and therefore differences in this region may be consequential for infection. Overall, B/Massachusetts/02/2012 HAs were predominantly mannosylated. The MDCK cell-derived HA contained larger, more highly branched forms of complex glycans. The head region contained two highly mannosylated sites, suggesting that it may be a good ligand for SP-D during infections with this and related strains. Influenza B virus strains have been associated with SP-D activity (35). The HA is likely to be moderately shielded specifically at sites BA, BB2, and BE, based on placement of glycosites relative to antigenic sites (Fig. 7C). Shielding would not be predicted to differ significantly between egg- and MDCK cell-derived HAs, since glycan sizes and compositions within antigenic regions were similar.

The contribution of glycosylation to antigenic differences between HAs is well documented (9, 33, 38, 39). As shown here and elsewhere, the cell system used for antigen production can have a significant impact on glycosylation (10). Furthermore, loss or gain of glycosylation and other mutations can occur during virus adaptation to the cell system (40). The question of whether the differences in glycosylation of the major antigen, HA, affect SRID potency results is relevant. To address the question of whether standard reagent sheep antiserum generated against HA purified from egg-grown influenza virus is impacted by glycosylation status, antigens in this study were carefully deglycosylated with endoglycosidases F1, F2, and F3, which are active on proteins in their native conformation. In all cases, the antigenicity, as detected by the SRID potency assay, was not sensitive to glycosylation status. Removal of an estimated 70% to 75% of glycans from the HAs did not significantly alter the magnitude or slope of the SRID responses. Future studies will be conducted to determine whether antisera generated against HA that is not glycosylated react equally well with glycosylated and nonglycosylated forms of HA in SRID assays. Given that several glycosylation sites shield important antigenic sites, it is possible that glycosylation might impede the reactivity of some antibodies generated by the naked HA.

Testing that is sensitive to glycosylation may provide an advantage. During the 2016-2017 influenza season, loss of a glycosylation site at T160 in egg-adapted H3N2 virus (A/Colorado/15/2014) resulted in a T160K mutation (6). This change altered antigenic site B and reduced neutralization titers in ferrets. Comparison of vaccine recipient sera supported the notion of antigenic mismatch in the adapted strain compared to circulating H3N2 viruses.

The observation that the whole-virus and subunit reference antigens studied here were predominantly decorated with high-mannose glycans may be significant. A recent study by de Vries et al. reported that highly mannosylated recombinant H5N7 HA produced lower antibody titers in mice than those obtained with HA glycosylated with complex glycans (41). The highly mannosylated HAs bound more strongly to recombinant DC-SIGN, suggesting a role for the lectin in modulating the immune response. Interactions between influenza virus and DC-SIGN, a mannose- and fucose-specific lectin, mediate viral entry into dendritic cells (DCs) (42–44) and facilitate major histocompatibility complex (MHC) class II processing, presentation, and modulation of the balance between Th1 and Th2 responses (45). DC-SIGN was shown to facilitate H3N2 and H1N1 infections of CHO cells expressing DC-SIGN (43). In the same study, virus with low mannosylation did not efficiently infect cells, whereas highly mannosylated virus did. A recent study reported induction of higher IgG and IgA responses from intranasal mannan-adjuvanted H1N1 vaccination than those induced by a vaccine conjugated with mannan or virus introduced alone (46). Both IgG2a and IgG1 were induced by H1N1 virus plus mannan, suggesting the induction of cellular as well as humoral immunity. If these observations extend more generally to influenza, tuning of antigen-linked glycosylation, free mannan, or another carbohydrate adjuvant and matching to the vaccine inoculation route may allow for a more optimized immune response.

Clearly, influenza virus utilizes glycosylation as part of its defensive strategy against host immunity, as evidenced by the gain of glycosites in antigenic regions as influenza viruses propagate through human populations. Some studies demonstrate that deglycosylated or minimally glycosylated antigen is more broadly protective. For example, murine models have shown that reduced-glycosylation influenza virus strains can generate protective antibody that is superior in cross-clade protection to that induced by highly glycosylated counterparts (47). Likewise, in humans, the H1N1 strain in use in vaccines before 2009 was highly glycosylated and showed poor vaccine efficacy during the 2009 H1N1 pandemic, in which the newly emergent strains were less glycosylated (4, 33, 48). Vaccine effectiveness was restored only when the low-glycosylation, antigenically exposed A/California strain was added to the vaccine (49).

A review by Rudd et al. reported that glycopeptide antigen processing may be different depending on the route of entry (50). Live attenuated vaccine may enter through invasion, be deglycosylated by cytosolic N-glycanase, and be subjected to proteasome processing prior to MHC-I presentation (51). This was demonstrated for the HLA-A2-restricted tyrosinase antigen on melanoma cells, where the peptide YMNGTMSQV was converted to YMDGTMSQV (52). In this case, the glycosidase and proteolysis routes are efficient. Alternatively, a killed virus would enter the endocytic pathway prior to MHC-II presentation. This pathway is populated by less efficient glycosidases and proteolytic enzymes. Glycosylation protects cleavage sites. Endopeptidases include cathepsins D, L, and S, and exopeptidases include cathepsins A, B, and H (53). The resultant antigen presented may retain more of the glycosylation, which may influence the immune response. Differential glycosylation may affect the eventual presented antigen (54).

Glycosylated antigens may be processed through specific routes. A T-cell repertoire exists that can specifically recognize glycopeptides complexed with both MHC-I and MHC-II molecules (55). Support for direct recognition of glycans by T-cell receptors (TCRs) of CD8⁺ cells came from crystal structures of MHC-I (H-2Db) molecules complexed with glycopeptides derived from an immunodominant epitope of the Sendai virus nucleoprotein. In that study, the GlcNAcβ1-O-Ser in a synthetic glycopeptide was solvent exposed and accommodated in the ternary complex of MHC, glycopeptide, and TCR (56). In another example, Kasper and coworkers tracked endosomal transport of group B streptococcal type III polysaccharides coupled to carrier protein. Resultant glycopeptides were presented by MHC-II molecules and activated CD4⁺ T cells. Downstream cytokine events led to T-cell help and to antibody-producing B cells that recognized polysaccharide antigen (55). These examples point out that the antibody response can accommodate antigen glycosylation. Therefore, the structure, orientation, and processing of the substituting glycan may be at play.

In the design of vaccines, it may be advantageous to consider the consequences of sequential infection. Mismatch of surface glycosylation patterns can prime the host for immunopathology if the antibody response does not match the T-cell response. Wanzeck et al. found this to be the case in mice infected with a highly glycosylated virus that was otherwise isogenic with the wild-type virus used in the challenge. The mice were not protected from infection and experienced significant T-cell-mediated immunopathology (33). The work was an attempt to model infection with the 2009 pandemic H1N1 strain, where a similar elicited response was seen in human populations. This may explain the severe disease in young adults infected with the 2009 H1N1 pandemic virus (57). A Th1/Th2 mismatch may occur, shifting the balance toward cytotoxicity. The fact that the strains studied were isogenic removed other antigenic drift as a possible contributor and focused the mismatch on the glycosylation.

In this work, we described the glycosylation patterns of reference antigens used in the 2014-2015 influenza vaccine. Cell substrate-dependent glycosylation differences were reported, and potential antigenic masking patterns were revealed to be localized to predicted antigenic sites. In the context of immune response and antigen processing, some glycosylation differences may be significant. Glycosylation mismatch may lead to a mismatch in the antibody response, as indicated by egg adaptations in seasonal vaccine antigens. The SRID potency assay is not sensitive to differences in glycosylation when performed using egg antigen-induced sheep sera. Further investigation into all aspects of how antigen glycosylation may impact influenza virus infection and vaccine performance is warranted.

MATERIALS AND METHODS

Reference antigens.

Three formalin-inactivated, egg-derived, whole-virus reference antigens representing three different viral strains (A/California/07/2009 X-179A [H1N1] [lot 76], A/Texas/50/2012 X-223A [H3N2] [lot 75], and B/Massachusetts/02/2012 wild type [WT] [lot 74]), two betapropiolactone (BPL)-inactivated, MDCK cell-derived, whole-virus reference antigens representing two viral strains {A/Texas/50/2012 X-223A [H3N2] [lot H3-Ag-1312] and B/Massachusetts/02/2012 WT [lot B(y)-Ag-1313]}, and a purified recombinant influenza virus hemagglutinin (rHA) reference antigen of A/California/07/2009 (H1N1) (lot H1-Ag-1303), produced in insect cells, were obtained from the Center for Biologics Evaluation and Research (CBER), Food and Drug Administration. A BPL-inactivated, cell-derived, whole-virus reference antigen of A/Brisbane/10/2010 (H1N1) (NIBSC code 11/134) was obtained from the National Institute for Biological Standards and Control (NIBSC), United Kingdom.

Reference sera.

All serum lots were obtained from CBER. All sera were raised in sheep. The antiserum for H1N1 was raised against bromelain-cleaved egg-derived A/California/07/2009 reassortant X-181 (CBER lot H1-Ab-1503), which is essentially identical to the egg-derived A/California/07/2009 X-179 antigen which was characterized here. The antiserum for H3N2 was raised against A/Texas/50/2012 reassortant X-223 (CBER lot H3-Ab-1318), derived from egg and essentially identical to the egg-derived antigen studied here. The antiserum for B-type virus was derived from egg-grown B/Massachusetts/02/2012 reassortant BX-51B [CBER lot B(y)-Ab-1307], identical to the egg-derived antigen characterized here.

Chemicals and reagents.

HyperSep C₁₈ and porous graphitic carbon (PGC) cartridges were purchased from Thermo Fisher Scientific Inc. (Waltham, MA). TSKgel amide 80 particles were purchased from Tosoh Bioscience LLC (Montgomeryville, PA). Sequencing-grade modified trypsin, immobilized trypsin, and Glu-C were purchased from Promega Corp. (Madison, WI). N-glycosidase A was purchased from Roche Diagnostics Corporation (Indianapolis, IN). PNGase F was purchased from New England BioLabs, Inc. (Ipswich, MA). Iodomethane, dimethyl sulfoxide (DMSO), sodium hydroxide beads, and other chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA); solvents were of high-pressure liquid chromatography (HPLC) grade or higher. All other reagents were ACS grade or higher.

Protease digestion and microwave-assisted protein trypsin digestion.

Microwave-assisted proteolysis was performed as previously described (9). As needed, Glu-C digestions were performed on tryptic peptides suspended in phosphate-buffered saline (PBS) (pH 7.0) to a concentration of ∼1 μg/μl. Glu-C was added (enzyme/protein ratio, 1/50 [wt/wt]), and the incubation time was 18 h at 37°C.

Enrichment of glycopeptides by HILIC.

Intact glycopeptides were enriched via solid-phase extraction with TSKgel amide 80 hydrophilic interaction chromatography (HILIC) resin according to our previous report (58). Briefly, 200 mg (400 μl of wet resin) of amide 80 resin was placed into a Supelco fritted 1-ml column and washed with 1 ml of 0.1% trifluoroacetic acid (TFA)-water. The column was conditioned with 1 ml of 0.1% TFA-80% acetonitrile (ACN). The peptides were suspended in 0.1% TFA-80% ACN and applied to the column. The hydrophobic species were washed away with 3 ml of 0.1% TFA-80% ACN, and glycopeptides were eluted with 1 ml of 0.1% TFA-60% ACN and 1 ml of 0.1% TFA-40% ACN. The eluents were combined and vacuum dried.

Permethylation of released N-glycans.

Enriched glycopeptides were processed sequentially by C₁₈ and porous graphite solid-phase extractions as previously described (9). Solid-phase permethylation was conducted according to the published protocol of Mechref and colleagues (59), with modifications as previously described (9).

Reverse-phase NanoLC-MS^E analysis of glycopeptides.

Analysis of glycopeptides by LC-MS^E and data processing were performed as previously described (9). The NanoLC-MS^E data were processed by use of BiopharmaLynx 1.3 (Waters) to identify the site-specific glycosylation of HAs. Identified glycopeptides in MS^E spectra were confirmed manually. The oxonium ions (such as m/z 204.1, 366.1, and 528.2) present in MS^E spectra were used to help locate and determine the presence of glycopeptides. Assignment criteria included (i) manual observation of oxonium ions, peptide with a neutral loss of glycan fragment, and GlcNAc plus a peptide fragment; (ii) BiopharmaLynx identification; (iii) identification of peptide fragments; and (iv) mass accuracy of less than 20 ppm and 30 ppm in parent and collision ion scans, respectively. Glycosylation site occupancy was assessed using NetNGlyc (http://www.cbs.dtu.dk/services/NetNGlyc/), using a probability score of 0.5 or higher.

MALDI-TOF analysis of N-glycans.

An Autoflex Speed MALDI-TOF/TOF mass spectrometer (Bruker Daltonics) was used to analyze the permethylated N-glycans. After permethylation, the glycans were suspended in 30% acetonitrile-water, spotted onto a MALDI plate, and mixed 1:1 with 2,5-dihydroxybenzoic acid (2,5-DHB) matrix at 10 mg/ml in a 30% acetonitrile-water solution containing 1 mM sodium chloride. Samples were analyzed in positive-ion reflectron mode in the 900 to 5,000 m/z range. The MS data were processed by use of the FlexAnalysis program (Bruker Daltonics).

Three-dimensional modeling of glycosylated hemagglutinins.

The protein sequence of each HA studied here was aligned with those of similar HAs that have published X-ray crystal structures (16, 20, 25). The three-dimensional structures were aligned using Pymol (www.pymol.org). The glycosylation sites were mapped onto the surfaces of the HA 3-D structures under PDB numbers 3UBQ (H1N1), 4WE8 (H3N2), and 4M40 (B). The predominant glycan at each site, as detected by MS^E extracted ion chromatogram analysis, was displayed in association with each site. Relative abundances of glycoforms derived from eggs and from MDCK cells are reported as ratios of high-mannose to complex and hybrid glycans. The relative abundances of glycoforms produced on Sf9 cell-derived rHA are reported as ratios of high-mannose, paucimannose, and intermediate glycans.

SRID assay.

Reference antigens were prepared with or without prior treatment with an endoglycosidase cocktail containing Endo F1, Endo F2, and Endo F3 (Sigma-Aldrich). Briefly, 2 μl of each enzyme (0.03 U Endo F1, 0.015 U Endo F2, and 0.015 U Endo F3) and 10 μl of 5× buffer (250 mM sodium acetate, pH 4.5) were added to 200 μg of antigen with 33.5 μl of deionized water. The reaction mixtures were incubated at 37°C for 1 h. Reactions were confirmed by SDS-PAGE.

HA potency was measured using the SRID assay as previously described (8), with some modifications. Briefly, 1% agarose gels were prepared in phosphate-buffered saline, pH 7.2, and placed in a water bath to equilibrate to 50°C before adding the appropriate amount of HA-specific sheep antiserum. After gentle mixing, the gel was poured onto GelBond film (Cambrex, East Rutherford, NJ, USA) and allowed to solidify. Equally spaced 4-mm wells were punched into the gel. Reference antigens and samples were reconstituted and then incubated with an equal volume of 1% Zwittergent 3 to 14 (Calbiochem, San Diego, CA, USA) at room temperature for 30 min. Several dilutions of the reference antigens were made, spanning 8 to 35 μg/ml HA. Reference antigen dilutions and samples were loaded into wells on replicate gels and incubated in a sealed, humidified chamber at room temperature for 18 to 24 h. Gels were then washed in saline and rinsed in water before drying at 40°C. Gels were stained with 0.5% Coomassie brilliant blue for 10 min, destained, and then dried before measurement of the diameter of precipitant rings by use of an Immunolab scanner. Antigen potency was computed from the reference antigen linear dose-response curve. The test validity was based on the correlation coefficient (r) and equality of slopes (t) between test and reference antigens. Averages and standard deviations (SD) were calculated for at least two independent tests of each sample.