Successful replication of influenza viruses in eggs was first reported in 1940 (6
). The allantoic sacs of embryonated eggs became an ideal substrate for influenza vaccines because they supported high antigen yields. The vast majority of current influenza vaccines are also produced in eggs from viruses propagated exclusively in eggs because vaccines produced in cell culture systems have not yet been licensed in most countries (22
). Although the egg allantoic sac has been used successfully to produce influenza vaccines for 60 years, recent human H3N2 viruses have largely lost the ability to replicate in this host unless they acquire certain mutations in the HA (48
). This would not become a major problem if a vaccine virus seed could be used for many years after its derivation in eggs. However, rapid antigenic drift mandates frequent updates in influenza vaccine composition, which requires derivation of new virus seeds in eggs within a short period of time (21
Host selection of influenza A virus mutants has been known for over 65 years (7
), but we have not yet fully characterized the role of egg propagation on amino acid changes and virus receptor specificity. Cell-grown viruses have been demonstrated to accumulate fewer mutations than viruses present in the respiratory tract (29
) and thus are more representative of the wild-type virus circulating in the human population.
Previous studies have identified 16 positions at which substitutions emerged in the HA of H3N2 viruses isolated in eggs. However, the majority of these studies reported using viruses isolated between 1986 and 1990 (19
), while one studied a single 1997 virus (43
) and two studies looked at a A/Fujian/2002-like virus (38
). Our studies with 9 pairs of viruses isolated in 2006 and 2007 identified substitutions at 8 of these 16 previously reported positions, and all were close to the RBS (Fig. ). Formal demonstration of their role in viral replication in the egg host system would require detailed analysis of isogenic mutant viruses. However, their potential significance for replication in eggs can be inferred from the nonrandom distribution of the 16 changes in the HA (Table ) with regard to spatial and population diversity considerations. First, the majority of the variation occurred at only eight amino acid positions, all of which are vicinal to the receptor binding site in the HA1. Second, most changes involved the 16 amino acids previously reported to vary in relation to isolation in eggs (19
On the basis of their binding profiles to glycan microarrays, the virus pairs studied could be divided into two groups. Group 1 HAs had only minimal differences in glycan microarray binding, and this was perhaps reflected in the somewhat reduced number of substitutions observed between the cell- and egg-derived HAs. Glycan array analysis of these cell isolates revealed that they already possessed a restricted sialoside binding profile. This property was associated with specific amino acids at critical positions; e.g., cell-derived NH06 had a mixed population at position 194 (Leu/Pro), while the HO06 cell-derived isolate already had a Leu at position 183, previously reported to be egg-selected (38
). Group 2 virus isolates from eggs, however, had more dramatic changes with respect to their glycan binding profiles; receptor recognition was restricted to a smaller subset of glycans compared to the size of the subset in their cell grown counterparts. These changes suggested that egg isolation was frequently associated with a Gly186Val substitution in combination with a substitution in the 220 loop of HA. The NY306 egg isolate was the exception, having a Leu194Pro substitution besides the 220-loop modification.
Substitutions were also observed at seven additional sites that had not been previously reported for H3N2 viruses isolated in eggs (colored blue in Fig. ). Five of the seven sites (190, 195, 198, 223, and 225) are within or proximal to the RBS. Position 190 substitutions have been associated with the replication of H1N1 viruses in eggs (39
), while position 198 was deleted in one of the BR07 clones (discussed below). Residues 223 and 225 are in the 220 loop, which was previously reported to be subject to changes during growth in eggs (44
). Collectively, changes at these five positions merit further studies to determine their role in efficient viral replication in the egg host system.
Our results from the BR07 egg-isolated high-growth reassortants can shed light on virus-host interactions in this system. The X-171A clone had a single change (Leu194Pro), while the X-171B clone had two amino acid substitutions and one deletion (Leu186Pro, Ala198del, and Ser219Tyr) (Table ). The uncloned BR07 isolated in eggs revealed heterogeneous populations at three of the four positions (Gly/Val at position 186, Leu/Pro at position 194, Ala/del at position 198), confirming the presence of most of these changes prior to selection of clonal reassortant viruses. Asp/Asn at 190 was the only change not seen in the cloned reassortants. Interestingly, two of the three changes observed in the high-growth reassortant (X-171B) were present in the egg passage 5 virus and predominated in the passage 9 virus (data not shown), suggesting that the Gly186Val/Ala198del combination is beneficial for growth in eggs. These two changes, in combination with an additional Ser219Tyr substitution (near the 220 loop), resulted in a dramatic switch in receptor specificity with a highly restricted binding profile (Fig. ). Furthermore, a high-growth virus with a deletion in the HA (Ala198del) was obtained in eggs, despite the scarcity of trinucleotide deletions relative to single base substitutions in the original virus population, suggesting that the Ala198del/Gly186Val combination imparted a replication advantage in eggs. Antigenic analyses of the cell and egg isolates by hemagglutination inhibition tests using a panel of ferret antisera to A/Brisbane/10/2007 wild-type virus and each of its reassortants failed to reveal a ≥8-fold and reciprocal titer differences, although a partial (nonreciprocal) ≤8-fold difference was noted in these studies (data not shown).
Disruption of the helix at position 190 resulting from the introduction of a Leu194Pro substitution in BR07 (as well as in NH06 and NY306) appears to be correlated with a glycan binding profile similar to that of the X-171B high-yield reassortant (Fig. and ). The same appears to happen for the deletion at position 198, although its true effect cannot be determined due to the additional substitutions at positions 186, 219, and 226, which may also play a role (as seen with the other group 2 viruses). Position 186 and the 220-loop substitutions all introduce bulkier and/or hydrophobic residues to the left side of the RBS (Fig. ). Comparing glycan microarray binding of X-171 and X-171B, which differ only in the position of the substitution on the 220 loop (Ile226Ser in X-171 and Ser219Tyr in X-171B), suggests that position 219 is the critical residue responsible for a more restrictive binding profile. Modeling a sialyl Lewis x glycan (Neu5Acα2-3Galβ1-4(Fucα1-3)GlcNAcβ) into a hypothetical BR07 binding pocket indicates that the fucose could directly interact with the hydroxyl of the phenolic side chain of the Tyr and might therefore selectively stabilize these glycans (Fig. ). In contrast, an Ile226Ser substitution introduces a potential hydrogen bond to the O-4 of galactose, which is less selective.
FIG. 5. Structural model of BR07 X-171B HA receptor binding site with a bound fucosylated sialoglycan. The avian receptor analog Neu5Acα2-3Galβ1-4(Fucα1-3)GlcNAcβ (sialyl Lewis x, 29 in the microarray; see Table S1 in the supplemental (more ...)
The narrow receptor specificity of the high-growth X-171B reassortant raises the possibility that these glycans might be present and used by the virus during egg growth: α2-3 sialylated LacDiNAc (glycan 21), sialyl Lewis x (glycans 29 to 33), α2-6 sialyl LacDiNac (glycan 55), α2-6 sialyl fucosyl tri-lactosamine (glycan 57), and sialylated N
-glycolyl Lewis x (glycan 68). However, to date, these structures have not been identified as terminal sequences on N-linked glycans of either embryonated chicken egg chorioallantoic or amniotic cells (63
), though neutral nonsialylated glycans containing LacDiNAc as well as fucosylated N
-acetyllactosamine have been described in the N-glycan structures of a chicken eggshell protein, ovocleidin-116 (47
), indicating that chickens have the biosynthetic capability to make the corresponding sialylated structures.
Kumari et al. recently analyzed MDCK cell-grown viruses isolated between 2003 and 2005 and found restricted specificity to only α2-6 glycans (35
). Except for WY03, all the virus pairs presented in this study are isolated from 2006 onwards and revealed a much broader specificity. A possible reason for this difference may be due to a Ser193Phe substitution that emerged in 2005 and that is currently the dominant residue in circulating strains. Analysis of the HA sequences from the study of Kumari et al. reveals a serine at this position. Position 193 is in the alpha helix that sits atop the RBS, and the bulky aromatic side chain is predicted to extend over the receptor pocket. This, in conjunction with substitutions (Ser189Asn and Ser227Pro) on the left side of the RBS in regions of importance for egg adaptation, appeared in viruses from 2004 onwards and may contribute to the increased α2-3 binding observed in this study. Interestingly, Gulati et al. reported that cell-derived Brisbane/10/07-like viruses bound to very few sialoglycans on the array, primarily α2-6 sialyl-polylactosamines (25
). The reason why their findings are so different from those for the viruses analyzed here is unknown. An analysis of the virus dilution of both cell- and egg-grown A/Brisbane/10/2007 (H3N2) virus at 2-fold incremental HA titers between 32 and 512 yielded an overall reduction in overall glycan signals rather than enrichment of the single long polylactosamine reported previously (see Fig. S3 in the supplemental material), suggesting that a drastic difference in the HA titers used between these different laboratories is not the reason. The rapid evolution of the receptor binding preferences of H3N2 antigenic variants that emerged between 2003 and 2007 (Fujian/410-like, Wisconsin/67-like, California/7-like, and Brisbane/10-like viruses) highlights the need to reevaluate functional predictions with any future variant viruses of interest (25
Compared to viruses isolated in MDCK cell culture, egg isolates from 2006 and 2007 can have both modestly and dramatically different glycan microarray profiles. In addition, substitutions at different positions, such as 194 (X-171A) and 186/198 (X-171B), appear to produce similar binding profiles, thus illustrating the complexity of the system that we are attempting to understand. Williams and Robertson (79
) reported that substitutions in the vicinity of the RBS appear to be responsible not only for modulating attachment to receptors but also for triggering virus penetration into the target cells. The mutations acquired by group 1 egg-derived isolates (and some of those of group 2-derived egg isolates) may modulate postbinding events mediating endocytosis and membrane fusion.
Although early studies showed that the sequences of HA genes of H3N2 viruses isolated in cell culture are generally identical to those sequenced directly from nasal secretions of patients, recently circulating viruses may also undergo host selection during isolation in mammalian cells (30
). Isolation of H3N2 viruses in MDCK cells has been progressively more difficult in recent years (49
). This obstacle has been overcome by utilizing a transfectant MDCK cell line expressing higher levels of α2-6 sialosides, suggesting that the paucity of these receptors in regular MDCK cells may consequently exert selective pressure for selection of viruses with enhanced binding of α2-3 sialylated glycans. Most group 2 viruses isolated in cells show moderate to high levels of α2-3 sialoside specificity. Viruses in this group with strong α2-3 binding specificity also bind to α2-3 linked N-glycans as well as mixed 2-3/2-6 biantennary glycans (glycans 60 and 61, respectively). In contrast, this is not seen in many of the egg isolates. This raises the possibility that the α2-3 specificity seen in the MDCK cell-grown virus could reflect selection of viruses with increased specificity for α2-3 N-linked glycans to compensate for the low level of α2-6 N-linked glycans. Thus, the differences in glycan specificity seen between the egg- and cell-grown viruses could result from selective pressures exerted by both laboratory hosts.
These results also provide a snapshot of the receptor specificity of human H3N2 influenza viruses circulating in 2006 and 2007. Two long α2-6 sialylated glycans (glycans 56 and 57) were the ligands preferred by most viruses, a finding consistent with that from a previous report of a study that used recombinant HA (66
). Three trisaccharides less frequently considered influenza virus receptors were the ligands widely used by these viruses. Two were α2-6 sialylated LacDiNAc and sulfated N
-acetyllactosamine (glycans 55 and 41, respectively). Interestingly, the sialyl Lewis x glycan, whose presence in airway epithelial cells has been well documented (1
), was also an excellent ligand for many of the viruses in this panel, suggesting that α2-3-linked sialic acids should not be excluded as potential receptors for human influenza viruses. The consensus glycans for the 2006 and 2007 viruses includes the three previously reported ligands for the recombinant HA of A/Moscow/10/1999 (H3N2) virus (66
): sulfated α2-6 sialyl N
-acetyllactosamine (glycan 41), α2-6 sialyl di-N
-acetyllactosamine (glycan 56), and α2-6 sialyl fucosylated tri-N
-acetyllactosamine (glycan 57), indicating substantial conservation of receptor specificity (Table ). The use of glycan arrays to analyze the receptor specificity of influenza viruses promises to advance our understanding of a very complex and dynamic virus-host interface. However, the body of information on influenza virus receptor specificity derived from glycan array analyses is still insufficient to interpret the general significance of the data.
Eggs will remain the most economical and practical vaccine production substrate until a cell-based alternative is fully established. Understanding how receptor binding is evolving in human H3N2 viruses circulating globally may explain the basis of poor replication in eggs and the significance of more radical structural changes, such as the Ala198 deletion or Leu194Pro substitution by BR07. Such changes may affect secondary structures in the HA and thus could impact the antigenicity of site B. Structural analysis of these changes will help to clarify their effects in both regards. Further studies will establish the feasibility of analyzing the receptor binding profiles of cell isolates to identify those that match a previously determined glycan binding profile correlated with permissive growth in eggs. As shown with group 1 viruses, the original sequences in certain clinical specimens may be less prone to the type of changes acquired by the HA1 of BR07 (group 2) to generate the high-growth reassortant in eggs. Such viruses would reduce the risk of potential antigenic changes in seed viruses selected for vaccine production.