Abstract

INTRODUCTION

Influenza virus is a member of the family Orthomyxoviridae, which contains a segmented and negative-stranded RNA genome in an enveloped virion. This virus is classified into types A to C, and influenza A virus has caused four global pandemics in humans during the last hundred years (18, 22). Influenza A virus is further classified into subtypes based on the genetic and antigenic characteristics of hemagglutinin (HA) and neuraminidase (NA), which constitute envelope glycoproteins. To date, 16 HA (H1 to H16) and 9 NA (N1 to N9) subtypes have been identified, and influenza A viruses bearing H1N1 and H3N2 subtypes (A/H1N1 and A/H3N2 viruses) are currently cocirculating in humans.

Influenza A virus escapes from host immune responses by changing the antigenicity of HA and NA both gradually (antigenic drift) and abruptly (antigenic shift) (16). The HA consists of signal peptide, HA1 (amino acid positions 1 to 328 in the H3 numbering system), and HA2 (amino acid positions 330 to 550), and it forms trimers in virions (30). The ectodomain of an HA trimer can be divided into globular head and stem regions (30). The globular head region is composed of a part of HA1 (amino acid positions 58 to 272), while the stem region is composed of the other parts of HA1 and HA2. Antigenic sites (AS), which are targets of antibodies (Ab), constitute five epitopes (A to E) in A/H3N2 viruses. Epitopes A, B, D, and E reside in the globular head region, while epitope C is in the stem region of HA1 (24, 29).

Although the escape from host immune responses occurs through changes in amino acids at AS that are recognized by Ab (14, 18, 23), it has been proposed that the attachment of an oligosaccharide to the N-glycosylation site (NGS), which is the Asn residue of the sequon (Asn-Xaa-Ser/Thr, where Xaa is any amino acid except for Pro), in the globular head region also contributes to the escape (1, 21, 27, 28). This is based on observations in experimental studies that some N-glycans attached to the globular head region interfere with the binding of Ab to AS by masking the surface of HA (1, 21, 27, 28).

In the evolutionary studies of A/H1N1 and A/H3N2 viruses, it has been observed that the number of NGS in the globular head region has increased continuously for several decades after the emergence of these viruses in the human population (4, 6–8, 11, 32). It was speculated that the addition of NGS to the globular head region conferred selective advantages to the viruses by preventing the binding of Ab to AS. However, this hypothesis has been questioned, because the gain of NGS in the globular head region of HA did not appear to influence the amino acid variation at AS, and no correlation was observed between the transitions of antigenic clusters and gains or losses of NGS during the evolution of human A/H3N2 virus (4, 8).

If N-glycans attached to the globular head region contribute to preventing the binding of Ab to AS, it is expected that natural selection for amino acid substitutions at AS is weakened (strengthened) when AS that are not covered (covered) by N-glycans become covered (not covered) through gains (losses) of NGS. Natural selection for amino acid substitutions may be identified from a comparison of the rate of nonsynonymous substitution (d_N) to that of synonymous substitution (d_S). Positive, negative, and no selection are inferred when the d_N/d_S value is >1, <1, and 1, respectively. It has been observed that d_N/d_S at AS is often >1, whereas that at other sites (NAS) is usually <1 in HA of human A/H1N1 and A/H3N2 viruses (20, 23, 24, 31).

The purpose of the present study was to examine the effect of gains and losses of NGS on d_N/d_S at AS and NAS that are covered and not covered by N-glycans during the evolution of human A/H3N2 virus.

MATERIALS AND METHODS

Estimation of d_N/d_S at AS and NAS.

For each amino acid position that was found to be an NGS during some period in the evolutionary history of human A/H3N2 virus, d_S and d_N at AS_10c, AS_15c, AS_10uc, AS_15uc, NAS_10c, NAS_15c, NAS_10uc, and NAS_15uc were estimated for the branches where the position was an NGS and a non-NGS separately. When an amino acid site in AS_10c, AS_15c, NAS_10c, and NAS_15c of a particular position was considered to possibly be covered by an N-glycan attached to another position on some branches where the former position was a non-NGS, these branches were eliminated from the computation of d_N and d_S for the non-NGS branches of the former position. In addition, three branches connecting HA sequences derived from human infections of triple-reassortant A/H3N2 viruses were excluded from all computations of d_S and d_N (Fig. 1) (2).

Open in a separate window

Fig 1

Gains and losses of NGS in the globular head region of HA in the evolutionary history of human A/H3N2 virus. The phylogenetic tree was constructed for 1,904 HA-coding nucleotide sequences by the NJ method with GTR+G+I. The red and blue arrows indicate the trunk branches where gains and losses of NGS in the globular head region occurred, respectively, with the amino acid positions of NGS indicated. The numbers in parentheses indicate the earliest (for gains) or most recent (for losses) year when the virus with the NGS was isolated. The long branches colored in blue were excluded from the computation of d_S and d_N, because the HA sequences at the ends were derived from human infections of triple-reassortant A/H3N2 viruses.

The d_S and d_N values were computed as Sd/S and Nd/N, respectively, where Sd and Nd denote the numbers of synonymous and nonsynonymous differences summed over the branches of sites included in each category, and S and N denote the numbers of synonymous and nonsynonymous sites averaged over the branches of sites included in each category with the weight proportional to the branch length. The numbers of synonymous and nonsynonymous differences as well as the numbers of synonymous and nonsynonymous sites for each branch of individual codon sites were obtained by comparing the sequences at the ancestral and descendant nodes using the modified Nei-Gojobori method with κ = 10. Since positions 165 and 285 appeared to be NGS throughout the evolutionary history of human A/H3N2 virus, the analysis of d_S and d_N was not conducted with these positions.

RESULTS

Gains and losses of NGS in HA of human A/H3N2 virus.

A phylogenetic tree was constructed for 1,903 complete HA-coding nucleotide sequences of A/H3N2 viruses isolated from humans in the period between 1968 and 2010 (Fig. 1), and the ancestral nucleotide sequence was inferred at each interior node. When the nucleotide sequences at the interior and exterior nodes were translated into amino acid sequences, some of them contained NGS at amino acid positions 63, 81, 122, 126, 133, 144, 165, and 246 in the globular head region and at positions 6, 7, 8, 22, 38, 45, 276, 285, 483, and 498 in the stem region (Fig. 2). N-glycans appeared to cover AS when they were attached to positions 45, 63, 81, 122, 126, 133, 144, 165, 246, 276, and 285 (Table 1).

Open in a separate window

Fig 2

Amino acid positions of NGS in the 3D structure of human A/H3N2 virus HA (PDB no. 3HMG) identified in the present study. RasMol (3) was used to visualize the 3D structure.

Table 1

Amino acid sites possibly covered by N-glycan in the 3D structure of trimeric HA

Antigenic sitea and NGSb	Amino acid site (epitope: amino acid position[s]) possibly covered by N-glycanc
AS10c
45	C: 44, 45, 46, 47, 297, 312
63	D: 96; E: 62, 63, 75, 78, 91, 92, 94
81	A: 150; D: 117, 121; E: 78, 80, 81, 82, 83
122	A: 122, 124, 150; D: 121
126	A: 124, 126, 130, 132, 168; B: 128, 129, 164, 165; D: 167
133	A: 131, 132, 133, 135, 146, 152; B: 155
144	A: 135, 137, 138, 140, 142, 143, 144, 145, 146
246	B: 163, 164, 165, 186; D: 201, 203, 212, 216, 217, 218, 219, 244, 246, 247, 248
276	C: 50, 51, 53, 54, 275, 276, 278
AS15c
45	C: 44, 45, 46, 47, 48, 294, 297, 299, 300, 305, 307, 308, 309, 310, 311, 312
63	D: 96; E: 59, 62, 63, 67, 75, 78, 80, 82, 87, 88, 91, 92, 94
81	A: 122, 150, D: 117, 121, 177, 179; E: 57, 59, 67, 78, 80, 81, 82, 83, 87, 260, 261
122	A: 122, 124, 126, 132, 133, 150, 152, 168; D: 121, 167, 170, 172, 174, 176, 177, 179; E: 80, 81
126	A: 122, 124, 126, 130, 131, 132, 133, 152, 168; B: 128, 129, 155, 157, 163, 164, 165; D: 167, 244, 246, 247
133	A: 122, 124, 126, 130, 131, 132, 133, 135, 137, 140, 142, 144, 145, 146, 150, 152; B: 128, 129, 155, 156, 157, 158, 194
144	A: 133, 135, 137, 138, 140, 142, 143, 144, 145, 146; D: 96, 226; E: 75
246	A: 126, 130, 168; B: 128, 129, 163, 164, 165, 186, 187, 188, 190, 197; D: 167, 182, 201, 203, 209, 212, 213, 214, 215, 216, 217, 218, 219, 227, 228, 229, 242, 244, 246, 247, 248
276	C: 48, 50, 51, 53, 54, 273, 275, 276, 278, 279, 280; E: 57, 59

Open in a separate window

^aAS_10c and AS_15c denote the antigenic sites located within 10 and 15 Å from NGS, respectively.

^bAmino acid positions of NGS analyzed in the present study.

^cEpitopes and amino acid positions possibly covered by N-glycans attached to the indicated amino acid positions. Amino acid positions used for the computation of d_N/d_S are denoted in boldface. Amino acid positions with underlines indicate those constituting sequons during some period in the evolutionary history of human A/H3N2 virus. The numbering system of amino acid sites follows the H3 numbering system with reference to A/Memphis/102/72 (PDB no. 3HMG).

The occurrences of gains and losses of NGS that may cover AS in the evolutionary history of human A/H3N2 virus was inferred from the analysis of nucleotide sequences at the interior and exterior nodes of the phylogenetic tree. When A/H3N2 virus emerged in the human population in 1968, only positions 81 and 165 were NGS in the globular head region of HA (19). The NGS at position 165 has been maintained throughout the evolutionary history of the virus. However, position 81 became a non-NGS in 1974 (Fig. 1). The NGS generated at positions 63, 122, 126, 133, 144, and 246 have been retained during evolution. In the stem region of HA, position 285 has been maintained as an NGS throughout the evolutionary history of the virus. The NGS at position 276 reverted to a non-NGS 3 years after its generation in 1993. At position 45, the gains and losses of NGS occurred sporadically since 1997.

Effect of N-glycosylation on natural selection at AS and NAS.

N-glycans appeared to cover AS when they were attached to positions 45, 63, 81, 122, 126, 133, 144, 165, 246, 276, and 285, as indicated above. For each of these amino acid positions, the effect of N-glycosylation on natural selection at AS was examined by computing d_N/d_S at AS_10c, AS_15c, AS_10uc, and AS_15uc for the branches where the amino acid position under consideration was an NGS and a non-NGS separately. When these positions were non-NGS, d_N/d_S values at AS_10c and AS_15c were estimated to be 0.42 to 2.25 and 0.36 to 2.20, respectively (Table 2). (Note that d_N/d_S was incalculable for some cases because d_S = 0.) Positive selection was detected at AS_10c of position 144 (P = 0.03 by χ² test). When the positions mentioned above became NGS, d_N/d_S at AS_10c and AS_15c generally decreased to 0.11 to 1.54 and 0.23 to 1.47, respectively, and no positive selection was detected. A reduction of d_N/d_S was observed at AS_10c of 3 out of 5 positions and at AS_15c of 6 out of all 6 positions where the comparison of d_N/d_S between the NGS and non-NGS branches could be made, and the latter observation was statistically significant (P = 0.03 by binomial test). Furthermore, the reduction of d_N/d_S at AS_15c of position 144 was statistically significant (P < 0.05 by χ² test).

Table 2

d_N/d_S values estimated at antigenic sites possibly covered by N-glycan

Antigenic site,a NGS class,b and NGSc	dN/dSd
	Non-NGS	NGS
AS10c
Maintained
63	NA	1.26
122	NA	NA
126	0.73	0.11
133	0.43	0.44
144	2.25	1.54
246	1.01	0.82
Avg	1.30	0.82
Lost
45	0.95	NA
81	0.42	0.50
276	1.52	NA
Avg	1.07	0.94
AS15c
Maintained
63	NA	1.47
122	0.40	0.39
126	0.99	0.66
133	2.20	1.21
144*	1.93	0.74
246	0.97	0.33
Avg**	1.42	0.73
Lost
45	0.36	NA
81	0.54	0.23
276	1.75	NA
Avg	0.70	0.45

Open in a separate window

^aAS_10c and AS_15c denote the antigenic sites located within 10 and 15 Å from NGS, respectively.

^bNGS class of the indicated amino acid positions. Maintained indicates that the amino acid positions were maintained as NGS after the gains of NGS during evolution. Lost indicates that the NGS at the amino acid positions were lost during evolution.

^cAmino acid positions of NGS analyzed in the present study. An asterisk is attached to the amino acid position when the difference in d_N/d_S values between the non-NGS and NGS branches was statistically significant (*, P < 0.05; **, P < 0.01).

^dThe d_N/d_S value at AS when the indicated amino acid positions were non-NGS and NGS. The d_N/d_S values are indicated in boldface when positive selection (d_N/d_S > 1) or negative selection (d_N/d_S < 1) was detected (P < 0.05). NA indicates that d_N/d_S was incalculable because d_S = 0.

The NGS analyzed above may be classified into two groups according to whether they were maintained (positions 63, 122, 126, 133, 144, and 246) or lost (positions 45, 81, and 276) during the evolution of human A/H3N2 virus. When the average d_N/d_S values at AS_10c and AS_15c for the positions included in each class was computed for the NGS and non-NGS branches separately, it was observed that the d_N/d_S value for the NGS branches generally was smaller than that for the non-NGS branches (Table 2). In particular, the difference was statistically significant at AS_15c of the class of positions maintained as NGS during evolution (P < 0.01 by χ² test).

For AS_10uc and AS_15uc, the d_N/d_S values were estimated to be 0.45 to 0.62 and 0.39 to 0.75, respectively, when the positions 45, 63, 81, 122, 126, 133, 144, 246, and 276 were non-NGS (Table 3). Negative selection was detected in most cases, and no positive selection was detected at all. When these positions became NGS, the d_N/d_S values at AS_10uc and AS_15uc did not change to any large extent (they were 0.32 to 0.73 and 0.47 to 0.76, respectively), except for AS_15uc of position 276, where an increase in d_N/d_S (from 0.67 to 2.48) was observed. Negative selection again was detected in most cases.

Table 3

d_N/d_S values estimated at antigenic sites possibly not covered by N-glycan

Antigenic site,a NGS class,b and NGSc	dN/dSd
	Non-NGS	NGS
AS10uc
Maintained
63	0.45	0.56
122	0.62	0.52
126	0.52	0.56
133	0.60	0.52
144	0.60	0.51
246	0.56	0.55
Avg	0.59	0.54
Lost
45	0.56	0.32
81	0.56	0.49
276	0.54	0.73
Avg	0.55	0.57
AS15uc
Maintained
63	0.39	0.73
81	0.73	0.47
122	0.74	0.66
126	0.72	0.70
133	0.75	0.67
144	0.65	0.76
246	0.53	0.73
Avg	0.67	0.71
Lost
45	0.69	NA
81	0.73	0.47
276*	0.67	2.48
Avg	0.70	0.91

Open in a separate window

^aAS_10uc and AS_15uc denote the antigenic sites located farther than 10 and 15 Å away from NGS, respectively.

^bNGS class of the indicated amino acid positions. Maintained indicates that the amino acid positions were maintained as NGS after the gains of NGS during evolution. Lost indicates that the NGS at the amino acid positions were lost during evolution.

^cAmino acid positions of NGS analyzed in the present study. An asterisk is attached to the amino acid position when the difference in d_N/d_S between the non-NGS and NGS branches was statistically significant (*, P < 0.05).

^dThe d_N/d_S value at AS when the indicated amino acid positions were non-NGS and NGS. The d_N/d_S values are indicated in boldface when negative selection (d_N/d_S < 1) was detected (P < 0.05). NA indicates that d_N/d_S was incalculable because d_S = 0.

It should be noted that the d_N/d_S value at AS may be influenced by natural selection operating not only on the specific function of AS but also on the HA as a whole. To eliminate the possibility that the reduction of d_N/d_S at AS_10c and AS_15c observed above reflected a change in functional constraints on the entire region of HA, e.g., due to a change in the effective population size of human A/H3N2 virus, d_N/d_S was also computed for NAS_10c, NAS_15c, NAS_10uc, and NAS_15uc.The d_N/d_S values obtained at NAS generally were much smaller than those obtained at AS, and negative selection was detected in most cases whether the amino acid positions 45, 63, 81, 122, 126, 133, 144, 246, and 276 were non-NGS or NGS (see Tables S2 and S3 in the supplemental material). Although a significant difference in d_N/d_S was observed between NGS and non-NGS branches in some cases, d_N/d_S was usually greater in the former branches than in the latter, which was the opposite of the pattern observed at AS_10c and AS_15c.

DISCUSSION

After the emergence of A/H3N2 virus in the human population in 1968, NGS were generated at amino acid positions 63, 122, 126, 133, 144, and 246 in the globular head region of HA. Since the glycosidic linkage of oligosaccharides may be flexible, N-glycans may have multiple conformations in solution (26). In addition, the structure of N-glycans may vary among NGS in HA (13, 15). For these reasons, two different distances from NGS, 10 and 15 Å, were used as the possible ranges of amino acid positions covered by N-glycans in the present study. When positions 63, 122, 126, 133, 144, and 246 were non-NGS, the d_N/d_S value was >1 on average for both AS_10c and AS_15c and positive selection was detected at AS_10c of position 144, suggesting that AS_10c and AS_15c of these positions contained target sites of Ab. However, the d_N/d_S values at AS_10c and AS_15c decreased, and the evidence of positive selection disappeared when the positions mentioned above became NGS. In contrast, the d_N/d_S values at AS_10uc and AS_15uc generally were <1 and did not change to any large extent whether the positions mentioned above were NGS or non-NGS. In addition, the d_N/d_S values at NAS_10c, NAS_15c, NAS_10uc, and NAS_15uc often were greater when these positions were NGS than when they were non-NGS, which was the opposite of the pattern observed at AS_10c and AS_15c. These results indicate that natural selection for amino acid substitutions at AS_10c and AS_15c was weakened due to the gain of NGS in the globular head region of HA, supporting the hypothesis that the N-glycans attached to the NGS in the globular head region of HA prevented the binding of Ab to AS in the evolutionary history of human A/H3N2 virus.

The results obtained in the present study may contradict those obtained in a previous study, where a reduction of amino acid variability was not observed after gains of NGS in the globular head region at ±5 amino acid sites from the NGS in the linear sequence of HA for human A/H3N2 virus (8). There are several differences between the present and previous studies that may have caused conflicting results. First, in the present study, amino acid sites possibly covered by N-glycans were identified using the distance from NGS in the 3D structure of trimeric HA, taking into account the size of N-glycans, whereas in the previous study, the ±5 amino acid sites from NGS were assumed to be covered by N-glycans. In addition, the AS and NAS were analyzed separately in the present study, whereas they were mixed in the previous study. It is therefore likely that the effect of N-glycans on natural selection for amino acid substitutions at AS could be detected more efficiently and specifically in the present study than in the previous study. Second, in the present study, the change in natural selection for amino acid substitutions was measured using d_N/d_S, whereas the number of different amino acids was used in the previous study. Since the amino acid variability is more strongly influenced by the mutation rate than the d_N/d_S value, it is likely that the change in natural selection was more reliably identified in the present study than in the previous study.

The N-glycans attached to NGS at positions 45 and 276 in the stem region of HA appeared to have covered AS in the evolutionary history of human A/H3N2 virus. However, the NGS at these positions have been generated only sporadically and have not been maintained for a long evolutionary period. It is known that natural selection for amino acid substitutions operates strongly around the receptor binding pocket (RBP), which is located in the globular head region of HA, because Ab may interfere with viral entry into the host cell more efficiently when they bind to the amino acid positions closely located to the RBP (24, 29). Indeed, positive selection was detected at AS_10c of amino acid position 144, which was located close to the RBP, and the NGS generated in the globular head region appeared to be maintained longer than those generated in the stem region during evolution. It is therefore conceivable that N-glycans attached to the globular head region of HA were advantageous to the virus by preventing the neutralization by Ab, whereas those attached to the stem region were less advantageous.

Furthermore, the NGS at amino acid position 81, which was located in the globular head region, also was lost during evolution. This event apparently occurred after the generation of NGS at position 63, which was located adjacent to position 81 in the 3D structure of HA. The d_N/d_S values at AS_10c and AS_15c of position 63 were incalculable when this position was a non-NGS. However, experimental studies have shown that N-glycosylations at positions 63 and 81 both block the binding of Ab to AS (1, 21). Since some of the AS that were likely to be covered by N-glycans attached to positions 63 and 81 overlapped (29), and since the NGS at position 63 has been maintained after its generation, it is speculated that the generation of NGS at position 63 compensated for the advantageous effect of N-glycosylation at position 81 by shielding the overlapping amino acid sites from the access of Ab, which may have allowed for the loss of the NGS at position 81 after the generation of the NGS at position 63.

Supplementary Material

ACKNOWLEDGMENTS

Footnotes

REFERENCES