Antigenic and Receptor Binding Properties of Enterovirus 68

ABSTRACT

Increased detection of enterovirus 68 (EV68) among patients with acute respiratory infections has been reported from different parts of the world in the late 2000s since its first detection in pediatric patients with lower-respiratory-tract infections in 1962. However, the underlying molecular mechanisms for this trend are still unknown. We therefore aimed to study the antigenicity and receptor binding properties of EV68 detected in recent years in comparison to the prototype strain of EV68, the Fermon strain. We first performed neutralization (NT) and hemagglutination inhibition (HI) tests using antisera generated for EV68 strains detected in recent years. We found that the Fermon strain had lower HI and NT titers than recently detected EV68 strains. The HI and NT titers were also significantly different between strains of different genetic lineages among recently detected EV68 strains. We further studied receptor binding specificities of EV68 strains for sialyloligosaccharides using glycan array analysis. In glycan array analysis, all tested EV68 strains showed affinity for α2-6-linked sialic acids (α2-6 SAs) compared to α2-3 SAs. Our study demonstrates that emergence of strains with different antigenicity is the possible reason for the increased detection of EV68 in recent years. Additionally, we found that EV68 preferably binds to α2-6 SAs, which suggests that EV68 might have affinity for the upper respiratory tract.

IMPORTANCE Numbers of cases of enterovirus 68 (EV68) infection in different parts of the world increased significantly in the late 2000s. We studied the antigenicity and receptor binding properties of recently detected EV68 strains in comparison to the prototype strain of EV68, Fermon. The hemagglutination inhibition (HI) and neutralization (NT) titers were significantly different between strains of different genetic lineages among recently detected EV68 strains. We further studied receptor binding specificities of EV68 strains for sialyloligosaccharides using glycan array analysis, which showed affinity for α2-6-linked sialic acids (α2-6 SAs) compared to α2-3 SAs. Our study suggested that the emergence of strains with different antigenicities was the possible reason for the increased detections of EV68 in recent years. Additionally, we revealed that EV68 preferably binds to α2-6 SAs. This is the first report describing the properties of EV68 receptor binding to the specific types of sialic acids.

INTRODUCTION

Human enterovirus 68 (EV68) (species, Human enterovirus D; genus, Enterovirus; family, Picornaviridae) was first isolated in the United States from pediatric patients hospitalized with lower-respiratory-tract infections in 1962 (1). Since its first detection, EV68 has only rarely been identified. In an enterovirus surveillance conducted in the United States from 1970 to 2005, only a total of 26 strains were reported over 35 years (2). However, the number of reported EV68 cases increased dramatically in the late 2000s, in different parts of the world (3–8). The underlying molecular mechanism for this sudden increase of EV68 detections in recent years is still unknown.

On the basis of nucleotide sequences of the capsid coding region VP1, the EV68 strains detected in recent years were categorized into three genetic groups, lineages 1, 2, and 3 (6, 9). According to our previous reports, most of the EV68 strains detected in the Philippines belonged to lineage 2, and a few of them belonged to lineage 3 (4, 10). The VP1 sequences detected in the Philippines and other countries demonstrated that the amino acid sequence patterns in BC and DE loops were correlated with the categorization of EV68 into the 3 lineages (6, 8, 10). Generally, loop structures, including BC and DE loops, are located on the viral surface and associated with antigenic epitopes of enteroviruses (11, 12). Moreover, it was previously demonstrated that amino acid changes in the BC loop of coxsackievirus B4 resulted in loss of virus neutralization by serotype specific antisera, suggesting that the BC loop has an important role in determining antigenicity of the virus (13). Therefore, it was speculated that amino acid changes in the BC and DE loops of EV68 might also be associated with altered antigenicity of EV68 strains detected in recent years (6, 8).

Enterovirus 71 (EV71), which also belongs to the genus Enterovirus, is known to have 3 genogroups and several genotypes (14, 15). It was previously reported that the reemergence of B5 genotype resulted in a large EV71 outbreak in Taiwan, and detected B5 strains were shown to possess genetic and antigenic diversity compared to previously circulating strains (16). Similarly, the emergence of strains with different antigenicity might have been involved in the worldwide occurrence of EV68 outbreaks in recent years. However, the presence of antigenic differences between EV68 strains detected recently and those detected before the recent outbreaks has yet to be investigated. In addition, amino acid sequences of VP2 and VP3 of EV68, both of which comprise the outer surface of enteroviruses together with VP1 protein, have still not been fully investigated.

It was previously demonstrated that removal of sialic acids (SAs) from the surfaces of HeLa cells by Arthrobacter ureafaciens sialidase resulted in a 90% reduction of virus binding to the cell surface, from which it was suggested that SAs might be the receptor for human rhinovirus 87 (HRV87) (the former name of EV68) (17). However, the specific type of SA as the receptor for EV68 is still not determined. The SAs have been intensively studied as the viral receptors for influenza viruses, and it was previously shown that among influenza A viruses, human viruses have affinity for oligosaccharides containing N-acetylneuraminic acid α2,6-galctose (NeuAcα2,6-Gal), and avian viruses specifically bind to NeuAcα2,3-Gal (18, 19). It is known that in the human respiratory tract, NeuAcα2,6-Gal is distributed on the upper respiratory tract and, in contrast, NeuAcα2,3-Gal is found in the lower respiratory tract (20). It has been reported that the α2-3 SA-specific highly pathogenic avian influenza virus (H5N1) acquired increased affinity for human type receptors (α2-6 SAs) by amino acid substitutions in hemagglutinin (HA) (21). This mutant H5N1 was shown to be transmissible only via direct contact in the ferret model (21). Moreover, the mutant H5N1 with affinity for human type receptors acquired transmissibility via respiratory droplets by introducing human N2 neuraminidase (NA) (21). The study indicates the crucial roles of receptor specificity and NA functions in the transmissibility of influenza viruses. It might be possible that such phenotypic changes were also associated with the increased EV68 cases in recent years. However, it is still unknown whether any differences in SA binding patterns are present among EV68 strains and, if such differences exist, whether those differences are associated with the increased detection of the virus in recent years. In addition, it also remains unknown if EV68 has any sialidase activities corresponding to NA of influenza viruses. In order to clarify if such differences in receptor recognition patterns and sialidase functions are present between EV68 strains detected recently and those detected before the recent outbreak, the binding affinity of EV68 to different types of SAs and the presence of sialidase activities needs to be investigated.

In the current study, we therefore investigated the underlying mechanisms for the increased detection of EV68 in recent years, including antigenic properties and receptor binding patterns as well as the sialidase activities of EV68 strains detected in recent years, in comparison to the properties of the prototype strain of EV68 detected in 1962.

MATERIALS AND METHODS

Viruses.

EV68 strains were isolated from nasopharyngeal swabs collected from children with acute respiratory infections at Yamagata Prefectural Institute of Public Health, Japan, in 2010, and at Research Institute for Tropical Medicine (RITM), the Philippines, in 2011. Isolation of EV68 at Yamagata Prefectural Institute of Public Health was described previously (9). At RITM, clinical samples collected at a tertiary-care hospital located in Tacloban, the Philippines, as previously described (4) were inoculated on monolayer cell culture of rhabdomyosarcoma A (RD-A) cells (22) in Eagle's minimum essential medium (EMEM) (Sigma-Aldrich, Inc., St. Louis, MO, USA) containing 2% calf serum (CS) and 1.7% glucose and incubated at 34°C. After ∼48 to 96 h of incubation, cultured fluid was harvested and stored at −80°C. EV68 isolates from Yamagata Prefectural Institute of Public Health and RITM were passaged on a monolayer culture of RD-18S cells 6 times, and harvested culture fluid was stored at −80°C until the experiments were performed.

The prototype strain of EV68, Fermon (GenBank accession number AY426531), was purchased from American Type Culture Collection (ATCC), Manassas, VA, after which the virus was passaged on an RD-18S cell culture up to 6 times, and harvested culture fluid was stored at −80°C until use.

In the study, strains of influenza A virus (H1N1) pdm, human parainfluenza virus type 1, and enterovirus 71 were used as control viruses. The strain of influenza A virus (H1N1) pdm (A/IWATE/1003/2009) was isolated at Research Institute for Environmental Sciences and Public Health of Iwate Prefecture, Morioka, Japan, in 2009 and stored at −80°C until use, after being passaged on Madin-Darby canine kidney (MDCK) cell cultures three times. This strain was kindly provided by Masaki Takahashi of the Research Institute for Environmental Sciences and Public Health of Iwate Prefecture. The strain of parainfluenza virus was isolated at Sendai Medical Center, Virus Research Center, Sendai, Japan, in 2008 and stored at −80°C until use, after being passaged on Macaca mulatta kidney (LLC-MK2) cell cultures 9 times. This strain was kindly provided by Hidekazu Nishimura, Sendai Medical Center, Virus Research Center, Sendai, Japan.

The strain of enterovirus 71 was isolated at Research Institute for Tropical Medicine (RITM), Muntinlupa, the Philippines, in 2011 as part of the acute flaccid paralysis surveillance in the Philippines (23) and stored at −80°C until use, after being passaged on RD cell cultures twice.

Preparation of antisera.

Three strains of EV68 isolated in Yamagata in 2010 were used as antigens to immunize guinea pigs, including Y10-2013 (GenBank accession number AB614407), Y10-2082 (GenBank accession number AB614430), and Y10-2076 (GenBank accession number AB614440), which were selected as representative strains of lineages 1, 2, and 3, respectively. These 3 strains were propagated on RD-18S cell cultures. After harvesting, culture fluids were purified by precipitation with 8% polyethylene glycol 8,000 (MP Biomedicals) and then centrifuged on a 40% sucrose gradient at 40,000 rpm for 3 h. Specific-pathogen-free female Hartley guinea pigs (6 weeks of age) were intraperitoneally injected with 1.0 ml of formalin-inactivated virus mixed with 1.0 ml of incomplete Freund's adjuvant (Difco). Guinea pigs were injected with the mixture of antigens and adjuvant 3 times in total, with intervals of 1 week and 2 weeks sequentially. One week after the third injection, serum was collected and stored at −80°C after heat inactivation at 56°C for 30 min.

Antisera was also generated with specific-pathogen-free female New Zealand White rabbit (12 weeks of age) following the same protocol, but with Y10-2082 only and with injection via the subcutaneous route.

Hemagglutination and hemagglutination inhibition tests.

The hemagglutination (HA) test was performed at 4°C by adding 50 μl of 0.75% guinea pig erythrocyte suspension to each well of a V-shaped bottom plate containing 50 μl of virus 2-fold serially diluted in phosphate-buffered saline (PBS). After mixing, the plate was incubated at 4°C for 2 h. The hemagglutination units (HAU) of the virus were determined as the reciprocal value of the last dilution exhibiting complete hemagglutination.

In hemagglutination inhibition tests, 25 μl of 2-fold serially diluted antiserum in saline was mixed with 25 μl of 8 HAU of virus and incubated at 37°C for 1 h, prior to the addition of 50 μl of 0.75% guinea pig erythrocytes. Before being mixed with virus, antiserum was mixed with receptor-destroying enzymes (RDE) (Denka Seiken Company, Ltd., Tokyo, Japan) at a ratio of 1:3 and incubated at 37°C for 18 to 20 h and then 56°C for 30 min to stop the reaction. Hemagglutination inhibition (HI) titers were determined as the reciprocal value of the last dilution of the antiserum inhibiting virus-induced hemagglutination, after incubation at 4°C for 2 h.

For both HA and HI tests, hemagglutinin antigens of influenza A virus A/Brisbane/59/2007 (Denka Seiken Company, Ltd., Tokyo, Japan) were also included in the analysis.

Neutralization tests.

In neutralization tests, 25 μl of 2-fold serially diluted antiserum in saline was mixed with 25 μl containing 100 50% tissue culture infective doses (TCID₅₀) of virus and incubated at 37°C for 1 h, prior to inoculation onto RD-18S cell cultures with 100 μl of EMEM containing 2% CS and 1.7% glucose. The neutralization antibody (NT) titer was determined by observing reciprocal value of the last dilution of the antiserum inhibiting cytopathic effect (CPE) completely, after incubation at 34°C for 6 days.

For neutralization tests, one isolate of enterovirus 71 was also included in the analysis.

Synthesis of glycoproteins.

A total of 6 kinds of oligosaccharides were prepared from various biomaterials as described previously (24). After the catalytic reduction, 2.5 μmol of each oligosaccharide was mixed with 160 μmol of N-(3-meleimidobenzoyloxy)succinimide (MSB) dissolved in N,N-dimethylformamide (DMF) and incubated at 30°C for 30 min. After chloroform extraction, the products were mixed with 100 nmol of reduced bovine serum albumin (BSA) and incubated at room temperature for 2 h. Then, the products were purified by dialysis with PBS.

Glycan array.

Twelve isolates of EV68, including the Fermon strain, 9 isolates from Yamagata, Japan, and 2 isolates from Tacloban, the Philippines (TTa-11-Ph224 and TTa-11-Ph344), were diluted to 2 HAU in binding buffer (Rexxam Company Ltd., Kagawa, Japan). Five nanograms of synthesized glycoproteins was spotted onto a microarray-grade epoxy-coated glass slide, and 80 μl of each diluted virus solution was applied to the glass chambers.

The glass chambers containing virus were incubated at room temperature, with gentle shaking, for 1 h, and thereafter the virus was removed. Then 80 μl of polyclonal antibody generated for one of the EV68 strains, Y10-2082, using New Zealand White rabbits was diluted to 1:1,000 with binding buffer (Rexxam Company Ltd., Kagawa, Japan) and added to the chambers. The chambers were incubated at room temperature for 1 h with gentle shaking. After removal of the polyclonal antibodies, 80 μl of anti-rabbit IgG sera conjugated with Cy3 (Bioss Inc., MA, USA), which was diluted to 1:1,000 with binding buffer (Rexxam Company Ltd., Kagawa, Japan), was added to each chamber, and the chambers were incubated at room temperature with gentle shaking for 1 h. After the incubation, fluorescence of the glass chambers was measured with a Bio-Rex Scan 200 instrument (Rexxam Company Ltd., Kagawa, Japan).

Preparation of modified erythrocytes.

All α2-3-, α2-6-, α2-8-, and α2-9-linked sialic acids were removed from the surface of guinea pig erythrocytes (GPE), by incubating 100 μl of 10% GPE with 5 mU of Arthrobacter ureafaciens sialidase (Roche Applied Science, Penzberg. Germany) at 37°C for 3 h. Removal of sialic acids was confirmed by a complete loss of HA titer for the control viruses, human parainfluenza virus type 1 for removal of α2-3-linked sialic acids and influenza A virus (H1N1) pdm for that of α2-6-linked sialic acids.

After sialidase treatment, GPE were washed twice with PBS, and then 100 μl of 10% sialidase-treated GPE was incubated with ∼1.25 to 10 mU of α2-3 or α2-6 sialyltransferase (Sigma-Aldrich, Inc., St. Louis, MO, USA) and 1.0 mM cytidine-5′-monophospho-N-acetylneuraminic acid sodium salt (Sigma-Aldrich, Inc., St. Louis, MO, USA) at 37°C for 4 h. Resialylation was confirmed by complete recovery of the HA titer for the control viruses.

Detection of sialidase activity.

The 2′-(4-methylumbelliferyl)-α-d-N-acetylneuraminic acid sodium salt hydrate (MUNANA) (Sigma-Aldrich, Inc., St. Louis, MO, USA) was diluted to 0.2 mM with 32.5 mM MES buffer (Wako Pure Chemical Industries, Ltd., Japan), and 50 μl of diluted MUNANA substrate was mixed with 50 μl of EV68 or influenza A virus diluted to 1 to 16 HAU with EMEM (Sigma-Aldrich, Inc., St. Louis, MO, USA). The mixture was incubated in black 96-well plates (Nunc, Thermo Fisher Scientific Inc., USA) at 37°C for 3 h. After the incubation, 100 μl of glycine buffer (0.1 M; pH 10.7) was added to stop the reaction. The fluorescence was measured at wavelengths of 365 nm (excitation) and 450 nm (emission) by with a Fluoroskan Ascent fluorometer(Thermo Fisher Scientific Inc., USA).

Sequence analysis.

RNA was extracted from clinical specimens using a QIA viral RNA minikit (Qiagen Inc., CA, USA) according to the manufacturer's instructions. The cDNA was synthesized using random primers (Invitrogen Life Technologies, Carlsbad, CA, USA) and MMLV reverse transcriptase (Invitrogen Life Technologies, Carlsbad, CA, USA). The portions of the viral genome corresponding to the viral protein 1 (VP1) 1, VP2, and VP3 regions were amplified by reverse transcriptase PCR (RT-PCR), using primer pairs VP4F and VP2R, VP2F and VP3R, VP3F and VP1R, 484 and 222 (25), and EV68-VP1F and EV68-VP1R or 485 (25), respectively (Table 1). The PCR amplicons were purified using a SUPREC-PCR kit (TaKaRa, Shiga, Japan) or QIAquick PCR purification kit (Qiagen Inc., USA) and used as templates in cycle sequencing (ABI Prism BigDye terminator cycle sequencing ready reaction kit, version 1.1; Applied Biosystems) in an automated sequencer (3130/3130 xl genetic analyzers and 3730/3130 xl DNA analyzers; Applied Biosystems). Sequence analysis was conducted using MEGA 5 software and BioEdit version 7.0.0 software (Isis Pharmaceuticals, Inc., USA).

TABLE 1.

Primers used for analysis of sequences of enterovirus 68

Primer	Sequence (5′ → 3′)	Location (positiona)
EV68- VP4F	GGACCCATCAAAATTCACTG	VP4 (876–895)
EV68- VP2R	CCATTGATGTGGAAATATTG	VP2 (1451–1470)
EV68- VP2F	CCAGGGTTCGATGATATCATG	VP2 (1360–1380)
EV68- VP3R	GGCCCGTCTAACTGTATGTC	VP3 (1944–1963)
EV68- VP3F	GCACATTCCAGGGCAGGTCC	VP3 (1785–1804)
EV68-VP1RFH	CACCAAGTTCGGGCGTTAATC	VP1 (2469–2488)
484 (25)	GGRTCYCAYTACAGGATGT	VP1 (2197–2215)
222 (25)	CICCIGGIGGIAYRWACAT	VP1 (2933–2951)
EV68-VP1F	ACCATTTACATGCAGCAGAGG	VP1 (2393–2413)
EV68-VP1R	GACAAGAACTTTTTCAAATGGACAA	VP1 (2683–2707)
EV68-VP1LTHF	CAAAGAGAAGTTTCGAATAC	VP1 (2633–2652)
485 (25)	ACATCTGAYTGCCARTCYAC	2A (3425–3406)

^aNumbers correspond to the genome of EV68 strain Fermon (AY426531).

Selection analysis.

The partial VP1 sequences (corresponding to amino acid positions 20 to 241 of the VP1 sequences of strain Fermon) of 127 strains, including 47 of lineage 1, 18 of lineage 2, 46 of lineage 3, and 16 of EV68 detected before 2005, were obtained from GenBank and used for selection analysis. The ratio of the nonsynonymous substitution rate to the synonymous substitution rate (dN/dS) and statistical significance were calculated by hypothesis testing using the Phylogenies (HyPhy) package, based on the single-likelihood ancestor counting (SLAC) method with neighbor-joining trees and TN93 as the substitution model. dN/dS values higher than 1 were defined as indicating positive selection.

Statistical analysis.

Differences in HI and NT titers between EV68 strains belonging to three genetic clusters were analyzed by using JMP Pro 9.0.2 software (SAS Institute Inc., Cary, NC, USA). Statistical significance was calculated using the Wilcoxon rank sum test (α = 0.05).

RESULTS

Hemagglutination inhibition tests.

Hemagglutination inhibition (HI) titers were measured for guinea pig antisera generated for Y2013 (lineage 1), Y2082 (lineage 2), and Y2076 (lineage 3), using 12 strains of EV68 and hemagglutinin (HA) antigens of influenza A virus (H3N2) A/Brisbane/59/2007. The HI titer against EV68 isolates was higher than a serum dilution of 1:10, while that against A/Brisbane/59/2007 was not higher than 1:10 against antisera generated for EV68 of any of the three lineages detected in recent years. Among EV68 strains, Fermon had a lower HI titer than EV68 strains detected in recent years against any of the 3 antisera. In HI test using antisera generated for Y2082 (lineage 2), the titer of strains belonging to the homologous lineage; lineage 2 (Y2082, Y2086, Y2163, and TTa-11-Ph344) were significantly higher than those of heterologous lineages, including lineage 1 (Y2013, Y2071, and Y2167) (P = 0.042) and lineage 3 (Y2076, Y2256, Y2336, and TTa-11-Ph224) (P = 0.027) (Fig. 1b). Similarly, in HI tests using antisera generated for Y2076 (lineage 3), the HI titer of the homologous lineage, lineage 3, was higher than those of heterologous lineages, including lineage 2 (P = 0.036).

FIG 1.

NT and HI titers of EV68 strains against antisera generated for EV68 detected in recent years. The circles indicate the mean HI titers (a to c) and NT titers (d to f) among strains of each lineage, and whiskers indicate the 95% confidence intervals for the mean values. The P value for the difference of titers between lineages are indicated above the brackets. Samples were tested in duplicate in HI tests and quadruplicate in NT tests, and each experiment was performed separately 2 to 4 times.

Neutralization tests.

Neutralization antibody (NT) titer was measured for guinea pig antisera generated for 3 strains of Y2013 (lineage 1), Y2082 (lineage 2), and Y2076 (lineage 3), using 12 strains of EV68 and one strain of EV71. The NT titer against EV71 was no higher than any NT titer against EV68 by antisera for any of the three lineages. Among EV68 strains, the NT titer of the Fermon strain was no higher than that of EV68 strains detected in recent years against any of the 3 antisera. Notably, the NT titer of Fermon was almost identical to that of EV71 in the NT test using antisera generated for Y2013 (lineage 1). In NT tests using antisera generated for Y2082 (lineage 2), the NT titer of strains belonging to the homologous lineage; lineage 2 (Y2082, Y2086, Y2163, and TTa-11-Ph344) was significantly higher than those of heterologous lineages, including lineage 3 (Y2076, Y2256, Y2336, and TTa-11-Ph224) (P = 0.042) (Fig. 1e). Similarly, in NT tests using antisera generated for Y2076 (lineage 3), the NT titer of the homologous lineage, lineage 3, was significantly higher than those of heterologous lineages, including lineage 2 (P = 0.049) (Fig. 1f).

Glycan array.

Six kinds of glycan were tested for binding to EV68 using glycan array analysis. As a result, fluorescence due to the binding of EV68 to the glycan was observed for NeuAcα2,6LacNAc-BSA (glycan 1 [G1] in Fig. 2), NeuAcα2,6LacNAcβ1,2Manα1,3(NeuAcα2,6LacNAcβ1,2Manα1,6)Manβ1,4GlcNAcβ1,4GlcNAc-BSA (G2 in Fig. 2), and NeuGcα2,6Lac-BSA (G3 in Fig. 2). The absorbance for NeuAcα2,6LacNAc-BSA and NeuAcα2,6LacNAcβ1,2Manα1,3(NeuAcα2,6LacNAcβ1,2Manα1,6)Manβ1,4GlcNAcβ1,4GlcNAc-BSA decreased as the amount of coated glycan got smaller (Fig. 2). In contrast, only a limited level of fluorescence was observed for glycans with α2-3-linked sialic acid terminals, including NeuAcα2,3LacNAc-BSA (G4 in Fig. 2), NeuAcα2,3LacNAcβ1,2(NeuAcα2,3LacNAcβ1,4)Manα1,3(NeuAcα2,3LacNAcβ1,2(NeuAcα2,3LacNAcβ1,6)Manα1,6)Manβ1,4GlcNAcβ1,4GlcNAc-BSA (G5 in Fig. 2), and NeuGcα2,3Lac-BSA (G6 in Fig. 2).

FIG 2.

Receptor binding specificities of EV68. The receptor binding specificities of 12 strains of EV68 was measured for 6 types of sialyloligosaccharides by glycan array analysis. Samples were tested in triplicate, and three separate experiments were performed. The fluorescent intensity for each glycan was calculated as the average of 3 determinations.

Agglutination of modified erythrocytes.

The hemagglutination for enzymatically modified guinea pig erythrocytes were tested for a total of 8 strains, including 6 isolates of EV68, one isolate of influenza A virus (H1N1) pdm, and one isolate of parainfluenza virus type 1 (PIV1). The hemagglutination of those 8 strains originally ranged from 4 to 256 HAU (Table 2). After removal of SAs from the erythrocytes using Arthrobacter ureafaciens sialidase, hemagglutination for the 8 strains was no higher than 2 HAU (Table 2). When NeuAc was added to the sialidase-treated erythrocytes using α2,6-sialyltransferase, hemagglutination recovered completely to the original level for all EV68 strains and influenza A virus (H1N1) pdm but not for PIV1. In contrast, after addition of NeuAc to the sialidase-treated erythrocytes using α2,3-sialyltransferase, hemagglutination recovered completely to the original level only for PIV1, while that for EV68 and influenza A virus (H1N1) pdm was not altered (Table 2).

TABLE 2.

HA titers for the modified erythrocytes

Virus	Strain	HA titera
		Original	SD+
			ST−	α2-6 ST+	α2-3 ST+
EV68	Fermon	256	2	256	1
	Y10-2013	32	2	32	1
	Y10-2071	16	1	16	1
	Y10-2086	64	1	64	1
	Y10-2336	128	<1	128	1
	TTa-11-Ph224	4	<1	4	<1
H1N1pdm		16	<1	16	<1
PIV1		16	<1	<1	16

^aSD, Arthrobacter ureafaciens-derived sialidase; ST, sialyltransferase.

Detection of sialidase activity.

The sialidase activity was measured for 4 strains of EV68, including Fermon, Y10-2013 (lineage 1), Y10-2076 (lineage 3), and Y10-2082 (lineage 2). In addition to these EV68 strains, sialidase activity of one isolate of influenza A virus (H1N1) pdm was measured as a positive control, as well as the same volume of EMEM as a negative control. As a result, the fluorescence intensity for influenza A virus (H1N1) pdm strain increased in a virus titer-dependent manner, ranging from 438.13 at 1 HAU to 1,215.67 at 16 HAU, while that for the negative control was merely 40.80 (Fig. 3). The fluorescence intensity for EV68 did not increase in a virus titer-dependent manner and was not higher than 40 for any of the tested virus titer (Fig. 3).

FIG 3.

Sialidase activity of EV68 for the MUNANA substrate. The sialidase activities of four strains of EV68 for MUNANA were measured with 5 different HA levels. One strain of influenza A virus, (H1N1) pdm, was used as a positive control. Samples were tested in triplicate, and three separate experiments were performed. Error bars indicate standard errors of the mean.

Sequence analysis.

In complete VP2 and VP3 amino acid sequences, there were 32 amino acid positions where the recently isolated strains had amino acid residues different from those in strain Fermon, including 15 in VP2 and 17 in VP3 (Fig. 4). Among those 32 positions, recent strains had identical residues at 19, while EV68 strains of lineage 3 had different residues from the other 2 lineages at 12 positions (VP2, positions 73, 74, 98, 138, 144, 156, and 222; VP3, positions 60, 65, 73, 166, 208, and 216) from and those of lineage 2 at 1 position (VP2, position 135) (Fig. 4a and b).

FIG 4.

Complete VP1, VP2, and VP3 amino acid sequences of EV68 strains used for analysis. The complete amino acid sequences of the regions VP2 (a), VP3 (b), and VP1 (c) of EV68 strains used for analysis were aligned with that the prototype strain, Fermon (AF081348). The positions where the sequences of strains used for this analysis had amino acid residues identical to those of strain Fermon are indicated with dots. The BC and DE loop regions in VP1 are boxed.

In complete VP1 amino acid sequences, there were 35 amino acid positions where the recently isolated strains had unique amino acid residues compared to strain Fermon (Fig. 4c). Among those 35 positions, recent strains had identical residues at 18 positions, while EV68 strains of lineage 3 had different residues compared to the rest of the lineages at 6 positions (VP1, positions 46, 92, 131, 140, 145, and 178), those of lineage 2 at 5 positions (VP1, positions 5, 90, 95, 243, and 280), and those of lineage 1 at 2 positions (VP1, positions 168 and 169) (Fig. 4c). Among those 35 positions, 5 (VP1, positions 90, 92, 95, 97, and 99) were in the BC loop region and 8 (VP1, positions 140, 141, 142, 143, 144, 145, 148, and 152) were in the DE loop region (Fig. 4c). All strains of lineage 3 had one amino acid deletion at position 140 of VP1 (Fig. 4c). At positions 141 of VP1 and 152 of VP1, TTa-11-Ph344 of lineage 2 had Gly (G) and Asp (D), respectively, which were identical to all 4 strains of lineage 3, while all other strains of lineage 2 had Asp (D) and Asn (N) at those positions (Fig. 4c).

Codons under positive selection in VP1 capsid region.

Selection analysis using the SLAC method was conducted for 3 sets of sequence data consisting of partial VP1 sequences of lineages 1, 2, and 3. When the analysis was conducted for the data set of lineage 1, 28 amino acid positions were positively selected without statistical significance (Fig. 5a). Out of those 28 positions, 5 (positions 90, 92, 95, 97, and 99) were found in the BC loop region and 4 were in DE loop region (Fig. 5a). In the data set for lineage 2, total of 28 positions were under positive selection without statistical significance, and 3 positions (VP1, positions 92, 97, and 99) were in BC loop and 4 positions (VP1, positions 141, 142, 143, and 148) were in DE loop (Fig. 5b). In the data set for lineage 3, a total of 29 positions were under positive selection without any statistical significance, and 6 positions (VP1, positions 90, 92, 95, 97, 98, and 99) were in BC loop and 2 positions (VP1, positions 143 and 148) were in DE loop (Fig. 5c). A total of 17 positions, including VP1 positions 76, 84, 92 (BC loop), 97 (BC loop), 99 (BC loop), 143 (DE loop), 216, 218, 222, 224, 225, 226, 227, 228, 229, 235, and 236 were under positive selective pressure in all 3 data sets. Within the BC loop, VP1 positions 90 and 95 were positively selected for the data sets of lineage 1 and 3. In the DE loop, VP1 position 141 was positively selected for the data sets of lineage 1 and 2, and VP1 position 148 was positively selected for the data sets of lineage 2 and 3 (Fig. 5).

FIG 5.

Codons under positive selection in region VP1. Selection analysis was conducted for 3 sequence data sets, including EV68 strains of lineages 1, 2, and 3, by the SLAC method. Codons with dN/dS values higher than 0 were defined as being under positive selection. The amino acid sequences of the BC and DE loop regions are boxed.

DISCUSSION

We conducted HI and NT tests using 12 EV68 isolates and antisera generated for EV68 of lineage 1, 2, and 3. Among 12 EV68 strains, strain Fermon, the prototype strain isolated in California in 1962, showed limited cross-reactivity in both HI and NT tests using antisera for recently detected EV68 strains compared to other EV68 strains detected in Japan and the Philippines in recent years. This indicates that antigenically different EV68 strains compared to strain Fermon were circulating in Japan and the Philippines in 2010 and 2011. In one previous study from Finland, it was reported that the mean NT titers of serum from pregnant women against strain Fermon were lower in samples which were collected more recently (26). In addition, some of the previous studies using the Markov chain Monte Carlo approach suggested that EV68 went through extensive genetic evolution in the 1990s and 2000s, which may have resulted in the emergence of three lineages (6, 8). It may be possible that those genetic evolutions resulted in the emergence of strains with antigenic differences from previously circulating strains in recent years.

In the current study, we compared the HI and NT titers of strain Fermon with those of EV68 strains recently detected in Japan and the Philippines. It was presumed that the possibility of antigenic differences due to geographical distance is minimal, since the previous studies from other parts of the world identified strains with VP1 sequences similar to those detected in our study in Asia, among each of the genetic groups of lineages 1, 2, and 3 (6, 27). Therefore, it is assumed that our results can also be generalized for EV68 strains detected in other countries. However, no EV68 isolates detected before the recent outbreak were available except for strain Fermon. Therefore, the degree of antigenic changes that occurred just before the increased detection in recent years could not be studied here. Further analysis using EV68 detected in 1990s and the early 2000s is therefore required to make conclusions about such changes.

When the HI and NT titers were compared between EV68 strains of three lineages detected from Japan and the Philippines in 2010–2011, significant differences in the titers were found between lineages by HI and NT tests. This suggests that although each of the antisera generated for three lineages was cross-reactive to the strains of other lineages, there were some antigenic differences between lineages 1, 2, and 3 of EV68, although they belong to the single homologous serotype. Several previous studies have shown that multiple lineages of EV68 were detected in each outbreak (6, 8, 10). The emergence and cocirculation of strains with various antigenic properties in the same community may be a possible mechanism for the rapid increase of EV68 cases in recent years. However, the epidemiological significance of cocirculation of different lineages of EV68 has not been clearly defined, due to the low detection rate of the virus as well as the limited number of reports on the long-term epidemiological study of EV68. Further studies are therefore required to define the epidemiological importance of cocirculation of strains with the antigenic differences. Moreover, in this study, we used EV68 strains that we successfully isolated, as well as the Fermon strain purchased from ATCC, so the number of available EV68 isolates was limited to 12. In Fig. 1, error bars indicate 95% confidence intervals, and some of them have wide ranges, presumably because the number of isolates included in each lineage group was small (no more than 4). Therefore, a larger number of EV68 isolates is further required to obtain more representative data.

In this study, two guinea pigs were immunized with EV68 strains to generate polyclonal antibodies for each of three strains. Between the guinea pigs in each pair, there was no difference in the reactivity of obtained polyclonal antibodies against viruses tested in the study (data not shown). Therefore, we assume that effect of differences in the capacity of producing reactive antibodies between immunized animals was minimal in the study.

When the amino acid sequences in VP2, VP3, and VP1 detected in recent years were compared with those of strain Fermon, amino acid substitutions compared to Fermon were observed in all 3 of the regions VP2, VP3, and VP1. Notably, more than half of these amino acid substitutions were found in VP1. When the amino acid sequences were compared among the EV68 strains detected in recent years, strains of lineage 3 were most likely to have unique amino acid mutations which differentiated them from other lineages in any of the 3 regions. On the contrary, unique amino acid mutations for lineage 1 strains were found only at 2 amino acid positions in VP1, but not in VP2 and VP3. In HI and NT tests, strains of lineage 3 were shown to have different antigenicity from all other lineages. These unique amino acid mutations for lineage 3 might be associated with the varied antigenic properties of lineage 3 strains.

In the present study, we used EV68 strains which were passaged up to 6 times on RD cells. However, it is worth mentioning that amino acid substitutions were not observed in capsid regions after passages (data not shown). Therefore, we assume that the effect of virus passages on RD cells was minimal in the study.

In selection analysis for VP1 amino acid sequences, a total of 17 amino acid positions were under positive selection in all 3 lineages, among which 6 positions (positions 76, 84, 92, 97, 99, 143, and 218) had amino acid substitutions compared to the VP1 sequence of the Fermon strain. Those 6 positions might be associated with antigenic epitopes of EV68; however, the presence of other epitopes is also possible, since strains of 3 lineages had identical amino acid residues at all of those 6 positions despite the antigenic differences between lineages shown in this study. A total of 11 positions at the 3′ end of VP1 (positions 216, 218, 222, 224, 225, 226, 227, 228, 229, 235, and 236) were found to be under positive selection in all 3 lineages; however, only one position of VP1, 218, had amino acid mutations. Therefore, the remaining 10 positions (positions 216, 222, 224, 225, 226, 227, 228, 229, 235, and 236) were assumed to be not associated with antigenicity determining sites of EV68. It is worth mentioning that a total of 8 positions in VP1, including positions 46, 90 (BC loop), 95 (BC loop), 141 (DE loop), 144 (DE loop), 145 (DE loop), 148 (DE loop), and 168, were positively selected in some but not all of the 3 lineages and had amino acid substitutions which differentiated the strains of one lineage from the others. Those 8 positions might also be associated with antigenic epitopes of EV68 and the antigenic differences among the strains of 3 lineages. In a previous study in Thailand, selection analysis was conducted for the data set containing VP1 sequences of all lineages, and as a result, a total of 7 amino acid positions were reported to be under positive selection in the region that was tested in our study (8). Out of those 7 positions, 5 were also found to be positively selected in our study. We additionally found several positively selected sites in the analysis of each of three data sets. Those variations between our study and the study conducted in Thailand might be due to the differences in the analyzed data set. Moreover, since we conducted the selection analysis on lineages 1, 2, and 3 separately, the number of sequences in each data set was relatively small compared to the size of data set analyzed in that study, which also might have contributed to the differences.

In this study, we also demonstrated that EV68 has higher affinity for α2-6 SAs than α2-3 SAs. In the glycan array assay, a total of 12 EV68 strains were tested, including the prototype strain, Fermon, detected in 1962, and 11 strains detected in Yamagata in 2010 and the Philippines in 2011, all of which showed specific affinity for α2-6 SAs. Despite the differences in amino acid sequence in the capsid region and the antigenicity between Fermon and EV68 strains detected in recent years, there were no differences in sialooligosaccharide-binding patterns between them. We also demonstrated that the hemagglutination activity of EV68 was fully dependent on the presence of α2-6 SAs but not α2-3 SAs. Considering that the HA titer of EV68 for sialidase-treated GPE was completely recovered by the addition of α2-6 SAs, it was presumed that α2-6 SA is a viral receptor of EV68.

It is noteworthy that there was minimal reactivity to α2-3 SAs detected in the glycan array analysis; however, affinity of EV68 for α2-3 SAs was not detected by assays using enzymatically modified erythrocytes. It could be possible that the assay using enzymatically modified erythrocytes was not sensitive enough to detect strains with reduced affinity. However, the clinical significance of the limited affinity of EV68 for α2-3 SAs present in glycan array analysis remains elusive.

In this study, we tested only for the binding affinity of the viruses for trisaccharides by using a glycan array. It was previously reported that a wide range of sialylated O- and N-glycans are extensively expressed on the human airway tract (28). Therefore, it might be possible that our method using trisaccharides failed to detect the differences in binding patterns which are present only in the naturally existing glycans in the human airway tract. Moreover, our study addressed only the binding specificity of EV68. Further studies are therefore required to reveal the role of sialic acids in the internalization and replication processes of EV68.

In the glycan array analysis, all strains of EV68 tested were shown to have affinity for both NeuAcα2,6LacNAc and NeuGcα2,6Lac. It was previously shown that human tissue possesses only a small concentration of NeuGc, which is equivalent to less than 0.1% of total SAs, while NeuGc is commonly found among nonhuman species (29, 30). Although EV68 has been detected only from humans, our result suggests that there might be some animal species that are susceptible to EV68 other than humans. However, we tested only for the artificial binding of the virus to trisaccharides; therefore, our results are not conclusive on this point.

Considering that all EV68 strains tested showed binding specificity to α2,6-linked sialic acid terminals, EV68 might have affinity for the upper respiratory tract. However, several previous studies reported that the detection rate of EV68 was significantly higher among patients hospitalized with severe lower-respiratory-tract infections than those visiting outpatient clinics with mild upper-respiratory-tract infections (6, 8, 10). Therefore, there might be unknown mechanisms for severe infections with EV68 other than the distribution of viral receptors.

In the species human enterovirus D (HEV-D), three serotypes, including EV68, EV70, and EV94 are known to causes illness in human (31, 32). The VP1 sequences, the most variable genome region of enteroviruses, EV68 were shown to be closely related to those of EV70 and EV94 (26, 32). It was previously reported that EV70 has affinity for α2-3 SAs (33). Although the genome regions of EV68 and EV70 associated with their affinity for SAs are still unknown, it might be possible that EV68 and EV70 share similar amino acid sequences, which are responsible for the receptor binding.

It is well known that influenza viruses have neuraminidase (sialidase) activity, by which the newly synthesized viral proteins are removed from SAs, and the virus progenies are released from the host cell surface (34). However, in our study, we did not detect such enzymatic activities for EV68 by the MUNANA assay. Several members of the family Picornaviridae are reported to have affinity for SAs (33, 35); however, it is unknown if they have sialidase activities. It was previously shown that, in contrast to enveloped viruses, nonenveloped viruses are released from the host cells by disintegration of the cells by using mechanisms such as cell lysis (36). On the other hand, influenza viruses are enveloped viruses, and the budding of the viral progenies is dependent on the interaction between the host cell membrane and the viral proteins, including HA and NA (34). The reasons for the lack of sialidase activity in EV68 strains might lie in those differences in the virus-release mechanisms.

In conclusion, our study showed that EV68 detected in recent outbreaks possesses highly divergent antigenicity compared to strain Fermon. The emergence of strains with different antigenicity was suggested as a possible mechanism for the recent increase in the rate of detection. However, further studies using strains detected at different time points between 1962 and 2010 are required for drawing any conclusions. Moreover, the antigenic differences were also detected among the strains of 3 genetic groups. The roles of those antigenic differences in the transmission dynamics of EV68 in the community also need to be further clarified.

We also demonstrated that EV68 preferably binds to α2-6 SAs without sialidase activities. In the study, we did not find any differences in the receptor binding properties among the EV68 strains; therefore, the association of the receptor recognition patterns with the recent EV68 outbreaks remains elusive. Furthermore, although our findings suggested that EV68 receptors are commonly distributed in the upper respiratory tract in humans, there is epidemiological and clinical evidence suggesting that EV68 is more likely to cause severe respiratory infections. Therefore, further studies are needed to gain an overview of pathogenesis for the severe respiratory illnesses associated with EV68 infections.