Virulence of Enterovirus A71 Fluctuates Depending on the Phylogenetic Clade Formed in the Epidemic Year and Epidemic Region

ABSTRACT

Although epidemics of hand, foot, and mouth disease (HFMD) caused by enterovirus A71 (EV-A71) have occurred worldwide, the Asia-Pacific region has seen large sporadic outbreaks with many severe neurological cases. This suggests that the virulence of the circulating viruses fluctuates in each epidemic and that HFMD outbreaks with many severe cases occur when highly virulent viruses are circulating predominantly, which has not been experimentally verified. Here, we analyzed 32 clinically isolated strains obtained in Japan from 2002 to 2013, along with 27 Vietnamese strains obtained from 2015 to 2016 that we characterized previously using human SCARB2 transgenic mice. Phylogenetic analysis of the P1 region classified them into five clades belonging to subgenogroup B5 (B5-I to B5-V) and five clades belonging to subgenogroup C4 (C4-I to C4-V) according to the epidemic year and region. Interestingly, clades B5-I and B5-II were very virulent, while clades B5-III, B5-IV, and B5-V were less virulent. Clades C4-II, C4-III, C4-IV, and C4-V were virulent, while clade C4-I was not. The result experimentally showed for the first time that several clades with different virulence levels emerged one after another. The experimental virulence evaluation of circulating viruses using SCARB2 transgenic mice is helpful to assess potential risks of circulating viruses. These results also suggest that a minor nucleotide or amino acid substitution in the EV-A71 genome during circulation causes fluctuations in virulence. The data presented here may increase our understanding of the dynamics of viral virulence during epidemics.

IMPORTANCE Outbreaks of hand, foot, and mouth disease (HFMD) with severe enterovirus A71 (EV-A71) cases have occurred repeatedly, mainly in Asia. In severe cases, central nervous system complications can lead to death, making it an infectious disease of importance to public health. An unanswered question about this disease is why outbreaks of HFMD with many severe cases sometimes occur. Here, we collected EV-A71 strains that were prevalent in Japan and Vietnam over the past 20 years and evaluated their virulence in a mouse model of EV-A71 infection. This method clearly revealed that viruses belonging to different clades have different virulence, indicating that the method is powerful to assess the potential risks of the circulating viruses. The results also suggested that factors in the virus genome cause an outbreak with many severe cases and that further studies facilitate the prediction of large epidemics of EV-A71 in the future.

INTRODUCTION

Enterovirus A71 (EV-A71) is one of the serotypes belonging to the Enterovirus A (EV-A) species within the genus Enterovirus of the family Picornaviridae (1). The viral icosahedral capsid contains a positive-sense, single-stranded RNA genome. EV-A71 particles infect cells expressing human scavenger receptor class B member 2 (SCARB2) on the cell surface (2, 3). SCARB2 mediates three steps of viral infection: attachment of EV-A71 virions to the cell surface, internalization into cells, and initiation of uncoating, which releases genomic RNA into the cytoplasm. Next, one large open reading frame, called the polyprotein, is translated from the internal ribosomal entry site in the 5′ untranslated region (UTR) of the viral RNA. Proteolysis of the polyprotein generates three precursor proteins (P1 to P3), and further cleavage results in maturation into 11 different viral proteins. Four proteins (VP1 to VP4), arising from the P1 region, constitute the viral capsid, while seven proteins arising from P2 and P3 (nonstructural proteins) exert functions essential for viral replication (4).

Several EV-A members, including EV-A71, are causative agents of hand, foot, and mouth disease (HFMD). HFMD is a disease characterized by a rash on the hands, feet, and mouth of infants and young children. Although largely self-limiting, EV-A71-induced HFMD can cause central nervous system complications and even death. HFMD caused by EV-A71 is a disease of significant public health concern due to repeated large-scale outbreaks with fatal cases, especially in the Asia-Pacific region. EV-A71 is classified into eight genogroups (A to H) based on the nucleotide sequence of VP1. Genogroups B and C diverged into subgenogroups B1-5 and C1-5, respectively. In past epidemics, different genogroups were prevalent in different years and regions (5). An outbreak of subgenogroup B3, B4, and C1 occurred in Sarawak, Malaysia, in 1997, with 29 reported deaths (6, 7). In 1998, an outbreak of subgenogroup C2 caused 78 deaths in Taiwan (8–10). Between 2000 and 2002, Taiwan experienced another major outbreak caused by subgenogroup B4, which killed 129 people (11, 12). From 2008, China had the largest outbreak of subgenogroup C4, with 3,667 deaths reported by 2018 (13). Vietnam in 2011 and Cambodia in 2012 also experienced outbreaks of subgenogroup C4, with 166 and 64 deaths, respectively (14–16). Thus, EV-A71 outbreaks are caused not by a specific genogroup but rather by various genogroups. However, Japan has seen epidemics caused by various genogroups (B2, B4, B5, C1, C2, and C4) since the 1980s (17, 18), but no outbreaks of EV-A71 with a large number of deaths have occurred.

In Japan, epidemics of EV-A71 occur every 3 to 4 years. A simulation study suggested that the leading cause of such periodicity is the increase or decrease of susceptible individuals due to immune acquisition and birth (19). Other factors, such as climatic conditions, geographic factors, cultural customs, genetic background, and the medical systems, may also be associated with the region-specific epidemic cycle. In addition, viral factors such as alterations in virulence, transmissibility, and antigenicity may also play important roles. We consider that increased viral virulence is the most critical factor underlying the occurrence of epidemics with large numbers of fatalities; these outbreaks deviate significantly from the cyclical nature of regional epidemics.

Therefore, we hypothesized that the virulence of EV-A71 fluctuates independently of subgenogroup during circulation. When we consider the dynamics of viral transmission and genome sequence, the viral lineage maintains itself by accumulating mutations while infecting a small number of people before an epidemic occurs. At a certain point, the common ancestor strain spreads and forms a phylogenetic clade comprising genetically similar strains. In epidemics occurring at different times and in different regions, different strains become the common ancestors. Therefore, a single EV-A71 clade, or a mixture of a few clades, can cause an epidemic. The virulence of the circulating virus may fluctuate due to critical mutations and outbreaks, with many severe neurological diseases occurring when the common ancestor happens to acquire the virulent mutation or when a strain that acquired the virulent mutation from the avirulent ancestor emerges and becomes predominant during the early expansion phase of the epidemic.

To assess the viral virulence levels, we need to evaluate the virulence of each virus strain within each clade using an infection animal model with a uniform genetic and environmental background. To this end, we developed two essential techniques to accurately assess the neurovirulence of EV-A71: a cell line in which EV-A71 can grow with minimal attenuation and human SCARB2 transgenic (hSCARB2-tg) mice that are susceptible to infection by EV-A71. By propagating EV-A71 in standard cell lines, amino acid mutations associated with adaptation to heparan sulfate (HS) attachment receptors occur, thereby attenuating virulence in animals. To overcome this problem, we established RD-ΔEXT1+hSCARB2 to minimize these mutations (20). EV-A71 does not infect wild-type adult mice, but hSCARB2-tg adult mice are susceptible to EV-A71 and exhibit neurological disease upon EV-A71 infection similar to that in humans (21). Using these systems, we previously analyzed the virulence of EV-A71 isolated from patients with HFMD, which was prevalent in Hanoi, Vietnam, in 2015 to 2016 (22). During this period, most circulating strains belonged to subgenogroup B5, with a small number belonging to C4. The C4 strains were closely related to C4a lineage 3, isolated in China in 2013 to 2015. Patients infected with C4 strains were more likely to require hospitalization than those infected with B5. Furthermore, when we assessed the virulence of these two contemporaneous and differently derived genogroups in hSCARB2-tg mice, the C4 strains were more virulent than the B5 strains. The results suggest that virulence differs between genetically different EV-A71 lineages.

The present study aimed to explore differences in virulence between groups forming different clades in phylogenetic trees according to epidemic year and region for subgenogroup C4 and B5 strains isolated in Japan and Vietnam. The results suggest that different clades have different neurovirulence, even if they belong to the same subgenogroup. These findings may help to unravel the mechanism underlying severe outbreaks of EV-A71 showing fluctuating virulence in future studies.

RESULTS

Sample collection.

We focused on virus strains isolated during three epidemics of C4 and B5, which occurred in Japan in 2002 to 2003, 2006 to 2007, and 2012 to 2013 (18). EV-A71 strains isolated from nasopharyngeal or throat swab samples from patients in Miyagi and Yamagata prefectures diagnosed with HFMD, herpangina, viral exanthema, or stomatitis were collected. In Japan, there is no mandatory reporting system for complications of HFMD or herpangina, but no neurological complications were identified in the physicians' diagnoses. Miyagi and Yamagata prefectures are adjacent to each other and lie in the northeastern region of Japan; there is much human interaction between them. Most of the strains were isolated using green monkey kidney (GMK) cells, but some were isolated using human embryo fibroblast (HEF) cells or rhabdomyosarcoma (RD)-18S cells (Table 1). We performed two to four blind passages during virus isolation. Eleven and 21 clinical isolates belonging to subgenogroups B5 and C4, respectively, were grown once to a sufficient titer in RD-ΔEXT1+hSCARB2 cells (20) before further analysis. All Japanese strains were isolated from patients without neurological complications. We used strains from Vietnam, which were characterized previously, as controls (22). These strains were isolated from the throat or rectal swabs of patients with mild HFMD or herpangina (grade 1) or patients with neurological manifestations (grade 2 to 4) and propagated once in RD-ΔEXT1+hSCARB2 cells.

TABLE 1.

List of EV-A71 strains analyzed in this study

Strain	Accession no.	Place	Subgenogroup	Clade	Clinical information					Origin	Virulence		πc
					Collection yr	Gender	Age (mo)	Diagnosisa	Isolationb		Paralysis rate (%)	Mortality rate (%)
2716-Ymg-03	LC375766.1	Yamagata	B5	I	2003	F	6	HFMD	G3	Clone	90	90	0.00074
2000-Ymg-12	LC626900	Yamagata	B5	II	2012	F	1	Herpangina	G4	Clone	100	90	0.00055
K282-Snd-12	LC626873	Miyagi	B5	III	2012	M	1	HFMD	V2G1	Clone	80	20	0.00010
N343-Snd-12	LC626872	Miyagi	B5	III	2012	F	10	HFMD	V2	Clone	80	10	0.00011
N362-Snd-12	LC626875	Miyagi	B5	III	2012	F	1	HFMD	V2	Clone	90	50	0.00007
N377-Snd-12	LC626901	Miyagi	B5	III	2012	F	5	HFMD	V2	Isolate	70	50	0.00022
N380-Snd-12	LC626876	Miyagi	B5	III	2012	F	21	HFMD	V2	Clone	80	60	0.00011
N384-Snd-12	LC626877	Miyagi	B5	III	2012	M	5	HFMD	V2	Clone	70	60	0.00017
1859-Ymg-12	LC626879	Yamagata	B5	III	2012	F	2	HFMD	G4	Clone	70	10	0.00005
1888-Ymg-12	LC626878	Yamagata	B5	III	2012	F	0	Viral exanthem	G4	Clone	100	20	0.00014
1499-Ymg-13	LC626874	Yamagata	B5	III	2013	M	3	Stomatitis	G3	Clone	90	50	0.00007
15NHP019	LC627069	Hanoi	B5	IV	2015	F	30	1	E1	Isolate	70	30	0.00028
15NHP105	LC627066	Hanoi	B5	IV	2015	M	28	1	E1	Isolate	70	10	0.00034
16NHP373	LC627078	Hanoi	B5	IV	2016	M	29	2	E1	Isolate	100	60	0.00051
16NHP399	LC627079	Hanoi	B5	IV	2016	M	14	2	E1	Isolate	10	0	0.00066
16NHP450	LC627081	Hanoi	B5	IV	2016	M	14	2	E1	Isolate	40	30	0.00017
16NHP453	LC627082	Hanoi	B5	IV	2016	F	31	2	E1	Isolate	60	20	0.00030
16NHP456	LC627083	Hanoi	B5	IV	2016	F	8	2	E1	Isolate	100	40	0.00025
15NHP003	LC627067	Hanoi	B5	V	2015	M	38	1	E1	Isolate	90	30	0.00033
15NHP009	LC627068	Hanoi	B5	V	2015	M	28	1	E1	Isolate	70	40	0.00014
15NHP023	LC627070	Hanoi	B5	V	2015	M	25	1	E1	Isolate	70	10	0.00021
15NHP088	LC627071	Hanoi	B5	V	2015	F	37	1	E1	Isolate	90	30	0.00008
15NHP108	LC627072	Hanoi	B5	V	2015	M	27	1	E1	Isolate	60	0	0.00038
15NHP120	LC627073	Hanoi	B5	V	2015	M	18	4	E1	Isolate	80	10	0.00027
15NHP128	LC627074	Hanoi	B5	V	2015	M	20	1	E1	Isolate	90	70	0.00039
15NHP152	LC627075	Hanoi	B5	V	2015	F	15	1	E1	Isolate	90	20	0.00017
15NHP157	LC627076	Hanoi	B5	V	2015	F	17	1	E1	Isolate	80	50	0.00025
16NHP226	LC627077	Hanoi	B5	V	2016	F	24	2	E1	Isolate	70	10	0.00023
16NHP400	LC627080	Hanoi	B5	V	2016	M	1	2	E1	Isolate	50	20	0.00037
16NHP462	LC627084	Hanoi	B5	V	2016	M	16	2	E1	Isolate	100	60	0.00030
2779-Ymg-02	LC626892	Yamagata	C4	I	2002	M	6	Viral exanthem	G4	Isolate	85	85	0.00018
100-Ymg-03	LC626899	Yamagata	C4	I	2003	F	9	HFMD	G3	Isolate	10	0	0.00014
452-Ymg-03	LC626891	Yamagata	C4	I	2003	M	4	HFMD	G4	Isolate	90	40	0.00045
1138-Ymg-03	LC626898	Yamagata	C4	I	2003	F	5	HFMD	G3	Isolate	30	20	0.00039
K258-Snd-03	LC626889	Miyagi	C4	I	2003	M	3	HFMD	H1G1	Clone	50	10	0.00002
K367-Snd-03	LC626888	Miyagi	C4	I	2003	M	2	HFMD	V1G1	Isolate	70	40	0.00080
K399-Snd-03	LC626886	Miyagi	C4	I	2003	M	2	HFMD	H2G1	Isolate	70	50	0.00027
K404-Snd-03	LC626885	Miyagi	C4	I	2003	F	2	HFMD	H2G1	Isolate	80	50	0.00090
N803-Snd-03	LC626881	Miyagi	C4	I	2003	F	2	HFMD	H2	Isolate	90	70	0.00022
N826-Snd-03	LC626880	Miyagi	C4	I	2003	F	8	HFMD	H2G1	Isolate	50	40	0.00020
K493-Snd-03	LC626884	Miyagi	C4	I	2003	M	1	HFMD	H1G1	Isolate	60	30	0.00047
1934-Ymg-03	LC626896	Yamagata	C4	I	2003	M	1	HFMD	G3	Isolate	30	30	0.00054
K528-Snd-03	LC626883	Miyagi	C4	I	2003	M	3	HFMD	V1G1	Isolate	100	80	0.00023
N772-Snd-06	LC506513.1	Miyagi	C4	II	2006	M	4	Herpangina	H2	Isolate	90	90	0.00062
N133-Snd-06	LC626882	Miyagi	C4	III	2006	M	4	HFMD	H1G1	Isolate	100	90	0.00058
2317-Ymg-06	LC626895	Yamagata	C4	III	2006	M	3	Viral exanthem	G3	Isolate	90	80	0.00025
2488-Ymg-06	LC626894	Yamagata	C4	III	2006	M	1	HFMD	R2G1	Isolate	90	60	0.00076
2498-Ymg-06	LC626893	Yamagata	C4	III	2006	M	2	HFMD	G3	Isolate	100	90	0.00041
K398-Snd-06	LC626887	Miyagi	C4	III	2006	F	1	HFMD	H2	Isolate	80	20	0.00041
114-Ymg-07	LC626897	Yamagata	C4	III	2007	F	2	HFMD	R2G1	Isolate	100	100	0.00030
56-Ymg-07	LC626890	Yamagata	C4	III	2007	M	2	HFMD	G3	Isolate	85	85	0.00054
15NHP107	LC627085	Hanoi	C4	IV	2015	M	10	1	E1	Isolate	100	100	0.00053
16NHP379	LC627086	Hanoi	C4	V	2016	M	29	2	E1	Isolate	90	90	0.00036
16NHP391	LC627087	Hanoi	C4	V	2016	M	13	2	E1	Isolate	60	60	0.00042
16NHP417	LC627088	Hanoi	C4	V	2016	M	16	2	E1	Isolate	80	80	0.00010
16NHP436	LC627089	Hanoi	C4	V	2016	M	23	2	E1	Isolate	70	60	0.00015
16NHP442	LC627090	Hanoi	C4	V	2016	M	12	2	E1	Isolate	90	90	0.00012
16NHP455	LC627091	Hanoi	C4	V	2016	F	23	2	E1	Isolate	90	80	0.00023
16NHP475	LC627092	Hanoi	C4	V	2016	F	21	2	E1	Isolate	90	80	0.00032

^aFor strains isolated in Japan (Miyagi and Yamagata), the name of the clinicians' diagnoses was noted, and for strains isolated in Vietnam, the clinical grade according to the Vietnamese HFMD Diagnostic Guidelines 2012 was noted. Grade 1 is the presence of herpangina or skin rash only; grade 2 is the presence of myoclonic jerk; grade 3 is the presence of autonomic dysfunction, with intractable fever and shock symptoms in the early stages; and grade 4 is the presence of cardiopulmonary collapse.

^bG, GMK; H, HEF; R, RD-18S; E, RD-ΔEXT1+hSCARB2. The numbers indicate the number of passages.

^cThe mean pairwise difference per site.

Phylogenetic analysis of the genome sequence of the EV-A71 isolates.

We hypothesized that viral virulence does not vary according to subgenogroup but depends on clades that appear one after another during each epidemic. Therefore, we decided to analyze the virulence of each clade. To examine clade organization, we performed next-generation sequencing (NGS) and assembled consensus sequences for the nearly complete length viral genomes of all 32 strains (11 and 21 were B5 and C4 strains, respectively). Next, we constructed a phylogenetic tree using these nucleotide sequences for each genetic region (Fig. 1). We also included 27 strains from Hanoi, Vietnam, which were isolated in 2015 to 2016 and previously analyzed by us (22), along with reference strains from each genogroup. The phylogenetic tree of the P1 region (Fig. 1A) revealed that subgenogroups B5 and C4 each comprised five clades. These were defined as clades B5-I to B5-V and C4-I to C4-V. The median nucleotide sequence similarity for P1 genome regions within the same clade was 99.76% (maximum, 99.90%; minimum, 99.67%) (Fig. 2A), whereas the median sequence similarity for these sequences among different clades was 97.07% (maximum, 99.59%; minimum, 94.88%) (Fig. 2B). Based on these results, it can be assumed that strains with more than 99.6% sequence similarity in the P1 region are the same clade.

FIG 1.

Phylogenetic analysis of the EV-A71 strains analyzed in this study. Phylogenetic trees were constructed using the maximum likelihood (ML) method for the P1 region (A), the 5′ UTR region (B), the P2 region (C), and the P3 region (D) of the 32 EV-A71 isolates collected in Japan in this study, for the 27 isolates collected in Vietnam in our previous study, and for the reference sequences for each subgenogroup. The number shown for each node in panel A is the bootstrap value of 1,000 replicates. Each clade is shown in a different color, and the clade name is shown on the right.

FIG 2.

Percent sequence similarity in nucleotide sequence of P1 region within clades (A) and between clades (B). Values with the highest similarity in each panel are shown in red, and values with the lowest similarity are shown in blue.

This method successfully grouped the strains in these clades by epidemic year and epidemic region. Each epidemic analyzed in this study comprised more than two clades (Fig. 3). In the 2002–2003 epidemic, C4-I was the predominant clade (14 strains), and B5-I, which contained only one strain, was the minor clade. In the 2006–2007 epidemic, C4-III was the major clade (seven strains), and C4-II was the minor clade (one strain). In the 2012–2013 epidemic, B5-III was the major clade (nine strains), and B5-II was the minor clade (one strain). In the 2015–2016 epidemic in Vietnam, we identified B5-IV (seven strains), B5-V (12 strains), C4-IV (one strain), and C4-V (seven strains).

FIG 3.

Summary of the year of isolation of strains belonging to each clade. The numbers shown indicate the number of strains.

Phylogenetic trees based on the sequence of the 5′ UTR and P2 regions maintained the clade organization shown for the P1 region (Fig. 1B and C). However, two strains (16NHP436 and 16NHP442), classified as C4-V, formed a different clade in the phylogenetic tree for the P3 region (Fig. 1D). Our previous study showed that the 3D polymerase-coding regions of these two strains are recombinants with coxsackievirus A8 (22). Therefore, except for these two strains, none of the virus strains analyzed here contained a genome that had recombined with different EV-A71 genogroups or different EV-A serotypes.

To better understand the Japanese strains that contained only one strain per clade in the above clade classification (2716-Yamagata-2003 [2716-Ymg-03] in B5-I; 2000-Ymg-12 in B5-II; and N772-Sendai-2006 [N772-Snd-06] in C4-II), we explored viruses closely related to these strains to identify the origin of the strains. We used the sequences of the VP1 region of these strains as a query in a nucleotide BLAST search for homologous EV-A71 sequences deposited in the GenBank database. We constructed a molecular phylogenetic tree by combining the sequence data for the top 100 sequences, the sequence data for the VP1 region of the strains from Japan and Vietnam in this study, and subgenogroup A as an outgroup (Fig. 4). As a result, we identified strains closely related to each of the three strains. Only the Japanese strains isolated in 2003 are closely related to B5-I (2716-Ymg-03) (green area in Fig. 4) (18, 23). In 2003, C4 caused an epidemic in Japan, but a small amount of B5 was detected (18). The 2716-Ymg-03 strain is likely one of the few B5 strains detected in Japan in 2003. In contrast, B5-II (2000-Ymg-12) showed a close relationship with two strains (AB936545.1 and AB936555.1) isolated in Japan in 2012 and with six strains isolated in Taiwan in 2012 (yellow area in Fig. 4); these are genetically different from other B5-III strains detected in Japan in the same year. Therefore, 2000-Ymg-12 is likely imported from Taiwan. Strain N772-Snd-06 (C4-II) is closely related to strains isolated in mainland China and Hong Kong from 2003 to 2009 (cyan area in Fig. 4); we found no Japanese strains that were very similar to N772-Snd-06 in the database. Therefore, it is probable that the N772-Snd-06 strain was imported into Japan from mainland China or Hong Kong but did not spread much. In summary, the prevalence of EV-A71 in Japan may involve both Japan-specific clades and clades imported from abroad.

FIG 4.

Analysis of viruses closely related to 2716-Ymg-03 (B5-I), 2000-Ymg-12 (B5-II), and N772-Snd-06 (C4-II). We inferred a phylogenetic tree using the ML method for the VP1 sequences of 32 Japanese strains, 27 Vietnamese strains, subgenogroup A, and the sequences of about 100 strains with VP1 sequences similar to those of each strain, as identified by nucleotide BLAST. The magenta and orange lines indicate Japanese strains identified in this study and the Vietnamese strains identified in the previous study, respectively. The cyan, green, and yellow areas indicate strains closely related to N772-Snd-06, 2716-Ymg-03, and 2000-Ymg-12, respectively. The magnified views are shown bottom left.

Virulence analysis.

The virus strains obtained in this way may contain several mutations that are artifacts caused during the propagation in the cell culture (20) and give bias in the virulence analyses (24, 25). To find out such mutations in the samples for virulence assessment, we conducted a single-nucleotide variation (SNV) analysis of the viral genomes of all Japanese and Vietnamese strains based on NGS data. The result showed that 10 of 30 B5 strains, and 1 of 29 C4 strains, contained detectable (maximum, 50.0%; minimum, 0.408%) glutamic acid (E) to glycine (G) or glutamine (Q) mutations at amino acid 145 of the capsid protein VP1 (VP1-145), or an E to lysine (K) mutation at VP1-98. Previous reports show that viruses harboring E to G or Q mutations at VP1-145, an E to K mutation at VP1-98, and a leucine to arginine mutation at VP1-97 can bind to the HS attachment receptor (26–29). We also showed that cultivation of EV-A71 in RD or other human cells, or monkey cells, often resulted in such HS-binding mutations and that the presence of even a tiny number of mutants in a sample results in a significant bias in the viral titer, thereby reducing the apparent virulence of EV-A71 in vivo (20). We consider these variants to be artifacts that occurred during conventional isolation procedures in HEF or GMK cells. Therefore, we constructed cDNA clones for the strains with detectable levels of HS-binding mutations (10 strains of B5 and one strain of C4; Table 1) and prepared new virus stocks by transfecting in vitro-transcribed RNA into RD-ΔEXT1+hSCARB2 cells. This method can minimize the introduction of unnecessary mutations (20). We confirmed the absence of HS-binding mutations in the clone-derived viral stocks by NGS analysis followed by SNV analysis. We used the viral stocks that contained HS-binding variants below the detection limit for virulence analyses without cloning (one strain of B5 and 20 strains of C4 subgenogroups; Table 1).

There are reports that heterogeneity of viral genomes is associated with virulence (30). Viruses isolated from patients have a certain degree of heterogeneity, but cloning is expected to reduce heterogeneity. Thus, we examined the genome heterogeneity of viruses derived from the clones and those not from the clones. NGS reads of each virus sample were analyzed by the SNPgenie software (31) to obtain the mean pairwise difference per site (π), which was used as an index of genome heterogeneity (Table 1). As a result, the mean value of π for isolates not derived from clones was 0.000349, and that for the clone-derived strain was 0.000195. The Student's t test comparison suggested a significant difference (P = 0.022), but the π values were quite comparable (the difference was less than twofold). Thus, we consider that the possible bias caused by cloning is slight and that cloning is a much better strategy than using heterogeneous virus-containing HS-binding mutations, which is known to give a significant bias to the results.

We then evaluated their virulence by representing the mortality after inoculating (500,000 50% tissue culture infective doses [TCID₅₀]/mouse) them intraperitoneally into a group of 10 (5 male and 5 female) 6- to 7-week-old hSCARB2-tg mice. We noted that the mortality varied from 0% to 100% depending on the virus strains (Fig. 5). In the analysis of Vietnamese strains, we found a tendency that the C4 strains that had higher disease severity in humans showed more than 60% mortality in hSCARB2-tg mice, while B5 strains that had low disease severity in humans showed less than 60% mortality in hSCARB2-tg mice (22). In this study, we defined the viruses with a mortality rate of more than 60% as virulent, and viruses with a mortality rate of less than 60% were avirulent based on the result of our previous experiment. We found that viruses belonging to some clades showed similar virulence levels. All seven strains of clade C4-V caused high mortality, ranging from 60 to 90% (median, 80%). Six out of seven strains of C4-III also showed high mortality, ranging from 60 to 100% (median, 80%). Clades B5-I, B5-II, C4-II, and C4-IV each contained a single strain that showed high virulence (mortality rates of 90%, 90%, 90%, and 100%, respectively). These results suggest that the virulence of these clades is relatively homogeneous and potent. In contrast, clades B5-III, B5-IV, and B5-V resulted in less than 60% mortality (median, 50%, 30%, and 25%, respectively). Unlike other clades, C4-I strains showed variable mortality, ranging from 0 to 85% (median, 40%), suggesting that this clade is heterogeneous.

FIG 5.

Evaluation of neurovirulence of EV-A71 strains isolated in Japan and Vietnam. Thirty-two strains of EV-A71 collected in Japan were inoculated (50,000 TCID₅₀/mouse) into 10 hSCARB2-tg mice aged 6 to 7 weeks. Mice were monitored for 2 weeks after that. The plots show the mortality rates for each strain by clade, including data from isolates collected previously in Vietnam and studied under the same conditions (22). Circles indicate experimental groups in which hSCARB2-tg mice were directly inoculated with strains isolated from patients, and triangles indicate experimental groups where HS binding mutants were excluded by cloning. Pairwise comparisons were made using the Wilcoxon rank-sum test to determine statistically significant differences among clades containing multiple strains. Pairs with P values of <0.05 are shown.

Finally, we used the Wilcoxon rank-sum test to perform a pairwise comparison of clades containing multiple strains (clades B5-III, -IV, and -V and C4-I, -III, and -V). The results showed that C4-III and C4-V were significantly more virulent than any of the B5 clades. The C4-V strains were significantly more virulent than the C4-I strains, and although it did not meet the 0.05 level of significance (P = 0.064), there was a twofold median difference in virulence between C4-I and C4-III (40% and 80% for C4-I and C4-III, respectively). These results suggest that virulence differs among clades.

DISCUSSION

Large outbreaks of HFMD associated with many severe cases occasionally occur by various subgenogroups of EV-A71. We hypothesized that viral virulence fluctuates from epidemic to epidemic, and that there may be a large number of severe cases when virulent strains appear and become prevalent and when the number of infected people is high. However, there is no experimental evidence to support this hypothesis. The patient outcome depends on many factors, including viral virulence, the dose of the virus to which they are exposed, the genetic background of the hosts, the temporal status of innate and acquired immunity, climate conditions, and medical treatment. For these reasons, even if viral virulence is the most fundamental factor, it is difficult to evaluate it when analyzing epidemiological data and human cases. In addition, many patients showed only mild clinical signs despite being infected with a highly virulent strain because of the low incidence of developing severe neurological diseases in EV-A71-infected patients. Therefore, it is not possible to evaluate the virulence of the viruses using the symptoms of patients as an indicator. To accurately analyze the virulence of individual epidemics, we established a cell culture method that enables the propagation of unattenuated EV-A71 strains and an hSCARB2-tg mouse model to evaluate the virulence of EV-A71. These systems make it possible to examine intrinsic viral virulence in a genetically homogeneous model without bias caused by host and environmental factors.

Using these systems, we compared the virulence of 32 Japanese and 27 Vietnamese EV-A71 strains belonging to 10 clades within subgenogroups B5 and C4. The results showed that we could discriminate the virulence levels of the clinically isolated strains experimentally, and there are both virulent and less virulent strains. It is of note that the virulence level was almost constant within a clade. Clades B5-I and B5-II were virulent, while B5-III, B5-IV, and B5-V were less virulent. Clades C4-II, C4-III, C4-IV, and C4-V were virulent, while C4-I was not. We analyzed only a limited number of virus strains and clades, none of which were involved in a large outbreak with severe cases. However, the results clearly show that the virulence of EV-A71 is independent of subgenogroup and that it fluctuates clade by clade, supporting our hypothesis of dynamic viral virulence during epidemics. The results indicate that we can directly detect the circulation of dangerous viruses using this system. Epidemiological studies with a sufficient number of infected people, supported by accurate clinical data from past or future outbreaks, will further confirm the validity of this experimental method and our conclusions.

It is worth discussing the relationship between epidemics and the virulence of viruses circulating in Japan. No large outbreaks were associated with a large number of severe cases in Japan. In the epidemic of 2002 to 2003, clade C4-I, which is not virulent, was the primary circulating virus. Clade B5-I, which is virulent, did not become predominant. The predominance of less virulent clades may be one reason why many severe cases did not occur. In the 2006–2007 epidemic, the relatively virulent clade C4-III dominated but without causing many severe cases. According to the National Epidemiological Surveillance of Infectious Diseases in Japan, there were fewer reports of HFMD caused by EV-A71 in 2006 to 2007 compared with the epidemics of EV-A71 of 2003 and 2010 (32, 33), which may have contributed to the small number of severe cases. Clades B5-II and B5-III caused the epidemic in 2012 to 2013. Strain 2000-Ymg-12 (B5-II) is likely to be an imported strain from Taiwan. Taiwan experienced a relatively large outbreak of EV-A71, with many severe cases, in 2012 (34), suggesting that B5-II strains, including 2000-Ymg-12, were highly virulent. However, this strain did not spread to the majority of the population. N772-Snd-06 (C4-II clade) is, presumably, a strain imported into Japan from mainland China or Hong Kong; however, it did not spread much in Japan. Previously, we demonstrated the importation of some EV-A71 strains from foreign countries and that subgenogroup exchanges occurred in a local region (Yamagata) as a result (23). The present study shows another possibility: that the imported strains contribute to changes in virulence levels.

After we confirmed that the virulence of the viruses fluctuates along with circulation, the next goal is to determine what mutations are responsible for this fluctuation. Since the sequence homology is very high, even among the different clades, a small number of mutations in the viral genome, apart from the distinction of genogroup and subgenogroup, may control viral virulence. The virulence determinants of poliovirus are most successfully analyzed (35–41). However, the situation in the search for virulence determinants of EV-A71 is different from that of poliovirus. Poliovirus virulence determinants were identified by comparing the highly attenuated Sabin strains and corresponding parental virulent strains (or revertant strains). Both belong to the same lineages (type 1 and type 3) or closely related virulent viruses (type 2). The studies revealed a common mutation among three serotypes; the mutations in stem-loop V in the 5′-UTR decrease translation initiation efficiency. In addition to this, several mutations are found in a certain serotype that is specific to each lineage. The mechanism of attenuation by these mutations is not common among the three lineages. On the other hand, representative strains for virulent and avirulent phenotypes of EV-A71 should be selected first. The virulent EV-A71 strains found in this study are not as virulent as wild-type poliovirus, and avirulent EV-A71 strains are not as highly attenuated as Sabin strains. Even if we could choose sufficiently virulent and avirulent EV-A71 strains, their relationship is not that of an ancestor and its descendant but rather two or more lineages evolved in parallel under different environments. These situations might make it difficult to perform experiments and draw a clear conclusion. We may find a virulent mutation common to all EV-A71 lineages if one or a few specific mutations repeatedly appear in virulent clades. Alternatively, we may find different mutations for each clade if different kinds of critical mutations occur in every lineage if the virulence levels continue to drift. In any case, we need to try at least several combinations of virulent and avirulent strains to elucidate the mechanisms of EV-A71 virulence comprehensively.

To this end, we are currently working to identify genomic positions responsible for the differing virulence of a pair of two C4 strains and another pair of B5 strains. Comparing multiple combinations of different lineages may be necessary before we can find a set of mutations that can reasonably explain the overall EV71 pathogenicity. If the factors that determine the virulence of EV-A71 in animal models become apparent, retrospective epidemiological studies and surveys of future emerging EV-A71 outbreaks may reveal whether these factors can be applied to virulence in humans. If such information is confirmed, it may be possible to foresee and prevent, at an early stage, large outbreaks caused by EV-A71.

MATERIALS AND METHODS

Ethics statements.

All experiments using recombinant DNA and pathogens were approved by the Committee for Experiments using Recombinant DNA and Pathogens at the Tokyo Metropolitan Institute of Medical Science (approval numbers 15-023, 15-049, 18-033, 18-035, and 19-009). All animal experiments were approved by the Animal Use and Care Committee (approval numbers 17036, 18047, and 19063) and performed by following the Guidelines for the Care and Use of Animals (42). Mice exhibiting severe paralysis in more than two limbs or losing more than 30% of their body weight during the infection experiments were sacrificed using an overdose of isoflurane or by cervical dislocation. Inclusion of patient information (sex, age, and diagnosis) in the article was approved by the Research Ethics Committee of the Tokyo Metropolitan Institute of Medical Science (approval number 17-4), the Yamagata Prefectural Institute of Public Health (approval number YPIPHEC 20-05), Sendai Medical Center (approval number 26-04 and 29–35), Kanazawa University (approval number 1611-1), and the National Hospital of Pediatrics and Research Institute for Child Health (approval number 14-012).

Cells.

RD-A cells were cultured in Dulbecco's modified Eagle medium (DMEM) containing 5% fetal calf serum (FCS). RD-ΔEXT1+hSCARB2 cells were established previously (20) and cultured in DMEM containing 5% FCS and 1 μg/ml puromycin. GMK, HEL, and RD-18S cells were cultured at 33°C in MEM containing 2% FCS.

Virus isolation and propagation.

The virus isolation process was described previously (18). Briefly, EV-A71 strains were isolated in GMK, HEF, and/or RD-18S cell lines. The blind passage was repeated several times until cytopathic effects appeared. Isolates were identified as EV-A71 using a neutralization method. The VP1 region of some of the strains analyzed in this study was already sequenced (18). The isolated virus was inoculated into RD-ΔEXT1+hSCARB2 cells to produce approximately 20 ml of virus stock used for all titrations, animal infection experiments, and NGS analyses. Viral titers were measured using the microplate method in RD-A cells (43).

For viruses that harbor HS-binding variants, infectious cDNA clones were constructed to eliminate these variants. Full-length cDNAs were cloned into a pSVA vector as described previously (24). When cloning a viral genome, it is difficult to obtain a clone that matches the consensus sequence perfectly because the viral stocks contain a mixture of mutants that comprise quasispecies. Therefore, clones that matched the consensus sequence at the amino acid level (synonymous substitution was accepted) were prepared. For the noncoding region, clones with the same sequence as the consensus sequence were selected. The virus was recovered from infectious cDNA as described previously (24).

NGS analysis.

NGS analysis was undertaken as described previously (20). Briefly, RNA was extracted from virus stocks using the QIAamp viral RNA extraction minikit (Qiagen). All libraries were prepared using the NEBNext RNA library prep kit for Illumina and the NEBNext multiplex oligonucleotides for Illumina dual-index primer set 1 (from New England Biosciences). The libraries were sequenced on a MiSeq instrument (Illumina) using a 150-bp paired-end kit and then trimming adaptor and low-quality sequences from the raw sequence reads using PRINSEQ software, version 0.20.4 (http://prinseq.sourceforge.net/index.html). The remaining reads were mapped to the hg38 human reference genome, and reads derived from host cells were removed using Bowtie 2, software version 2.3.4.1, with default settings (44). Nearly complete consensus viral genomic sequences were assembled from the hg38 unmapped reads using IVA software, version 1.0.8, with default settings (45). The remaining reads were mapped to the assembled sequence using Smalt software version 0.7.6 (https://www.sanger.ac.uk/science/tools/smalt0) and converted to BAM files using SAMtools software version 1.9–4-gaelf9d8. The generated BAM files were used for variant calling using Lofreq software, version 2.1.2 (46), followed by annotating SNVs using SnpEff software version 4.3t (47). To evaluate the heterogeneity of the genome, we used SNPgenie software (31) to calculate the mean pairwise difference per site (π).

Phylogenetic analysis.

The nearly complete viral genome sequences assembled from the NGS reads and the viral nucleotide sequences belonging to each subgenogroup obtained from GenBank were used as references for the following analyses. These sequences were subjected to multiple-sequence alignment using the G-INS-i method in MAFFT, version 7 (48). The phylogenetic tree was inferred using the ML method in RAxML, version 8.2.12 (49). The nucleotide substitution model was GTRGAMMA, and 1,000 bootstraps verified all nodes. The VP1 region was entered into nucleotide BLAST as a query, and sequences of strains closely related to the specific strains shown in Fig. 4 were obtained. The nucleotide collection was chosen as the database and Enterovirus A71 (taxid 39054) as the organism. The phylogenetic tree was inferred as described above by adding the 100 most homologous sequences, excluding the query sequences themselves, registered in the database.

Calculation of sequence homology over sequence pairs within clades and between clades.

Using MEGA X (50), the number of base differences per site for all sequence pairs within and between each clade and averaged. Sequence homology was calculated based on these values.

Virulence analysis.

hSCARB2-tg mice (6 to 7 weeks old [n = 10 per strain]) were used for the infection experiments. Mice were anesthetized with isoflurane and inoculated intraperitoneally with a virus solution (50,000 TCID₅₀/mouse). Changes in body weight and clinical signs of disease were monitored for 2 weeks.

Statistical analysis.

Pairwise multiple comparisons were made using the Wilcoxon rank-sum test in R, version 3.5.1, and P values were adjusted by the false discovery rate (Fig. 5).

Data availability.

All nucleotide sequence data are available in GenBank under accession numbers LC375766.1, LC506513.1, LC626872–LC626901, and LC627066–LC627092.