Am. J. Trop. Med. Hyg., 65(3), 2001, pp.

242–251
Copyright 䉷 2001 by The American Society of Tropical Medicine and Hygiene

PHYLOGENETIC ANALYSIS OF JAPANESE ENCEPHALITIS VIRUS: ENVELOPE GENE

BASED ANALYSIS REVEALS A FIFTH GENOTYPE,
GEOGRAPHIC CLUSTERING, AND MULTIPLE INTRODUCTIONS OF THE
VIRUS INTO THE INDIAN SUBCONTINENT
PRADEEP D. UCHIL AND VIJAYA SATCHIDANANDAM
Department of Microbiology And Cell Biology, Indian Institute of Science, Bangalore, India

Abstract. We report the analysis of the complete nucleotide sequence for the Indian isolate (P20778; Genbank
Accession number AF080251) of Japanese encephalitis virus (JEV). The phylogenetic tree topology obtained using
thirteen complete genome sequences of JEV was reproduced with the envelope, NS1, NS3, and NS5 genes and
revealed extensive divergence between the two Indian strains included. A more exhaustive analysis of JEV evolution
using 107 envelope sequences available for isolates from different geographic locations worldwide revealed five
distinct genotypes of JEV, displaying a minimum nucleotide divergence of 7% with high bootstrap support values.
The tree also revealed overall clustering of strains based on geographic location, as well as multiple introductions of
JEV into the Indian subcontinent. Nonsynonymous nucleotide divergence rates of the envelope gene estimated that
the ancestor common to all JEV genotypes arose within the last three hundred years.

INTRODUCTION nese SA-14 isolate.15 Most of these studies utilized uncor-

Japanese encephalitis virus (JEV) is the cause of the most
prevalent viral encephalitis of man in terms of morbidity and
mortality.1 Approximately 50,000 human cases of JE occur
annually in Asia.2 Japanese encephalitis virus belongs to the
family Flaviviridae, whose members include several patho-
gens of humans and animals. In India, epidemics of JEV
have been reported since the mid-1950s, and the virus is now
endemic in most parts of the country. The genomes of fla-
viviruses comprise single-stranded RNA of positive polarity
approximately 11,000 nucleotides (nt) in length. The struc-
tural proteins: capsid, premembrane, and envelope, are en-
coded in the 5⬘ third of the genome and are followed by the
genes specifying the seven nonstructural proteins.
Japanese encephalitis virus is one of the recently diverged
members of the family Flaviviridae, and most JEV isolates
are estimated to have evolved over the last 130 years.3 The
study of flavivirus phylogeny has primarily utilized the
structural glycoprotein envelope sequences.4 These, in ad-
dition to serological studies,5 have elegantly shown differ-
ences in evolutionary dynamics of the tick- and mosquito-
borne subgroups of this genus. Phylogenetic analysis of se-
quences from the nonstructural RNA dependent RNA poly-
merase (NS5) gene of mosquito- and tick-borne
flaviviruses6,7 have also yielded tree topologies similar to that
obtained with envelope sequences. More recently, NS3, the
full-length genome sequence, and NS5 were used with equal
success to classify the nonvector-borne and arthropod-borne
flaviviruses.8
To study JEV evolution, on the other hand, a stretch of
240 or 198 nt from the premembrane (prM) region have been
predominantly used.9–13 These studies have classified JEV
strains into four distinct genotypic groups. Another tree
building exercise using a limited number of twenty JEV en-
velope sequences identified four clusters which however did
not match the four genotypes mentioned above and also ‘‘did
not correspond to geographic origin, isolation host, or viru-
lence’’.14 Yet another analysis which compared six full-
length JEV sequences suggested that the Indian strain GP78
(Accession number AF075723) may be related to the Chi-
rected p-distances (proportion of nucleotide or amino acid
sites at which the two sequences compared are different) for
building phylogenetic trees.
Our objective was three-fold. Firstly, to study the evolu-
tion of JEV, which required identification of good phyloge-
netic markers to replace the 240 nt stretch used previously
(see below). We have determined and analyzed the complete
nucleotide sequence for an Indian isolate P20778 of JEV and
carried out phylogenetic analysis of this and twelve other
independently isolated full-length JEV sequences available
in the database at the nucleotide and amino acid sequence
level. The ten individual genes were separately analyzed us-
ing the nucleotide sequences to assess their utility as phy-
logenetic markers.
Our second objective was to analyze the extensive genetic
diversity of Indian JEV strains9,16 and to see if strain varia-
tion of JEV from different parts of the globe could be linked
to their geographic origin. The literature on this topic is con-
flicting.9,17 Therefore, an exhaustive tree building exercise
was undertaken with all 107 unique envelope sequences
available in the database of JEV strains from geographically
diverse regions of Asia. This revealed the hitherto unrec-
ognized fifth genotype of JEV.
Thirdly, we determined the date of the JEV ancestor com-
mon to all 5 E-gene based genotypes we have identified.
This study thus represents an up-to-date, rigorous, and in-
depth phylogenetic analysis of JEV isolates.
MATERIALS AND METHODS
Cells and viruses. The Aedes albopictus cell line C6/36
(obtained from NCCS, Pune, India) was maintained in Ea-
gle’s minimum essential medium (MEM) with 10% fetal bo-
vine serum (FBS), and was used for growth of virus. Japa-
nese encephalitis virus strain P20778 from Vellore18 was
used for sequence determination reported here. Confluent
monolayers of C6/36 were infected with virus at a multi-
plicity of infection (m.o.i.) of 0.1. Medium containing virus
was harvested twice, at day four and again at day eight post-
infection (p.i.). Virus titer was determined on monolayers of

242
 PHYLOGENY OF JAPANESE ENCEPHALITIS VIRUS
TABLE 1
Details of virus isolates compared in the study
Strain Genbank accession #
P20778 (P20) AF080251
GP78 AF075723
Beijing-1 (BP1) L48961
Beijing P3 (P3) U47032
SA-14 (SA14) U14163
SA-14-2-8 (VAC) U15763
JaGAr 01 (JA01) AF069076
JaOArS892 (S892) M18370
JE RP-9 (RP9) AF014161
HVI AF098735
TC AF098736
TL AF098737
FU AF217620
K94P05 (K94) AF045551
YFV 17D (YFV) X03700
* Year of isolation not available.
the porcine kidney cell line PS (obtained from NCCS, Pune,
India) maintained in MEM with 10% FBS by the TCID50
method.19
RNA isolation and cDNA synthesis. Virus was pelleted
from the medium of infected cells by centrifugation in a
Beckman L8–80 ultracentrifuge at 100,000 ⫻ g for 90 min-
utes. The crude virus pellet was then centrifuged at 100,000
⫻ g for 16 hr through a gradient of 15–60% sucrose. Pure
virus banding at a density of 1.17 g/cm3 was solubilized in
4 M guanidinium isothiocyanate and used for isolation of
viral RNA. cDNA synthesis was carried out using random
hexanucleotide primers and Avian myeloblastosis virus re-
verse transcriptase (Promega Corporation, Madison, WI).
The cDNAs were tailed with EcoRI linkers and ligated to
EcoRI-digested pUC18. Overlapping recombinant clones
were ordered by hybridization to one another as well as by
determining the sequences at the two ends of the inserts.
The following regions of the viral genome were obtained by
RT-PCR of total RNA from virus infected cells: 1 to 154 nt,
1,200 to 2,000 nt, and 9,200 to 10,976 nt. The primers for
the ends of the genome spanning positions 1 to 31 (primer
A) and complementary to position 10,958 to 10,976 (primer
B) were synthesized based on the sequence published for the
JaOArS982 isolate.20 The sequence of P20778 at the 3⬘ end
of the genome was identical to that of the JaOArS982 strain
based on northern hybridization under stringent conditions
of labeled primer B to viral RNA.
DNA sequencing. pUC18 recombinants were sequenced
using M13 forward and reverse primers in the dideoxy chain
termination method with Sequenase version 2.0 from US
Biochemicals (Cleveland, OH). Some of the sequencing was
also carried out using an ABI (Foster City, CA) Prism DNA
sequencing kit for dye terminator cycle sequencing with
AmpliTaq-FS enzyme. The sequence of the P20778 JEV iso-
late has been deposited in Genbank (Accession number
AF080251). Every segment was sequenced on both the
strands with several stretches being done more than once on
each strand. For PCR products, sequencing was carried out
for two independent clones on either strand.
Sequence analysis. The full-length and individual gene
and protein sequences of thirteen independent JEV isolates
listed in Table 1 were analyzed. The average proportions of
243
Year and origin Genome length
1958, India, Vellore 10,977
1978, India, Gorakhpur 10,976
1949, China, Beijing 10,976
1949, China, Beijing 10,977
1954, China 10,976
China* 10,969
1959, Japan, Gunma 10,976
1982, Japan, Osaka 10,976
Taiwan* 10,976
1965, Taiwan 10,976
Taiwan* 10,976
Taiwan* 10,976
1995, Australia 10,964
1994, Korea, Wando 10,963
Derived from Asibi strain Africa* 10,862
synonymous (Ps) and nonsynonymous substitution (Pn) were
estimated using the Nei and Gojobori method with Jukes-
Cantor correction available in the MEGA package version
1.01 (obtained from the website, http://evolgen.biol.
metro-u.ac.jp/MEGA).21
Phylogenetic analysis. The sequences of the P20778
strain and the twelve other independently isolated full-length
JEV sequences available in the database were used in this
study are listed in Table 1. Yellow Fever virus (YFV) 17D
strain22 was used as reference taxon for the analyses. Mul-
tiple alignments of the amino acid sequence sets were gen-
erated using the Clustal W version 1.8 software (obtained
from the web site, http://www.ebi.ac.uk/clustalw)23 assuming
default alignment parameters. Conserved motifs and putative
cleavage sites were used as control of validity for alignments
as described earlier.8 This alignment was used to direct the
alignment of the nucleotide sequence sets using the
TransAlign version 1.0 Java program package (obtained
from the website, http://life.anu.edu.au/molecular/software/
transalign).24 Alignment of the 5⬘ and 3⬘ untranslated regions
was performed at the nucleotide level. Phylogenetic analyses
of the JEV isolates was carried out in every case using max-
imum likelihood (ML), neighbour-joining (NJ) and parsi-
mony methods using PAUP* version 4.0b4a (obtained from
Sinauer Associates, Sutherland, MA).25 When ML and NJ
methods were employed, the general time-reversible (GTR)26
or Hasegawa-Kishino-Yano (HKY85) substitution model27
was used with the shape of the gamma distribution for
among-site variation and ratio of transitions to transversions
estimated from the data. In the case of analysis using par-
simony method Goloboff-fit criterion28 with default concav-
ity parameter was assumed. Phylogenetic analyses for amino
acid sequences were also performed with the help of MEGA
version 1.01.21 Jukes-Cantor and gamma distance (a ⫽ 2)
algorithms were used with NJ method for construction of
trees. Phylogenetic relationships for the nucleotide and pro-
tein sequences of the complete viral genome as well as nu-
cleotide sequences of individual genes were determined. The
robustness of phylograms was evaluated by 1,000 bootstrap
resampling applying random heuristic search with 100 rep-
licates each. Phylogenetic trees were drawn using the pro-
244 UCHIL AND SATCHIDANANDAM
gram TreeView version 1.6.1 (website: http://
taxonomy.zoology.gla.ac.uk/rod/rod.html).29
Estimation of lineage divergence times. Initial trees
were obtained using the maximum parsimony and distance
methods available in the PAUP* version 4.0b4a. These trees
were then used as input for the maximum likelihood method
available in PAUP* and allowed to be perturbed using either
NNI (nearest-neighbour interchange) or TBR (tree bisection-
reconnection) in a heuristic search using a GTR model for
distance estimation. The shape parameter of the discrete
gamma distribution for among-site variation was determined
from the sequence. The number of rate categories were as-
sumed to be ten. The envelope gene was used for this anal-
ysis with only the nonsynonymous sites taken into consid-
eration because saturation plots drawn using synonymous
distances revealed that it was saturated. Pairwise nonsynon-
ymous substitutions per site were obtained by using the GTR
model under the distance settings present in PAUP*. Isola-
tion dates where known are listed in Table 1 and Figure 1.
The estimation of pairwise nonsynonymous substitution
per site for multiple JEV clades from the predicted ancestor
for that clade, was done using the branch lengths obtained
from the maximum likelihood tree as mentioned above.
These values were regressed on the time interval that sepa-
rates the years of isolation. The slope of such a plot is the
regression coefficient and is the average rate of nonsynon-
ymous substitution per site, which is referred to as ‘‘k’’. The
divergence date was estimated by dividing the branch length
(BL) towards a given node (or the sum of BLs, if the node
is next to a tip, i.e., patristic distance) by the value of k to
obtain the amount of time it took for the amount of change
represented by that BL to take place.
RESULTS AND DISCUSSION
Determination and comparative analysis of the se-
quence of P20778. A summary of the relevant information
relating to the thirteen JEV strains compared in this study
along with YFV as the outgroup is given in Table 1. Sur-
prisingly, the capsid, NS1 and NS2a protein sequences
showed the highest percentage of amino acid differences
among isolates (11.81%, 11.89% and 10.17% variable sites,
respectively, Table 2). This was despite the lowest percent-
age of variable nucleotides (16.79%) for the capsid gene,
and was correlated with the presence in the capsid gene of
the highest proportion of nonsynonymous substitutions
among all ten genes analyzed (Table 2). A comparison of
the amino acid sequences of all the proteins of the thirteen
JEV strains listed in Table 1 showed that the nonstructural
proteins NS2b and NS4a were the most conserved, with per-
centage amino acid variability of 6.1 and 6.7. In keeping
with its non-neuroinvasive phenotype, P20 did not have any
of the nine amino acids found to be unique to the envelope
FIGURE 1. Maximum likelihood phylogenetic trees for Japanese encephalitis virus: maximum likelihood trees were derived from the
sequences of each of the three genes envelope (A), NS1 (B), NS3 (C), NS5 (D) and full-length sequence (E) using PAUP* version 4.0b4a for
fourteen JEV sequences available in the database listed in Table 1. Branch lengths were derived using the general time reversible (GTR) model
with the shape parameter for among-site variation estimated from the input sequences. Corresponding YFV sequences were used as reference
taxa. The scale shows a genetic distance of 10, which is equivalent to 0.01% nucleotide divergence. The internal node numbers indicate
bootstrap support values expressed as percentage for 1,000 replicates.
protein of the neuroinvasive isolate Beijing P3.30 Also, none
of the five amino acids unique to the envelope protein of the
attenuated vaccine strains (derived from SA-14)31 was com-
mon to P20.
We observed the insertion of a G residue at position
10,701 in the P20 sequence. The same insertion was also
seen in the JaGAr01 strain from Japan and the FU strain
from Australia. Secondary structure prediction of the 3⬘ un-
translated region (UTR) sequence using the program
MFOLD version 3.0 (obtained from the website, http://
bioweb.pasteur.fr/docs/softgen.html#MFOLD)32 revealed
that the insertion of this G residue fell within a predicted
loop and consequently did not affect the secondary structure.
The 3⬘ UTR of K94 and the FU strains were 13 and 12 bases,
respectively, shorter than the usual 585 bases present in most
of the wild type JEV strains analyzed. The percentage of
variable sites in the 595 nt 3⬘ UTR was 2.79-fold more than
in the 95 nt 5⬘ UTR. Complete identity was observed in the
5⬘ UTRs of nine of the thirteen JEV sequences compared.
This suggested that the mode of recognition of the 5⬘ and 3⬘
UTRs which function as viral promoter regions by the viral
polymerase complex may be different. Extensive secondary
structure has been predicted for the 3⬘ UTR of flaviviruses,33
which appears to be vital for its function in viral replication.
One might speculate that for the 5⬘ UTR, in addition to sec-
ondary structure,34 nucleotide sequence may also play an im-
portant role in recognition by the viral polymerase complex
and/or by the capsid protein during packaging.
Identification of candidate phylogenetic markers. The
premembrane (prM) region has been extensively used to
study JEV evolution. The choice of a 240 nt stretch of the
prM gene was based on a comparison of the genes encoding
the prM, membrane (M), envelope (E), and the nonstructural
1 (NS1) proteins among a limited data set of five JEV iso-
lates, all of which belong to genotype III.9 These workers
concluded that the prM region harbored the largest number
of third position or synonymous mutations ‘‘implying that
this region was free of selective pressures that might obscure
long-term evolutionary relationships’’9 and thus is a good
phylogenetic marker. Indeed, our analysis with 13 full-length
JEV sequences belonging to genotypes I, II and III (Tables
1 and 2) also confirmed this high nucleotide divergence of
the prM region. Subsequent to the analysis of 46 JEV strains
based on this 240 nt stretch,9 several other workers have
found this region valuable for classifying newer isolates
among the 4 genotypes.10–13 To date, the database contains
sequences for this stretch from 167 JEV isolates, testifying
to its suitability for genotype-based molecular phylogenetic
classification of JEV.
The 240 nt prM tree obtained in the above study9 was a
simple p-distance based tree, and was not subjected to sta-
tistical analysis. When we carried out a tree-building exer-
cise with this same stretch using the rigorous GTR model
→
PHYLOGENY OF JAPANESE ENCEPHALITIS VIRUS 245
246 UCHIL AND SATCHIDANANDAM

TABLE 2
Analysis of sequences of Japanese encephalitis virus genes*

% Variable sites

Gene Total sites AA NT Ps Pn

prM (240 nt) 240 8.75 24.16 0.4867 ⫾ 0.0975 0.0111 ⫾ 0.0037
Capsid 381 11.81 16.79 0.1366 ⫾ 0.0188 0.0108 ⫾ 0.0027
Premembrane 501 7.78 22.75 0.3022 ⫾ 0.0349 0.0090 ⫾ 0.0023
Envelope 1,500 8.20 21.53 0.2638 ⫾ 0.0161 0.0067 ⫾ 0.0010
NS1 1,236 11.89 20.22 0.2220 ⫾ 0.0158 0.0104 ⫾ 0.0015
NS2a 501 10.17 23.35 0.2551 ⫾ 0.0275 0.0106 ⫾ 0.0026
NS2b 393 6.10 20.35 0.2697 ⫾ 0.0351 0.0068 ⫾ 0.0023
NS3 1,857 7.75 22.50 0.2922 ⫾ 0.0164 0.0068 ⫾ 0.0010
NS4a 447 6.71 21.02 0.2409 ⫾ 0.0268 0.0056 ⫾ 0.0020
NS4b 765 7.05 22.61 0.2572 ⫾ 0.0205 0.0071 ⫾ 0.0016
NS5 2,715 8.06 19.07 0.2282 ⫾ 0.0108 0.0067 ⫾ 0.0008
5⬘UTR 95 – 5.26 – –
3⬘UTR 585 – 14.70 – –
Full-length ORF 10,296 8.50 23.49 0.2471 ⫾ 0.0058 0.0076 ⫾ 0.0004
* Individual genes from the thirteen JEV isolates listed in Table 1, which excluded the attenuated derivative SA-14-2-8, were used for sequence analysis. Ps and Pn ⫽ average proportion
of synonymous (Ps) and nonsynonymous (Pn) mutations per synonymous and nonsynonymous site (mean ⫾ standard error); AA ⫽ amino acids; NT ⫽ nucleotides.

for distance estimation followed by bootstrap resampling,
(see Methods), no statistically reliable tree topology was ob-
tained for members of genotype III. This indicates that the
nucleotide differences among isolates within this region
were not sufficiently meaningful to resolve the phylogenetic
relationship among closely related isolates. The functional
significance of this high level of nucleotide variation within
this 240 nt prM region has therefore yet to be established.17
Indeed, earlier reports have shown that the use of short
stretches of nucleotides or amino acids in phylogenetic anal-
ysis35,36 results in incorrect tree topologies, suggesting that
the 240 nt prM region is not the ideal candidate for studying
the molecular evolution of JEV.
We therefore evaluated possible phylogenetic markers that
would best reflect JEV evolution. We used the simple cri-
terion that a good marker would be a region of the JEV
genome that could faithfully reproduce tree topologies ob-
tained using the full-length genome sequences with high
confidence values. As a first step toward this goal we ob-
tained a phylogenetic tree using the thirteen independently
isolated full-length JEV sequences from Genbank (Table 1)
using YFV as the outgroup. Of the remaining seven full-
length sequences available in the database for attenuated de-
rivatives of original isolates, we included only SA-14-2-8 in
our analysis. This strain consistently paired with its parent,
thus functioning as an additional indicator of the correctness
of the tree topology. The resulting tree (Figure 2E) revealed
that the Korean strain (K94) and the Australian strain (FU)
were the most divergent of all the JE strains studied, and
separated from the rest as the first branch after the outgroup.
This was in keeping with the position of these two strains
within genotypes I and IV in the envelope tree (Figure 1).
The trees also placed the Indian strain P20 distant from the
other Indian strain GP78, and the Taiwanese strains close to
the Japanese strains. A similar tree topology was obtained
using the full-length amino acid sequences (data not shown).
Although full-length sequences gave reliable trees, rigor-
ous phylogenetic analysis of large numbers of long sequence
stretches requires enormous computing capacity. We there-
fore explored the utility of individual structural and non-
structural genes as phylogenetic markers, which reliably re-
produce the tree topology, obtained with the full-length se-
quence. Among the nonstructural and structural genes ana-
lyzed, polytomies were observed with capsid, prM, NS2a,
NS2b, NS4a, and NS4b (data not shown), possibly due to
lack of sufficient phylogenetically informative sites. NS1,
NS3, NS5, and envelope genes alone reproduced the topol-
ogy obtained using the full-length sequence with reliable
bootstrap support values (Figure 2, A–E). When we used the
240 nt stretch from the prM region to study JEV evolution,
polytomies surfaced with respect to isolates from genotype
III, making these trees unreliable (data not shown).
The trees obtained during the above analysis reflected the
wide divergence of the Indian JEV strains P20 and GP78
(Figure 2, A–E) despite their geographic proximity. An ear-
lier report had observed genetic diversity for JEV strains
from Japan and India.9 Immunotyping and oligotyping data
suggest that strain variation among JE viruses is not related
to geographic location.16,37–39 Comparison of nucleotide and
amino acid sequence of envelope genes of thirteen JEV
strains17 showed no clustering based on geographical loca-
tion. However, Chen and coworkers9 used forty-six prM se-
quences to show that JEV isolates from the same geographic
region and time period are similar.
To resolve this issue, and to gain insight into the genetic
diversity of JEV from different geographic areas, we carried
out an exhaustive phylogenetic analysis using all 107 en-
velope sequences from unique JEV isolates in the database.
E-gene sequences for JEV are available in the database,
making it possible to use this gene for phylogenetic analysis.
These isolates span a sixty-year period and represent strains
from almost all the countries where JE virus is prevalent.
Japanese encephalitis virus has been reported in Pakistan;40
however, no sequence for this isolate is available. The da-
tabase contained repeats for 32 envelope sequences, which
were also included in our analysis. The 107 E-gene sequenc-
es were first analyzed using distance and parsimony meth-
ods, as described in Methods. The tree classified JEV strains
into five distinct genotypic groups differing by a minimum
nucleotide divergence of 7% with high bootstrap support
values (Figure 1). Four of these groups broadly matched the
four genotypes that were proposed earlier using the prM
 PHYLOGENY OF JAPANESE ENCEPHALITIS VIRUS 247

FIGURE 2. Phylogenetic tree derived from Japanese encephalitis virus envelope sequences from 107 unique isolates: neighbor joining method
available in PAUP* using Hasegawa-Kishino-Yano (HKY) distances with transition/transversion (ts/tv) ratios estimated empirically was used
to draw the tree. Homologous sequences from Murray Valley encephalitis virus, St. Louis encephalitis virus (sister groups to JEV10) and yellow
fever virus were used to root the tree. Numbers on the internal nodes refer to the bootstrap support values expressed as percentage for 1,000
replicates. The scale shows a genetic distance of 10, or 1% nucleotide divergence. The E-gene sequences from the clade of Taiwanese strains
depicted as node ‘‘A’’ in genotype III were used for regression analysis to estimate the JEV evolutionary rate.

gene region,9 while a fifth genotype surfaced in our analysis,
containing the lone Singapore Muar isolate.41 The Muar
strain was the most divergent of the JEV isolates analyzed,
differing from genotypes I, II, III, and IV by a minimum
nucleotide divergence of 21%, 20%, 19.3%, and 22% re-
spectively. Using a 7% cut-off value of nucleotide diver-
gence, the Muar strain could clearly be placed in a newly
formed fifth genotype. The reason for the presence of this
unique strain in the midst of countries populated by strains
of 3 other genotypes is unclear. Without exception, all of the
repeat envelope sequences of any isolate grouped together
as expected. The vastly larger number of envelope sequences
analyzed, compared to the prM region9, and with greater
diversity, may contribute to the 7% cut-off in nucleotide di-
vergence for this gene. Thus, while the tree based on the
240 nt prM region could not reliably display statistically
significant topologies (data not shown), it could distinguish
between the different genotypes. This may be due to the 12%
nucleotide divergence between genotypes in this nucleotide
stretch9. We therefore propose the use of envelope sequences
to classify all future JEV isolates into the various genotypes
and more importantly, to obtain information on precise re-
lationships among closely related isolates.
We also observed striking differences between the prM-
and E-based trees. In the prM-based tree9 JEV strains from
Thailand were found only in genotypes I and II. The large
E-gene data set-based tree placed Thai strains KPP034–
35CT and Chiang Mai in genotype III. Similarly, Indonesian
strains JKT646 and 208335, previously classified in geno-
type IV, and JKT1724 previously classified in genotype II,
as well as a Korean strain (K82P01) previously classified in
genotype I, were all now placed in genotype III. The Korean
strain showed a minimal difference of 4.4% and 5.3% from
genotypes III and I respectively. Using the 7% nucleotide
divergence cut-off mentioned above, this strain could be
placed with equal confidence in both genotypes I and III.
One of the hallmarks of the tree was the overall clustering
of strains according to their geographic origin (Figure 1).
For example, the Chinese, the Japanese, and the Taiwanese
strains were found to group together on this basis. Similarly,
JEV strains from the Indian subcontinent, i.e., from India,
Sri Lanka, and Nepal also clustered together. The tree also
revealed the genetic diversity among the JEV strains within
the same geographic boundary. For instance, the Japanese
strains grouped into four different sub-geographic clusters
suggesting the co-existence of at least four genetically di-
248 UCHIL AND SATCHIDANANDAM

TABLE 3
Genome sequence analysis of the two Indian strains P20 and GP78 of Japanese encephalitis virus

% Variable sites

Gene Total sites AA NT Ps Pn

prM (240 nt) 240 1.25 5.41 0.2524 ⫾ 0.0802 0.0107 ⫾ 0.0076
Capsid 381 3.93 4.72 0.1530 ⫾ 0.0437 0.0176 ⫾ 0.0079
Premembrane 501 2.39 4.99 0.1927 ⫾ 0.0448 0.0132 ⫾ 0.0059
Envelope 1,500 1.60 4.53 0.1862 ⫾ 0.0249 0.0071 ⫾ 0.0025
NS1 1,236 1.45 2.75 0.1051 ⫾ 0.0203 0.0063 ⫾ 0.0026
NS2a 501 1.19 4.19 0.1643 ⫾ 0.0389 0.0054 ⫾ 0.0038
NS2b 393 1.52 4.83 0.2102 ⫾ 0.0532 0.0067 ⫾ 0.0047
NS3 1,857 2.26 5.33 0.2196 ⫾ 0.0250 0.0107 ⫾ 0.0028
NS4a 447 2.01 4.92 0.1847 ⫾ 0.0439 0.0091 ⫾ 0.0053
NS4b 765 1.96 4.70 0.1723 ⫾ 0.0320 0.0089 ⫾ 0.0040
NS5 2,715 1.48 4.15 0.1820 ⫾ 0.0189 0.0062 ⫾ 0.0017
5⬘UTR 95 – 1.05 – –
3⬘UTR 585 – 3.58 – –
Full-length ORF 10,296 1.80 4.37 0.2220 ⫾ 0.0158 0.0082 ⫾ 0.0010
Ps and Pn ⫽ average proportion of synonymous (Ps) and nonsynonymous (Pn) mutations per synonymous and nonsynonymous site (mean ⫾ standard error).

verse groups in this country where JEV was first reported.
The JEV strains in Taiwan and the Indian subcontinent
grouped into three distinct clusters each, whereas the strains
from China were homogeneous and found in one large clus-
ter. This last mentioned feature was reported previously.9,42
Our analysis thus revealed that while genetic diversity within
a defined geographic region resulted in distinct clades, there
was overall clustering of JE isolates within the genotypes
based on place of isolation. Notable exceptions to this pat-
tern of clustering were also observed. The P1 strain from
China for example was found in a cluster which contained
isolates from Japan (Sagiyama), Vietnam (VN118), and Tai-
wan (TL). Two strains from Thailand falling within genotype
III along with Chinese and Indian strains have been previ-
ously mentioned. Strains from Indonesia were distributed
among genotypes II, III, and IV, along with viruses from
other geographic regions, making the Indonesian strains the
most diverse among the JEV prevalent areas.
In a second analysis performed for dating the ancestral
JEV lineage, we carried out a refinement on a subset of 40
E genes containing representatives from the 5 genotypes.
Here we used the more rigorous and time-consuming ML
method, which can overcome possible substitutional super-
imposition, as described in the Methods section. This tree
faithfully reproduced the topology obtained from 107 se-
quences, proving the reliability of the method used to obtain
the initial tree (data not shown).
Genetic diversity of JEV strains in the Indian subcon-
tinent. Our full-length genome sequence analysis indicated
the significant differences between the two Indian strains
P20 and GP78. A comparison of these strains revealed that
the percent nucleotide and amino acid divergence was 4.37
and 1.80, respectively (Table 3). The NS2a protein was the
most conserved between the two isolates with a difference
of only 1.19%. The capsid protein, on the other hand was
least conserved with 3.93% divergence. Interestingly a com-
parison of the capsid protein of GP78 with other JEV iso-
lates listed in table 1 revealed that two of the four amino
acid residues that were unique to GP78 capsid sequence fell
in the 85 to 102 amino acid stretch, resulting in abolition of
the putative ‘‘nuclear localization signal’’ identified by the
program ProfileScan (obtained from the website, http://
www.isrec.isb-sib.ch/software/PFSCAN㛮form.html).43 This
sequence presumably represents the RNA binding region of
the capsid protein, judging from the presence of an arginine-
rich stretch as well as RGG motifs.44 Prediction of secondary
structure of the capsid protein using the program Predator
(obtained from the web site,http://www.embl-heidelberg.de/
argos/predator)45 revealed that in this region of interest, in
contrast to the helical structure assumed by the capsid of all
other JE isolates, GP78 displayed a random-coil structure.
Significantly, the RGG box of the RNA binding protein
hnRNP U44 also displayed a random-coil secondary structure
in this region. Hence in addition to envelope playing an im-
portant role in the slow release of RNA into the cytoplasm
following viral entry for GP78 as hypothesized,46 the capsid
may also contribute to this defect by binding strongly to the
viral RNA.
The large data set of 107 E-gene sequences contained
eight independent Indian isolates from various parts of India.
In addition 2 strains each from Nepal and Sri Lanka were
also present. This analysis grouped these strains into three
different clusters (Figures 1 and 3). The first group desig-
nated as Vellore group contained the two Vellore strains P20
(isolated in 1958 from human brain), and G8924 (isolated
in 1956 from mosquito) plus a strain 782219 from Tamil
Nadu (isolated in 1978). The Sri Lankan strain, 691004 (iso-
lated in 1969 from human brain) which was placed with the
Nakayama strain from Japan, was 2.9% divergent from the
Vellore group, reflecting the geographic proximity of Sri
Lanka to South India. The second group designated as Bank-
ura group comprised four Indian strains, which included
strains from northern India (GP78 from Gorakhpur, isolated
in 1978 and 733913 from Bankura isolated in 1973) and
western India (826309 from Goa isolated in 1982). In ad-
dition, this group contained a strain H49778 from Sri Lanka
(isolated in 1987). The third group, designated as Nepal
group, was formed by an Indian strain 7812474 from Assam
(isolated in 1978 from human brain) and the B2524 strain
from Nepal (isolated in 1985). Our calculations using the E-
gene based tree revealed that an arbitrary cut-off of 3.4%
divergence could be set to define these three groups (Figure
1). The Assam strain 7812474 could be placed either in the
Bankura or Nepal groups, based on the 3.4% divergence cut
 PHYLOGENY OF JAPANESE ENCEPHALITIS VIRUS
FIGURE 3. Map of the Indian subcontinent showing the geo-
graphic expanse of the different Japanese encephalitis virus groups
classified in Figure 2.
off. We chose to keep this strain in the Nepal group since it
was closer to the B2524 strain from Nepal (1.5% diver-
gence) than to any of the strains present in the Bankura
group. Availability of more E-gene sequences from this geo-
graphic region may further consolidate this group. The Vel-
lore group was phylogenetically closer to the Nakayama
strain from Japan, whereas the Bankura group was closer to
the Chiang Mai strain from Thailand. The Nepal group was
distant from most other JEV clusters and contained the
JKT1724 strain from Java, Indonesia (Figure 1). The E-gene
sequence of the Nepalese strain B2524 differed from that of
the Indonesian strain JKT1724 by only two nucleotides.
Thus, the phylogenetic relatedness may indicate the country
of origin for the viruses present in the Indian subcontinent,
and the possible involvement of migratory birds in trans-
mitting JEV into the Indian subcontinent from these geo-
graphically distant regions. Interestingly, the ability of sera
from vaccinees given the Nakayama strain JE vaccine to
cross-neutralize JEV strains in the Vellore group more ef-
fectively than those from the Bankura group47 shows the
relatedness of the Vellore group strains to the Nakayama
strain from Japan. This similarity between the serology data
and our phylogenetic classification further validates the
grouping of the JEV strains in the Indian subcontinent. Per-
haps additional genetically diverse groups exist in the Indian
subcontinent, the extent of which will become known as
more E-gene sequences become available. Phylogenetic
analysis may therefore serve to point to the serological char-
acteristics of the envelope proteins and may help to guide
the planning of vaccination strategies, whose success will be
influenced by the extent of diversity seen in natural isolates.
The geographic expanse of JEV strains from the Bankura
group ranges from the east (West Bengal) and north of India
(Uttar Pradesh) to the west coast of South India (Goa), and
249
TABLE 4
Dating the common ancestor of Japanese encephalitis virus (JEV)*
Date for the JEV ancestral node
in years before present
Genotype (mean ⫾ standard error)
I 133.00 ⫾ 56.24
II 166.21 ⫾ 58.81
III 154.17 ⫾ 45.01
IV 259.69 ⫾ 37.33
V 158.59 ⫾ 68.81
* The rate of nonsynonymous substitution per site (k ⫽ 7.5147 ⫻ 10⫺4) obtained from
regression analysis of clade A (Figure 1) was used to date the JEV ancestor with respect
to each of the five genotypes.
Sri Lanka. (Figure 3). The strains, which form the Bankura
group, were isolated between 1973 and 1982. Except for the
Tamil Nadu strain isolated in 1978, members of the Vellore
group were isolated in 1956 and 1958. This suggests that
JEV was introduced into India on two occasions separated
by atleast 17 years. The earlier introduction that formed the
Vellore group appears to have been restricted in its ability
to spread, presumably due to non-availability of suitable
host-vector combinations. However, the virus obviously was
not lost and appears to be the progenitor of the more recent
1978 isolate from Tirunelveli (782219) in Tamil Nadu.
Therefore, the present day endemicity of JEV in India ap-
pears to be predominantly due to the robust spread of strains
from the Bankura group.
Dating the ancestor common to all JEV genotypes. Es-
timation of accurate branch lengths is critical to obtain valid
numbers for divergence dates. The ML method gives the
best estimate of branch lengths,48,49 while the distance and
maximum parsimony methods are best suited for obtaining
reliable tree topologies. We therefore obtained tree topolo-
gies for the envelope genes using the latter 2 methods avail-
able in PAUP* version 4.0b4a. These trees faithfully repro-
duced the topologies obtained earlier, and then served as user
trees for the maximum likelihood method. PAUP* allows
estimation of transition to transversion ratios and shape pa-
rameters for among-site variation from the sequence data.
This approach yields the best estimates of branch lengths
and consequently a more accurate dating of a particular
node. Of the 107 envelope sequences mentioned earlier, forty
JEV isolates representing each of the five genotypes were
chosen for this rigorous analysis.
Regression analyses was subsequently carried out using
sequences of multiple clades from this tree belonging to the
different genotypes, each derived from a geographically re-
stricted region. Pairwise nonsynonymous substitutions per
site obtained by comparing each member of the clade to its
predicted ancestor were plotted versus the time period in
years that separates the isolates. The slope of the linear re-
gression is the rate of nonsynonymous substitution per site,
referred to as ‘‘k’’. We obtained the highest confidence level
of P ⬍ 0.0001 (R2 ⫽ 0.899) that the slope is significantly
different from zero by using the t-test available in EXCEL,
for the clade comprising the Taiwanese strains in genotype
3 (node ‘‘A’’ in Figure 1; the attenuated strains CH2195LA
and CH2195SA were not included in this analysis). The cor-
responding k value of 7.5147 ⫻ 10⫺4 was used to date the
common ancestor of JEV with respect to each genotype (Ta-
ble 4). The value of 2.6 ⫻ 10⫺4 for k obtained earlier3 relied
250 UCHIL AND SATCHIDANANDAM
on a limited number of 4 taxa and the use of the method of
Li and coworkers.50 We also obtained similar values for k
upon linear regression analysis of some clades, but with poor
confidence values. Our estimations assumed the operation of
a molecular clock for evolution of JEV, which may not be
entirely valid. The value for divergence times from the com-
mon ancestor ranged from 259 ⫾ 37 (mean ⫾ standard error)
years for genotype IV, the oldest of the genotypes, to 133 ⫾
56 years for genotype I, one of the most recently diverged
of all five genotypes of JEV. Thus, the global ancestor of all
JEV strains appears to have arisen in the last 300 years.
Acknowledgments: We are indebted to Niranjan V. Joshi from the
Centre for Ecological Sciences, Indian Institute of Science, for valu-
able guidance especially in the initial stages. We thank the staff of
the Supercomputer Education and Research Centre for help with use
of the supercomputer. We thank Jim Wilgenbusch, Paolo Zanotto,
and Ernest Gould for clarifications and suggestions during the course
of these analyses.
Financial Support: PDU is a recipient of the Senior Research Fel-
lowship of the Council of Scientific and Industrial Research.
Authors’ addresses: Pradeep D. Uchil and Vijaya Satchidanandam,
Department of Microbiology and Cell Biology, Indian Institute of
Science, Bangalore-560012, Karnataka, India.
Reprint requests: Vijaya Satchidanandam, Department of Microbi-
ology and Cell Biology, Indian Institute of Science, Bangalore-
560012, Karnataka, India. Telephone: 91-80-3092685. FAX: 91-80-
3602697. Email: vijaya@mcbl.iisc.ernet.in
REFERENCES
1. Gourie-Devi M, Ravi V, Shankar SK, 1995. Japanese encepha-
litis: an overview. Rose FC, ed. Recent Advances in Tropical
Neurology. Amsterdam, Netherlands: Elsevier Science Pub-
lishers B. V., 217–235.
2. Burke DS, Leake CJ, 1998. Japanese Encephalitis. Boca Raton,
FL: CRC Press.
3. Zanotto PM, Gould EA, Gao GF, Harvey PH, Holmes EC, 1996.
Population dynamics of flaviviruses revealed by molecular
phylogenies. Proc Natl Acad Sci USA 93: 548–53.
4. Zanotto PM, Gao GF, Gritsun T, Marin MS, Jiang WR, Venu-
gopal K, Reid HW, Gould EA, 1995. An arbovirus cline
across the northern hemisphere. Virology 210: 152–159.
5. Calisher CH, Karabatsos N, Dalrymple JM, Shope RE, Porter-
field JS, Westaway EG, Brandt WE, 1989. Antigenic relation-
ships between flaviviruses as determined by cross-neutrali-
zation tests with polyclonal antisera. J Gen Virol 70: 37–43.
6. Kuno G, Chang G-JJ, Tsuchiya R, Karabatsos N, Cropp CB,
1998. Phylogeny of the genus Flavivirus. J Virol 72: 73–83.
7. Marin MS, Zanotto PM, Gritsun TS, Gould EA, 1995. Phylog-
eny of TYU, SRE, and CFA virus: Different evolutionary
rates in the genus Flavivirus. Virology 206: 1133–1139.
8. Billoir F, Chesse R, Tolou H, Micco P, Gould EA, Lamballerie
X, 2000. Phylogeny of genus Flavivirus using complete cod-
ing sequences of arthropod-borne viruses and viruses with no
known vector. J Gen Virol 81: 781–790.
9. Chen WR, Tesh RB, Rico-Hesse R, 1990. Genetic variation of
Japanese encephalitis virus in nature. J Gen Virol 71: 2915–
2922.
10. Chen W-R, Tesh RB, Rico-Hesse R, 1992. A new genotype of
Japanese encephalitis virus from Indonesia. Am J Trop Med
Hyg 47: 61–69.
11. Huong VTQ, Ha DQ, Deubel V, 1993. Genetic study of Japa-
nese encephalitis viruses from Vietnam. Am J Trop Med Hyg
49: 538–544.
12. Ali A, Igarashi A, Paneru LR, Hasebe F, Morita K, Takagi M,
Suwonkerd YT, Wada Y, 1995. Characterization of two Jap-
anese encephalitis virus isolated in Thailand. Arch Virol 140:
1557–1575.
13. Chung Y, Nam J, Ban S, Cho H, 1996. Antigenic and genetic
analysis of Japanese encephalitis viruses isolated from Korea.
Am J Trop Med Hyg 55: 91–97.
14. Paranjpe S, Banerjee K, 1996. Phylogenetic analysis of enve-
lope gene of Japanese encephalitis virus. Virus Res 42: 107–
117.
15. Vrati S, Giri RK, Razdan A, Malik P, 1999. Complete nucleotide
sequence of an Indian strain of Japanese encephalitis virus.
Am J Trop Med Hyg 61: 677–680.
16. Banerjee K, Ranadive SN, 1989. Oligonucleotide fingerprint
analysis of Japanese encephalitis virus strains of different geo-
graphical origin. Indian J Med Res 89: 201–216.
17. Ni H, Barrett ADT, 1995. Nucleotide and deduced amino acid
sequence of the structural protein genes of Japanese enceph-
alitis viruses from different geographical locations. J Gen Vi-
rol 76: 401–407.
18. Webb JKG, Pavri K, George S, Chandy J, Jadhav M, 1964.
Japanese B encephalitis in South India: isolation of virus from
human brain. Bose SK, Dey AK, eds. Asian Paediatrics.
Bombay: Asia Publishing House.
19. Gould EA, Clegg EA, 1985. Growth, titration and purification
of Togaviruses. Mahy BWJ, ed. Virology: A Practical Ap-
proach. Oxford, England: IRL Press Limited, 43–78.
20. Sumiyoshi H, Mori C, Fuke I, Morita K, Kuhara S, Kondou J,
Kikuchi Y, Nagamatu H,, Igarashi A, 1987. Complete nucle-
otide sequence of the Japanese encephalitis virus genome
RNA. Virology 161: 497–510.
21. Kumar S, Tamura K, Nei M, 1993. MEGA: Molecular Evolu-
tionary Genetics Analysis. Version 1.01. University Park, PA:
The Pennsylvania State University.
22. Rice CM, Lenches EM, Eddy SR, Shin SJ, Sheets RL, Strauss
JH, 1985. Nucleotide sequence of yellow fever virus: impli-
cations for flavivirus gene expression and evolution. Science
229: 726–733.
23. Thompson JD, Higgins DG, Gibson TJ, 1994. CLUSTAL W:
Improving the sensitivity of progressive multiple sequence
alignment through sequence weighting, positions-specific gap
penalties and weight matrix choice. Nucleic Acids Res 22:
4673–4680.
24. Weiller GF, 1999. TransAlign: a Java program package for align-
ing coding nucleotide sequences according to the amino acid
sequence encoded. Version 1.0. Canberra A.C.T. 26O1, Aus-
tralia: Bioinformatics Laboratory, Research School of Biolog-
ical Sciences, Australian National University.
25. Swofford DL, 1998. Paup*. Phylogenetic Analysis Using Par-
simony (* and Other Methods). Version 4. Sunderland, Mas-
sachusetts: Sinauer Associates.
26. Rodriguez R, Oliver JL, Marin A, Medina JR, 1990. The general
stochastic model of nucleotide substitution. J Theor Biol 142:
485–501
27. Hasegawa M, Kishino H, Yano T, 1985. Dating the human-ape
split by a molecular clock of mitochondrial DNA. J Mol Evol
22: 106–174.
28. Goloboff PA, 1993. Estimating the character weights during tree
search. Cladistics 9: 83–91.
29. Page RDM, 1996. TREEVIEW: an application to display phy-
logenetic trees on personal computers. Comput Appl Biosci
12: 357–358.
30. Ni H, Barrett ADT, 1996. Analysis of molecular basis of high
neuroinvasiveness for mice of wild-type Japanese encephalitis
virus strain P3. J Gen Virol 77: 1449–1455.
31. Ni H, Chang GJ, Xie H, Trent DW, Barrett ADT, 1995. Molec-
ular basis of attenuation of neurovirulence of wild-type Jap-
anese encephalitis virus strain SA14. J Gen Virol 76: 409–
413.
32. Zuker M, Mathews DH, Turner, DH, 1999. Algorithms and
Thermodynamics for RNA Secondary Structure Prediction: A
Practical Guide. Dordrecht, Netherlands: Kluwer Academic
Publishers.
33. Brinton MA, Fernandez AV, Dispoto JH, 1986. The 3⬘-nucleo-
tides of flavivirus genomic RNA form a conserved secondary
structure. Virology 153: 113–121.
34. Brinton MA, Dispoto JH, 1988. Sequence and secondary struc-
 PHYLOGENY OF JAPANESE ENCEPHALITIS VIRUS
ture analysis of the 5⬘-terminal region of flavivirus genome
RNA. Virology 162: 290–299.
35. Nei M, Kumar S, Takahashi K, 1998. The optimization principle
in phylogenetic analysis tends to give incorrect topologies
when the number of nucleotides or amino acids used is small.
Proc Natl Acad Sci USA 95: 12390–12397.
36. Gascuel O, 2000. On the optimization principle in phylogenetic
analysis and the minimum-evolution criterion. Mol Biol Evol
17: 401–405.
37. Okuno Y, Okuda T, Kondo A, Suzuki M, Kobayashi M, Oya A,
1968. Immunotyping of different strains of Japanese enceph-
alitis virus by antibody-absorption, haemagglutination-inhi-
bition and complement-fixation test. Bull World Health Organ
38: 547–563.
38. Wills MR, Sil BK, Cao JX, Barrett ADT, 1992. Antigenic char-
acterization of live attenuated Japanese encephalitis vaccine
virus SA14–4–2: a comparison with isolates of virus covering
a wide geographic area. Vaccine 10: 861–872.
39. Hori H, 1986. Oligonucleotide fingerprint analysis of Japanese
encephalitis (JE) virus strains of different geographic origins.
Trop Med 28: 179–190.
40. Igarashi A, Tanaka M, Morita K, Takasu T, Ahmed A, Akram
DS, Waqar MA, 1994. Detection of West Nile and Japanese
encephalitis viral genome sequences in cerebrospinal fluid
from acute encephalitis cases in Karachi, Pakistan. Microbiol
Immunol 38: 827–30.
251
41. Hasegawa H, Yoshida M, Fujita S, Kobayashi Y, 1994. Com-
parison of structural proteins among antigenically different
Japanese encephalitis virus strains. Vaccine 12: 841–844.
42. Huang CH, 1982. Studies on Japanese encephalitis in China.
Advances Virus Res 27: 71–101.
43. Hofmann K, Bucher P, Falquet L, Bairoch A, 1999. The PROS-
ITE database, its status in 1999. Nucleic Acids Res 27: 215–
219.
44. Burd CG, Dreyfuss G, 1994. Conserved structures and diversity
of functions of RNA-binding proteins. Science 265: 615–621.
45. Frishman D, Argos P, 1996. Incorporation of non-local inter-
actions in protein secondary structure prediction from the
amino acid sequence. Protein Eng 9: 133–42.
46. Vrati S, Agarwal V, Malik P, Wani SA, Saini M, 1999. Molec-
ular characterization of an Indian isolate of Japanese enceph-
alitis virus that shows an extended lag phase during growth.
J Gen Virol 80: 1665–1671.
47. Banerjee K, 1986. Certain characteristics of Japanese encepha-
litis virus strains by neutralization test. Indian J Med Res 83:
243–250.
48. Nei M, 1996. Phylogenetic analysis in molecular evolutionary
genetics. Annu Rev Gen 30: 371–403.
49. Yang Z, 1996. Phylogenetic analysis using parsimony and like-
lihood methods. J Mol Evol 42: 294–307.
50. Li W-H, Tanimura M, Sharp PM, 1988. Rates and dates of di-
vergence between AIDS virus nucleotide sequences. Mol Biol
Evol 5: 313–330.