J o = ~ a ! . . o f ~ne~al..~'i,~!~gy..(!9?% .77~..30!3.-!024:.....~r!9~e~. in G~e~.Br!ta.~n..........................................................................................................................
Evolutionary analysis of variants of hepatitis C virus found in
South-East Asia: comparison with classifications based upon
sequence similarity
P. S i m m o n d s , 1 J. M e l l o r , 1 T. S a k u l d a m r o n g p a n i c h , 2 C. N u c h a p r a y o o n , z S. T a n p r a s e r t , z E. C. H o l m e s 3
a n d D. B. S m i t h 1
Department of Medical Microbiology, University of Edinburgh, Teviot Place, Edinburgh EH8 9AG, UK
2National Blood Centre, Thai Red Cross Society, Henri Durant Road, Bangkok 10330, Thailand
3Department of Zoology, University of Oxford, South Parks Road, Oxford OXI 3PS, UK
Variants of hepatitis C virus (HCV) have been
classified by nucleotide sequence comparisons in
different regions of the genome. Many investigators
have defined the ranges of sequence similarity
values or evolutionary distances corresponding to
divisions of HCV into types, subtypes and isolates.
Using these criteria, novel variants of HCV from
Vietnam, Thailand and Indonesia have been classified as types 7, 8, 9, 10 and 1 1, many of which can
be further subdivided into between two to four
subtypes. In this study, this distance-based method
of virus classification was compared with phylogenetic analysis and statistical measures to establish the confidence of the groupings. Using
bootstrap resampling of phylogenetic trees in several subgenomic regions (core, E l , NS5) and with
complete genomic sequences, we found that one set
of novel HCV variants ('types 7, 8, 9 and 1 1 ')
consistently grouped together into a single clade
Introduction
Hepatitis C virus (HCV) shows a degree of genetic
heterogeneity similar to that of other RNA viruses with
different isolates of HCV showing as much as 30 % nucleotide
sequence divergence over the entire genome. It is suspected
that such divergence may lead to serological and biological
Author for correspondence:PeterSimmonds.
Fax +44 131 650 6531. [email protected]
Sequencesobtainedin this studyhavebeensubmittedto GenBankand
bear the accessionnumbersL50535-50542, L50544-50552,
L50554 and L50556.
0001-4214 © 1996 SGM
that also contained type 6a, while 'type l O a '
grouped with type 3. In contrast, no robust higherorder groupings were observed between any of the
other five previously described HCV genotypes
(types 1-5). In each subgenomic region, the distribution of pairwise distances between members of
the type 6 clade were consistently bi-modal and
therefore provided no justification for classification
of these variants into the three proposed categories
(type, subtype, isolate). Based on these results, we
propose that a more useful classification would
regard all these variants as subtypes of type 6 or
type 3, even though the level of sequence diversity
within the clade was greater than observed for other
genotypes. Classification by phylogenetic relatedness rules out simple sequence similarity measurements as a method for assigning HCV genotypes,
but provides a more appropriate description of the
evolutionary and epidemiological history of a virus.
differences between variants of HCV. For example, the
diversity of the amino acid sequence of the HCV E1 and E2
glycoproteins of HCV (around 34-40% for El, and 26--29%
for E2) is similar to that found in the E gene of different
serotypes of dengue fever virus (26-40 %) which are known
not to cross-neutralize. In the absence of biological or true
neutralization assays for HCV, classification of HCV is
currently based upon nucleotide sequence comparisons. Sequence relationships appear equivalent in different parts of the
genome. For example, the six main groups of variants found in
the NS5 gene (Simmonds et al., 1993) are paralleled by
equivalent groupings in EI (Bukh et al., I993), core (Bukh ¢t al.,
1994) and NS4 (Bhattacherjee et al., 1995), and by comparison
of available complete genomic sequences (types la, lb, lc, 2a,
2b, 3a, 3b) (Sakamoto et al., 1994). So far, there is no evidence
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
301~
=-
(a)
Z
c~
,.4 BB9
PB
~8893
PE
NGI
n
L ~ EUTH86
_____~
B492
[
PC
m
EUBURI
~ / _ . ,VN507
.
VN530
8a
VN606
YN655
V'N405
"[
--I
I
]
I
HPCNSS011
EUTH22
,
EUTHIO0
j-1
E rHso
Eu~.
I-[PCN85013
i'
8b
9c
EUTH1
1
k
J
i 9b
t._ HPCNSS0 t 2
_~9a
lla
VNO04
VN085
"-[
1489
CJK085C
EUTH36
d ! s
i't
g
----
EUTH85
EUTH89
NS5009
NS5008
52A
$5018
7c/NGII f
EUTIr7
[I
L~-
HPCNS5007
BB7 LC47/93
YN968V]~87
~ V N 843
TM
VN711
1
882VN998
VN235
HPCNS5006
HPCNS5004
i 7a
-1 7b
7a
"1 6b
d
6a
a
C
b
VN865
H?-N748
VN710
VN824
8
3
5O6
.,[.~
541
VN555
VN569
VN889
930
VN893
l a Pt
Fig. 1. Phylogenetic analysis of nucleotide sequences of HCV variants from Hong Kong, Vietnam, Thailand, Burma and
Indonesia. (a) A 222 base fragment (positions 7975-8196) ; (b) a 366 base fragment (positions 8 4 6 4 - 8 8 2 9 ) of
~01 z
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
t-,
(b)
6a VN506
6a VN571
6a VN569
8aVN538
HPCNS5004
8a VN507
F
3
8a VN530
9aVN004
9a VN085
L_ HPCNS5011
HPCTH33P3
~H
Z
6a
a
6b
b
8a
!
8b
k
9a
h
J
8bVN405
m----
Z
9c/VIII
I
HPCNS5012
HPCTH3PI2
HPCTH9PI2
PCNS5013
~__
HPCJK065C
HPCJKI48D
HPCJK046C
Ta VN998
VN540
N787
9b/VIII
i
~77a~
VN843
HPCNSS006
7bVN235
HPCTH35P3
"~HPCTH37PI
..~HPCNS5008
HPCTH26P3
---HPCNS5007
L
HPCTH22P3
HPCTH15P3
HPCTH18P3
HPCNS5009
HPCNS52A
,
HPCTH13P3
l a HCVl
m
11a
g
7a
e
7d
7b
C
d
7c/VII
~
NS5. The sequence of type 1 a (Choo et o/., 1991 ) was used as an outgroup. Previously described genotype designations
(coLumn 1) or provisional clade identifiers (column 2) are shown on the right.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
~0
for recombinants of HCV that have sequences of one genotype
in one part of the genome and of a different one elsewhere
(Smith et al., 1995; Stuyver et al., 1994; Simmonds et al.,
1994 b).
How long and from where in the genome a region
should be to enable reliable classification by this method has
been a matter for debate. Although it is generally true that
longer sequences are more informative for classification, it is
usually possible to identify genotypes by sequence comparison of relatively short subgenomic regions of HCV (Mellor
et al., 1995; Tokita et al., 1994; Bukh, 1995). However, the
identification of novel genotypes of HCV is generally
considered to require sequences from at least two parts of the
genome (Simmonds et al., 1994a). The 5' non-coding region
(5'NCR) is too conserved to be useful for this purpose,
although some subtypes and almost all types of HCV show
distinct sequences in this region (Mellor et al., 1996; Tokita et
al., 1995; Smith et al., 1995).
Another more fundamental discussion concems the
methods by which sequences should be compared. To date, it
has been possible to obtain equivalent classifications of HCV
by either phylogenetic analysis of sequences or by measures of
pairwise sequence similarity. In the former method, genotypes
have been identified by the branching order of a phylogenetic
tree. For example, sequences of subgenomic fragments of types
1--6 always produce trees containing six main branches (clades)
which split further into groupings corresponding to the
subtypes of HCV.
The basis for the second method is that the sequences of the
different types are all similarly divergent from each other, as
are virus subtypes, and so it is possible to define ranges of
sequence similarities that correspond to the type and subtype
categories. For example, pairwise comparisons of 76 sequences
from a 222 base fragment of NS5 (positions 7975-8196)
produce ranges of similarity values from 56-2-72"1% between
types of HCV, 74"8-86"0 % between subtypes and greater than
87"8 % between variants of the same subtype (Simmonds et aI.,
1993). Different regions of the genome produce different
ranges depending on how conserved or variable the region is,
but in each case the ranges are non-overlapping. For E1 and
NS4, the maximum and minimum pairwise distances between
type and subtype are greater than for NS5 (Bhattacherjee et al.,
1995; Bukh et al., 1993), while they are smaller for core (Bukh
et al., 1994). For many investigators the sequence similarity of
a novel sequence to those of established genotypes has
become the primary means to identify and classify new
genotypes.
Recently, a series of novel variants of HCV has been found
in Vietnam, Thailand, Burma and Indonesia (Doi et al., 1996;
Mellor et al., 1995, 1996; Tokita et al., 1994, 1995, 1996;
Sugiyama et al., 1995; Apichartpiyakul et al., 1994). On the
basis of previously established ranges for type and subtypes in
NS5 and core/E1, it was proposed that they should be
classified as types 6a, 6b, 7a, 7b, 7c, 7d, 8a, 8b, 9a, 9b, 9c, 10a
01(
and 11a (Tokita et al., 1994, 1995, 1996). Sugiyama et al. (1995)
described similar variants from Thailand as types VII and VIII.
In the current study, we have carried out phylogenetic
analysis of several subgenomic regions of HCV, followed by
bootstrap resampling to determine the robustness of the
observed groupings. The classification based upon this method
was compared with that derived from pairwise sequence
measurements in order to critically evaluate which method
provides the most suitable method for the classification of
HCV.
Methods
• Samples. Nucleotide sequences in the core and NS5 regions were
amplified by PCR from plasma from blood donors from Bangkok,
Thailand and Burma as previously described (Mellor et aI., 1995).
Amplified DNA was directly sequenced between base positions 15-322
(core) and 7975-8196 (NS5; sequencesnumbered as in Choo et al., 1991)
as previously described. The analysis also included previously published
sequences from core, E1 and different regions of NS5 (Doi et aL, 1996;
Tokita et al., 1994, 1995, 1996; Mellor el al., 1995, 1996; Sugiyamael al.,
1995; Apichartpiyakulel al., 1994).
In order to compare the ability of fragments of different sizes to
differentiate between genotypes, as well as to include as much of the
published sequence information as possible, we carried out analysis on a
total of six datasets. Sequences corresponding to 308 bases at the 5' end
of the core region have been obtained from 56 Thailand, Burmese and
Vietnamese samples (Mellor el al., 1995, 1996; Tokita et aL, 1994, 1995);
more sequences (n = 66) were availablefrom a shorter region (positions
27-277; length 251 bases) (Sugiyama et at., 1995). A total of 26
sequences were obtained from Vietnamese, Thailand and Indonesia
samples in the E1 region (positions 574-1149) (Tokita et al., 1994, 1995,
1996). Three sets of sequences were compared in the NS5 region. The
first consisted of the 3' terminal 1093 bases of NS5B, for which 25
sequences were compared (Tokita et al., 1994, 1995, 1996). Thirty-six
sequences were available from a 366 nucleotide fragment of NS5B
between positions 8464-8829 (Tokitaet al., 1994, 1995, 1996; Sugiyama
et al., 1995). The largest dataset for NS5 was obtained between positions
7975-8196 (n = 65) (Doi et al., 1996; Tokita el al., 1994, 1995, 1996;
Mellor et al., 1995, 1996; Apichartpiyakul el al., 1994). Because of the
very large number of sequences available for core, E1 and 222 base NS5
regions for genotypes 1-5, it was necessary to use a subset in the
phylogenetic analyses. Consequently, phylogenetic trees were reconstructed using two representatives, where available, of every other
previously described type and subtype (listed in Mellor et al., 1995).
• Sequenceanalysis.Evolutionarydistances (i.e. those corrected for
multiple substitution) between pairs of sequences were estimated using
the DNADIST program in the PHYLIP package (Felsenstein, 1993).
Frequency analysis of pairwise distances was assisted by the use of
standard statistical software (SYSTAT).Phylogenetic analysiswas carried
out using the programs DNADIST, PROTDIST, NEIGHBOR and
DRAWTREE in the PI-IYLIPpackage (Felsenstein, 1993). Trees (rooted
and unroofed) are shown with branch lengths drawn to scale. The
robustness of the groupings was assessed using bootstrap resampling of
100 replicate neighbour-joining trees (programs SEQBOOT, DNADIST,
NEIGHBOR and CONSENSE in the PHYLIP package; Felsenstein,
1993).
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
iiiiiiii!iii:iiiii:iiii
iiii ii! ii i
iiiiiiiiiiiiiiiii!iiiiil,
iiii!ii!
4d
4a(E)
/ 5a
6a
~t~ 7c/NGII/VII
2(I)
/
_~lla
g
~
NGI
2
3
h~ i'N,J "9c
9a 9b
lc(E)
/
-
l(II) la
/
/
3e 3d
3b
3(m)
TD:
10a
Results
Identification of HCV genotypes in South-East Asia
Eight previous studies describe the detection of novel
genotypes of HCV from samples in Thailand, Burma, Vietnam
and Indonesia (Doi et al., 1996; Tokita et al., 1994, 1995, 1996;
Mellor et al., 1995, 1996; Sugiyama et al., 1995; Apichartpiyakul et al., 1994). These independent studies have sometimes
analysed different regions of the HCV genome which complicates the analysis of their inter-relationships. However,
phylogenetic trees of sequences in the 222 base and 366 base
NS5 regions allowed almost all of variants to be compared (Fig.
I).
Sequences described as NG(II) (Mellor et al., 1995) were the
same genotype as those described as 7c (Tokita et al., 1995),
type VII (Sugiyama et al., 1995) and as BB7 and LD47/93 (Doi
et al., 1996), and represent one of the most frequently detected
variants in the four different studies. Some variants found in
northern Thailand (Apichartpiyakul et al., 1994) were equivalent to NG(I) (Mellor et al., 1995), while others (B492, PC)
were similar to a variant found in Burma, EUBURI (Mellor et
at., 1996). Neither of these variants was found in Vietnam
(Tokita et al., 1994). Similar analysis of a 251 base fragment of
the core region allowed comparison of sequences from all nine
previous studies, and of the 13 new core sequences obtained in
this study. Each of the groupings observed in the two regions
of NS5 were closely reproduced in the core region (data not
shown).
Combining the results from NS5 and core, a total of 14
distinct genotypes could be identified in four South-East Asian
Fig. 2. Phylogenetic analysis of
nucleotide sequences of HCV variants
from Hong Kong, Vietnam, Thailand,
Burma and Indonesia in the NS5 region
(222 bases; positions 7975-8196).
South-East Asian variants were compared
with one or two representative examples
of type 6a and other genotypes I - 5
using programs DNADIST and NEIGHBOR
in the PHYLIP package. Previously
published names for HCV genotypes are
indicated at tips of branches; clade
identifiers (Fig. I) are indicated along
branches within the type 6 group.
countries. Clusters a, d, e, k, l and h (described as types 6a, 7a,
7b, 8a, 8b and 9a) were found in Vietnam, while clusters b, c,
f, i, j, m and n were found in Thailand [described as 6b, 7d,
7c/NG(II)/VII, 9b, 9c, B492/PC and NG(I)] and in Burma
(EUBUR1). Cluster g (type 11a) was found in Indonesia. So far,
none of these variants show an overlapping geographical
distribution with any other variant.
Classification by phylogenetic analysis
Evolutionary relationships between the South-East Asian
variants and genotypes 1-5 and with type 6a were analysed by
phylogenetic analysis of nucleotide sequences in different
subgenomic regions. In order to compare all of the published
sequences, five datasets were used corresponding to core (308
bases), E1 and three different regions in NSS. The unrooted
phylogenetic tree for the 222 base region of NS5 was the most
complete (Fig. 2) but otherwise showed a topology typical of
each of the regions analysed, i.e. six major branches corresponding to genotypes 1-5 and a sixth branch that contains
type 6a as well as all of the variants from Vietnam and Thailand
and the variant from Indonesia described as 'type 1 la' (Tokita
et al., 1996) (Fig. 2). Another Indonesian variant described as
'type 10a' grouped with type 3 sequences and corresponded
to TDd, a variant previously found in Indonesia (Hotta et al.,
1994a).
Bootstrap resampling of phylogenetic trees was carried out
to test the robustness of the observed major clades (Table I).
Although there has been discussion about the statistical
meaning of bootstrap values (Li et al., 1994; Felsenstein &
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
31q
Table 1. Bootstrap resampling of phylogenetic groupings in different regions of the HCV genome
Next most frequent grouping
Clade*
Region
Core
E1
NS5
NS5
NS5
Complete
genome
Nucleotide
positions
1
2
3t
4
5
6
15-3220
574-1149
7975-8196
8464-8829
7938-9030
--341-9033
78
100
98
100
100
100
I00
100
I00
I00
I00
I00
84
100
90
100
100
100
100
100
96
NA
NA
NA
100
I00
100
NA
NA
NA
90
I00
90
92
100
NA
Branching
order
2(3(5(4,1)))
4((1,5)(2,3))
3(4,(2,(I,5)))
2,(1,3)
1,(2,3)
1,(2,3)
Genotypes¢
Bootstrap§
5---(4,1)
2-3
1-5
1--2
2--3
2--3
46/100
76/100
66/I00
49/100
73/100
64/100
* Number of 100 bootstrap resamplings that group variants into six clades corresponding to HCV genotypes I-5 and a sixth group containing
type 6a and variants from South-East Asia.
t Indonesian variants TD3 and 'type 10a' included in type 3 clade.
:1: Next most frequent grouping of genotypes.
§ Bootstrap value of next most frequent grouping.
N^, Not available.
Table 2. Distribution of evolutionary distances between type, subtype and isolates of HCV genotypes 1 - 6 in different
regions
Start*...
End*...
Length (bases)...
No. of sequencest ...
Isolate
Subtype
Type
Minimum value
Maximum value
Median
Minimum value
Maximum value
Median
Minimum value
Maximum value
Median
Core
15
322
308
93
El
574
1149
576
87
NS5
222 bases
7975
8196
222
173
0"000
0"082
0"033
0"057
O'153
0"099
0"084
0"295
0"192
0"000
0"I85
0'080
0'260
0'450
0"329
0"434
0"873
0"592
0"000
0-I51
0-052
0-I49
0"349
0"235
0"317
0'743
0"532
NS5
NS5
366 bases 1093 bases
8464
7938
8829
9030
366
1093
37
33
0"011
0"075
0"057
0"I07
0'246
0"192
0'291
0"573
0"377
0"008
0"074
0"060
0"I4I
0"227
0"197
0-340
0'487
0"410
Complete
genome
-- 341
9030
9371
21 t
0"009§
0"099
0"200
0"231
0-312
0'343
* Sequences numbered as in Choo et aI. (1991).
t Number of sequences of established genotype used to define ranges.
:1: Genotypes Ia (n = 3), lb (n = 13), lc(O) (n = 1), 2a (n = 1), 2b (n = 1), 3a (n = 2) and 3b (n = 1).
§ Uncorrected pairwise distances used to compare complete genomic sequences.
Kishino, I 9 9 3 ; Hillis & Bull, 1993), a cut-off value of a r o u n d
75 % is generally considered to establish confidence in the
phylogenetic groupings. Bootstrap values obtained for the
main clades were almost always much greater than this.
Whichever subgenomic region was analysed, variants of H C V
corresponding to g e n o t y p e s 1 - 5 consistently formed separate
groups with bootstrap values usually of 100% (the lowest
~01~
being for the type I clade in the core region with a bootstrap
value of 78 out of 100 replications).
Supporting the classification of South-East Asian variants
(apart from type lOa) as type 6 variants, bootstrap resampling
showed consistent g r o u p i n g of these variant within the type 6
clade with values r a n g i n g from 90 to 100 per 100 replications.
Similarly, T D 3 and ' t y p e lOa' consistently g r o u p e d with other
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
iiiii iiii iiii ii!iiiiii i i iiii ii i!iiiiiiii!iiiiiiiiiiiiiiiiiiii
iii
0.20
0.021
(a)
Core - 308 b.
0.149
0.074
1~
0.20
(b)
E1
0.477
0.15~
0.15
0.10
O.lO
0.05
l
0.05 I
r~
0.00 0.06 0.12 0.18 0.24 0.30
Evolutionarydistance
(c)
NS5 - 222 b.
0.067
0.414
0.15 q U
0.00 0.20 0.40 0.60 0.80
Evolutionarydistance
(d)
NS5 - 366 b.
0.039 0.248
0.25
11
U
(e)
o301°°4°
0.20
i
0.15
0.20 d
o.Io !
1.00
NS5 - 1093 b.
0.291
i
0.10 ]
0.05
0.10 t
k
0.05
0.00 0.15 0.30 0.45 0.60 0.75
Evolutionarydistance
0.00 0. i5 0.30 0.'~5 0.60 0.75
Evolutionarydistance
0.00 0.12 0.24 0.36 0.48 0.60
Evolutionarydistance
Fig. 3. Frequency histograms of pairwise distances between nucleotide sequences of variants in the type 6 clade in different
subgenomic regions: (a) Core, 308 bases, positions 15-322; (b) El, 576 bases, 574-1149; (c) NS5, 222 bases,
7975-8196; (d) NSS, 366 bases, 8464-8829; (e) : NS5, 1093 bases, 7938-9030. Median values for the two distributions
of sequence distances in each subgenomic region are indicated by arrows.
type 3 variants (bootstrap values 84 to 100 %). In contrast, no
robust higher-order grouping of other genotypes was found.
For example, the grouping of type 3 with type 4 sequences
(Fig. 2) was not supported by bootstrap analysis (Table 1). In
this region, the most commonly observed grouping was
between type 1 and 5 sequences (66% of trees, a value
insufficient to confer significance). Bootstrap resampling of
sequences in other subgenomic regions also failed to reveal
associations between genotypes 1-5, and the groupings that
were observed varied between core, E1 and NS5. For example,
the most frequent pairwise grouping in E1 was between
genotypes 2 and 3 (76 %), compared with 5 and 1 + 4 combined
for core (46 %) and I and 5 for NS5 (66 %). Even using complete
genomic sequences, no grouping other than into genotypes 1,
2 and 3 and a separate branch for 'type 11a' was observed.
A separate bootstrap analysis of trees produced by the
program PROTDIST using the inferred amino acid sequences
from each NS5 region and E1 produced similar results to those
in Table I for nucleotide sequences. Each region supported the
grouping of South-East Asian variants with type 6a into a
single clade, and none provided consistent evidence for
significant grouping of other genotypes. The amino acid
sequences in the core region were too invariant to produce
reproducible groupings.
Classification by sequence similarity
Using collections of sequences of established genotype
(types 1-5 and 6a), the distributions (minimum and maximum
values, medians) of pairwise evolutionary distances between
type, subtype and isolate for different subgenomic regions
(core, 308 bases; El; NS5, 222 bases; NS5, 366 bases; NS5,
1093 bases) and of complete genomic sequences were
calculated (Table 2).
The distribution of evolutionary distances within the
different clades observed in the subgenomic regions has been
used as an alternative method to determine how variants may
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
;01!
(a)
10a JK049~
3b Tr
/
\
1la JK046
/
!
3a NZL1"
2b HCJ8"
I
2a HCJ6
(b)
1la JK046
10a JK049\
3b Tr
/
\
l
3a NZL1"
"~]~.~ ~
/
lb
HCV-BK
2b HC-J8"
I
2a HC-J6
Fig. 4. Phylogenetic analysis of (a) nucleotide or (b) inferred amino acid sequences of complete genomes of types 1,2, 3,
'type lOa' and 'type 1 l a ' (Tokita et al., 1996), shown as unrooted trees. Bootstrap values for the type 1, 2 and 3 (including
JK049) clades are shown in Table 1.
be classified. For example, the distribution of sequence
distances between the currently classified genotypes (1-5, 6a)
is tri-modal. The first peak corresponds to the set of pairwise
distances within types, the second to distances between
subtypes, while the third corresponds to distances within
subtypes. This analysis was used previously to substantiate the
proposed type, subtype and isolate categories used for HCV
classification (Simmonds et al., 1994a).
This analysis was extended to the five subgenomic regions
described above to investigate the relationships between
sequences falling within the type 6 clade (Fig. 3). The finding
of three separate peaks of sequences distances would support
the previous proposal to classify them into a novel series of
types, subtypes, and isolates (types 6a, 6b, 7a, 7b, 7c, 7d, 8a,
8b, 9a, 9b, 9c and lla), while two distributions would support
their classification as subtypes of type 6.
~02C
In practice, each of the subgenomic regions produced a
distribution with two distinct peaks (Fig. 3 a-e). For example, in
the core region the first peak of evolutionary distances had a
median value of 0"021, similar to that observed between
isolates within established HCV subtypes (0"033; Fig. 3a;
Table 2). The second peak had a median value of 0"I49,
intermediate between the values previously calculated between
subtypes and between genotypes of currently classified
variants of HCV (0"099 and 0"192 respectively; Fig. 3 a; Table
2). Similar bimodal distributions of sequence distances were
observed in each of the other four subgenomic regions
amongst the set of sequences obtained from South-East Asia.
For example, the median of the second peak for the 222 base
fragment of NS5 was 0"414, almost halfway between the
median values for subtype (0"235) and type (0-532; Table 2).
Even using the longest fragment of NS5 (I093 nucIeotides),
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
Type Subtype
(a)
Type
Isolate
Subtype
Isolate
(b)
)
j
Core
I
m
I
i
I
I
I
I
m
I
I
~
i
I
l
I
t
I
E2
I
l
I
I
I
NS2
J
NS3
i
~
I
I
I
I
,I
I
t
40
I
I
I
i
m
NS4a
l
•
NS5a
J
NS5b
30
20
10
% Sequence Divergence
genome
I
0
i
I
I
I
I
I
I
I
I
I
J
I
I
m
L
1
I
40
I
II
I
E
I
i
I
m
j
1
I
I
l
I
Complete
i
. I
m
[
NS4b
|
I
I
I
I
m
I
.
m m
1
j
) ]
I
•
F-]
I
mira
t.
I
i
I
I
El
I
F---3
I
I
~
)
i
i
I
)
i
i
i
30
20
10
% Sequence Divergence
I
0
Fie3. 5. Ranges of pairwisedistances betweentype, subtypeand isolateof 22 complete9enomicsequences of types 1a (n =
3), I b (n = 13), Ic(0) (n = I), 2a (n = I), 2b (n = I), 3a (n = 2) and 3b (n = I), and 'type lOa'. In (a), ranges for
complete genomic sequences and for subgenomic fragments corresponding to each of the HCV genes, core through NS5B,
were calculated based upon the classification of 'type I Oa' as a separate major genotype. The ranges obtained by its
alternative classification as a subtype of type 3 allow a much clearer discrimination of types and subtypes (b). Vertical lines on
scale bars represent increments of 10% sequence divergence.
the median value of distances between South-East Asian
genotypes was intermediate (0"291 compared with medians of
0"197 and 0"410 for subtype and type respectively). In this
region, none of the pairwise distances between variants within
the type 6 group overlapped with distances between genotypes 1-3 (maximum value for type 6 sequences 0"336,
minimum distance between previously defined genotypes
0"340; Table 2), showing that the intermediate status of the
South-East Asian variants did not arise from simply comparing
inadequate lengths of sequence.
Analysis of complete genomic sequences
Tokita et al. (1996) have suggested that the current
uncertainty and difficulty in classifying type 6-group variants
may be resolved by obtaining complete genomic sequences.
Although the complete sequence of the variant from Indonesia
('type 1Ia') was described, this remains of little use in the
absence of any other variant in the type 6 group with which to
compare it. However, an analogous comparison can be made
between the other Indonesian variant ('type 10a') with the
complete genomic sequences of type 3a and 3b, as the
sequences show similar ambiguous inter-relationships to those
found within the type 6 clade.
By phylogenetic analysis in any of the subgenomic regions,
'type 10a' consistently grouped within the type 3 clade (Table
I). Similarly, analysis of complete nucleotide (Fig. 4a)
sequences showed the close evolutionary relationship of 'type
10a' with type 3a and 3b. The branching point for 'type 10a'
with type 3 was only slightly displaced from the branch of
type 3a and 3b. In contrast, no grouping is found between any
other pair of genotypes (bootstrap values in Table I), and each
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
m
clade appears to radiate from a single point. Similar relationships exist between amino acid sequences translated from the
complete genomes (Fig. 4 b). Using evolutionary or phenotypic
analysis, there is no justification for regarding 'type 10a' as a
separate genotype from type 3.
However, Tokita et al. proposed that 'type 10a' belonged
to a separate major genotype because of its greater divergence
from all known complete genome sequences (25 %) than was
previously found between subtypes (range 20"0--23"1%; Table
2). Similarly, pairwise distances outside the normal subtype
range were also found in each of the subgenomic regions (data
not shown). However, the pairwise distance between type 3b
and 'type 10a' was also considerably smaller than between
other genotypes (25 % compared with 31"2 %). Indeed, it might
be more parsimonious to regard 'type 10a' as a subtype of
type 3 (and extend the subtype range by 1"9%) rather than to
regard 'type 10a' as a separate genotype [and therefore reduce
the range of distances between types by 6"2 % (from 31"2 % to
25 %)].
The effect of these two alternative classifications on the
ranges between type and subtype in subgenomic regions is
shown in Fig. 5. Classifying 'type 10a' as a separate type
reduces the lowest pairwise distance between types to a point
where it becomes dose to or even overlaps the upper limit of
the subtype range (Fig. 5a) in each subgenomic region. In
contrast, classification of 'type 10a' as a subtype of 3 slightly
increases the subtype range in each subgenomic region, but
maximizes the separation between categories (Fig. 5 b). Clear
differentiation of type and subtype can therefore be observed
in each of the larger and more variable subgenomic regions
(such as NS5B, NS5A, NS4B, NS3 and El).
Discussion
Implications for HCV classification
The main finding of this analysis of HCV variants from
South-East Asia was the consistent grouping of most of these
variants with type 6a upon phylogenetic analysis of several
subgenomic regions. This grouping was found in each
subgenomic region analysed and each appeared to be equally
informative in the phylogenetic analysis. In contrast, no robust
grouping was observed between other genotypes (types 1-5 ;
Table 1). The second finding was that the set of pairwise
distances between different genotypes amongst the South-East
Asian variants formed a single distribution in each subgenomic
region. Minimum and maximum values ranged continuously
from those observed between subtypes of other genotypes
through to distances found between different major genotypes
(Table 2).
Clearly, variability within this group is structured differently from that between other HCV variants, making classification into types and subtypes according to previous criteria
difficult to apply (Simmonds et al., 1994a). For example,
genotypes from Vietnam were classified as types 7a, 7b, 8a, 8b
~02~
and 9a by comparison of their pairwise distances with ranges
of distances previously found between other genotypes and
subtypes. Variants with pairwise distances below the previously established 'cut-off' value of 20 % (Tokita et aI., 1994)
(or evolutionary distances of 0"23; Mellor et al., 1995) for the
1093 base fragment of NS5, would be subtypes (7a, 7b and 8a,
8b) while those with values greater would be different types.
This simple approach is justified for classification of sequences
when it is applied to discrete distributions of pairwise distances
(Simmonds et al,, 1993). However, the same 'cut-off' value
cannot be used to classify sequences with a unimodal
distribution of pairwise distances that has a median distance
close to this cut-off value, as is the case for the type 6 clade.
Indeed, we have previously found it impossible to consistently
classify as types or subtypes variants from Thailand previously
described as NG(I) and NG(II). For example, the pairwise
distance between NG(II) and type 7a for the 222 nucleotide
region of NS5 was 0"31 (in the subtype range), but 0"35 with
type 7b (type range). Similar difficulties will undoubtedly be
encountered as more sequences from the type 6 clade become
available, and perhaps for other clades such as type 3.
To resolve this difficulty, we have previously suggested
that this group might more usefully be regarded as a single
major genotype that contains members that are more divergent
from each other than are subtypes of other genotypes (Mellor
el al., 1995). This proposal fits better with the observation of a
single distribution of sequence distances, and their consistent
grouping by phylogenetic analysis. As an evolutionary
classification it would also be more appropriate for the
description of a continuous range of variants found in a
relatively restricted geographical area (Vietnam, Thailand,
Burma, Indonesia; Mellor et al., 1996). However, this proposal
would effectively rule out the straightforward use of sequence
similarity as a means of virus classification.
The nomenclature for South-East Asian variants is influenced by the method of sequence comparison. Whereas
analysis of these variants on the basis of sequence similarity
has led to the suggestion of four new major genotypes (types
7, 8, 9 and 11), each with three or four subtypes, as well as a
new subtype of type 6, 6b (Tokita et al., 1994, 1995),
phylogenetic analysis implies a single genotype (type 6)
comprising 14 relatively diverse subtypes (Figs 1 and 2).
Although most of the analysis described in this study has
been confined to variants of HCV grouping within the type 6a
clade, there is evidence that the phenomenon of highly
divergent subtypes may also occur amongst other HCV
genotypes. For example, the variants 'type I0a' and TD3
obtained from Indonesia (Tokita et al., 1996; Hotta et al.,
1994b) consistently grouped with type 3 variants upon
phylogenetic analysis (Fig. 2; Table 1), although they showed
sequence similarity values with type 3 variants generally in the
type range (evolutionary distances in the NS5 222 base
fragment of 0"323-0"412). It is possible that more extensive
analysis of HCV variants in Indonesia and in neighbouring
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
iiiiiiiiiiiiiiiiiiiii iii ii ii iii ii@
!iiiiiiiiiii!iiiiiiiiiiiiiiiiii
countries might reveal a similarly diverse range of genotypes
phylogenetically related to type 3, and pose the same problems
for classification as do the type 6 variants.
Aims of HCV classification
The classification of virus variants must not only be
accurate in describing reaI and consistent grouping of variants
as discussed above, but should also be of some practical
benefit. So far, biological differences have been consistently
observed between individuals infected with HCV type 1 and
types 2 and 3 in terms of the response to interferon treatment
(Miyakawa et al., 1995; Davis, 1994; Bukh eta]., 1994), but
there are no convincing data for clinical or virological
differences between virus subtypes (Lau et al., 1996; Mahaney
et al., 1994). Similarly, serological differences have been
demonstrated between virus types 1--6 (Bhattacherjee et aL,
1995; Tsukiyama Kohara et al., 1993; McOmish et al., 1993)
but not yet between virus subtypes.
As yet, clinical or virological information about type 6
clade viruses is sparse, and it is unclear whether members of the
clade behave as diverse subtypes with common biological
properties, or as genotypes with distinct characteristics.
However, we have recently observed that samples from
individuals infected with type 6 clade variants (other than 6a)
generally show a specific serological reactivity to type 6a
peptides in a serotyping ELBA (Bhattacherjee et al., 1995)
(and unpublished results). This suggests that the phylogenetic
grouping of variants within the type 6 clade may also reflect
serological properties of the variants. In the future it will be
possible to investigate whether other biological and clinical
similarities are also found amongst members of the type 6
clade.
the repeated detection of certain genotypes such as type
7c/VII/NG(II) in Thailand indicates the more recent and more
rapid spread of certain variants possibly in association with
new risk groups for HCV infection (Holmes et al., 1995).
Over much longer periods, it is possible that the process of
diversification within the type 6 clade will proceed further to
produce HCV genotypes with no detectable residual commonality, as currently appears to be the case for genotypes
1-5. However, this process is relatively slow. Estimates of the
rate of sequence change of HCV from a large cross-sectional
study of individuals infected with HCV 17 years previously
from a common source ranged from 0"11% per site per year in
NS5 to 0"16% in E1 at silent sites. Using these rates, the current
diversity between different isolates within a subtype such as
type Ib predicts a common ancestor approximately 70--80
years ago (D. B. Smith and others, unpublished data). Although
constraints on sequence change make it more difficult to
extrapolate divergence times between different subtypes, such
as type la and lb, a minimum estimate would be 300 years,
correcting for differences in the transitions and transversions
and silent and non-silent substitutions.
The current diversity observed in Vietnam and Thailand
therefore represents a period of transmission within the local
population considerably longer than any of these estimates.
The existence of ancient lineages heightens the difficulties in
understanding the persistence of a virus infection within
human societies in which parenteral exposure has been
documented to be the only efficient route of transmission.
References
Apichartpiyakul, C., Chlttivudikarn, C., Miyajima, H., Homma, H. &
Hotta, H. (1994). Analysis of hepatitis C virus isolates among healthy
blood donors and drug addictsin Chiang Mai, Thailand.Journal of Clinical
Microbiology 32, 2276--2279.
Molecular epidemiology of HCV g e n o t y p e s
Although a number of issues remain concerning the
nomenclature of HCV variants based upon different methods
of sequence comparison, the analysis presented in this paper
provides a new insight into the process of HCV sequence
diversification. We have previously suggested that the observed diversity of HCV variants in different geographical
regions may reflect the history of transmission within a
community. For example, the numerous subtypes of type 4 in
central Africa and of type 3 in the Indian subcontinent may
reflect a long-standing endemic pattern of virus transmission,
while the restricted subtype distribution elsewhere (such as 4a
in Egypt, and type 3a in drug-users in Europe) is evidence for
their recent spread into new risk groups (Mellor et al., 1995).
In the case of the variants found in South-East Asia, even
highly limited sampling of the population provides an
unusually diverse range of sequences. The existence of
numerous co-existing lineages suggests a long-term stability in
virus transmission within these populations. At the same time,
Bhattacherjee, V., Prescott, L.E., Pike, I., Rodgers, B., Bell, H,
Elzayadi, A. R., Kew, M. C., Conradie, J., Lin, C. K., Marsden, H., Saeed,
A.A., Parker, D., Yap, P. L & Simmonds, P. (1995). Use of NS-4
peptides to identifytype-specificantibody to hepatitis C virus genotypes
1, 2, 3, 4, 5 and 6. Journal of General Virology 76, 1737-1748.
Bukh, J. (1995). Geneticheterogeneity of hepatitis C virus: quasispecies
and genotypes. Seminars in Liver Disease 15, 41--63.
Bukh, J. & Miller, R. H. (1994). Diagnostic and clinical implications of
the differentgenotypes of hepatitis C virus. Hepatology 20, 256--259.
Bukh, 1., Purcell, R. H. & Miller, R. H. (1993). At least 12 genotypesof
hepatitis C virus predictedby sequenceanalysis of the putative E1 gene
of isolates collected worldwide. Proceedings of the National Academy of
Sciences, USA 90, 8234-8238.
Bukh, J., Purcell, R. H. & Miller, R. H. (1994). Sequenceanalysis of the
core gene of 14 hepatitis C virus genotypes. Proceedings of the National
Academy of Sciences, USA 91, 8239-8243.
Choo, Q. L., Richman, K. H., Han, J. H., Berger, K., Lee, C., Dong, C.,
Gallegos, C., Colt, D., Medina Selby, R., Barr, P.J., Weiner, A.J.,
Bradley, D. W., Kuo, G. & Houghton, M. (1991). Genetic organization
and diversity of the hepatitis C virus. Proceedingsof the National Academy
of Sciences, USA 88, 2451-2455.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47
302~
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii
iiiiiiiiiiiiiiiiiii
iiii!iiii
Davis, G. L. (1994). Prediction of response to interferon treatment of
chronic hepatitis C. Journal of Hepatology 21, 1-3.
Doi, H., Apichartpiyakul, C., Ohba, K. I., Mizokami, M. & Hotta, H.
Sakamoto, M., Akahane, Y., Tsuda, F., Tanaka, T., Woodfleld, D. G. &
Okamoto, H. (1994). Entire nucleotide sequence and characterization of
a hepatitis C virus of genotype v/3a. Journal of General Virology 75,
1761-1768.
(1996). Hepatitis C virus (HCV) subtype prevalence in Chaing Mai,
Thailand and identification of novel subtypes of HCV in major type 6.
Journal of ClinicalMicrobiology 34, 569-574.
Simmonds, P., Holmes, E. C., Cha, T. A., Chan, S.-W., HcOmish, F.,
Irvine, B., Beall, E., Yap, P. L., Kolberg, J. & Urdea, H.S. (1993).
Felsenstein, J. (1993). PHYLIP inference package version 3.5. Seattle:
Department of Genetics, University of Washington.
Classification of hepatitis C virus into six major genotypes and a series of
subtypes by phylogenetic analysis of the NS-5 region. Journal of General
Virology 74, 2391-2399.
Felsenstein, J. & Kishino, H. (1993). Is there something wrong with
bootstrap on phylogenies? A reply to Hillis and Bull. Systematic Biology
42, 193-200.
Hi,is, D. M. & Bull, J. J. (1993). An empirical test of bootstrapping as a
method for assessing confidence in phylogenetic analysis. Systematic
Biology 42, 182-192.
Holmes, E.C., Nee, S., Rambaut, A., Garnett, G. P. & Harvey, P.H.
(1995). Revealing the history of infectious disease epidemics using
phylogenetic trees. PhilosophicalTransactionsof the Royal Societyof London
Series B BiologicalSciences349, 33-40.
Hotta, H., Doi, H., Hayashi, T., Purwanta, M., Soemarto, W., Mizokami,
M., Ohba, K. & Homma, l . (1994o). Analysis of the core and e1
envelope region sequences of a novel variant of hepatitis C virus
obtained in Indonesia. Archives of Virology 136, 53-62.
Hotta, H., Handajani, R., Lusida, M. I., Soemarto, W., Doi, H., Miyajima,
H. & Homma, M. (1994b). Subtype analysis of hepatitis C virus in
Indonesia on the basis of NSSb region sequences. Journal of Clinical
Microbiology 32, 3049-3051.
Lau, I. Y. N., Davis, G. L., Prescott, L. E., Maertens, G., Lindsay, K. L.,
Qian, K. P., Mizokami, M., Simmonds, P., Perrillo, R. P., Schiff, E. R.,
Bodenheimer, H.C., Balart, L.A., Regenstein, F., Dienstag, J.L.,
Katkov, W. N., Tamburro, C. H., Gaff, J. S., Everson, G. T., Goodman, Z.
& Albrecht, J, (1996). Distribution of hepatitis C virus genotypes
determined by line probe assay in patients with chronic hepatitis C seen
at tertiary referral centers in the United States. Annals of InternalMedicine
124, 868.
Li, W.H. & Zharkikh, A. (1994). What is the bootstrap technique?
Systematic Biology 43, 424-430.
Mahaney, K., Tedeschi, V., Maertens, G., Dibisceglie, A. H., Vergalla,
J., Hoofnagle, J. H. & Sallie, R. (1994). Genotypic analysis of hepatitis
C virus in American patients. Hepatology 20, 1405-1411.
McOmish, F., Chan, S.-W., Dow, B.C., Gillon, J., Frame, W.D.,
Crawford, R. J., Yap, P. L., Fotlett, E. A. C. & Simmonds, P. (1993).
Detection of three types of hepatitis C virus in blood donors:
investigation of type-specific differences in serological reactivity and rate
of alanine aminotransferase abnormalities. Transfusion 33, 7-13.
Mellor, J., Holmes, E. C., Jarvis, L. H., Yap, P.-L., Simmonds, P. & The
International HCV Collaborative Study Group (1995). Investigation of
the pattern of hepatitis C virus sequence diversity in different
geographical regions: implications for virus classification. Journal of
General Virology 76, 2493-2507.
Mellor, J., Walsh, E. A., Prescott, L. E., Jarvis, L. M., Davidson, F., Yap,
P. L., Simmonds, P., Nowlcki, M. J., Mosley, J. W., Lin, C. K., Lai, C. L.,
Deolim, G., Martins, |. A., Ong, Y. W., Teo, D., Lin, M., Nuchprayoon, C.
& Tanprasert, S. (1996). Survey of type 6 group variants of hepatitis C
virus in southeast Asia by using a core-based genotyping assay. Journal
of ClinicalMicrobiology 34, 417-423.
Miyakawa, Y., Okamoto, H. & Mayumi, M. (1995). Classifying hepatitis
C virus genotypes. MolecularMedicine Today 1, 20-25.
02~
Simmonds, P., Alberti, A., Alter, H. ]., Bonino, F., Bradley, D.W.,
Brechot, C., Brouwer, J.T., Chan, S.W., Chayama, K., Chen, D.S.,
Choo, Q. L., Colombo, M., Cuypers, H. T. M., Date, T., Dusheiko, G. H.,
Esteban, J. I., Fay, 0., Hadziyannis, S. J., Han, J., Hatzakis, A., Holmes,
E. C., Hotta, H., Houghton, H., Irvine, B., Kohara, M., Kolberg, J. A.,
Kuo, G., Lau, J.Y.N., Lelie, P.N., Haertens, G., HcOmish, F.,
Hiyamura, T., Hizokami, H., Nomoto, A., Prince, A. H., Reesink, H. W.,
Rice, C., Roggendorf, H., Schalm, S. W., Shikata, T., Shimotohno, K.,
Stuyver, L., Trepo, C., Weiner, A., Yap, P. L. & Urdea, H. S. (19940).
A proposed system for the nomenclature of hepatitis C viral genotypes.
Hepatology 19, 1321-1324.
Simmonds, P., Smith, D.B., McOmish, F., Yap, P.L., Kolberg, J,
Urdea, H. S. & Holmes, E. C. (1994b). Identification of genotypes of
hepatitis C virus by sequence comparisons in the core, E1 and NS-5
regions. Journal of General Virology 75, 1053-1061.
Smith, D. B., Mellor, J., Jarvis, L H., Davidson, F., Kolberg, J., Urdea,
H., Yap, P.-L., Simmonds, P. & The International HCV Collaborative
Study Group (1995). Variation of the hepatitis C virus 5' non-coding
region: implications for secondary structure, virus detection and typing.
Journal of General Virology 76, I749-1761.
Stuyver, L., Vanarnhem, W., Wyseur, A., Hernandez, F., Delaporte, Eo
& Haertens, G. (1994). Classification of hepatitis C viruses based on
phylogenetic analysis of the envelope I and nonstructural 5b regions and
identification of five additional subtypes. Proceedings of the National
Academy of Sciences, USA 91, 10134-10138.
Sugiyama, K., Kato, N., Nakazawa, T., Yonemura, Y., Phornphutkul, K.,
Kunakorn, H., Petchclai, B. & Shimotohno, K. (1995). Novel genotypes
of hepatitis C virus in Thailand.Journalof GeneralVirology76, 2323-2327.
Tokita, H., Okamoto, H., Tsuda, F., Song, P., Nakata, S., Chosa, T.,
lizuka, H., Hishiro, S., Miyakawa, Y. & Mayumi, H. (1994). Hepatitis C
virus variants from Vietnam are classifiable into the seventh, eighth, and
ninth major genetic groups. Proceedingsof the NationalAcademy of Sciences,
USA 91, 11022-11026.
Tokita, H., Okamoto, H., Luengrojanakul, P., Vareesangthip, K.,
Chainuvati, T., lizuka, H.~ Tsuda, F., Hiyakawa, Y. & Mayumi, t .
(1995). Hepatitis C virus variants from Thailand classifiable into five
novel genotypes in the sixth (6b), seventh (7c, 7d) and ninth (9b, 9c)
major genetic groups. Journal of General Virology 76, 2329--2335.
Tokita, H., Okamoto, H., lizuka, H., Kishimoto, J., Tsuda, F., Lesmana,
L.A., Miyakawa, Y. & Mayumi, M. (1996). Hepatitis C virus variants
from Jakarta, Indonesia classifiable into novel genotypes in the second (2e
and 2f), tenth (10a) and eleventh (11a) genetic groups. Journal of General
Virology 77, 293-301.
Tsukiyama Kohara, K., Yamaguchi, K., Haki, N., Ohta, Y., Miki, K.,
Mizokami, M., Ohba, K., Tanaka, S., Hattori, N., Nomoto, A. & Kohara,
M. (1993). Antigenicities of group I and group II hepatitis C virus
polypeptides-molecular basis of diagnosis. Virology 192, 430-437.
Received 17 June 1996; Accepted 6 August 1996
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 17:07:47