Genomic epidemiology of a dengue virus epidemic in urban Singapore.
ABSTRACT Dengue is one of the most important emerging diseases of humans, with no preventative vaccines or antiviral cures available at present. Although one-third of the world's population live at risk of infection, little is known about the pattern and dynamics of dengue virus (DENV) within outbreak situations. By exploiting genomic data from an intensively studied major outbreak, we are able to describe the molecular epidemiology of DENV at a uniquely fine-scaled temporal and spatial resolution. Two DENV serotypes (DENV-1 and DENV-3), and multiple component genotypes, spread concurrently and with similar epidemiological and evolutionary profiles during the initial outbreak phase of a major dengue epidemic that took place in Singapore during 2005. Although DENV-1 and DENV-3 differed in viremia and clinical outcome, there was no evidence for adaptive evolution before, during, or after the outbreak, indicating that ecological or immunological rather than virological factors were the key determinants of epidemic dynamics.
- SourceAvailable from: Julia M Martínez Gómez[Show abstract] [Hide abstract]
ABSTRACT: Dengue (DEN) represents the most serious arthropod-borne viral disease. DEN clinical manifestations range from mild febrile illness to life-threatening hemorrhage and vascular leakage. Early epidemiological observations reported that infants born to DEN-immune mothers were at greater risk to develop the severe forms of the disease upon infection with any serotype of dengue virus (DENV). From these observations emerged the hypothesis of antibody-dependent enhancement (ADE) of disease severity, whereby maternally acquired anti-DENV antibodies cross-react but fail to neutralize DENV particles, resulting in higher viremia that correlates with increased disease severity. Although in vitro and in vivo experimental set ups have indirectly supported the ADE hypothesis, direct experimental evidence has been missing. Furthermore, a recent epidemiological study has challenged the influence of maternal antibodies in disease outcome. Here we have developed a mouse model of ADE where DENV2 infection of young mice born to DENV1-immune mothers led to earlier death which correlated with higher viremia and increased vascular leakage compared to DENV2-infected mice born to dengue naïve mothers. In this ADE model we demonstrated the role of TNF-α in DEN-induced vascular leakage. Furthermore, upon infection with an attenuated DENV2 mutant strain, mice born to DENV1-immune mothers developed lethal disease accompanied by vascular leakage whereas infected mice born to dengue naïve mothers did no display any clinical manifestation. In vitro ELISA and ADE assays confirmed the cross-reactive and enhancing properties towards DENV2 of the serum from mice born to DENV1-immune mothers. Lastly, age-dependent susceptibility to disease enhancement was observed in mice born to DENV1-immune mothers, thus reproducing epidemiological observations. Overall, this work provides direct in vivo demonstration of the role of maternally acquired heterotypic dengue antibodies in the enhancement of dengue disease severity and offers a unique opportunity to further decipher the mechanisms involved.PLoS Pathogens 04/2014; 10(4):e1004031. · 8.14 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Coalescent methods are widely used to infer the demographic history of populations from gene genealogies. These approaches-often referred to as phylodynamic methods-have proven especially useful for reconstructing the dynamics of rapidly evolving viral pathogens. Yet population dynamics inferred from viral genealogies often differ widely from those observed from other sources of epidemiological data, such as hospitalization records. We demonstrate how a modeling framework that allows for the direct fitting of mechanistic epidemiological models to genealogies can be used to test different hypotheses about what ecological factors cause phylodynamic inferences to differ from observed dynamics. We use this framework to test different hypotheses about why dengue serotype 1 (DENV-1) population dynamics in southern Vietnam inferred using existing phylodynamic methods differ from hospitalization data. Specifically, we consider how factors such as seasonality, vector dynamics and spatial structure can affect inferences drawn from genealogies. The coalescent models we derive to take into account vector dynamics and spatial structure reveal that these ecological complexities can substantially affect coalescent rates among lineages. We show that incorporating these additional ecological complexities into coalescent models can also greatly improve estimates of historical population dynamics and lead to new insights into the factors shaping viral genealogies.Molecular Biology and Evolution 10/2013; · 10.35 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Long-term disease surveillance data provide a basis for studying drivers of pathogen transmission dynamics. Dengue is a mosquito-borne disease caused by four distinct, but related, viruses (DENV-1-4) that potentially affect over half the world's population. Dengue incidence varies seasonally and on longer time scales, presumably driven by the interaction of climate and host susceptibility. Precise understanding of dengue dynamics is constrained, however, by the relative paucity of laboratory-confirmed longitudinal data.PLoS neglected tropical diseases. 07/2014; 8(7):e3003.
JOURNAL OF VIROLOGY, May 2009, p. 4163–4173
Copyright © 2009, American Society for Microbiology. All Rights Reserved.
Vol. 83, No. 9
Genomic Epidemiology of a Dengue Virus Epidemic in Urban Singapore?†
Mark J. Schreiber,1* Edward C. Holmes,2,3Swee Hoe Ong,1,4,5Harold S. H. Soh,6Wei Liu,1
Lukas Tanner,1Pauline P. K. Aw,4Hwee Cheng Tan,7Lee Ching Ng,7
Yee Sin Leo,8Jenny G. H. Low,8Adrian Ong,8Eng Eong Ooi,9
Subhash G. Vasudevan,1‡ and Martin L. Hibberd4
Novartis Institute for Tropical Diseases, 10 Biopolis Road, Chromos 05-01, Singapore 1386701; Center for Infectious Disease Dynamics,
Department of Biology, The Pennsylvania State University, University Park, Pennsylvania 168022; Fogarty International Center,
National Institutes of Health, Bethesda, Maryland 208923; Genome Institute of Singapore, A*STAR, 60 Biopolis Street,
Genome 02-01, Singapore 1386724; Swiss Tropical Institute, Socinstrasse 57, P.O. Box CH-4002, Basel, Switzerland5;
Institute for High Performance Computing, A*STAR, 1 Science Park Road, 01-01 The Capricorn, Singapore Science Park II,
Singapore 1175286; National Environment Agency, 40 Scotts Road, 13-00, Singapore 2282317;
Tan Tock Seng Hospital, 11 Jalan Tan Tock Seng, Singapore 3084338; and
DSO National Laboratories, 27 Medical Drive, 09-01, Singapore 1175109
Received 26 November 2008/Accepted 3 February 2009
Dengue is one of the most important emerging diseases of humans, with no preventative vaccines or antiviral
cures available at present. Although one-third of the world’s population live at risk of infection, little is known
about the pattern and dynamics of dengue virus (DENV) within outbreak situations. By exploiting genomic
data from an intensively studied major outbreak, we are able to describe the molecular epidemiology of DENV
at a uniquely fine-scaled temporal and spatial resolution. Two DENV serotypes (DENV-1 and DENV-3), and
multiple component genotypes, spread concurrently and with similar epidemiological and evolutionary profiles
during the initial outbreak phase of a major dengue epidemic that took place in Singapore during 2005.
Although DENV-1 and DENV-3 differed in viremia and clinical outcome, there was no evidence for adaptive
evolution before, during, or after the outbreak, indicating that ecological or immunological rather than
virological factors were the key determinants of epidemic dynamics.
The phylogenetic analysis of gene sequence data is com-
monly used to determine the origins of disease outbreaks,
particularly those caused by rapidly evolving RNA viruses (1, 2,
7, 9, 16, 26). Historically, such molecular epidemiological stud-
ies have usually utilized a small number of genes and have
largely concentrated on determining the origins of disease out-
breaks and retracing their pathways of spread. As such, these
studies have rarely been able to shed light on the precise
spatial and temporal dynamics of viral transmission. Recently,
whole-genome sequencing of viruses has been utilized to pro-
vide greater phylogenetic resolution on outbreak dynamics (12,
14, 15, 20) and is likely to become the benchmark in the near
One disease where genomic sequence may be especially
informative is dengue fever, an acute febrile disease caused by
a mosquito-borne RNA virus (dengue virus [DENV]; single-
strand positive-sense, family Flaviviridae) and the most com-
mon vector-borne viral infection of humans; some 100 million
dengue cases are reported on an annual basis, with epidemics
especially common in Southeast Asia (10, 11). The potential
expansion of the viable geographic range for Aedes aegypti
mosquitoes following global warming coupled with the current
lack of an effective vaccine or antiviral therapies make under-
standing the epidemic dynamics of this important emerging
virus a key priority.
Dengue has the ability to cause major outbreaks in urban
settings, often with high levels of morbidity. The Singapore
dengue outbreak of 2005 was the largest of its kind in this
locality, with a case rate of 335 per 100,000 population (17).
The 2005 outbreak was also notable in that it was characterized
by the cocirculation of two of the four viral serotypes—
DENV-1 and DENV-3—combined with a low level of
DENV-2 transmission. The resurgence of dengue in Singapore
is particularly striking given that an aggressive vector control
program started in 1970 has resulted in a very low household
index of mosquito breeding sites (21). Lessons learned in Sin-
gapore, with its long history of commitment to dengue control,
may therefore be vital to the overall global effort in dengue
By undertaking a detailed analysis of whole-genome data
sampled from a major outbreak of DENV in Singapore during
2005—the first of its kind—we demonstrate how a synthesis of
comparative genomics and fine-scale spatial and temporal ep-
idemiological data provides unprecedented power to reveal the
origins, causes, and dynamics of this important emerging virus
in a densely population urban environment.
MATERIALS AND METHODS
Collection of viral samples. Viral samples were collected as described in Low
et al. (17). Briefly, blood samples were collected by a research nurse from
* Corresponding author. Mailing address: Novartis Institute for
Tropical Diseases, 10 Biopolis Road, Chromos 05-01, Singapore
138670. Phone: (65) 672-2973. Fax: (65) 672-2910. E-mail: mark
† Supplemental material for this article may be found at http://jvi
‡ Present address: Emerging Infectious Diseases Program, Duke-
NUS Graduate Medical School, 2 Jalan Bukit Merah, Singapore
?Published ahead of print on 11 February 2009.
consenting patients presenting with fever of ?38°C for less than 72 h. Portions (1
ml) of serum from samples confirmed as dengue positive by reverse transcrip-
tion-PCR (RT-PCR) were inoculated onto the Aedes albopictus mosquito (C6/
36) cell line (ATCC CRL-1660). Cells were incubated at 37°C for up to 10 days
or until 75% of the cell monolayer showed cytopathic effects. Isolation of the
virus was confirmed and serotyped by indirect immunofluorescence using DENV
group-specific and DENV serotype-specific monoclonal antibodies.
Molecular analysis. Viruses were propagated by two passages in C6/36 mos-
quito cell culture. Virus titer was measured by using a plaque assay. Virus titers
of at least 106were found to be required for optimal success in subsequent
RT-PCR steps. Viral RNA was extracted from the culture supernatant by using
a Qiagen QiaAmp kit and extraction protocol.
The extracted RNA was copied to cDNA using an RT reaction, followed by
PCR amplification. The virus was amplified as five overlapping fragments. Unless
specified below the same conditions were used for all five fragments. RT of
fragments 1, 3, and 4 was performed in a single tube. RT of fragments 2 and 5
was performed in separate tubes. Samples were kept on ice and pipetting was
carried out using RNase free filter tips. PCR primers used in the RT reaction are
detailed in Table S1 in the supplemental material and reaction conditions are
presented in Table S2 in the supplemental material. The five RT fragments
were subsequently amplified by PCR. The PCR primers that were used to
amplify the RT fragments are detailed in Table S3 in the supplemental material.
PCR of the RT fragments was carried out in a thermal cycler using the program
shown in Table S4 in the supplemental material.
Gel electrophoresis was used for visualization of the PCR products as well as
gel extraction and purification of products. The products were separated in 1%
agarose Tris-borate-EDTA gels after visualization of ethidium bromide-stained
bands under UV light. DNA was extracted from bands excised from agarose gels
by using a Qiagen QiaQuick extraction kit.
Finally, gel-purified fragments were sequenced by using an Applied Biosys-
tems BigDye ddNTP capillary sequencer. Viral genome sequences generated in
the present study are deposited in GenBank with accession numbers EU081177
to EU081281. All genome sequences, their GenBank accession numbers, and
their standard strain names (27) used in the analyses in the present study are
detailed in Table S5 in the supplemental material.
Evolutionary analysis. To reveal the origins of the Singapore viruses we
conducted a phylogenetic analysis of the complete coding region of the genome
sequences of all those viruses sequenced here, as well as those already available
in GenBank. This resulted in data sets of the following sizes: DENV-1, 145 taxa,
10,176 nucleotides (nt); DENV-2, 116 taxa, 10,173 nt; and DENV-3, 122 taxa,
10,173 nt. To determine the best-fit model of nucleotide substitution we used
MODELTEST (24). In all cases, the most general GTR?I??4model, where
GTR is generalized time reversible, I is proportion of invariable sites, and ? is
the shape parameter of the gamma distribution, was favored. Maximum-likeli-
hood (ML) trees were then inferred under this model using PAUP* (30), with
tree bisection reconnection branch-swapping in each case. Finally, a neighbor-
joining bootstrap analysis (1,000 replications), but using the ML substitution
model, was used to determine the robustness of key nodes on each phylogeny.
To determine the population dynamics of DENV-1 and DENV-3 during the
2005 dengue outbreak in Singapore, we analyzed isolates that clearly diversified
during the course of the epidemic. For DENV-1, this meant our analysis was
confined to 53 genome sequences from Singapore (genotype I), while 42 ge-
nomes (genotype III) were used in the case of DENV-3. There were insufficient
sequences for an analysis of DENV-2. Demographic and evolutionary parame-
ters for both serotypes were estimated by using the Bayesian Markov Chain
Monte Carlo (MCMC) approach implemented in the BEAST package (4). Be-
cause of the typically complex population dynamics we utilized the Bayesian
skyline plot as a coalescent prior. This provides a piecewise graphical depiction
of changes in the effective number of infections (Ne?), where Neis the effective
population size and ? is the generation time. We also utilized both strict and
relaxed (uncorrelated lognormal) molecular clocks. The GTR??4model of
nucleotide substitution was used in all cases, with the invariant-sites parameter
(I) excluded since it tended to overfit to the data. Uncertainty in the data for each
estimated parameter is reflected in values of the 95% high probability density
(HPD), with all MCMC chains run for sufficient time (50 million steps, with 10%
removed as burn-in) to ensure statistical convergence.
To determine the nature of selection pressures acting on each gene of
DENV-1 and DENV-3 sampled from Singapore, we computed the mean ratio of
nonsynonymous (dN) to synonymous (dS) substitutions per site (dN/dS) using the
single likelihood ancestor counting method available through the DATAMONKEY
web interface (23) and assuming the GTR model of nucleotide substitution and
an input neighbor-joining tree. This analysis also allowed us to compute the tree
length (TL) in substitutions per site for each gene. In addition, we used the
CODEML program within the PAML package (33) to estimate an overall dN/dS
for the entire coding region of both serotypes (the “one-ratio” model). This was
compared to the case in which a separate dN/dSvalue was estimated for the
external and internal branches of each data set (the “two-ratio” model).
Finally, to determine the strength of spatial structure in both DENV-1 and
DENV-3, we first obtained the physical address of each viral isolate and pro-
duced clusters according to their geographical proximity by K-means clustering.
For DENV-1, the physical address was available for 48 isolates, which were then
placed into one of six different spatial groups (with a single-letter character state
code identifying each group). In the case of DENV-3, address information was
available for 42 isolates, and these were separated into four spatial groups.
Although there are a variety of methods by which the extent of spatial structure
in phylogenetic data can be determined, particularly utilizing parsimony charac-
ter state mapping (28), we used a Bayesian MCMC approach (22), thereby
accounting for any error in the underlying phylogeny. This analysis was based on
the trees output from the previous BEAST analysis (with a new BEAST analysis
conducted on the 48-sequence DENV-1 data set), using 1,000 replications and
with 10% of trees removed as burn-in. From these trees we computed the mean
values, credible intervals, and significance of the parsimony score (PS) and
association index (AI) statistics of the strength of geographical clustering (22).
RESULTS AND DISCUSSION
Fortuitously, the 2005 outbreak coincided with the launch of
the longitudinal EDEN (early dengue infection and outcome)
study in the central Ang Mo Kio district of Singapore (17). Of
133 RT-PCR dengue-positive patients enrolled during the
EDEN study collected between April and November 2005,
serotyping determined 66 (48.9%) to be DENV-1, 62 (46.6%)
to be DENV-3, and 5 to be DENV-2 (3.8%). No cases due to
DENV-4 were observed (17), and one patient was found to be
coinfected with serotypes 1 and 3. The detection of large num-
bers of DENV-3 in the Ang Mo Kio area was unusual as
DENV-1 was the predominant serotype in the rest of Singa-
pore. We were able to isolate 112 (84.2%) viruses, correspond-
ing to 57 DENV-1, 50 DENV-3, and 5 DENV-2 isolates. Com-
plete genome sequences were obtained for 54 DENV-1, 44
DENV-3, and 4 DENV-2 viruses. The remaining samples, al-
though shown to be dengue positive by RT-PCR, did not yield
sufficient viral RNA for genome sequencing.
To determine the origins of the viral isolates responsible for
the 2005 dengue epidemic in Singapore, we conducted a phy-
logenetic analysis of the complete genomes of the viruses sam-
pled here combined with representative DENV isolates taken
from GenBank. To assist in this comparison, we also com-
pleted whole-genome sequences of some historical DENV
samples from Singapore. All but one of the DENV-1 genomes
from this epidemic were classified as genotype I, which com-
monly circulates in Southeast Asia (Fig. 1). The single outlier
belongs to genotype III, which is predominantly found in Latin
America and West Africa (8), although a genotype III virus
was previously sampled in Singapore in 1993. The closest rel-
atives of the 53 genotype I Singapore DENV-1 isolates are the
Chinese isolates DENV-1/CN/Fj231/2004 and DENV-1/CN/
ZH1067/XXX, suggesting that frequent transfer of DENV
may occur between Singapore and China, and DENV-1/JP/20-
Feb/XXX from Japan. Importantly, since three historical Sin-
gapore DENV-1 samples isolated in 2003 fell at the base of the
2005 cluster, it is possible that this particular lineage of
DENV-1 genotype I has been circulating continuously in Sin-
gapore since at least 2003.
The four DENV-2 genomes form part of the “cosmopolitan”
genotype (Fig. 2), which has a wide distribution in tropical and
4164 SCHREIBER ET AL.J. VIROL.
FIG. 1. Phylogenetic relationships of 145 complete genomes DENV-1 sampled globally determined by using a ML method. Isolates sampled
from Singapore are shown in red, and individual genotypes are shown. Bootstrap values (?80%) are shown next to key nodes, and all horizontal
branch lengths are drawn to scale.
FIG. 2. Phylogenetic relationships of 116 complete genomes DENV-2 sampled globally determined by using a ML method. Isolates sampled
from Singapore are shown in red, and individual genotypes are shown. Bootstrap values (?80%) are shown next to key nodes, and all horizontal
branch lengths are drawn to scale.
4166SCHREIBER ET AL.J. VIROL.
FIG. 3. Phylogenetic relationships of 122 complete genomes DENV-3 sampled globally determined by using a ML method. Isolates sampled
from Singapore are shown in red, and individual genotypes are shown. Bootstrap values (?80%) are shown next to key nodes, and all horizontal
branch lengths are drawn to scale.
VOL. 83, 2009GENOMIC EPIDEMIOLOGY OF DENGUE VIRUS4167
4168SCHREIBER ET AL. J. VIROL.
subtropical localities (32). Close relatives of these strains in-
clude DENV-2/ID/BA05i/2004 and DENV-2/ID/TB16i/2004,
which were isolated during a dengue fever epidemic in Jakarta
in 2004 (29), as well as three strains from Brunei Darussalam,
one from China, and an older isolate from Queensland, Aus-
tralia (DENV-2/AU/TSV01/1993) (13), possibly introduced
from the nearby Indonesian islands. Since a number of these
viruses were isolated between 2004 and 2006, it seems likely
that this lineage was relatively common in this geographical
area at the time of the outbreak.
The majority of DENV-3 genomes fell into genotype III
(Fig. 3). This genotype was originally associated with the In-
dian subcontinent until the mid-1990s, when it was introduced
into Latin America and the Caribbean (19). Of more impor-
tance from the perspective of this outbreak was that an isolate
from genotype III was first detected in Singapore in 2003
(DENV-3/SG/S221/2003) and which fell basal to the 2005 out-
break viruses in our phylogenetic analysis. Such a phylogenetic
pattern is compatible with the in situ evolution of this lineage
in Singapore since at least 2003. Hence, as is also likely the case
with DENV-1, the 2005 outbreak of DENV-3 may also be due
to the amplification of a preexisting viral lineage rather than
the invasion of an “exotic” DENV strain. To further test this
hypothesis, we obtained the additional genome sequence of a
Singaporean DENV-3 genotype I isolate (DENV-3/SG/SS710/
2004) sampled in 2004. As expected under the hypothesis of in
situ evolution, this isolate occupies an intermediate position
between the 2003 and 2005 strains. Finally, two of the 44
DENV-3 isolates from the 2005 outbreak in Singapore fall into
genotype I, which is endemic in the Malay archipelago. This
observation provides an additional point of similarity between
the DENV-1 and DENV-3 components of the 2005 DENV
outbreak in Singapore: that individual epidemic serotypes can
be composed of multiple viral genotypes.
To determine the evolutionary processes that enabled the
emergence of multiple DENV genotypes in a single outbreak
we examined each of the amino acid changes associated with
these viruses. Remarkably, no amino acid changes were com-
pletely fixed on the branches leading to the DENV-1 or
DENV-3 isolates sampled in Singapore during 2005 (Fig. 4A).
In DENV-1 residue 76 of NS4A was observed to be K in all
samples from 2003 and R in many 2005 samples. However, the
FIG. 4. Observed amino acid changes in DENV-1 (A) and DENV-3 (B). Viral isolates are plotted on the x axis. The aligned polyproteins of
each virus were compared to count the number and distribution of amino acid changes. Positions that were not completely conserved are shown
with the individual protein name and the amino acid position within the protein on the y axis. The color of each square indicates the type of amino
acid residue found in isolate x at position y.
VOL. 83, 2009GENOMIC EPIDEMIOLOGY OF DENGUE VIRUS 4169
penetration of the mutation was incomplete and is quite con-
servative in nature and therefore unlikely to be significant.
Similarly, no amino acid variants in DENV-1 observed during
2005 were found in the two genomes isolated and sequenced
from the 2006 nonepidemic year, indicating that no mutations
that occurred in 2005 became fixed. Some substitutions were
observed in DENV-3 between the 2003 and 2004 nonepidemic
strains, although all are conservative, except for the change
from serine to lysine at residue 895 in NS5. Position 895 is not
conserved in the four serotypes and is usually S, P, or E. In the
recently solved structure of the DENV-3 RNA-dependent
RNA polymerase (34), position 895 is near the C terminus
(position 900) and does not appear to be functionally signifi-
cant. More notably, there are no fixations between the 2004
isolate and the 2005 epidemic isolates. Although a number of
nonconservative substitutions were observed within the 2005
isolates of DENV-3, that this serotype was not detected in 2006
indicates that none were capable of perpetuating the clade
(Fig. 4B). Finally, there was no evidence for positive selection
in any gene of the Singapore viruses, with a relatively low ratio
of nonsynonymous to synonymous substitutions per site (dN/
dS) in all genes and no evidence for site-specific positive se-
lection (Table 1; which also gives a variety of other gene-
specific measures of genetic diversity).
To infer the epidemiological dynamics of DENV-1 and
DENV-3 during the Singapore outbreak, we used a Bayesian
coalescent approach (3, 5) incorporating data on the exact day
of viral sampling. For both serotypes, we estimated the chang-
ing patterns of relative genetic diversity through time as re-
flected in the effective number of infections (Ne?) using a
Bayesian skyline plot and assuming a relaxed (uncorrelated
lognormal) molecular clock (although very similar results were
observed under a strict molecular clock; these results are avail-
able from the authors on request). Similar epidemic profiles
were observed in both viruses, comprising a rapid growth phase
followed by a constant population size, although the mean age
of the common ancestor was significantly greater in DENV-1
(1,740 days; 95% HPD ? 741 to 3,222 days) compared to
DENV-3 (298 days; 95% HPD ? 225 to 387 days), indicating
that already diverse lineages of DENV-1 were present in Sin-
gapore at the outset of the 2005 outbreak (Fig. 5). Similarly,
mean estimates of peak Ne? were smaller in DENV-3 (1,632
days; 95% HPD ? 56 to 4,334) than DENV-1 (7,211 days; 95%
HPD ? 637 to 23130), although with overlapping HPD values.
Interestingly, DENV-3 was not widely reported in Singapore
during the outbreak and appeared to be mainly contained to
the sampling area, which supports this result. The substitution
dynamics of both serotypes were also similar, with mean evo-
lutionary rates of 1 ? 10?3substitutions/site/year (95% HPD,
0.4 ? 10?3to 1.6 ? 10?3substitutions/site/year) and 1.3 ?
10?3substitutions/site/year (95% HPD, 8.7 ? 10?4to 1.8 ?
10?3substitutions/site/year) for DENV-1 and DENV-3, re-
spectively, and equivalent to those estimated previously for
As an additional analysis of evolutionary dynamics, we de-
termined the genetic distance (under the ML substitution
model) for each pair of DENV-1 and DENV-3 sequences and
compared these values to time intervals of sampling (based on
day of fever onset) (Fig. 6). Interestingly, it appears that ge-
netic distances are often higher between isolates separated by
shorter periods of time but then decline between pairs sampled
over a longer time period (although this analysis does not take
into account phylogenetic structure). This may in part be due
to the presence of transient deleterious mutations in samples
that are only separated by short time periods (such that genetic
distances strongly reflect the background mutation rate), which
are later purged by purifying selection, so that longer-term
genetic distances are more indicative of the population substi-
tution rate (12). This is supported by the observation that
dN/dSis higher on external (0.130 and 0.157) than internal
(0.081 and 0.030) branches of the complete coding region
phylogenies for both DENV-1 and DENV-3, respectively, as
expected if most nonsynonymous polymorphisms are deleteri-
Interestingly, our analysis of DENV spatial dynamics in Sin-
gapore revealed significant population substructure (i.e., the
existence of distinct spatial clusters) in both DENV-1 and
DENV-3. This spatial dynamic was especially strong in the case
of DENV-3 (P ? 0.001 for both the PS and AI statistics)
compared to DENV-1 (P ? 0.008 and 0.037 for the PS and AI
statistics, respectively). Hence, although these viruses were
sampled from a relatively restricted region within Singapore,
the movement of hosts and/or vectors is sufficiently limited that
spatial structure is present in the data.
TABLE 1. Phylogenetic and evolutionary patterns among the proteins of DENV-1 and DENV-3 sampled from Singapore during 2005a
Protein Length(s) (bp)b
1,479 and 1,485*
744 and 747*
2,679 and 2,700*
aIS, number of parsimony informative sites; TL, tree length expressed as the number of substitutions per site.
b*, Lengths for DENV-1 and DENV-3, respectively.
cNA, not applicable. That is, no significant evidence for positive selection was observed in any gene in either DENV-1 or DENV-3.
4170 SCHREIBER ET AL.J. VIROL.
Finally, the viremia from each patient’s sera was estimated
from a crossover threshold (Ct) calculated using quantitative
RT-PCR at 1 to 3 days and 4 to 7 days after fever onset. A low
Ct value of the RT-PCR, indicating high viremia levels, in the
first sampling has been previously shown to be predictive of
severe thrombocytopenia in our cohort (31). The Ct value for
DENV-1 was significantly lower (P ? 0.002) than for DENV-3
at both the first (17.07 versus 19.76) and the second (26.87
versus 29.57) serum samplings, indicating a higher viremia
level for DENV-1. DENV-1 also resulted in a significantly
(P ? 0.021) higher ratio of hospitalizations among the sampled
population compared to DENV-3 (0.74 versus 0.50) and may
also be reflected in the higher values of Ne? for DENV-1 than
Overall, these results suggest that ecological and/or immu-
nological factors, rather than aspects of viral evolution, were
central in shaping the dynamics of this dengue outbreak. Most
notably, two different serotypes, and multiple cocirculating ge-
notypes, emerged simultaneously, experienced similar epide-
miological dynamics, and seemingly spread without the aid of
positive selection, accumulating no amino acid fixations. Sim-
ilarly, it is clear that viral evolution did not succeed in extend-
ing the epidemic; despite numerous mutations (Table 1), the
number of dengue cases declined rapidly in 2006, suggesting
FIG. 5. Population dynamics of DENV-1 and DENV-3 in Singapore during 2005 depicted using Bayesian skyline plots. The plots show changes
in relative genetic diversity, depicted as the effective number of infections (Ne?), through time. The black line represents the mean estimate of Ne?,
while the 95% HPD intervals are shown in blue. Time is shown as the number of days from the most recent sample. To aid interpretation, DENV-1
and DENV-3 have been shown on the same time axis.
VOL. 83, 2009GENOMIC EPIDEMIOLOGY OF DENGUE VIRUS 4171
that few, if any, of these mutations provided any selective
advantage in the face of rising immunity. However, both
DENV-1 and DENV-3 appear to possess an inherent robust-
ness that allows them to persist at a low level of infection at
times when ecological and immunological conditions do not
favor an outbreak. This is supported by the observation that
the DENV-1 and DENV-3 lineages that characterized this
outbreak were found in Singapore as early as 2003 but did not
result in an outbreak until 2005. However, a wider sampling of
viral isolates from neighboring geographic areas is needed to
fully test this analysis. Furthermore, dengue epidemics in Sin-
gapore follow a regular 6- to 7-year periodic cycle, which is
difficult to explain by patterns of viral evolution or by vector
population density and might be more attributable to changing
levels of herd immunity (6). In addition, in Colombia it has
been observed that epidemic years are correlated with an in-
crease in the infection rate of mosquitoes and not the total
number of mosquitoes per household (18), indicating surveil-
lance of mosquito populations may also be important in un-
Together, our results have important implications for the
future study and control of DENV epidemics. In particular, the
epidemic surveillance of viral genome sequences in this case
would not have been sufficient to predict the 2005 outbreak.
Hence, incumbent strains with apparently inherent epidemic
potential are required but apparently not sufficient to spark an
outbreak. Concurrent surveillance of viral isolates, mosquito
vectors (including the proportion of mosquitoes infected with
DENV), and periodic surveys of seroprevalence rates of the
population may therefore provide the additional required pre-
dictive information. The chance discovery of the DENV-3 out-
break also highlights the value of comprehensive city-wide
fever surveys in detecting rare events.
We thank the staff involved in undertaking the EDEN study, to-
gether with Lai Yee Ling from the Environmental Health Institute, Fu
Xiuju from the Institute for High Performance Computing, and Leng
Marcel from the National University of Singapore. We thank Edison
Liu and Duane Gubler for useful comments on the manuscript.
M.J.S., W.L., and S.G.V. are employed by the Novartis Institute for
Tropical Diseases, which is funded by Novartis AG for the purpose of
developing anti-dengue therapies. S.H.O. is a Novartis-funded Ph.D.
candidate at the University of Basel. L.T. was enrolled in the Joint
M.Sc. Programme in Infectious Diseases organized in conjunction with
the National University of Singapore, the Novartis Institute of Tropical
Diseases, the Swiss Tropical Institute, and the University of Basel.
This study was in part funded by NIH grant number GM080533-01
to E.C.H., a Singapore BioMedical Research grant to A.O., the Sin-
gapore Agency for Science Technology and Research, the Singapore
Tissue Network, and the Novartis Institute for Tropical Diseases.
1. Bennett, S. N., E. C. Holmes, M. Chirivella, D. M. Rodriguez, M. Beltran, V.
Vorndam, D. J. Gubler, and W. O. McMillan. 2006. Molecular evolution of
dengue 2 virus in Puerto Rico: positive selection in the viral envelope ac-
companies clade reintroduction. J. Gen. Virol. 87:885–893.
2. de Oliveira, T., O. G. Pybus, A. Rambaut, M. Salemi, S. Cassol, M. Ciccozzi,
G. Rezza, G. C. Gattinara, R. D’Arrigo, M. Amicosante, L. Perrin, V. Colizzi,
and C. F. Perno. 2006. Molecular epidemiology: HIV-1 and HCV sequences
from Libyan outbreak. Nature 444:836–837.
3. Drummond, A. J., S. Y. Ho, M. J. Phillips, and A. Rambaut. 2006. Relaxed
phylogenetics and dating with confidence. PLoS Biol. 4:e88.
4. Drummond, A. J., and A. Rambaut. 2007. BEAST: Bayesian evolutionary
analysis by sampling trees. BMC Evol. Biol. 7:214.
5. Drummond, A. J., A. Rambaut, B. Shapiro, and O. G. Pybus. 2005. Bayesian
coalescent inference of past population dynamics from molecular sequences.
Mol. Biol. Evol. 22:1185–1192.
6. Egger, J. R., E. E. Ooi, D. W. Kelly, M. E. Woolhouse, C. R. Davies, and P. G.
Coleman. 2008. Reconstructing historical changes in the force of infection of
FIG. 6. Distribution of pairwise nucleotide, temporal, and geographic distances among complete coding sequences of DENV-1 and DENV-3
combined. Histograms show the frequency of temporal and nucleotide distances on the x and y axes, respectively. Each pair is represented as a
point, and the color of the point indicates the physical distance between the home address of the patients from whom the viruses were isolated.
4172SCHREIBER ET AL. J. VIROL.
dengue fever in Singapore: implications for surveillance and control. Bull.
W. H. O. 86:187–196.
7. Gilbert, M. T., A. Rambaut, G. Wlasiuk, T. J. Spira, A. E. Pitchenik, and M.
Worobey. 2007. The emergence of HIV/AIDS in the Americas and beyond.
Proc. Natl. Acad. Sci. USA 104:18566–18570.
8. Goncalvez, A. P., A. A. Escalante, F. H. Pujol, J. E. Ludert, D. Tovar, R. A.
Salas, and F. Liprandi. 2002. Diversity and evolution of the envelope gene
of dengue virus type 1. Virology 303:110–119.
9. Guan, Y., B. J. Zheng, Y. Q. He, X. L. Liu, Z. X. Zhuang, C. L. Cheung, S. W.
Luo, P. H. Li, L. J. Zhang, Y. J. Guan, K. M. Butt, K. L. Wong, K. W. Chan,
W. Lim, K. F. Shortridge, K. Y. Yuen, J. S. Peiris, and L. L. Poon. 2003.
Isolation and characterization of viruses related to the SARS coronavirus
from animals in southern China. Science 302:276–278.
10. Gubler, D. J. 2002. Epidemic dengue/dengue hemorrhagic fever as a public
health, social and economic problem in the 21st century. Trends Microbiol.
11. Halstead, S. B. 2007. Dengue. Lancet 370:1644–1652.
12. Jarman, R. G., E. C. Holmes, P. Rodpradit, C. Klungthong, R. V. Gibbons,
A. Nisalak, A. L. Rothman, D. H. Libraty, F. A. Ennis, M. P. Mammen, Jr.,
and T. P. Endy. 2008. Microevolution of Dengue viruses circulating among
primary school children in Kamphaeng Phet, Thailand. J. Virol. 82:5494–
13. Johansson, M., A. J. Brooks, D. A. Jans, and S. G. Vasudevan. 2001. A small
region of the dengue virus-encoded RNA-dependent RNA polymerase, NS5,
confers interaction with both the nuclear transport receptor importin-beta
and the viral helicase, NS3. J. Gen. Virol. 82:735–745.
14. King, C. C., D. Y. Chao, L. J. Chien, G. J. Chang, T. H. Lin, Y. C. Wu, and
J. H. Huang. 2008. Comparative analysis of full genomic sequences among
different genotypes of dengue virus type 3. Virol. J. 5:63.
15. Klungthong, C., C. Zhang, M. P. Mammen, Jr., S. Ubol, and E. C. Holmes.
2004. The molecular epidemiology of dengue virus serotype 4 in Bangkok,
Thailand. Virology 329:168–179.
16. Lanciotti, R. S., J. T. Roehrig, V. Deubel, J. Smith, M. Parker, K. Steele, B.
Crise, K. E. Volpe, M. B. Crabtree, J. H. Scherret, R. A. Hall, J. S. Mac-
Kenzie, C. B. Cropp, B. Panigrahy, E. Ostlund, B. Schmitt, M. Malkinson,
C. Banet, J. Weissman, N. Komar, H. M. Savage, W. Stone, T. McNamara,
and D. J. Gubler. 1999. Origin of the West Nile virus responsible for an
outbreak of encephalitis in the northeastern United States. Science 286:
17. Low, J. G., E. E. Ooi, T. Tolfvenstam, Y. S. Leo, M. L. Hibberd, L. C. Ng,
Y. L. Lai, G. S. Yap, C. S. Li, S. G. Vasudevan, and A. Ong. 2006. Early
dengue infection and outcome study (EDEN): study design and preliminary
findings. Ann. Acad. Med. Singapore 35:783–789.
18. Mendez, F., M. Barreto, J. F. Arias, G. Rengifo, J. Munoz, M. E. Burbano,
and B. Parra. 2006. Human and mosquito infections by dengue viruses
during and after epidemics in a dengue-endemic region of Colombia. Am. J.
Trop. Med. Hyg. 74:678–683.
19. Messer, W. B., D. J. Gubler, E. Harris, K. Sivananthan, and A. M. de Silva.
2003. Emergence and global spread of a dengue serotype 3, subtype III virus.
Emerg. Infect. Dis. 9:800–809.
20. Ong, S. H., J. T. Yip, Y. L. Chen, W. Liu, S. Harun, E. Lystiyaningsih, B.
Heriyanto, C. G. Beckett, W. P. Mitchell, M. L. Hibberd, A. Suwandono,
S. G. Vasudevan, and M. J. Schreiber. 2008. Periodic re-emergence of
endemic strains with strong epidemic potential-a proposed explanation for
the 2004 Indonesian dengue epidemic. Infect. Genet. Evol. 8:191–204.
21. Ooi, E. E., K. T. Goh, and D. J. Gubler. 2006. Dengue prevention and 35
years of vector control in Singapore. Emerg. Infect. Dis. 12:887–893.
22. Parker, J., A. Rambaut, and O. G. Pybus. 2008. Correlating viral phenotypes
with phylogeny: accounting for phylogenetic uncertainty. Infect. Genet. Evol.
23. Pond, S. L., and S. D. Frost. 2005. Datamonkey: rapid detection of selective
pressure on individual sites of codon alignments. Bioinformatics 21:2531–
24. Posada, D., and K. A. Crandall. 1998. MODELTEST: testing the model of
DNA substitution. Bioinformatics 14:817–818.
25. Pybus, O. G., A. Rambaut, R. Belshaw, R. P. Freckleton, A. J. Drummond,
and E. C. Holmes. 2007. Phylogenetic evidence for deleterious mutation load
in RNA viruses and its contribution to viral evolution. Mol. Biol. Evol.
26. Rico-Hesse, R., L. M. Harrison, A. Nisalak, D. W. Vaughn, S. Kalayanarooj,
S. Green, A. L. Rothman, and F. A. Ennis. 1998. Molecular evolution of
dengue type 2 virus in Thailand. Am. J. Trop. Med. Hyg. 58:96–101.
27. Schreiber, M. J., S. H. Ong, R. C. Holland, M. L. Hibberd, S. G. Vasudevan,
W. P. Mitchell, and E. C. Holmes. 2007. DengueInfo: a web portal to dengue
information resources. Infect. Genet. Evol. 7:540–541.
28. Slatkin, M., and W. P. Maddison. 1989. A cladistic measure of gene flow
inferred from the phylogenies of alleles. Genetics 123:603–613.
29. Suwandono, A., H. Kosasih, Nurhayati, R. Kusriastuti, S. Harun, C.
Ma’roef, S. Wuryadi, B. Herianto, D. Yuwono, K. R. Porter, C. G. Beckett,
and P. J. Blair. 2006. Four dengue virus serotypes found circulating during
an outbreak of dengue fever and dengue haemorrhagic fever in Jakarta,
Indonesia, during 2004. Trans. R. Soc. Trop. Med. Hyg. 100:855–862.
30. Swofford, D. L. 2003. PAUP*: phylogenetic analysis using parsimony (* and
other methods), version 4. Sinauer Associates, Sunderland, MA.
31. Tanner, L., M. Schreiber, J. G. Low, A. Ong, T. Tolfvenstam, Y. L. Lai, L. C.
Ng, Y. S. Leo, L. Thi Puong, S. G. Vasudevan, C. P. Simmons, M. L. Hibberd,
and E. E. Ooi. 2008. Decision tree algorithms predict the diagnosis and
outcome of dengue fever in the early phase of illness. PLoS Negl. Trop. Dis.
32. Twiddy, S. S., E. C. Holmes, and A. Rambaut. 2003. Inferring the rate and
time-scale of dengue virus evolution. Mol. Biol. Evol. 20:122–129.
33. Yang, Z. 2007. PAML 4: phylogenetic analysis by maximum likelihood. Mol.
Biol. Evol. 24:1586–1591.
34. Yap, T. L., T. Xu, Y. L. Chen, H. Malet, M. P. Egloff, B. Canard, S. G.
Vasudevan, and J. Lescar. 2007. Crystal structure of the dengue virus RNA-
dependent RNA polymerase catalytic domain at 1.85-angstrom resolution.
J. Virol. 81:4753–4765.
VOL. 83, 2009GENOMIC EPIDEMIOLOGY OF DENGUE VIRUS 4173