Dengue 1 Diversity and Microevolution, French Polynesia 2001–2006: Connection with Epidemiology and Clinics

Background Dengue fever (DF) is an emerging infectious disease in the tropics and subtropics. Determinants of DF epidemiology and factors involved in severe cases—dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS)—remain imperfectly characterized. Since 2000, serotype 1 (DENV-1) has predominated in the South Pacific. The aim of this study was (i) to determine the origin and (ii) to study the evolutionary relationships of DENV-1 viruses that have circulated in French Polynesia (FP) from the severe 2001 outbreak to the recent 2006 epidemic, and (iii) to analyse the viral intra-host genetic diversity according to clinical presentation. Methodology/Principal Findings Sequences of 181 envelope gene and 12 complete polyproteins of DENV-1 viruses obtained from human sera in FP during the 2001–2006 period were generated. Phylogenetic analysis showed that all DENV-1 FP strains belonged to genotype IV–“South Pacific” and derived from a single introduction event from South-East Asia followed by a 6-year in situ evolution. Although the ratio of nonsynonymous/synonymous substitutions per site indicated strong negative selection, a mutation in the envelope glycoprotein (S222T) appeared in 2002 and was subsequently fixed. It was noted that genetic diversification was very significant during the 2002–2005 period of endemic DENV-1 circulation. For nine DF sera and eight DHF/DSS sera, approximately 40 clones/serum of partial envelope gene were sequenced. Importantly, analysis revealed that the intra-host genetic diversity was significantly lower in severe cases than in classical DF. Conclusions/Significance First, this study showed that DENV-1 epidemiology in FP was different from that described in other South-Pacific islands, characterized by a long sustained viral circulation and the absence of new viral introduction over a 6-year period. Second, a significant part of DENV-1 evolution was observed during the endemic period characterized by the rapid fixation of S222T in the envelope protein that may reflect genetic drift or adaptation to the mosquito vector. Third, for the first time, it is suggested that clinical outcome may be correlated with intra-host genetic diversity.


Introduction
Dengue fever is the most common vector-borne viral disease affecting humans and represents an archetypal emerging infectious disease whose epidemiological landscape has been substantially modified during the past century [1,2]. Each year, an estimated 100 million people contract dengue fever (DF) in the tropics and subtropics [3] with an increasing incidence of the severe forms, i.e. at least 500,000 cases annually of dengue haemorrhagic fever (DHF) or dengue shock syndrome (DSS).
In this study, we performed an analysis of the E-gene sequence of 181 DENV-1 viruses and the nearly complete coding sequence of 12 DENV-1 viruses collected over a 6-year period from patients experiencing various clinical presentations in the five FP archipelagos. In addition, we performed a comprehensive comparative analysis of intra-host viral genetic diversity in 16 patients. This study enabled us to predict the precise geographic origin and evolutionary relationships, during both endemic and epidemic periods, of the DENV-1 isolates that circulated in FP from the severe 2001 outbreak to the recent 2006 epidemic. Original patterns of intra-host genetic diversity were also identified in association with the clinical severity of infection.

Specimen data
We analyzed serum samples from 181 DENV-1 infected patients from FP. Sampling was conducted in the five FP archipelagos ( Figure 1): Society (Windward and Leeward islands), Tuamotu, Gambier, Austral and Marquesas, from January 2001 to December 2006 ( Table 1). The study period included the 2001 and 2006 DENV-1 outbreaks, separated by four years of low-level transmission (2002)(2003)(2004)(2005). From a total of 181 cases, 152 patients experienced DF, 19 DHF and ten DSS with one death. Dengue disease severity was graded according to the World Health Organization (WHO) classification guidelines [20]. The time of serum collection relative to infection ranged from one to six days in documented cases. All human sera analyzed in this study had been preserved at 280uC at the Institut Louis Malardé (Tahiti, FP).

Ethics statement
All samples were obtained from sera initially sampled for diagnostic purpose, and archived at the Institut Louis Malardé (Tahiti, FP). The use of biological samples and the collection of information were performed with the authorization of the ''Direction des affaires juridiques et des droits des patients, Centre Hospitalier Territorial de Polynésie Française (Tahiti)'' and in accordance with French regulations.

Molecular characterization
Virus RNA was extracted from acute-phase sera of DENV-1 infected patients using the QIAamp Viral RNA Mini Kit (Qiagen) according to manufacturer's instructions.
DENV-1 sequences were retrieved from public databases and used to design oligonucleotide primers for reverse transcriptionpolymerase chain reaction (RT-PCR) amplification and sequencing of FP viruses.
Genetic characterization of the E-gene was conducted using the Qiagen OneStep RT-PCR kit together with primers E1F-E4R, followed by a nested PCR using primers E2F-E3R (Table S1) to produce a 1,759 nt fragment including the complete E-gene (1,485 nt) which was subsequently sequenced directly using amplification primers.
For characterization of full-length coding sequences, 12 overlapping cDNA fragments were generated by RT-PCR using 12 sets of oligonucleotide primers (Table S1). Fragments C1, C3-C8, C10 and C11 were obtained using the same one-step RT-PCR protocol as described above. Fragments C2, C9 and C12 were synthesized using a two-step protocol: cDNA was generated using a mixture of random hexaprimers (RT Taqman Applied Biosystems) followed by PCR amplification using Taq Polymerase (Invitrogen). Sequencing using amplification primers resulted in the characterization of a 10,075 nt sequence.
For the analysis of intra-host genetic diversity, the Qiagen OneStep RT-PCR kit was used together with primers Q1F-Q1R (Table S1) to produce a 758 nt fragment within the E-gene, which was subsequently purified using the QIAquick PCR Purification Kit, ligated into the cloning vector pCR 2.1 and transformed into TOP10 competent cells, according to the manufacturer's protocol (TA Cloning, Invitrogen). Approximately 40 clones per serum were generated and sequenced using the T7 promoter primer (59-CCCTATAGTGAGTCGTATTA-39). To estimate the error rates of our amplification system, we carried out a control experiment using a fully sequenced clone of the 758 nt fragment. Serial dilutions were produced and the last dilution providing a clear positive signal was used as a control. It was submitted to onestep RT-PCR amplification and clones (n = 90) were characterized under identical conditions as viral RNA extracted directly from acute phase DENV-1 patient sera. In order to evaluate the influence of viral load in DENV-1 genetic diversity within patients, viral RNA was quantified by real-time RT-PCR, as described previously [21] in 11 of the 17 analyzed sera corresponding to five DHF/DSS cases and five DF cases (including sequential blood samples for one patient: 47.2002 and 49.2002).

Phylogeny and sequence analysis
Sequence data from sequencing reactions were combined for analysis and edited using the Sequencher 4.7 software (Gene

Author Summary
The molecular characterization of 181 serotype 1 Dengue fever (DENV-1) viruses collected regularly during the 2001-2006 period in French Polynesia (FP) from patients experiencing various clinical presentations revealed that the virus responsible for the severe 2001 outbreak was introduced from South-East Asia, and evolved under an endemic mode until a new epidemic five years later. The dynamics of DENV-1 epidemics in FP did not follow the model of repeated virus introductions described in other South Pacific islands. They were characterized by a long sustained viral circulation and the absence of new viral introduction over a six-year period. Viral genetic variability was not observed only during outbreaks. In contrast with conventional thinking, a significant part of DENV-1 evolution may occur during endemic periods, and may reflect adaptation to the mosquito vector. However, DENV-1 evolution was globally characterized by strong purifying selection pressures leading to genome conservation, like other DENV serotypes and other arboviruses subject to constraints imposed by the host-vector alternating replication of viruses. Severe cases-dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS)-may be linked to both viral and host factors. For the first time, we report a significant correlation between intra-host viral genetic variability and clinical outcome. Severe cases were characterized by more homogeneous viral populations with lower intra-host genetic variability.   Codes Corporation). Nucleotide sequences used for phylogenetic analyses were aligned using Clustal W [22], and then imported into the MEGA 3.1 package [23]. Nucleotide genetic distances were calculated using the Kimura 2 algorithm [24] and Neighbor-Joining was used for phylogenetic reconstructions. Robustness of phylogenetic trees was assessed using bootstrap resampling analysis (1000 replications). Supplementary maximum likelihood phylogenetic analyses were performed using the Bayesian method available in MrBayes v3.1.2 [25] with a minimum of ten million generations and a burnin of 10%. Stationary was assessed at effective sample size (ESS).400 using Tracer v1.4.1 (part of the BEAST package [26]).'' Phylogenetic analysis of E-gene sequences was conducted using a sample of 240 DENV-1 sequences. This included 181 FP sequences generated in this study together with three sequences of viruses that were previously characterized during the 1988-89 and the 2001 DENV-1 outbreaks in FP [27]: D1.French Polynesia/89, GenBank accession number AY630408; D1.French Polynesia/01, GenBank accession numbers AY630407 and AB111070. These FP sequences were combined with a sample of 56 viruses representing the global genetic variability of DENV-1 available from GenBank. In addition, we conducted a phylogenetic analysis based on the complete coding regions of 41 DENV-1 strains isolated worldwide (available from GenBank) and the corresponding sequences of 12 FP strains characterized in this study.
Differences in nucleotide and protein sequences were analyzed and compared according to the geographical origin, the sampling period and the clinical presentation. The extent of sequence divergence was evaluated using the pairwise distance among the nucleotide sequences (p nt) and the amino acid sequences (p aa). The mean ratio of nonsynonymous (d N ) to synonymous (d S ) substitutions per site was estimated using the pairwise method of Nei and Gojobori [28] as implemented in the MEGA 3.1 package.
For the analysis of intra-host genetic diversity, the sequence of each clone was compared to all other clones for each human serum. The percentage of variable nucleotide sites (number of variable nt sites/number of nt sites), of nucleotide mutations (number of nt mutation/number of nt sequenced), and of mutant clones (number of clones with mutation/total number of clones) was calculated, as well as the p nt, p aa, d N , d S and d N /d S parameters. Results were then compared according to the clinical presentation of dengue infection.
To explore the selection pressures acting on DENV-1 at different levels of viral evolution, distinct datasets were analyzed as follows: (i) ''FP intra-host'' dataset: this group included 17 series of cloned sequences obtained from 16 patients infected with DENV-1 (eight DF, eight DHF/DSS) in FP between 2001 and 2006. For one patient (DF -Moorea, Windward islands, Society archipelago -December 2002) two series of clones were produced from sequential blood samples obtained at day one and day four of the disease ; (ii) ''FP inter-host'' dataset: this group included the 181 sequences generated in FP between 2001 and 2006 (this study) ; (iii) ''genotype IV inter-host'' dataset: this included 26 sequences representing the genetic diversity of the ''South Pacific'' genotype; (iv) ''serotype 1 inter-host'' dataset: this included 59 sequences that reflect the worldwide diversity of DENV-1 isolates. For each dataset, the same parameters (percentage of variable nucleotide sites, p nt, d S , d N , d N /d S ) were analyzed.

Statistical analysis
All statistical analyses were performed using the R software package (R development Core Team version 2.6.0). Categorical and binary variables were compared using a Fisher's exact test. A Mann-Withney test was used for continuous variables (p values below 0.05 were considered to indicate statistical significance).
To evaluate differences between endemic and epidemic periods, For the analysis of intra-host genetic variability according to the clinical severity of dengue infection, we compared the percentage of variable nucleotide sites, the percentage of nucleotide mutation, the percentage of mutant clones, the average pairwise distance (p nt) and the mean d N , d S , d N /d S ratio obtained for each group of clones in nine DF sera (including two sequential blood samples for one patient) versus eight DHF or DSS sera.
The phylogenetic reconstruction based on E-gene nucleotide sequences showed that all DENV-1 strains that circulated in FP between 2001 and 2006 fall into genotype IV -''South Pacific'' (Figures 2 and S1). This genotype also includes DENV-1 viruses originating from other locations in the Pacific (Australia, Malaysia, Philippines, Palau, Yap, Nauru, Samoa, Hawaii), from South-East Asia (Thailand, Myanmar, China, Indonesia, Timor), and from the Indian Ocean (Seychelles, Reunion) between 1974 and 2006. According to a previous study [27], the D1.French Polynesia/89 strain isolated during the 1988-89 DENV-1 epidemic (preceding the 2001 outbreak) belonged to a different genotype (genotype V ''Americas/Africa'') and was very close to D1.French Guyana/89. DENV-1 strains recovered in FP during the severe 2001 epidemic shared a common ancestor with D1.Indonesia/98, a strain isolated in 1998 in a patient from Indonesia (Figures 2 and S1). However, they were more distantly related to D1.Palau/00, a strain isolated in Palau (Micronesia), the first island affected by DENV-1 in the Pacific Ocean in the 2000's [30]. The D1.Palau/ 00 strain was found to be more closely related to strains isolated during DENV-1 epidemics in the Philippines or Samoa islands in 2001 and 2002. Altogether, these results strongly suggest that DENV-1 that circulated in FP in 2001 originated from Indonesia rather than from Palau. This finding is further supported by phylogenetic analysis of 53 complete polyprotein sequences (Figures 3 and S2). In addition, phylogenetic analysis showed that most DENV-1 strains recovered during the 2001 outbreak in Hawaii clustered in the same lineage as FP 2001 strains (Figures 3 and S2), suggesting a Polynesian origin of the DENV-1 epidemic that occurred in Hawaii in 2001 [31].
A more detailed phylogenetic analysis of FP 2001-2006 sequences suggested that all FP viruses characterized in this study derived from a common ancestor, i.e. originated from a single introduction event in FP followed by a 6-year in situ evolution  In this condensed tree, branch length is not proportional to genetic distance. Numbers on branches represent bootstrap support for each branch. Five DENV-1 genotypes were identified. The validity of these genotypes, in particular genotype II ''Thailand'' and genotype III ''sylvatic/Malaysia'', is supported by previous phylogenetic analyses based on maximum likelihood method [5,29].'' doi: 10 Figure S1).

Analysis of sequence divergence
The polyprotein sequences of 12 FP 2001-2006 DENV-1 viruses were studied. Nonsynonymous mutations were observed in all genes, except NS4A. Amongst them, four mutations located in the E, NS4B, and NS5 genes have been fixed during viral evolution ( Table 2).
Additional analyses were conducted on a 1,759 nt region that encompassed the complete E-gene for 181 FP 2001-2006 sequences ( Table 3) nonsynonymous mutations were observed both during epidemic and endemic periods. The number of variable sites was found to be significantly higher during endemic period than during epidemics. The nucleotide sequence divergence (p nt) was also higher during endemic than during epidemic periods (p,0.001). When focusing on the first appearance of amino acid changes,

Intra-host genetic diversity of DENV-1 and the estimation of selection pressures
To examine the extent of genetic diversity of DENV-1 in vivo at the intra-host level, we sequenced 662 clones corresponding to partial E-genes of DENV-1 populations from 16 human sera at a single time point during acute infection. For one DF patient, clones corresponding to sequential samples (day one and day four of the symptoms) were sequenced and compared. Approximately 40 clones from each sample were analyzed, and the results are summarized in Table 4.
We carried out a control experiment to evaluate the sequence variation due to in vitro polymerase errors (see Methods). Among 90 clones of the 758 nt fragment studied within the E-gene, 55 nucleotide substitutions were found, corresponding to an error frequency of 0.10% or 2461026 changes/nt/PCR cycle. This result was significantly lower than the mean levels of intra-host diversity (percentages of nucleotide mutations) observed in our samples (0.25% or 6061026 changes/nt/PCR cycle, p,0.001).
In the 17 human sera studied, a high proportion of mutant clones was observed (mean 69%) with no significant difference in terms of clinical presentation, endemic or epidemic period, and time of sampling: analysis of sequential blood samples indicated that DENV-1 viraemia comprised a genetically heterogeneous mixture of variants that were present at the time of first appearance of the symptoms.
Mutations occurred in 15 (2%) to 81 (11%) sites of the 758 nucleotides sequenced. The proportion of nonsynonymous mutations was very high in each group of clones (63% on average). Most mutations were observed only once. However, identical mutations were sometimes observed in several clones from the same serum and/or in different sera. For instance, E269K was observed in 16 clones (strains 41.2002, 47.2002, 49.2002, 10.2003, and 10.2004) and E309K in 45 clones (strains 37.2001, 41.2002, 42.2002, 47.2002, 49.2002, 10.2003, 10.2004, and 32.2006). They were present simultaneously in 14 clones (strains 41.2002, 47.2002, 49.2002, 10.2003 and 10.2004). Of note, the mutation S222T was recovered in all clones of the 42.2002 strain, the first DENV-1 strain that expressed the mutation in August 2002 in Tahiti, and it was absent in all clones tested from sera of patients who were infected previously.
Overall, clones with in-frame stop codons were identified in 9 of the 17 sera studied, with a frequency ranging from 0% to 12%  Table 5). The percentage of nucleotide mutations (number of nt changes/number of nt sequenced) was significantly lower in severe (DHF and DSS) clinical presentations (mean 0.17%, range 0.10%-0.26%) than in classical forms (DF) of dengue infection (mean 0.32%, range 0.15%-0.57%, p = 0.015). Moreover, the mean sequence divergence was found to be lower in severe cases than in DF cases (p = 0.014 for p nt, p = 0.025 for p aa). Despite a similar proportion of mutant clones in DF and severe cases (69%), d N and d S were significantly lower in the latter cases (p = 0.014 and p = 0.011, respectively). When error frequencies calculated in our control experiment were subtracted from the results obtained for DF, DHF and DSS clones, differences between severe and classical cases remained significant (data not shown). Altogether, these findings indicate that the level of intra-host genetic diversity is lower in severe presentations than in classical forms of DENV-1 infection. In order to evaluate the influence of Table 3. Genetic diversity of DENV-1 at different levels and at different times of viral evolutionary divergence based on a 1,759 nt fragment including the E-gene.    . No correlation was found between the level of intrahost genetic diversity and viral load (range 0.12*10 5 -4.5*10 5 , mean 1.10*10 5 RNA copies/mL): linear regression analysis showed that the percentage of nt mutations and p nt were not correlated with viral load (p = 0.51 and 0.61, respectively). Moreover, in these serum samples, viral loads were not significantly different in severe cases than in DF cases (p = 0.31, Mann-Withney test). Finally, the mode of evolution of DENV-1 in FP was investigated by analysing the mean ratio of nonsynonymous to synonymous substitutions per site (d N /d S ) in our different dataset: d N /d S was 0.100 for complete genome sequences, and 0.091 for E gene sequences, indicating (d N /d S ,1) a strong negative (purifying) selection pressure [32]. This was confirmed by the study of the genetic variability at different levels of evolutionary divergence, i.e. in the four datasets: ''FP intra-host'', ''FP inter-host'', ''genotype IV inter-host'', and ''serotype 1 inter-host'' (Table 5). Within the group of FP viruses, the genetic variability of DENV-1 was higher within hosts than between hosts, as indicated by p nt, and d N /d S values which were higher in the intra-host dataset than in the inter-host dataset. At the inter-host level, the genetic divergence increased with the scale of the population studied (p nt ''FP'',''genotype IV'',''serotype 1'') whereas the proportion of nonsynonymous mutations decreased (d N /d S ''FP''.''genotype IV''.''serotype 1''), reflecting strong purifying selection pressures.

Discussion
In this study, DENV-1 evolution was analyzed during two recent outbreaks in FP separated by a four-year period of low-level transmission. Original dynamics of epidemics were revealed in the FP ecosystem. Our results suggest that a significant part of DENV-1 evolution occurred during the 2002-2005 endemic years. Despite evidence for strong negative selection, we report mutations that could reflect viral adaptation, particularly S222T that has been fixed by viral evolution in the envelope glycoprotein. Importantly, we report for the first time a significant correlation between levels of intra-host DENV genetic variability and clinical outcome.
Historically, FP has experienced successive dengue epidemics that involved the four DENV serotypes [12][13][14][15][16]19]. In contrast with most endemic countries and other islands such as those in the Caribbean, where different DENV serotypes circulate, prolonged co-circulation of several serotypes has never been detected in FP. Most Polynesian DENV epidemics were due to the introduction of a new serotype originating either from the Americas, South-East Asia or the Pacific region. Since 2000, serotype 1 has predominated in the South Pacific region and a significant increase in the number of DENV-1 cases has been observed since spring 2006 in several Pacific islands, particularly in FP and in the neighbouring Cook islands [18,19,33,34]. Classically, dengue fever is not believed to be endemic in the Pacific region and outbreaks are usually linked with the importation of a new virus: it has been shown that multiple and repeated introductions of DENV-1 occurred in the Pacific between 2000 and 2003 from a variety of locations in Asia [35].
Accordingly, our first objective was to identify the origin of the DENV-1 strain responsible for the 2001 outbreak in FP. In accordance with a preliminary study [27], phylogenetic analysis based on a large number of either complete polyprotein or E-gene sequences indicates that the most probable source of this epidemic was an Asian strain, as suggested by the close genetic relationship with a strain isolated in Indonesia in 1998. This finding is in contradiction with the hypothesis that the first DENV-1 outbreak observed in the Pacific Ocean in 2000 in Palau (Micronesia) dispersed secondarily to Polynesia and Melanesia [30,33,34] and emphasizes the relation between DENV-1 viruses in Asia and those responsible for recent outbreaks in the Pacific [35]. Figure 2 shows that the strain implicated in the Palau outbreak is only distantly related to FP strains and cannot be implicated as the origin of DENV-1 circulation in FP.
Our second objective was to determine whether or not the 2001 and 2006 FP outbreaks followed the model of iterative reintroductions evoked above. Our results indicate that the Polynesian dynamic of DENV-1 is different from that previously described in other Pacific islands such as New Caledonia [35].  Table 2).
The observation that viral evolution also occurred during periods of endemic transmission was expected, but the extent of the phenomenon deserved further investigation. Accordingly, a detailed analysis of complete E-gene sequences was performed, which allowed to include a much higher number of sequences (93 Statistical analysis showed that the number of variable sites (nt and aa) and the percentage of sequence divergence (p nt) were not higher during outbreaks than during endemic periods (Table 3) -and even suggested the opposite. It may appear to be in conflict with conventional thinking since the total virus replicative turnover would be expected to be higher during epidemics, and thus it would be expected that viral genetic variability occurred mainly during the 2001 and 2006 outbreaks. However, the distribution of viral genetic variability between endemic and epidemic periods was considered carefully, since the delineation between endemic and epidemic periods may appear simplistic. For example, although the 2001 outbreak was considered to end in November, the number of confirmed DENV-1 cases reported monthly by the Altogether, it stands out from our analyses that a significant part of Dengue virus evolution occurred during periods of endemic transmission and not only during outbreaks. Moreover, the majority of amino acid changes were observed during the early stages of the endemic period (Figure 4), suggesting adaptation to new specific environmental conditions. This is notably the case for S222T, the most frequent substitution identified in 88 strains, which appeared in August 2002 and was subsequently fixed by viral evolution. This mutation concerns the envelope protein, a major component at the virion surface implicated in the interaction with host cells, membrane fusion and induction of a protective immune response. Residue 222 is localized in domain II which is implicated in the dimerization of the envelope protein at acidic pH preceding membrane fusion and viral entry into the host cell [36]. This mutation is not described in the literature and it is not present in DENV-1 sequences available on GenBank.  2005 and 2006). This mutation in the envelope glycoprotein of FP DENV-1 viruses may be the result of genetic drift but it may be explained by positive selection also. S222T appears to have been fixed rapidly (10 months) which is not suggestive of a simple genetic drift. Its appearance during a period of endemic transmission (August 2002) and its rapid stabilization through time suggest that S222T would confer a selective advantage to the virus and may possibly be associated with adaptation to the mosquito vector.
Another event suggesting possible virus adaptation to the vector is the mutation K363R. This mutation was present in seven strains recovered in FP from March to December 2006. Residue 363 is localized within the ''immunoglobulin-like'' domain III of the envelope protein which contains regions thought to be important for receptor binding [36]. This residue belongs to a B-cell epitope (293-402) identified in DENV-1 [37]. As DENV infection confers a prolonged type-specific protective immunity, the hypothesis of an immune selection of this variant in humans is unlikely [4,5]. Rather, K363R may be the consequence of adaptation to the mosquito vector. Importantly, this mutation occurred only in patients originating from Moorea, Raiatea, Tahaa or the Austral archipelago and was not observed in Tahiti where the majority of cases occurred. Since Aedes (Stegomyia) polynesiensis, an endemic mosquito specie widespread in most of islands from the Polynesian Triangle connecting Hawaii and Easter Island to New Zealand, is thought to be an important vector of Dengue virus in rural areas [38,39], whereas Aedes (Stegomyia) aegypti is a major vector in urban and sub-urban zones, the K363R mutation may possibly reflect viral adaptation to Aedes polynesiensis in FP islands less urbanized than Tahiti.
Although we provide tentative evidence for the existence of a few adaptive mutations during the 2001-2006 period, DENV-1 evolution over this period is globally characterized by strong negative selection, in accordance with previous studies on DENV-2 and DENV-3 evolution [40,41]. The low d N /d S values (0.100 for polyprotein sequences and 0.091 for E-gene sequences) denote purifying selection and may reflect constraints imposed on Dengue virus evolution by the alternating replication of viruses in humans and mosquitoes. Further striking evidence for negative selection is provided by the analysis of genetic variability of DENV-1 at different levels of evolutionary divergence (Table 5): viral diffusion is associated with increasing purifying constraints as illustrated by the decrease in the d N /d S ratio measured in intra-host viral populations (d N /d S ''FP intra-host'' = 0.620), in a population of epidemiologically related viruses (d N /d S ''FP inter-host'' = 0.333), in viruses belonging to the same genotype (d N /d S ''genotype IV'' = 0.058) or to the same serotype (d N /d S ''serotype 1'' = 0.045). These results indicate that only a small proportion of nonsynonymous mutations observed at a given level of evolution are likely to persist at a higher time-and space-scale.
Dengue virus, like other RNA viruses, exhibits extensive intrahost genetic diversity [40][41][42][43][44][45]. We analyzed 662 clones from 16 patients infected with DENV-1 in the study period and observed that the structure of intra-host genetic diversity represents an extreme situation in which purifying selective constraints are lower than at higher levels of evolutionary divergence. As noted in a previous study on DENV-2 and DENV-3 [40,41,43,44], most nonsynonymous mutations occurred in single cases (not identified in more distantly related DENV-1) and genome-defective viruses (with stop codons) were identified (3% of clones) in human sera. Similar results were previously reported in a study of 70 clones obtained from four mosquitoes and 220 clones obtained from 13 patients infected with DENV-1 in Myanmar [42]. Defective viruses may interfere with viral evolution but long term transmission of a stop-codon lineage has been described within humans and mosquitoes infected with DENV-1 [42]: complementation mechanisms may occur in host cells coinfected with both functional viruses and defective viruses.
The large number of samples studied here allowed for the first time a comparative analysis of intra-host DENV-1 diversity according to the clinical presentation of the disease. We found that the extent of sequence diversity varied among infected patients. The composition of DENV-1 populations was different in classical (DF) and in severe infections (DHF and DSS). Although intra-host sequence variability was probably overestimated due to in vitro artefacts [46], genetic divergence was significantly lower in severe cases than in classical cases. In severe cases, d N and d S values were significantly lower than in classical presentations. In other words, DENV-1 populations were more genetically homogeneous in DHF or DSS cases than in DF cases. In our study, no correlation was found between the level of intra-host genetic diversity and viral load. Moreover, viral loads were not significantly different between the two groups, in a sample of five severe cases and five DF cases. It is therefore not likely that the lower intra-host genetic diversity observed in severe cases would have been influenced by larger amounts of template DNA in amplification reactions (associated with a more rapid saturation of PCR reaction and thus lower error rates).
The mechanisms that lead to different structures of DENV-1 intra-host genetic diversity according to the clinical severity remain undetermined. We do not know if the differences observed are the cause or the consequence of disease severity. Our findings suggest that further analysis of viral variation in both mosquitoes and human samples may in the future shed new light on dengue infection, pathogenesis and the existence of predictive factors of clinical outcome. www.plosntds.org length is not proportional to genetic distance. Numbers on branches represent bootstrap support for each branch. Found at: doi:10.1371/journal.pntd.0000493.s001 (0.28 MB TIF) Figure S2 Phylogenetic tree based on 53 nucleotide sequences of complete coding region of DENV-1 (Neighbor-Joining method, Kimura 2 algorithm). Taxon names of FP sequences correspond to D1.FP/sample number.year (month, geographical origin, clinical presentation). Taxon names of GenBank sequences correspond to D1.country/last two digits of year of isolation and GenBank accession number. In this condensed tree, branch length is not proportional to genetic distance. Numbers on branches represent bootstrap support for each branch.