We analyzed the phylogenetic relationships of DENV-2 strains isolated from countries of Central America from 1999 to 2009. To the best of our knowledge, this is the first report on the phylogeny, molecular clock and selection pressure analysis of DENV-2 in this region. We also report here the first complete genome sequence of a Guatemalan DENV strain, and the second DENV-2 ever sequenced from that country. Our phylogenetic analysis suggests that two lineages of DENV-2 of AM/AS genotype have circulated in Central America and in Nicaragua, they have co-circulated at a certain point in the recent past.
An inherent limitation of this analysis is the fact that most sequences of DENV-2 strains from Central America available in the GenBank are from Nicaragua. Only a very small number of sequences available are from the rest of the countries of Central America, and those are mostly partial sequences. Thus, additional fully sequenced strains of DENV from other countries of Central America are needed in order to perform analyses to better understand the intricacies on the dynamics of the circulation of DENV in that region.
The phylogeny of DENV-2 in Central America has been largely hypothesized to be reflected by the Nicaraguan scenario, from where an important number of sequences spanning more than a decade of collections are available for study. The exact year in which the AM/AS genotype of DENV-2 was introduced into countries of Central America other than Nicaragua, and its complete phylogenetic history is hard to trace due to the unavailability of sequences spanning the last 15 years, time in which that genotype started to replace all other genotypes in the continent 
. Data from our molecular clock analysis suggest the TMRCA for the AM/AS genotype strains from Central America existed around 19 years ago (95%HPD 15–22 years).
The first reference to the origin of Nicaraguan DENV-2 strains was made during the epidemic of 1998, where Balmaseda et al 
using restriction site-specific PCR (RSS-PCR) identified the isolates circulating as belonging to the “Jamaica” subtype (AM/AS genotype). More recently within the Pediatric Dengue Cohort Study (PDCS) conducted in that country since 2004, a number of strains have been sequenced and phylogeographic analyses have been conducted identifying two clades of DENV-2 circulating in its capital city Managua. These clades, denominated “clade 1” and “clade 2” by Balmaseda 
, correspond respectively to clades 2a and 2b in this study. One of these circulating clades (“clade 2” in Balmaseda's nomenclature) has been associated with severe dengue and also with the appearance of a higher proportion of symptomatic dengue cases in the cohort during 2007-2008 
. In those studies, the authors suggest that the older “clade 1” circulated during 2004–2005 and 2005–2006 and “clade 2” was only found during 2006–2007 and 2007–2008. In our analysis, we have found that a number of strains from 2004 and 2005 (therefore corresponding to their “clade 1”) clustered together with strains from 2007 and 2009 ( and ). That finding suggest that the co-circulation of both clades still happened in later years, perhaps in lower frequency than before, but in fact the new “clade 2” has not totally replaced “clade 1” in Nicaragua. In our analysis, the TMRCA for clade 2b (that includes strains from Nicaragua, Honduras and Guatemala) appears between 9 and 12 years ago (mean 11 years), coinciding with the occurrence of large DENV epidemics associated with circulation of DENV-2 with any of the other serotypes in the countries of that region 
While in Nicaragua strains from this AM/AS genotype has been linked to severe dengue, this has not been the case in other countries of that region 
. However, the low number of cases of severe dengue reported in Belize, Costa Rica, El Salvador, Guatemala, and Panama most likely result from under-reporting or misdiagnosis, which have been proven to occur in other dengue endemic countries in the Americas 
. On the other hand, Honduras is by far the country that has reported more cases of dengue, dengue hemorrhagic fever/severe dengue and deaths associated with the disease in Central America in the last decade (Figures S1
, and S3). Interestingly, it may have happened that opposed to other Central American countries that may have been underreporting, in Honduras cases of other febrile diseases may have been accounted as dengue cases, which were most likely not confirmed by laboratory tests. Consequently, the number of symptomatic dengue cases in that country may be over reported. This interesting epidemiological question deserves to be studied in more detail.
Another limitation encountered when performing such epidemiological analyses in that the official statistics available from PAHO, which are received directly from the Ministries of Health from the countries of the region, sometimes reflects the number of suspected dengue cases while in others refer to confirmed cases only, as is the case of Nicaragua 
. Even in the case of Nicaragua, recent reports show significant differences in the number of confirmed symptomatic dengue cases obtained in the PDSC study in comparison with data provided by the Nicaraguan Ministry of Health, which is used as source for the official statistics compiled by the PAHO 
Central America represents an interesting place to the evolutionary dynamic of dengue. The region is surrounded by the Caribbean and South America, both of which have been continuous foci of dengue transmission in the last decades 
. DENV-2 strains from Asian origin have replaced the autochthonous AM strains that were circulating in the region during the 80s and 90s 
. This has been linked, at least in a part, to the increasing number of severe dengue cases reported in the continent 
Circulation of DENV-2 strains from three genotypes and its ultimate replacement by strains of the AM/AS genotype has been reported in Central America. Bennet et al. 
reports phylogenetic analyses that includes a number of Central American strains of the AS-I genotype (strain ES 12 00 from El Salvador), AM genotype (strains CR A and B from Costa Rica) and the AM/AS genotype (CR 6530/2000 and CR 7945/2000 from Costa Rica and strain ES 18 99/1999 from El Salvador). All these AM/AS genotype sequences were included in our E protein gene phylogenetic analysis. Similarly, Foster et al 
found four Honduran strains from 1991 (HON A–D) and a single strain from El Salvador (SAL 1987) belonging to the AM genotype, while a single Honduran strain (HON 1986) was identified to be closely related with the prototype strain New Guinea C (NGC), from the AS-II genotype.
Circulation of strains of the AS-I or AS-II genotypes in the Americas has been previously reported in countries outside Central America (Venezuela, Cuba, Haiti and Mexico) 
. However, that seems to be unlikely and possibly the resultant of laboratory contamination with the strain NGC, a commonly used prototype strain present in most of the laboratories involved in dengue research throughout the world. Cases of NGC-like strains have been detected and reported even in different regions outside the Americas (i.e. China and Vietnam 
), but none of these NGC-like strains have the amount of variation expected based on the consistently reported substitution rates to which epidemic DENV-2 would be subjected in nature 
. On the other hand, a single strain isolated in 1996 (strain BC17) from the Mexican state of Yucatan which is a neighbor of Guatemala, has been reported to cluster within the COS genotype 
. That finding has been considered as evidence of limited circulation of this genotype in that region, although as in the case of the strains of AS-I and AS-II genotypes identified circulating in the continent, it may also be the result of erroneous identity and potential contamination.
Our analysis shows that strains GU/FDA-GUA09/2009 and HN/F07-075/2007 constitute a different lineage within sub-clade 2b of the AM/AS genotype of DENV-2. This finding is supported by high bootstrap values in ML analysis and Bayesian high posterior probability values. Thus, our analysis confirmed that this linage of the AM/AS genotype represented by GU/FDA-GUA09/2009 and HN/F07-075/2007 strains, circulated in 2009 in Guatemala. These findings agree with the suggestion by Tang et al. 
that the strains currently present in Guatemala and Honduras may be the result of re-emergence of autochthonous DENV-2 strains introduced sometime in the past, instead of being newly introduced viruses. If this is true, it is possible to speculate that there is more than one lineage of DENV-2 circulating in these countries, and that idea can be supported by our results that also identified at least two lineages of DENV-2 circulating simultaneously in neighboring Nicaragua. There is a possibility suggested by the positioning of strain NI/BID-V2924/2000 basal to clade 2b ( and ) that in the past, strains closely related to NI/BID-V2924/2000 were introduced from Nicaragua to neighboring Guatemala and Honduras, and that continuous introduction resulted in clade replacement by strains with higher fitness, followed by in situ evolution of these lineages. This may be an epidemiological feature of DENV-2 transmission dynamics in Central America.
In relation to the selection pressure exerted over DENV-2, a number of reports 
and our own results shown that overall the amino acid sequences in DENV-2 are highly conserved, therefore suggesting that the virus is subjected to strong purifying selection. However, in our analysis we were able to detect codons under weak positive selection in the Central American DENV-2 strains analyzed: two when analyzing E protein gene datasets alone (codons 83 and 131) and nine when analyzing ORF datasets (codons C41
The findings of positive selection in codons of E are expected since this protein is subjected to immune pressure and a number of codons in E have been reported to be subjected to this type of selection. From the three codons found under positive selection in the Central America datasets used in our analysis two (129 and 131) have been reported before by Twiddy et al and Bennet et al 
when analyzing datasets of Cosmopolitan, Asian and AM/AS DENV-2 strains, but codon 83 identified by our analysis appears to be a novel addition to the already numerous sites identified under positive pressure in the E protein gene to date (reviewed in 
). In congruence with Vasilakis et al. 
, we did not find evidence of positive selection on E protein position 390, which has been suggested as a virulence marker for DENV-2 
. However, we found that the mutation D→N390
was fixed in all the analyzed DENV-2 strains from Central America since 1999 (G. Añez, information not published).
We also detected for the first time positive selection in a codon of the capsid of the virus (C41
), close to the hydrophobic conserved region of the protein 
, although only when using the IFEL method and very close to the statistical significant threshold. The capsid of DENV contains the genome viral, and no site under positive pressure had been reported for any of the DENV serotypes to date. However, it may be possible that in the in situ evolution scenario that the AM/AS genotype of DENV-2 is undergoing in Central America, viruses with amino acids residues responsible for advantageous replication are been selected in this protein, and that the role of the adaptive pressure over this site in the capsid protein needs to be properly addressed by detailed studies.
We are also reporting for the first time all sites detected by our analyses under positive pressure in the NS2A, NS3 and NS5 genes. Since the function of the small NS2A protein is not well understood, it is hard to speculate about the significance of codons found under positive pressure in this protein. NS2A, NS4A and NS4B has been shown to antagonize the interferon response during DENV infection 
, and it would be evolutionary advantageous for the virus to select for strains with strong innate immunity suppression mechanisms. As for the sites detected in NS3, there have not been reports on positive selection in this protein in none of DENV serotypes before. NS3 is important protein for the replication of the virus, since it possesses protease and ATPase/helicase functions necessary for viral replication 
, and as in the case of the capsid protein, this finding needs to be further investigated.
Three codons were identified under positive pressure in the NS5 protein, which is crucial for viral replication for its RNA- dependent RNA polymerase 
. All these sites were detected only by one of the methods employed, except for one site (NS5523
) that was detected by both FEL and the IFEL methods in both ORF datasets. In the case of the other two sites reported in the NS5 protein (269 and 850), these sites were detected by either one of the methods (FEL and IFEL) with statistical significance, while p values from the other method were very close to the established significance limits (). Jointly, these results should be studied in greater detail to elucidate the possible role of these residues in the pathogenesis of the disease.
Nucleic acid amplification technology (NAT) assays has been developed and implemented in the detection of DENV in endemic countries, as well as in places in which the disease is imported and can cause autochthonous transmission, as it happened recently in the US 
. Even though primers and probes for these assays are designed to target conserved regions of the viral genome, there is always a risk that the assay could fail to detect variants bearing mutations located at the target area. Monitoring of the occurrence of mutations in DENV appears to be a solution that allows for updating diagnostic NAT assays when needed. The use of complete genome sequencing appears to be more suitable than partial sequencing in order to detect variations that could appear all over the DENV genome and could negatively affect the performance of NAT tests.