Background

Dengue fever, caused by dengue virus serotypes 1–4 (DENV-1–4), represents a significant public health challenge in tropical and subtropical regions [1]. A vector-borne disease predominantly transmitted by Aedes aegypti mosquitoes, annually affects millions with symptoms ranging from mild fever to severe conditions like dengue hemorrhagic fever and dengue shock syndrome [1]. In the past 3 years, dengue activity has surged in the Americas. Approximately 3.4 million cases were reported in 2023, exceeding the 2.8 million total for 2022 [2]. Situated in the Caribbean basin, the Dominican Republic has been frequently grappled with dengue outbreaks over the past two decades. The high incidence of dengue in this country is thought to impact the overall epidemiology of dengue in the region.

In 2023, 28,078 cases have been reported by the country’s surveillance system which captures physicians’ reporting of dengue cases [2]. Though this surveillance system does not provide a complete picture of dengue transmission, it allows health authorities to maintain awareness of, and response to, dengue transmission throughout the country. In fact, the number of RT-PCR positive samples normally obtained per year, and exceeding 7,000 in 2023, is sufficiently high to provide an opportunity to further characterize transmission using genomic surveillance approaches or through additional analysis of related epidemiological data. In this regard, we implemented a genome-based surveillance approach in partnership with the Pan-American Health Organization and the Central Public Health Laboratory in Santo Domingo to sequence and analyze DENV whole genome sequences from the 2023 epidemic. We aimed and identifying existing lineages within the context of recent emergence of novel DENV-2 and -3 variants in the Americas [3]. In addition, by integrating genomic, epidemiological, and climatic data, we provide a historical overview of the epidemiological trajectory of DENV in the Dominican Republic. This comprehensive analysis lays the groundwork for establishing a genomic surveillance model in the Caribbean, with potential to monitor and respond to dengue transmission in the region.

Material and methods

Sample collection and whole genome sequencing

A total of 85 samples were obtained from patients exhibiting clinical symptoms consistent with dengue viral infection. All cases were classified as mild, with the most common symptoms being fever and cutaneous rash. These samples were sent to the Laboratorio Central de Salud Pública of the Dominican Republic in Santo Domingo for genomic sequencing. Initially, nucleic acid extraction was performed using the QIAamp Viral RNA Mini Kit (Qiagen). Subsequently, real-time reverse transcription PCR (RT-qPCR) for DENV 1–4 was conducted on the samples, following previously described methods [4]. Samples that tested positive (n = 85) with a cycle threshold (Ct) value of ≤ 36 underwent whole genome amplification using a set of tiled primers described elsewhere [5, 6]. The resulting DNA amplicons were purified using AMPure XP beads (Beckman Coulter, Brea, USA). Library preparation of the purified amplicons was carried out using the COVIDseq Kit (Illumina, San Diego, USA), originally designed for SARS-CoV-2 genomic research but adapted for other viruses [7, 8]. Subsequent sequencing was performed using the Illumina MiSeq platform (Illumina, San Diego, USA), following the manufacturer's recommended protocols. Consensus sequence generation was conducted using the ViralUnity pipeline (available at https://github.com/filiperomero2/ViralUnity) and typing assignment of dengue virus sequences was accomplished using the dengue virus typing tool available at https://www.genomedetective.com/app/typingtool/dengue/ [9].

Phylogenetic and phylodynamic inferences

We generated phylogenetic trees to investigate the relationship between the sequenced genomes from the Dominican Republic and those from other regions worldwide. Alignment of all sequences was performed using MAFFT [10] and manually adjusted using AliView [11]. Initial maximum likelihood phylogenies were constructed using IQ-TREE 2 software with the HKY + G4 substitution model [12]. Time-scaled trees were inferred using TreeTime [13] and BEAST software [14], preceded by TempEst [15] analysis to assess temporal signal. A rigorous model selection process, employing path-sampling (PS) and steppingstone (SS) procedures, was utilized to determine the optimal molecular clock model for Bayesian phylogenetic analysis [16]. The uncorrelated relaxed molecular clock model was chosen based on estimation of marginal likelihoods, incorporating the codon-based SRD06 model of nucleotide substitution and the nonparametric Bayesian Skyline coalescent model. To simulate the geographic spread of the identified 2022–2023 transmission clade, we utilized a flexible relaxed random walk diffusion model [17, 18], accounting for variation in dispersal rates across branches with a Cauchy distribution and a jitter window site of 0.01 [19, 20]. Each sequence was assigned geographic coordinates of latitude and longitude. Bayesian phylogenetic inference analyses were performed in BEAST v1.10.4, with two runs of 50 million Markov Chain Monte Carlo chains (MCMC) each, and samples were taken every 10,000 steps in the chain. Convergence was assessed using Tracer, ensuring effective sample size for all significant model parameters was greater than 200. The maximum clade credibility (MCC) trees for each run were summarized using TreeAnnotator after discarding the initial 10% as burn-in. Finally, the R package 'seraphim' version 1.0 [20] was used to extract and visualize spatiotemporal data contained within the posterior trees.

Eco-epidemiological modelling

Weekly reported cases of DENV in the Dominican Republic from 2013 to 2023 were gathered and structured from the PAHO database for dengue [21]. Notified infections were defined as dengue cases where a diagnostic test was performed and yielded a positive result, as per the PAHO platform criteria. The theoretical suitability of climate-driven dengue virus transmission was assessed using the mathematical expression of index P, incorporating humidity (u) and temperature (t) variables [21]:

$${P}_{(u,t)}=\frac{a^v_{(u)}{^{2}}\phi^{v{\rightarrow{h}}}_{(t)}\phi^{h{\rightarrow{v}}}_{(t)}\gamma^{v}_{(t)}\gamma^h}{\mu^{v}_{(u,t)}(\sigma^{h}+\mu^{h}(\gamma^{h}+\mu^{v}_{(u,t)}) }$$

Briefly, the index utilizes mathematical formulations based on empirically established connections between DENV and Ae. aegypti characteristics alongside meteorological factors. Climate-dependent characteristics encompass the extrinsic incubation period (〖γ〗((t))^v), adult mosquito lifespan (〖μ〗((u,t))^v), adult mosquito biting rate (〖a〗((u))^v), transmission probability per mosquito bite from infected human to susceptible mosquito (〖ϕ〗((t))^(h → v)), and from infected mosquito to susceptible human (〖ϕ〗_((t))^(v → h)). Climate-independent traits include intrinsic incubation period (〖γ〗^h), human lifespan (〖μ〗^h), and human infectious period (〖σ〗^h). Comprehensive methodological details and technical validation of Index P can be found in Nakase et al. [22, 23]. Monthly climate data for the Dominican Republic was sourced from Copernicus.eu satellite climate data [24]. Temperature data summary was derived by calculating yearly lowest, mean, and maximum values.

Results

A total of 85 complete genomes were obtained from 100 PCR positive samples with sufficient DNA (2 ng/L) for library preparation. The average cycle threshold (Ct) value for PCR was 24.6, with values ranging between 12.6 and 36 (Table S1). The ages of the patients sampled ranged from 2 to 48 years, with a median age of 17 years. Of these patients, 53% (n = 45) were male (Table S1), and all cases examined were classified as autochthonous. The sequencing procedure yielded an average coverage of 84%, ranging from 60 to 99% (Table S1). This allowed for the identification of the DENV-2 genotype III variant (n = 29) and the DENV-3 genotype III (n = 56) variant. Genome sequences were obtained from all three macro-regions of the Dominican Republic, which are further segmented into ten specific regions (Fig. 1A).

Fig. 1
figure 1

Dynamics of DENV-2 and DENV-3 in the Dominican Republic. A Map of the Dominican Republic (DR) depicting the sampling of new DENV-2 and DENV-3 genome sequences by region and district. The color and size of the circles represent the number of new genomes generated in this study: black for DENV-2 and white for DENV-3. B Time series of dengue cases reported monthly in Dominican Republic, presented as incidence per 100,000 inhabitants (gray bars) versus the climate-based suitability (index P, blue line) for the period data where available. The shaded areas in pink mark visually identifiable waves of cases (between April of consecutive years, see main text) and green marks the year 2023 from which all genomic samples were obtained. C Long-term time series of monthly climate-based suitability (index P, 1981–2023, blue line). The moving average of suitability (plus-minus 18 months) is presented in black before the year 2001 and in red thereafter (the shaded gray area highlights this time period)

Full size image

The distribution of the DENV-2-III and DENV-3-III genotypes reflected current DENV transmission patterns. DENV-2-III has been circulating predominantly in the north-central parts of the country since 2014. In contrast, DENV-3-III, recently introduced through population-dense areas, was found in both the north-central and southeastern regions (Fig. 1A).

While climate-based suitability for DENV transmission presented clear, yearly oscillations demarking time periods with potential for dengue activity, notified DENV cases (between 2014 and 2023) revealed only three waves, of cases (defined as increases in the number of cases above the historical median) occurring in 2015–2016, 2019–2020 and 2021–2022 (Fig. 1B). The reasons why transmission is low in some years are unclear, though we can consider a mix of local factors including accumulated herd-immunity to specific serotypes, yearly changes in mosquito populations not captured by the climate-based suitability measure, and population cross-immunity post Zika virus emergence after 2016 [25,26,27,28]. Nonetheless, during the three epidemic waves, Pearson’s correlation between cases and suitability was high, at 0.88 for 2015–2016, 0.76 for 2019–2020 and 0.93 for 2021–2022. Together, these results suggest that, although climate suitability alone is insufficient to promote waves of DENV transmission, local temperature variation tends to be highly associated.

We also explored the historical trends in climate-based suitability in the Dominican Republic (over the past 40 years, Fig. 1C). This study revealed recent trends towards higher suitability for DENV transmission, particularly after the turn of the century, which derives from the emergence of much higher seasonal peaks of suitability; before 2001, the average yearly peak of suitability was 1.42 (min 0.95; max 1.89), while after 2001, the average peak was 2.08 (min 1.45; max 2.76). However, the time period analyzed is too short for a statistical assessment of significance. Local climate change, driven by rises in temperature, is theoretically favoring the transmission of DENV in the Dominican Republic, as in other countries of the region.

To understand the phylogenetic history of DENV-2-III, we combined our recently sequenced samples (n = 29) from 2023 with an additional set of 14 DENV-2-III viral genomes that were sequenced in the Dominican Republic in 2022 [29]. This set included samples collected from the Southern, Eastern regions, and the National District (Santo Domingo) [29]. These sequences were then combined with other sequences of the same genotype (n = 647) retrieved from GenBank for context. Our analysis revealed that the new genomes from this study cluster with sequences recently obtained by others [29] and form a distinct monophyletic clade with robust statistical support that is basal to the DENV-2-III BR4 variant that emerged in 2019 in Brazil (Fig. 2) [4].

Fig. 2
figure 2

Dispersion dynamics of DENV-2-III in the Dominican Republic. A Maximum likelihood (ML) phylogenetic analysis of 29 new complete genome sequences of DENV-2-III generated in this study combined with 661 sequences from GenBank. The scale bar represents units of nucleotide substitutions per site (s/s) and the tree is mid-pointed rooted. Colors represent geographic sampling locations. B The highlight on the right (panel B) shows the phylogeographic reconstruction of the Dominican Republic Clade (n = 43). Solid curved lines denote the links between nodes and the directionality of movement. Circles represent nodes of the MCC phylogeny and are colored according to their inferred time of occurrence. Shaded areas represent the 80% highest posterior density interval and depict the uncertainty of the phylogeographic estimates for each node

Full size image

Additionally, we reconstructed the viral movements of this clade (n = 43) within the different regions in the Dominican Republic. We estimate that the mean time of origin of this variant was early-January 2014 with a 95% highest posterior density (HPD), ranging from mid-June 2013 to late-January 2014. These results suggest that the 2023 epidemic may have not been caused by a novel introduction but could be the result of continual transmission within the Dominican Republic of a viral strain that was introduced in the region in early 2014. This variant spread from the southern part of the country (Ozama region) toward the southeastern, northern and midwestern, as demonstrated in Fig. 2.

To explore the phylogenetic history of DENV-3-III in the Dominican Republic, we combined our new genome dataset (n = 56) with 1,760 sequences from GenBank for global context. Our analysis revealed that the new genomes group within the recently reported American lineage II of genotype III; however, our sequences formed a distinct monophyletic clade that separates from the main lineage, which includes viral sequences recently identified from other Caribbean and Latin American countries, such as Cuba, Puerto Rico, Brazil, Suriname, and as well as in North America (specifically Florida) [3, 29]. These results propose a complex transmission scenario, with introduction events likely mediated by trans-continental travel and underscores the significant influence of human movement in facilitating the spread and introduction of viral lineages to new regions [8] (Fig. 3).

Fig. 3
figure 3

Dispersion dynamics of DENV-3-III in the Dominican Republic. A Maximum Likelihood (ML) phylogenetic analysis of 56 new complete genome sequences of the DENV-3-III genotype from this study, in addition with 1,760 sequences from GenBank representing all different DENV-3 genotypes. The scale bar represents units of nucleotide substitutions per site (s/s) and the tree is mid-pointed rooted; B Maximum clade credibility tree (MCC) of the DENV-3-III emerging American lineage II from 2022–2023, including n = 56 complete genome sequences from the Dominican Republic from this study combined with 168 additional genomes for context. Sequences are colored according to sampling location. C Phylogeographic reconstruction of the Dominican Republic Clade (n = 56). Circles represent nodes of the MCC phylogeny and are colored according to their inferred time of occurrence. Shaded areas represent the 80% highest posterior density interval and depict the uncertainty of the phylogeographic estimates for each node. Solid curved lines denote the links between nodes and the directionality of movement

Full size image

We investigated the spatial–temporal dynamics of the Dominican Republic DENV-3-III variant in more detail using a smaller data set (n = 56) derived from the American lineage II. Phylogeographic analyses allowed the reconstruction of viral movements across different regions in the Dominican Republic (Fig. 3) and suggested a mean time of origin in early April 2023 (95% highest posterior density (HPD): 2 to 27 April 2023). This variant spread from the south of the country (Yuma region) towards the north and later to the southern part of the country, as indicated by isolates from the Cibao Noroeste, El Valle, and Enriquillo regions (Fig. 3).

Conclusion

In conclusion, our genome-based surveillance system, integrated with other methodologies, provided a nuanced view of the recent dengue epidemiology in the Dominican Republic. By elucidating a key transmission pathway from the South to the North for both genotypes, identifying co-circulation of different DENV serotypes and genotypes, and highlighting the cumulative effect that local climate change appears to be having on the transmission potential of DENV, we underscore the importance of advanced monitoring techniques. The identification of the South as point of entry for viral variants requires further study, in particular given the presence of likely introduction points (international airports and cruise ports) being present both in the North and the South. Seroprevalence studies are also urgently needed to characterize the local immunity landscape against different serotypes in search for a better understanding of the actual serotype spatial distribution in the island, and its possible impact on regional public health in future serotype switching events. Only through active tracking of viral evolution and spread can we enable timely interventions, ensuring communities are fortified against the persistent threat of dengue and fostering adaptability in response to the changing landscape of infectious diseases.

Limitations

In this study, we elucidate the recent transmission dynamics of the dengue virus in the Dominican Republic by combining genomic sequencing, epidemiological data analysis, and an examination of historical climate patterns. However, the scarcity of complete DENV genome sequences in Latin America limits our ability to characterize the molecular epidemiology of viral strains at a regional level, highlighting the importance of increasing sequencing efforts to improve real-time data generation, sharing, and representativeness. Despite this limitation, our phylodynamic analysis offers valuable insights into the dynamics of DENV in the country, providing a foundation for better-informed public health strategies and interventions.