Article | Open | Published:

Genomic and epidemiological characterisation of a dengue virus outbreak among blood donors in Brazil

Abstract

Outbreaks caused by Dengue, Zika and Chikungunya viruses can spread rapidly in immunologically naïve populations. By analysing 92 newly generated viral genome sequences from blood donors and recipients, we assess the dynamics of dengue virus serotype 4 during the 2012 outbreak in Rio de Janeiro. Phylogenetic analysis indicates that the outbreak was caused by genotype II, although two isolates of genotype I were also detected for the first time in Rio de Janeiro. Evolutionary analysis and modelling estimates are congruent, indicating a reproduction number above 1 between January and June, and at least two thirds of infections being unnoticed. Modelling analysis suggests that viral transmission started in early January, which is consistent with multiple introductions, most likely from the northern states of Brazil, and with an increase in within-country air travel to Rio de Janeiro. The combination of genetic and epidemiological data from blood donor banks may be useful to anticipate epidemic spread of arboviruses.

Introduction

Dengue virus (DENV) is one of the most important vector-borne diseases worldwide, causing an estimated 390 million infections annually in over 100 countries1. DENV belongs to the Flavivirus genus and is classified into four genetically and antigenically distinct serotypes, DENV1-42,3. The geographic range of DENV types has dramatically increased in recent decades4, fueled by the expansion of its main vector species, the Aedes aegypti mosquito, human population growth, international travel and trade, and increasing urbanization in the tropics and subtropics5,6.

Brazil has reported the highest number of dengue cases worldwide, with >7.3 million notified infections between 2000–20137, and transmission occurring in most of the country8. Current estimates of DENV burden in Brazil may however be largely underestimated; recent work suggests that actual number of cases may be 12- to 17-fold higher than those captured by passive case reporting implemented in the country9. In Latin America DENV disease affects mostly adults and manifests as primary dengue fever (DF). However, the clinical and epidemiological profile of DENV in Brazil has been shifting in recent years. As a result of the increased co-circulation of different DENV serotypes in the country, DENV infection is now becoming associated with more severe cases in younger populations10. In addition, Brazil has experienced recurrent severe DENV outbreaks of increasing magnitude over the past three decades, with particularly high incidence peaks in 2002, 2008, 2010, and 201211.

After more than two decades of absence from Brazil, DENV4 has been re-introduced at least six times into the northern part of the country until 201112. This region is characterized by a high suitability for Aedes mosquitoes13,14 but has the lowest human population density in Brazil and comparatively poor transportation connections to the rest of the country. Although most of these DENV4 introductions resulted in short-lived transmission chains, in 2012–2013 DENV4 spread beyond the northern region and caused large urban outbreaks in central and southern Brazil15. Today, the four DENV serotypes co-circulate in an increasing number of Brazilian regions4,16,17,18.

Despite the morbidity and mortality caused by DENV infection, there is limited information about what drives the transmission potential and spatial spread of novel DENV lineages, since most epidemiological studies rely on clinical case reports, which rarely discriminate infection by DENV serotype and are sometimes confounded by infections caused by other arthropod-borne viruses, for example Chikungunya, Zika, Oropouche or Mayaro virus. Further, current investigations of DENV dynamics often do not fully exploit the information contained in genome sequences about virus introduction, invasion and persistence and report only cross-sectional case data, whilst longitudinal data are needed for detailed investigation of viral population dynamics19. Blood donor screening, combined with retrospective analysis of circulating virus strains represents a valuable alternative approach for epidemic characterisation and the early detection of newly emergent flaviviruses20,21,22,23.

The first report of DENV4 genotype II (DENV4-II) in Brazil was in 1981–1982 in Boavista, the capital city of the northernmost state of Roraima24. After a limited outbreak there, DENV4 was absent from the country until 2005–2007, when it was detected in the northern Amazon state that neighbours Roraima24. In 2010 DENV4 was found again in other northern states (Fig. 1A). However, it was only in 2012 that DENV4-II caused large outbreaks in more densely populated states of Brazil, such as Rio de Janeiro (RJ) and Pernambuco (PE) (Fig. 1B). The municipality of Rio de Janeiro is an important tourism destination, with well-connected air, road and railway networks and one of the largest harbours in the Americas. Local Aedes spp. populations are highly susceptible to DENV25 and previous studies suggest that Rio de Janeiro has played a key role in the national dissemination of several dengue serotypes, including DENV1 in 198626, DENV2 in 199027 and DENV3 in late 200028.

In order to characterize and understand the epidemic behaviour of the lineage that caused the 2012 outbreak of DENV4 we undertook retrospective molecular surveillance of the largest Brazilian study of transfusion-transmitted DENV22. This included a total of 40,278 blood donors and recipients in the capital cities of RJ and PE states, namely in Rio de Janeiro city (RJc), and Recife. In our study, individuals were recruited between January and July 2012 and a mean DENV RNA + prevalence (per unit) of 0.51% was found in Rio de Janeiro, and 0.80% in Recife22. From these RNA + samples, we obtained a total of 92 complete or near-complete DENV4 genomes using next generation sequencing. We then combined evolutionary analysis with ento-epidemiological modelling to investigate the introduction, spread and impact of environmental factors on the 2012 DENV4-II outbreak. Our study demonstrates the potential of blood bank screening for understanding the epidemiology and molecular evolution of DENV and other arboviruses currently circulating in Brazil.

Results

A total of 147 DENV RNA + samples were obtained and submitted to amplification of whole genome, of those 92 were PCR + for at least one of the 5 fragments. The length of the consensus genome sequences obtained varied between 1,215 and 10,066 nt in length, with a median of 8,752 nt, of which 42 were whole genomes (for details see Table S1). From the remaining 57 DENV RNA + samples that failed to amplify, viral load was available for 42, and only 2 samples had a viral load > 103 copies/mL (1,116 and 3,100 copies/mL). As expected, viral loads were positively correlated with the length of the consensus genome obtained (Pearson’s correlation, adjusted r2 = 0.62, p-value < 0.0001).

There was a strong correlation between the total number of DENV RNA + samples in blood bank donors and recipients in each month and the number of notified dengue infections in RJ municipalities during that month (Pearson’s correlation for RJ municipalities: adjusted r2 = 0.97, p-value = 0.0002) and in RJc neighbourhoods (adjusted r2 = 0.86, p-value = 0.011). Further, we found a moderate but significant spatial correlation between the number of cases in each municipality in RJ or RE and the estimated human population density in that municipality (Spearman’s rank correlation, $$\rho$$ = 0.72; p-value < 0.001; human densities were obtained from high-resolution gridded datasets for Brazil29, Fig. 1D,E). This suggests that the demographics of the blood donor-recipient cohort analysed here was roughly representative of the population under study.

To place our newly generated DENV genomes into a global evolutionary context, we estimated maximum likelihood (ML) and Bayesian molecular clock phylogenies from an alignment comprising a total of 218 DENV4 genomes sampled worldwide sampled from 1972–2013, including the 92 novel Brazilian genomes reported here (Fig. 1F); a fully annotated phylogeny is shown respectively as Fig. S1. Both the ML and Bayesian phylogenies show that 98% (88 of 90) of the novel Brazilian DENV isolates fall within a single well-supported monophyletic clade (maximum likelihood bootstrap score, BS, = 87; Bayesian posterior probability, PP, = 0.99). This clade is closely related to viruses from Venezuela and from the northern Brazil region that contain the capital cities of Roraima (RR) and Amazonas (AM), that have a high suitability for Ae. aegypti and high human population density (Fig. 1D,F).

There was a strong correlation between the genetic distance of each sequence to the root of the global DENV4 phylogeny and the date of sequence sampling for the global dataset (Fig. 1F inset, r2 = 0.76). Despite this, the relaxed molecular clock model exhibited a high among-lineage coefficient of variation (1.74; 95% Bayesian credible interval = 1.38–2.14), indicating significant rate heterogeneity among branches. Such high among-lineage rate variability is not unexpected for flavivirus phylogenies that comprise genomes that have been longitudinally sampled over 40 years19,30,31. The estimated genomic evolutionary rate for the global DENV4 dataset was 1.36 × 10−3 substitutions per site per year, s/s/y (95% Bayesian credible interval = 1.17–1.53 × 10−3 s/s/y).

Our data also reveals for the first time the circulation of DENV4 genotype I (DENV4-I) in Rio de Janeiro. The two DENV4-I isolates from RJ cluster together with maximum statistical support (BS = 100, PP = 1.00) and are related to isolates sampled in Southeast Asia in 2002–2011 (Fig. 1F). In addition, these two isolates are polyphyletic with respect to a single DENV4-I isolate previously detected in Bahia state in 201112. One DENV4-I isolate was also detected in 2012 in the harbour city of Santos, São Paulo state (AC da Costa, personal communication). Together, these data suggest that DENV4-I appears to have been introduced to Brazil from Southeast Asia on at least two separate occasions.

The time of the common ancestor of the outbreak clade was estimated to be February 2010 (95% Bayesian credible intervals, BCI: April 2009 to September 2010, Fig. 2A). In February 2011, DENV4 was confirmed for the first time in the northeast of Brazil (states of Bahia, Pernambuco and Piauí) and in May 2011 the virus was confirmed in the southeast (São Paulo and Rio de Janeiro)32,33, but fewer than 100 cases were detected in total in these locations in 2011.

Crucially, DENV4 had never been confirmed in Rio de Janeiro municipality before mid-May 2011, raising the following question: where did the outbreak lineage reside before its earliest detection in the city? One hypothesis is that the most recent common ancestor of the outbreak strains in Rio descends from a single lineage that circulated undetected in the city prior to the ignition of the outbreak (scenario A). Alternatively, it is possible that the outbreak in Rio de Janeiro has multiple ancestors, each of them originating from a pool of unsampled diversity outside of Rio. Likely locations of this source population are the urban areas of the Brazilian or Venezuelan Amazon with high suitability for DENV transmission (Scenario B). Consistent with this we observe that (i) phylogenetic analyses place the ancestors of past Brazilian DENV4-II outbreaks in Venezuela (Fig. 1F and Fig. S1), which is connected by the Branco and Amazon rivers to Boavista (Roraima state) and Manaus (Amazonas state), and (ii) two isolates sampled in mid-2010 in the city of Roraima are consistently placed amongst the early lineages of the outbreak clade (Fig. 1F and Fig. 2A).

We next sought to investigate why and how the outbreak in Rio de Janeiro ignited only in January 2012, and not before. Due to the extensive genetic sampling of the outbreak, we were able to estimate temporal fluctuations in its effective reproduction number, R e, using the birth-death skyline (BDSKY) model, based on a phylogenetic birth-death process34. R e is defined as the average number of potential secondary infections from an infected individual in the presence of herd-immunity at each point in time during the outbreak. R e  > 1 indicates increasing number of cases, and a potential epidemic, while R e  < 1 indicates that the number of cases decreases with each generation, and if the decrease is maintained the epidemic with be eliminated. R e equals the basic reproduction number, R 0, at the start of an epidemic when the host population is completely susceptible.

The BDSKY results are shown in Fig. 2B. Although the BDSKY uncertainty intervals are larger in the past, the analysis shows seasonal variations that are consistent with epidemiological information (Fig. 1C). R e started to increase around December 2011 and peaked around mid-March 2012, at a value of R e  = 1.9 (95% BIC: 1.5–2.6) (Fig. 2B). Around mid-June 2012, estimated R e declines to <1, coinciding with the observed end of the epidemic. More generally, the BDSKY suggests a rise during the first 3 months of 2012 followed by a fall in R e < 1 in the second quarter of the year.

The detection of seasonal fluctuations across multiple years is encouraging given that almost all sequences used in analysis were sampled in the first half of 2012. Using the BDSKY approach we also estimated the generation time, i.e. the period between two infections mediated by the vectors, to be 21 days (95% BCI: 11 to 34 days). The proportion of cases that were sampled was estimated to be 3.42 × 10−5 (95% BCI: 2.12 × 10−8 to 1.22 × 10−4). If we consider human population sizes of the three sampled federal states from which genome data was analysed (RJ, PE and MS), this corresponds to an infected population of at least 940,000 individuals until the end of July 2012, which is around the triple of the number of official cases notified during this period in Brazil (Fig. 1b). The phylogeny is significantly structured by geographic location (federal state) of sampling (Association index = 0.43; bootstrap significance = 1.00), but this spatial structure does not appear to impact our BDSKY estimates for 2012 since the model captures well the observed oscillations in notified cases (Fig. 1C).

To investigate whether the ignition of the DENV4 outbreak in Rio de Janeiro was linked with an increase in the number of travelers from other destinations within Brazil, we collated data on the number of passengers arriving on commercial flights into Rio de Janeiro from all others airports across Brazil between January 2010 and December 2012. These data were obtained from the International Air Transportation Association, which collects data on an estimated 90% of all passenger trips worldwide. We find that between 2010 and July 2012, the number of passengers per month to Rio de Janeiro nearly doubled, increasing on average 77% per year during this period (Fig. 2C).

Finally, we used a susceptible-exposed-infected-recovered (SEIR) mathematical model to explore in detail the ecological conditions and vector transmission dynamics underlying the 2012 outbreak, whilst explicitly accounting for temperature variation as a major determinant of vector capacity35 (see Methods). A time series of incidence during the outbreak was compiled from the DENV4 RNA + weekly cases in Rio de Janeiro and fitted to the model. Figure 3 shows the daily estimates of key epidemiological parameters from the SEIR model. Notably, the initial increase and peak in R e in 2012 observed from genetic data using the BDSKY approach (Fig. 2B) matches the estimates of R e from the ento-epidemiological SEIR model (Fig. 3A). However, the lower R e until end-2011 estimated from the genetic data may reflect the absence of pre-outbreak dengue isolates (Fig. 1B). We obtain a similar trend using the ento-epidemiological SEIR model; since the latter does not depend on pre-outbreak case counts, these findings support the notion that mosquito and temperature factors were insufficient for outbreak transmission between mid-2011 and early 2012. According to the SEIR model, the date of the first human case of the DENV4-II outbreak in Rio de Janeiro occurred around Nov-Dec 2011. This timing is consistent with our molecular clock estimates and with multiple concurrent introductions to Rio de Janeiro from northern Brazil with high suitability for DENV endemicity (Scenario B).

Besides capturing the dynamics of the DENV4 outbreak (Fig. 3A), the ento-epidemiological SEIR model recovers several other parameters such as the life expectancy and extrinsic incubation period of the mosquito vector (Fig. 3B). In line with previous observations from field mosquitoes in Rio de Janeiro36,37, we estimated the extrinsic incubation period of Aedes (i.e. time necessary for the virus to be detected in saliva and ready for transmission after being ingested with the blood meal) to be around 7 days (95% CI: 4.4–10 days), and its life span around 10.5 days (95% CI: 10.1–12.3).

Discussion

In this study, we characterized the dynamics of a large urban dengue virus outbreak in Brazil. By combining genetic and epidemiological models applied to blood bank surveillance data, we show that the establishment of the DENV4-II outbreak in Brazil was contingent upon the introduction of new strains to large population centres during a time window of high transmission potential (Re > 1). We obtained similar epidemic trends and parameter estimates using two fundamentally different approaches based on independent data sources, a birth-death phylodynamic model and an ento-epidemiological SEIR model. Notably, the estimated peak of epidemic transmission in March 2012 (Re 2) coincides with the observed peak in DENV detection in blood-donor banks, highlighting the potential of blood bank screening for surveillance and early outbreak detection. More generally, the urban and peri-urban areas of residence of blood donor/recipient population under study may be highly suitable for arboviral outbreaks, as measured by Aedes suitability weighted by human population density (Fig. 1E).

Our study emphasises that sequence data can support and corroborate epidemiological data when the former is weak, absent, or affected by under-reporting. The current DENV surveillance system in Brazil is based on passive surveillance, with clinical testing through antigen detection assays. There is often serological cross-reactivity, not only among different DENV lineages but also with other circulating Flaviviruses38,39,40. Approximately 213,000 dengue cases were notified to the national surveillance system in RJ and PE during 2012. Consistent with previous seroincidence studies22,41, we estimate at least a 3.4-fold greater number of dengue infections compared to those that resulted in reported disease. The approach described here is directly applicable to other arboviruses that spread via the same Aedes vector such as Chikungunya, Zika and urban yellow fever virus.

Temperature-based models and entomological surveys have shown that the Amazon region of Brazil is highly suitable for year-round DENV transmission42,43,44. Since the clade responsible for the 2012 outbreak seems to descend from viral strains circulating in the Amazon region, the most parsimonious explanation for the absence of notified cases in urban centres under study until 2012 (Fig. 1B) is that DENV4-II circulated undetected in the northern tropical areas of South America. A similar pattern has been observed for specific genotypes of DENV types 1, 2 and 330. This suggests a sink-source dynamic for DENV in Brazil, with locations in the northern region acting as a DENV reservoir, or as a stepping-stone location for virus spread from countries outside Brazil to the rest of Brazil. Further, the fact that Re is only >1 for part of the year in Rio, as shown by the SEIR model and the BDSKY, indicates that Rio consistently receives new DENV virus lineages. In other words, transmission chains within Rio are unlikely to be sustained across multiple seasons and will therefore require re-seeding from a source location; this re-seeding is a shaped by local vector ecology and human mobility30,45. While the incorporation of vector species dynamics into a viral evolutionary framework remains challenging46, the modeling approach employed here could be extended to capture not only multi-vector dynamics but to generate daily estimates of key parameters using real-time temperature fluctuations.

Human population density in Brazil is highest in the eastern and southern coastal urban centres29, in climatic areas where the risk of DENV establishment after its introduction is usually limited to a particular yearly window. For Rio de Janeiro, which hosted the Olympic and Paralympic Games in August and September 2016, our results suggest that the risk that an introduced arbovirus lineage (against which there is little herd immunity) will generate a sustained chain of transmission is highest between November and December, which coincides with the peak in annual passengers arriving on commercial flights into Rio de Janeiro (Fig. 2). Thus, whilst introduction of new DENV lineage or other arboviruses to Rio de Janeiro is a continual process, there was little risk that introductions would ignite large local outbreaks within Rio as a result of these popular sporting events.

Our study also provides the first report of an introduction of DENV4-I in Rio de Janeiro. A single genome isolate of DENV4-I had been previously collected in March 2011 in Bahia12, but the pair of sequences in Rio de Janeiro identified here reveals a second independent introduction of this genotype into Brazil, and there is some evidence that DENV4-I exhibited limited circulation in the Amazonas state in 200847.

Our estimates of viral and vector parameters obtained using mathematical modelling are consistent with estimates previously published for DENV and Aedes vectors. Namely, our within-season (years 2009–2012, months December to April) estimates of Re for DENV4-II are around 2.2 (0.62–4.3), in line with previous estimates35,48,49,50. Also, we found temperature variation to be a key driver of parameters dictating vector capacity, consistent with other studies35,42; estimated means of vector lifespan and EIP were well within that expected for the tropical wet and dry climate of Rio de Janeiro36,37. Both the genetic and time series data captured well seasonal oscillations between January and July 2012. However, estimates of transmission potential (Re) under the two approaches varied considerably before January 2012. Both models are likely to lose predictive power prior to that date due to the absence of genetic or case count data to be fitted.

Genomic surveillance of blood banks is particularly useful for viral epidemics that may spread undetected in immunologically naïve populations. For example, the explosive spread of Zika virus in Brazil went undetected for at least one year56,75 before first being reported in March 201551. Further, there is the possibility that West Nile virus (recently found in humans in Piauí state52), the ECSA genotype of Chikungunya virus spreading in the northeast of Brazil53,54, or the Japanese encephalitis virus currently circulating in South Asia may cause future arboviral epidemics in Brazil. There is a need for continuous pathogen surveillance to protect blood supplies and reduce the risk of transfusion-associated transmission. We expect identical or near identical viruses to be transmitted from the blood donor-recipient pairs. Unfortunately, of the six donor-recipient DENV RNA + sample pairs, we could not obtain sequences for both donor and recipient to inspect whether these were phylogenetically linked.

Although we investigated the epidemic of a DENV4 urban outbreak, we did not evaluate the impact of ecological and demographic factors on the geographic spread of the virus. Moreover, the blood donor/recipient population studied here may be biased in respect to geographic distribution, age, gender or socio-economic status. However, we remain confident that the sample presented here is mostly representative of the population as sero-conversion follows the epidemic curve observed in Rio de Janeiro from surveillance data22,41. More comprehensive spatial and genomic sampling from unsampled regions are needed to unambiguously identify the drivers and hotspots of virus transmission55,56,57. Such quest could be achieved with analyses of large-scale blood banks combined with a real-time genomic surveillance network capable of discriminating between virus species and serotypes58. This would allow effective preventive measures to be designed and implemented efficiently in regions of greatest risk of importation and exportation at particular periods of elevated outbreak risk. Additional work could further evaluate intra-host diversity in symptomatic and asymptomatic patients.

In conclusion, joint epidemiological and genomic surveillance of blood banks as the potential to elucidate the drivers of arboviral transmission and to detect cryptic transmission of newly introduced mosquito-borne viral lineages and species, particularly into regions where generalist Aedes vectors circulate.

Materials and Methods

Patient cohort

The study protocol was approved by the Institutional Ethics Committee (Comissão de Ética em Pesquisa em Seres Humanos da Faculdade de Medicina da Universidade de São Paulo). During Feb-Jun 2012 all donors from the blood centers in Hemório (Rio de Janeiro municipality, Rio de Janeiro state) and Hemope (Recife municipality, Pernambuco state) were asked to participate in the study. An extra Plasma Preparation Tube (PPT; Becton Dickinson, Franklin Lakes, NJ) was collected from each consenting donor and recipient. Recipients who received components from a consenting donor were identified and invited to participate in the study. Once accepted, samples were collected 3–21 days after the index transfusion. All pre- and post-transfusion samples from enrolled recipients (n = 1,144), all donor samples from Rio de Janeiro (n = 15,866) and donor samples from Recife that were linked to enrolled recipients (n = 4,051) were tested by a transcription-mediated amplification (TMA) assay (Hologic/Grifols) that has a 50% detection limit of ~5 RNA copies/mL for each of the four DENV serotypes [19]. The study detected 149 DENV RNA + samples. Of those, 142 samples (119 donors and 23 recipients) had sufficient plasma volume to proceed to viral RNA extraction and whole genome PCR amplification (Table S1). Ethics committees in Brazil and the USA approved this study. The study design is described in detail in a previous study REDS-III22.

Extraction, cDNA synthesis, PCR amplification and sequencing

All donor samples collected from individuals in RJc (n = 15,866), and the samples obtained from donors in Recife that were linked to enrolled recipients (n = 4,051) and 657 recipient samples were tested by a transcription mediated amplification assay (TMA; Hologic/Grifols). Plasma was centrifuged and viral RNA was extracted from supernatant with the Agencourt RNAdvance Tissue Kit (Beckman Coulter, Indianapolis IN, US) according to the manufacturer’s instructions. cDNA was synthesized with SuperScript III Reverse Transcriptase (Invitrogen, Carlsbad, CA, US), according to the manufacturer’s instructions using specific primer decamers (10 bp) complementary to the conserved region of 3’UTR of all dengue serotypes. Degenerate primers (described in Table S2) distributed along the complete genome were designed to amplify five overlapping fragments (2–4kb-long) covering the full genome (F1: 1–3790 bp, F2: 2884–4993 bp, F3: 4830–7330 bp, F4: 5982–8704 bp and F5: 7800–10649 bp, based on a DENV4 reference genome (Genbank NC_002640). PCR products were purified using the QIAquick PCR Purification Kit (QIAGEN, Stockach, Germany), following the manufacturer’s instructions and DNA concentrations were estimated using the Qubit dsDNA HS Assay Kit (Invitrogen, Carlsbad, CA, USA). The amplicons from a single viral genome were pooled at equimolar ratios. Prior to sequencing, amplicon mixtures (200ng) were sonicated to produce 200bp-long fragments using the Bioruptor (Diagenode Inc Denville, NJ, USA). A library was then built using the AB Library Builder System, Ion Xpress Barcode Adapters kit and Ion Plus Library Preparation kit (Invitrogen, Carlsbad, CA, USA), following the manufacturer’s instructions. Clonal amplification was performed using the Ion OneTouch, Ion PGM Template OT2 200 kit (Invitrogen, Carlsbad, CA, USA). Sequencing was performed using the PGM Instrument, the Ion PGM Sequencing 200 Kit v2 and the Ion 318 Chip. Datasets were then trimmed according the quality (99% coverage) and length (reads < 30 bp were removed) of each read using Geneious R7 software (Biomatters). The sequences of primers used for PCR were also removed. Full genome sequences were then reconstituted by mapping the reads to a reference sequence from GenBank using Geneious (majority base at each position)59. Sequences obtained were examined to ensure that the mapping to a reference sequence did not generate a biased consensus sequence.

Dataset collation, sequence quality control and alignment

One hundred and fifty three DENV4 publicly-available and published near full-length genomes (>8000 bp) were retrieved in Feb 2016 using the ACNUC biological database system60. The genome sequences newly generated in this study were added to this global full-length DEN4 dataset. Multiple nucleotide sequence alignment was performed with MAFFTv.761. After manually editing of the alignment62, a maximum likelihood (ML) tree (n = 244) was estimated using RAxML under general time reversible model with gamma-distributed rates ($$GTR+{\rm{\Gamma }}$$) nucleotide substitution model63. TempEst64 was used to inspect the molecular clock signal in the phylogeny and to identify potential contaminants. Three strains (N28351-Recife, N26684-Recife, N15054-Rio de Janeiro) were removed from subsequent analysis as they had an unusually high divergence compared to sequences collected within the same time frame. In addition, two sylvatic strains were removed from subsequent analysis to avoid introducing biases in evolutionary rate estimates. DENV types were classified using the Dengue, Zika and Chikungunya Viruses Typing Tool v.0.9 (http://bioafrica2.mrc.ac.za/rega-genotype-alpha/aedesviruses/typingtool/).

Global molecular clock phylogeny of DENV4

Time-calibrated phylogenies were generated using the Bayesian Markov chain Monte Carlo (MCMC) framework implemented in BEAST v.1.8.4 package65. The BEAGLEv.2 library66 was used to enhance computational speed and a ML tree was used as a starting tree. For computational tractability, the best-fitting nucleotide substitution model was identified using path-sampling and stepping stone model selection (chain length 10 million steps and 100 path steps) using a subset of 40 isolates random chosen from the global dataset. The $$GTR+{\rm{\Gamma }}$$ outperformed both the HKY and the TN93 (not shown). To model changes in the effective population size over time, two non-parametric coalescent models were used, a Bayesian Skygrid model67 with 50 grid points and a Bayesian skyline model with 10 groups68. Markov chain Monte Carlo (MCMC) sampling was performed in duplicate for 250 million steps and convergence of the chains was inspected using Tracer v.1.6. After removing 10% burn-in, log and tree files were combined and a maximum clade credibility (MCC) tree was generated using TreeAnnotator69.

Estimating epidemic spread using a birth-death skyline model

Dated phylogenies were reconstructed for the clade representing the DENV4 outbreak in Brazil (Fig. 2A). As expected, very little temporal signal was obtained using only the sequences belonging to the outbreak clade (root-to-tip regression analysis; r2 = 0.044; slope = 2.52 × 10−3 substitutions per site per year, s/s/y). Therefore, to perform time-calibrated phylogenetic reconstruction, we placed an informative lognormal prior on the TMRCA of the outbreak clade (mean 3.26 years before July 2012 and standard deviation 0.5) as obtained from analyses of the global dataset with strong temporal signal (Fig. 1F inset).

To estimate changes in epidemic spread through time, we used a Bayesian birth-death skyline plot model70 implemented in BEAST271. In this model, each infection may transmit at a rate $$\lambda$$ and becomes noninfectious at a rate $$\delta$$. Upon becoming infected, each individual is sampled with a probability $$s$$ and included in the dataset. The model enables the piecewise estimation of R e (the effective reproduction number of the outbreak), $$\delta$$ and $$s$$ through time. We assumed sampling proportion s to be constant over time. Relaxing this assumption to allow parameter s to be zero for the periods when no sequence data was available resulted in similar trends for R e with wider Bayesian credible intervals (not shown). The rate $$\delta$$ was modelled using a lognormal prior with a mean of 14 days and a standard deviation of 0.5, which roughly corresponds to the sum of the intrinsic and extrinsic incubation period of dengue virus. The BDSKY analysis was run for 3 independent MCMC chains of 20 × 106 steps, with parameters and trees being sampled once every 20000 steps. After removal of 10% burn in, sampled parameters were combined using LogCombiner69.

Ento-epidemiological Susceptible-Exposed-Infected-Recovered model

To model the daily dynamics of DENV4-II infections in Rio de Janeiro, we used a Bayesian MCMC modelling approach based on framework previously used to investigate the first DENV1 outbreak in the Island of Madeira35. A time series for DENV4-II in Brazil was compiled from the number of positive cases from February to July 2012 within the RED-II cohort in Rio de Janeiro (Fig. S3). Weekly minimum temperature data for RJ was obtained from SEOMA/RJ, the Brazilian weather repository. In brief, the approach uses ordinary differential equations (ODE) to describe the dynamics of infection within the human and mosquito populations. The human population is assumed to be fully susceptible before the introduction of DENV4 and constant in size throughout the period of observation. Upon challenge with infectious mosquito bites, individuals enter the incubation phase, later becoming infectious, and finally recovering with life-long immunity (SI, Equations 1 to 5). For the dynamics of mosquito populations (SI, Equations 6 to 10), individuals were divided into two life-stages: aquatic and adult females. The adults are further divided into the epidemiologically relevant stages for DENV transmission: susceptible, incubating, and infectious for life.

Following the approach in35, the ODE model has 8 entomological parameters that are temperature-dependent (aquatic to adult transition rate, aquatic mortality rate, adult mortality rate, oviposition rate, incubation period, transmission probabilities per infectious bite, hatching success rate; Table S3). The MCMC approach estimates four unknown model parameters: the time of first human case, the aquatic carrying capacity, and two linear factors used to scale the effects of temperature on mosquito mortality and incubation period. Five other parameters are not estimated by the MCMC nor considered to be temperature-independent, for which we used literature-based estimates (mosquito biting rate, human incubation period, human infectious period, mosquito sex ratio, donor observation rate; Table S4). The Bayesian MCMC implementation and the ento-epidemiological SEIR model are described in detail in Supplementary Information, including the expressions for the effective and basic reproduction numbers of DENV72,73.

Data availability

New sequence data have been deposited in Genbank under accession numbers (to be included). Genomic position, sequencing statistics, viral load and geographic location of the newly sequenced data is shown in Table S1. Code for the ODE model is available from the authors upon request.

Ethical statement

All participants were informed of the study aims, and written informed consent was received from each patient before sample collection. The study protocol was approved by the Institutional Ethics Committee (Comissão de Ética em Pesquisa em Seres Humanos da Faculdade de Medicina da Universidade de São Paulo) and was in accordance with the Declaration of Helsinki for Human Research of 1974 (last modified in 2000). All experimental protocols were approved by the Institutional Ethics Committee (Comissão de Ética em Pesquisa em Seres Humanos da Faculdade de Medicina da Universidade de São Paulo).

Biosafety

All protocols and procedures were conducted within the enhanced laboratory biosafety level 2 (ABSL-2) facility of Institute for Tropical Medicine, Sao Paulo University. The ABSL-2 facility consists of a laboratory in which all in vitro experimental work is carried out in class 3 biosafety cabinets, which are also negative pressurized (<−200 Pa). Although all experiments are conducted in closed class 3 cabinets and isolators, special personal protective equipment, including laboratory suits, gloves and FFP3 facemasks is used. Air released from the class 3 units is filtered by High Efficiency Particulate Air (HEPA) filters and then leaves via the facility ventilation system, again via HEPA filters. Only authorized personnel that have received the appropriate training can access the facility. The facility is secured by procedures recognized as appropriate by the institutional biosafety officers and facility management at Sao Paulo University and Brazilian National Technical Biosafety Commission (CTNBio).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. 1.

Bhatt, S. et al. The global distribution and burden of dengue. Nature 496, 504–507, https://doi.org/10.1038/nature12060 (2013).

2. 2.

Holmes, E. C. & Twiddy, S. S. The origin, emergence and evolutionary genetics of dengue virus. Infect Genet Evol 3, 19–28 (2003).

3. 3.

Twiddy, S. S., Holmes, E. C. & Rambaut, A. Inferring the rate and time-scale of dengue virus evolution. Mol Biol Evol 20, 122–129 (2003).

4. 4.

Messina, J. P. et al. Global spread of dengue virus types: mapping the 70 year history. Trends Microbiol 22, 138–146, https://doi.org/10.1016/j.tim.2013.12.011 (2014).

5. 5.

Gubler, D. J. D. Urbanization and Globalization: The Unholy Trinity of the 21(st) Century. Trop Med Health 39, 3–11, https://doi.org/10.2149/tmh.2011-S05 (2011).

6. 6.

Weaver, S. C. Urbanization and geographic expansion of zoonotic arboviral diseases: mechanisms and potential strategies for prevention. Trends Microbiol 21, 360–363, https://doi.org/10.1016/j.tim.2013.03.003 (2013).

7. 7.

WHO. World Health Organization Dengue and severe dengue., (who.int/mediacentre/factsheets/fs117/en/, 2014).

8. 8.

Saúde, S. d. V. e. S. M. d. Boletim Epidemiológico (http://portalsaude.saude.gov.br/). http://u.saude.gov.br/images/pdf/2015/dezembro/11/svs-be-2015-047-dengue-se47-final.pdf 46 (2016).

9. 9.

Silva, M. M. et al. Accuracy of Dengue Reporting by National Surveillance System, Brazil. Emerg Infect Dis 22, 336–339, https://doi.org/10.3201/eid2202.150495 (2016).

10. 10.

Rodriguez-Barraquer, I. et al. From re-emergence to hyperendemicity: the natural history of the dengue epidemic in Brazil. PLoS Negl Trop Dis 5, e935, https://doi.org/10.1371/journal.pntd.0000935 (2011).

11. 11.

Teixeira, M. G., Siqueira, J. B. Jr., Ferreira, G. L. & Bricks, L. & Joint, G. Epidemiological trends of dengue disease in Brazil (2000-2010): a systematic literature search and analysis. PLoS Negl Trop Dis 7, e2520, https://doi.org/10.1371/journal.pntd.0002520 (2013).

12. 12.

Nunes, M. R. et al. Phylogeography of dengue virus serotype 4, Brazil, 2010-2011. Emerg Infect Dis 18, 1858–1864, https://doi.org/10.3201/eid1811.120217 (2012).

13. 13.

Kraemer, M. U. et al. The global compendium of Aedes aegypti and Ae. albopictus occurrence. Scientific data 2, 150035, https://doi.org/10.1038/sdata.2015.35 (2015).

14. 14.

Kraemer, M. U. et al. The global distribution of the arbovirus vectors Aedes aegypti and Ae. albopictus. Elife 4, e08347, https://doi.org/10.7554/eLife.08347 (2015).

15. 15.

Villabona-Arenas, C. J. et al. Epidemiological dynamics of an urban Dengue 4 outbreak in Sao Paulo, Brazil. PeerJ 4, e1892, https://doi.org/10.7717/peerj.1892 (2016).

16. 16.

Andrade, E. H. et al. Spatial-Temporal Co-Circulation of Dengue Virus 1, 2, 3, and 4 Associated with Coinfection Cases in a Hyperendemic Area of Brazil: A 4-Week Survey. Am J Trop Med Hyg 94, 1080–1084, https://doi.org/10.4269/ajtmh.15-0892 (2016).

17. 17.

Teixeira, M. G., Costa Mda, C., Barreto, F. & Barreto, M. L. Dengue: twenty-five years since reemergence in Brazil. Cad Saude Publica 25(Suppl 1), S7–18 (2009).

18. 18.

Villabona-Arenas, C. J. et al. Detection of four dengue serotypes suggests rise in hyperendemicity in urban centers of Brazil. PLoS Negl Trop Dis 8, e2620, https://doi.org/10.1371/journal.pntd.0002620 (2014).

19. 19.

Pybus, O. G. et al. Unifying the spatial epidemiology and molecular evolution of emerging epidemics. Proc Natl Acad Sci USA 109, 15066–15071, https://doi.org/10.1073/pnas.1206598109 (2012).

20. 20.

Busch, M. P. et al. Screening the blood supply for West Nile virus RNA by nucleic acid amplification testing. N Engl J Med 353, 460–467, https://doi.org/10.1056/NEJMoa044029 (2005).

21. 21.

Musso, D. et al. Potential for Zika virus transmission through blood transfusion demonstrated during an outbreak in French Polynesia, November 2013 to February 2014. Euro Surveill 19 (2014).

22. 22.

Sabino, E. C. et al. Transfusion-Transmitted Dengue and Associated Clinical Symptoms During the 2012 Epidemic in Brazil. J Infect Dis 213, 694–702, https://doi.org/10.1093/infdis/jiv326 (2016).

23. 23.

Tambyah, P. A. et al. Dengue hemorrhagic fever transmitted by blood transfusion. N Engl J Med 359, 1526–1527, https://doi.org/10.1056/NEJMc0708673 (2008).

24. 24.

Acosta, P. O. et al. Dengue virus serotype 4, Roraima State, Brazil. Emerg Infect Dis 17, 1979–1980, https://doi.org/10.3201/eid1710.110776 (2011). author’s reply 1980–1971.

25. 25.

Lourenco-de-Oliveira, R., Vazeille, M., de Filippis, A. M. & Failloux, A. B. Aedes aegypti in Brazil: genetically differentiated populations with high susceptibility to dengue and yellow fever viruses. Trans R Soc Trop Med Hyg 98, 43–54 (2004).

26. 26.

Schatzmayr, H. G. & Nogueira, R. M. R. Travassos da Rosa, A. P. A. An outbreak of dengue virus at Rio de Janeiro - 1986. Memórias do Instituto Oswaldo Cruz 81, 245–246 (1986).

27. 27.

Nogueira, R. M. R. et al. Dengue epidemic in the state of Rio de Janeiro, Brazil, 1990–1991. Epidemiol. Infect 111, 163–170 (1993).

28. 28.

Nogueira, R. M. R., Miagostovich, M. P., Filippis, A. M. B., Pereira, M. A. S. & Schatzmayer, H. G. Dengue virus type 3 in Rio de Janeiro, Brazil. Mem Inst Oswaldo Cruz 96, 925–926 (2001).

29. 29.

Sorichetta, A. et al. High-resolution gridded population datasets for Latin America and the Caribbean in 2010, 2015, and 2020. Scientific data 2, 150045, https://doi.org/10.1038/sdata.2015.45 (2015).

30. 30.

Nunes, M. R. et al. Air travel is associated with intracontinental spread of dengue virus serotypes 1-3 in Brazil. PLoS Negl Trop Dis 8, e2769, https://doi.org/10.1371/journal.pntd.0002769 (2014).

31. 31.

Raghwani, J. et al. Endemic dengue associated with the co-circulation of multiple viral lineages and localized density-dependent transmission. PLoS Pathog 7, e1002064, https://doi.org/10.1371/journal.ppat.1002064 (2011).

32. 32.

Rocco, I. M. et al. First isolation of Dengue 4 in the state of Sao Paulo, Brazil, 2011. Rev Inst Med Trop Sao Paulo 54, 49–51 (2012).

33. 33.

Nogueira, R. M. R. & Eppinghays, A. L. Dengue virus type 4 arrives in the state of Rio de Janeiro: a challenge for epidemiological surveillance and control. Mem Inst Oswaldo Cruz 106, 255–256 (2011).

34. 34.

Stadler, T. et al. Estimating the basic reproductive number from viral sequence data. Mol Biol Evol 29, 347–357, https://doi.org/10.1093/molbev/msr217 (2012).

35. 35.

Lourenco, J. & Recker, M. The 2012 Madeira dengue outbreak: epidemiological determinants and future epidemic potential. PLoS Negl Trop Dis 8, e3083, https://doi.org/10.1371/journal.pntd.0003083 (2014).

36. 36.

Maciel-de-Freitas, R., Codeco, C. T. & Lourenco-de-Oliveira, R. Daily survival rates and dispersal of Aedes aegypti females in Rio de Janeiro, Brazil. Am J Trop Med Hyg 76, 659–665 (2007).

37. 37.

Brady, O. J. et al. Modelling adult Aedes aegypti and Aedes albopictus survival at different temperatures in laboratory and field settings. Parasit Vectors 6, 351, https://doi.org/10.1186/1756-3305-6-351 (2013).

38. 38.

Cardoso, C. W. et al. Outbreak of Exanthematous Illness Associated with Zika, Chikungunya, and Dengue Viruses, Salvador, Brazil. Emerg Infect Dis 21, 2274–2276, https://doi.org/10.3201/eid2112.151167 (2015).

39. 39.

Faria, N. R. et al. Epidemiology of Chikungunya Virus in Bahia, Brazil, 2014–2015. PLoS Currents Outbreaks 1, doi:https://doi.org/10.1371/currents.outbreaks.c97507e3e48efb946401755d468c28b2. (2016).

40. 40.

Pessoa, R. et al. Investigation Into an Outbreak of Dengue-like Illness in Pernambuco, Brazil, Revealed a Cocirculation of Zika, Chikungunya, and Dengue Virus Type 1. Medicine (Baltimore) 95, e3201, https://doi.org/10.1097/MD.0000000000003201 (2016).

41. 41.

Busch, M. P. et al. Duration of Dengue Viremia in Blood Donors and Relationships Between Donor Viremia, Infection Incidence and Clinical Case Reports During a Large Epidemic. J Infect Dis 214, 49–54, https://doi.org/10.1093/infdis/jiw122 (2016).

42. 42.

Brady, O. J. et al. Global temperature constraints on Aedes aegypti and Ae. albopictus persistence and competence for dengue virus transmission. Parasit Vectors 7, 338, https://doi.org/10.1186/1756-3305-7-338 (2014).

43. 43.

Degener, C. M. et al. Temporal abundance of Aedes aegypti in Manaus, Brazil, measured by two trap types for adult mosquitoes. Mem Inst Oswaldo Cruz 109, 1030–1040, https://doi.org/10.1590/0074-0276140234 (2014).

44. 44.

Codeco, C. T. et al. Seasonal dynamics of Aedes aegypti (Diptera: Culicidae) in the northernmost state of Brazil: a likely port-of-entry for dengue virus 4. Mem Inst Oswaldo Cruz 104, 614–620 (2009).

45. 45.

Tian, H. et al. Increasing airline travel may facilitate co-circulation of multiple dengue virus serotypes in Asia. PLoS Negl Trop Dis 11, e0005694, https://doi.org/10.1371/journal.pntd.0005694 (2017).

46. 46.

Rasmussen, D. A., Boni, M. F. & Koelle, K. Reconciling phylodynamics with epidemiology: the case of dengue virus in southern Vietnam. Mol Biol Evol 31, 258–271, https://doi.org/10.1093/molbev/mst203 (2014).

47. 47.

de Melo, F. L., Romano, C. M. & Zanotto, P. M. A. Introduction of Dengue Virus 4 (DENV-4) Genotype I into Brazil from Asia? PLoS Negl Trop Dis 3, e390 (2009).

48. 48.

Marques, C. A., Forattini, O. P. & Massad, E. The basic reproduction number for dengue fever in Sao Paulo state, Brazil: 1990–1991 epidemic. Trans R Soc Trop Med Hyg 88, 58–59 (1994).

49. 49.

Favier, C. et al. Early determination of the reproductive number for vector-borne diseases: the case of dengue in Brazil. Trop Med Int Health 11, 332–340, https://doi.org/10.1111/j.1365-3156.2006.01560.x (2006).

50. 50.

Coelho, G. E., Burattini, M. N., Teixeira Mda, G., Coutinho, F. A. & Massad, E. Dynamics of the 2006/2007 dengue outbreak in Brazil. Mem Inst Oswaldo Cruz 103, 535–539 (2008).

51. 51.

Zanluca, C. et al. First report of autochthonous transmission of Zika virus in Brazil. Mem Inst Oswaldo Cruz 110, 569–572, https://doi.org/10.1590/0074-02760150192 (2015).

52. 52.

Vieira, M. A. et al. West Nile Virus Encephalitis: The First Human Case Recorded in Brazil. Am J Trop Med Hyg 93, 377–379, https://doi.org/10.4269/ajtmh.15-0170 (2015).

53. 53.

da Costa, A. C. et al. Spread of Chikungunya Virus East/Central/South African Genotype in Northeast Brazil. Emerg Infect Dis 23 (2017).

54. 54.

Nunes, M. R. et al. Emergence and potential for spread of Chikungunya virus in Brazil. BMC Med 13, 102, https://doi.org/10.1186/s12916-015-0348-x (2015).

55. 55.

Dudas, G. et al. Virus genomes reveal factors that spread and sustained the Ebola epidemic. Nature 544, 309–315, https://doi.org/10.1038/nature22040 (2017).

56. 56.

Faria, N. R. et al. Establishment and cryptic transmission of Zika virus in Brazil and the Americas. Nature 546, 406–410, https://doi.org/10.1038/nature22401 (2017).

57. 57.

Kraemer, M. U. et al. Spread of yellow fever virus outbreak in Angola and the Democratic Republic of the Congo 2015–16: a modelling study. Lancet 17(3) 330–338 (2017).

58. 58.

Quick, J. et al. Multiplex PCR method for MinION and Illumina sequencing of Zika and other virus genomes directly from clinical samples. Nat Protoc 12, 1261–1276, https://doi.org/10.1038/nprot.2017.066 (2017).

59. 59.

Kearse, M. et al. Geneious Basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics 28, 1647–1649, https://doi.org/10.1093/bioinformatics/bts199 (2012).

60. 60.

Gouy, M. & Demotte, S. Remote access to ACNUC nucleotide and protein sequence databases at PBIL. Biochimie 90, 555–562 (2008).

61. 61.

Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol 30, 772–780, https://doi.org/10.1093/molbev/mst010 (2013).

62. 62.

Larsson, A. AliView: a fast and lightweight alignment viewer and editor for large datasets. Bioinformatics 30, 3276–3278, https://doi.org/10.1093/bioinformatics/btu531 (2014).

63. 63.

Stamatakis, A. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313, https://doi.org/10.1093/bioinformatics/btu033 (2014).

64. 64.

Rambaut, A. Path-O-Gen v1.4, http://tree.bio.ed.ac.uk/software/pathogen/ (2014).

65. 65.

Drummond, A. J., Suchard, M. A., Xie, D. & Rambaut, A. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol Biol Evol 29, 1969–1973, https://doi.org/10.1093/molbev/mss075 (2012).

66. 66.

Ayres, D. L. et al. BEAGLE: an application programming interface and high-performance computing library for statistical phylogenetics. Systematic biology 61, 170–173, https://doi.org/10.1093/sysbio/syr100 (2012).

67. 67.

Gill, M. S. et al. Improving Bayesian population dynamics inference: a coalescent-based model for multiple loci. Mol Biol Evol 30, 713–724, https://doi.org/10.1093/molbev/mss265 (2013).

68. 68.

Drummond, A. J., Rambaut, A., Shapiro, B. & Pybus, O. G. Bayesian coalescent inference of past population dynamics from molecular sequences. Mol Biol Evol 22, 1185–1192, https://doi.org/10.1093/molbev/msi103 (2005).

69. 69.

FigTree v1.4.2 (2015).

70. 70.

Stadler, T., Kuhnert, D., Bonhoeffer, S. & Drummond, A. J. Birth-death skyline plot reveals temporal changes of epidemic spread in HIV and hepatitis C virus (HCV). Proc Natl Acad Sci USA 110, 228–233, https://doi.org/10.1073/pnas.1207965110 (2013).

71. 71.

Bouckaert, R. et al. BEAST 2: a software platform for Bayesian evolutionary analysis. PLoS computational biology 10, e1003537, https://doi.org/10.1371/journal.pcbi.1003537 (2014).

72. 72.

Wearing, H. J. & Rohani, P. Ecological and immunological determinants of dengue epidemics. Proc Natl Acad Sci USA 103, 11802–11807, https://doi.org/10.1073/pnas.0602960103 (2006).

73. 73.

Lourenco, J. & Recker, M. Natural, persistent oscillations in a spatial multi-strain disease system with application to dengue. PLoS computational biology 9, e1003308, https://doi.org/10.1371/journal.pcbi.1003308 (2013).

74. 74.

Janeiro, S. d. E. d. S. d. R. d. Boletim Epidemiologico 07/2015: Dengue Cenario Epidemiologico e Perspectiva para o Ano de 2016. (2015).

75. 75.

Samia N. Naccache, et al. Distinct Zika Virus Lineage in Salvador, Bahia, Brazil. Emerging Infectious Diseases 22 (10):1788–1792 (2016).

Acknowledgements

N.R.F. is funded by a Sir Henry Dale Fellowship (grant 204311/Z/16/Z). ACC is funded by FAPESP #2012/03417-8, 2014/03417-7 and 2017/00021-9. JL received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no. 268904 – DIVERSITY. MUGK is supported by the Society in Science, The Branco Weiss Fellowship, administered by the ETH Zurich and acknowledges funding from a Training Grant from the National Institute of Child Health and Human Development (T32HD040128) and the National Library of Medicine of the National Institutes of Health (R01LM010812, R01LM011965). We thank Shirley Cavalcante Vasconcelos Komninakis (UNIFESP), Luciano Monteiro da Silva and Lucy Santos Vilas Boas, (IMT-USP) and Paulo Brito, director from SEOMA/RJ, for daily meteorological data from Rio de Janeiro municipality. This work was funded by FAPESP # 2016/01735-2 and 2011/18955-1, and the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/European Research Council grant agreement number 614725-PATHPHYLODYN.

Author information

Author notes

1. Nuno R. Faria, Antonio Charlys da Costa and Jose Lourenço contributed equally to this work.

2. *A comprehensive list of consortium members appears at the end of the paper.

Affiliations

1. Department of Zoology, University of Oxford, Oxford, United Kingdom

• Nuno R. Faria
• , José Lourenço
• , Moritz U. G. Kraemer
• , Julien Thézé
•  & Oliver G. Pybus
2. Instituto de Medicina Tropical, Universidade de São Paulo, São Paulo, Brazil

• Antonio Charlys da Costa
• , Roberto Ribeiro
• , Camila Romano
• , Ester C. Sabino
•  & Ligia Capuani
3. LIM46, Departamento de Moléstias Infecciosas e Parasitárias, Faculdade de Medicina, Universidade de São Paulo, São Paulo, Brazil

• Antonio Charlys da Costa
• , Roberto Ribeiro
• , Ester C. Sabino
• , João Eduardo Ferreira
• , Marcio Oikawa
•  & Pedro Losco Takecian
4. Faculdade de Ciências Médicas, Fundação Hemope, Recife, Brazil

• Paula Loureiro
• , Divaldo de Almeida Sampaio
•  & Clarisse Lobo
5. Fundação Hemório, Rio de Janeiro, Brazil

• Maria Esther Lopes
•  & Anna Bárbara Carneiro-Proietti
6. Laboratório de Investigação Médica LIM03, Hospital das Clínicas, Sao Paulo, Brazil

• Cecilia Salete Alencar
7. Institut de Recherche pour le Développement, Université de Montpellier, Montpellier, France

• Christian J. Villabona-Arenas
8. Department of Statistics, University of Oxford, Oxford, United Kingdom

• Chieh-Hsi Wu
9. Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Toronto, Ontario, Canada

• Kamran Khan
•  & Shannon E. Brent
10. Division of Infectious Diseases, University of Toronto, Toronto, Ontario, Canada

• Kamran Khan
11. Blood Systems Research Institute, San Francisco, California, USA

• Eric Delwart
• , Brian Custer
• , Michael P. Busch
• , Shannon Kelly
•  & Thelma T. Gonçalez
12. University of California San Francisco, San Francisco, California, USA

• Eric Delwart
• , Brian Custer
•  & Michael P. Busch
13. Fundação Pró-Sangue Hemocentro de São Paulo, São Paulo, Brazil

• Cesar de Almeida Neto
14. Fundação Hemominas, Belo Horizonte, Minas Gerais, Brazil

• Alfredo Mendrone-Junior
15. Universidade Universidade Federal de São João del-Rei, São João del-Rei, Minas Gerais, Brazil

• Cláudia Di Lorenzo Oliveira
16. RTI international, Rockville, Maryland, USA

• Donald Brambilla
•  & Christopher McClure
17. National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland, USA

• Simone A. Glynn

Contributions

Manuscript preparation: N.R.F., A.C.d.C., J.L., M.U.G.K., J.T. Performed the experiments: A.C.d.C. Technical assistance: R.R., C.S.A., C.M.R. Conceived the study: E.D., B.C., M.P.B., O.G.P., E.C.S., N.R.F., A.C.d.C. Performed the analyses: N.R.F., A.C.d.C., M.U.G.K., C.J.V.A., J.T., C.H.W., K.K., S.E.B., C.M.R. Obtained funding and study supervision: E.D., B.C., M.P.B., O.G.P., E.C.S. Providing blood samples and clinical data: P.L., M.E.L., E.C.S. All authors reviewed, contributed to, and approved the final version of the manuscript.

Competing Interests

The authors declare that they have no competing interests.

Corresponding authors

Correspondence to Nuno R. Faria or Antonio Charlys da Costa or Ester C. Sabino.

Electronic supplementary material

DOI

https://doi.org/10.1038/s41598-017-15152-8