European Commission logo
English English
CORDIS - EU research results
CORDIS
Content archived on 2024-05-27

Dengue research Framework for Resisting Epidemics in Europe

Final Report Summary - DENFREE (Dengue research Framework for Resisting Epidemics in Europe)

Executive Summary:
DENFREE combines expertise from different disciplines. The overall objective is to understand a role of asymptomatic dengue infection in dengue transmission and their virological and immunological profile.
To understand better dengue transmission dynamics and their relation to climate change, we generated a calibrated climate dataset, performed database inter-comparison for the different climate product variables and created a user-friendly tutorial. An online near-operational skill finder tool has been developed to assess where and in which seasons climate information is going to be most useful for disease prediction purposes. We established a database with climate information for improving Public Health decision-making. We studied the spatial-temporal pattern of propagation of annual epidemics in Cambodia, which highlighted the role that human movement plays. We have implemented a simplified ABM/metapopulation mixed model where a single strain vector borne infection travelling through a network of cities and villages is deployed. At a larger geographical scale, we generated a database of dengue cases in French Polynesia and demonstrated that maintenance of dengue can only occur by island hopping. In addition, we demonstrated that not only asymptomatic people are infectious to mosquitoes, but also are significantly more infectious to mosquitoes than symptomatic dengue at a given viremia level.
Concerning basic biology of interaction among DENV, host and vector, we showed that viral dissemination was similar between the viruses from France or Thailand but the differences were associated with mosquito populations. We experimentally enhanced transmission potential of DENV for European Ae. albopictus. When the viruses were passaged in Ae. albopictus mosquitoes and cell culture alternatively, one variant reached consensus level within two passages and became almost fixed soon after, consistent with a strong selective sweep.
We demonstrated that greater variability occurred in the non-structural genes compared with structural genes, in terms of significant minority variants, and particularly towards the end of the epidemic. It was noteworthy that patients with secondary infection showed greater variability than patients with primary infections. Our results suggest that changes in the viral population swarm occurred with the epidemic’s progression, and these changes could have an impact on viral fitness.
We identified and structurally characterized new type of antibodies, which cross-neutralize against all four dengue serotypes. In addition, we identified T cells were highly activated in asymptomatic viremic individuals while plasma blasts were increasingly developed in dengue patients. These findings provide a new concept for dengue vaccine development. Genome wide study of dengue patients from Cuba and Thailand identified genes in lipid metabolism and xenobiotic signaling pathways involved in pathogenesis of classical dengue. We calculated the genetic risk using these genetic markers and showed that European populations present an even higher risk than Southeast Asian populations for severe dengue, while they are the best protected ones against classical dengue.
For dengue preventive intervention study, we explored the effect of vaccine intervention through our dynamical modelling approach. Such methods can be used to evaluate optimal controls via vaccination and mosquito control where both methods alone have their limitations. In addition, we demonstrated that the auto-dissemination of the insect growth hormone, pyriproxyfen led to ~20% higher larval mortality rates in sentinel sites but an even greater reduction in adult densities. Finally, we developed a new point-of-care diagnostic test based on reverse transcription-recombinase polymerase amplification to detect DENV genome. We have developed DNA-tagged Dengue antigen for antibody detection method, which allows a highly sensitive test for DENV antibodies and will be combined with a virus type specific amplification method in the future.
Project Context and Objectives:

Project context
Dengue is a major international public health concern and one of the most important arthropodborne diseases. Approximately 55% of world’s population are at risk of dengue virus (DENV) infection.
The current state-of-the-art strategy to manage the burden of dengue in countries with endemic DENV has focused on improving clinical care of hospitalized severe cases. Whilst this has reduced the case fatality rate, alone it is insufficient in the face of an ever-increasing population at risk. Mosquito control, although successful in some cases, has largely failed to make an impact over the long term. Epidemiological studies have shown that most DENV infections are asymptomatic or subclinical and thus in a completely naïve population, the first hospital cases will be the tip of the iceberg. Hospital-based surveillance will be inadequate – too little and too late.
Currently we know very little about the epidemiology of subclinical infections. How frequent are they? What factors determine the symptomatic/asymptomatic outcome of infection? Can they infect mosquitoes? Is the duration of infection different from symptomatic episodes? The role of individuals with asymptomatic infection in spreading the virus must be addressed. Such epidemiological information is crucial to predict local dengue epidemiology and determine whether intervention strategies, such as anti-viral treatment targeting asymptomatics, are useful.
DENFREE combines expertise from different disciplines. Each participant generates expert data using samples from the same individuals with well-characterized clinical and epidemiological data. This multidisciplinary and complementary approach will allow each participant to access to data in other relevant fields and analyze them together. From the knowledge acquired, we can make serious efforts at reducing the impact of dengue in endemic areas and at preventing its extension into new areas.
Project objectives
The overall objective of DENFREE program is to understand a role of asymptomatic dengue infection in dengue transmission and their virological and immunological profile. We plan to estimate risk of having dengue outbreaks in Europe. We documented climate prediction models, European vector competence, viral genetic markers for adaptation to European mosquitoes and population genetic susceptibility to clinical and severe dengue. In addition, we developed new tools for point of care diagnostic test for detecting asymptomatic DENV infected individuals, tested new potential therapeutic agents, identified new vaccine candidates and evaluated innovative vector control strategies suitable for European countries.

Description of the work performed since the beginning of the project and the main results achieved so far
Both field study sites successfully accomplished their missions of generating biological samples within defined epidemiological conditions for use by the other studies.
Diagnostic test:
We have published an isothermal amplification method – reverse transcription- Recombinase Polymerase Amplification to detect DENV genome in clinical specimens. During this period, we have developed DNA-tagged Dengue antigen for antibody detection method. The road-block of DNA-antigen cross-linking has been overcome and antigens have been DNA-coded. Amplified products are now detected sensitively by fluorescence. This will allow a highly sensitive test for DENV antibodies and will be combined with a virus type specific amplification method in the future.
Calibrated climate dataset:
WP3 has generated a calibrated climate dataset that can extensively be used by other WPs and projects, performed database inter-comparison for the different climate product variables and created a user-friendly tutorial (see Climate Explorer tutorial posted in the DENFREE website). The tutorial intends to illustrate how even a naive user can visualize and download different types of climate observations and simulations for research purposes. Similarly, an online near-operational skill finder tool has been developed that can be used to assess where and in which seasons climate information is going to be most useful for disease prediction purposes.
Descriptive and Predictive models:
In WP4 we developed models and improved and extended already existing models including include secondary infection, temporary cross-immunity, seasonality, climatic variables and vector dynamics. We then explored the effect of vaccine intervention through our dynamical modelling approach. Because of the non-protective effect of the vaccine in naïve individuals, we recommend a screening on sero-status before administering the vaccine. Using optimal control methods we showed that such methods can now be used to evaluate optimal controls via vaccination and mosquito control where both methods alone have their limitations.
The spatial-temporal pattern of propagation of annual epidemics in Cambodia highlighted the role that human movement plays. Particularly we revealed a great spatial heterogeneity in the propagation of the annual epidemic. Each year, epidemics are highly synchronous over a large geographic area along the busiest national road of the country whereas travelling waves emanate from few rural areas and move slowly along the Mekong River towards the capital, Phnom Penh. Then, we have implemented a simplified ABM/metapopulation mixed model where a single strain vector borne infection travelling through a network of cities and villages is deployed. A major finding is that a complete stationary arrest of symptomatics has little impact on the peak of disease incidence at virtually every ratio of symptomatic-asymptomatic infected individuals in the population.
At a larger geographical scale, we generated a database of dengue cases in French Polynesia and demonstrated that maintenance of dengue can only occur by island hopping; even the large island of Tahiti is unable to maintain the virus for extended periods and thus the virus persists in this island complex by hopping from one island to the next.
Finally we established a database with climate information for improving Public Health decision making and validated this with dengue data from southern coastal Ecuador. Oceanic Niño Index (ONI), rainfall and temperature were positively associated with dengue and the ONI was identified as the most important climate variable. Our structural time series model is specifically tailored to forecast at long lead times of 24 months or more, going well beyond the traditional "spring barrier" of ENSO prediction, thus generating more lead time for prediction and preparedness.
Asymptomatic transmission, European vector competence and innovative vector control strategy
Most of dengue human cases are asymptomatic or result in too few symptoms to be detected by existing surveillance systems. These clinically inapparent cases were not considered to develop a high enough viremia to infect mosquitoes and therefore were assumed to be dead-ends for virus transmission. We demonstrated that not only asymptomatic people are infectious to mosquitoes, but also that they are significantly more infectious to mosquitoes than people with symptomatic infections at a given viremia level (Duong et al. 2015 PNAS).
In Europe, the Asian tiger mosquito has been introduced in 1990 and is now present in 20 European countries and was responsible for local dengue cases in France, Croatia, and Madeira. Owing to the increasing number of autochthonous dengue cases, we evaluated the ability of European populations of Ae. albopictus to transmit imported strains of DENV. We showed that (i) viral dissemination increased with the dpi, (ii) viral dissemination was similar whatever the DENV-1 (France or Thailand), and (iii) differences were associated with the mosquito population.
Aedes albopictus is not considered as a main vector of DENV. However owing to the recurrent detections of local dengue cases in Europe, the question of its long-term adaption for a better transmission of DENV has been addressed. We have designed a protocol to experimentally enhance transmission potential of DENV for European Ae. albopictus. When the viruses were passaged in Ae. albopictus mosquitoes and cell culture alternatively, some new variants reached consensus level (frequency > 50%). The variant reached consensus level within two passages following initial detection and became almost fixed (frequency > 99%) soon after, consistent with a strong selective sweep.
The auto-dissemination of the insect growth hormone, pyriproxyfen (PPF) was tested in Madeira Island, through a small pilot study in the village of Paul do Mar and Funchal. PPF implementation led to ~20% higher larval mortality rates in sentinel sites but an even greater reduction in adult densities.
Viral genetic diversity and adaptation
Concerning viral population study using DENV-3 samples collected at different time points during the 2001-2002 Cuban epidemic, it was demonstrated that greater variability occurred in the non-structural genes compared with structural genes, in terms of significant minority variants, and particularly towards the end of the epidemic. Still greater variability was observed in the non-structural genes compared with structural genes; but it was noteworthy that patients with secondary infection showed greater variability than patients with primary infections. In addition, secondary patients presented minority variants in the structural genes (PrM and E), some of which were non-synonymous. By contrast, patients suffering primary infections only had mostly synonymous minority variants (>1%) in non-structural genes. Our results suggest that changes in the viral population swarm occurred with the epidemic’s progression, and that these changes could have an impact on viral fitness. Therefore, the dynamics of evolving viral populations in the context of heterotypic antibodies could be related to the increasing clinical severity observed during dengue epidemics.
Key immunological factors determining DENV clinical outcome
We have published two important articles showing cross-neutralization against all four dengue serotypes. The findings could lead to new type of dengue vaccine with one antigen against all four dengue serotypes. In addition, these antigens can be used as biomarkers for protective immunity against dengue. We used the same approach to screen our human monoclonal antibodies collection. We identified both cross neutralizing and enhancement antibodies. In addition, in collaboration with partners in WP8, we identified that T cells were highly activated in asymptomatic viremic individuals while plasma blasts were increasingly developed in dengue patients. These findings provide a new concept for dengue vaccine development.
Role of antigen presenting cells, T cells and metabolic pathway in DENV infection outcome; Risk of dengue in European population
The whole genome expression analysis suggested higher antigen presentation activity and activation of T cells in asymptomatic viremic individuals. It is interesting to note that while several immune systems are higher activated in asymptomatic DENV infected individuals, the pathways, which controlled over-reactive immunity were also highly activated. This finding suggested a balance in immune activation is important in asymptomatic infection and support over reactive immunity underlying pathogenesis of dengue diseases.
Genome wide study of dengue patients from Cuba and Thailand identified new interesting distinct pathways involved in pathogenesis of classical dengue and severe dengue. While confirming a role the phospholipase C pathway in severe dengue as previously reported in the Vietnamese genome wide study, we identified genes in xenobiotic signaling pathway involved in pathogenesis of classical dengue. We calculated the genetic risk using the newly and confirmed susceptible/resistant haplotypes. We showed that European populations (as well as South Asian and USA) present an even higher risk than Southeast Asian populations for severe dengue, while they are the best protected ones against classical dengue.

The expected final results and their potential impact and use
Major results achieved:
Dengue transmission dynamics and dengue control strategies
We generated several results that have potential impact on vector surveillance and dengue control strategies including
1) Individuals who are infected but either pre-symptomatic or asymptomatic are just as infectious to mosquitoes as individuals who have symptoms.
2) A significant proportion of the force of infection of dengue was not occurring in and around the house of an infected individual.
3) Development of an online climate skill interactive tool
4) Provision of all climate information and datasets for the epidemiological modeling.
5) Development of a three module virtual environment simulation system.
6) Development, implementation and validation of an Optimal Control Theory
7) Successful implementation of innovative mosquito control strategy
8) Point of care DENV diagnostic test
Host, viral and vector interactions
We identified several new human and viral factors, which could protect and enhance dengue disease severity and dengue transmission.
9) The demonstration that European Aedes albopictus mosquitoes are as competent for Thai and imported dengue viruses
10) Artificial selection revealed the rapid spread of a single nucleotide variant.
11) Identification and structural characterization of highly potent broadly neutralizing anti-DENV antibodies
12) Highly activated T cell response in asymptomatic DENV infection
13) Cross-enhancement and neutralization of anti-DENV antibodies against Zika virus (ZIKV)
We emphasize the importance of community involvement in research and control of dengue especially in the light of asymptomatic DENV infections. Our findings on asymptomatic DENV infections change the current paradigm of dengue transmission and vaccine development, which hitherto focused on hospitalized dengue patients. Our results have significant impact on future dengue research activities, dengue surveillance and control.

Project Results:
- WP1 Index case community study of dengue epidemiology
As detailed below, WP1 involves two field study sites that were responsible for conducting field studies in human populations with dengue for generating samples which the other workpackages could use. Both sites successfully accomplished their missions of generating biological samples within defined epidemiological conditions for use by the other Workpackages. In Cambodia index cases were recruited in the villages of the prospective community study and in three referral district hospitals. When index cases were identified who originated from the study area, serosurveys and entomological studies were conducted in the index case’s home and the 20 closest neighbouring homes or in the households located within a 100 meter-radius. In the community setting, a prospective study was conducted from 24 June to 31 October 2012. This part of the study involved 4 nurses from Kampong Cham hospital and 22 village volunteers to form the investigation team.In the hospital setting, in 2012 cases were recruited in three hospitals of Kampong Cham province: Kompong Cham Province Referral hospital, Prey Chor Operational District (OPD) Hospital and Tbong Khmom OPD Hospital. Consecutive testing of admitted pediatric patients for suspected dengue provided a cohort of 167 symptomatic dengue patients originating from villages located in the study area. In the household serosurvey, a total of 98 serosurveys were conducted in 45 villages.

To conduct these serosurveys, a field team of 6 interviewers aided by 44 nurses from 23 commune health centers was assembled. The number of interviewers was increased to 16 during the peak transmission weeks to 20, 55, and 25 respectively for a total of 85 investigator-months as of 31 October 2012. The teams were coordinated by staff at the Epidemiology and Public Health unit and the Virology units of the Institut Pasteur du Cambodge. During the day, these teams were based in a rented house. The furthest household investigated was ~40 km from the base. Depending on the visit and subject status, types of specimen collected for testing include venous blood, dry blood spot by finger prick, saliva and urines. The number of home visits varies from 3 visits for either negative or positive subjects to 8 visits per asymptomatic dengue subject. As of 31 October 2012, a total of 3,008 persons living in or within 20 households of a confirmed index case’s house were tested. Of these, 125 were found positive for DENV by either onsite NS1 rapid test or PCR.

In Thailand, enrolment of index cases and household members occurred in two Bangkok sites, Vajira Hospital and the Hospital for Tropical Diseases, and Thansongyang Hospital. Due to the seasonal nature of dengue transmission, enrolment took place between July and September 2015, with follow ups continuing until the end of November 2015. During the enrolment period there was an outbreak of dengue in Tak Province, the location of Thansongyang Hospital, leading to a large number of enrolments. During the reporting period we enrolled 290 index cases, 148 households and 192 household members, achieving our target. After the 2015 enrolment and follow up period, we requested the full data from the collection sites. At the beginning of this second reporting period we also had on hand the 2014 data and we started data checking with both sets of data. To ensure quality data, each entry must be checked individually. Data dissemination and sharing: Once a set of data has been through the quality control process it is passed to colleagues in Faculty of Science, Mahidol University, who have responsibility for other objectives and tasks in other WPs.

- WP2 New detection method
Since Dengue virus (DENV) infections are caused by one of the four viruses named DENV1, DENV2, DENV3, and DENV4, a highly type specific neutralizing antibody response is induced in patients. The aim of WP2 was to develop new diagnostic tools to identify DENV serotype specific antibodies, to test them by in--field studies and to correlate the obtained diagnostic results to other clinical parameters.
The principle strategy was the expression and purification of the three E protein domains ED1, ED2 and ED3 as separate antigens. These antigens were used in ELISA and a simple dot-assay to test serum antibodies. A second strategy was the study of serotype specific neutralization by a classical virus neutralization test. Here we used DENV variants recommended by WHO and virus isolates from endemic regions like Cambodia. This allows a highly autologous study, where antibody test results by ELISA or dot-assay could be compared to virus neutralization results and the type of virus present in acute infection. This has not been done in such detail before.
Before the start of the project, current knowledge implies that when a DENV-specific antibody response is diagnosed the exact matching virus must have been present, or when a DENV was identified by RT-PCR the patient will develop an appropriate and type-specific antibody response against exactly this virus. It was thought that diagnostic tests that will not find such a comparison are without any value and if possible have to be improved.
This knowledge, that for example DENV-1 is inducing a DENV-1 type specific immune response, was now challenged by our study using the new diagnostic tool set. For a clear explanation in this report viruses directly identified in patients sera by RT-PCR are called virotypes and DENV analysis by an indirect antibody test like our ELISA or dot-assay are designated serotypes.
The E protein contains three distinct domains, with the third domain (ED3) as a well-known target for DENV serotype-specific antibodies. ED3 is encoded by a single open reading frame within the E gene. This is one of the reasons why ED3 was used as a diagnostic tool in many studies starting from the early days of cloning. In contrast, ED1 is encoded by three and ED2 by two open reading frames making it difficult to express these domains as single separated units. To solve this problem, we have linked the ED1 and ED2 open reading frames by a Gly-Pro-Gly coding sequence. This leads to expression of functionally ED1 and ED2 domains. Functionally means, that these domains reacted positively with DENV-specific sera in field studies using sera from Vietnam, Colombia and Cambodia (n=2000).
In Summary, ED1 was identified as a DENV-specific antigen. It reacted positively with DENV-positive sera. Not with serum controls positive for JE or WNV. It is therefore suited to diagnose DENV infection in general. ED2 was characterized as a highly cross reactive antigen. ED2 reacted with DENV as well as sera positive for JE, HIV-1 and Plasmodium. It was therefore not used to study DENV or DENV-specific immune responses in human serum samples. ED3 was characterized as a highly type-specific antigen. ED3 showed single and double positive reactions that were identical to the serotyping result obtained by the classical virus neutralization test. The same results can be obtained by the endpoint dilution EED3 ELISA. The ED3 ELISA was also be advanced by other ED3 antigens than DENV. To test DENV antibody and to differentiate them from Zikavirus antibodies we have also constructed Zika ED3 antigens and tested them with sera from Columbia and Africa. Sera that were diagnosed for DENV but showed low antibody titres and no direct virus detection by RT-PCR were found to be positive for Zika antibodies. On the other hand, sera tested positive for Zika showed high titres to one of the DENV. This implicates that serotyping by ED3 antigens, especially by the endpoint dilution ELISA is a convenient tool to differentiate flavivirus infections. All antigens tested in our studies (ED3, ED2, ED1) were expressed as a fusion to maltose binding protein (MBP). Such fusions can be expressed in bacteria with high yields. Purification from bacterial extracts by affinity chromatography is a well-established technique and we have developed and second purification step to obtain highly purified MBP-fusion proteins. Serotyping by the ED3 ELISA identifies the peaking antibody response in DENV-positive sera which was identical to the peaking neutralization titre identified in virus-cell culture experiments. Thus, our ED3 ELISA is in clear coincidence with the results obtained by classical virus neutralization. The ED3 ELISA is therefore a very useful tool to diagnose the DENV-specific humoral immune status in patient sera. Since it is material and serum consuming it can be used preferentially for daily diagnostics or for the study of DENV serotypes in small cohorts. The ED3 ELISA is also suited to differentiate DENV from other flavivirus infection, for example Zika infections.
Since the ED3 ELISA is a more time and serum consuming technique we have developed a very simple test for serotyping lager cohorts. The aim was to study between 300 and 480 sera per day. This can be done by a simple dot assay. Our test system contains 24 antigens on a nitrocellulose strip that is treated in analogy to a western blot strip. This can be done manually in 30- or 48-well trays or by special robot machines, especially develop for the western blot technology. Antigens for DENV, WNV, YF, JE, Zika, positive and negative controls are dotted on one of the nitrocellulose strips. To analyse the test strips we have developed a software tool to obtain digital data for each antigen dot. Our aim was to design the test as simple as possible. The test can be evaluated using the software tool or even by naked eye. This makes it possible to handle the test even under simple and low-level laboratory conditions. The test gave serotype specific results, identical to those found by ELISA or virus neutralization in 85% of the samples tested (n=1700). It is therefore suited to monitor the antibody status in lager cohorts. One advantage of the test is that the dot assay gives clear black and white results for single positive sera when using ED3 in a modified version. This modified version was claimed by a patent application. The modified ED3 together with ED3 and control antigens allows an easy DENV serotyping. Using this test we have studied serotypes in stored serum samples from the IPC biobank in Phnom Penh (Cambodia). Analysis of sera taken in different years showed that a shift to the DENV-2 serotype occurred in 2012 indicating that peaking antibody was now mainly directed to DENV-2. Since the serotyping study showed that other responses like the response against DENV-1 were reduced, DENV-1 will spread more frequently. Principally, the ED3 dot assay can replace the time consuming neutralization test and the maintenance of an expensive BSL 3 lab. It can be performed in a fast and easy manner and is suited to diagnose peaking DENV-specific neutralizing antibody responses in patients.
For a simple and rapid DENV diagnostic test we have used ED1 as antigen linked to a DNA molecule that is one of the primers used in an isothermal amplification strategy. By using this antigen (MBP-ED1) we have developed a diagnostic tool to study DENV-antibody in general. The antigen identifies antibody but uses DNA amplification and detection technology to analyse the antigen-antibody complex. The ED1 antigen itself was generated for each of the DENV and tested for reactivity to serum antibody in two different field studies in Cambodia. The ED1 and ED2 antigens, expressed using the Gly-Pro-Gly linker, were claimed by a second patent application and can be used by the cooperating partner AmpTec to develop a commercial DENV-test.
Results of ELISA and dot assay data, the serotype analysis, were compared to the virotype present during acute infections. No perfect match was observed, showing that the serotype was not matching the virotype. The DENV-specific antibody response measured by ELISA or ED3 dot assay was not identical to the virotype analysed by RT-PCR. This disagreement (virotype doesn’t match serotype) was 100% when tested sera and viruses were both from the same endemic area, i.e. Cambodia. From these data we concluded that an analysis of DENV-specific antibody responses by diagnostic tests must be specially tailored for exactly those viruses present in an endemic region. DNA technology allows easy cloning of such tailored ED3 antigens to obtain area-tailored ED3 ELISA or dot assays. The gold standard, the virus neutralization test is more complicated to adapt. One major problem is that not all the virus variants present in endemic areas are available in the labs, making it impossible to compare and evaluate the antigen test accordingly. To make DENV variants available for neutralization tests we have developed a vector system to generate Flavivirus M/E envelope pseudotyped HIV-1 particles. Co-transfection of an M/E expression vector and an env deleted HIV-1 vector leads to particle release. These particles carry M/E envelopes and are infectious for VeroB4 or other DENV permissive cells. Since the HIV-1 carries a reporter gene (luciferase), infection can be measured by standard luciferin assays.
Taken together, two test systems have been developed. (i) A system based on DNA amplification and detection technology to detect antibodies to DENV in general by a ED1 antigen linked to a DNA molecule and (ii) a diagnostic system for serotyping of DENV infections. This second test system can replace the virus neutralization test by using ED3 antigens.

WP3 Seasonal climate predictions for the dengue study sites
WP3 had as a major objective setting up a series of research quality public health user environmental and climate-based oriented products. The accomplished goals now make it possible to have the best observational and prediction climate datasets for the target regions at the easy access by any interested user. Similarly, an online climate skill interactive tool (Skill Finder) depicts coupled regions where for each key climate covariate, incorporation of climate information is both possible and useful for dengue and other arboviral diseases months ahead of the outbreaks (Fig. 1). This online near-operational skill-finder tool has been developed that can be used to assess where and in which seasons climate information is going to be most useful for disease prediction purposes. A move to operational products for dengue prediction was initiated and it is therefore possible in the near future for interested public agencies and local users. Similarly, WP3 furnished and calibrated all needed climate information and datasets (observational, new reanalysis, seasonal forecasts and future longer-term projections) for the use of epidemiological models developed in WP4.

Figure 1: A QGIS project to interactively overlap skill contours for different variables, seasons and terciles. The forecast system is ECMWF System4 with 15 ensemble members. (see figure in the “figures Final report DENFREE” document).

A new methodology to assess both the intra-seasonal and the year-to-year variations of the East Asia Summer Monsoon (EASM) variability was also developed. This method estimates the main regimes of the temperature and rainfall over Thailand and Cambodia using monthly data zonally averaged over the region and covering the latitudes between the Equator and 22ºN and the period from June to October. The aim of the longitudinal averaging is to take into account the information on the latitudinal migration and temporal distribution of the summer monsoon precipitation over the region. This approach represents a process-oriented assessment of both the variability and predictability of the ITCZ-related EASM effects. The spatial patterns related to the leading modes are effectively used to describe the EASM regimes. The results obtained from a suite of different global coupled climate models and the multi-model ensemble suggest that this methodology can be useful to characterize the predictability and skill of the intraseasonal variability of the EASM. This intreaseasonal variability was assessed for the region within the period 1980-2015, in conjunction with the interval for which dengue datasets were available for different countries in the region.
The same approach was carried out for the East Asian winter monsoon (EAWM), which in contrast to the EASM is characterized by a strong surface northerly, a strong west-east pressure gradient between the Siberian high and the Aleutian low, an upper tropospheric East Asian jet stream, and a deep long wave trough along the East Asian coastal region. The intraseasonal and interannual variations of the EAWM play a major role in determining the evolution in temperature, known to be one of the main external covariates of use in arboviral disease models. Along these lines, predictive model frameworks for chikungunya in Cambodia and for dengue in Brazil, Ecuador and Thailand were developed and extended in collaboration with WP4 that used this climate information and seasonal forecasts for the first time ever as one of the driving covariates in both statistical and dynamical disease models. For instance this allowed the exploration of how seasonal climate forecasts could be seamlessly linked to a statistical dengue model, to provide early warnings of dengue risk to public health authorities. The same approach was developed but with a dynamical dengue framework in the case of both Thailand (also for climate change scales) and as part of the US dengue Challenge Initiative in 2014 and 2015. In collaboration with European and Brazilian climate services, Universities and the Brazilian Climate and Health Observatory, data from different sources and spatial/temporal scales (e.g. dengue, climate, cartographic, demographic, socio-economic) was collated to formulate probabilistic dengue predictions for the 553 microregions of Brazil. By assessing the past performance of the model, optimum trigger alert thresholds were identified to maximise "successful prediction" and minimise "false alarms" for scenarios of medium-risk and high-risk of dengue, according to incidence alert levels defined by the Ministry of Health. This work was done in combination with final WP4 objectives to be accomplished through the integration within a suite of different dengue models of a seasonal multimodel climate prediction system and for the first time, ahead of a mass-gathering event.
Work conducted in this WP further helped define the role ENSO plays on southern Indochina’s climate through the study of the characteristics of simultaneous rainfall and near surface temperature variations. In association with ENSO extreme events on the basis of the selected simultaneous RI and ONI years. negative (positive), rainfall anomalies over southern Indochina generally link to warm (cold) SSTAs in the central tropical Pacific, though the link is clearly evident between the decaying phases of both ENSO events and the corresponding rainfall response. Moreover, the slow-varying SSTA in the central tropical Pacific can totally advance rainfall variations in southern Indochina up to 4 months if the overall growth-decaying event is considered, confirming the essential role of climate driver that ENSO phenomenon has over the region with correlations accounting for nearly the entire variability at these frequency ranges (e.g. a complete coupling in 1997). Effects and associated mechanisms by which El Niño-Southern Oscillation affects climate in the Indochina Peninsula and South America and the ways in which they can be useful for dengue and chikungunya prediction models were equally assessed in DENFREE.
Towards the aim of incorporating the role of climate in the simulation of dengue epidemics, first estimates of the seasonality and the dynamics of monsoon active and break phases were explored that can be of use in the future for very short lead time prediction (one to two weeks ahead). Similarly at the seasonal and interannual scales, good prospects exist to develop a useful early-warning system for regional public health authorities, as the core science and implementation information is directly available as a result of DENFREE. Future work will consist in applying the same methodology to daily values instead of monthly means, although this initiative will encounter the difficulty of accessing robust daily data for the region. Besides, climate data from the new version of the NMME, NMME-2, should be used for a better outcome. Finally, this methodology has been extended to three-dimensional data, where the variables are not zonally averaged. This approach better resolves the characteristics of the spatial variability required to improve prediction skill.
The deterministic skill of the dynamical forecast systems described above has been assessed at each grid point over the Southeast Asian region (Equator-22ºN, 96ºE -110ºE) to evaluate to what measure they are able to simulate the seasonal rainfall and temperature variations at the interannual time scale. The correlation coefficient is used to assess the degree of linear correspondence between the predicted ensemble-mean and the reference values (observation based), which are taken from GPCP and ERA Interim. Figure 2 shows the correlation of the ensemble mean for temperature and precipitation for the eight forecast systems considered. The skill of the temperature forecasts is higher than for precipitation, a general feature in seasonal forecasting. Most forecast systems have a positive statistically significant skill for temperature. For precipitation, skill is also positive, especially away from the Equator, and suggests that there is an ability to predict their changes from year to year.
The forecast systems are able to reproduce many of these characteristics, especially the meridional shift of the temperature and precipitation along the monsoon season.
Of the several deterministic and probabilistic scores displayed, both GFDL and CMC2 show substantial probabilistic skill in terms of BSS (considering the events “rainfall regime above the median” or “above the upper quartile”) and, in particular, in terms of the reliability of the forecasts, measured by the reliability component of the BSS. The link between the first modes with ENSO is clearly observed in particular for precipitation, and can explain a large part of the skill found. For the climate change timeframe, a database with regionalized simulations from CMIP3+ from the whole ensemble of model runs available is made publicly available as a repository at IC3.
Climate services are now ready to be developed and are demonstrated in several countries to effectively make a big difference in terms of disease prediction thanks to the climate skill available. As an example, generalized linear and additive mixed models (GLMM/GAMM) were formulated to assess the importance of climate variables as drivers of spatial variation and interannual variability in dengue transmission across Thailand and Brazil. A negative binomial model was used to account for over-dispersion found in the dengue count data. By including the expected number of cases of dengue as an offset, we estimated the relative risk (SMR) of dengue using a combination of spatio-temporal structures and linear and nonlinear functions of climate. Added value (linearly or nonlinearly) of climate information can be clearly seen to make a large difference for dengue prediction in Thailand (Fig.3).
Very recent approaches taking advantage of the leading capacity to predict el Niño phenomenon within DENFREE by means of nonlinear and nonstationary statespace models allow to successfully anticipate dengue outbreaks in Ecuador more than two years in advance (Fig. 4). The different model frameworks developed within WP3 are extremely flexible and could be applied to model spatio-temporal dengue variation in any geographical setting. This would facilitate between-country comparisons of the impact of climate on dengue fever and contribute towards a more global approach to assessing and predict the impact of climate variability and climate change on dengue risk many months ahead.

Figure 2: Average correlation of the ensemble mean for temperature and precipitation for the eight forecast systems considered. The predictions shown are for lead time 1 month (i.e. predictions starting in May). (see figure in the “figures Final report DENFREE” document).

Figure 3: Value added from using climate information to explain dengue relative risk variation. (see figure in the “figures Final report DENFREE” document).

Figure 4: Long-lead forecast of dengue in Ecuador obtained nesting a statistical dengue prediction system with an El Niño state-space Kalman filter model. (see figure in the “figures Final report DENFREE” document).

- WP4 Descriptive and Predictive models
Task 4.1: Study of the interaction of dengue strains in the build up of immunity with dynamical models for dengue
We developed models and improved and extended already existing models including include secondary infection, temporary cross-immunity, seasonality, climatic variables and vector dynamics. These models gave a qualitatively good fit to empirical incidence data for the Province of Chiang Mai, North of Thailand. We further collected data on various aspects of the epidemiology of dengue fever. Especially, large data sets of severe dengue over 30 years and for all provinces in Thailand were obtained for subsequent model comparison.
We investigated spatial extended stochastic systems for spreading of disease, which are a form of individual based models, in the sense that each individual at any location in physical space has as state variables the property to be infected or susceptible, eventually recovered etc.. The neighbourhood between individuals is specified by a so-called adjacency matrix, which easily allows us to incorporate geographical neighbourhood, weighting with distances, or any other network structure, like the famous small world networks or scale free networks. We concentrated up to now on superdiffusive spreading, which preserves some notion of geographical distance, also called spatially restricted networks. Levy exponents for superdiffusion have been investigated for human contact networks via surrogate processes like spreading of money bills and their distributions in space and time. From such studies it has become clear that human contact networks are likely to have superdiffusive structures, with Levy exponent as low as 1.5 or less (ordinary diffusion has Levy exponent 2, and superdiffusion with Levy exponent 1 is also known as Cauchy process). For spatially extended models on dengue fever and especially its spreading into new areas, such notions of human contact patterns will be of crucial importance.

Task 4.2: Diagnose and reassess climate-disease relationships for each targeted DENV- location.
The spatial-temporal pattern of propagation of annual epidemics in Cambodia highlighted the role that human movement plays. Particularly we revealed a great spatial heterogeneity in the propagation of the annual epidemic. Each year, epidemics are highly synchronous over a large geographic area along the busiest national road of the country whereas travelling waves emanate from few rural areas and move slowly along the Mekong River towards the capital, Phnom Penh. Preliminary results based on the Thai dataset also showed similar results corroborating our observations from Vietnam and Cambodia. The main result of the published papers concerns the refutation of the dominant paradigm that highlights the influence of the urban centre in the dengue propagation. All epidemics at a given scale were originated from a large urban centre. Our published results have clearly shown the reverse with low incidence in large urban centres as Phnom Penh and Ho Chi Minh City and the fact that each epidemic wave arrived always, these last years, later in these two capitals (with a non-neglected delay around 1-3 months). To illustrate this point, one can focus on the Cambodian example (Teurlai M. et al. 2012).

Task 4.3: To integrate a dynamic disease model within a seamless ensemble prediction system using seasonal hindcasts and projections
The role of climate variability in modulating the local dynamics of diseases is a subject of great interest and continued investigation, though its degree of relevance still a highly disputed issue. In the case of dengue in Thailand, climate variability has long been recognized to exert an influence on either dengue exacerbation or dampening.
The climate of Thailand and Cambodia in the Indochina Peninsula is largely modulated by the meteorological swing of the Asian monsoons at intra-seasonal to interannual and longer timescales, with the region receiving most of the rainfall during the summer monsoon season. In the dynamics and strength of each year’s monsoon the land-sea thermal contrast plays a key role, with a myriad of other regional and local factors contributing to the modulation of this seesaw in temperatures. Large-scale processes do also modulate the ocean-atmosphere coupling in the Indian Ocean, responsible for a year-to-year difference in ocean sea-surface temperatures. Among those, the snow-pack in the Tibetan Plateau, linked to large-scale interhemispheric phenomena, the northwest Pacific and the El Niño-Southern Oscillation (ENSO) in the tropical Pacific, TPac, stand-out among the main contributors to both land and Indian ocean temperature anomalies.
Extraction of local variability signatures is possible with the use of selected statistical techniques specifically aimed at determining both the position, structure and duration of temporary climate forcing. A well-developed cold horseshoe is evident in the western Pacific, exerting also an effect on the climate of the region. The remote climate effect exerted by the TPac can be tracked several months in advance and offers prospects for regional predictability. Similar diagnostic analyses were performed but for the dengue incidence data for the interval 1975-2009. Cold SST anomalies prevail over the TPac indicating a La Niña (LN) condition. Higher than normal SSTs can be observed around the South China sea to Indonesia, with larger temperatures over mainland Asia. These ocean and atmosphere conditions are very stable months before and up to the peak in dengue, suggesting predictability potential for indices capturing these patterns in a forecast model for dengue in the region. Maximum correlation occurs at lead times of four months prior to the peak.
Concomitant to the decrease in incidence after the peak in dengue in the region, warm SST anomalies develop in the southern Indian Ocean and over the TPac, the latter typically conducive to an EN event. This suggest that on average, EN conditions favour the termination of dengue epidemics in the region, despite considerable nonlinearity exists both in the presumed forcing regions and in the outcomes as anomalies in dengue incidences.
Despite the average conditions depicted above (with essentially cold anomalies over the TPac for most of the dengue peak events), considerable nonlinearity exists in the climatic signatures leading to the rise of the disease incidence over Indochina. These variable oceanic and atmospheric situations act modulating the climate of the region in a different way and highlight the complexity and the lack of a unique climatic configuration favouring dengue increases (e.g. both drought and flooding situations if associated to temperature anomalies appear linked to dengue variability). Environmental conditioning of the breeding sites, vector population dynamics, modulation of human-mosquito interactions and the influence on the social behaviour might ultimately interplay in the population dynamics of the disease.
Therefore, once accounting for spatial-temporal confounding factors, non-linear functions of temperature and precipitation were found to have a statistically significant positive contribution to the relative risk of dengue in the following month. Therefore, forecast climate information, which predicts anomalous climate patterns several months in advance, has potential utility in a dengue decision support system for Thailand. Taking advantage of climate forecast lead times, public health authorities will be better able to put into place community communication campaigns to increase public awareness of the ensuing epidemic season and encourage community based environmental hygiene to reduce the mosquito breeding sites. The model framework presented here is extremely flexible and could be applied to model spatio-temporal dengue variation in any geographical setting. This would facilitate between country comparisons of the impact of climate on dengue fever and contribute towards a more global approach to assessing the impact of climate variability and climate change on dengue risk.

Task 4.4: Spatio-temporal analysis and modeling of Dengue diffusion at micro-level
Development of a tripartite Agent Based model enabled us to explicitly explore dengue diffusion in real-time and real-space. MO3 is a spatialy explicit dynamical model to explore emergence and diffusion of dengue in urban areas. MO3 is composed by three sub-models: one for environment – MODE (Model Of Dynamical Environment) –, one for the vector - MOMA (Model Of Mosquito Aedes) - and one for the host – MOMOS (Model Of Mobility Simulation)-. The last two sub-models are coupled with a contagious-infectious model. We have calibrated MOMA, implemented MODE and explored mobility data for MOMOS.
MOMA is a spatially explicit agent-based simulation model of Aedes aegypti female mosquito, the dengue vector. The model aims to produce statistical data on mosquito behaviour and population dynamics that are difficult to obtain through field surveys such as population densities in various geographical and climatic conditions. Study of simulated mosquito behaviour reveals the model’s ability to produce the mosquito’s realistic life cycle. The first results revealed a significant relation between urban topology, human densities and adult mosquito flight. We showed that this virtual laboratory, constructed in collaboration with entomologists, make it possible to represent the mosquito’s living areas, which are either restricted or facilitated by spatial amenities conditioning its needs and its capacity of movement during its lifetime.
In order to simulate the mosquito behavior of MOMA in different geographical and climatic environment, we then developed a methodological framework (MODE) to generate synthetic and representative environment at different spatial scales. MODE’s objective is to identify environmental dynamics that contribute to the spatio-temporal distribution of Ae. aegypti in urban areas. We developed the resource-based habitat concept and its application to Ae. aegypti, in order to build generic methods for estimating the environmental factor of this habitat and the factors dynamics. We developed an agent-based model of the host (MOMOS), focusing on the daily mobility at the scale of the city. Through statistical analyses of social network data and of socio-economic data, links were made to human mobility with intra-urban amenities.
Finally we integrated these 3 sub-models into a single (MO3) to produce large-scale simulations. We can now apply these methods locally or globally in order to study various effects on hosts infections, on vector population dynamics or on dengue diffusion. In addition, we integrate into MO3 the “Virus” component, enabling implementation of the main characteristics of different viruses transmitted by Aedes aegypti, notably dengue and Zika virus. MO3 is implemented in the GAMA platform.

Task 4.5 Integrative Investigation of invasion scenarios and intervention strategies
We investigated intervention strategies into our modelling approaches, and concentrated on the development of tetravalent vaccines against all 4 serotypes of dengue virus simultaneously to avoid elevated risks of secondary infection in case of incomplete coverage against any of the serotypes. The front runner in the development of these tetravalent vaccines is the one developed by Sanofi Pasteur, Lyon, which has completed the phase three of trials and already has been licensed at the beginning of this year in several endemic countries. However, the efficacy of this ”tetravalent” vaccine varies significantly from serotype to serotype with medians between 30 and 80 %. As opposed to previous results, negative efficacies can be excluded for all strains, but close to no efficacy is still statistically possible for two of the four serotypes. As a most striking result we can observe a clear grouping into high efficacy against two serotypes, 3 and 4, and low efficacy against the other two serotypes, 1 and 2. Statistically, from this analysis there is no probability any more that still there might be the same efficacy against all strains and just the fluctuations in the trial results would give larger or smaller efficacies.
These results and new information from recent publications on the phase 3 results from Sanofi-Pasteur have been now implemented in our multi-strain models. Highly negative efficacy was reported for the vaccine in the age class from 2 to 5 years, where negative efficacy means that in the vaccine group there were more dengue cases observed than in the control group, indicating that the vaccine actually puts young children at higher risk than naturally observed. Also in the age groups up to 9 years of age the vaccine was found to incur risk, and only in age groups beyond 9 years the vaccine showed some advantageous effect over the placebo control group. The published numbers also indicate that seronegative people at baseline are at risk from the vaccine, and only positive effects appear for people, who have already antibodies from previous dengue infections, i.e. seropositive individuals. Hence in our modelling we had to introduce, alongside primary versus secondary infection, age classes in order to compare with the recommendations given by Sanofi-Pasteur and subsequently also by the WHO. However, every numerical evidence indicates that age class is only a surrogate for sero-status, since in younger age classes more individuals are sero-negative than in older age classes. However, the sero-negatives in older age classes are as endangered by the vaccine as the seronegative in the young age classes. Hence we recommend a screening on sero-status before administering the vaccine.

Task 4.6 Development of metapopulation models for dengue data in French Polynesia
French Polynesia is located in the middle of the Pacific. French Polynesia is composed of 119 islands spreading across 5 million km2 of ocean and has intermittent dengue epidemics with long periods of low circulating endemic dengue. Given the small but “island” fractured population, we developed metapopulation approaches to modelling dengue in this scenario to address whether this is the best approach when dealing with a sub-structured population. To study the metapopulation system, we used an SEIR model using Ordinary Differential Equations to determine the states of the groups in small populations connected into a meta-population and coupled by human inter-island migration (by boat and air).
Such a metapopulation approach captured the dynamics of dengue much better than treating French Polynesia as a single population and suggest that dengue can only be maintained in French Polynesia because of island-hopping of the virus from island to island and as importantly, a very sub-structured population within the main island Tahiti, that for the majority of the time acts as a virus source for out islands. A particularly important and novel parameter as to add in a viscosity parameter that restricts the number of people a mosquito can bite. This thus incorporates an element of agent-based modelling into an otherwise deterministic framework.

- WP5 Asymptomatic transmission, European vector competence and innovative vector control strategy

The WP5 is dedicated to examine the role of mosquito vectors and more specifically, Aedes albopictus in sustaining a dengue outbreak in the European context. To reach this goal, WP5 is divided into 3 tasks: (1) estimate the contribution of asymptomatic patients in the transmission of DENV, (2) evaluate the ability of European Ae. albopictus to transmit DENV carried by patients returning from DENV-endemic regions, and (3) identify genetic changes conferring an enhanced transmission of DENV by European Ae. albopictus.

5.1 Contribution of asymptomatic patients in the transmission of DENV
Most of dengue human cases are asymptomatic or result in too few symptoms to be detected by existing surveillance systems. These clinically inapparent cases were not considered to develop a viremia high enough to infect mosquitoes and therefore were assumed to be dead-ends for virus transmission. We showed that not only asymptomatic people are infectious to mosquitoes, but also that they are significantly more infectious to mosquitoes than people with symptomatic infections at a given viremia level (see publication below). Our results will change significantly the current paradigm of dengue epidemiology as well as the response to a dengue outbreak. The asymptomatic people may behave as viral reservoirs ensuring the dissemination and maintenance of the virus.

5.2 Evaluate the ability of European mosquito species to transmit imported Dengue virus strains
Arboviral emergence requires competent mosquitoes. In Europe, the Asian tiger mosquito has been introduced in 1990 and is now present in 20 European countries. It was responsible for local dengue cases in France (2010, 2013, 2014, 2015), Croatia (2010), and Madeira island (2012). Owing to the increasing number of autochthonous dengue cases, we have evaluated the ability of European populations of Ae. albopictus to transmit imported strains of DENV. Six populations of Ae. albopictus have been established from eggs collected in ovitraps: Genova (Italy), Alessandria (Italy), Cornella (Spain), Martorell (Spain) Nice Jean Archet (France), and Saint-Raphael (France). We used two Dengue-1 strains isolated from DF cases: D1-1806 from an authochthnous case from Nice in 2010 (provided by the National Reference Center of Arboviruses, France) and D1-30A from a patient in Kamphaeng Phet, Thailand in 2010 (provided by the Afrims, Thailand). Mosquito adults were fed on an infectious blood-meal at a titer of 106.8 FFU/mL and were examined at days 14 and 21 post-infection (pi). Mosquito susceptibility to DENV-I infection was measured using common vector competence indices such as midgut infection rates and viral dissemination to other organs. Midgut infection rates were determined via positive/negative titration of mosquito bodies, while dissemination was determined via titration of viral loads in the head.

Figure 5. Infection Rates and Viral Titers at Sites of Disseminated Infection: Adult female mosquitoes were infected with DENV-I from Thailand (30A) or southern France (1806). At indicated days pi, mosquitoes were sacrificed and decapitated. Bodies and heads were homogenized and titered on C6/36 cells. Infection rates were determined using positive/negative scoring, while viral titers at sites of dissemination were quantified via focus-forming assay. (see figure in the “figures Final report DENFREE” document).

Figure 6. Analysis of Interactions between Mosquito Populations and Virus Genotype: Interactions between mosquito population and virus genotype were analyzed using statistical software. Vector competence is driven by mosquito populations and appears largely independent of viral genotype. (see figure in the “figures Final report DENFREE” document).

Our experiments show that within mosquito populations the viral genotype does not seem to play a role in either infection rate or viral loads at sites of dissemination. In contrast, we observed that the most striking difference was between mosquito populations suggesting that population genetics may play a larger role (Figures 1 and 2).

Temperature is a factor which is carefully controlled in laboratory environments, however during natural infection cycles, temperatures fluctuate with day/night cycles, affecting mosquito metabolism and consequently, viral replication. To determine the effect of temperature on the vector competence of Ae. albopictus, mosquitoes were orally infected with DENV-I isolated from a locally acquired infection in southern France in 2014. Infected females were incubated for 11 or 18 days pi under three different temperature conditions: constant 28°C, constant 20°C, or variable temperatures between 17-23°C. Mosquitoes were examined as previously described. We observed that temperature play a role in both infection rate and viral dissemination (Figure 3).

Figure 7. Temperature Effects Infection Rate and Impacts Viral Loads at Sites of Dissemination: Field derived Aedes albopictus mosquitoes from Nice, France were orally infected with DENV-I isolated from a locally infected patient from southern France. At 11 and 18 days pi, adult females (n= 17 – 20) were sacrificed and decapitated. Body and head homogenates were made and infection rate was quantified using positive/negative selection in C6/36 cells. Viral loads in disseminated organs were quantified via focus-forming assay. (see figure in the “figures Final report DENFREE” document).

5.3 Identify viral genetic polymorphisms associated with experimentally enhanced transmission by European vector populations.
Aedes albopictus is not considered as a main vector of DENV. However owing to the recurrent detections of local dengue cases in Europe, the question of its long-term adaption for a better transmission of DENV has been addressed. We have designed a protocol to experimentally enhance transmission potential of DENV for European Ae. albopictus. An imported viral strain may acquire one or several mutations conferring enhanced transmission by a European mosquito population.

DENV isolates from Thailand (30A) and France (1806) have been passaged serially in a single Aedes albopictus population originally isolated in Nice, France, and Ae. albopictus (C6/36) cells in vitro (to amplify virus recovered from mosquito saliva during the previous passage). Ten passages have been completed and passages 1, 5, were deep sequenced to identify mutations associated with enhanced transmission in European Aedes albopictus. Control isolates were serially passaged in C6/36 cells to identify mutations resulting from genetic drift or adaptation to insect cell line.

Results show no major change in single nucleotide variants (SNV) frequency in the viral population for isolate 1806 and 30A when serially passaged in C6/36 cells. However, when the viruses were passaged in Ae. albopictus mosquitoes and cell culture alternatively, some new variants reached consensus level (frequency > 50%) as soon as passage 4 (isolate 1806) and passage 5 (isolate 30A). Only one SNV reaching consensus level was shared between replicates for isolate 1806, and none for isolate 30A. Its frequency increased over the passages, first detected at passage 2 or passage 6 depending of the replicate (Figure 4). The variant reached consensus level within two passages following initial detection and became almost fixed (frequency > 99%) soon after, consistent with a strong selective sweep.
This variant presents a nucleotide change in the highly structured 3’ untranslated region of the DENV genome. The presence of two secondary and tertiary level RNA structures, called pseudo-knots is involved in the generation of subgenomic RNA called sfRNA (subgenomic flavivirus RNA) through resistance to a cellular enzyme involved in messenger RNA degradation. The nucleotide variant we identified is present in the second of these pseudo-knots and may affect its tertiary structure. Disturbance of tertiary structure of the second pseudo-knot has been involved in diminution of sfRNA production. In mosquito cells, accumulation of mutations in DENV second pseudo-knots sequence can be associated with a decrease of sfRNA production in mosquitoes. We are attempting to create an infectious clone of DENV isolates used in this study (30A and 1806). These clones will allow us to perform mapping studies, and correlate mutations identified from deep sequencing to viral replication and transmission kinetics in mosquitoes.

Figure 8: Observed rise in frequency of a nucleotide variant located in the 3’ UTR of isolate 1806 during the 10 selection cycles. (See figure in the “figures Final report DENFREE” document).

5.4 Vector control method: The use of PPF as a measure for Ae. aegypti control focuses on a strategy that exploits the innate behaviors of mosquitoes. This approach, called auto-dissemination, is based on the assumption that mosquitoes exposed to a contaminated surface with PPF are able to spread this larvicide to their own breeding sites, through oviposition and rest (Devine et al 2009; Caputo et al. 2012). The exposure of the mosquito population to PPF is achieved through the use of artificial structures, called dissemination stations. Mosquitoes, after coming into contact with the contaminated surface with this larvicide in dissemination stations, will transport, in their body and legs, the particles of PPF that will transfer to other breeding sites. This strategy, when directed to dengue mosquitoes, particularly Ae. aegypti, it’s enhanced by their oviposition behavior, since the eggs are distributed in the same posture for several breeding sites, thus allowing greater breeding site coverage (Caputo et al 2012).
Between September and December of 2014, the auto-dissemination approach was first tested in Madeira Island, through a small pilot study in the village of Paul do Mar, located in the parish of Paul do Mar, Calheta county. The results of this study suggested that this may be a promising method for the control of mosquitoes Ae. aegypti in Madeira (See Deliverable 5.4). Thus, it was proposed to direct the PPF auto-dissemination approach to the parish of Imaculado Coração de Maria, in Funchal, where densities have remained high. Similar results were obtained in Funchal as were achieved in the previous year’s study in Paul do Mer. PPF implementation led to ~20% higher larval mortality rates. The results are currently being analysed and effects may be even greater than those observed via larval mortality; adult densities as measured in BG sentinel traps were significantly decreased and this was especially notable in Imaculado Coração de Maria where mosquito densities are generally much higher.

- WP6 Viral genetic diversity and adaptation
The activities within WP6 is still ongoing. DENV strains obtained from human and mosquitoes during Kampong Cham DENFREE study in Cambodia in 2012-2013 will be subject to deep sequencing at IP in Paris. We aim to study interhost genetic diversity of DENV (i) among household dengue cases and natural infected mosquitoes, (ii) among dengue index case (DIC) and mosquitoes infected through direct mosquito feeding on (DIC). We also plan to look at the genetic diversity of DENV in human with different clinical patterns such symptomatic (DF, DHF, DSS) and asymptomatic cases. Finally, we aim to study intrahost genetic diversity of DENV through analysis of DENV population in serial blood samples collected from one patient (first sample collected during the pre-symptom phase then blood collection during the clinical phase of the disease). The final result should deliver very interesting data which should lead to a high impact article.

- WP7 Key immunological factors determining DENV clinical outcome
We characterized a panel of human monoclonal antibodies reacting to the DENV envelope protein, generated from plasmablasts isolated from DENV infected patients. Detailed epitope mapping of these antibodies demonstrated three broad reactivities. Around 1/3 of the antibodies reacted to the well described fusion loop epitope (FLE), 1/3 were not definitively mapped, but like the fusion loop antibodies they reacted to envelope protein by Western Blot (these are termed non-FLE as they were not sensitive to mutation of envelope residue W101). Finally, a group of around 40 antibodies did not react to envelope protein by western blot and only bound to intact virus particles. We termed this new epitope the E dimer epitope (EDE), which were subdivided into two groups EDE1 and EDE2 based on the sensitivity to the removal of the N-linked glycan N153 in E (EDE2 binding was reduced by removal of N153, EDE1 not). Some EDE antibodies were fully crossreactive to all four DENV serotypes and could neutralize infection in the picomolar range.
Zika virus (ZIKV) is an arbovirus belonging to the family flaviviridae and is transmitted to man by Aedes mosquitos and differs from Dengue virus (DENV) by 41-16% (E protein). Recent reports have shown difficulty in distinguishing DENV and ZIKV infections serologically implying a degree of antigenic similarity between the viruses. We therefore tested the binding of the panel of anti-DENV monoclonal antibodies to ZIKV by capture ELISA. The profile of binding between the African (HD78788) and French Polynesian (PF13) strains of ZIKV was highly similar, all of the fusion loop antibodies cross reacted with ZIKV, 36/37 of the non fusion loop antibodies crossreacted whereas the crossreaction of the EDE antibodies was variable with 27/33 EDE1 and 8/17 EDE2. Interestingly, the epitope recognized by some EDE antibodies is also conserved in the ZIKV E-dimer, leading to equally potent neutralization making the EDE a potential target in ZIKV.
Structural characterization of these antibodies with Felix Rey, a co-PI on this WP, has shown that anti-EDE mAbs bind in a valley formed between the two E- subunits of the head-to-tail dimers present at the surface of the virion (Fig. 1). The antibodies make contact with a conserved surface patch at the dimer surface, including atoms of the fusion-loop main chain but not its side chains. This conformational site is the also responsible for the interaction of the E-dimer with prM during virus maturation, explaining its conservation within the DENV-ZIKV group (Fig. 2). In addition to their broad neutralizing potential, the anti-EDE mAbs also efficiently neutralise virus produced in insect as well as primary human cells. The latter are a probable surrogate of viruses produced in the infected human host, contain low levels of prM and are the most difficult to neutralize.

Figure. 9 The structures of DENV-2 in complex with anti-EDE-mAb showing the epitope of anti-EDE antibody lies across 2 E within a dimer. A) side view and B) top view. Domain I, II and III of E protein are indicated in red, yellow and blue. On the top view, grey and green ovals show the binding areas of heavy and light chains of the anti-EDE mAb. C) Exposed main-chain atoms in the epitope. Surface representative of DENV-2 sE as viewed from outside the virion with exposed main-chain atoms orange (top) or with main-chain atoms plus conserved side chains in orange, and highly similar side chains in yellow (bottom). The epitopes of two EDE mAbs are indicated. (See figure in the “figures Final report DENFREE” document).

Figure 10 The structures of ZIKV in complex with anti-EDE1-mAb a, Overall view of the complex, with the sE moiety coloured by domains (I, II and III in red, yellow and blue, respectively); the antibodies in grey and dark green for light and heavy chains, respectively. The CDRs are coloured (H1, light blue; H2, sand; H3, pink; L1, light grey; L2, magenta; L3, orange). The inset shows a comparison with the corresponding DENV-2 complex. For clarity, the variable region of the C8 Fab fragment of the DENV-2 sE–C8 complex was superposed on the C8 scFv in complex with ZIKV sE to draw the Fab axes and show the docking angles better. b, Zoom of the ZIKV sE–C8 interaction to show the recognition of the b strand. Hydrogen bonds are shown as dotted lines and immobilized water molecules as red spheres. c, Same region on the DENV-2 sE–C8 Fab complex. Note that the N67 glycan on DENV also interacts with the antibody. d, The footprint of EDE1 C8 is outlined on the ZIKV sE dimer shown in surface representation (looking from outside the virion) coloured according to conservation of surface-exposed amino acids. Atoms from the main-chain and conserved side chains are orange, highly similar side chains are yellow, and all the others are white. e, f, Footprints of EDE1 C8 on a surface representation of ZIKV sE (e) and DENV-2 sE (f) shown in purple. FL, fusion loop. The two protomers of sE in the dimer are in light and dark grey. Relevant antigenic sE regions are labelled. Note the more confined interacting surface in the ZIKV sE dimer than in DENV-2, for example, N67 glycan is absent in ZIKV sE. (See figure in the “figures Final report DENFREE” document).

- WP8 Role of antigen presenting cells, T cells and metabolic pathway in DENV infection outcome; Risk of dengue in European population

Role of antigen presenting cells and T cells
In order to find host mechanisms that result in control of DENV infection without clinical symptoms we used an integrated approach of transcriptional profiling and immunological analysis of a Cambodian population of acute DENV infected, strictly asymptomatic individuals in comparison to clinical dengue patients. A total of 84 dengue-infected individuals were included in the study. Dengue infection was confirmed by detection of viral RNA in serum: 8 individuals were strictly asymptomatic during inclusion and the 10-day follow-up period.
We revealed no major differences in activation of innate immune responses. Pathways such as regulation of innate immunity, antiviral innate immunity, activation of pattern recognition receptors, IL-8 signaling and type I and II IFN regulated pathways did not differentiate between symptomatic or asymptomatic clinical outcome.
The most significantly activated pathway in asymptomatic individuals was NFAT mediated regulation of immune response. NFATs are at the crossroads of the adaptive immune response and are expressed after antigenic stimulation of lymphocytes. Pathway analysis revealed an up regulation of the antigen presentation pathway and dendritic cell maturation in asymptomatic dengue-infected individuals. These data were confirmed by ex vivo phenotypic analysis of peripheral blood mononuclear cells (PBMCs) collected from both groups. Taken together these transcriptional and protein expression data suggested that antigen-presenting cells are highly activated in asymptomatic individuals.
In accordance to the observed increase in antigen-presentation, PKCθ Signaling in T Lymphocytes is highly activated in asymptomatic dengue-infected individuals. PKCθ is an essential component of the T-Cell supramolecular activation cluster and mediates several crucial functions in TCR signaling. By phenotypic analysis of PBMC, CD69 expression was significantly higher in both CD4+ and CD8+ T cell populations in asymptomatic dengue infected individuals. One of the most significantly activated pathways in asymptomatic dengue infected individuals is implicated in TCR-mediated apoptosis: calcium-induced T Lymphocyte Apoptosis, which might be due to an attempt to counteract the proliferative response after TCR stimulation. Taken together, an asymptomatic outcome of DENV infection is associated with increased T cell activation and T cell controlling mechanism.
We investigated in detail the association between clinical outcome and B cell responses after DENV infection. Pathway-based analysis indicated an activation of B cell receptor (BCR) signaling pathway in asymptomatic individuals. Of major interest, FcγRIIB signaling, mediating inhibition of BCR signaling after ligation, was activated in asymptomatic individuals. In addition, molecules such as FCRL1 and FCRL6 implicated in inhibition of B cell differentiation/activation were highly up regulated in asymptomatic infection. This could be reconciled with the observed strong inhibition of PI3K signaling in B cells in asymptomatic dengue infected individuals. Moreover, key transcription factors for plasma blast differentiation are down regulated in asymptomatic infection compared to DENV infected individuals with symptoms. These data indicate that whereas B cells are activated in asymptomatic individuals, inhibitory mechanisms are in place preventing the differentiation to plasma cells. In symptomatic patients on the other hand, differentiation to plasma cells is increased.
Role of metabolic pathway in DENV infection outcome
The results came from two independent population genetics/genome wide association studies (GWAS) of dengue patients collected from Cuba and Thailand. We conducted a GWAS of 2.5 million SNPs in 274 Cubans, including patients (DF and DHF) of the 2006 dengue epidemic, from Havana (west) and Guantanamo (east) cities, and in geographically matched asymptomatic individuals and population controls. The high level of admixture in Cuba enabled us to apply the first admixture mapping, thus facilitating the identification of candidate markers ethnically associated with dengue infection. Firstly, we conducted a global admixture analysis, which allowed us to confirm the statistically significant decrease of African ancestry in the DHF cohort and to fine tune the association analysis in the admixed Cubans. Secondly, we performed a local ancestry assignment along chromosomes which identified the precise regions (and genes) where the African ancestry was significantly higher in asymptomatic compared with DHF. The fine-tuned association analysis can identify a few SNPs located in a small chromosomal region that confer reasonable individual risk/protection. We identified OSBPL10 and RXRA candidate genes, with most significant SNPs placed in inferred weak enhancers, promoters and lncRNAs. OSBPL10 had significantly lower expression in Africans than Europeans, while for RXRA several SNPs may differentially regulate its transcription between Africans and Europeans. Their expression was confirmed to change through dengue disease progression in Cuban patients and to vary with disease severity in a Thai transcriptome dataset. These genes interact in the LXR/RXR activation pathway that integrates lipid metabolism and immune functions, being a key player in dengue virus entrance into cells, its replication therein and in cytokine production. Knockdown of OSBPL10 expression in THP-1 cells by two shRNAs followed by DENV2 infection tests led to a significant reduction in DENV replication, being a direct functional proof that the lower OSBPL10 expression profile in Africans protects this ancestry against dengue disease.
We explored the admixture of the Thai dengue cohort and of a previously published Vietnamese DSS cohort4 in the nexus between South, Northeast and Southeast Asia. We observed a statistically significant increase in the Southeast Asian background in Thailand for both DF (4.1% increase; p-value = 1.25 x 10-7) and DSS (4.8% increase; p-value = 5.90 x 10-8) compared to Thai control. By applying a joint ancestry and association test (BMIX) analysis, we were able to replicate the two previously reported genes in the Vietnamese cohort4, MICB and PLCE1 genes. For the Thai DSS vs. control analysis, we obtained a reliable signal of six linked significant SNPs for PLCB4 (phospholipase C, beta 4), a gene in the same family as PLCE1, and participating in many common pathways, such as dendritic cell maturation, PI3K signalling in B lymphocytes and PPARA/RXRA activation.
When comparing Thai DF vs. control, a distinctive genetic signature was obtained. Three genes located on different chromosomes had at least two SNPs above the BMIX significant posterior probability threshold, forming haplotypes: CHST10 codes for carbohydrate sulfotransferase 10, AHRR codes for aryl-hydrocarbon receptor repressor, and PPP2R5E codes for protein phosphatase 2, regulatory subunit B', epsilon isoform. Interestingly, the proteins coded by these three genes, and by another gene, GRIP1 that codes for glutamate receptor interacting protein 1 are involved in the xenobiotic metabolism signalling pathway. Our findings suggest that xenobiotics and lipid metabolism, as well as interaction of virus proteins to these molecules and to its phosphatases, are critical in the development of DF, but not of DSS. DSS would require independent pathways involving MHC locus and phospholipase C gene family, being more dependent on the activation of immune cells and inflammation of blood vessels.
Risk of dengue in European population
European candidate gene allele/haplotype frequency
Response to infections has been proved to be one of the main drivers of selection in the human population, with different genetic diversity being positively selected in different populations along time, as pathogens emerge and become important agents in that specific geographical region. Thus, ethnicity is a major confounding factor in risk measuring. In order to fill a gap of lack of information, in a predictive way, for the naïve regions of Europe and North America, and for regions where registries/diagnosis mix-up predominates as in Africa, we performed a genetic risk evaluation at the worldwide scale to dengue fever.
We began by focusing on published candidate markers, ideally for which some functional information was available, ending up by selecting 13 candidate SNPs from 8 genes (CCR5, CD209, CD32, MICB, OAS1, OAS3, PLCE1 and TNFA). Around 1,312 individuals from 15 countries were genotyped, which added up to published data made up between 4,055 and 4,830 individuals of between 43 and 61 populations. When considering these 13 alleles in an additive model of genetic risk, all populations would have a similar risk to dengue disease, with, curiously, a lower risk of East and Southeast Asian and South American populations than the other regions of the world. In DENFREE, we were able to identify at least one gene (OSBPL10) that confers protection against dengue hemorrhagic fever, which is associated with African ancestry. European populations have the causative haplotype in high frequency, which increases the genetic risk of the hemorrhagic phenotype in these populations when considering only this gene.
The genome wide association study also conducted by us in Thailand indicated four genes (CHST10, AHRR, PPP2R5E and GRIP1) of the xenobiotic metabolic pathway associated with dengue fever phenotype, and the extra PLCB4 gene (Figure 6 – risk includes PLCB4 and previously associated genes PLCE1 and MICB) associated with severe dengue. Attending to these genes, European populations have a low risk of dengue fever when compared with East Asian populations, but the risk of the severe phenotype is as high in the two groups, with African ancestry protecting against the hemorrhagic and shock phenotype.

Figure 11-Genetic risk for the phenotype dengue fever based on additive model for the four genes (CHST10, AHRR, PPP2R5E and GRIP1) of the xenobiotic metabolic pathway

Figure 12- Genetic risk for the phenotype severe dengue based on an additive model for PLCB4, PLCE1 and MICB genes
In conclusion, the evidence collected so far for the candidate genes associated with dengue are indicating that European populations have probably a genetic risk as high as East Asian populations to dengue disease, for both dengue fever phenotype and severe form. And we were already able to confirm that African populations have genetic protection against the severe form, despite the risk to dengue fever being equivalent to other geographical regions.

(See figures in the “figures Final report DENFREE” document).
Potential Impact:
- WP1 Index case community study of dengue epidemiology
WP1 is an epidemiological workpackage with a major objective of identifying and recruiting dengue infected individuals for biological samples for other workpackages as well as testing diagnostic tests developed by colleagues in other WPs. The samples from the Thai study sites were used extensively and predominantly by the immunology WPs as well as the human genetics and virology WPs. The Cambodia study site not only contributed samples to the human genetics and virology WPs but also provided a study site to test diagnostic methods developed in the Diagnostics WP and to test the infectiousness of infected but asymptomatic individuals to mosquitoes. The burden of work of identifying asymptomatic infected individuals, followed up for several weeks, was huge but generated the possibility to deliver a major result: individuals who are infected but are either pre-symptomatic (asymptomatic at the time but who later became symptomatic) or completely asymptomatic (with no symptoms ever) are just as infectious to mosquitoes as individuals who have symptoms. This result and other results pertaining to the immunology, virology and human genetics studies are more amply described in the relevant WPs.

Finally, a detailed risk factor analysis was carried out and demonstrated that even in this rural community, a significant proportion of the force of infection of dengue was not occurring in and around the house of an infected individual. Given that it is children in this community who are at risk of infection, the other most likely place that they can become infected is at school. This has led us, in collaboration with the health authorities, to develop a protocol to strengthen school based mosquito control strategies to provide evidence that such mosquito control can significantly limit dengue transmission. If successful, this would provide the local health authorities with evidence that such mosquito control efforts can be implemented, are cost-effective and efficient in controlling dengue, despite the existence of asymptomatic but infectious individuals who cannot be readily detected in the community.

- WP2 New detection method
The newly developed diagnostic tools can be used to analyse the serotype, the peaking DENV-specific antibody response in patients. Such a test provides information of the immune status against DENV and can be helpful to diagnose which DENV is neutralized and which one is not. Such information cannot be concluded from the direct virus diagnostics since we have shown that the virotype is not matching the serotype in acute infections. This observation will have some impact on strategies fighting DENV infections. The test allows also to monitor such neutralizing antibody responses in human populations over time to find out against which DENV a population is mainly protected and to which DENV people might be at risk. A new diagnostic tool for surveillance studies.
Today several vaccines are in clinical trials. One problem is, that the vaccine is administered in endemic areas were people are primed against DENV by natural symptomatic and asymptomatic infection. This makes it difficult to evaluate vaccine efficiencies when vaccine candidates are grouped only by age. Especially in very young children (age 2-9 years) a vaccine was considerably less efficient. The effect of the vaccine was moreover contra productive since young children vaccinated developed a 2-3 times higher risk for DENV hospitalized cases. Thus, the DENV vaccine was only helpful when people were exposed to DENV and mosquitos for longer periods (> 9 years). This implicates that the vaccine can boost an established anti-DENV immune response but is not efficient without natural anti-DENV priming. Here our serotyping test could help to group vaccine candidates into groups defined by serotyping and not by age. This is one of benefits of our serotyping assays. Additionally, this test can also analyse not only the immune status before vaccination it can also be helpful to analyse the anti-DENV immune status during and after vaccination procedures. A vaccine inducing antibodies to DENV but not the type-specific neutralizing ones against the endemic DENV strains might be contra productive. This might be seen by the studies in young children not naturally primed against DENV. A vaccine that boost such a establish immune response will be helpful but must be controlled for efficiency against all four DENV. This can be achieved by the new diagnostic tools, our ED3 ELISA and ED3 dot assay.
The Test in general for DENV antibodies was developed together with AmpTec. They have all the rights to develop an antibody detection kit based on their isothermal amplification technology using ED1, -2 and -3 antigens.
Future plans are to test the single domains of the E protein as vaccine candidates. It is known that ED3 induces very type-specific immune responses that make it complicated to get full DENV protection by just one set of DENV-1-4 ED3 antigens. On the other hand ED1 was found to be DENV specific, not showing serotype specificity. Since we have constructed ED1 domains for DENV-1-4 by our Gly-Pro-Gly linking strategy we can now test ED1 and ED2 as single and separated domains as vaccine candidates against DENV.
The DENV pseudotyped HIV-1 particles are a new tool to study flavivirus infection and neutralization. Based on RT-PCR or sequence data DENV pseudotypes can be specifically tailored in any combination for any need. This is in principal not only a new method to study immune responses against DENV it is also a new tool for fundamental research in virology.

- WP3 Seasonal climate predictions for the dengue study sites
The building of a set of climate covariates-based operational products for dengue prediction was initiated and these are therefore possibly be available in the near future for interested public agencies, local users, and policy makers. A remarkable result obtained is the finding of a suite of locally-derived climate indicators that differ between temporal scales which has a possibly differential effect on dengue evolution in the region of interest. The indicators span from the rapid weekly intra-seasonal active/break monsoon stages, to the year-to-year regional and the projected changes span time ranges that occur at climate change scales. This multiple-scale climate framework can be very effectively incorporated into national adaptation and mitigation strategies to better fight against the threat of extensive arboviral disease outbreaks. WP3 has produced a calibrated climate datasets for Thailand and Cambodia that was of use by other WPs, and performed database inter-comparisons for the different climate products available to date. Additionally WP3 created a user-friendly tutorial (i.e. the “Climate Explorer tutorial” that can be found in the DENFREE website). The latter tutorial intends to illustrate how even a naive user, who is not familiar with climate datasets, can visualize and download different types of climate observations and simulations for research purposes. Moreover, work conducted in this WP helped to further define the role ENSO plays on southern Indochina’s climate through the study of the characteristics of simultaneous rainfall and near surface temperature variations, and it was found that rainfall anomalies over southern Indochina generally link to warm (cold) SSTAs in the central tropical Pacific, though the link is clearly evident between the decaying phases of both ENSO events and the corresponding rainfall response. Initial estimates of the seasonality and the monsoon dynamics that may be helpful to produce short lead time dengue incidence prediction in the region have been scrutinized, as well some hints have started to show at the seasonal and inter-annual scales, which means that a path leading to early-warning system for strategy and control purposes can be though feasible in the near future. Recent approaches taking advantage of the leading capacity to predict el Niño phenomenon within DENFREE by means of nonlinear and nonstationary statespace models allow to successfully anticipate dengue outbreaks in Ecuador more than two years in advance. Such an approach can be used in the next future to tune up further dynamical dengue models in Thailand and Cambodia (such as these produced in WP4) and facilitate in-between country comparisons of the impact of climate on dengue fever and contribute towards a more global approach.

- WP4 Descriptive and Predictive models

Optimal control methods for linear and quadratic cost functions
Optimal control theory is in mathematics most of the time performed by quadratic cost functions, historically due to the origin of the optimization coming from physics with quadratic energies extremized. However in economic terms and hence public health economics costs are always per patient I or per vaccine dose and not quadratic in I or the control variable u. On the other hand an optimal control strategy can be sought in terms of manageability when several control measures are possible but none of them on its own is sufficient to reduce the population infection level below the threshold to disease extinction (like it is reached by near perfect vaccines e.g. in the case of childhood diseases). Hence both ways of designing cost functions, economically or idealized for a certain goal, have some rational to be used, especially when we have like in the case of dengue quite imperfect vaccines and as well sub-optimal mosquito control measures, which have experienced little success up to now in combating dengue.
Using the SISUV model, which describes vector borne diseases with human hosts susceptible S, infected I and mosquitoes uninfected U and infected and acting as disease vector V with parameters suitable for Aedes mosquitos transmitting dengue and other flaviviruses. Due to the non-linearities not only in the disease dynamics but also in the control signal, we can obtain similar optimal control results for the above given quadratic cost function and for a more economically reasonable linear cost function. Such methods can now be used to evaluate optimal controls via vaccination and mosquito control where both controls individually might have their limitations.

Optimising mobility interventions to stem the regional spread of epidemics
Using network-SEIR fitted to the CHIKV outbreak in the Caribbean we make intervention actions by cutting the down the mobility by disconnecting the traffic airline regional network. It is clear that for any strategic plan of intervention there are costs associated to implementing the actions at different times. For example taking actions in the first stages of an outbreak could be costly if the prediction is not good enough and risk of having an epidemic is overestimated. Also taking actions early in an epidemic can be difficult because it takes time to coordinated the actions among different partners, the epidemic can be undetected during the very first stages, etc. On the other hand, the model shows too that if the actions are taken later and later in time the effectiveness of the control actions are less and less effective. At some time around day 130 after the day 0 of the outbreak, the model predicts a threshold, i.e. that one can do little, regarding cutting down commercial airplane, to help control the epidemic. At this threshold cutting traffic all way down to zero has the same effectiveness than performing no intervention at all and leave the traffic at its fullest and the network fully connected. By implementing these kind of models, by for example associating real calculated costs to realistic intervention actions, one can find optimal solutions that can bring together costs and benefits to the most favorable balance for public health results.

Climatic information for improving Public Health Decision making
Certain diseases are particularly sensitive to climate extremes. For example, a previous study found that the timing and magnitude of dengue outbreaks in El Oro province in Ecuador were associated with El Niño. In this study, long-lead forecasts of equatorial Pacific SST are used within a dengue prediction model, to assess the extent to which dengue epidemics can be predicted well in advance. The ENSO model could have helped to predict the dengue epidemic that occurred in the region in 2010 as early as 24 months ahead. Thus, long-lead ENSO forecasts could be incorporated into dengue prediction models, to enhance the development of a dengue warning system for Ecuador and other tropical and subtropical countries sensitive to the ENSO phenomenon.

- WP5 Asymptomatic transmission, European vector competent and innovative vector control strategy
The demonstration of the infectiousness of asymptomatic/inapparent infections has far reaching significance for vector control measures. We can no longer simply focus on transmission reduction around symptomatic cases. Our results will change significantly the current paradigm of dengue epidemiology as well as the response to a dengue outbreak. The asymptomatic people may behave as viral reservoirs ensuring the dissemination and maintenance of the virus.
The demonstration that European Aedes albopictus mosquitoes are as competent for Thai and the imported dengue viruses confirms our deepest suspicions that virus spread will be not be blocked by incompatibility issues. However, there were distinct differences among mosquito populations at a European scale, suggesting that the same expansion of the virus is unlikely to happen everywhere within Europe.
Artificial selection experiments revealed the rapid spread of a single nucleotide variant. This variant corresponds to a substitution in the subgenomic flavivirus RNA (sfRNA) region, of the untranslated 3’ end of the DENV genome. This sfRNA is highly structured at the secondary and tertiary levels, and has been involved in pathogenicity in the vertebrate host and escape of antiviral responses in mosquitoes. The impact of viral evolution on the rate of viral spread and outcome of disease is likely to be highly complex and significant for public health and will require constant surveillance.

- WP6 Viral genetic diversity and adaptation
NA

- WP7 Key immunological factors determining DENV clinical outcome

The discovery of the EDE opens up a number of interesting future possibilities in dengue vaccine research. Current vaccination strategies use tetravalent formulations with the aim of raising single serotype specific responses against all four serotypes. An overall efficacy of a larger scale Phase III trial of Sanofi Pasteur was less than 70% and it also rose a concern of enhancing infection. The demonstration that anti-EDE mAbs, highly potent, broadly neutralising antibodies, are produced in dengue infection, which can also potently neutralize ZIKV, means that the generation of such antibodies should be a goal for the next generation vaccines. Importantly, as the response is limited to the E-dimer it opens the way for subunit vaccines consisting of E-dimers alone and furthermore, it may be possible to design a single universal immunogen, rather than a multivalent formulation to achieve this response. Alternatively, heterologous prime boost strategies may be used to focus the response to the EDE, potentially following LATV priming.
The work led to a patent, 4 publications in Nature and Nature Immunology journals, 1 review in Nat Rev Immunology Journal, news in newspapers and conference presentations. Furthermore, the work is formed a basis of a grant application which has been awarded under Wellcome Trust Collaborative program.

- WP8 Role of antigen presenting cells, T cells and metabolic pathway in DENV infection outcome; Risk of dengue in European population

A role of T lymphocytes in the outcome of DENV infection has been debated. Our finding of significantly higher activation of T cells in asymptomatic DENV infection compared to clinical dengue supported a protective role of this T cell population. In addition to presenting this result in several international meetings including the 9th European Congress on Tropical Medicine and International Health: 6-10 September 2015, Basel, Switzerland, the International Conference on Dengue/DHF: 24-26, February 2016, Colombo, Sri Lanka, the manuscript is under revision for publication in Science Translational Medicine journal.
This finding has implication in dengue vaccine development. All dengue vaccines, either the licensed one or the ones under development, have focused on induction of neutralizing antibodies and avoiding ADE phenomenon. The result of recent dengue vaccine phase III trial suggested that ADE could occur during a long-term follow up when level of anti-DENV antibodies is waning, especially in young children. Most dengue vaccines lack T cell epitopes, which are enriched in the non-structural protein regions of the virus. Our finding suggested that dengue vaccine containing T cell epitopes could increase its protective effect. In addition, DENV specific T cells induced by the vaccine may circumvent the effect of ADE as they can eliminate viral infected cells while antibodies cannot. Together with the E-dimer epitope described above, we believe that our next generation dengue vaccine will be more efficient than the current one.
We thus designed a new “PolyNS” epitope, containing part of the NS regions of the virus, which are conserved among all four serotypes and strongly activate T cells (Patent EP-305984.8 filed on June 23, 2014). The patent is extended by demand of PCT/EP2015/064010, filed on June 22, 2015, and published as WO2015/197565 on December 30, 2015. We have shown that the PolyNS is highly potent in inducing DENV specific T cells in a mouse model and confer complete protection against DENV infection in interferon-α receptor knock-out mouse model (Ifnar1 -/-) when vaccinated with measles vaccine vector containing the dengue PolyNS epitope.
We obtained funding from the Global Health Innovative Technology fund of Japan (https://ghitfund.org/about/orgnization/mission) to develop the next generation dengue vaccine based on this concept. In addition, we are collaborating with several companies which have a vaccine delivery system known to stimulate T cells with their own funding. Two of the new dengue vaccine candidates are under investigation for its efficacy using non-human primate model. We are searching for funding to support good manufacturer practice (GMP) production, toxicology and clinical phase I trial. We will increase the number and the quality of dengue vaccine candidates, which will proceed into further development and clinical testing within the context of the European and Developing Countries Clinical Trials Partnership (EDCTP2).
The protection conferred by OSBPL10, RXRA and related lipid metabolism against dengue illness supported in this study points out potential therapeutic applications. The generation of synthetic ligands of LXR should be pursued, given its central role in dimerization with RXRA. The possible PI3P-dependent activation of OSBPL10 could be tested through the various PI3K inhibitors being developed in cancer treatment. In addition, the African protection against DHF through kinases supports pursuing the development of kinase inhibitors as they seem to be able to block DENV assembly.

- WP9 Project Management and Coordination
-
The DENFREE management tasks scheduled to be performed during the lifetime of the project (period 01/01/2012- 31/2/2016) were successfully conducted and completed in accordance with the provisions of Articles II.2.3 and II.16.5 of the Grant Agreement and of the Annex I–Description of Work.

Annual meetings with attendance of member of scientific advisory board as well as dissemination meetings were organized :

• The Kick-off meeting was organised on 23-24/01/2012 at KohChang, Thailand.
• The 2nd annual meeting was organised in Chateau Saint-Just, Belle-Eglise, FRANCE; (MS19);
• Following discussion amongst the PIs of the three FP7 Dengue consortia (IDAMS, DengeTools and DENFREE), it was decided that the consortia should meet and develop a strategy to work in collaboration. On 19-20/04/2012, DENFREE organised the first 3Dproject initiative meeting in Chateau des Ravatys, France. During this meeting the three consortia presented overviews of their projects and workshops were conducted in major areas of research (epidemiology/disease, entomology, virology.
• The 3rd annual meeting was organised in Cambodia; the agenda and list of participants were prepared and communicated to the consortium before the meeting (MS22).
• The 4th annual meeting was organised on 31 January 2016 to 2nd February 2016 in Siem Reap, Cambodia. All Beneficiaries attended. This meeting allowed consortium members to exchange information and results on the project.
• The final meeting, “the final stakeholder meeting” organized by thee IDAMS and DENFREE consortium in collaboration with WHO-TDR took place at the Institut Pasteur, Paris on the 21-23 November 2016. This joint meeting was part of a common dissemination strategy from these two projects.

The specific project website (denfree.eu) serving as a communication tool among the DENFREE Consortium members was regularly updated throughout the project based on inputs provided by the consortium. In September 2016, our partner in charge of the maintenance of DENFREE.EU informed us that the hosting had been hacked and our page taken down. We thus decided to move the website to www.DENFREE.org .

The DENFREE Consortium has also initiated dissemination mechanism to promote the project to the general public, through press released (example : press realeased on asymptomatic people transmit the virus to mosquitoes) or training sessions in collaboration with the BMA teams. In fact, a series of meetings has been jointly organised with BMA Health Department in order to introduce our research project and methodologies to local teams of health workers, public health specialists and community health workers in Thailand. The consortium has also developed toward political stakeholders. On Wednesday April 22nd, 2015 a meeting was held at the office of Dr. Pusadee Tamthai, Deputy Governor of Bangkok. The purpose of this meeting was to discuss forthcoming fieldwork operations in Bangkok, and future collaborations between the different partners in the field of mosquito control and epidemiological surveillance.
List of Websites:
www.denfree.org