Ae. albopictus and Ae. aegypti were sampled in cities, villages, natural reserves and forests in eight regions of Madagascar (Figure 1). The characteristics of the sampled areas are summarized in Table 1 and Figure 2. The Analamanga region is located in the highlands of Madagascar at an altitude of 1200-1433 m, with a tropical climate. In this area, two seasons are observed: a hot (21°C on average) and rainy period (November to April with about 200 mm precipitation in a month), and a relatively dry cool season for the rest of the year (10°C on average and rainfall not exceeding 20 mm precipitation in a month). The Zoological and Botanic Park of Tsimbazaza is located in the Center of Antananarivo at 1200 m altitude. As Analamanga region, Amoron'i Mania region (Ambositra) and Haute Matsiatra region (Fianarantsoa) are also in the highlands of Madagascar and have similar features. Ranomafana forest is situated in a region between Amoron'i Mania and Vatovavy Fitovinany in the western side of Madagascar. The climate is tropical, humid and rainy (with an annual rainfall of 2600 mm) and the temperature varies from 14 to 20°C. The Antsinanana region (Toamasina, Foulpointe) is in the eastern plains of the island. The climate is hot and wet and rains occur all the year with the mean annual rainfall of about 3200 mm and a relative humidity around 87%.

The Boeny region (Mahajanga) is located in the western plains. It has an equatorial climate with an arid and warm summer (mean temperature of 27°C) with moderate rainfall (the mean precipitation is about 400 mm per year). The Menabe region (Morondava, Kirindy forest) is also situated in the western plains. It has a similar climate to the Boeny region. However, the temperature can reach more than 40°C during the rainy season (late December until the end of March). Rainfall is not constant during this period but spaced with dry periods.

The Vatovavy-Fitovinany region (Mananjary) is in the southwest coast of Madagascar and has a tropical climate along the coast, is temperate inland and arid in the south. The Diana region is situated in the north-western plains of Madagascar, where the sampling sites of Diego-Suarez and Ambanja cities are located, as well as the reserved forests of Ankarana and Montagne d'Ambre. The climate is equatorial dry and this area is situated between 4 m altitude (Ankarana) and 1400 m (Montagne d'Ambre). It is characterized by an alternation of wet and dry seasons (May to November) and a hot and humid season (December to April). The average temperature is 27°C and rainfall varies from 985 to 2171 mm per year.

Adults and immature stages of mosquitoes were collected monthly from October 2007 to October 2009 in 8 regions corresponding to a total of 15 different sites (Figure 2). In each site, collections were performed daily over a period of one week by visiting three to four spots previously identified as hosting adult breeding sites and larval habitats of Aedes (Table 1). The same spots were visited, unless changes occurred due to human activities, thus obliging capturers to search for other potential new breeding sites and habitats in the area. As Aedes albopictus and Aedes aegypti are diurnal mosquitoes, adults were collected during the two peaks of biting activities, i.e. 7:00 to 10:30 am and 15:30 to 18:00 pm. Two methods were used to catch adult mosquitoes. The first method consisted of the use of butterfly nets that allows collection of both females and males flying near the grass. This was done by applying ten rounds of netting near the grassland and underneath bushes, bamboo and fruit trees. Then, capturers moved to another site. The second method was performed using oral aspirators to capture host-seeking females landing on the legs of human volunteer-baits who gave their agreement. Malagasy administrative authorities, particularly Madagascar National Parks (formely ANGAP), were informed and approved mosquito collections which were conducted in conformity to the declaration of Helsinky on ethical principles for medical research involving human subjects. To avoid biting, three capturers performed collection at one time. Exposed legs of one volunteer were under monitoring of two non-exposed capturers who catch mosquitoes immediately after landing. When trapping in an area, capturers spent 15 to 30 min at one spot then moved to another. Each spot was visited twice during each peak of biting activity. Collected adult mosquitoes were stored in cups covered with netting. Aedes spp. specimens were identified morphologically 1 , then males and females were separately desiccated, in silicagel. In order to keep some individuals alive during transfer to the laboratory, impregnated cotton with 6% sucrose solution was placed at their disposal. Larval habitats consisted of artificial containers and natural sites containing standing water (Table 1). When larvae and pupae were present, they were collected using a dipper and transferred to a plastic bottle using a wide-mouthed pipette. Collected specimens were brought back to the laboratory where they were reared until adult stage and then identified. Usually, larvae and pupae were sampled at the same time as adults. For each species, the number of individuals, date of capture and location site were recorded.

Wild-caught females were blood-fed on chickens and allowed to lay their eggs on absorbent paper in insectaries under standard conditions (28°C ± 2°C, 80% ± 12% humidity, photoperiod: 12 h/12 h) in Madagascar. Batches of eggs were brought to biosafety insectaries in the Pasteur Institute in Paris using a safety transport procedure. After hatching, larvae were reared to adults in standard conditions as described 20 .

To infect mosquitoes, two independent experiments were carried out at an interval of three months with the same batches of eggs. The strain CHIKV (E1-226 V), which has an A- > V amino acid substitution at position 226 in the E1 glycoprotein, was used 8 . Virus stock and mosquito infection were performed as described 9 20 . Briefly, 1 mL of viral suspension was added to 2 mL of washed rabbit erythrocytes supplemented with ATP (5 × 10^-3 M) as a phagostimulant. The resulting infectious blood at a titer of 10^7.5 PFU/mL was transferred to a glass feeder at 37°C on top of the mesh covering a plastic box containing female mosquitoes from each of the six populations that were previously starved for 24 h. After 15 min of oral feeding, mosquitoes were sorted on ice. Fully engorged females were transferred to cardboard containers then fed with 10% sucrose at 28°C for 14 days. After the incubation period, disseminated infection rates of CHIKV were determined (using surviving live mosquitoes), by immunofluorescence assay (IFA) on head squashes as described 21 . In brief, mosquito heads were placed between two glass slides and squashed by hand pressure. Squashed tissues were fixed by immersing in acetone for 20 min at 20°C, then incubated with a first antibody, anti-CHIKV diluted in PBS 1X (1:200); the antibody was obtained from a mouse ascite. After incubation for 30 min at 37°C, slides were washed three times in PBS 1X and incubated with an anti-mouse conjugate diluted in PBS 1X (1:80) and supplemented with Evan blue. Slides were observed under an epifluorescence microscope (Leitz, Larbolux K). In positive samples, the head tissue appeared in green whereas blue-colored tissue is negative. Therefore, the disseminated infection rate corresponds to the ratio between positive heads to the total number of surviving individuals examined.

To extract genomic DNA, five males and five females from each site were washed three times with sterilized water, then surface-disinfected with ethanol 80% for 5 min. After five washes with sterilized water, each individual mosquito was homogenized and the genomic DNA extracted using the procedure previously described 22 .

For PCR, amplification was performed using a T Gradient Thermocycler (Biometra, France) with the primers targeting two mtDNA gene fragments: a 597-bp fragment of COI (cytochrome-oxydase subunit 1) and a 450-bp fragment of ND5 (NADH dehydrogenase subunit 5). The two sets of primers used were: for COI, CI-J-1632 (5'-TGATCAAATTTATAAT-3') and CI-N-2191 (5'-GGTAAAATTAAAATATAAACTTC-3') 23 ; and, for ND5, ND5FOR (5'-TCCTTAGAATAAAATCCCGC-3') and ND5REV (5'-GTTTCTGCTTTAGTT-CATTCTTC-3') 24 . Reactions were made in 50 μl volume containing 90 ng of DNA template, 1× PCR buffer, 50 mM MgCl2, 10 μM of each primer, 2 mM of dNTP mix (INVITROGEN, France), 0.4 U of Taq DNA polymerase (INVITROGEN, France). The temperature profile for ND5 consisted of an initial denaturation at 98°C for 2 min, followed by 5 cycles of 95°C for 30 s, 45°C for 30 s, 72°C for 45 s, then 25 cycles of 95°C for 30 s, 46°C for 45 s, 72°C for 45 s, and a final extension at 72°C for 5 min. For COI the amplification program consisted of an initial denaturation at 95°C for 5 min, followed by 35 cycles of 97°C for 30 s, 40°C for 45 s, and 72°C for 1 min, and a final extension at 72°C for 5 min 25 . An aliquot of 10 μl of each PCR product was subjected to electrophoresis on a 1% agarose gel at 50 V for 30 min, then stained with ethidium bromide and photographed with Gel Doc 2000 system (BioRad, USA). When bands with expected size were visualized, the remaining PCR products (approximately 40 μl) were sent to sequencing at BIOFIDAL-DTAMB (FR BioEnvironment and Health, Lyon, France).

The COI and ND5 gene fragments were sequenced in both strands and deposited in Genbank with accession numbers JN406654-JN406797, JN406804-JN406809, JN406822-JN406832, JN406839-JN406844. Sequences were then aligned with those of mosquitoes available in databases by using BioEdit and Multialn softwares. The two gene sequences were concatenated to improve the reliability of the phylogenetic analysis. Phylogenetic analysis was carried out by using the Seaview software (http://pbil.univ-lyon1.fr/software/seaview.html) based on maximun likelihood (ML), maximum parsimony (MP) and neighbor-joining (NJ) methods. Trees were then constructed with the general time reversible (GTR) model, and branch supports were estimated by bootstrapping with 1000 replicates.

The effects of the sampling region, year, period, main habitat type and presence of one mosquito species to the abundance of the other species were investigated using linear regression models. The variable "period" corresponded to subdivision in a three-months sampling period per year, which can be linked roughly to a more rainy season from October to March and a more dry season from April to September. The variable "area" was categorized as highlands, eastern plains, western plains, southeast coast, north plains and north east (Table 1). All possible models (with all subsets of variable numbers and combinations) were generated. A square transformation was applied to ensure the normality of the residuals. Due to the sample size, only the region*year interaction was tested. Area and region were never put together in a model because they were correlated. The most appropriate model was selected using the Akaike Information Criterion adjusted for small sample size (AICc, 26 ). The models were ranked according to the smallest AICc differences (denoted ΔAICc) between the focal model and the lowest AICc model. When ΔAICc was larger than 2, the model with the smallest AICc was selected. On the contrary, when ΔAICc was smaller than 2 and models nested, the most parsimonious model was kept. When models were not nested but AICc smaller than 2, both models were kept.

Immature stages and adults of Ae. albopictus predominated in most sampled sites (13 out of 15). During the 27 months of sampling, a total of 12,639 adults (9891 females and 2748 males) and 7891 immature specimens of Ae. albopictus were collected (Table 2). Aedes albopictus was found in diverse ecological areas, from highland to the coast and under contrasting climates from arid to humid. Breeding sites and habitats included natural and human-made environments. In contrast, the species Ae. aegypti was found in very limited numbers as imago (269 females and 46 males) and was mostly confined to sylvatic zones (Ranomafana, Kirindy, Ankarana and Montagne d'Ambre forest), with only a few specimens also captured in Diego Suarez city (Table 3). In some areas, the current distribution of the two species was significantly different from what was recorded in the 1980s (Figure 1). For instance, in Ambanja town (Diana region), the two species were reported to be sympatric, but during our investigation, only Ae. albopictus was found. This coincided with the extension of the urbanization zone and an increase of man-made breeding sites. Interestingly, a converse pattern was observed in neighbouring villages of Ankarana where the sympatry was substituted by the allopatry of Ae. aegypti. In the same line, Ae. aegypti has recently colonized the natural reserve of Ranomafana, whereas Ae. albopictus has extensively invaded the cities of Mahajanga (Boeny region) and Morondava (Menabe region).

As Ae. albopictus has expanded since the 1980's in Madagascar and now predominated in most sampled sites, further analyses were focused on that species. Statistical analysis showed that the abundance of Ae. albopictus was best described by a linear regression model including a regional effect (F = 62.00, df = 7.58, p < 2.2 × 10^-16) and a period effect (F = 36.22, df = 3.58, p = 2.548 × 10^-13) (Table 4). There was an increasing abundance in sequential order Haute Matsiatra, Amaoron'I Mania, Menabe, Boeni, Diana, Analamanga, Vatovavy Fitovinany, and Antsinanana, whereas less Ae. albopictus was captured from July to September and April to June, that corresponded to more dry season, than from January to March and October to December (Table 5). There was no significant variation of Ae. albopictus from year to year during the study period (Table 6).

The sequences of two genes (449 bp for ND5 and 597 bp for COI) were obtained from 80 Ae. albopictus (5 females and 5 males for each site) and were aligned with similar sequences retrieved from the Genbank database (Figure 3). The alignment revealed substitutions consisted of both transitions and transversions (not shown). The tree built from concatenated sequences (1046 bp) identified two well-defined groups (Figure 3), with a strong support bootstrap value (84%). One group consisted of specimens from South America (Brazil) and South-Asia (Cambodia, Thailand, Vietnam), and the other clustered all the sequences from Madagascar together with those of the Indian Ocean (La Reunion), North America (USA and Hawaii) and Europe (France). In Madagascar, the data of COI and ND5 separately (not shown) or concatenated (Figure 3) did not cluster Ae. albopictus specimens according to neither geographical location nor urban, suburban and sylvatic habitats. For example, individuals from Montagne d'Ambre forest (North West) were not distinguishable from those collected in Toamasina city (Eastern plains) or Ankazobe village (Highlands). As females have a dispersal behaviour driven by the search for oviposition sites, which are in turn influenced by natural or domestic environments, a phylogenetic analysis was performed according to sex. However, similar tree topologies were obtained (not shown). Sequences were intermixed between individuals sampled in contrasted ecological niches (bamboos, bushes, fruit trees, containers, tires etc.).

Among eggs of the 13 Ae. albopictus populations transported from Madagascar to France, only six successfully hatched and gave adults for vector competence testing. The susceptibility of each population to infection was evaluated using artificial infectious CHIKV blood-meal. Overall, from a total of 1018 females tested in two trials, 1005 individuals (98.7%) were diagnosed positive as measured by IFA on head squashes (Table 7). No fluorescent signal was found in mosquitoes engorged with non-infectious blood-meal as expected. At the population level, the disseminated infection rate ranged from 94 to 100%.