Urbanization is a main driver for the larval ecology of Aedes mosquitoes in arbovirus-endemic settings in south-eastern Côte d'Ivoire

Background Failure in detecting naturally occurring breeding sites of Aedes mosquitoes can bias the conclusions drawn from field studies, and hence, negatively affect intervention outcomes. We characterized the habitats of immature Aedes mosquitoes and explored species dynamics along a rural-to-urban gradient in a West Africa setting where yellow fever and dengue co-exist. Methodology Between January 2013 and October 2014, we collected immature Aedes mosquitoes in water containers in rural, suburban, and urban areas of south-eastern Côte d’Ivoire, using standardized sampling procedures. Immature mosquitoes were reared in the laboratory and adult specimens identified at species level. Principal findings We collected 6,159, 14,347, and 22,974 Aedes mosquitoes belonging to 17, 8, and 3 different species in rural, suburban, and urban environments, respectively. Ae. aegypti was the predominant species throughout, with a particularly high abundance in urban areas (99.374%). Eleven Aedes larval species not previously sampled in similar settings of Côte d’Ivoire were identified: Ae. albopictus, Ae. angustus, Ae. apicoargenteus, Ae. argenteopunctatus, Ae. haworthi, Ae. lilii, Ae. longipalpis, Ae. opok, Ae. palpalis, Ae. stokesi, and Ae. unilineatus. Aedes breeding site positivity was associated with study area, container type, shade, detritus, water turbidity, geographic location, season, and the presence of predators. We found proportionally more positive breeding sites in urban (2,136/3,374, 63.3%), compared to suburban (1,428/3,069, 46.5%) and rural areas (738/2,423, 30.5%). In the urban setting, the predominant breeding sites were industrial containers (e.g., tires and discarded containers). In suburban areas, containers made of traditional materials (e.g., clay pots) were most frequently encountered. In rural areas, natural containers (e.g., tree holes and bamboos) were common and represented 22.1% (163/738) of all Aedes-positive containers, hosting 18.7% of the Aedes fauna. The predatory mosquito species Culex tigripes was commonly sampled, while Toxorhynchites and Eretmapodites were mostly collected in rural areas. Conclusions/significance In Côte d’Ivoire, urbanization is associated with high abundance of Aedes larvae and a predominance of artificial containers as breeding sites, mostly colonized by Ae. aegypti in urban areas. Natural containers are still common in rural areas harboring several Aedes species and, therefore, limiting the impact of systematic removal of discarded containers on the control of arbovirus diseases.


Introduction
Several Aedes species act as vectors of arboviral diseases, such as yellow fever, dengue, chikungunya, Rift Valley fever, and Zika virus infections that are of considerable public health relevance [1]. The transmission patterns of these arboviruses and their geographic expansion are expected to change due to environmental transformation, including urbanization [2,3]. Besides yellow fever, other arboviruses are likely underestimated and underreported in Africa because of low awareness by health care providers, other prevalent non-malarial febrile illnesses, lack of diagnostic tests, and absence of systematic surveillance [4]. Nevertheless, yellow fever, dengue (DENV1-4), chikungunya, and Zika viruses are currently circulating in West Africa through the sylvatic, rural, and epidemic cycles maintained by wild and urban vectors [5,6]. Côte d'Ivoire has been repeatedly facing yellow fever and dengue outbreaks involving several vectors such as Aedes africanus, Ae. furcifer, Ae. luteocephalus, Ae. opok, and Ae. vittatus in rural, and Ae. aegypti in urban areas [7,8]. These outbreaks have often occurred in foci characterized by high rate of urbanization due to economic development supported by palm oil and rubber farming, trade, and traffic [7]. Arboviral disease transmission is influenced by community-level effects of container-dwelling Aedes mosquito larvae by regulating the production and fitness of adult vectors [9]. Aedes mosquito larvae are highly sensitive to environmental changes, including urbanization [10]. Some Aedes species (e.g., Ae. aegypti) inhabit a wide variety of containers ranging from natural containers (e.g., tree holes) to artificial containers (e.g., tires, discarded items, and other water containers) due to their ecologic plasticity [11], while others are restricted to specific breeding sites because of the higher sensibility of their offspring to environmental changes [12]. The ecologic plasticity allows Ae. aegypti and Ae. albopictus to spread worldwide by sea, air, and land transportation networks, and to adapt to new and changing environments [10].
The choice of breeding sites is governed by competition and predation among immature stages of Aedes and other mosquitoes that co-exist in the same breeding site [11,12]. For example, intra-and interspecific competition between Ae. aegypti and Ae. albopictus [13] and among several Aedes species [12] has been reported. Moreover, mosquito species such as Toxorhynchites spp., Eretmapodites spp., and Culex tigripes predate on the larvae of Aedes [12,13]. The biotic factors may also interact with abiotic factors, such as the climate [13]. As larvae directly depend on water, precipitation is the most important physical factor. The complex patterns of flooding and drying of larval breeding sites govern arboviral transmission [14].
In Côte d'Ivoire, yellow fever has been a key factor that forced the transfer of the colonial capital from Grand-Bassam to Bingerville near Abidjan in 1899 [15]. However, more than a century later, yellow fever and dengue outbreaks still remain an unresolved public health issue [7,8,15]. During arbovirus epidemics, vector controls are mostly based on the systematic removal of artificial Aedes breeding sites in urban areas.
The most effective vector control strategy is the control of immature stages in their aquatic habitats [12]. Hence, effective larval control requires a deep understanding of larval ecology. Our study aimed to characterize the dynamics of Aedes larval breeding sites, species composition, and biological associations in terms of geographic and seasonal variations along a ruralto-urban gradient in south-eastern Côte d'Ivoire. As Aedes mosquito larvae are highly sensitive to environmental changes [10], we hypothesized that larval breeding sites differ in species composition between urban and rural areas.

Ethics statement
The study protocol received approval from the local health and other administrative authorities. All entomologic surveys and sample collections carried out on private lands or private residential areas were done with the permission and written informed consent of the residents. This study did not involve endangered or protected species.
Natural and artificial containers such as tree holes, bamboo, fruit husks, tires, discarded items, and water storage receptacles that may serve as potential breeding sites for Aedes mosquitoes vary according to human habitation and activities. The rural area is surrounded by farms of palm oil trees (Eleasis guineensis) covering 11,444 ha and a preserved rainforest of 100 ha, while the suburban area is located about 2 km away from the Banco National Park with over 3,750 ha of rainforest. The rainforest is inhabited by a diverse fauna (e.g., primates and birds) that serve as hosts for Aedes mosquitoes.
The climate is characterized by high temperature and precipitation with two rainy seasons. The seasons are distinguished by rainfall rather than temperature. The main rainy season extends from May to July, while the shorter rainy season occurs from October to November, with distinct dry seasons in between. The average annual precipitation ranges from 1,200 to 2,400 mm. The annual average temperature and relative humidity are around 26.5˚C and 80-90%, respectively. The larval breeding sites of Aedes mosquitoes were monitored in three areas: Ehania-V1 (A), Blockhauss (B), and Treichville (C), representing rural, suburban, and urban settings, respectively. The study site of Ehania-V1 includes the villages of Ehania-V1 and Akakro and represents the rural area without major and secondary paved roads. The site is in close proximity to a primary rainforest. The study site of Blockhauss covers the villages of Blockhauss and Petit-Cocody and represents the suburban area with only secondary paved roads. It is about 5 km away from the rainforest of the Banco National Park. The study site of Treichville comprises the sections of Jacques-Aka and Biafra and is an urban area with numerous major and secondary paved roads. It is located in the center of Abidjan and is separated from Blockhauss by the Ebrié lagoon.

Study design
Aedes larval breeding sites were sampled quarterly in domestic (space inhabited by humans) and peri-domestic (surrounding vegetated environment within a 600 m radius from the domestic areas) sites in rural, suburban, and urban areas from January 2013 to October 2014. While waterholding containers, tree holes, and bamboo were repeatedly sampled, other potential breeding sites were sampled for the presence of immature stages of Aedes mosquitoes. All accessible properties were surveyed simultaneously in the three settings. Some properties could not be sampled because the residents refused to provide access or because there were physical barriers of access.

Characterization of Aedes breeding sites
Potential larval breeding sites of Aedes mosquitoes were sampled in all three study sites by teams consisting of four trained mosquito collectors in each study area. Each mosquito collector team was composed of the same persons during all surveys. The number of experienced mosquito collectors was constant on any one day in each study area, whereas the teams made rotations from one study area to another in order to ensure similar sampling efforts and efficiency in the three study areas, and minimize potential biases. The collectors worked from 08:00 to 16:00 hours, and spent proportionally equal time periods searching for potential mosquito breeding sites in the study areas.
Readily visible and accessible containers in the selected households and surrounding premises were examined for the presence of water and mosquito larvae. In a preliminary survey, existing larval breeding sites, such as natural and artificial cavities or containers with a potential to contain water were kept in an inventory and assigned a unique label. Based on this preliminary survey, potential breeding sites were classified into two categories, three sub-categories, and 16 types, depending on their location, origin, material, and container type (Table 1 and S1 Fig). The breeding sites were assessed for abiotic and biotic characteristics, including geographic location (domestic and peri-domestic sites), color, exposure to sunlight (full shade, no exposure to sunlight; partial shade, partial exposure to sunlight; no shade, permanent exposure to sunlight), turbidity (transparent/clear, colored, opaque), substrate type (no substrate, foliage, moss, soil), surface of water, depth, presence of mosquito larvae, and predators (larvae of Cx. tigripes, Eretmapodites spp., and Toxorhynchites spp. mosquitoes, toad tadpoles, and arachnids).

Mosquito sampling
Larvae and pupae of Aedes mosquitoes were sampled using the World Health Organization (WHO) standard equipment adapted to the aperture and the depth of larval habitats. A flexible collection tube connected to a manual suction pump was used to sample water from bromeliads and bamboo holes. Scoops of 350 ml capacity were used to collect immature mosquitoes from larger breeding sites (e.g., tree holes, discarded containers, tires, and puddles). The collected Aedes mosquito were counted using a pipette and classified as young larvae (1-2 instar), old larvae (3-4 instar), and pupae. Non-Aedes mosquito larvae such as Anopheles spp., Coquelitidia spp., Culex spp., Eretmapodites spp., Filcabia spp., Toxorhynchites spp., and Uranotenia spp. were also recorded. The predacious larvae of mosquitoes, such as Cx. tigripes, Eretmapodites spp., and Toxorhynchites spp. were removed from the samples to avoid predation on the other species and preserved separately. All mosquito samples were stored separately in plastic boxes and transported in a coolbox to a field laboratory.

Laboratory procedures
In the laboratory, mosquito larvae were reared until they reached the adult stage. In order to minimize mortality, a maximum of 20 larvae were placed in 200 ml plastic cups, filled with 150 ml distilled water and covered with netting. Larvae of Aedes and other mosquitoes were fed each morning between 07:00 and 08:00 hours with Tetramin Baby Fish Food. Predacious larvae of Toxorhynchites spp. and Cx. tigripes were fed with larvae from colonies sampled from the study areas. Emerging adult mosquitoes were identified to species level using a morphological key [17]. As larval mortalities were low, the proportion of mosquito species was estimated on the basis of emerging adults. Adult specimens were stored by species and recorded in an entomology collection database.

Statistical analysis
The frequency of Aedes-positive breeding sites (FP) was calculated as the percentage of water holding containers with at least one larva or pupa (numerator) among the wet containers (denominator). The proportion of Aedes-positive breeding site types among the Aedes-positive breeding sites (PP) was expressed as the percentage of each Aedes-positive container type (numerator) among the total Aedes-positive containers (denominator) in each study area. To test whether there was a difference in the number of positive breeding sites and the number of available wet containers in each category, we used Fisher's exact test and χ 2 , as appropriate, to test for differences in the frequency of Aedes-positive breeding sites across the three study areas, between the domestic and peri-domestic sites, and between dry and rainy seasons.
Aedes species proportions were calculated as the percentage of specimens belonging to the genus Aedes for each study area and then compared between breeding sites as above. Larval abundances of Aedes mosquitoes were standardized as the mean numbers of larvae per liter of water, expressed as the geometric mean, known as Williams' mean (i.e., log[number of mosquito larvae + 1]) [18], and compared using the Kruskal-Wallis test, followed by Mann-Whitney. The Mann-Whitney U test was also performed to compare pairs of study areas when the Kruskal-Wallis H test showed a significant difference or only two habitats. Aedes species richness was defined as the number of collected species in each study area and compared using a one-way analysis of variance (ANOVA), followed by the Tukey post-hoc test for post-hoc pairwise comparisons [19]. Aedes species diversity and dominance were estimated using the Shannon-Weaver index [20] and Simpson index [21] and analyzed using a Kruskal-Wallis test. Kruskal-Wallis test was performed because a test for normality showed a significant difference in the variances after log-transforming the data. A significance level of 5% was set for statistical testing. All statistical analyses were conducted using Stata version 14.0 (Stata Corporation; College Station, TX, United States of America). Table 2 shows the species composition of adult mosquitoes that emerged from the larvae and pupae sampled from the breeding sites along the rural-to-urban gradient in south-eastern Côte d'Ivoire and reared after transfer to the laboratory. In total, 7,661, 16,931, and 26,968 adult mosquitoes emerged from the collected larvae in rural, suburban, and urban areas, respectively. The rural setting had the highest mosquito species diversity (eight genera and 37 species), followed by the suburban setting (four genera and 14 species), and the urban setting (three genera and nine species). The genus Aedes predominated throughout, with proportions of 80.40% (n = 7,661) in rural, 84.75% (n = 16,931) in suburban, and 85.19% (n = 26,968) in urban settings. The rural setting had the largest number of Aedes species (17 species), followed by the suburban (eight species) and urban settings (three species).

Mosquito species composition
The predacious mosquito species Cx. tigripes was sampled in each of the three study settings, while the predators Eretmapodites chrysogaster, Er. inornatus, and Toxorhynchites brevipalpis were primarily collected in rural settings. Moreover, several other vector competent mosquito species, namely Anopheles coustani, An. gambiae, Coquelettidia fuscopennata, Cx. quinquefasciatus, and Cx. poicilipes were sampled. Table 3 summarizes the species composition of Aedes mosquitoes collected as larvae among different types of breeding sites in the rural, suburban, and urban areas. Ae. aegypti and Ae. vittatus were commonly encountered in the three settings. Ae. aegypti was the most prevalent species in the all study areas, and exhibited rising abundance from rural (n = 6,159; 75.12%) to suburban (n = 14,347, 93.94%), and urban (n = 22,974, 99.37%) areas. The highest prevalence of Ae. vittatus (5.18%) was found in suburban areas. In rural areas, Ae. furcifer (4.53%), Ae. palpalis (3.96%), Ae. dendrophilus (3.83%), Ae. vittatus (2.83%), Ae. africanus (2.31%), Ae. were collected at frequencies above 1%. We also found two specimens of Ae. albopictus (0.01%) in the urban settings. The presence of Aedes mosquito larvae in breeding sites significantly varied by species (Table 3). For example, Ae. aegypti were found in all types of Aedes-positive breeding sites sampled in all the three study areas. Moreover, Ae. dendrophilus, Ae. furcifer, and Ae. luteocephalus were found in all container types in the rural areas, while Ae. vittatus and Ae. metallicus were collected from both natural and artificial containers in the suburban areas. Ae. africanus, Ae. lilii, Ae. unilineatus, and Ae. usambara were mostly present in natural containers such as tree holes, bamboo, and fruit husks in rural settings.

Associations among different Aedes species
Several species were found together in the same breeding sites. For example, Ae. aegypti, Ae. dendrophilus, Ae. furcifer, and Ae. africanus shared the same breeding sites in the rural areas, whereas Ae. aegypti co-existed with Ae. vittatus in suburban settings (n = 1,295, 12.8%). These two species co-occurred, albeit at low frequency (n = 57, 0.3%) in urban breeding sites. Additionally, Cx. quinquefasciatus and An. gambiae were often collected together with Ae. aegypti in tires and discarded containers in peri-domestic environments in the three study areas.
Mosquito predators, such as Cx. tigripes, Er. chrysogaster, and Tx. brevipalpis were found in the same breeding sites as Ae. aegypti, Ae. dendrophilus, Ae. furcifer, and Ae. africanus in rural settings. These ecologic associations were most present in tree holes, discarded containers, and tires in the rural areas and in peri-domestic breeding sites during the rainy season.

Dynamics of Aedes breeding sites
Fig 2 shows that the Aedes-positive microhabitat rate varied widely from one breeding site type to another in all three areas. The rural area showed the largest variability in Aedes breeding sites grouped into 16 types, followed by the suburban and urban areas presenting 15 and 12 microhabitat types, respectively. S1  Urbanization and Aedes mosquito larval ecology stages of Aedes mosquitoes in the urban than in the suburban (χ 2 = 17.3, df = 1, p < 0.001) or rural settings (χ 2 = 57.3, df = 1, p < 0.001). Furthermore, there was a statistically significant difference in Aedes mosquito positivity rate in water storage container between the suburban and rural settings (χ 2 = 15.8, df = 1, p < 0.001). Besides the variations in the frequency in the colonization of Aedes breeding sites, the most abundant Aedes breeding sites were tires and discarded containers in all the study areas (all p < 0.05) (Fig 3). Also frequently positive were natural breeding sites such as tree holes (62/738, PP = 8.4%), fruit husks (59/738, PP = 8.0%), industrial containers such as tarps (41/738, PP = 5.6%), vehicle tanks (41/738, PP = 5.6%), and vehicle carcasses (68/738, PP = 9.2%) in the rural area, and water storage containers (141/2,136, PP = 6.6%) in the urban area (Fig 3). Table 4 summarizes the abundance, richness, diversity, and dominance of Aedes mosquito species according to the breeding site types among sites and study areas. The Shannon's diversity and Simpson's dominance indices highly varied between the study areas and breeding sites, showing higher overall values in peri-domestic environments. The highest larval abundances of Aedes mosquitoes were recorded in tires in all study areas (p < 0.05). In addition, tree holes and metallic pots in the rural, vehicle tanks and building tools in the suburban, and discarded containers, vehicle tanks, and vehicle carcasses in the urban areas were also highly productive breeding sites for Aedes mosquito (S2 Fig). Aedes species richness was significantly different among the microhabitats in the rural (F = 4.3, df = 16, p < 0.001), suburban (F = 9.2, df = 7, p < 0.001), and urban settings (F = 11.1, df = 2, p < 0.001). Significantly higher numbers of species (13 species) were found in tree holes in the rural areas. The rural areas showed the  Geographic shifts in Aedes breeding sites

Discussion
When designing strategies to monitor and control Aedes arbovirus vectors in their breeding sites, failure to identify the broad spectrum of potentially available breeding sites will bias the results from field sampling and will thus negatively affect the impact of larval control interventions. Our study pertaining to larval habitats of Aedes mosquitoes alongside a rural-to-urban gradient within yellow fever and dengue co-endemic areas in the south-eastern part of Côte d'Ivoire provided strong evidence for influence on species structure, breeding sites, and biological interactions among the immature forms (Fig 6). Compared to a previous study conducted in the same area of Côte d'Ivoire [16], the current study identified 11 additional Aedes species (i.e., Ae. albopictus, Ae. angustus, Ae. apicoargentus, Ae. argenteopunctatus, Ae. haworthi, Ae. lilii, Ae. longipalpis, Ae. opok, Ae. palpalis, Ae. stokesi, and Ae. unilineatus) and 16 additional non-Aedes species that may influence arbovirus transmission patterns. To our knowledge, Aedes mosquito species such as Ae. lilii, Ae. stokesi, and Ae. unilineatus, and others such as Cq. fuscopennata and Tx. brevipalpis appear to be reported for the first time in Côte d'Ivoire. Ae. albopictus is not native to Côte d'Ivoire, but has previously been reported [22]. Presumably this species has been introduced through the seaport bordering the urban municipality of Treichville. The higher numbers of Aedes species is likely due to abundant presence of natural and artificial breeding sites, and their potentials to provide suitable microenvironments. Gravid Aedes females select oviposition sites according to their physical, chemical, and biological characteristics [11,12] and these may change in space and time over the year [16]. The public health relevance of Aedes mosquitoes results from their invasiveness and ecologic plasticity, competence for multiple pathogens, potential as bridge vectors due to their opportunistic feeding behavior and adaptation to urban, rural, and forest areas [23]. Almost all of the container-specialist Aedes mosquitoes collected as larvae such as Stegomyia subgenus, including Ae. aegypti, Ae. africanus, Ae. albopictus, Ae. angustus, Ae. apicoargenteus, Ae. luteocephalus, Ae. metallicus, Ae. opok, Ae. vittatus, Ae. unilineatus, and Ae. usambara species, and Diceromyia and Aedimorphus subgenera comprising respectively Ae. furcifer and Ae. stokesi species have been shown to carry and/or to transmit in nature over 24 viruses, including yellow fever, dengue, Zika, chikungunya, and Rift Valley in tropical regions [5,6]. In addition, Ae. (Aedimorphus) argenteopunctatus in South Africa [24] and Ae. (Neomelaniconion) palpalis [25]   in Malaysia [26] may potentially transmit arboviruses in Côte d'Ivoire. Among non-Aedes mosquitoes, Er. chrysogaster, Er. inornatus and Cq. fuscopennata have been found to have natural infection, while Er. quinquevittatus has exhibited laboratory competence with yellow fever virus in Africa [6]. Moreover, An. coustani has been found to be infested with Zika virus [27], while O'nyong-nyong and chikungunya viruses have been isolated from An. gambiae [28]. Cx. quinquefasciatus [25] and Cx. poicilipes [26] have been shown susceptible to transmit Rift Valley fever virus. In conclusion, as in Senegal [12], the collections of immature stages of nonanthropophagic, unexpected and new potential vectors in rural areas suggest the co-existence of several still unidentified arbovirus cycles in south-eastern Côte d'Ivoire.
Our results also revealed that, urban areas showed higher capacity to support Aedes breeding sites and larvae than suburban and rural areas. The higher numbers of positive breeding sites and higher abundance of Aedes mosquito larvae may be due to the destruction of natural vegetation coverage for infrastructure buildings in the urbanized areas that may affect biological factors (e.g., fauna and flora), and increase the radiation budget thus modifying the microenvironments within and around the microhabitats [29]. Increased exposure to sunlight probably accelerates Aedes mosquito larval development and thus increases the size of adult vectors that possibly find more opportunities of blood feeding sources from larger human populations in urban areas [16,29]. Still, urban Aedes populations are probably less exposed to the pressures from agricultural insecticide and predators (e.g., Eretmapodites spp. and Toxorhynchites spp.) compared to rural communities. We also found that less than two-thirds of breeding sites were infested with Aedes larvae thus suggesting that not all available containers filled with water were occupied by at least one larva or pupa of Aedes mosquitoes and the immature Aedes mosquitoes were not randomly distributed [12]. The presence of empty containers might imply that the gravid females of Aedes mosquitoes select their egg-laying sites carefully according to their physical characteristics (e.g., depth, color, clearance, surface, location, height, shade, sun exposure, and food sources) [12,29], and biological interactions (e.g., competition and predation) [10,11,30] at play within the water-holding container systems.
In our larval surveys, we documented distinct geographic and seasonal variations in terms of the proportions of positive breeding sites and abundance of Aedes mosquitoes in all areas. Indeed, the highest proportions and relative abundance of Aedes mosquitoes were observed among vegetated peri-domestic breeding sites and during the rainy seasons in all areas. The shade of the vegetation reduces the water temperature [12], thus protecting breeding sites from drying out. Moreover, leaves supply organic detritus and associated microorganisms that may serve as food sources for the mosquito larvae [10]. The geographic and seasonal patterns in Aedes breeding sites are important from an epidemiologic perspective and suggest that the rainy season is the best period of time to identify breeding sites, while during the dry season it would be an ideal period of time to control immature Aedes mosquitoes, with particular attention for peri-domestic environments.
Our data revealed that the pattern of Aedes mosquito breeding sites changes substantially from natural containers to artificial containers along a rural-to-urban gradient. Although artificial breeding sites dominate in all areas, there is a higher proportion of natural containers (e.g., rock holes, animal detritus, leaf axils, fruit husks, bamboo, and tree holes) in rural areas, traditional containers (e.g., clay pots, wood-containers, and metallic pots) in suburban areas. However, in the urban areas, the most productive breeding sites for Aedes mosquito were industrial containers (e.g., tarps, discarded tires, vehicle tanks, carcasses, building tools, and water storage containers). The availability of, and the segregation among, Aedes breeding sites probably result from the strong impacts of human activities on the environment, while the natural breeding sites are provided by the natural landscape and agriculture [12]. We observed that tree holes, tires, and water storage containers showed higher Aedes species richness in rural, higher Aedes abundances in all areas, and high Ae. aegypti infestation rates in urban areas, respectively. Tree holes, found in the preserved rainforest, seem to provide ideal larval habitats for several species due to their greater stability, various trophic inputs, and retention of rainwater for longer periods of time [12]. Used tires are mostly associated with the palm oil industry in rural areas, production of the local dish "Attiéké" in suburban areas, and selling of tires and car repairs in urban areas. Tree holes and tires have bigger volumes and are expected to better protect the immature forms of Aedes mosquitoes against flushing during heavy rains [12,14]. Moreover, tires are black-colored containers that are highly attractive to the gravid Aedes females searching for oviposition sites [11,31]. The high number of water barrels infested with Aedes larvae might be due to the water being held for longer periods in uncovered receptacles [32].
Taken together, Aedes species diversity, richness, abundance, and dominance significantly changed from rural to urban settings. The variations in Aedes mosquito species may be explained by the sensitivity of their larvae to environmental changes induced by urbanization [10,12]. Native species such as Ae. africanus, Ae. argenteopunctatus, Ae. longipalpis, Ae. stokesi and Ae. usambara were restricted to natural breeding sites in the rural areas. However, other wild species, such as Ae. furcifer, Ae. dendrophilus, Ae. palpalis, Ae. vittatus, Ae. luteocephalus, and Ae. metallicus were also surprisingly frequent in artificial containers. In contrast, our surveillance failed to sample Ae. fraseri that have been collected by ovitraps in the rural areas previously [16], probably due to its possible cryptic breeding sites or potential height-dependent oviposition behavior. The existence of multiple types of behavior in the same Aedes mosquito species may indicate the existence of generalist species or sibling strains of individuals from various origins [6,11] that have experienced different selective urbanization pressures.
Lastly, our study showed that urbanization acts as a series of ecological filters for Aedes mosquitoes by advantaging Ae. aegypti, the primary vector of yellow fever, dengue, chikunguya, and Zika viruses [1][2][3]. Ae. aegypti was the most prevalent species in all study areas, exhibiting an increasing abundance along rural-to-urban gradient towards an higher abundance in urban areas where larvae mostly inhabit in anthropogenic containers (e.g., tires, discarded containers). Ae. aegypti displayed behavioral plasticity in that the females lay eggs in a vast array of containers ranging from natural containers such as rock holes, tree holes, and bamboo to a wide range of man-made containers [11], including water storage containers in urban areas [32]. The ecologic variations in oviposition behavior of Ae. aegypti and other Aedes mosquitoes may be discussed in ecologic, evolutionary, and epidemiologic approaches [11], and suggest possible overlaps of sylvan and urban vector distributions thus linking several potential mixed arbovirus transmission cycles [5,6,12,16]. In addition, if highly infested microhabitats are targeted for removal, Aedes mosquito females may possibly adapt to changes in breeding habitats and alternatively oviposit in other containers previously unoccupied [33]. The ability of Ae. aegypti to adapt ovipositional behaviors to changing environments possibly enabling to overcome ecological constraints (e.g., instability and disturbance of the breeding sites) imposed by urbanization [10,11]. Ae. aegypti-transmitted yellow fever outbreaks are historically well known in Côte d'Ivoire to have forced the transfer of the capital from Grand-Bassam to Abidjan in 1899 [15]. Since then, several unpredictable resurgences of yellow fever and dengue have been occurring in rural and urban areas causing many suspected, confirmed and fatal cases, and remain presently an unresolved major public health concern [7,15,34], with the current outbreak of dengue DENV-3 resulting in one confirmed and 17 suspected cases recorded in Abidjan in May 2017. Our study suggests that the unique removal of artificial containers that is a common practice in arbovirus control programs in Côte d'Ivoire might not effectively control diseases in the south-eastern part of the country. Vector control measures should combine removals of artificial containers [6] and autocidal gravid ovitrap-based on mass trapping [35], and insecticide auto-dissemination approaches [36].

Conclusions
In south-eastern Côte d'Ivoire, urbanization is associated with larval habitats of Aedes species at a finer scale by driving their breeding sites from natural to artificial containers, and at the larger scale by transforming rural to urban areas. Ae. aegypti is most prevalent in urban areas, suggesting that urbanization is a driver for producing suitable breeding sites for this mosquito species, and hence related disease outbreaks. However, rural settings still support irremovable containers such as natural breeding sites (e.g., tree holes) that host several wild Aedes species and Ae. aegypti. Therefore, even effective removal of discarded containers in urban areas (a common practice in arbovirus control programs) might not be sufficient to control arboviral diseases. Instead, vector control strategies should embrace a more holistic approach, combining different tools and methods of proven efficacy [6,35,36].