Habitat characterization and insecticide susceptibility proles of Aedes aegypti mosquitoes in Ifakara area, south-eastern Tanzania

Background: Aedes-borne diseases such as dengue and chikungunya constitute constant threats globally. In Tanzania, these diseases are transmitted by Aedes aegypti, which is widely distributed in urban areas, but whose ecology remains poorly-understood in small towns and rural settings. Methods: We surveyed aquatic habitats of Ae. aegypti mosquitoes in and around Ifakara, a fast-growing town in south-eastern Tanzania. The area was divided into 200m × 200m search grids and habitats containing immature Aedes were characterized. Field-collected Aedes were tested for susceptibility to common public health insecticides (deltamethrin, permethrin, bendiocarb and pyrimiphos methyl) in dry and rainy seasons. Results: Of 1,515 and 1,933 aquatic habitats examined in dry and rainy seasons respectively, 18.87% and 14.64% contained Aedes immatures (container index (CI): 286-283). In the 2,315 and 2,832 houses visited in dry and rainy seasons, 4.9% and 6.6% had at least one Aedes-positive habitat (house index (HI): 114-186). The main habitat types included: a) used vehicle tires and discarded containers, b) ower pots and clay pots, and c) holes made by residents on trunks of coconut trees to support climbing harvesters. Used tires had highest overall abundance of Aedes immatures, while coconut tree holes had highest densities per habitat. Ae. aegypti adults were susceptible to all tested insecticides in both seasons, except bendiocarb, against which resistance was observed in rainy season. Conclusion: This is the rst study on ecology and insecticide susceptibility of Aedes in Ifakara area, and will provide a basis for future studies on its pathogen transmission activities and its control. The high infestation levels observed indicate signicant risk of Aedes-borne diseases, requiring immediate action to prevent potential outbreaks in the area. While used tires, discarded containers and ower pots are key habitats for Aedes here, this study also identied coconut harvesting as an important risk factor, and the associated tree-holes as potential targets for Aedes control. Since Ae. aegypti mosquitoes


Background
In recent decades, signi cant attention has been put on controlling mosquitoes that transmit malaria, leading to signi cant progress since 2000 [1,2]. However, other mosquito-borne diseases, such as dengue, yellow fever, chikungunya and zika, which are transmitted by Aedes mosquitoes remain largely neglected. Golding et al.,, 2015 showed that more than 90% of persons at risk of vector-borne diseases are affected by at least two such diseases, malaria and dengue fever being the commonest [3]. The WHO Global Vector Control Response (GVCR) initiative therefore recommended integrated approaches to address multiple vectors and vector-borne diseases [4]. Unfortunately, unlike malaria, for which effective prevention and treatment options are widely available, the Aedes-borne diseases still rely mostly on personal protection measures [5], even though vaccine trials are increasingly advanced as well [6].
In Tanzania, concerns about Aedes-borne diseases have become increasingly prominent in recent years, due to multiple outbreaks, detection of the viruses in humans, and the wide distribution of the Aedes mosquitoes [7][8][9][10]. Dengue cases have been reported in multiple regions in the country, including Dar es Salaam city, the islands of Zanzibar and Pemba, Mbeya and Iringa areas in the southern Tanzania, and Kilimanjaro in the north [11][12][13]. The most recent outbreak occurred in May 2019, when 1,012 new cases were con rmed over just two weeks [14]. By September 2019, 6912 cases had been reported, including 13 deaths [14].
Most outbreaks of Aedes-borne diseases have been observed in urban areas, where densities of both the vector and humans are high [11]. However, human mobility has also led to introduction of the viruses in rural areas and small towns [7]. Unfortunately efforts against these diseases are hampered by lack of proper medication or diagnostics [15][16][17]. Effective vector surveillance and control to prevent potentially-infectious mosquito bites therefore remain core components of programs targeting such diseases [5].
Current understanding of Aedes mosquitoes is largely based on studies in urban areas where the vector is most widespread [18]. Ae. aegypti, the most important of the Aedes species, is considered highly anthropophilic, and is a frequent breeder in arti cial containers [19], common in urban settings [8]. Improper disposal of waste containers provides perfect breeding environment for Ae. aegypti mosquitoes. For example, in coastal Tanzania, used tires and disposed containers were identi ed as commonest aquatic habitats for Ae. aegypti [11,20]. However, less is known regarding the ecology of these vectors in inland Tanzania, including small towns and rural settings. This is important to understand distribution of the vectors across the country, but more importantly to prevent introduction or spread of Aedes-borne arboviruses. To ensure effective control, such ecological studies should be complemented with investigations on susceptibility to commonly-used public health insecticides [21,22].
This study was therefore conducted to ful l three key objectives: a) investigate spatial distribution of Ae. aegypti in Ifakara town and surrounding wards in south-eastern Tanzania, b) characterize aquatic habitats of the mosquitoes in the area, and c) assess susceptibility of the mosquitoes to insecticides commonly used for vector control.

Study area
Surveys for Aedes immatures were conducted in Ifakara town and surrounding wards, namely, Lipangalala ( Sitini are characterized as urban, while the other three are rural. The area has an average of 270m altitude, annual rainfall of 1200 -1800mm, relative humidity of 51% -71%, and daily temperatures of 20˚C -32.6˚C [23].
The area experiences short rains in November and December, which is interrupted by dry months from January to March. Heavy rains continue from April to May or June, followed by dry July and September. It is a rapidly growing area with total population now estimated at 67,500, based on the 2.7% annual growth from the last census in 2012 [24].

Selection of sampling sites
The study area was divided into grids measuring 200m × 200m, in ArcGIS 10.4 environment (ESRI, USA) as previously done by Mwangungulu et al.,,[25], and each grid assigned a unique identi er ( Figure 2). We overlaid the grids with household geo-location data initially collected by Ifakara Health Institute's Health and Demographic Surveillance System [26]. This data was updated using population density maps from Google satellite imagery, and a high resolution settlement layer (HRSL) obtained from the Facebook Connectivity Lab and Centre for International Earth Science Information Network (CIESIN) [27].
From each ward, 34 grids containing human habitation or other actively-used buildings were selected as search grids. For each search grid, houses or buildings nearest to the centroid were identi ed as starting points for Aedes habitat searches. Where no informed consent was obtained, the next nearest consenting household was selected. From the starting points, we searched all potential aquatic Aedes habitats within 100m radii, visiting each search grid twice in dry season and twice in rainy season. We also mapped important features such as schools, marketplaces, worship areas, health facilities and water pumps using handheld GPS receivers (Magelan eXplorist GC, USA).

Sampling of mosquito immatures and characterization of their aquatic habitats
Sampling for Aedes immatures and characterization of their habitats was focused on natural and arti cial water-holding objects such as tree holes, used tires, wells and discarded containers and animal feeding containers. Others included coconut shells, tarpaulins, broken grasses and other small objects that could potentially hold water longer than three days. All sites with Ae. aegypti larvae or pupae were geo-referenced using handheld GPS. The habitats were characterized by: a) location, b) size, c) apparent water color, d) presence of vegetation, e) presence of shading f) source of water in the habitat, g) whether the habitat was movable or not, and h) environmental and social activities surrounding the habitats.
We sampled larvae and pupae from each of the identi ed habitats using standard 350ml dippers, or a smaller 70ml dipper in cases where habitats were too small to sample using the standard dipper. The larvae and pupae were placed in white trays for morphological identi cation, using pictorial keys created by US Centres for Disease Control [28]. They were then sorted, counted and data recorded by habitat type, location and survey instance.

Mosquito rearing and identi cation of emergent adults
The Aedes larvae were transferred to the vector biology laboratory (VectorSphere), at Ifakara Health Institute (IHI) for rearing, and eventual morphological identi cation of emergent adults. The larvae were fed on Tetramin® baby sh food, and maintained at temperatures of 26°C ± 2 °C and relative humidity of 82% ± 10%.

Measurements of mosquito wing lengths
We also assessed whether mosquitoes from different wards or habitat types varied in size, by assessing their wing lengths. Emergent adults were anaesthetized at -10°C. Wings were removed from male and female mosquitoes (one wing/mosquito). Drops of distilled water were used to x the wings onto glass slides. Wing lengths were measured, as distance from the apical notch to the auxiliary margins, under stereo zoom microscope using a micrometer ruler.

Data analysis
Statistical analyses above were done in the open-source R statistical software, version 3.231 [32]. Descriptive analysis was done to compare larval densities in different wards and seasons. Densities obtained from the 70ml dipper were compared to those from the standard 350ml dipper and a correlation coe cient calculated across all collections. Using this coe cient, the densities assessed by small dipper were all converted into standard dipper, so that all subsequent analyses were done on the standard dipper.
Generalized Linear Models [33] following Poisson distributions for count data were used to model number of larvae collected per dipper as response variable against season and habitat type as xed factors. Logistic regression was also used to assess associations between positivity of different habitat types for Aedes larvae (proportion of individual habitats of one type that were positive with Aedes larvae). The relative risk (RR), odds ratios (OR) and their 95% CI were estimated. The dabestr package was used to assess mean differences of larval abundance between wards and seasons.
Larval indices, namely Container Index (proportion of containers infested with Ae. aegypti larvae or pupae), House Index (proportion of houses infested with Ae. aegypti larvae or pupae) and Breteaux Index (number of infested containers per 100 houses) were also calculated by ward and season [22,34]. Mosquito wing lengths were compared using one-way ANOVA, followed by Tukey's post-hoc tests to assess mean differences betweenward for both male and female mosquitoes. Susceptibility status of Ae. aegypti was computed according to WHO guidelines [30], log-probit analysis was used to compute mean duration at which 50% (KD 50 ) and 95% (KD 95 ) of the exposed mosquitoes were knocked down. Spatial and seasonal distribution of Aedes immatures were analyzed by geostatistical in ArcGIS 10.4 (ESRI, USA). Inverse Distance Weighted (IDW) interpolation technique [35,36] was used to visualize the areas with high larval densities. Representation of IDW maps show patterns based on the distance from one observed point to another. Known values (number of larvae) were used as key input feature to estimate unknown locations within 400m range based on estimated average ight range of Aedes mosquitoes [37,38]. Geo-processing extents and masks were de ned to match the study area.

Larval indices
A total of 1,515 breeding sites were visited in the dry season and 1,933 in rainy season. Of these, 286 (18.87%) in dry season and 283 (14.64%) in rainy season were positive with Aedes immatures. The proportion of infestation varied across wards and seasons as summarized in Table 1. In the dry season, high Container Indices (CI) were observed in Katindiuka, Viwanja Sitini and Ifakara Town wards, while in rainy season, high CIs were in Ifakara town, Viwanja sitini and Lipangalala wards. With regard to House Indices (HI), 2,315 and 2,832 houses were visited in dry and rainy season surveys, of which, 114 (4.9%) and 186 (6.6%) had at least one positive habitat respectively. Lipangalala ward had the highest HI during the dry season, while Ifakara town had highest HI in rainy season. Compared to dry season, HI increased during rainy season in all wards expect Lipangalala (Table 1). It was also observed that Viwanja Sitini ward had highest Breteaux Index (BI) in both seasons.
Densities of Ae. aegyptiimmatures, their distribution, and their aquatic habitats A total of 63,470 larvae or pupae were collected from all wards. Of these, 76.3% (n = 48,459) were Ae. aegypti, 20.9% (n = 13,253) were Culex and 2.8% (n = 1,758) were identi ed as other Aedes species mosquitoes. In the dry season surveys, Ifakara town produced nearly one third of all immature Aedes and more than one third of immature Culex. In the rainy season however, Viwanja Sitini had more than one third of the Aedes immatures, while Katindiuka produced more than half of all Culex. Most Culex were found in dry season, while Aedes were more prevalent in wet season (Table 2).
Overall, most Aedes larvae were from used tires and clay pots followed by other containers such as discarded tins, buckets, drums and animal feeding pots ( Figure 3). However, coconut tree holes and ower pots had far higher numbers of larvae per dip compared to all other habitat types, in the dry season ( Table 3). Likelihood of getting larvae in individual tree holes was three times higher than in used tires (RR = 3.00 [2.58-3.50], P<0.01).
However, in the rainy season, higher larval densities were observed in other habitats ( Table 3).

Positivity of different habitat types for Aedes immatures
Positivity of the habitats for Aedes are summarized in Table 4. By assessing proportions for each type of habitat, it was determined that used tires were the most commonly infested with Ae. aegypti (89% positivity), followed by containers (86% positivity) and clay pots (82% positivity), garage pits (64% positivity) and others (90% positivity). Majority of the positive breeding sites were movable, associated with human activities, or were found in and around residential areas, commercial places and garages. We also observed signi cantly higher Aedes positivity in rainy season than dry season. Also, number of positive habitats were higher if they had clear water than turbid water.

Spatial and seasonal distribution of Aedes immatures
The spatial distribution of Aedes immatures varied between dry and rainy season (Figure 4). In dry season, the highest infestation was from the center of Ifakara town toward western parts of Katindiuka ward. In the rainy season on the other hand, most infested locations were in southern Lipangalala and in Viwanja sitini ( Figure 4).
Generally, fewer breeding sites were observed in dry season compared to rainy season in all study sites, though actual abundance varied signi cantly between sites. Ifakara town consistently had higher mean number of larvae than the other wards across seasons ( Figure 5). We also estimated the residual mean differences of larval abundance between study ward.

Susceptibility of adult Aedes aegypti mosquitoes to insecticides
Ae. aegypti females were generally susceptible to all four classes of insecticides. Only in few instances did Ae. aegypti show reduced susceptibility to carbamates, and pyrethroids ( Figure 6). Con rmed resistance was detected against only bendiocarb in the rainy season tests (Figure 6).
Overall knockdown KDT 50 and KDT 95 ranged from 7 to 112 minutes and 13 to 159 minutes respectively ( Table   5). The knock down analysis revealed spatial and seasonal variation. Dieldrin and pirimiphos-methyl consistently achieved slower knock-down across wards, while bendiocarb and deltamethrin had quick knockdown. Knock-down times were not predictive of overall 24hr mortality. Often, mosquitoes were not affected by the insecticides during rst 60 mins but mortality after 24 hours was still high.

Wing lengths of adult Aedes aegypti mosquitoes
Wing lengths, used here as a proxy for adult sizes of male and female Ae. aegypti ranged from 1.9mm to 3.5mm (Figure 7) Post hoc analysis also revealed differences between pairs of wards ( Figure 7). Also, the mean wing length of female Ae. aegypti were generally larger than those of male Ae. aegypti (ANOVA: F-statistic: 365.9 df = 1, p<0.001).

Discussion
In Tanzania, majority of studies conducted on arbovirus vectors are in response to outbreaks, and are often concentrated in large urban areas [11]. Basic ecological studies to understand distribution and behaviors of the vectors, as well as their responses to interventions remain very few. This current study involved an exploratory survey of Ae. aegypti mosquitoes in a small town and its surrounding wards in south-eastern Tanzania. The ndings therefore constitute essential baseline data on Aedes mosquitoes in this area where no outbreak has previously been reported, yet the risk is high. Given that there have been reports of arboviral infections such as Dengue and Chikungunya in neighboring districts [7], it is crucial to invest on studies to improve our understanding of the ecology of the vectors, so as to boost control options.
This study therefore assessed three important aspects, namely: a) spatial distribution of Ae. aegypti mosquitoes in Ifakara town and its surrounding wards in south-eastern Tanzania, b) characteristics of key aquatic breeding habitats of these mosquitoes, and c) the susceptibility of the mosquitoes to insecticides commonly used for vector control.
The main nding was that, larval indices (container index (CI), house index (HI) and breteaux index (BI)) are high enough to signal signi cant risk of Aedes-borne diseases in the area. In the rainy season in particular, house and container indices in all wards exceeded the value of 5.0, speci ed by WHO for actionable arboviral infections risk [39][40][41]. Dry season risk was however con ned to fewer wards though not completely absent from the rest of the wards. Immature Ae. aegypti infestation varied between wards and seasons, but remained signi cant even in dry season. This is expected since Aedes mosquitoes typically breed in man-made containers not fully dependent on rainfall. Besides, the vectors have fewer options of breeding sites in dry season hence elevating container level of infestation with immature Ae. aegypti (Table 1). On the contrary, aquatic habitats were relatively large in number during the rainy season, resulting in lower positivity rates (Table   1). This higher level of container infestation in the dry season concur with the study conducted in northern regions of Ghana which showed that, indices in the dry season was aggravated by poor water supply system in the area. As a result, facilitated the storing of water in pots and barrels for a period enough to bred Aedes mosquitoes [42].
We noted that Ae. aegypti prefers breeding in clean and stagnant waters. Similar to other studies [19,20,43]. Common habitats for Ae. aegypti were used tires, clay pots, ower pots, containers, coconut tree holes, pits, and on rare occasion disposed shoes, cooking pans, broken grasses and tarpaulins. Majority of these habitats were easy to discard, indicating an opportunity for proper waste management and environmental management as effective options for Aedes control, especially if used alongside traditional larviciding. As already highlighted by several previous studies, tires in particular serve as important breeding sites for Ae. aegypti because they can hold water for long periods even in dry season [11,19,44]. The multiple applications of used tires in the area will however complicate efforts to effectively dispose of the tires. For example, we observed that people use these tires as make-shift chairs, for playing by kids, for planting trees (residents believed that tires prevent plant pests) and for vehicle repairs.
A major natural breeding site in the area was coconut trees, which had arti cial holes created for climbing during the coconut harvesting period. These holes served as the perfect breeding sites for Ae. aegypti mosquitoes. We recommend that coconut tree holes be lled with sands to prevent rainwater from stagnating [22]. Clay pots were also common in Katindiuka and Lipangalala wards where they were mostly used for collecting rainwaters for various domestic purposes. Unfortunately, residents did not know these pots bred mosquitoes. We also observed rare habitats such as disposed coconut shells, broken glass, animal feeding containers, tarpaulins and discarded plastic shoes which produced high larval abundance (larvae/dipper). Higher abundance was in uenced by size of habitats and the volume of water present breeding sites.
During data collection period, we raised awareness in surrounding communities about mosquito breeding behaviors and diseases they transmit. This led to a better understanding for them, and greater engagement of the communities in our work. Some breeding sites observed during the rst visit were not there during subsequent visit as people became aware of the risks and hence proactively removed or covered potential habitats. This observation highlights the potential of educating communities about Ae. aegypti mosquito habitat sources and participatory control efforts. In Tanzania, the government is already implementing monthly clean-up campaigns, which could be leveraged to achieve such gains. Moreover, efforts to reduce mosquito population can prioritize areas identi ed with higher risk.
Mosquito sizes play an important role in overall vector competence, vectorial capacity and ability to disseminate viruses [45,46]. Smaller mosquitoes tend to have high contacts with hosts as they need more frequent blood meals than bigger mosquitoes, a phenomenon which could increase transmission [46]. In the other hand, bigger mosquitoes have been demonstrated to be more resistant toward insecticides [47]. Here, the wing length measurements for Ae. aegypti was done as previously documented by Nasci in 1986 [48], and showed a range of 1.9mm to 3.5mm. We also observed differences between administrative wards, though the extent to which such variations affect pathogen spread remains to be determined.
Lastly, we assessed how Ae. aegypti mosquitoes in the area would responds to control by commonly available insecticides. Fortunately, this study showed that the mosquito populations here are still generally susceptible to most insecticide classes except for bendiocarb against which there was resistance during the rainy season.
Since this study is the rst in the area of its kind, there are no immediate comparisons for the resistance pro le. However, in studies done in Dar es salaam, Peru and Burkina Faso, resistance to pyrethroids and organophosphate was marked [20,49,50]. In our study, we have also observed notable spatial and seasonal variation toward Bendiocarb. Similar observation was previously documented for Anopheles arabiensis and Culex pipiens in south-eastern Tanzania [51,52]. Reduced susceptibility to pyrethroids observed in some of our assays, and the resistance seen against bendiocarb in the rainy season are however signs that we must remain vigilant as insecticide resistance could rapidly spread among the vector populations once active control programs begin. This would therefore mean that environmental management, including larval habitats search and removal, should be an important component of any anti-Aedes campaigns. As most habitats are those that can be discarded, combinations of insecticidal and non-insecticidal approaches would likely be effective.
Though the main objectives were successfully completed, this study also had various limitations. First, larvae and pupae were only collected in the selected grids (34 grids per ward) but these wards are not of the same surface area (Figure 2), thus some might have been underestimated. We therefore recommend the future studies should consider all the grids occupied by human habitations and building. Second, we adopted WHO standard dose speci ed for Anopheles mosquitoes, as we still do not have a comprehensive guideline for Aedes. However, some of these insecticides, such as pirimiphos methyl, permethrin and deltamethrin already have diagnostic concentrations speci c for Aedes mosquitoes. Therefore, If the right concentration were used the results might have been different. For instance, results for permethrin (0.75%) demonstrated susceptibility toward standard concentration for Anopheles, which is three times the Aedes standard concentration (0.25%).
This mean Aedes mosquitoes might be resistant toward this concentration but susceptible toward Anopheles concentration. We recommend therefore that future studies should incorporate appropriate guidelines for the species.

Conclusion
This is the rst study on ecology and insecticide susceptibility of Aedes mosquitoes in this area, and will provide a basis for future evaluation of its role in pathogen transmission, as well as options for its control.
Infestation levels observed indicate that immediate actions should be taken to prevent outbreaks. The larval indices (container index, house index and breteaux index) are high enough to signal signi cant risk of Aedesborne diseases in the area. Fortunately, the Ae. aegypti in the area are still susceptible to majority of insecticides used in public health, indicating available opportunities to include insecticides in the control programs. Since most habitats were those that can be discarded, integrating concepts of environmental management, insecticide use and community engagement could yield signi cant progress. While used tires, discarded containers and ower pots are key habitats for Aedes in the area, this study also identi ed coconut harvesting as an important risk factor, and the associated tree-holes as vital targets for Aedes control.     OR=odds ratio, CI=con dence interval, N=number of breeding sites, Category used as reference R=1, social and environmental factors were dropped in the analysis they had less impact.   Selected grids in the study area, which were sampled for conducting Ae. aegypti larval surveys in dry and rainy seasons. Estimated population densities are also shown.

Figure 3
Various breeding sites identi ed in the study area: A) used vehicle tires, here repurposed by residents as seats B) used tires kept for protecting trees from pests, C) disposed coconut shells, D) ower pots, E) animal feeding container, F) broken grasses, G) disposed containers, H) coconut tree holes, I) clay pots, J) small containers, and J) pits such as those at construction sites, in garages, or inspection chambers in waterworks.  Estimated means of Aedes larvae/dip in Ifakara town and surrounding wards in: A) Dry season, and B) Rainy season. Estimation plots are used to portray the distribution of residual mean differences of larval abundance between study wards. The vertical lines represent mean ± con dence levels (the gap in the line is the mean).
The lled curves indicate the resampled mean difference distribution of the larval abundances with reference to Ifakara town. Black vertical line indicates 95% con dence level. Black dot indicates mean difference to the reference group. The signi cance is considered depending on how far the means of residual deviated from the refence line. Differences in mean wing lengths between wards. Pairwise comparisons are shown at 95% Con dence Levels for A) Female and B) Male Aedes aegypti mosquitoes.