Electric nets and sticky materials for analysing oviposition behaviour of gravid malaria vectors

The study was carried out a semi-field system [41] located at the International Centre of Insect Physiology and Ecology, Thomas Odhiambo Campus (icipe-TOC), Mbita, on the shores of Lake Victoria, Kenya (geographic coordinates 0° 26’ 06.19” S, 34° 12’ 53.13”E; altitude 1,137m above sea level). This area is characterized by a very consistent tropical climate with an average minimum temperature of 16°C and an average maximum temperature of 28°C (based on data from icipe-TOC meteorological station for 2010–2011). The area experiences two rainy seasons, the long rainy season between March and June and the short, and less reliable rainy season between October and December. The average annual rainfall for 2010–2011 was 1,477mm.

The semi-field system was a screened greenhouse-like building (Figure 1) 7.1m wide, 11.4m long and 2.8m high at the wall and 4.0m high at the highest point of the roof [15]. The two opposite shorter walls and the roof were made of glass and the two longer walls were screened with black fibreglass netting gauze (1.7×1.5mm). The floor was covered with sand to a depth of 30cm so that artificial aquatic habitats could be dug in to the ground to simulate a natural breeding site for the mosquitoes. To increase the relative humidity inside the semi-field system to 60-70% for experiments the sand floor was watered from 15:00–16:00h prior to the experiment. Care was taken to ensure that no pooling of water occurred on the floor and that the upper layer of sand was dry by the time mosquitoes were released into the system. When treatments were positioned in the corners of the semi-field system (Site 1–4) mosquitoes were released in the centre and when the treatment was positioned in the centre (Site 5) mosquitoes were released 1.5m from the wall at Site 6 (Figure 2). Treatments in the corners were always placed 1.5m from the two adjacent walls.

Two types of artificial habitats were used for experiments. For most experiments round ponds were constructed by positioning a black plastic bowl of 15L capacity (36cm diameter and 18 cm depth) into the ground so that the upper lip was at the same level as the sand floor. The pond was then filled with 9L of water originating from Lake Victoria and filtered through a charcoal-sand filter [42] henceforth called filtered lake water. Rectangular ponds were constructed by positioning black plastic containers of 17Lvolume (50cm long, 37cm wide and 18cm deep) into the ground and filled with 9L of filtered lake water.

E-nets consist of high-tension wires stretched in parallel, across an aluminium frame (1.0 m high×0.5 m wide) with aluminium rods fixed to the two shorter opposite sides of the frame (Figure 3A). Electricity flows between the two ends of each wire generating differentials of >2.5 kV between adjacent wires [23, 27], which kills insects that touch the wires. The wires are invisible to flying insects and do not have significant impact on air movement [29].

The rods had holes at a distance of 8mm for fixing the wires into the rods, to enable the electric wires to be arranged in a vertical position. Small nylon loops (Damyl® fishing lines) were tied of the same size as Fabory® zinc-plated draw springs (0.5×3.5×20mm). Copper wires (diameter of 0.2mm) 1m in length were tied to the fish line loops (insulator) from one end and to the springs (conductors) on the other end (Figure 3B). The ends of the wires with the springs and with the fish lines were alternately fitted to the holes 8mm apart on rods to enable the flow of opposite charges in opposite directions (Figure 3A). Torr et al.[28] assessed different spacing of wires in the electric nets and observed no difference in mosquito catch size between 4, 6 and 8mm spacing. The e-nets were held upright by using clamps on metal stands with base (Figure 3A). Alternate wires in each row were charged by a 12V car battery via a transformer (spark box). In the spark box, the 12V direct current (DC) is first converted to an alternating current (AC) that is stepped up to 400 volts peak AC. It is then rectified and converted back to 400V DC. The 400V DC voltage is used to charge a bank of capacitors that are then discharged into the primary of the ignition coil. The voltage output to the e-nets can be reduced by an energy dial lowers the 12 DC input to the spark box, that in turn lowers the 400V output that charges the capacitors. The dial position roughly equates to the energy reduction not to a direct conversion of the voltage outputs to the nets. The output is 400V at 100% spark energy setting and approximately 300V at the 50% spark energy setting of the dial.

A range of different materials were used as trapping devices for mosquitoes in the experiments. Throughout the manuscript reference is made to the materials listed in Table 1.

Insectary-reared An. gambiae s.s. mosquitoes were used throughout. Gravid mosquitoes were prepared as follows;300 female and 300 male mosquitoes, two to three days old, were kept in 30×30×30cm netting cages and provided with 6% glucose solution ad libitum at 25-28°C and a relative humidity from 68-75%. Saturated cotton towels (50x25cm) were folded and placed over the cages to avoid mosquito desiccation. Mosquitoes were starved from sugar for 7 h and allowed to feed on a human arm for 15 minutes at 19.00h on the same day the same procedure was repeated 24 hours later. After the first blood meal unfed mosquitoes were removed from the cages. Fed mosquitoes were kept together with males for two more days after the second blood meal before they were utilized in an experiment (i.e. females aged 4–5 days after first blood meal). Host-seeking mosquitoes were prepared by selecting 300 two to three days old females on the day of experiment. Mosquitoes were starved for 6 h before the experiment commenced at 18.00h.

All experiments were implemented in a single semi-field system. Two hundred gravid female mosquitoes were selected from the holding cages based on their abdominal stages (whitish in colour and oval in shape) and were released into the semi-field system between 17.30h and 18.00h. Experiments were terminated at 08.00h the following morning. Experiments with more than one treatment followed a randomized complete block design. Treatments were assigned randomly (using a random number generator) to the corners of the semi-field system and rotated randomly across corners until all treatments were run once in each of the corners included in the respective experiment. This block of experiments was then repeated. Experiments were carried out for 8 or 12 nights.

A second problem associated with e-nets is how to collect the stunned mosquitoes. Insects killed or stunned after colliding with the e-net fall to the ground. For ease of collection and to prevent them from being eaten by ants and other predators a catching device on the ground was needed. Water-filled trays under the e-nets worked well in experiments with host-seeking mosquitoes [27, 28], however when studying the behaviour of gravid mosquitoes, water-filled trays cannot be used since they might attract gravid mosquitoes in search of an oviposition site and divert them from the intended target. A series of experiments with e-nets positioned over sticky boards were carried out to find the most suitable material for collecting mosquitoes when stunned by the electric net.

One e-net was set up in a corner of a semi-field system (Sites 1–4) and a round pond placed 70 cm from the e-net, between the net and the corner to attract gravid females. Transparent sticky film was mounted on two 50x80cm cardboard rectangles. A grid of two rows, each row with 4 cells (20x25cm), was marked on the boards. One board was placed on each side of the e-net (Figure 3A). The e-net was charged using 50% spark energy and the experiment carried out for 8 nights. The number of mosquitoes that got stuck on the film was counted separately for each cell and direction towards the e-net.

A collection device under an e-net should not attract gravid mosquitoes otherwise the number of mosquitoes approaching a target will be overestimated. Shiny sticky surfaces may, to a gravid mosquito, look like a water body. Accordingly, studies were undertaken to assess whether gravid mosquitoes landed on the shiny surfaces of the transparent sticky films. Four boards (50×80cm) were prepared with transparent sticky film. Two of the boards were placed on the ground in one corner and the other pair of boards in the opposite corner of the semi-field system. In order to test if landing on the sticky boards is associated with a resting behaviour close to a water source just prior or after egg-laying or if it is an actual attraction towards the surface an artificial pond was added to one of the two treatments. A round artificial pond was dug into the sand 20cm behind one of the pairs of the sticky boards. The experiment was carried out for 8 nights. The number of mosquitoes that landed on the boards was recorded.

To find a non-attractive device for the collection of electrocuted mosquitoes, three sticky surfaces different in texture and colour were compared. Three cardboard squares of 50x50cm were covered with one of the following treatments (Figure 4): (1) transparent sticky film; (2) black netting painted with 100 g insect glue dissolved in 25 ml hexane; and (3) yellow sticky film. Boards were positioned in three different corners of the semi-field system. One corner remained empty but was included in the random allocation of treatment location. Round artificial ponds were dug into the sand at a distance of 20cm behind each of the boards (Figure 4).The experiment was carried out for 12 nights. The number of mosquitoes that landed on the boards and the number of eggs laid in their respective ponds were recorded.

A complete square of four e-nets was mounted around a rectangular pond set up in the centre of the semi-field system in order to estimate the number of gravid females approaching water. Adjacent e-nets were held together by clamps on stands and two of them shared one battery and a spark box (Figure 5). E-nets were charged with 50% spark energy. Four yellow sticky boards of 50×50cm were placed in front of each of the e-nets. Any open space inside the square of e-nets was also covered with yellow sticky board (Figure 5). Boards were divided into two horizontal rows (25×50cm) for further evaluation of the efficacy of the net and of the board as a collection device. The number of mosquitoes collected in the two rows of the board and inside the square of e-nets was recorded separately. Any eggs in the ponds were counted.

A prerequisite for the development of new monitoring control tools targeting oviposition site seeking mosquitoes e.g. with gravid traps [18‐20, 38, 47‐53] is to know if and where gravid females land during oviposition. Very few studies have assessed this particular behaviour and a variety of different modes of oviposition have been described. Here the use of different sticky materials and a detergent are evaluated to analyse potential landing of gravid females on the water surface or habitat edge for laying eggs.

The edges of three round artificial ponds were made sticky to trap any landing mosquito by applying one of the three treatments: (A) yellow sticky film, (B) spray glue or (C) transparent double-sided sticky film to their inner walls. The sticky edge was 7cm wide and bordered the water surface. The ponds were set up in three corners of a semi-field system. The empty corner was included in the randomization of the treatments. The experiment was run for 12 nights. The number of mosquitoes stuck to the sticky edges and the number of eggs laid in the ponds were recorded.

Four round artificial ponds were prepared. One of the following four treatments were applied on the water surfaces: (1) two A4 overhead projection transparencies were overlain on each other with colourless adhesive tape to form a cross-shaped surface; the transparencies were coated on one side with 100 g insect glue dissolved in 30 ml hexane and placed on the water surface leaving 8areas of approximately 105cm² free water access at the edges (Figure 6A); (2) a circle of dark-green wire screen of the same area as the pond was prepared and coated with 100g insect glue dissolved in 30ml hexane; the wire screen was mounted on a square of wire and placed horizontally inside the pond 5cm below the edge of the pond and 2cm above the water surface (Figure 6B); (3) 225ml (2.5%) detergent was added to the water (Figure 6C); (4) insect spray glue was uniformly sprayed on the water surface (Figure 6D). Ponds were set up in the corners of the semi-field system and the experiment carried out on 12 nights. The number of mosquitoes caught and the number of eggs laid in each pond were recorded.

Finally the best catching tools from the previous two experiments were combined, to assess whether there is a sequence in the landing behaviour (e.g. landing on surface for egg-laying and then resting on the edge of the pond). One round artificial pond was prepared. On the edge spay glue was applied and 225ml detergent added to the water. The artificial pond was set up in the centre of a semi-field system. The experiment was carried out for 8 nights. The number of mosquitoes caught and the number of eggs in the pond were recorded.

The data generated in this study were count data, i.e. either the number of gravid mosquitoes recollected or the number of eggs laid in artificial ponds, and were not normally distributed. Therefore, multivariable analyses were done with untransformed data using generalized linear models [54, 55]. Data analyses were done with R statistical software version 2.13.1, including the contributing packages MASS, effects, epicalc, multcomp, lme4, gee, geepack, aod [56]. Experiments with one treatment tested in the semi-field system each day were analysed using generalized linear models with negative binomial data distribution using the glm.nb function and a log link function. Data collected for two or more treatments in the same semi-field system on the same day were not independent and were therefore analysed with generalized estimating equations using the gee or geepack function. In this case the repeated measure was the day of experiment. Here, a Poisson distribution of the data was used in the model and an exchangeable working correlation matrix. The fixed factor variables included in this model were the treatments of interest and the corner of the semi-field system (site) in which a treatment was placed. It was thought possible that the probability of catching mosquitoes might differ between the four corners (sites 1–4) of the greenhouse, independently from the test treatment, due to slightly different environmental factors such as light intensity, wind direction and microclimate. If the effect of site was insignificant this variable was removed from the final model. The output presented in the tables includes only significant factors from the final model.

The parameter estimates of the models were used to predicted the mean counts per treatment and their 95% confidence intervals (CIs) by removing the intercept from the models [54]. Similarly, multiple comparisons of treatments were calculated based on the model parameter estimates.

With the low energy setting, twice as many An. gambiae s.s. mosquitoes were collected than with the high energy setting (Table 2). Thus, the low energy setting was chosen for all subsequent experiments with gravid females.

In the first e-net experiment with gravid females, an average of 104.1 females (95% CI 78.0-138.9) were collected per night on the transparent film of the collection boards, representing around 50% of females released. Similar numbers were caught on both sides of the e-net, with greatest numbers close to the net in the centre (Table 2, Figure 7). This distribution indicated that most mosquitoes were electrocuted by the net but many females on the row furthest from the e-net appeared to ‘sit’ on the board rather than lay on the side as was the case when stunned, some were even still alive in the morning. This suggested that some females were not stunned by the net but had been attracted by the shiny film and landed on it. If this was true the number of mosquitoes on the collection board overestimated the number attracted by the water and stunned by the e-net. It was, therefore, necessary to evaluate the potential attractiveness of the collection device in the next experiments.

In this experiment mosquito collections were significantly affected by the corner in which the treatments were presented in the semi-field system. If any of the two treatments was set up in site 2 it was 2.5 times more likely to catch a mosquito than if it was set up in site 4 (Table 2). Adjusting for corner, the analyses showed that the sticky board alone caught approximately 15% of the released mosquitoes, while 24% were collected when the sticky board was placed next to water. These results suggest that the sticky board alone was attractive to gravid females and their landing on it was not associated with resting around a potential habitat otherwise females should not have been trapped by the boards without pond. This experiment also confirms that water vapour is a strong attractant for oviposition site seeking mosquitoes.

Gravid females were equally attracted by the transparent sticky film and the sticky black fly-screen, yet few were collected on the yellow sticky film. Furthermore, a significantly higher number of eggs were laid in the pond behind the yellow boards than in the ponds behind the other sticky materials (Table 2, Figure 8). Yellow sticky boards did not interfere with the approach of the gravid female towards a pond and consequent egg-laying and were therefore chosen as routine collection device under e-nets.

In order to estimate the number of gravid females approaching an aquatic habitat a complete square of e-nets was used that surrounded an artificial pond (Figure 5). On average one third (65.3 (95% CI 55.9 – 76.10)) of the 200 released mosquitoes were collected. Over 81% of these were found on the outside of the ring indicating that only few gravid females might have approached the oviposition site from a height above 1m from the ground or passed through the 8 mm gaps between the vertical aluminium frames and the wires. Low numbers of eggs were found on 6 out of 8 days. The average number of eggs was 80.3 (95% CI 43.6 – 147.8). On average over nine times as many mosquitoes were found on the sticky board close to the e-nets than further away suggesting that they were stunned by the electric nets and hence fell close to the base of the net (Table 2).

On average the number of females trapped on the water surfaces was over four times higher than on the edges (103.3, 95% CI 93.0-115 and 23.7, 95% CI 20–28.2, respectively) irrespective of collection device (Tables 3 and Figure 8). The detergent and the spray glue caught about three times more mosquitoes than the sticky wire screen or transparencies (Table 3). The detergent lowered the water surface tension to such an extent that mosquitoes that landed on the water surface sunk, presenting little opportunity to lay eggs. On the other hand, a large proportion of the mosquitoes stuck on the surface with spray glue laid eggs, leading to more than 11 times higher mean egg numbers than other treatments (Table 3).

From those treatments applied to the edge of the pond, the yellow and transparent films trapped similar numbers of mosquitoes but less than half of the spray glue (Table 3). Similar egg numbers in all the treatments indicate that a similar number of gravid females approached these ponds and laid eggs. It is unlikely that all these eggs were laid by the few mosquitoes trapped on the edge. The mean number of eggs in these ponds is comparable with the mean number laid by mosquitoes stuck on the spray glue on the water surface (Table 3).

Finally, when detergent in the water was combined with spray glue on the edge of a pond, most mosquitoes were drowned in the water with only 15% stuck on the pond edge (Table 3). Notably, approximately a quarter of the released mosquitoes were collected with this method. This is only slightly less than the figures obtained from the square of e-nets where approximately one third of all released mosquitoes were collected. Eggs were not found in the pond throughout the test nights suggesting that oviposition did not take place in flight.