Villages
Experiments were carried out in the following villages, located in the Koulikoro Province of Mali, West Africa: Tiko (12.13444, -8.396860), Balala (11.96599, -8.468310), Niaganabougou (12.15466, -8.308260), Sambadani (12.14454, -8.316880), Korea (12.04576, -8.399230), Krekrelo (11.98836, -8.551460), Cissebougou (12.09628, -8.372850), and Trekorou (12.068577, -8.314414). The villages, accessible by car, were within 10 km of the Niger River, and had high densities of An. gambiae s.l.
Bait station construction and ASB Composition
ASB consisted of the following: ~22% (w/w) date syrup as the attractant component, 77% (w/w) brown sugar ("Nature Sugar" brown, Louis Dreyfus, Israel) as the feeding stimulus, 10% (w/w) of a proprietary mixture of slow-release substances, (BaitStab ™, Westham Innovations Ltd., Tel Aviv, Israel) which is a proprietary preservative component, and 0.5% (w/w) orange food dye (Carmoisine E122, Stern, Natanya, Israel). A similarly prepared solution with green food dye (Tartrazine 19140, Stern, Natanya, Israel) instead of orange was used for the 1.0 m height experiment. Bait stations were constructed using a white, rectangular plastic frame, 24 cm w by 36 cm l, with the ASB inside a proprietary, permeable, black SEBS (Styrene Ethylene Butylene Styrene) membrane (100 g of ASB inside 16 cells of membrane Westham Innovations LTD, Tel Aviv, Israel).
Optimal bait station number on external house walls
Monitoring in each of eight villages was carried out with ten CDC UV (John W. Hock Co., Gainesville, FL, USA) light traps, set in the center of the village, 5 m from houses and at least 10 m away from each other in a rough grid pattern for eighteen nights in 2016. ASB station coverage consisted of one, two or three bait station(s) with dye on every house in the village; if more than one bait station was used, they were situated on opposite walls at a height of 1.8 m (Supplemental Fig. 1). On August 17, 24, 30, a single ASB station was hung at 1.8 m above the ground on each house as well as on September 04, 10, 16, and 27. Two bait stations were hung per house on August 19 and 26, and on September 02, 05, 12, and 20. three bait stations were hung per house on August 21 and 29 as well as on September 09, 13 and 25. CDC Traps were set at 18:00 h and emptied at 06:00 h the following day.
Effect of bait station height.
Monitoring with CDC Traps from 18:00 h to 06:00 h as described above was carried out on nine nights: July 25, 27, 29, August 05, 09, 08, 11, 12, and 15 with ASBs deployed in which 2 ASBs were situated on the same wall at 1.0 m and 1.8 m above the ground on two opposing walls of each house in the village. ASBs at 1.0 m contained green stain while the stain in the ASB positioned 1.8 m was orange.
Performance of fresh versus aged baits.
ASB coverage consisted of 2 bait stations on every house, side by side, 2 m apart, both 1.8 m above the ground, one fresh, the other aged under field conditions (by hanging outside near the study area) for 6 months each with differently coloured bait (orange and green). The trials ran for six days and nights, during which feeding rates of An. gambiae s.l. were evaluated and dates included October 1, 2, 6, 8, 11 and 13. Mosquitoes were collected in each village with 10 CDC-UV light traps set in the center of the village, 5 m from houses but at least 10 m away from each other. Traps were set at 18:00 h and emptied at 06:00 h the following day.
NTO monitoring
NTOs were monitored after the ASB deployment in each village with 50 yellow plates (yellow disposable plastic plates 25-cm diameter filled with water and a drop of Triton X-100 as detergent), four Malaise traps (6 m; John W. Hock Co., Gainesville, FL, USA), two ultraviolet light traps (generator powered 250W ML light bulb mounted in front of a white 2 m × 5 m white linen sheet), six ultraviolet tray traps [13], 50 pitfall traps (500 ml plastic cups buried to the rim in the ground, baited with 10 ml vinegar), sweep nets (BioQuip, Rancho Dominguez, CA, USA) (operated by two collectors), and aerial hand nets (BioQuip, Rancho Dominguez, CA, USA). Collected insects were stored in paper envelopes and petri dishes at -20°C before being processed for identification.
Stained specimens were separated from unstained specimens in the catches, stained specimens were pooled and identified with assistance from experts from the Entomological Department at the ZSM Natural History Museum (Zoologische Staatssammlung München). ASB feeding evaluation focused on seven of the most common orders. Feeding was determined by dissecting and examining guts for food dye under a 10X dissecting microscope. The insect orders were: Hymenoptera [ants (Formicidae) bees (Apidae), and wasps (Vespidae)], Lepidoptera (Rhopalocera, Bombyces, Geometroidea, Noctuoidea, Sphingidae, Pyraloidea), Coleoptera (Carabidae, Tenebrionidae, Scarabaeidae, Cerambycidae, and Chrysomelidae), Diptera (Brachycera, Chironomidae), Hemiptera (Cicadomorpha and Heteroptera), Neuroptera (Myrmeleontiformia) and Orthoptera (Caelifera and Ensifera).
Statistical Methods.
Bait Station Number Analysis. We used a generalized linear mixed model to compare the effect of the number of bait stations on the number of mosquitoes that fed on the bait and were caught in the traps. Female and male mosquitos were analyzed separately. The fixed effects in the model were village and number of bait stations which was a repeated measure. The random term was traps nested within villages that formed the error term for the repeated measure of number of bait stations. A compound symmetric covariance matrix was used to represent the correlated data structure. We included the total number of males and females trapped as an offset to produce model mean percent, standard error, and 95% confidence bounds. We also present p-values for comparisons of the mean percent of mosquitoes that fed on the bait among the number of bait stations.
Bait Station Height Analysis. This analysis compared the number of mosquitoes that fed on the bait and were caught by traps at a height of 1.0m and 1.8m. We used a generalized linear mixed model for Poisson outcome: the number of mosquitoes caught. Female and male mosquitos were analyzed separately. The fixed effects in the model were village and trap height which was a repeated measure. The random term was traps nested within villages that formed the error term for the repeated measure of height. A compound symmetric covariance matrix was used to represent the correlated data structure. We used the total number of males and females trapped as an offset to produce model mean percent, standard error, and 95% confidence bounds. We also present p-values for the comparison of the mean percent of mosquitoes that fed on the bait at each trap height.
Aged versus fresh baits. We used a generalized linear mixed model to compare the effect of bait age on the relative number of mosquitoes that fed on fresh and 6-month old ASB stations. Female and male mosquitoes were analyzed separately. The fixed effects in the model were village and bait age which was a repeated measure. The random term was traps nested within villages that formed the error term for the repeated measure of number of bait stations. A compound symmetric covariance matrix was used to represent the correlated data structure. We included the total number of males and females trapped as an offset to produce model mean percent, standard error, and 95% confidence intervals. We also present p-values for comparisons of the mean percent of mosquitoes that fed on the bait between the two ages of the bait.
Effect of bait station number on non-targets: We used a generalized linear mixed model to compare the effect of the number of bait stations on the number of insects that fed on the bait and were caught in the traps. The fixed effects in the model were village, type of insect, number of bait stations, and the interaction of insect type with number of bait stations. Type of insect and number of bait stations were repeated measures. The random term was traps nested within villages that formed the error term for the repeated measures. A compound symmetric covariance matrix was used to represent the correlated data structure. We included the total number of insects trapped as an offset to produce model mean percent, standard error, and 95% confidence bounds. For each insect group, we also present p-values for comparisons of the mean percent of insects that had fed on the bait, between number of bait stations.
Effect of height on non-targets: This analysis compared the number of insects that fed on the bait at a height of 1.0m and 1.8m. Separate analyses were performed for the different heights because the type of insects caught were different, depending on the height of the bait station. We used a generalized linear mixed model for Poisson outcome: the number of insects that fed on the bait. The fixed effects in the model were village and type of insect which was a repeated measure. The random term was traps nested within villages that formed the error term for the repeated measure of type of insect. A compound symmetric covariance matrix was used to represent the correlated data structure. We included the total number of insects trapped as an offset to produce model mean percent, standard error, and 95% confidence bounds. We also present p-value for the comparison among the mean percent of each type of insect that had fed on the bait.
Analyses were carried out using SAS 9.4 (SAS Institute, Inc.; Cary, NC). The two-tailed alpha level was used to determine the statistical significance of all statistical tests.