Testing a pyriproxyfen auto-dissemination station attractive to gravid Anopheles gambiae sensu stricto for the development of novel attract-and-kill strategies for malaria vector control

Background Larval source management is an effective supplementary tool for malaria vector control although it is not used widely in sub-Saharan Africa. This study explored whether an attract-and-kill strategy could contaminate gravid Anopheles gambiae sensu stricto with the insect growth regulator, pyriproxyfen, at a bait-station, for dissemination to larval habitats. Methods A bait-station comprising an artificial pond, containing water was treated with 20 ppm cedrol, an oviposition attractant, was covered with pyriproxfen-treated netting. Three identical semi-field cages were used to assess the potential of gravid Anopheles gambiae sensu stricto to transfer pyriproxyfen from the bait-station to three open ponds. Gravid females were released in the test and one of the control cages that had no pyriproxyfen on its bait-station. No mosquitoes were released in the third cage with a pyriproxyfen-treated station. Transfer of pyriproxyfen to open ponds was assessed by monitoring emergence of late instar insectary-reared An. gambiae sensu stricto larvae introduced into the open ponds. Liquid chromatography-mass spectrometry was used to quantify the amount of pyriproxyfen carried by a mosquito and the amount transferred to water. Results 86% (95% CI 81-89%) of larvae introduced into the open ponds in the two control cages developed into adults. Transfer of pyriproxyfen to the test cage depended on the distance of the pond from the bait-station. While only 25% (95% CI 22-29%) adult emergence was observed in larvae introduced into ponds 4.4 m from the bait-station, the emergence rates increased to 92% (95% CI 89-94%) in larvae introduced in ponds 10.3 m away. Each mosquito was contaminated with 112 µg (95% CI 93-123 µg) pyriproxyfen, whilst 230 ng/L (95% CI 180-290 ng/L) was transferred by a single female to 100 ml of water. Conclusions Pyriproxyfen was auto-disseminated by gravid females from attractive bait-stations, but mainly to

3 delivery systems are needed.
Background Improved access to the core malaria control interventions namely vector control, effective diagnosis and prompt treatment have greatly contributed to the global reduction in malaria morbidity and mortality (1-3). However, recent World Malaria Reports from 2017 and 2018 indicate that this remarkable progress has stalled (1,4). This worrying trend emphasizes the need to explore additional tools for malaria prevention to supplement the current frontline measures and ensure that the gains achieved in the last decade are sustained (5)(6)(7).
Malaria control programs are encouraged to adopt integrated vector management strategies to increase efficacy, cost-effectiveness and sustainability of disease control (7,8) . Larval source management (LSM) targeting immature mosquito vectors in their aquatic habitats, such as larviciding and environmental management, can effectively serve as complementary vector control measures (9,10). Studies in East Africa highlighted the benefit of combining long-lasting insecticidal nets (LLINs) and larviciding with microbial larvicides for reducing transmission (11)(12)(13)(14). However, the challenge associated specifically with larviciding is the need to reach all available and potentially suitable aquatic habitats in an intervention area (15)(16)(17)(18), many of which might only be reached by aerial application. Whilst it has been suggested that larviciding might be targeted only at a proportion of most productive habitats (19,20), habitat productivity is still poorly understood and not easily predicted by application teams (21)(22)(23).
Auto-dissemination, a novel strategy that exploits the adult insect as a 'vehicle' to deliver insecticide, might be one way of addressing this challenge. This strategy has been shown to be effective in the control of Aedes mosquito (24)(25)(26)(27) leading to an increased interest in its exploration for control of Afrotropical malaria vectors (28,29). Semi-field studies 4 conducted in Tanzania provide the first evidence of the potential of An. arabiensis to transfer the insecticide pyriproxyfen (PPF) from resting surfaces to larval habitats, consequently inhibiting larval development (30)(31)(32). However, these studies were implemented at extremely high mosquito population densities that are unlikely to occur under natural conditions. Furthermore, the targeting of host-seeking mosquitoes before or shortly after bloodmeals for contact with PPF is likely to cause a large proportion of the females to get sterilized and not develop into gravid females (33)(34)(35)(36)(37). We recently showed that a female that is not gravid is significantly less likely to visit an oviposition site and hence transfer PPF to a water body, and that the optimum time for exposing female Anopheles gambiae sensu stricto to PPF for auto-dissemination is close to oviposition (35).
Consequently, the aim of this study was to design an attractive bait-station to contaminate gravid An. gambiae s.s. with PPF and to test the efficiency of PPF transfer to open ponds under semi-field conditions.

Study site
Experiments were carried out in large netting-screened semi-field cages (10.8 m long × 6.7 m wide × 2.4 m high) at the International Centre of Insect Physiology and Ecology, Thomas Odhiambo Campus (icipe-TOC), located on the shore of Lake Victoria in Mbita, Homa Bay county, western Kenya (geographic coordinates 0⁰ 26' 06.19" S, 34⁰ 12' 53.13"E; altitude 1,137 m above sea level). The cages had a sand floor and did not contain any vegetation. Mbita is characterized by tropical climate with an annual average minimum temperature of 16 ºC and maximum temperature of 29ºC. The area experiences two rainy seasons; the long rains between March and June and the short rains between October and December.

Mosquitoes
Anopheles gambiae s.s. (Mbita strain) larvae and adults used in this study were obtained from the mosquito insectaries at icipe-TOC. Immature stages were reared in a semi-field cage at ambient conditions with average daily temperature of 25-28ºC, relative humidity of 68-75% and natural lighting. Mosquito larvae were reared in round plastic tubs (diameter 60 cm) filled with 5 L water (5 cm deep) from Lake Victoria filtered through a charcoal-sand filter. Mosquito larvae were fed with a pinch of fish food (Tetramin©Baby) twice daily. Third (late) instar mosquito larvae were randomly selected from different tubs to ensure that cohorts of larvae used in experiments were a representative sample of the size distribution of the experimental larval population. Adult mosquitoes were held in mosquito-netting covered cages (30 cm x 30 cm x 30 cm) in a holding room with ambient climate conditions and provided with 6% glucose solution ad libitum. Three-day old females were allowed to feed on a human arm on two consecutive nights. Gravid mosquitoes were used for experiments in this study.

Development of a bait-station
Contamination of adult An. gambiae s.s. with PPF. Water vapour has been shown to attract gravid malaria vectors (38) and hence it was considered essential to include water in the bait-station. Females were prevented from accessing the water to lay eggs using fly gauze (black fibre-glass netting gauze, mesh size 1 mm x1 mm). To determine the best method to treat netting surfaces with PPF for efficient contamination of mosquitoes, preliminary cage tests were conducted in small-sized cages (30 cm x 30 cm x 30 cm). Two methods of applying PPF on the netting gauze that served as the dissemination platforms were 6 compared. First, the netting gauze (diameter 7 cm) was treated with 1 g of PPF dust applied with a soft brush to ensure uniform spreading of the PPF over the netting surface.
Second, 1 g of PPF dust was mixed with 2 ml of cooking oil and applied to the netting gauze with a soft brush and left to dry in the air for 30 minutes. The rationale for this was to test a formulation that would be easier to apply and less likely to be distributed by wind.
Each experimental cage was provided with two glass cups (Pyrex®, capacity 100 ml, diameter 7 cm) and the cups were placed at the diagonal corners of the cage, approximately 26 cm apart. The first cup in each cage was filled with 100 ml nonchlorinated tap water and left open for gravid mosquitoes to lay eggs. The second cup, serving as the bait-station, was filled with six-day old soil infusion previously shown to attract gravid An. gambiae s.s. (39). Soil infusion was prepared by incubating 15 L of nonchlorinated tap water with 2 kg of soil collected from a known An. gambiae s.l. habitat which was dry at the time of collection (39). Infusions were prepared in round plastic tubs (diameter 0.42 m) and left for six days before use in experiments as described in detail previously (39). The top of the bait-station in the control cages was covered with untreated netting gauze while in the test cages it was covered with netting gauze treated with either PPF dust or PPF dust formulated in oil. In each cage five gravid An. gambiae 10.3 m away from the bait-station ( Figure 2). The set-up in the second semi-field cage was the same as the first, except that no mosquitoes were released in the cage. The aim here was to investigate if PPF might be distributed by air movement to neighbouring ponds, rather than mosquitoes. In the third semi-field cage, mosquitoes were released but the netting gauze placed on top of the bait-station was left untreated. This set-up served to investigate natural adult mosquito emergence rates from ponds when no insecticide was present in the semi-field cage. Two hundred gravid An. gambiae s.s. were released at 18.00 h per experimental night inside the hut and allowed to disperse through the open eaves.
The following morning, all open ponds in the three semi-field cages were visually assessed for the presence of eggs laid. Eggs were not counted since an exact estimate would have required removing the eggs from the ponds using a sieve or similar tools potentially interfering with the amount of PPF transferred. To ensure sufficient replication of the experiment, the impact of PPF was not assessed by monitoring the development of eggs that were laid by the exposed females which would have taken approximately two weeks to complete a single replicate and over half a year to complete 12 replicates (42). Instead, the possible transfer of PPF by females to the ponds was assessed by monitoring the adult emergence of 50 insectary-reared late instar An. gambiae s.s. larvae that were introduced into open ponds in all three experimental set-ups in the morning after gravid females were released. Introduced larvae were fed daily with a pinch of Tetramin®Baby Fish food. Any pupae that developed in the three ponds were transferred with a small volume of water from the pond into 200 ml plastic cups (diameter 7 cm) and monitored for adult emergence. It took 6-7 days for all introduced larvae to develop into adults or die.
Thereafter the ponds and hut were cleaned and all remaining flying adult mosquitoes in cages aspirated using a motorized backpack aspirator (John W. Hock Company, USA). A new round of replicates was set-up with fresh oviposition substrates and fresh batches of gravid mosquitoes and mosquito larvae. The experiment was replicated over 12 rounds with each round lasting seven days. The four ponds were randomly allocated in all four corners of the three semi-field cages in a randomized complete block design. To avoid contamination, the semi-field cages in which the test and the two control experiments were conducted were not exchanged.

Liquid-chromatography-mass spectrometry (LC-MS) quantification of the amount of PPF carried by an individual mosquito and transferred to a water sample
An enamel bowl (diameter 0.42 m) filled with 7 L of non-chlorinated tap water was introduced into a 60 x 60 x 60 cm cage (BugDorm-2120F; MegaView Science Taiwan). The top of the bowl was covered with black fibre-glass netting gauze treated with 3.5g PPF dust (average 20.3 g PPF/m 2 retained on gauze on weighing) as described above. Gravid An. gambiae s.s. were introduced into the cage and monitored for contact with the PPFtreated netting gauze. At any time, there were only two females in the cage. Females that contacted the PPF-treated netting material at least twice were gently aspirated from the cage into holding containers.
Two different tests were conducted with females that contacted the PPF-treated netting.
First, 200 potentially contaminated females were individually transferred into 1.5 ml Eppendorf tubes and frozen at -70°C until they were used to quantify the amount of PPF on their body. Secondly, 30 potentially contaminated females were used to determine the amount of PPF that a single mosquito transfers to water during oviposition. For this, bioassays were conducted by introducing these females individually into 15 cm x 15 cm x 15 cm cages provided with a glass cup (Pyrex®, 100 ml, diameter 7 cm) filled with 100 ml of non-chlorinated tap water. The females were left overnight in the cages to lay eggs.
The following morning the glass cups were assessed for presence of eggs laid.
To confirm the transfer of PPF into the oviposition water, 10 insectary-reared late instar An. gambiae s.s. larvae were introduced into all cups in which females had laid eggs and monitored for adult emergence as described above. When all introduced larvae had died or emerged as adults, the water from the cups was transferred into 50 ml glass jars. The water samples were frozen at -70°C awaiting chromatographic quantification of PPF in the samples. Comparisons were made to a control group of gravid females that were unexposed to PPF. Thirty replicates of test and control cages were done. source temperature, 110ºC; nitrogen gas temperature for desolvation, 350ºC; and nitrogen gas flow for desolvation, 400 L/h.

Data analysis
Data were analysed in R statistical software package version 2.13. Generalized estimating equations (GEE) were used to analyse all data with experimental round/night included as repeated measure in the models (43,44). Data collected in cage and semi-field experiments to determine the transfer of PPF to water were analysed as proportions.
Proportions were analysed by fitting a binomial distribution with a logit function and an exchangeable correlation matrix. Preliminary cage bioassays testing the two PPF formulations were analysed by including treatment (control or test) as fixed factor with the control cage used as the reference (45).

Gravid Anopheles gambiae s.s. pick up more PPF when the PPF is only dusted on the netting than when PPF is formulated in oil
Both application methods of PPF on the netting of the bait-stations led to the transfer of PPF to the open cup by gravid females and significant reduction in the emergence of adults from introduced larvae ( Figure 3 and Table 1). However, the adult mosquito emergence rates were five times lower when the PPF was applied on the netting as dust 14 than when it was formulated in oil.

Oviposition attractants can lure gravid An. gambiae s.s. to a bait-station under semi-field conditions
The number of mosquitoes trapped on the sticky screens placed over ponds containing soil infusion or cedrol at 5 ppm or 20 ppm was significantly higher than the number trapped over ponds with untreated water ( Table 2). The attractiveness of soil infusion and water treated with 5 ppm cedrol was similar and not very strong; a female was only approximately 1.3 times more likely to land on the test pond than on a control pond (Table   2). However, when the water was treated with 20 ppm of cedrol, it was approximately twice as likely for a female to be trapped on the single test pond compared to any of the three control ponds in the experiment (Table 2).

Transfer of PPF by gravid An. gambiae s.s. is dependent on the distance of the habitat from the bait-station
In all semi-field cages where gravid females were released inside the hut, eggs were observed the following morning in all open ponds on any experimental night. In the absence of PPF on the bait-station as well as in the absence of gravid females in the semifield cage, on average 82% (95% CI 81-89%) of larvae introduced emerged as adults. For some unexplained reason there were differences in the emergence rates of larvae introduced into the ponds in the two control set-ups.  The emergence of larvae in the experiment where no mosquitoes were released but PPF was present on the bait-station was consistently higher than in control experiment where mosquitoes were released but no PPF was present in the cage (Table 3). This might have been due to some microclimate differences in the two semi-field cages or an unexplained interaction between the early instars originating from eggs laid by gravid females and  (Table 4). It was 17 times less likely for a larvae to emerge when it was introduced into water in which PPF exposed female had laid eggs than when introduced into a cup in which unexposed female had laid eggs (  Test -PPF-exposed females 0.45 (0.39-0.51) 0.06 (0.03-0.10) 0.007 Based on the control emergence, the corrected percent emergence inhibition observed was 52% (95% CI 46-56%). In other words, an individual female transferred the concentration of PPF to 100 ml of water that inhibited emergence of approximately 50% of larvae (EI 50 ).
The preliminary chemical analysis showed that it was impossible to detect the low concentration of PPF that was washed off the body of a single mosquito, thus, samples which is equivalent to 0.00023 mg/L (0.00023 ppm). Similarly, this is likely to be an overestimate since in three of the test water samples PPF was below the detection limit.
Based on the bioassay results, this is the concentration that provided around 50% emergence inhibition of introduced larvae.

Discussion
To our knowledge this is the first study to have developed a prototype bait-station for gravid An. gambiae s.s. for the auto-dissemination of PPF to aquatic habitats. We show in principle that gravid females can be attracted to a target, can successfully pick up PPF and then transfer it to an aquatic habitat while laying eggs consequently killing immature mosquito stages. However, even though 200 gravid females were released in a relatively small space of approximately 170 m 3 , adult emergence from larval habitats was only inhibited by around 70% from ponds located less than 5 m from the bait-station but not from habitats 10 m away. These results strongly suggest that even if females can be lured successfully to a bait-station, they are most likely to transfer the PPF only to the closest available and detectable oviposition sites.
This study highlights a number of challenges that need to be taken into account for the development of an efficient auto-dissemination approach for African malaria vectors.
Based on previous work we consider exposing gravid females to PPF the most effective way to ensure transfer to an aquatic habitat given that exposure earlier in a female's life has significant impact on her fertility and fecundity (33,34,36,46), lowering the females 20 predisposition to visit an aquatic habitat. Attracting the gravid female is however more challenging than attracting a host-seeking or resting female, due to the scarcity of synthetic oviposition attractants (38,47,48).
Our study confirms previous findings that the chemical compound, cedrol (38), attracts gravid An. gambiae s.s. and might be used in a novel attract (-release) -and-kill approach.
However, contrary to earlier findings by Lindh et al. (38) where twice as many gravid females were recovered with 5 ppm cedrol than with water alone, we only achieved the same attractiveness with 20 ppm cedrol. This difference is likely due to the absence of Gravid females transferred more PPF from treated surfaces when PPF was applied as a dust than when formulated in oil. There are two possible explanations for this. First, the oil might reduce the transfer of PPF to mosquitoes, with more of it adhering to the netting 21 surface. Second, it might also be that the oil contributed to a larger proportion of PPF remaining on the mosquito's body, thus limiting the chance of PPF getting in contact with water. Presumably, the hydrophobic oil attaches more strongly to the hydrophobic cuticle than to water (45). In the approach used in this study, a lot of the active ingredient remained on the netting gauze and was wasted since not all the material was taken up.
Furthermore, as shown in the chromatographic analyses, an individual female transferred >4,800 times less PPF to the oviposition cup during egg-laying than the amount the female picked up from the treated netting. PPF on the insect cuticle is likely to decrease with time due to loss during flight, resting and penetration through the insect cuticle (50)(51)(52). The chromatography confirms our findings from the bioassay, that a single female could transfer sufficient PPF to inhibit the emergence of 50% (EI50) of the larvae in 100 ml of water. The average concentration of PPF detected in water used in the bioassays was 0.00023 mg/L (95% CI 0.000180-0.000290 mg/L), which correlates well with previously published results from laboratory assays when the EI 50 was established at 0.000120 ng/L (95% CI 0.000090-0.000160 /L) (49). The findings are also consistent with previous cage bioassays where females were contaminated in a plastic jar coated with PPF and a single female caused approximately 50% of the introduced larvae not to emerge (33). Taken together it appears that this is the average amount of our test formulation that a gravid An. gambiae s.s. can transfer to an aquatic habitat in a test system like ours. In our system, approximately 500 gravid females would be required to visit a water body 1 m 2 in area and 10 cm deep, to transfer the optimal lethal concentration determined in the laboratory, which is highly unlikely under most natural conditions as gravid females comprise a very small proportion of the overall mosquito population. For largescale application and cost-effective use of the relatively expensive active ingredient there 22 is need to investigate strategies that use PPF more efficiently. Improved technologies such as electrostatic charging of PPF to enhance adherence of insecticide particles on contact with contaminated surface and ensuring delivery of larger amounts of insecticide will be beneficial (49). Whether this would improve the amount transferred to water or only increase the overall amount carried by the individual mosquito would need to be tested.
Mathematical models show that the success of auto-dissemination for malaria vector control is dependent on the abundance of adult vectors, the number and stability of larval habitats and persistence of the insecticide used (27,28). The auto-dissemination approach for controlling immature malaria vectors has received attention specifically to target habitats that are not easily accessible for insecticide application teams on the ground (10). This study highlights that the transfer of PPF to a larval habitat is dependent on the distance of the pond from the bait-station, with the closest suitable habitats being the most likely to be visited and hence the most likely to receive sufficient amounts of PPF to have an impact on immature mosquito development. It is likely that vegetation cover, which was on purpose excluded from the experimental design might have further impacted on the ability of finding a suitable habitat and on the amount of PPF lost due to resting (52,53). Numerous bait-stations would therefore be required in the field for gravid mosquitoes to transfer sufficient lethal doses of PPF to their larval habitats. This is a substantial challenge considering the large number and extensive nature of water bodies utilized by African malaria vectors (18,22).
Comparable semi-field studies to evaluate the possibility of auto-dissemination with PPF were done targeting host-seeking and resting An. arabiensis for PPF exposure using treated clay pots (29)(30)(31). The efficiency of transfer reported from these studies were significantly higher, with an emergence inhibition of over 80% from the provided aquatic habitats in these studies. Factors that might explain the greater impacts in these studies include the use of numerous resting posts as dissemination stations for PPF and fewer, smaller-sized aquatic habitats. For instance, in two of these studies the authors placed eight resting pots treated with PPF and provided only 2-4 small aquatic habitats of a 2.5 L capacity (29,30). This is a ratio of dissemination station to oviposition sites of 4:1 or 2:1.
In comparison, our study had a ratio of 1:3. Furthermore, in the studies with An.
arabiensis, a very large number of 1,500-5000 females were either introduced into the system, or reared inside the system (30)(31)(32), maximising the likelihood of a mosquito visiting a PPF-treated resting pot and the number of oviposition events in a single aquatic habitat.
A limitation of our study was the introduction of insectary-reared late instar larvae into test ponds rather than monitoring the larvae that hatch from the eggs laid by PPF-exposed females for adult emergence inhibition. While PPF has greater pupicidal effects, the impact in reducing adult emergence is higher on immature stages that have prolonged exposure to the insecticide during larval development (54). It is likely that greater adult emergence inhibition rates would have been observed if the effect on emergence inhibition was assessed on mosquito larvae exposed to the insecticide since hatching from eggs.

Conclusions
Our study carried out under controlled conditions highlights potential limitations of the auto-dissemination strategy for the control of Afrotropical malaria vectors. Our results emphasize the need to investigate the required ratio of bait-stations to aquatic habitats for adult gravid females to transfer sufficient amounts of PPF for efficient control of immature malaria vectors in all aquatic habitats. The skip-oviposition behaviour recently observed in An. gambiae s.s. in cages (45) and in An. arabiensis in the field (55) is likely to benefit the auto-dissemination approach for malaria vector control since gravid females visit several habitats to lay eggs. Nevertheless, significantly more work is required in designing highly attractive bait-stations for gravid malaria vectors by identifying more attractants and composing highly attractive chemical blends, determining better mechanisms for optimum release of the attractants, identifying better and more costeffective mechanisms for retaining and dispensing of PPF as well as improving the physical components of the bait-station to provide protective barriers from rain.
Additionally, insecticides of greater persistence in the environment than PPF, such as novaluron (26)