**4. Discussion**

There is a growing array of vector control tools based on insecticides which act via a range of different exposure routes. The Attractive Targeted Sugar Bait (ATSB) is the only one that involves ingestion by adult mosquitoes. One ATSB currently under evaluation includes dinotefuran, which mosquitoes feed on in a sugar-based bait. As with any insecticidebased intervention there is a need to monitor for the emergence of resistance in the target population, which conventionally has been done using testing methods which expose field caught mosquitoes of the target population via tarsal contact to a treated bottle [5] or filter paper [4]. Because the exposure route of an ingested insecticide is different to a contact insecticide, the results of these tests may not be an accurate indicator of resistance and risk of failure of ATSBs.

This study considered two alternative methods to screen for resistance. The first was an oral ingestion assay developed to prevent tarsal contact while allowing ingestion of a spiked sugar meal, the most direct test for resistance to an oral insecticide. The assay was demonstrated to be robust and quantitative enough to be able to establish a dose response in laboratory strains of *Anopheles*, including in two strains that are highly resistant to pyrethroids. By using a uranine marker, individual mosquitoes that fed were identified, and in the controlled laboratory setting of these experiments the feeding rate was high. However, the applicability of the sugar feeding assay used here in the field is limited due to the large variation of sugar feeding behavior—and thus insecticide uptake—when applied to field caught mosquitoes and using a less controlled laboratory environment. It is likely that the feeding rate, which was high in laboratory strains adapted to feeding on an artificial sugar source, would be much lower in field caught adults or adults emerging from field collected larvae. A low feeding rate would further increase the resources required to produce significant data which could be relied upon in a screen for emerging resistance. Even in these experiments conducted in tightly controlled laboratory conditions and with mosquitoes reared using standardized protocols [9], the results were varied. The methodology would be difficult to standardize sufficiently that it could be performed in multiple field sites, likely with less controlled environments, and achieve robust and comparable data. The ingestion bioassay method also requires greater resources in terms of space and time than topical application, and access to a fluorescent microscope.

The consistently high feeding rate across treatments in this study, which was not correlated with concentration of dinotefuran, suggests a lack of any detectable repellent effect of the dinotefuran. However, in adapting this method for other insecticides, there is a risk of a repellent effect reducing the feeding rate. To avoid this as a possible confounding factor, it is important to use some methods to eliminate individuals that do not feed from mortality scoring, either including uranine and scoring fluorescence as done here or using an alternative such as Trypan blue dye [6,7]. Another possible confounding factor is that it has not been established exactly how long mosquitoes may survive after ingesting particularly lower concentrations of insecticide without direct observation for the whole exposure period. It is possible mosquitoes may have fed just before collection and be scored as survivors when they may have died even minutes later. However, the assay still demonstrated sufficient sensitivity to measure a difference in mortality between concentrations in a dose responsive manner, so this does not appear to prevent the ingestion assay being applicable for this purpose.

Because of the logistical challenges of the ingestion assay, a practical and well established topical application bioassay [14] was also used, applying dinotefuran to the thorax of adult *Anopheles* mosquitoes with organic carrier to bypass lipid cuticle barriers. The direct application of insecticide to the mosquito thorax bypasses the need for uptake of insecticide from a surface and penetration of the insecticide through the cuticle, and mortality as a result of this exposure route has been shown in other insects to correlate well with oral toxicity [8]. The variability of the data between replicates is less with topical application because parameters which define the dose taken up by the mosquito are less variable than for the ingestion assay, producing more robust data.

This study compared the topical application dose response with the respective response of the oral toxicity bioassays, across four *Anopheles* strains. The dose response curves plotted for the same strains were very similar, and there was similar relative susceptibility between the two strains tested with both methods. Topical application is a well-established method and relatively easy to apply [15]. The similarity of results between the two methods demonstrated here suggests that a topical application-based DC, determined based on WHO guidelines, could be used as a proxy for monitoring the development of resistance in field populations to orally ingested dinotefuran from ATSB stations. However, tissuespecific resistance mechanisms are not well studied and there is risk that topical application will not pick up on the emergence of an as ye<sup>t</sup> unidentified ingestion specific mechanism. If potential resistance is observed in results of susceptibility monitoring using topical application, further investigation would be warranted, including exploring such possible mechanisms using ingestion assays in the laboratory. Similarly, the results of topical testing could be affected by the presence of cuticular resistance, through cuticular thickening, altered cuticle composition, or alterations in receptors that affect uptake and penetration of active ingredients. Such resistance mechanisms may be primarily overexpressed in the tissues of the mosquito that are typically in contact with insecticides, such as the tarsi. The direct application of acetone to the thorax (the solvent used to deliver insecticides during topical application) is believed to bypass these mechanisms, and no correlation between cuticular resistance and reduced mortality by topical application has ye<sup>t</sup> been reported. The WHO approach to establishing a DC is to perform dose response experiments and establish LC values for a range of strains susceptible to the insecticide being tested, and then to select the highest DC established for the least susceptible strain, based either on the calculated LC values (DC = 2 × LC99) or an observed LC100, defined as the lowest concentration tested which reliably produces 100% mortality in susceptible strains [4]. A pragmatic decision may be made as to whether to recommend a specific DC for each species or to select the highest DC to use for a group of species, and sometimes rounding the calculated DC to a value more easily applied in field testing [16]. Based on the dose response observed and LC values calculated in this study, a DC of 100 ng/mosquito would be recommended for topically applied dinotefuran for *An. gambiae* and *An. funestus*. However, a lower tentative DC of incipient resistance, at 10 ng/mosquito, is highly recommended as well, to collect baseline susceptibility data and capture possible variation in bioassay responses among populations in the ATSB trial sites. These trials will further validate the methodology, as well as define the most appropriate DC for screening field *Anopheles* populations.

The relative susceptibility of Kisumu and VK7 2014 strains was measured by each method by calculating a resistance ratio of 5.4 by ingestion and 1.8 by topical application. Both these ratios are very low, all below 10, and do not indicate that there is resistance, but rather inherent variability in susceptibility between strains. A more robust validation of the correlation of results between the two methods by repeating this study with a strain known to be resistant to dinotefuran, or neonicotinoids in general, would help to confirm comparability of results from the two methods but to date no such resistance has been reported in field caught mosquitoes and so no such laboratory strain is available. Another possibility would be to make use of transgenic strains which have resistance to neonicotinoids induced, using a method such as CRISPR/Cas9 [17,18].

The lethal dose of dinotefuran was lower by ingestion than by topical application (approximately 20 times). This points towards a higher toxicity when ingested, though there are several sources of variability in calculating the precise dose of dinotefuran ingested in the sugar feeding assay that mean a direct comparison cannot be made. These include different meal sizes taken by individual mosquitoes, related to body size and previous handling, some individuals taking full sugar meals and some only partial feeds, and the possibility that different volumes are ingested in treatments where insecticide is added.

The volume of bait ingested from an ATSB may be different to the volume of sugar water ingested in this assay, and the size of sugar meal may differ between mosquito populations. It is not, therefore, possible to directly compare the toxic ingested dose of dinotefuran in this assay with the actual dose of dinotefuran in the ATSBs and predict efficacy against target mosquitoes. We can, however, make some estimations based on the assumption that a similar volume of bait is ingested from an ATSB station by wild mosquitoes. The Westham ATSB stations currently under evaluation contain 0.1% dinotefuran, so a mosquito taking up 0.4 μL of bait (Appendix A) will ingest 400 ng of insecticide, 300 times the dose shown to kill 100% of mosquitoes in the ingestion assay. This means that if only 0.0013 μL of bait is ingested, it will be lethal to the mosquitoes. Based on the calculated LC50, an amount consumed 3000 times lower than the typical sugar meal would be sufficient to kill 50% of the mosquitoes which feed on it. The bait stations should continue to be effective in populations even where resistance is seen to have emerged through monitoring, using the relatively sensitive DC which has been established.

Practically, no cross-resistance between dinotefuran and pyrethroids was observed. The pyrethroid resistance ratio for these same strains tested with permethrin previously were 145.77 (149–397) via topical application and 128.23 (81.4–198.5) via a tarsal assay [9], but the response of the same strains to dinotefuran was essentially not different or indicated a very low cross-resistance. These data confirm the utility of chemicals belonging to different Insecticide Resistance Action Committee (IRAC) mode of action (MoA) classes [19] (i.e., different target sites and/or routes of uptake for insecticide resistance managemen<sup>t</sup> (IRM)). Neonicotinoids are nicotinic acetylcholine receptor competitive modulators (IRAC class 4A), with a high selectivity in binding to insect nicotinic acetylcholine receptor sites over that of mammal receptors [20]. The target of action is thus different to pyrethroids, which are sodium channel modulators (IRAC class 3A). None of the WHO Prequalified Vector Control Products contain dinotefuran, and so it is unlikely that mosquitoes have been exposed and developed resistance to dinotefuran. Fludora® Fusion and SumiShield 50WG IRS formulations contain clothianidin, also a neonicotinoid, and their potential for use against pyrethroid resistant insects has been demonstrated [21,22]. Extensive experiments with a proposed diagnostic concentration of 2% *w*/*v* clothianidin on filter papers failed to find conclusive evidence of resistance in 43 sites in sub-Saharan Africa [23], or in *Anopheles arabiensis* in Ethiopia [24]. No evidence of resistance to clothianidin was found in western Kenya using a DC of 150 μg/bottle [25]. However, once an insecticide is being deployed a selection pressure is applied and there is a risk of resistance evolving, and so once validated, the DC should be used to perform regular resistance monitoring in all sites where ATSBs are deployed. Validation of this methodology should also be carried out for any future insecticides used in new ATSB designs. There is also a risk of cross-resistance to dinotefuran as a result of exposure to other neonicotinoids used for vector control or in agriculture.
