A pyrethroïd-treated bed net increases host attractiveness for Anopheles gambiae s.s. carrying the kdr allele in a dual-choice olfactometer

The use of long lasting insecticide nets (LLINs) treated with pyrethroïd is known for its major contribution in malaria control. However, LLINs are suspected to induce behavioral changes in malaria vectors, which may in turn drastically affect their efficacy against Plasmodium sp. transmission. In sub Saharan Africa, where malaria imposes the heaviest burden, the main malaria vectors are widely resistant to pyrethroïds, the insecticide family used on LLINs, which also threatens LLIN efficiency. There is therefore a crucial need for deciphering how insecticide-impregnated materials might affect the host-seeking behavior of malaria vectors in regards to insecticide resistance. In this study, we explored the impact of permethrin-impregnated net on the host attractiveness for Anopheles gambiae mosquitoes, either susceptible to insecticides, or carrying the insecticide resistance conferring allele kdr. Groups of female mosquitoes were released in a dual-choice olfactometer and their movements towards an attractive odor source (a rabbit) protected by insecticide-treated (ITN) or untreated nets (UTN) were monitored. Kdr homozygous mosquitoes, resistant to insecticides, were more attracted by a host behind an ITN than an UTN, while the presence of insecticide on the net did not affect the choice of susceptible mosquitoes. These results suggest that permethrin-impregnated net is detectable by malaria vectors and that the kdr mutation impacts their response to a LLIN protected host. We discuss the implication of these results for malaria vector control.


Introduction 35
Anopheles gambiae is one of the major mosquito vectors of human malaria parasites in sub-36 Saharan Africa. Its remarkable vectorial capacity [1] mainly relies on its high degree of anthropophily. 37 Moreover, An. gambiae prefers to bite humans indoors and often rests inside houses after blood 38 feeding [2-4]. These behavioral preferences led to the development of insecticide-based indoor 39 vector control measures, such as insecticide-treated bed nets (ITNs) and indoor residual spraying 40 (IRS), to limit the human-vector contacts and reduce mosquito survival. To date, four insecticide 41 families are available for IRS (organochlorides, organophosphates, carbamates and pyrethroïds), 42 whereas only pyrethroïds are recommended for mosquito nets because of their low mammalian 43 toxicity and high insecticidal potency [5]. 44 To kill, insecticide molecules must contact and penetrate through the mosquito cuticle/gut 45 to then reach and interact with their target before being degraded. Any physiological or behavioral 46 mechanism that may interfere with one of these steps can lead to insecticide resistance. The 47 widespread use of pyrethroïd (PYR) insecticides in malaria vector control and agriculture has favored 48 the development of resistance in malaria vector species [6]. One of the most studied physiological 49 mechanisms involved in PYR resistance is the reduced sensitivity of the voltage-gated sodium 50 channels to PYR binding caused by non-silent mutations, known as knockdown resistance (kdr) 51 mutations [7]. Behavioral resistance is another mechanism involved in PYR resistance. This can be 52 defined as a modification of the mosquito behavior to avoid contact with a lethal dose of insecticide 53 [8]. To date, behavioral resistance to insecticides remains poorly documented, despite of its huge 54 potential impact on malaria transmission. 55 Behavioral adaptations to pesticides can be classified as stimulus-dependent or -independent 56 [9]. Stimulus-independent adaptations are not associated with the perception of chemicals, but more 57 probably with modifications of the vector intrinsic behavior, such as changes in host-seeking 58 Experimental set-up 106 The dual-choice olfactometer was adapted from Geier and Boeckh (1999) [30]. It was made of 107 Plexiglas and was divided in four parts: release zone (RZ), flight chamber (FC) and one collecting zone 108 in each of the two arms (A1 or A2) (Fig 1). Rotating doors made from mesh gauze in the RZ and in 109 both arms allowed mosquito release or capture. The upwind part of the experimental set-up was 110 composed of a wide chamber where an attractive host (a rabbit) can be placed, and that was 111 connected to two treatment boxes that contained or not the nets. Each treatment box was 112 connected to one arm of the olfactometer. In order to avoid any perturbation on the airflow by the 113 treatment, fans were placed on the downwind faces of the experiment boxes and extracted the air 114 from the treatment boxes to the olfactometer, providing the odor-laden air current. At the beginning 115 of each experiment, the airflow was measured in arm 1 and 2 and in the release zone using a Testo © 116 435-1 multifunctional meter (Testo, Forbach, France) and thermo-anemometric probe (m.s -1 ) and 117 adjusted at 0.20 ± 0.03 m.s -1 . During the experiment, a thick black tarpaulin covered the olfactometer 118 to keep all the system in darkness and avoid visual disturbance. 119  Assays for the four experiments were performed every day for 20 days between 10:00am and 140 14:00pm (corresponding to mosquito strain feeding time in laboratory). We always started with 141 assays of experiment 1, to check possible odor or insecticide contaminations. When possible (i.e., 142 when the insectary production was sufficient), females of the three genotypes were tested the same 143 day for the four experiments, otherwise at least two genotypes were tested the same day (a 144 summary of the assays is presented in supplementary data). Each day, in assays for experiments 3 145 and 4, treatments were rotated one time between boxes to prevent any arm effect. Between 146 rotations, the boxes were carefully cleaned with ethanol to avoid any residual insecticide effect. 147 Moreover, the olfactometer was cleaned with ethanol every day. The experimenter wore latex 148 gloves to avoid contamination. The same rabbit was used as odor source during all the study. It was a 149 1-year old female reared in the same conditions as those used in insectaries to feed mosquitoes. CO 2 150 concentration and relative humidity (RH) were monitored in each arms with a Testo © 435-  Relative attractiveness of UTN vs. empty box and ITN vs. UTN were analyzed using a similar model 214 that, in addition, allowed for random effects of the box that received the treatment: 215 , where RA ikl is the proportion RA exp3 or RA exp4 for genotype i on day k with the treatment placed in 217 box l, F C H#IJ45%# indicates the effect on the logit of the classification in category i (SS, RS or RR) of 218 Genotype; b l , the effect on the logit of the the box l that received the treatment (UTN or ITN for 219 RA exp3 and RA exp4 , respectively) and d k , the random intercept for day k. Each genotype was 220 successively used as reference class for multiple comparisons. Odds ratios and their 95% CI were 221 computed. 222 CO 2 concentrations were compared between arms using the Wilcoxon signed-rank test for paired 223 data. RH values were compared between arms using the paired T test. 224

225
Overall, 6286 mosquitoes were included in the assays (2621 SS, 1268 RS and 2397 RR) during 226 47, 49, 84 and 98 releases for experiments 1 to 4 respectively ( Table 2). 227   (Table 3). However, the comparison of the upwind 242 flight probability between genotypes show that for RS mosquitoes, UF probabilities became 243 significantly higher than for SS and RR individuals (Fig 2A;

255
To test whether the insecticide on net fibers affected mosquito progression, we compared 257 upwind flight and localization probabilities in the presence (experiment 4) or absence (experiments 2 258 and 3) of the ITN. The probabilities were similar in presence or absence of the ITN, regardless of the 259 genotype (Table 3; Fig 2A, 2B). Nevertheless, the comparison between genotypes showed that 260 upwind flight probabilities for heterozygous RS mosquitoes remained higher than those of the two 261 other genotypes, both in the presence or absence of insecticide (Fig 2A;

Is the experimental set up symmetric?
264 Analysis of the arms' relative attractiveness data from experiment 2 (Rabbit odor; two empty 265 boxes) showed no significant differences between the number of mosquitoes collected in A1 vs. A2, 266 regardless of the genotypes (Fig 3A; RA  Moreover, CO 2 concentration and RH were not different between arms (p>0.05; S1). These results 270 demonstrated that the olfactometer was symmetrical. Moreover, analyses of RA exp3 and RA exp4 , 271 (results described below), showed no effect relative to the box receiving the treatment (i.e. variable 272 no significant in the model), indicating that symmetry was maintained during experiments 3 and 4. 273

296
The host-seeking behavior of mosquitoes towards humans sleeping under a bed net is poorly 297 understood. Particularly, it is not known whether specific volatile chemicals emanating from treated 298 nets might modulate this behavior. Here, we used a dual-choice olfactometer to assess whether the 299 presence of permethrin-treated nets may influence the host attractiveness for mosquitoes of 300 different kdr genotypes. We found that nets represent both a physical and a chemical signal that 301 modulate mosquito activation and/or choice. Moreover, the three kdr genotypes behaved differently 302 in response to host odors in the presence of ITNs or UTNs. 303 In experiment 3, mosquitoes preferably chose the arm connected to the empty box rather 305 than the box with UTNs. No difference in CO 2 quantity was noted between arms. However, the 306 humidity level was slightly higher in the arm connected with the empty box. As humidity is known to 307 attract mosquitoes [33], the observed preference for the empty box (higher humidity) was not 308 surprising. This difference could have been caused by the physical barrier formed by UTNs that may 309 absorb humidity coming from the rabbit box. In addition, the net structure could also have retained permethrin respectively) indicated that such pyrethroid might be present in the air even they are 318 semi-volatile organic compounds. More accurately, a study by Bomann et al. [42] from the Bayer 319 company measured a mean concentration of cyfluthrin (a pyrethroid with a molecular structure close 320 to the permethrin) of 0.000021 mg/m 3 at 1m away from a treated net. Such concentration 321 corresponds to 3.46x10 9 molecules/cm 3 . Angioy et al [36] found that only 6 molecules of a 322 pheromone entered in contact with the olfactory sensillum of moth species may induce a 323 physiological response. We therefore hypothesize that mosquito may detect very low concentration 324 of pyrethroid in the air. 325 Moreover, some nasal trouble (i.e runny nose) have been recorded when LNs were used for 326 the first time [37]. Such observations reinforce the hypothesis that LNs emit chemicals in the air. 327 Regardless these chemicals are insecticide itself, additive chemicals (i.e. fragrances), degradation 328 products, that composed the net, they may be detected by mosquitoes and elicit behavioral 329 modulation. 330 The behavior of insects, such as mosquitoes, is driven by the perception and integration of 331 odorant signals in antennae and higher brain center. In our study, we observed that kdr resistant 332 mosquitoes were more attracted by host odors emanating behind a permethrin-treated net than 333 host odors behind an untreated net (Fig 3C), it indicates that they perceived at distance a difference 334 between ITN and UTN and behaved differently in response. We then hypothesize that mosquitoes 335 are able to detect chemicals released by net with olfactory receptors (Ors) tuned to respond to these 336 chemicals. As an example, authors recently identified one olfactory receptor activated and another 337 inhibited by synthetic pyrethroïd in Aedes aegypti [38], suggesting that such OR may also exist in 338 Anopheles mosquitoes. The major research perspective raised by our results is therefore to study the 339 chemical and behavioral ecology relative to vector control tools already widespread in endemic 340 country. 341  [42]. Therefore, if the neuronal excitability differs in homozygous kdr 354 genotypes, the response pattern of the olfactory neurons and subsequently the signal integration 355 and processing in the central nervous system could be altered, leading to a modified motor response, 356 in this case, a difference in host choice. 357

Insecticide resistance & host seeking behavior
The present study suggests the existence of interactions between the physiological 358 mechanisms that allow mosquitoes to survive a contact with insecticide and the behavioral response 359 to olfactory cues. These interactions may have major implications in malaria control. As an example, 360 chemicals emanating from the ITNs are strongly related to the presence of human beings. Should it 361 be integrated as an attractive cue for resistant mosquitoes? This may dramatically affects the 362 personal and community protection given by the massive use of ITNs. Our study only focused on the 363 Kdr mutation, but the resistance pattern in wild Anopheles populations is far more complex [43]. It 364 would be interesting to investigate the interaction between each resistance mechanisms isolated in 365 specific strains before going to study this interaction between resistance, behavior and ITNs in semi-366 field and natural conditions. Recent papers were modeling and questioning the risk conferred by 367 resistance, based on survival to insecticide exposure [44], but the impact of such resistance on 368 behavior is also to be investigated urgently [45]. 369 We used rabbit as an odor source because mosquitoes were fed on rabbits at the laboratory, 370 and were likely "selected" to respond to rabbit's odor. But in the field, Anopheles gambiae prefers to 371 bite human when available [46]. Whether the same experiment conducted with humans as an odor 372 source will provide similar results remain to be experimentally evaluated. If we used a human instead 373 of the rabbit we change the composition of odor plume (quantity and quality of semiochemicals). 374 Therefore the interaction between chemicals released by LLIN and human odor should induce a 375 different behavior. Nevertheless, our experiment highlighted the involvement of LLIN in host seeking 376 behavior and emphasized the need to studying the relation between LLIN, host odors and mosquito 377 host seeking behavior. 378

Kdr genotypes & behavior 379
Heterozygous RS mosquitoes were more active than SS and RR mosquitoes. The addition of 380 an attractant did not change the proportion of RS leaving the RZ, suggesting that this behavior might 381 be related to a better anemotactic response (i.e response to air flow) or spontaneous flight activity 382 than a better perception of odorants in RS mosquitoes. This hypothesis is strengthened by the 383 absence of difference in the progression towards the olfactometer arms among genotypes. In other 384 words, heterozygous mosquitoes fly more, but might not smell better. On the one hand, by flying 385 more they might increase the probability of encountering a host odorant plume which might be 386 advantageous. Such heterozygous advantage for the kdr locus in An. gambiae s.s. has been recently 387 documented also for another behavioral trait: the ability to find a hole in a piece of bed net [24] and 388 for male mating [47]. In other hand, it could represent a cost for mosquitoes if energy spent during 389 flight is no more available for other traits closely related to fitness as fertility, fecundity and 390 longevity. This trade off must be deeply investigated as this might have great influence on 391 Plasmodium transmission. 392 The behavior of kdr heterozygous individuals in our study must be interpreted with caution 393 because other loci, distinct from the kdr locus, could also influence this behavioral trait. Introgression 394 and selection the kdr allele to produce the homozygous resistant strain was indeed likely to also have 395 selected linked polymorphisms [45]. 396

Conclusion 397
In conclusion, our study showed that the Anopheles mosquitoes detected the presence of 398 both physical and chemical barriers of ITNS. Face to this results, it urges to decipher with the 399 interaction between host-seeking behavior, insecticide resistance and vector control tools. The most 400 overlooked part of the puzzle is the chemical ecology in a context of large vector control measure 401 deployment. This research avenue will be challenging for the vector control community but is crucial 402 not to waste forces in wrong directions. 403

404
We thank Teun Dekker for helpful discussion. We would like to thank the IEMTV staff in Abomey 405 Calavi (Benin) for technical assistance.