Behavioral Responses of Pyrethroid-resistant Anopheles Gambiae Mosquitoes to Insecticide-treated Bed Net

Background: Long-lasting insecticidal nets are an effective tool in reducing malaria transmission. However, with increasing insecticide resistance little is known about how physiologically resistant malaria vectors behave around a human-occupied bed net, despite their importance in malaria transmission. We assessed the host-seeking behavior of the major malaria vector Anopheles gambiae s.s, when an intact human-occupied treated bed net is in place, with respect to their insecticide resistance status under semi-eld conditions. Methods: Pyrethroid resistant and susceptible colonies of female Anopheles gambiae s.s aged 3-5 days that have been bred in our insectary, were color-marked with uorescent powder and released inside a semi-eld environment housing a hut which was occupied by a human host. Inside the hut, the occupant slept under an insecticide-treated bed net trap or untreated bed net trap. The window exit trap was installed to catch mosquitoes exiting the hut. A prokopack aspirator was used to collect indoor and outdoor resting mosquitoes in the morning. Clay pots were placed outside the hut to collect mosquitoes resting outdoors. Results: The proportion of resistant mosquitoes caught in the treated bed net trap was higher 43% (95% CI= [40.6-45.3]) compared to the susceptible mosquitoes 28.3% (OR=1.445; P<0.00019). The proportion of susceptible mosquitoes caught in the untreated bed net trap was higher 51.3% (95% CI= [48.8-53.6]) compared to the treated bed net trap 28.3% (95% CI= [26.3-30.5]) (OR=2.65; P<0.0001). Resistant mosquitoes were less likely to exit the house when a treated bed net was present (5.2%; 95% CI= [4.2-6.4]) compared to the susceptible mosquitoes (11.5%; 95% CI= [9.6-12.6]). The proportion of susceptible mosquitoes avoiding contact with the treated bed net and caught resting indoors in the hut (53.8%) and outdoors (64.5%) was higher compared to the resistant mosquitoes (indoors: 46.2%, outdoor: 35.4%). The susceptible females were 2.3 times more likely to stay outdoors away from the treated bed net (OR=2.25; 95% CI= [1.7-2.9]; P<0.0001). Conclusion: The results show that in the presence of a treated net, the host-seeking performance was not altered for the resistant mosquitoes, unlike the susceptible females that were observed to exit the house and remained outdoors when a treated net was used. However, further investigations of the behavior of resistant mosquitoes under natural conditions should be undertaken to conrm these observations and improve the current intervention which are threatened by insecticide resistance and altered vector behavior.


Introduction
Reduction in malaria morbidity and mortality over the past few years in sub-Saharan Africa is largely attributed to the effectiveness of long-lasting insecticidal nets (LLINs) 1 .This has been possible because the main malaria vectors primarily feed indoors at night, a behavioral pattern that coincides with the period when human hosts are indoors and asleep [2][3][4] . However, extensive use of insecticides has subjected mosquitoes to intensive selection pressure, resulting in the development of physiological and behavioral resistance, threatening the future of existing tools, hence the need for continuous monitoring of their e cacy and development of novel LLINs 5 .
The continued success of the current vector control interventions is dependent on the susceptibility of the target mosquito population to the insecticides used. Pyrethroids are one of the insecticide classes recommended for treating mosquito nets owing to their low mammalian toxicity, unique modes of action such as fast knockdown, and high insecticidal potency, although currently assessment of some innovative nets treated with a combination of a pyrethroid and either a non-pyrethroid compound is ongoing 6,7 . Over the past two decades, the use of insecticide-treated nets has increased, exerting greater selection pressure on the malaria vector populations resulting in higher incidences of pyrethroid insecticide resistance that is likely to affect the effectiveness of vector control 8 . Some studies have reported the spread of pyrethroid resistance and the mechanisms involved including target site insensitivity caused by kdr mutations 5,9 and detoxi cation enzymes that metabolize the insecticide before reaching its target site 10 . However, it is less clear how the observed resistance affects current control measures. Recent eld observations in Africa suggest changes such as increased outdoor hostseeking within the principal malaria vectors, An. gambiae and An. funestus species complexes 3,11−13 . These behavioral shifts have been largely linked to the increased use of vector control measures selecting for vector species with more exophilic behavior 14,15 .
Existing literature on behavioral changes comes mainly from pyrethroid susceptible mosquitoes but the data on the behavior of pyrethroid-resistant malaria vectors is sparse and, at times con icting, highlighting the need for additional research. It is suggested that avoidance behavior in mosquitoes that have become insensitive to pyrethroids may weaken due to increased selection pressure exerted by the insecticides used 16 . Some authors assert that physiologically resistant mosquitoes may use the recognition of insecticides as a proxy for host presence [17][18][19] . It is unclear if mechanisms related to insecticide resistance may in uence the behavior of malaria vectors, as any molecular change in the insect nervous system, may have a pleiotropic effect on nerve function and insect behavior 20 .
Given the important role of the current vector control interventions as a means of reducing the burden of malaria transmission and increasing insecticide resistance, the behavior of physiologically resistant malaria vectors should be well de ned. The host-seeking behavior of the major malaria vector, Anopheles gambiae s.s. (hereafter An.gambiae) was assessed when an intact human-occupied insecticide-treated and untreated bed net is in place, with respect to their insecticide resistance status under semi-eld conditions. We hypothesize that pyrethroid-resistant mosquitoes will seek and bite human hosts indoors despite the presence of indoor-based vector control interventions while the susceptible mosquitoes will leave the house through the windows or eaves and seek blood meal elsewhere. This study provides more information on how the behavior of physiologically resistant vectors might differ in comparison to their susceptible counterparts, an aspect that is poorly understood. There is an urgent need for evidence-based studies on the behavior of malaria vectors in the presence of vector control interventions, given the rapid development of insecticide resistance in a large number of malaria vectors, if the signi cant gains made in reducing malaria morbidity and mortality is to be maintained.

Mosquitoes strain used in the experiments
Mosquito strains used in this study consisted of a deltamethrin selected resistant colony and an unselected susceptible colony (hereafter referred to as resistant and susceptible mosquitoes) that were collected from Bungoma in western Kenya. These colonies were selected and maintained at the Centre for Global Health Research, Kenya Medical Research Institute (KEMRI) in Kisumu 21 , Western Kenya, under standard rearing conditions of 27 ± 2 °C and relative humidity (RH) of 80 ± 10% °C under a L12: D12 h light: dark cycle. During the process, each colonized strain had three independent lineages that started with 200-250 females at every new generation to limit bottleneck effects. The progeny of F1 wildcaught mosquitoes from the same site were also used to undertake these experiments.

Resistant strain:
This colony underwent deltamethrin selection after each generation. The 6 th generation used was highly resistant with 20% mortality according to the WHO criteria 22 . Resistance in this colony was mainly mediated by cytochrome P450 detoxi cation enzyme. The two kdr mutations 1014S and 1014F were present and at high frequencies 21 .

Susceptible strain:
This strain shares the same genetic background with the resistant colony, however, it was reared in the absence of insecticide selection pressure. After 9 generations without selection pressure, the population had almost lost resistance to deltamethrin (92%) and after 13 generations the population showed increased mortality (97.3%). The vgsc1014S was at a high frequency given that the allele was already xed in the parent population 21 . The 14 th generation was used in this study.
Wild population: F1 progeny obtained from wild-caught An. gambiae female mosquitoes from the same area where the resistant and susceptible colonies were selected were used for validation. Each female(mother) was identi ed by PCR as An. gambiae s.s according to the methods of Scott, et al. 23 . The wild population had 56% resistance to deltamethrin. The observed resistance is mediated by a mix of metabolic and kdr 9,24-27 .

Semi-eld set up
The study was carried out in Western Kenya at the Centre for Global Health Research, Kenya Medical Research Institute, Kisumu. The release and recapture studies were conducted in a MalariaSphere a closed system 20m long x 8m wide 28 with slanted roo ng (3m in the sides and 4.5m in the middle). The entire structure is covered with an insect-proof netting screen that prevents mosquitoes inside the system from escaping into the environment, or vice versa ( Fig. 1 A). The system is also double-doored for the same reason. Inside the systema3m x 3m mud-walled hut is erected resembling a typical house in the study village in terms of size, structure and mosquito exit/entry points (eaves, window,and door) (Fig. 1B). The structure has local vegetation and grass oor to mimic the natural vegetation and provide shelter for mosquitoes in the outdoor environment (Fig. 1B). Two round clay pots are installed in the enclosure but outside the hut to act as outdoor resting sites ( Fig 1C). Inside the hut either a treated LLIN inside a bed-net trap (Mbita trap)or an untreated net inside a bed-net trap as control was hanged ( Fig 1D). Treated and untreated nets were used at different nights in the same hut. For each night, a consented and remunerated human volunteer slept under the bed-net trap in the hut. To offset any personal bias due to differential sleeping habits or relative attractiveness to mosquitoes, two sleepers were recruited for this experiment and took turns to sleep under the bed net. They were instructed not to consume alcohol or smoke and avoid deodorants during the study period. The volunteer who slept under the bed net served as a bait to attract the mosquitoes into the hut but was not bitten because of the net shield.

Bed net trap for the collection of host-seeking mosquitoes
The Mbita trap which is a bed-net trap described by Mathenge, et al. 29 . This is a modi ed conical bednet made of light white cotton cloth instead of netting (Fig.2). The trap has two chambers; the upper trap chamber and the lower bait chamber. The upper chamber contained a netting panel xed halfway ( Fig.  2A) to prevent mosquitoes from reaching the human bait sleeping in the lower chamber (Fig. 2B). For this experiment, the netting panels were either treated or untreated. The treated netting panels were cut from DawaPlus 2.0 a long-lasting insecticidal net (LLIN) containing 80 mg/m 2 deltamethrin. The nets were selected for this experiment based on the fact that they were distributed in the largest proportion in the study site by the National Malaria Control Programme in Kenya during the 2017 mass net campaign. The untreated bed net was obtained from the local market in Kisumu, Kenya.

Mosquito release and recapture
Batches of 200 uninfected and unfed female mosquitoes aged 3-5 days from the resistant or susceptible colonies were gently mouth-aspirated into a clean paper cup. The mosquitoes were sugar-starved for 6 hours before releases and color-marked with green (for the susceptible colony) and pink (for resistant colony) uorescent powder to distinguish them after simultaneous release in the semi-eld environment. Mosquito releases were done outside the hut and at the same time of day (1840hrs) to avoid circadian cycle effects. The volunteer entered the bed net 30 mins after the release of the mosquitoes. Fifteen (15) tests were conducted with each net (treated or untreated net). The release was done every 3 days to allow for the wash-out period. Windows of huts were tted with exit traps to catch exiting mosquitoes (Fig. 1E). The oor of the hut was covered by white sheets to ease the collection of knocked-down mosquitoes.
Host-seeking mosquitoes trapped in the bed net trap were collected and recorded (Fig. 1F). Validation of these behaviors was done using the F1 progeny obtained from An. gambiae s.s caught from the eld.
Indoor and outdoor resting mosquito recapture Mosquitoes that were not caught in the bed net trap or window exit trap were collected from inside the hut and outside at 0700HRS using Prokopack aspirators (John W Hock, Gainesville, FL, USA). For mosquitoes resting indoors, walls and ceilings were systematically aspirated using progressive down and upward movements along its entire length. For outdoor resting mosquitoes, the collection was done from the clay pots ( Fig. 1 F) by placing a white mesh from a mosquito cage over the mouth and agitating the mosquitoes inside the pot, causing them to y and move to the cage 30 . The corners of the screen house and the vegetation cover were also scanned for resting mosquitoes using the Prokopack aspirator.
WHO cone bio-assay to determine bed-net e cacy The treated net insecticidal e cacy was con rmed by exposing mosquitoes for 3 mins according to the standard WHO cone bioassay procedure. This was done with 4-5day old, non-blood fed, An. gambiae s.s. The bioassays included 5 replicates from both the resistant and susceptible colony with an average of ve mosquitoes per tube. The cone bioassays were conducted using DawaPlus 2.0 long-lasting insecticidal net treated with deltamethrin. The F1 progeny of wild-caught mosquitoes were also used for this experiment for validation. After exposure, the groups of mosquitoes were placed in a single 1 L paper cup and provided with cotton wool soaked with 10% sugar solution for 24 hrs. Their knock-down status was measured 60 min post-exposure and mortality were recorded after 24 h. An untreated net was used as a negative control for the assay.

Scienti c and Ethical clearance
This study was approved by the Ethical Review Board of the Kenya Medical Research Institute (KEMRI) under the scienti c steering committee (SSC 3434). Prior to commencement of the study, volunteers were given an information sheet describing the aims, study procedures, risks and bene ts of their participation in this study. Written informed consent was obtained from individual volunteers before the experiments.
The experiments were performed in accordance with the institution guidelines and regulations.

Statistical analysis
The proportions of mosquitoes caught in the bed net trap were interpreted as host-seeking mosquitoes. The proportions were calculated by dividing the number of mosquitoes caught in the bed net trap/exit trap/resting with the total number of mosquitoes recaptured for each phenotype (resistant and susceptible mosquitoes) respectively. Observations of host-seeking behavior and exit behavior of resistant and susceptible phenotypes were compared between treatments using generalized linear model (GLM) with binomial distribution and logit link function. The LLINs were considered bio-effective when the percentage of mosquitoes knocked down after 60 min post-exposure was above 95% or mortality after 24 h was above 80% in the WHO cone bioassays 31 . Statistical analysis was done using the statistical program Stata (Version 14, StataCorp, College Station, Texas). times more likely to search for a host than when a treated bed net was present (OR = 2.65; 95% CI=[2.29-3.05]; P < 0.0001, Fig. 3). Overall, there was a signi cant effect of treatment type on mosquito hostseeking behavior, with more mosquitoes caught in the bed net trap when untreated net was present than when there was a treated bed net trap (OR = 1.908; 95% CI=[1.73-2.12]; P < 0.0001, Fig. 3).

Responses of mosquitoes to insecticide-treated and untreated bed-net trap
For the wild population, a total of 1013 (50.6%) mosquitoes were recaptured out of 2000 F1 females released. The proportion of the wild population caught in the untreated bed net trap was slightly higher 40% compared to treated bed net trap 33.5% (Fig. 3). Although, this was not statistically signi cant (OR = 0.773; 95% CI=[0.598-0.999]; P = 0.489). The mortality of the resistant population trapped in the treated bed net trap was 77.7 % (549/706) and 85. 2% (144/169) for the wild population. All the susceptible mosquitoes trapped in the insecticide-treated bed net were dead.

Insecticide induced exophily of resistant and susceptible populations
The proportion of mosquitoes that were caught in the window exit trap when treated bed net trap was present was 5.2% (95% CI= [4.2-6.4]) for the resistant colony and 11.5% (95% CI= [9.6-12.6]) for the susceptible colony. When the untreated bed net was present, the number of mosquitoes exiting reduced for both groups; 2.4% (95% CI= [1.8-3.3]) of resistant population exited and 3% (95% CI= [1.8-3.5]) for the susceptible population (Table 1). Overall, the resistant population was less likely to exit the house if any of the treatment was present compared to the susceptible population (GLM, OR = 0.54; 95% CI=[0.432-0.674]; P < 0.0001). The susceptible females were 4.6-fold more likely to exit the house when treated bed net (11.5%) was present than when untreated bed net (3%) was used (GLM, OR = 4.64; 95% CI=[3.3-6.5]; P < 0.0001). For the wild eld population, 4% (95% CI= [2.2-5.6]) of the recovered mosquitoes were caught in the exit trap when the treated bed net trap was present, while 3.4% (95% CI=[1.9-5.1]) when the untreated net was used. Even though the eld population was likely to exit the house when a treated bed net was present, this was not statistically signi cant (GLM, OR = 1.12; 95% CI=[0.58-2.15]; P = 0.719). Indoor and outdoor resting behavior in response to insecticide-treated and untreated bed net. net was used, was higher 64.5% compared to resistant females 35.4%. The susceptible population was 2.3 times more likely to stay outdoors away from the treated bed net (OR = 2.25; 95% CI= [1.7-2.9]; P < 0.0001; Fig. 4).
For wild population, the proportion of females caught resting indoor or outdoor when a treated bed net was present was 52.03% vs 52.2% respectively compared to the untreated bed net trap (47.9% vs 47.8% respectively, Fig. 4). Even though the proportion resting indoor or outdoor was high when the treated bed net was present, there was no signi cant difference (Indoor: OR = 1. LLIN bioassay and knockdown rates against resistant and susceptible colonies Prior to the semi-eld trials, the e cacy of the treated bed net was evaluated. The knockdown response of the resistant females exposed to DawaPlus2.0 for 60 minutes was 7% whilst 83% for the susceptible population. The mortality rate for the resistant colony was 13% (95% CI=[9.

Discussion
Physiological resistance in mosquito populations to common public health insecticides across Africa is widely reported but evidence of the actual impact of this, the functionality and e cacy of LLINs is scarcely discussed or documented 5 . Monitoring the host-seeking behavior of physiologically resistant mosquitoes in the presence of indoor vector control tools is necessary to determine whether the e cacy of the tools could be compromised with the resistance pro les or whether they can be optimized. This study provides insights into the behavior of pyrethroid-resistant An. gambiae when they encounter pyrethroid-based LLIN in a free-ight environment similar to the eld settings. The results demonstrate that in the presence of a treated net, the host-seeking performance was not altered for resistant females, unlike the susceptible females that were observed to exit the house and remained outdoors when a treated net was used.
One of the consequences of the massive roll-out of LLINs is the change in mosquito behavior where the interventions may select vectors with increased exophily (feeding outdoors early in the evening or morning hours when LLINs are not in use) because of the exposure to insecticides 12 . This study observed a large proportion of host-seeking susceptible females exiting the house and resting outdoors than resistant females when the treated net was present. The observed behavior suggests that in the presence of insecticides, susceptible mosquitoes may be pushed from indoor treated environments and resort to search blood meal outdoors or rest outdoors and initiate their search for a host soon after dusk, leading to increased outdoor transmission. Examples of spatial avoidance of insecticide-treated environment has been observed in malaria vectors in the eld, displaying increased outdoor host-seeking and resting outdoors following the implementation of IRS and ITNs 32,33 . This indicates that a substantial part of residual malaria transmission is occurring outdoors, raising the questions on the effectiveness of LLINs in reducing malaria infections when susceptible indoor feeding mosquitoes are diverted to feed outdoors when people are outside LLINs. On the other hand, the ndings suggest, physiologically resistant malaria vectors that have developed the capacity of blood-feeding or resting indoors in the presence of LLINs, may compromise the effectiveness of LLINs, maintaining the indoor malaria transmission.
The strategy of LLINs in malaria prevention is to deter mosquitoes from entering houses and reduction in blood-feeding rates, achieved as a consequence of excito-repellent effects of the pyrethroids 34 .In this study, a higher proportion of the resistant females were caught in the treated bed net trap compared to the susceptible population. The resistant females were less likely to avoid the search for a host when a treated net was present, unlike the susceptible population that was observed to avoid contact with the treated net. One plausible explanation for the difference in behaviors is the pleiotropic effects on nerve function associated with a point mutation in the voltage-gated sodium channels of resistant mosquitoes, as it interferes with the sensitivity of the sensory nervous system to pyrethroids resulting in reduced avoidance behavior 35,36 . This implies that in the eld, physiologically resistant mosquitoes are likely to spend more time in search of a host in the presence of insecticides increasing their probability of encountering a host unlike their susceptible counterparts. In nature, pyrethroid-resistant mosquitoes have been found resting inside holed LLINs 37 . Such behavior may compromise the e cacy of the current indoor-based vector control tools resulting in increases in malaria transmission indoors 38 . Recent studies from western Kenya observed high resistance levels, rates of human blood index and sporozoite rates in the mosquitoes resting indoors compared to the mosquitoes collected resting outdoors 26,39 . The study ndings are in agreement with similar studies that have observed reduced host-seeking performance of susceptible mosquitoes in the presence of LLIN unlike the resistant mosquitoes whose behavior was not altered 18,40−42 .
The ndings of this study show that despite the coverage of the indoor interventions, it is evident that not all malaria transmission can be controlled with the existing tools that are indoor-based. The population of vectors that move outdoors are not taken care of, a situation that creates a pressing need for supplementary vector control tools to control residual transmission.

Conclusion
The results show signi cantly reduced avoidance behavior of the resistant mosquitoes compared with the susceptible population. The susceptible females were more likely to exit and rest outdoors away from the treated environment. This might be a reason for increased outdoor malaria transmission in sub-Saharan Africa. This situation calls for urgent deployment of control tools that can complement the current vector control methods to tackle outdoor malaria transmission.      Percentage of mosquitoes resting indoors and outdoors when a treated and untreated bed net trap was present. Error bars indicate 95% con dence intervals.