Anopheles funestus mosquitoes from rural south-eastern Tanzania may have stronger resistance to pyrethroids than the other malaria vector, Anopheles arabiensis

Background: Long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS) have greatly reduced malaria transmission in sub-Saharan Africa, but are threatened by insecticide resistance. In south-eastern Tanzania, pyrethroid-resistant Anopheles funestus are now implicated in > 80% of malaria infections, even in villages where the species occurs at lower densities than the other vector species, Anopheles arabiensis. This study compared the intensities of resistance between the two malaria vectors, so as to improve options for control. Methods: The study used WHO assays with 1×, 5× and 10× insecticide doses to assess levels of resistance, followed by synergist bioassays to understand possible mechanisms of the observed resistance phenotypes. The tests involved adult mosquitoes collected from villages across two districts in south-eastern Tanzania and identied using morphological and molecular approaches. Findings: At baseline doses (1×), both species were resistant to the two pyrethroids (permethrin and deltamethrin) but susceptible to the organophosphate (pirimiphos-methyl). An. funestus, but not An. arabiensis was also resistant to the carbamate (bendiocarb) at baseline doses. Both species were generally resistant to DDT, except An.arabiensis from one village. An. funestus showed strong resistance to pyrethroids, surviving the 5× and 10× doses except in one village. Pre-exposure to the synergist, piperonyl butoxide (PBO), enhanced the potency of pyrethroid in both An. arabiensis and An. funestus achieving mortalities >98%, except for An. funestus from two villages for which permethrin-associated mortalities exceeded 90% but not 98%. Conclusions: In these communities where An. funestus dominates malaria transmission, this study may suggest that the species also have much stronger resistance to pyrethroids than its counterpart, An. arabiensis and can survive more classes of insecticides, including carbamates. The pyrethroid resistance in both species appears to be mostly metabolic and may be temporarily addressed using synergists, e.g. PBO. These ndings may

treated nets (ITNs) are being developed that contain either multiple insecticide classes [12] or pyrethroids and synergists [13,14] and are expected to improve the control of resistant mosquitoes.
Mosquito resistance involves different mechanisms, through which they can withstand exposures to insecticides. These include metabolic resistance, target-site resistance, behavioural resistance, and cuticular resistance [15][16][17]. The discriminating concentration is used to evaluate the phenotypic resistance which when detected, the level of resistance can be subsequentlly determined by using intensity assays. The intensity assays uses the 5× and 10× discriminating concentration in a stepwise manner aiming at providing the information on the range of resistance present in a target vector [18].
Mosquitoes that express metabolic forms of resistance produce large quantities of enzymes or alternate the enzyme catalytic centre to e ciently detoxify the insecticide. The speci c enzymes include monooxygenases (i.e. cytochrome P450s), which detoxify pyrethroids and carbamates, glutathione-Stransferases (GSTs), which detoxify organochlorides like DDT [17] and esterases, which detoxify pyrethroids and organophosphates [19,20]. The degree to which the enzymatic proteins are expressed, and the level of resistance can be assessed using quantitative polymerase chain reaction (qPCR). Phenotypic assays use synergists such as piperonyl butoxide (PBO), which enhance the potency of an insecticide by inhibiting the enzymes responsible for the insecticide metabolism [21]. On the other hand, some mosquitoes may have one or multiple target-site mutations due to modi cation of protein receptors usually targeted by insecticides (e.g. the voltage-gated sodium channels targeted by pyrethroids and organochlorides), thereby blocking or reducing the effectiveness of the insecticides [22][23][24]. Recently, scientists have also demonstrated that a sensory appendage protein (SAP2) enriched in the legs of malaria-carrying mosquitoes can also confer resistance to insecticides, thus allowing these mosquitoes to survive contact with ITNs [25]. Different Anopheles species have diverse levels of competencies in pathogen transmission, and also respond different to interventions based on their behavior and physiology [26,27]. With the rise of insecticide resistance in the vector populations, the choice of interventions will depend on characteristics of the local vectors. Comprehensive understanding of the distribution and underlying mechanisms of insecticide resistance is therefore important for planning and implementing vector control interventions.
In the villages of south-eastern Tanzania, where Anopheles funestus have been reported to implicate most of the ongoing malaria transmission [28][29][30], signs of resistance to most public health pesticides have been observed. An. funestus also appears to survive longer than other co-existing vector species (parity rates are higher than Anopheles arabiensis) [28]. However, since the species is highly anthropophagic (prefers to blood-feed on humans over other vertebrates) [31,32] and endophilic (prefers to bite indoors) [33], one would expect its populations and transmission activity to have been signi cantly reduced by ITNs now widely used in Tanzania for more than one decade [34][35][36][37]. Indeed, historical evidence from both East and Southern Africa suggests that effective insecticide-based indoor interventions can eliminate An. funestus on a local scale [38,39]. Anopheles gambiae s.s. which is generally considered the most competent malaria vector, shares similar behaviors with An. funestus, i.e. high degree of anthropophily and endophilly [31,32,40]. However, An. gambiae s.s. unlike An. funestus has been highly impacted by ITNs in Kenya and Tanzania [41][42][43].
An important question therefore is why and how An. funestus, despite being highly anthropophagic and endophilic, survived the ITN onslaught, and why it continues to mediate most malaria transmission in rural south-eastern Tanzania despite co-occurrence with a different malaria vector species, An. arabiensis. One hypothesis has been that An. funestus expresses higher intensities of resistance to most of the commonly used insecticides in comparison to other malaria vectors and are therefore far less impacted by insecticidal interventions. A meta-analysis of various datasets has veri ed this phenomenon at global scale [44], but speci c eld tests comparing resistance intensities in the different vector species are limited. As shown by Kaindoa et al in a study conducted in south-eastern Tanzania, An. funestus were resistant to pyrethroids, organochlorides, and carbamates [28]. Another study from the same study area demonstrated that resistance of An. arabiensis to diagnostic insecticide concentrations varied between nearby locations and seasons [45]. However, the intensity and mechanisms of these resistance phenotypes were not compared between species.
This study therefore compared the intensities and knockdown time of insecticide-resistance between the two main malaria vectors, An. funestus and An. arabiensis, in rural south-eastern Tanzania. Potential involvement of metabolic resistance and insecticide potency enhancement by Piperonyl Butoxide (PBO) synergists was also assessed

Study site
Mosquito collections were done in three different villages, namely; Ikwambi (7.98033°S, 36.81701°E) and Sululu (8.00324°S, 36.83118°E) in Kilombero district, and Tulizamoyo village (8.35747°S, 36.70664°E) in Ulanga district, south-eastern Tanzania ( Figure 1). The main malaria vectors in this area include An. arabiensis and An. funestus, with the latter driving more than 80% of the malaria transmission [28,29]. This area has had high coverage of pyrethroid-treated nets for several years, but no IRS is implemented. The villages are all in low altitude areas, rising not more than 500m above sea level. Mean daily temperatures are 20-33°C, annual rainfall, 1200-1800 mm and relative humidity ranged between 24 and 97% [46,47]. Most community members here are farmers, cultivating rice, maize and other crops in the Kilombero river valley.

Mosquito collection
The World Health Organization (WHO) protocol for insecticide susceptibility tests [18] was used with slight modi cations to conduct the basic bioassays and the resistance intensity assays. Since An. funestus mosquitoes were di cult to nd as larvae across all the study villages, young nulliparous adult females for both An. funestus and An. arabiensis were used instead of larval collections. Mosquitoes were collected from September 2018 to November 2019 using CDC light traps [48]. Collections were done from 07.00 pm to 07.00 am each night. To maximize probabilities of getting young unfed nulliparous females, houses near the edges of the villages and near potential habitats were selected for collections, based on previously-described heterogeneity of malaria transmission [49].
The traps were hung beside human occupied bed nets [50], but with extended catch bags to improve survival of the mosquitoes for subsequent assays. Each morning, after collections, the mosquitoes were transported to the Ifakara Health Institute's mosquito biology laboratory, VectorSphere, in Ifakara and maintained at 27 ± 2°C and 80 ± 10% relative humidity for 24 hours to acclimatize as previously described [45]. During the acclimatization period, mosquitoes were supplied with 10% glucose solution.
The mosquitoes were identi ed morphologically using the Gillies and Coetzee identi cation key [51], and non-target species were discarded. Tests were conducted using only non-blood-fed An. funestus and An. arabiensis females.

Bioassays
This study incolved three steps: i) susceptibility assays to examine phonotypic resistance against standard insecticide doses, ii) tests to assess intensity of resistance against chemicals to which resistance had been detected at baseline level, and iii) tests using synergists to assess possible mechanisms of the observed reistance.
Where resistance was observed in the baseline assays with standard diagnostic doses (i.e. 1×), additional tests were done to assess intensities of the resistance using 5× and 10× multiplicative doses of the insecticides. These included tests against 3.75% & 7.5% permethrin, and 0.25% & 0.5% deltamethrin. The procedures were similar to the baseline tests for assess mortality.
Lastly, 4% Piperonyl Butoxide (PBO), a synergist, was used to assess the possible resistance mechanism by attempting to reverse the observed mortality outcomes [18]. Each test had four groups, each with 80 mosquitoes (in groups of 20), treated as follows: the rst cohort was exposed to 4% PBO for one hour and immediately exposed to deltamethrin or permethrin for 60 minutes, a second group was exposed directly to the respective insecticides (i.e. deltamethrin, permethrin), a third group was exposed to the PBO only and the fourth group was exposed to control papers impregnated by silicone oil but no insecticide nor synergist. Given test kit limitations, the PBO tests were done only for pyrethroids.

Molecular identi cation of sibling species of the tested mosquitoes
Up to 10% of the mosquitoes from each bioassay were packed separately and labelled with information about experimental date, village name, type of insecticide, insecticide dose used, species of mosquito, replicate number and sample ID. The packed mosquitoes were sent to the laboratory for molecular species identi cation of sibling species in the An. funestus and An. gambiae s.l. complexes, using DNA extracted from the mosquito legs. Polymerase chain reaction assays were conducted based on speciesspeci c nucleotide sequences of the ribosomal DNA (rDNA) by relying on the intergenic spacer regions (IGS) for An. gambiae s.l. members and the non-coding internal transcribed spacer 2 region (ITS2) for An. funestus [52,53]. DNA bands were photographed under ultraviolet light using Kodak Gel Logic 100 imaging system [54].

Data analysis
The data on insecticide susceptibility was interpreted based on the WHO-speci ed thresholds for resistance determination [18]. Susceptibility was con rmed when mortality was ≥ 98%, possible resistance was determined when mortality ranged from 90% -97%, in which case the tests were repeated for con rmation, and resistance was con rmed when mortality was < 90%. When mortality greater than 10% was observed in controls, the test mortality was corrected using Abbott's formula to avoid the biased estimations [55]. Tests were discarded and repeated, whenever control mortality exceeded 20% [18]. Final results were plotted in graphs using R software version 3.0 [56]. Log-probity analysis was used to calculate mean duration at which 50% (KD 50 ) and 95% (KD 95 ) of mosquitoes exposed to speci c insecticides were knocked down.

Results
Phenotypic resistance at baseline insecticide concentrations Both species were resistant to the pyrethroids (permethrin and deltamethrin) and the organochloride (DDT), but susceptible to the organophosphate (pirimiphos-methyl) at standard baseline doses (1×). There was general susceptibility to the carbamate (bendiocarb) by both species across the study area, except in one of the villages, Tulizamoyo, where An. funestus were resistant to this insecticide. An. funestus generally showed lower mortalities to the insecticides in the baseline tests compared to An. arabiensis ( Figure 2). Figure 2: Percentage mortality of Anopheles funestus (right) and Anopheles arabiensis (left) exposed to baseline concentrations of candidate insecticides. Red-dotted and blue-dotted intercepts represent 90% and 98% mortalities indicative of resistance or susceptibility respectively. Phenotypic resistance at 5 and 10 times baseline concentrations An. funestus populations from Ikwambi and Tulizamoyo are resistant to both 5× and 10× concentrations of permethrin, but the same species from Sululu were susceptible to 10× permethrin concentrations ( Figure 3). For An. arabiensis on the other hand, resistance intensity declined with increasing insecticide concentrations. Their resistance to pyrethroids was already overcome at 5× doses in Ikwambi and Tulizamoyo villages, while the ones from Sululu village, which survived 5× doses, were overcome at 10× doses. At 10× doses, An. arabiensis from all the villages were completely susceptible to the two pyrethroids ( Figure 3). Because of the observed susceptibilities at baseline doses (Figure 1), no intensity assays were done against pirimiphos-methyl or bendiocarb on either of the species.

Knockdown times (KDT)
Knockdown time for both mosquito species varied between insecticide and study villages. Consistently, the knockdown time slowed as the concentration increased for pyrethroid-exposed mosquitoes. Prolonged KDT 95 (51-245min) was observed to An. funestus exposed to pyrethroids compared to An. arabiensis (16-76 min), similar trend was observed on KDT 50 . Furthermore, the lowest knockdown time was observed when the mosquitoes were exposed to bendiocarb both at KDT 50 and KDT 95 for An. funestus and An. arabiensis (Table 1). Effects of pre-exposure to the synergist, PBO Pre-exposure to the synergist, PBO, signi cantly reversed the pyrethroid resistance in both An. arabiensis and An. funestus. The PBO assays achieved mortalities > 98% in most cases, except for An. funestus populations from Sululu and Tulizamoyo villages, for which permethrin-associated mortalities were reversed past 95% but not 98%. The synergist assays on An. arabiensis from all study areas demonstrated highest restoration of susceptibility ( Figure 4). Figure 4: Proportion mortality of Anopheles funestus and Anopheles arabiensis to pyrethroids when preexposed to synergist. Red-dotted and blue-dotted intercepts represent 90% and 98% mortalities respectively

Molecular identi cation of species
After the bioassays, a total of 305 An. funestus and 144 An. arabiensis were sent to the laboratory for sibling species identi cation. Of all the An. funestus assessed, successful PCR ampli cation was 76% (n = 233). Of those that ampli ed, 99% were An. funestus s.s. (n= 232), while one was ampli ed as An. leesoni. For An. gambiae s.l. successful ampli cation was 92% (n = 132), all of which were identi ed as An. arabiensis. The rest did not amplify in the PCR assays (n=12).

Discussion
In this study, both An. arabiensis and An. funestus were resistant to pyrethroids and DDT. However An. funestus exhibited far lower mortalities when subjected to pyrethroids at either the baseline concentration, ve times concentration or the ten times concentration in the intensity bioassays. This suggests that while An. funestus is strongly resistant to the pyrethroids, the level of resistance in An. arabiensis was either low or moderate. This is the rst study to directly compare resistance intensities of these two vectors in the area, and therefore provides important information on potential performance of current or future interventions against malaria. Given the differential contribution of the two vectors to overall transmission, their responsiveness to insecticidal interventions is an important factor for consideration in the elimination efforts.
Initial ndings from standard WHO susceptibility assays by Kaindoa et al [28] and Matowo et al [45] on the two malaria vectors in the same study area observed that the baseline mortalities were higher in An. arabiensis than An. funestus. This was the initial indication that the intensity of resistance would be different between the two species, and necessitated additional tests according to standard WHO assays [18]. The duration at which either 50% or 95% of mosquitoes would be knocked-down varied between species, insecticides and study villages (Table 1). Knock-down time was high when mosquitoes were exposed to standard concentrations of pyrethroids, but slowed as the concentrations increased. Since no kdr mutation tests were done, it is not possible to determine whether these observations were associated with the voltage-gated sodium channel protein mutation [57]. The slowest knockdown time was observed when the mosquito species were exposed to bendiocarb, these ndings still support the potency of the insecticide ( Table 1). The new ndings clearly demonstrate that An. funestus populations, despite being the more dominant vector of malaria in the area, would be much more di cult to control using current pyrethroid-based interventions, in particular the LLINs.
As demonstrated by Matowo et al for both An. arabiensis and Culex mosquitoes [45,58], there were signs of ne-scale spatial variations in insecticide resistance. For example, An. funestus populations from Tulizamoyo were resistant to bendiocarb but populations of the same species from the other two villages were susceptible to the same chemical (Figure 2). Similarly, the mortality percentages observed at 5× and 10× doses varied between the villages (Figures 2 & 3). This might be attributed to differences in the use of agricultural pesticides for crop protection in these villages [59]. Surprisingly, An. arabiensis from Ikwambi were 100% susceptible to DDT, against which both species from the other study villages were resistant (Figure 2), which further suggests ne-scale spatial differences in resistance pro les.
The currently observed dominance of An. funestus is likely to be contributed by their well-documented resistance to commonly used insecticides [28,[60][61][62][63][64], their high survival probabilities in the wild [28,30] and high levels of anthropophily [28,31,33]. Its dominance in areas where insecticidal interventions such as ITNs are widely implemented is particularly surprising given that scale-up of ITNs has coincided with signi cant declines in populations of other anthropophilic vectors such as An. gambiae s.s. [41][42][43]. Today, in rural south-eastern Tanzania, An. funestus co-exists with other Anopheles species, namely An. arabiensis, Anopheles coustani, Anopheles squamosus, Anopheles leesoni, Anopheles rivulorum and Anopheles pharoensis [28]. However, it is known to carry most of the ongoing malaria transmission, sometime implicating to nearly nine in every ten new cases, even in areas where it occurs in lower densities than An. arabiensis [28,29]. It was thus hypothesized that its dominance may at least be partly driven by stronger insecticide resistance levels to insecticides commonly used for public health, notably the pyrethroids used on bed nets.
As insecticide resistance increases across Africa, some populations have been observed to withstand up to 1000 times the standard concentrations [65], making it an urgent need to nd new classes or combinations of insecticides [3][4][5]17]. In areas where An. funestus is dominant, such as in south-eastern Tanzania, the decisions on which insecticides to be implemented in vector control measures should re ect intensity of resistance in this species, even if it is di cult to nd its larvae. An. funestus were resistant up to ten times the WHO-recommended concentration of pyrethroids, clearly indicating that this class of insecticides can no longer be useful in the area and must be urgently replaced by other classes such as organophosphates, against which resistance is not yet detected.
The synergist tests in this study showed complete or almost complete restoration of susceptibility in the malaria vector mosquitoes nearly from all study areas. This full restoration is a likely indicator of metabolic resistance [58,66] and suggests that ITNs which have both PBO and pyrethroids such as PermaNet 3.0 [14] and Olyset Plus [13] may be suitable for malaria prevention in these areas, and could potentially provide better protection than standard LLINs [67]. Synergist pre-exposure combined with deltamethrin had a greater restoration in An. funestus than when the synergist was combined with permethrin ( Figure 4), but in both cases there was still substantial restoration. This could be likely due to different resistance levels against the two pyrethroid classes as observed by Rakotoson et al [68] when An. arabiensis were pre-exposed to PBO. Partial restoration of susceptibility observed in An. funestus mosquitoes might be a sign of multiple metabolic resistance forms or other resistance mechanisms including the target-site mutation [69]. This could also be a manifestation of the demonstrated high intensities of pyrethroid resistance (Figure 3). These ndings are in line with the previous studies on the resistance of malaria vectors to pyrethroids and organochlorides and incomplete susceptibility restoration after the synergist pre-exposure to pyrethroids [68]. Nonetheless, further exploration is needed to identify the speci c metabolic enzymes responsible for the observed resistance under biochemical tests. Additionally, the level of these resistant enzymes needs to be assessed using quantitative PCR assays in both An. arabiensis and An. funestus.
Despite largely achieving the stated aims, the ndings of this study should be considered only as indicative and not in any way conclusive. This is due to the various methodological limitations faced during the study. First, the overall collection of the specimen was distributed over several months, and may bave been in uenced by seasonal variations in resistance, as previoauly demonstrated [45]. Besides, given the scarcity of An. funestus specimen in some of the villages, the tests used just 120 mosquitoes. Third, this study was the use of wild mosquitoes which may have varying ages, which is an important factor long-demonstrated to impact resistance [70][71][72]. While this way of testing gives a true representation of the natural mosquito population in communities and their ability to withstand insecticidal interventions, it makes it di cult to compare the tests in a conclusive manner. The WHO guidelines recommend the use of age-synchronized F 1 generation, 3-5 days old [18]. In this study the challenge was minimized by: a) collecting the adult female mosquitoes at the edges of the village near potential aquatic habitats, thus maximizing the chances of getting young nulliparous mosquitoes [49], b) adding an acclimatization period of mosquitoes for 24 hours between the actual mosquito collection and the resistance tests, and c) using the CDC light trap for mosquito collection, thereby capitalizing collection of nulliparous host-seeking mosquitoes [73][74][75]. In addition, the tests did not combine collections from multiple days, but instead used synchronized days for each replicate, thus ensuring that the mosquito ages were approximately similar. It is recorgnized however that these improvements slightly improved the tests but are not adequate to enable conclusive determination or comparison of resistance levels.
Another limitation was the non-ampli cation of the samples where 8% (n=12) of An.arabiensis and 24% (n=62) of An. funestus complex were unidenti able. It is possible that either there were polymorphisms in the ITS2 region of rDNA ampli ed in these assays, which might have been the main contributor of the observed non-ampli cation (Mapua et al unpublished data), or there were a few other sibling species for which no primers were available in the assay. Future studies should involve a larger sample size, and possibly individual analysis of specimen to distinguish between species.

Conclusion
Overall, this study has demonstrated that other than the differential importance of malaria vector species and the multiplicity of malaria transmission in different settings, the responsiveness of these vectors towards different insecticides may also vary. In rural south-eastern Tanzania, An. funestus, which now dominates malaria transmission, also indicate strong resistant to pyrethroids commonly used on ITNs than its counterpart, An. arabiensis. Despite its rarity at aquatic stage, collection methods must endeavour to nd this vector with synchronized age and study its resistance pro le so that effective interventions can be mounted. Lastly, the study also emphasizes that decisions on which insecticidal interventions to apply should be informed by geographical and species-speci c studies rather than generalized studies. In this cases, it appears that PBO-based LLINs and IRS with non pyrethroids such as organophosphates may be appropriate for now, as the main vectors are still susceptible to these treatments. Authors' contributions PGP, CE and FOO objective conceptualization, designed and conducted experiments, analysed the data and drafted the manuscript; DM and HSN helped in data analysis, SA helped in molecular mosquito species identi cation edited and revised the manuscript, JK and HB helped with the eld mosquito collection and morphological identi cation. All authors read and approved the nal version of the manuscript.

Figure 1
Locations of the study sites in Kilombero and Ulanga districts, where mosquito collections were performed.

Figure 2
Percentage mortality of Anopheles funestus (right) and Anopheles arabiensis (left) exposed to baseline concentrations of candidate insectiicides. Red-dotted and blue-dotted intercepts represent 90% and 98% mortalities indicative of resistance or susceptibility respectively.

Figure 4
Proportion mortality of Anopheles funestus and Anopheles arabiensis to pyrethroids when pre-exposed to synergist. Red-dotted and blue-dotted intercepts represent 90% and 98% mortalities respectively