Interventions that effectively target Anopheles funestus mosquitoes could significantly improve control of persistent malaria transmission in south–eastern Tanzania

Malaria is transmitted by many Anopheles species whose proportionate contributions vary across settings. We re-assessed the roles of Anopheles arabiensis and Anopheles funestus, and examined potential benefits of species-specific interventions in an area in south-eastern Tanzania, where malaria transmission persists, four years after mass distribution of long-lasting insecticide-treated nets (LLINs). Monthly mosquito sampling was done in randomly selected households in three villages using CDC light traps and back-pack aspirators, between January-2015 and January-2016, four years after the last mass distribution of LLINs in 2011. Multiplex polymerase chain reaction (PCR) was used to identify members of An. funestus and Anopheles gambiae complexes. Enzyme-linked immunosorbent assay (ELISA) was used to detect Plasmodium sporozoites in mosquito salivary glands, and to identify sources of mosquito blood meals. WHO susceptibility assays were done on wild caught female An. funestus s.l, and physiological ages approximated by examining mosquito ovaries for parity. A total of 20,135 An. arabiensis and 4,759 An. funestus were collected. The An. funestus group consisted of 76.6% An. funestus s.s, 2.9% An. rivulorum, 7.1% An. leesoni, and 13.4% unamplified samples. Of all mosquitoes positive for Plasmodium, 82.6% were An. funestus s.s, 14.0% were An. arabiensis and 3.4% were An. rivulorum. An. funestus and An. arabiensis contributed 86.21% and 13.79% respectively, of annual entomological inoculation rate (EIR). An. arabiensis fed on humans (73.4%), cattle (22.0%), dogs (3.1%) and chicken (1.5%), but An. funestus fed exclusively on humans. The An. funestus populations were 100% susceptible to organophosphates, pirimiphos methyl and malathion, but resistant to permethrin (10.5% mortality), deltamethrin (18.7%), lambda-cyhalothrin (18.7%) and DDT (26.2%), and had reduced susceptibility to bendiocarb (95%) and propoxur (90.1%). Parity rate was higher in An. funestus (65.8%) than An. arabiensis (44.1%). Though An. arabiensis is still the most abundant vector species here, the remaining malaria transmission is predominantly mediated by An. funestus, possibly due to high insecticide resistance and high survival probabilities. Interventions that effectively target An. funestus mosquitoes could therefore significantly improve control of persistent malaria transmission in south–eastern Tanzania.

remaining malaria transmission is predominantly mediated by An. funestus, possibly due to high insecticide resistance and high survival probabilities. Interventions that effectively target An. funestus mosquitoes could therefore significantly improve control of persistent malaria transmission in south-eastern Tanzania.

Background
Malaria continues to be one of the most significant mosquito-borne parasitic diseases, affecting about 212 million people, causing 429,000 deaths annually [1], and adversely affect socioeconomic development in sub-Saharan African countries [2]. The World Health Organization (WHO) estimates that there has been a decline in malaria burden, and that morbidity worldwide reduced by 21% and mortality by 29% between 2010 and 2015, but sub-Saharan Africa still accounts for approximately 92% of all malaria deaths and cases [1] Tanzania has experienced a decline in malaria transmission following the introduction of insecticide treated nets (ITNs) and the scale-up of long-lasting insecticide treated nets (LLINs) [3]. By 2010, the country had made significant progress, and most areas that were experiencing prevalence above 50% in 2000, now had below 10% prevalence [4]. However, the most recent malaria indicator survey has returned an average prevalence of 14.8% prevalence in children under 5 years in Tanzania [5]. Getting to zero transmission therefore remains a major challenge, mainly because the persistent malaria transmission is mediated by mosquitoes that are not adequately responsive to control by existing indoor insecticidal interventions, such as LLINs and indoor residual spraying (IRS). The Tanzanian National Malaria Control Program is aiming at cutting down malaria transmission to 1% by 2020 [6], thus the need for new complementary approaches is even more urgent.
Despite the high coverage of LLINs, high indoor densities of potentially infectious mosquitoes remain common in some places [7,8]. Evidence suggests that indoor densities of major malaria vectors such as An. funestus and An. arabiensis depend on house design features, environmental conditions and demographic composition [8,9]. Previous studies have reported high levels of resistance in An. funestus to pyrethroids in Tanzania [10] and elsewhere in East Africa [11][12][13]. An. funestus is one of the major malaria vectors in Africa and is widely distributed across the continent [14,15]. The species is highly anthropophilic and endophilic, with higher vectorial capacity than most other vector species [14,16]. The An. funestus group includes several sibling species, which are difficult to distinguish morphologically, but can be separated by polymerase chain reaction (PCR) using a cocktail of species-specific primers [17]. Previous studies in Tanzania have reported presence of four member species in the An. funestus group, namely An. funestus s.s, An. rivulorum, An. leesoni and An. parensis [18]. All the four members of the An. funestus group have been found positive for P. falciparum though the infection rates are often higher in the predominantly anthropophagic An. funestus s.s than the rest. The importance of this vector is increasingly associated with the persistent malaria transmission [8,10], and it has been suggested that it could be responsible for resurgence of malaria transmission in rural Tanzania [10].
In this study, we assessed contributions of residual populations of major malaria vectors, An. arabiensis and An. funestus, in endemic villages in rural south-eastern Tanzania, where the last mass distribution of LLINs prior to our study had been completed in 2011 [19]. The study assessed prevalence of Plasmodium sporozoites in the vector species, survival rates of these vectors, and their resistance status to examine what factors might be mediating the role of these two major species in the ongoing malaria transmission. Inferences are made with regard to what would happen if appropriate species-specific interventions were deployed in this and similar settings. More specifically, interventions that effectively target mosquito species mediating most of the remnant malaria transmission should result in extensive declines in the transmission. Where one vector species dominates the transmission system, approaches against such vectors may offer opportunities to sufficiently disrupt local malaria transmission.

Study area
The study was conducted in three villages in the Kilombero floodplains in Ulanga district, south-eastern Tanzania (Fig 1). This area is perennially meso-endemic for malaria and has high mosquito densities throughout the year, peaking between January and May. Annual rainfall and temperatures range from 1200 to 1800 mm and 20˚C to 32.6˚C respectively. Malaria vector species in the area comprise primarily of the An. gambiae s.l (almost exclusively consisting of An. arabiensis), and An. funestus group. Several other Anopheles mosquitoes such as An. coustani, An. pharoensis, An. squamosus, An. ziemanni and An. wellcomei are also found, as well as several culicine species, mainly Mansonia mosquitoes, Aedes mosquitoes and Culex species. The major vector control intervention in this area is LLINs [19], and the last mass-distribution of nets, prior to this study had been conducted between 2010 and 2011 [19].

Study procedures
This study was part of a high-resolution longitudinal mosquito surveillance program, where we collected host-seeking and indoor-resting mosquitoes between January 2015 and January 2016, in randomly selected households, and also selected sentinel households within the Health and Demographic Surveillance System (HDSS) maintained by Ifakara Health Institute in the study area [20].
Monitoring indoor densities of host-seeking and resting disease-transmitting mosquitoes. Firstly, 1600 households were randomly selected from the HDSS households database in three study villages of Kivukoni, Minepa and Mavimba, and assigned to 16 geographical clusters each consisting of 100 households. The sampling clusters were assigned based on household latitudes so that clusters 1-16 were arranged in a north-south direction in the study area. From each geographical cluster, six households were selected for sampling. These included three households randomly selected on monthly basis, and another three households that were fixed as sentinel households, visited repeatedly every month. This way, 96 households were studied per month, each time consisting of 48 sentinel households across the study area, and 48 new households randomly selected each month. Geo-positions of all the households were recorded using hand held GPS receivers (Magellan eXplorist 110). Each household was visited only once each month, and mosquitoes were sampled from one room with occupant(s) sleeping under mosquito nets, using miniature Centre for Disease Control (CDC) light traps set near occupied bed nets at the foot end, at approximately 1.5 m from the ground [21,22]. The traps were operated from18:00 to 06:00 each night.
Indoor resting mosquitoes were collected using a backpack aspirator [23] from the walls, bed nets and floors in same households where the CDC light trap catches were conducted. Aspiration of resting mosquitoes was conducted for approximately 15 minutes in each household. Each morning, the collected mosquitoes were killed in a closed container using alcohol. The mosquitoes were identified morphologically and sorted by taxa and sex before they were identified by PCR. The identification was done following the taxonomic key developed by Gillies and Coetzee, i.e. Keys for the Anopheles of Africa, south of Sahara [24].

Identification of sibling species and detection of Plasmodium infection in malaria vectors.
All female An. gambiae complex mosquitoes, were examined by multiplex polymerase chain reaction (PCR) technique as described by Scott et al. [18]  [26] to distinguish sibling species. All Anopheles mosquitoes caught were segregated by species and examined in pools by enzyme-linked immunosorbent assays (ELISA), to detect Plasmodium sporozoites in their salivary glands [27]. To avoid false positives, the ELISA lysates were boiled for 10 minutes at 100˚C, effectively eliminating heat-labile non Plasmodium falciparum protozoan antigens [28]. We assessed all the collected Anopheles gambiae s.l and Anopheles funestus s.l mosquitoes by ELISA, using heads and thoraces. However, only a sub-sample was assessed by PCR for species identification, using legs.
Identification of mosquito blood meals. Blood meal analysis was also done on a subsample, working only with the mosquitoes that were blood-fed. Each mosquito abdomen was separated from the thorax and homogenized in phosphate buffer saline (PBS) following the method described by Beier et al. [29]. This way, all the blood-fed mosquitoes were screened for human, bovine, chicken, goat and dog blood using ELISA. Assessment of insecticide susceptibility and parity rates. The insecticide susceptibility tests were conducted using standard WHO guidelines [30]. Insectary-reared An. gambiae s.s (Ifakara strain) was used as a reference susceptible population. Because of the difficulties in finding An. funestus larvae in the study area, adult female mosquitoes were collected indoors from April to July 2016, by CDC light traps from the same households in three villages. The mosquitoes were transferred into netting cages and maintained on 10% glucose solution. An. funestus mosquitoes were then separated from other mosquito species by gently sucking them from the cages into a small-netting cage using mouth aspirators and all other mosquito species were killed.
Non blood-fed adult An. funestus s.l females were therefore exposed to 0.75% permethrin, 0.05% deltamethrin, 0.05% lambda cyhalothrin, 4% dieldrin, 4% DDT, 0.1% bendiocarb, 0.1% propoxur, 0.25% pirimiphos methyl, and 1% malathion, as recommended by the WHO. Up to 100 adult females of unknown age were exposed to each insecticide during the experiment, including at least four replicates. In each test, mosquitoes collected from the same villages at the same time were exposed to oil impregnated papers, constituting additional controls. The number of knocked-down mosquitoes in each tube was counted and recorded after 10, 15, 20, 30, 40, 50 and 60 minutes. After one hour of exposure, mosquitoes were transferred to the holding tubes and maintained on 10% sugar solution then monitored for survival and mortality after 24 hours. The post exposure observation was conducted after 80 minutes only if the knockdown was less than 80% [30].
Physiological age of the wild-caught female malaria vectors was approximated based on the status of their ovaries, i.e. whether they had previously laid eggs or not [31]. Abdomens of a randomly selected sub-samples of unfed female An. funestus and An. arabiensis collected from the selected households were dissected for parity [32]. Each unfed female mosquito was first anesthetized in a refrigerator. A drop of distilled water was added to a slide and each specimen kept still on the slide, then the seventh and eighth abdominal segment was pulled using a fine needle under a stereo dissecting microscope. The ovarial tracheoles were then observed under a compound microscope at 10X magnification, to determine whether they were parous or nulliparous [31].

Data analysis
Statistical analyses were carried out using R statistical software version 3.3.1 [33]. Mosquito abundance was over-dispersed so we modeled it following a negative binomial distribution using the generalized linear mixed model automatic differentiation model builder (glmmADMB) package [34] with log link function. Here, mosquito abundance (nightly total count) were treated as response variable while locality (villages) were treated as fixed effect.
Plasmodium infection rates and blood meal sources of malaria vectors were estimated as percentages of total mosquitoes assayed. Similarly, observed mean mortality after 24 hours of exposure to each insecticide was calculated as percentage of total mosquitoes exposed in all four exposure replicates [30]. The insecticide susceptibility test findings were interpreted according to WHO guidelines [30].
The EIR was calculated as a function of biting rates and proportion of tested mosquitoes infected with sporozoites. Since we initially calculated these biting rates for each trap night, the EIR was adjusted by multiplying the daily EIR values by 365 nights (Annual EIR = Nightly biting rate x Sporozoite rate x 365). We relied on values comparative trap evaluations done in the same study area, where CDC light traps had been directly compared to human landing catches inside households [35]. Those earlier trials had indicated that CDC light traps consistently caught 0.3 times as many An. gambiae s.l as human landing catches, and 0.68 times as many An. funestus mosquitoes as human landing catches [35]. In this current study, we have therefore adjusted the CDC light trap catches using these coefficients, but also presented the unadjusted results.

Ethical considerations
Meetings with local leaders were held in the study areas and the main aims of the study were explained by the scientists. Permission to conduct studies in the separate villages was obtained from respective leaders of each of the villages; Mr. Ayubu Salila, Mr. Kiluka Mdumba and Mr. Baalesa Ligola for Kivukoni, Minepa and Mavimba villages respectively. I confirm that the field studies did not involve endangered or protected species. A written and signed informed consent was obtained from each household head participating in the study. All information was given in Kiswahili, the local language. All the households were provided and protected by intact LLINs (

Indoor densities of Anopheles mosquitoes in the area
A total of 25670Anopheles mosquitoes were collected by CDC light trap and back pack aspirator, comprising An. funestus s.l (18.7%) and An. gambiae s.l, which in turn comprised of An. arabiensis sibling species (79.0%), An. pharoensis (0.5%), An. squamosus (0.4%), An. ziemanni (0.3%) and An. wellcomei (0.4%). The highest densities, i.e. 54% and 65% of all An. arabiensis and 65% of all An. funestus were observed in the middle of study area, i.e. Minepa village. Densities of An. peaked between May, while the density of An. funestus peaked in February and May.one month after onset of the long rainy season, which is the main malaria transmission season. The lowest vector densities were observed between September and November (Fig 2).

Sibling species of malaria vectors and Plasmodium infection rates
A total of 20,135 An. gambiae s.l and 4,759 An. funestus s.l mosquitoes were assayed in the laboratory by PCR or ELISA or both. All the An. gambiae s.l. mosquitoes tested by PCR were confirmed to be An. arabiensis (100%

Mosquito blood meal sources
A total of 82 mosquitoes that were identified by PCR also had blood meals in their abdomen, so these were assessed by way of blood-meal ELISA. These included 64 An. arabiensis, 17 An. funestus s.s, and 1 An. leesoni. The study identified a broad spectrum of host blood antigens, mostly humans (79.2%), followed by cattle (17.1%), dogs (2.4%) and chicken (1.2%). The analysis did not detect any mixed blood meals. The results for mosquito blood meals are summarized in Table 3. Sporozoite infected An. funestus were widely distributed across all the study villages, it was also observed that An. arabiensis were mostly confined to an area in the middle of the study area (Fig 4). The total number of sporozoite infected mosquitoes during the year was, however, too small to enable detailed analysis of spatial relations.

Discussion
Malaria prevalence is declining across most of African countries [1], due to improved vector control [1] and case management, but also urbanization, improved health care and better living standards [1]. This study has also determined that An. funestus vector is resistant to pyrethroids used on LLINs (Table 4), and survives unexpectedly longer than An. arabiensis, thus requiring new control approaches. Given the extensive resistance in the species population against a wide range of insecticides currently approved for public health use, it appears that we now have only a limited set of options, including organophosphates such as pirimiphos methyl. Thus, one of the main control methods that could be used against this species is focal IRS with long-  Targeting Anopheles funestus can improve malaria control in Tanzania lasting formulations of organophosphates (e.g. Actellic 1 CS-300), against which An. funestus are still highly susceptible. Other potential interventions may include house improvement and house screening, as well as use of eave baffles treated with organophosphates, which would take far less insecticide, but deliver equal impact on the mosquito populations as IRS (Killeen et al unpublished manuscript). Going forward, it may be desirable to conduct new evidencebased studies in well-characterized areas such as these study sites, to demonstrate possible vector elimination strategies leading to malaria transmission control. Approaches may include combinations of organophosphate-based IRS alongside other interventions like LLINs, improved housing and larval source management, especially where rice growing is commonly practiced. To achieve malaria elimination, we must identify and target all pockets of transmission, including geographically distinct areas of high transmission and demographically high-risk sub-populations [40,41]. In this study setting in rural south-eastern Tanzania, low-level transmission persists, despite over 80% LLIN coverage [38]. The persistent residual transmission is most definitely mediated predominantly by An. funestus s.s, suggesting possibly that targeted, high impact, species-specific interventions may improve control and accelerate efforts towards eventual malaria elimination. Our recent findings suggested that the overall distribution of malaria mosquitoes, including An. funestus is strongly influenced by and can be predicted based on numbers and spatial variability of household occupancy [8]. The densities of Anopheles mosquitoes were highest in rainy season even though transmission was evidently sustained at low densities even in dry season, mostly by An. funestus. Since aquatic breeding habitats for An. funestus are not easily identified, it may be necessary to identify other environmental characteristics that affect natural abundance of An. funestus within villages and between households. This approach can be feasible by involving the local community and volunteers. The evidence from Mwangungulu et al show that scientists can rely on community knowledge and experiences to identify areas where mosquitoes are most abundant [42]. The data presented here would be essential in developing control measures that target primarily the most important residual vector species, in this case An. funestus. In these specific study settings in Tanzania, the nucleated settlements, and spatial correlations between human biomass and vector densities [8], indicates this proposed study area renders itself well to spatially-targeted interventions. The study has rigorously demonstrated that An. funestus is now the primary vector of malaria in this area in rural-south eastern Tanzania, despite concurrently high densities of other species such as An. arabiensis. The area historically had very high malaria transmission, averaged 400 infectious bites/person/year (ib/p/y) until early 2000s [43], with prevalence as high as 70% in 1990s [44], but this study indicates that transmission has now dropped around 20 ib/p/y. Moreover, this study has also demonstrated that An. funestus s.s co-exists with An. rivulorum and An. leesoni, but malaria transmission was mainly mediated by An. funestus s.s. It is not possible to determine conclusively if the greater contribution of the An. funestus s.s. to malaria transmission is mostly due to its relatively higher abundance than that of the other sibling species, but it is clear that this species mediates most of the residual transmission.
Of all the An. funestus mosquitoes that tested positive for malaria parasites, An. funestus s. s. comprised 96% of those An. funestus s.l. which were positive for Plasmodium sporozoites., though their overall relative abundance was 76.6%, while An. rivulorum was responsible for 4% of the infections though their relative abundance was only 2.9% of all the An. funestus mosquitoes. Anopheles leesoni, constituted 7.1% of all mosquitoes in this group, but carried no infections at all. Given that the ELISA lysates were all boiled to eliminate any false positives [28], occasionally associated with non-Plasmodium protozoans transmitted by zoophilic mosquitoes [45], we can assume, based on the 4% sporozoites rates, that An. rivulorum is indeed an important vector in this setting, and that its role is limited mostly by its low densities, relative to the other species. Other studies in East Africa have previously demonstrated the importance of the species in residual transmission settings [46], its role increasing after the decline of An. gambiae s.s populations [47,48] (Table 5). Our study did not however,  [49], An. squamosus [50] and An. ziemanni [51], but no such evidence was observed in this current study. Table 5 shows some of the other localities where An. funestus mosquitoes significantly contribute to the residual malaria transmission. The extremely high proportion of Plasmodium infections currently mediated by An. funestus mosquitoes in this area is compounded by, and is likely also a result of the very high insecticide resistance in this species. This study found that An. funestus mosquitoes were not only carrying most of the residual malaria parasites, but they were also very highly resistant to common pyrethroids used for vector control, i.e. deltamethrin, permethrin, lambda-cyhalothrin, as well as to the organochlorine, DDT. Moreover, there was also reduced susceptibility to insecticides like bendiocarb and propoxur, which would normally be candidates for replacing the pyrethroids. An earlier study by Lwetoijera et a.l in the neighbouring Kilombero district, also found increasing importance of An. funestus in malaria transmission associated with high resistance [10]. Even though the insecticide susceptibility status for An. arabiensis was not evaluated in this study, a separate study from these villages by Matowo et al (unpublished findings) found An. arabiensis to be also highly resistant to pyrethroid chemicals.
While the high human blood index observed in the An. funestus clearly suggests that the species could still be best targeted by household-based interventions that protect humans. Observations of increasing outdoor biting by the different An. funestus sibling species has been reported in this study area [56,57], which warrants consideration of complementary interventions that also target outdoor-biting vectors. Also, given the high resistance to the common insecticides for vector control, it appears that use of organophosphates, against which the vector is still susceptible, notably pirimiphos methyl, could provide a temporary solution against residual malaria transmission in this area. Long-lasting formulations of pirimiphos methyl already exist [58], which could be used for IRS in this area to complement LLINs, as it has already been implemented in Zanzibar, northern region and lake zones in Tanzania  . This study did not assess the role of An. funestus on outdoor malaria transmission, but, a separate study conducted in the same villages by Ngowo et al (unpublished findings) yield evidence that some An. funestus collected outdoors were found to be sporozoite positive. Furthermore, tools should be included such as chemicals with new AIs available for IRS that are being evaluated for consideration by the WHO, including non-pyrethroid active ingredients like neonicotinoids as well as bednets that combine synergists with pyrethroids, or those that combine different insecticide classes to tackle insecticide resistance and lower vector survival.
One limitation of this study is that collections for insecticide susceptibility were conducted on wild caught females of variable age, since we could not find adequate larvae for the assays. WHO recommends where possible to use mosquitoes of the same age cohort (F1 progeny). Using wild mosquitoes that are older than the standard 3-5 days may lead to underestimation of resistance depending on the species distribution and the insecticide being tested. Blood meal analyses were carried out on visually blooded mosquitoes and so may have missed those with small amounts of blood in their abdomens, such as those that had interrupted feeding or had partially digested the blood. Another potential limitation is that parity assessments, such as the ones conduced here are greatly impacted by season, this study conducted parity assessment for mosquitoes collected for two months, we suggest that future further studies should be done to assess parity status across seasons. Lastly, there was also a significant nonamplification of mosquito samples during the PCR analysis to determine the sibling species. About 13.4% of all sample analyzed in the laboratory did not amplify. An analysis of the sample handling processes, the analysis process, and the control samples, suggests that the laboratory processes were unlikely to be the reason for the non-amplification. Instead, this situation might have arisen due to sub-optimal storage conditions of some samples, or minor discrepancies in the morphological identification. These specific processes therefore deserve greater attention in future analyses.

Conclusion
This study provides an update on the residual malaria transmission situation in rural southeastern Tanzania, where LLINs were already widely distributed. Though An. arabiensis is still the most abundant vector species here, ongoing residual transmission is predominantly mediated by An. funestus, possibly due to high resistance and high survival probabilities. Interventions that effectively target An. funestus could therefore significantly improve control of residual malaria transmission in the area.