Background
In Africa, current frontline strategies for reducing malaria transmission rely on the use of residual insecticides through application on insecticide-treated nets (ITNs) and indoor residual spraying (IRS). The contribution of these strategies to reduce child mortality and morbidity has been considerable [
1‐
4]. However, these approaches are facing challenges and limitations as the mosquito vectors they target are increasingly becoming resistant to insecticides [
5,
6] and many exhibit behavioural plasticity (e.g. biting and resting outside of houses, or early in the evening) that limits their contact with insecticides [
7], indicating that these strategies alone may not be sufficient and that new control strategies are needed to supplement them [
8].
One of the challenges undermining contemporary vector control strategies is the limited understanding of the ecological complexities that allow vector populations to persist and evade control approaches. Taking the example of African malaria vectors, there is insufficient understanding of mosquito life-history processes that occur outside of the domestic environment (e.g. houses) where they usually bite, including oviposition, larval development, sugar feeding, mating and dispersal [
8‐
10]. Most vector control studies are understandably focused on developing and evaluating specific interventions. While such studies provide the ultimate evidence for evaluating whether to adopt a particular strategy, failure to concurrently measure the ecological parameters of the target vector population during the trial means that little evidence is available to interpret why an intervention failed, and what aspects of its implementation could be modified to achieve greater success. Paying explicit attention to mosquito ecology is vital not only for interpretation of why some otherwise well proven interventions are less effective than expected in different ecological settings, but also for identifying other vulnerabilities in the mosquito life cycle that could be targeted by novel methods.
Gaining insight into the ecological processes of malaria vectors can be both logistically difficult and expensive in natural field settings. This is due to the lack of sampling tools for reliably measuring the abundance and behavioural diversity of different species, genotypes, sexes and life-history stages of mosquito vectors inside and outside of domestic environments, and because of the substantial heterogeneity in their density over time and space [
11,
12]. As a result of these inherent challenges, many researchers adopt a laboratory experimental approach to quantify key aspects of mosquito life-history and demography. However, it is recognized that laboratory conditions may be insufficient to adequately represent vector fitness and behaviour in nature. Furthermore, the artificial feeding and rearing regimes used in laboratory colonies have been associated with the appearance of behaviours [
13,
14] and phenotypic traits [
15] that are atypical of corresponding field populations. Consequently, in order to progress understanding of vector ecology beyond the limitations of current field and laboratory approaches, there is an urgent need for more environmentally realistic experimental systems where mosquito vector behaviour, ecology and population dynamics can be studied in a natural context over multiple generations.
Contained semi-field systems (SFS) have been proposed as more realistic and reliable experimental tools for the characterization and manipulation of vector ecology [
16,
17]. An SFS is defined as an experimental mesocosm, situated within the natural environment of the target vector population and exposed to similar climatic conditions, within which all natural dietary and habitat resources for their life-cycle completion are present [
16,
17]. The movement of insect vectors into or out of the SFS is typically prevented by netting which blocks their dispersal, but not natural airflow or climatic influences. A key benefit of SFS is that they permit the maintenance of relatively large vector populations in a situation where mating and other behavioural activities can occur more naturally than in the laboratory, and where inbreeding may be less likely to occur. It is thus expected that the demographics, genetic composition, behaviour and life-history of vectors maintained under SFS conditions will be much more representative of wild populations than laboratory colonies. Another advantage of such systems is that in contrast to field studies, the exposure of workers to pathogens such as malaria can be eliminated (e.g by restricting access to potential sources of infection). In the absence of such risk, researchers can conduct a wider range of experimental manipulations, including exposure to mosquito biting, that would be ethically unacceptable in the field. Finally by facilitating detailed study of a defined vector population over time, SFS provide a unique opportunity to investigate their evolutionary as well as ecological dynamics in response to experimental manipulations that mimic the effect of predicted environmental change or interventions; a feat very difficult to achieve under natural field conditions.
Although the value of SFS as an experimental tool for insect disease vectors is increasingly recognized [
17‐
21], few such systems have been successfully established [
16,
17], and none so far have reported the successful maintenance of a vector population for multiple generations. Here we report the first successful long-term establishment (> 24 generations) of an African malaria vector population under SFS conditions within an area of endemic transmission in southern Tanzania [
17]. The study focused on the establishment of
An. arabiensis, a widespread vector of malaria in Africa [
22,
23]. Historically,
Anopheles gambiae s.s, has been recognized as the most important vector of malaria in Africa. However, the abundance and distribution of this vector species is shrinking in many parts of the continent following the widespread use of ITNs, with its sister species
An. arabiensis playing an increasingly important role in maintaining transmission [
23‐
25]. This is because
An. gambiae s.s. is more endophilic while
An. arabiensis is more exophilic and less susceptible to indoor control measures. Given the growing importance of
An. arabiensis, there is increased interest in obtaining knowledge of its ecology to stimulate new approaches for its control.
Discussion
This study reports the first successful establishment of a self-replicating population of an African malaria vector,
An. arabiensis, in a contained semi-field system. It is suggested that a major contributing factor to the successful establishment of this population was the close concurrence between the environmental conditions of the SFS and those experienced by mosquitoes in nature. The mean daily temperatures recorded in all SFS larval habitats were within the range reported for natural
An. gambiae s.l. larval habitats (20°C-36°C) [
43,
44] and never exceeded their upper tolerance limit of 40°C [
34,
44,
45]. The time required for
An. arabiensis to develop from larvae to pupae within these habitats (6 - 17 days) was similar to the reported range (8-18 days) for
An. gambiae s.l. in the wild [
46]; with the median larval development under both our SFS (11.5 days) and field conditions (e.g. 11.9 days [
41]) being close. During their development, the only source of food that would have been available to larvae was microbes and algae that developed naturally within their aquatic habitats. These resources are the primary food source of larvae in natural populations [
35], and likely played a similar role in the SFS.
The survival of larvae to pupation within the SFS was highly dependent on larval habitat size, with the pupation rate in 'large' habitats being four times higher than in medium and small habitats. The pupation rate within large habitats corresponds to a daily larval survival of 96.2%, which is in line with what has been reported in other semi-field settings (95.7%) [
40]), but slightly higher than what has been reported in some field populations (e.g. 85% [
47]). The moderately higher pupation rate observed here may be a result of reduced predation and competition within our SFS relative to field conditions. It is hypothesized that the enhanced pupal productivity of the 'large' larval habitats in the SFS is a function of the greater amount of food resources (algae and microbes) they can support relative to smaller habitats. In the absence of predators and pathogens, food availability within larval habitats is a key predictor of the number of adults that emerge from them [
43]. The total quantity of microbial growth in larval habitats is related to both surface area and volume [
30] and thus are greater in the larger than small habitats where larvae may have experienced more intense resource competition.
Similar to larval development, the behaviour of adult mosquitoes within the SFS was also consistent with what has been reported in nature. The mean temperatures within available adult resting sites varied by no more than 1.3 °C, and all were within the tolerance range of
An. arabiensis [
48,
49]. However, there was considerable variation in resting preference. Both males and females were more likely to rest in clay pots (outdoors) than in indoor sites; confirming the previously demonstrated exophilic tendency of this species in the wild [
50,
51]. Although
An. arabiensis is known to be substantially more exophilic than
An. gambiae s.s., it has not previously been possible to estimate the relative proportion of outdoor resting within natural populations. If our SFS results are typical of natural populations, it suggests that up to 60% of resting adults may be missed by surveys and control measures targeted indoors. Restricting consideration to indoor resting sites, females were more selective than males. Specifically, females were more likely to rest in the cow shed than in other sites, whereas indoor resting males occurred with similar frequency in all three resting sites. The closer association of females with the cow shed suggests they prefer to rest close to the host blood source, and may be most efficiently targeted there.
Several other aspects of the adult behaviour and life history of
An. arabiensis within the SFS conformed to what is known of their natural ecology. For example, female
An. arabiensis in the SFS readily blood-fed on cows and were able to maintain their population exclusively on this host type as has been reported in other field populations in East Africa [
52]. In nature, plants are thought to be a major source of sugar for both male and female mosquitoes [
27,
28]. Mosquitoes can readily imbibe and digest plant juices and nectar to enhance their survival [
53,
54]. As males rarely survive for more than 48 h without a sugar source [
54], evidence that a large proportion of adult males (at generation 20) within the SFS (80%) were four days or older indicates that they were feeding on plant nectar sources within it. These plant sugar sources may also have been used as nutritional supplements by females. Further studies such as gut content analysis could confirm the extent of
An. arabiensis reliance on plant nectar, and which types are preferred.
Analysis of the age structure of male and female
An. arabiensis at generation 20 indicated that a significant proportion of both male and females survived beyond the minimum period required to reproduce. In this study, approximately 80% of males were estimated as being > 4 days, the period beyond the peak of
An. gambiae s.l. mating activity [
55,
56]. Similar analysis of adult females indicated that 28% of females had survived through their first gonotrophic cycle (estimated as < 4 days). The observed parous rate of 0.28 corresponds to an estimated daily survival rate of 0.53, which falls within the range of reported for adult
An. gambiae s.l. daily survival during the dry (0.49) and wet seasons (0.84) in East African populations [
25,
57,
58]. Although the age-grading methods used in this study provide a general indication of adult mosquito age, they could not precisely estimate how long individuals survived beyond four days. Further age-grading studies using chronological age estimation methods to more precisely estimate the life span are recommended [
59].
Another natural adult behaviour observed within this SFS population was swarming. Swarming has been suggested as the primary reproductive strategy of
An. gambiae s.l. mosquitoes [
60,
61], and has been documented in several wild populations in East [
62] and West Africa [
61,
63]. However in some parts of East Africa, male anopheline swarms have been difficult to observe, possibly due to the fact that they occur at dusk when visibility is poor [
61,
62,
64], or because these populations deploy alternative strategies such as mating indoors [
65]. Due to its inconsistent occurrence (observed on 58.3% of occasions), we could not establish whether swarming was the primary mating strategy of mosquitoes in the SFS. Further investigation is needed to identify the mating strategies of
An. arabiensis both in the SFS and the wild population from which they were established.
The SFS approach adopted provides useful opportunities for characterizing mosquito demography, life-history and behaviour traits that are difficult to measure in nature, and poorly represented in the laboratory [
11,
66]. This advantage will be particularly strong for vector species that are difficult to colonize under laboratory conditions (e.g.
An. funestus and
Mansonia annulata) and/or sample in the wild; as the relatively more natural conditions within the SFS may prove more amenable for their establishment. Although it is argued that biological inferences generated in SFS provide a more accurate representation of field populations than laboratory colonies, this approach also has limitations. For example in the SFS, there were no interactions between
An. arabiensis and other mosquito species in larval habitats as occurs in nature [
67]. Furthermore the high availability of aquatic sites within the SFS probably minimized cannibalism and other types of intraspecific competition [
68]. Also unlike field settings, pressure from insecticides or other vector control interventions was absent in the SFS, and the diversity of natural predators and competitors was probably under represented. Furthermore although humans are a common host for
An. arabiensis in many African settings, this host type was not available in the SFS during their typical host-seeking period (e.g. 10 pm -5 am [
69]). Consequently,
An. arabiensis within this system were not exposed to human malaria parasites, which may have eliminated another source of selection pressure that acts on natural mosquito populations. However, as only 1-2% of
An. gambiae s.l. become infected even in highly endemic settings, it is perhaps unlikely that parasites are a significant source of selection [
70]. Finally, there have been some accounts that the host preference of vectors (from an insectary population) when assayed under semi-field conditions give a biased representation of natural feeding preferences [
14]. Whether SFS populations that have been established directly from a wild population and consistently maintained on natural host types may also develop atypical preferences is not yet unknown, but careful and repeated monitoring is required to ensure that these and other behavioural traits of SFS populations reliably approximate natural vector ecology. Supplementing stable population within SFS routinely with fresh materials from the field may be another option to circumvent this, but requires further study. Thus it is cautioned that SFS studies should be seen as a bridging ground between lab and field, and not replacement of field studies. Furthermore, where possible observations from these SFS studies should be verified under natural conditions.
In addition to elucidating fundamental aspects of vector ecology, SFS studies have a variety of other potential uses including the preliminary evaluation and optimization range of vector control interventions and trapping methods. SFS can be used to develop and optimize new field sampling tools and trapping methods, and identify the most promising candidates to take forward for full field testing within a relatively short period of time (e.g. optimal doses for repellants/attractants [
71]). Additionally, the long-term establishment of vector populations allows the testing of some important evolutionary questions that are difficult to monitor in open field populations, such as prediction of the nature of behavioural and physiological resistance strategies against specific interventions. Of particular relevance is the use of SFS studies to determine the feasibility of disease control strategies based on the release of sterile and/or genetically-modified refractory mosquitoes [
72]. Despite the optimism that such novel control strategies have garnered [
72], so far investigation of their feasibility has been largely based on laboratory studies [
72]. It remains unknown whether mosquitoes carrying GM sterility or refractory traits would be fit enough to compete for female mates with their wild counterparts [
72]. As unrestricted field trials of GM mosquitoes are unlikely to be authorized before all potential biosecurity risks have been evaluated [
72], testing the viability of these mosquitoes under contained semi-field settings will be a mandatory first step before proceeding
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KRN, BGJK and HMF designed this study. KRN and DM carried out the laboratory and SFS work. KRN analyzed and interpreted the data. KRN, HMF, BGJK drafted the manuscript. All authors read and approved the manuscript.