The combination of a geographically extensive field survey of anopheline larvae, spatially stratified by hydro-ecology, and a genetic identification approach has revealed a complex and dynamic assemblage of potential malaria vectors across Barotseland in western Zambia. Anopheline larvae were found throughout this hydrodynamically complex area and in all ecological zones, but as expected were more widespread and abundant in the wet season (higher proportion of transect points). They were significantly more abundant in the Luena, Lui and dambo ecological zones than in the Zambezi floodplain and floodplain edge in the wet season. Within this widespread anopheline distribution there were substantial differences in abundance and distribution of different species, and differences between wet and dry seasons. A key finding was that in terms of both distribution and abundance, the anopheline population was dominated by secondary vector species. The An. coustani group dominated in both wet and dry seasons (55% and 62%, respectively), followed by An. squamosus which was more prevalent in wet than dry season (18% and 8%, respectively), and An. spp O/15 in the wet season only (5%). These secondary vector species / species groups were widespread across the region, in both wet and dry seasons, although several species (An. coustani, An. spp O/15) were more abundant in Zambezi and Luena floodplain habitats. Primary vector species (An. arabiensis, An. gambiae s.s. and An. funestus s.s.) were relatively rare in both seasons, and had very localised distributions consistent between seasons. A group of anopheline individuals that could not be identified to known species comprised 20% and 26% of the dry and wet season population respectively.
Application of molecular techniques consistent with previous studies ensured that species assignments are comparable with those from eastern Zambia [12] and western Kenya [11]. As in these study areas, we uncovered an unexpected diversity of potential vector species, but with a surprising scarcity of primary vectors (<2%) amongst 995 sequenced anopheline larvae. Bias for specific feeding or resting behaviours was avoided by larval sampling and so potentially better represents the whole anopheline community within 1.5 km of villages than adult trapping; although this sampling does not directly demonstrate exposure of the human population to the potential vector species encountered [7], persistent malaria prevalence in the region [16] indicates the presence of substantial numbers of vectors. Given the known diversity of the Anopheles genus, with over 140 species in sub-Saharan Africa [5], and the abundance of non-human blood sources across Barotseland (especially livestock [18]), it is expected that non-vector anophelines will be represented in larval sampling. Nonetheless, village-centred surveys encompassing all water body types encountered within known flight range [32] of human bloodmeals failed to reveal substantial numbers of primary vector species either in the dry season or following the peak of the wet season. Reduced average transect length in the wet season effectively further concentrated sampling on the peridomestic environment, which might have been expected to bias the sample towards more anthropophilic species, yet a smaller proportion of the sample in this season consisted of primary vector taxa.
The primary malaria vector species in Zambia are thought to be An. arabiensis, An. funestus s.s. and An. gambiae s.s., although detailed entomological studies have been undertaken only comparatively recently [46, 47]. Despite the continued high prevalence of malaria in Barotseland, An. gambiae complex and An. funestus s.s. represented only 1.7% and 0.1% of anophelines identified by genetic methods in this study. Our result is in contrast to the study by Lobo et al. (2015; [12]) which sampled adult anophelines in villages in eastern Zambia and described a predominance of An. funestus s.s. (55% of specimens) followed by An. arabiensis, and implicated both species in malaria transmission. An. funestus s.s. is also thought to dominate transmission in northern Zambia, followed by An. gambiae s.s. [48], whilst in southern Zambia An. arabiensis is thought to be the primary vector [47]. However, although An. funestus and An. gambiae s.l. accounted for 29% and 9%, respectively, of indoor and outdoor collections of adult anophelines in Western Province villages in 2013 [49], they were not the most abundant species in that collection (some of which originated in our study area). All three of these species exhibit large human blood indices in Zambia [50], and although An. arabiensis is generally considered to be less anthropophilic than An. gambiae s.s. and An. funestus, it is more anthropophilic in Zambia than elsewhere in Africa [47].
It is unknown whether the larval community we recorded is a product of the suppression of populations of endophilic and endophagic primary vectors by increased interventions [16], as has been documented elsewhere (e.g. [4, 51, 52]), or is representative of a natural species assemblage of anopheline vectors in Barotseland that has not been historically dominated by primary vector species. The distribution of An. gambiae s.l. found in the present study was highly clustered, with 8 of 10 specimens found within 140m of each other in the dry season. Sampling after the peak of the subsequent wet season found 3 of 7 An. gambiae s.l. associated with the same village, and all specimens across both seasons were closely associated with people (within 600 m of a village). This conforms to the established tendency of An. gambiae s.l. to breed in close proximity to human settlements [53, 54], and limit dispersal from these blood sources [32]. The range of aquatic habitats surveyed in our study encompassed streams and large, permanent and heavily vegetated larval habitats typically associated with An. funestus, and smaller, more ephemeral ones often favoured by An. gambiae s.l. [48, 55, 56, 57], so we do not think biased habitat sampling can explain the low abundance of primary vector species detected. Combined sampling of larval habitat and adults will be needed to further resolve the dynamic relationships among primary and secondary vector populations around villages and across the wider landscape.
An. coustani group species comprise up to 65% of surveyed larvae, which supports the limited sampling of adult mosquitoes in Western Province in 2013 that found 60% to be An. coustani [49]. The molecularly-identified An. coustani group in our study contains several closely-related species which cannot be further resolved from COI sequences; although some may be separable based on morphology, a number of species remain virtually indistinguishable [58]. In addition, molecular identification of anophelines in eastern Zambia by Lobo et al. has suggested the presence of further morphologically cryptic sibling species, denoted An. cf. coustani 1 and 2 [12] (one of which may represent An. crypticus, the most recently identified cryptic species in the group [59]).
An. coustani group was distributed across all ecological zones in this study, although consistently less abundant in floodplain edge habitats. Significantly higher abundance in permanently waterlogged dambo habitats across both seasons reflects the recognised preference of An. coustani s.s. for more established water bodies [60]. The species displays a preference for natural vegetated water bodies and an aversion to temporary, non-vegetated pools elsewhere [61], and this is supported by its ubiquitous presence in vegetated habitats [20] across Barotseland in this study. Until recently, An. coustani has been considered a secondary vector as it was seen as largely zoophilic [61], but it was recently demonstrated to be the main vector in a village in Madagascar [62] and its secondary vector status is increasingly being reconsidered. An. coustani exhibits high levels of anthropophily in some settings [15, 62, 63], practices endophagy as well as (predominantly) exophagy [64, 65, 66], and shows a markedly high rate of early biting [62, 63, 67]. It has tested positive for Plasmodium infection in Ethiopia [65], Kenya [66], and Madagascar [67]; in Zambia, An. coustani, An. cf. coustani 1 and An. cf. coustani 2 have all tested positive [12]. Even low infection rates, in combination with these behavioural traits and locally high abundance [62, 67], may allow An. coustani to play a substantive role in malaria transmission.
Amongst other potential members of the taxon designated as An. coustani group in this study, An. tenebrosus is assumed not to be a competent vector, due to its low parity and long gonotrophic cycle; it has not been detected with malaria parasites [7, 61], and records of presence in Zambia are historical [45]. These traits are also associated with An. ziemanni, and although it has occasionally been found to harbour Plasmodium, it has historically not been considered a significant vector [7, 61]. Nonetheless, in some areas it demonstrates anthropophily [68, 69] and it may be locally important as a vector in an area of low transmission in Cameroon [70]. Habitat preferences for both these species are also poorly characterised, and thought to conform broadly with those of An. coustani; both are typically associated with permanent water [61].
An. squamosus was also abundant in our larval collections during the wet season. This species is more zoophilic and exophagic than An. coustani, although an unexpectedly high degree of anthropophily has been revealed in southern Zambia [63] and Madagascar [71], and Plasmodium infection has incriminated the species as a vector in both countries [63, 71]. In Barotseland An. squamosus displayed a habitat-specific distribution, predominantly in floodplain and floodplain edge zones. Larvae have been recorded previously from a wide range of habitats, provided they are at least partially vegetated [61], but relatively little is known about this species’ habitat associations.
The third abundant non-primary vector anopheline identified in Barotseland was An. species O/15, previously identified in Kenya as a potential sibling species to An. coustani [11]. An. species O/15 is not considered a malarial vector in Kenya (negative for Plasmodium falciparum; [11]), and there is no further information on bionomics or larval habitat preferences. In the present study An. species O/15 constituted 5% of the wet season sample and was significantly more prevalent in floodplain habitats than in other ecological zones.
In eastern Zambia, 39% of 18 delineated anopheline taxa [12], and 53% of 17 taxa in Kenya [11], were designated as ‘unknown’ due to the lack of conclusive similarity to database sequences. Whilst some taxa may represent novel or cryptic species, the lack of an identity may also result from the absence of DNA sequences from known species. In the present study, identifications were based on matches at >95% similarity to published COI sequences, supported by ITS2 matches using the same threshold, and specimens that failed to yield an above-threshold match (but had consistent below-threshold matches to anopheline sequences - see Additional File 1) were designated as “unknown Anopheles species”. Some of these “unknown” individuals may represent further examples of the cryptic species diversity uncovered by recent studies [11, 12], as we know that these individuals do not match to the taxa identified in these studies even though we had high sequence coverage of the same DNA regions; others are likely to be known species which are poorly represented in the NCBI nt database. The occurrence of unidentified anophelines does not affect the key conclusion of this study, because where these individuals are matched (at <95% similarity) to a sympatrically occurring species, the majority cluster with secondary vector taxa, and we can be confident that they are not primary vectors well-represented in the published database. The occurrence of potential further cryptic anopheline species in this and other areas should be taken account of in future studies.
A key remaining question is how representative our comprehensive larval survey is of the distribution of adult Anopheles vectors and hence malaria transmission hazard. In some settings, larval densities have been found to be a poor predictor of adult abundance [72], whilst in others there is very close correlation [73]. If the larval composition is representative, there are significant implications for malaria vector control efforts in the region.