Aedes aegypti, the ‘yellow fever mosquito’, is the primary vector to humans of the four serotypes of dengue flaviviruses (DENV1-4) and the yellow fever flavivirus (YFV). Dengue is a major public health problem in the tropics, causing millions of dengue fever and hundreds of thousands of dengue hemorrhagic fever cases annually [1]. In endemic areas the annual number of cases has risen steeply since the 1950s [2]. With multiple serotypes circulating in endemic areas, 100 million infections of dengue fever (DF) occur annually, including up to 500,000 cases of the more severe form of disease called dengue hemorrhagic fever (DHF) with a case fatality rate of up to 5% [3]. Despite the development of a safe, effective YFV vaccine, yellow fever remains an important health risk in sub-Saharan Africa and tropical South America [4], [5]. The WHO estimates that there are 200,000 cases and 30,000 deaths attributable to YFV infection each year, most of which occur in Africa [6].

There are two recognized subspecies of Ae. aegypti s.l., the presumed ancestral form, Ae. aegypti formosus (Aaf), a sylvan mosquito supposedly limited to sub-Saharan Africa; and Ae. aegypti aegypti (Aaa), found globally in tropical and subtropical regions typically in association with humans. The designation of Ae. aegypti s.l. subspecies arose from observations made in East Africa in the late 1950's that the frequency of pale “forms” of Ae. aegypti was higher in populations in and around human dwellings than in adjacent forests [7], [8]. The implied correlation between color and behavior prompted Mattingly to revisit the biology and taxonomy of Ae. aegypti [9]. He described formosus (Walker) as a subspecies of Ae. aegypti that was restricted to sub-Saharan Africa and in West Africa “is the only form known to occur except in coastal districts and in one or two areas of limited island penetration.” However, this latter statement was based only on two collections, one from Ghana and the other from Burkina Faso. He also suggested that Aaf most frequently breeds in natural containers such as tree holes, and feeds primarily on wild animals. Mattingly also stated that in addition to the dark-scaled parts of the body being generally blacker, “ssp. formosus never has any scales on the first abdominal tergite.” The type form of Aaa was alternatively defined as “either distinctly paler and browner (at least in the female) than ssp. formosus or with pale scaling on the first abdominal tergite or both.” He also suggested that Aaa breeds in artificial containers provided by humans, will breed indoors, and has a preference for feeding on human blood [9].

The subsequent studies of Tabachnick, Powell, Munstermann and Wallis [10]–[21] on the population genetics and vector competence of Ae. aegypti s.l. showed that global collections fell into two clades. One clade contained Aaa from East Africa, South America and the Caribbean suggesting that these New World populations were derived from East Africa. The other clade contained Asian and Southeastern U.S. Aaa and a basal branch containing Aaf from both East and West Africa suggesting an independent New World and Asian introduction. Their parallel work on vector competence [11]–[13] showed that West African Aaf had lower competence for YFV than other global collections of Aaf and Aaa. A more recent study examined 24 worldwide collections of Ae. aegypti s.l. at 12 polymorphic microsatellite loci [22]. Two distinct genetic clusters were identified: one included all domestic populations outside of Africa and the other included both domestic and forest populations within Africa.

Fourteen papers published since 2000 [23]–[37] and 4 unpublished datasets on GenBank used sequence variation in the mitochondrial NADH dehydrogenase subunit 4 (ND4) gene to describe patterns of gene flow among Ae. aegypti s.l. collections within and among countries outside Africa (Table 1). For example, the first paper in Table 1 was a population genetic analysis of gene flow among 10 Aedes aegypti collections from seven cities along the northeastern coast of Mexico [23]. A total of 574 mosquitoes were examined and 9 novel ND4 haplotypes were discovered. Using Tamura-Nei distance [38] and neighbor joining [39], haplotypes were placed into two clades with 90% and 99% support. Table 1 documents that to date 95 novel ND4 haplotypes have been discovered and that two phylogenetic patterns were consistently noted: either mtDNA haplotypes were distributed on two well supported clades (pattern 1), seen in three published datasets [23], [25], [35], and three unpublished datasets (GB1, GB2, GB4) or as a basal group (more similar to the outgroup species) from which a second well supported derived (less similar to the outgroup species) clade arises (pattern 2 - publications [24], [28], [30], [32]–[34], [36]. These patterns are not limited to the mitochondrial ND4 gene. A study in Brazil utilized the mitochondrial Cytochrome Oxidase I (COI) gene to examine gene flow among 163 mosquitoes in 14 collections [40]. Their phylogenetic analysis identified two clades with 81 and 96% bootstrap support. Based upon comparison with GenBank COI sequences [41] from an Ae. aegypti strain collected from Kenya, another from West Africa and a third Aaf strain; they designated one clade as “East African” and the other as “West African.” A study in Argentina that included collections from Brazil, Paraguay, Uruguay and Bolivia utilized Restriction Fragment Length Polymorphism (RFLP) analysis of the ND4, ND5, COI and COII mitochondrial genes and identified three clades [34]. However, since that study did not include sequence data for these four genes they could not be compared to sequences in the present study. A combined study of Aaa, Aaf, Ae. albopictus and Ae. mascarensis from islands in the southwest Indian Ocean examined phylogenetic relationships within and among all four taxa [41]. Bayesian phylogenetic analysis clearly differentiated two clades; one (labeled GR1) had a credibility value of 0.81 and contained all mosquitoes identified as Aaf while a second clade (GR2) had a credibility value of 0.86 and contained all Aaa mosquitoes. Aaa and Aaf were monophyletic with Ae. mascarensis immediately basal. A study using microsatellites and the mitochondrial ND4 and COI genes in Bolivia detected two clades [36] with credibility values of 0.75–0.76.

Despite the large numbers of studies that have detected these two mitochondrial clades, no studies have been made of the clades in continental Africa. Assumptions about the African origin of Ae. aegypti s.l. are based upon the observation that 58 species of the subgenus Stegomyia are also endemic to Africa [42] and the greatest genetic diversity in allozymes markers [14], [15], [18] and microsatellites [22] in Ae. aegypti s.l. are found in African collections.

It is currently unclear if there is an association between the two well documented mitochondrial clades in the literature and the Aaa and Aaf subspecies or if the clades are differentially associated with East versus West Africa. To address this deficiency, the present study examines ND4 haplotypes among 426 Ae. aegypti s.l. collected at 10 locations in Senegal, West Africa and seven novel haplotypes collected in 7 locations in Kenya in East Africa. A comparison of these sequences was then made with the 95 existing haplotypes detected and reported globally in the literature (Table 1).

The phylogeny displayed in Figure 1, the phylogenetic analyses of all mtDNA ND4 haplotypes reported to date (Figs. S1 & S2) in addition to the fourteen independently derived phylogenies that appear in publications (Table 1) all support an hypothesis that Ae. aegypti populations from around the world consist of mosquitoes that arise from one of two matrilineages. Outgroups consisting of four related subgenus Stegomyia species and two additional African subgenera, consistently indicate that one of the Ae. aegypti matrilineages is basal while the second matrilineage arises from the first.

The purpose of this study was to trace the African origins of these two clades. Key observations are that all but one of the ND4 haplotypes from Senegal occur on the basal matrilineage whilst haplotypes from East Africa arise predominantly on the second, derived matrilineage (Fig. 1). However, samples from Kenya are only from the Rabaï area. Mbarakani, Bengo and Rabaï are approximately 100 m apart and this cluster is 4 km from Changombe. Further, Rabaï is 14 km inland from Mombassa on the coast. Mombassa is the second largest city of Kenya and a major port. Thus, as with Dakar in Senegal, Mombassa could easily be a place where Ae. aegypti immigrate through human commerce. It would be very interesting to sample Ae. aegypti from other locations further inland in East Africa to assess this possibility.

This pattern prompted us to re-examine all of the 215 Ae. aegypti s.l. ND4 sequences in GenBank (Table S1). Phylogenetic analyses of the 95 unique haplotypes confirmed that all but one West African haplotype occurred on the basal matrilineage. This matrilineage also contained many globally distributed haplotypes. Conversely, most east African haplotypes occurred on the derived matrilineage which also contained many globally distributed haplotypes. The phylogenies presented here demonstrate that Ae. aegypti populations outside Africa consist of “mixtures” of mosquitoes from both the basal and derived matrilineages.

Figure 1 is ambiguous as to whether Aaf or Aaa (sensu Mattingly) was the ancestor because basal haplotypes were detected in mosquitoes with and without scales on the first abdominal tergite [52]. This result is not surprising given that McClelland's 1974 study [53] also found collections of almost pure Aaf in Pensacola, Key West and Miami, Florida. Conversely, collections from Kenya, Nigeria, Tanzania, Senegal, Ghana, Burkina Faso, Sri Lanka, Calcutta, Jamaica, and Miami Airport contained diverse mixtures of Aaf and Aaa mosquitoes. Inferences about subspecies composition and West versus East African origins cannot be inferred from the earlier allozymes studies [10]–[21] nor from the recent microsatellite study [22] because they did not use McClelland's [53] scoring scheme nor did they include an outgroup.

The current study is unique in providing the first mitochondrial ND4 data from West Africa and definitively associating the two clades reported in the literature with West and East Africa. We strongly emphasize that the hypotheses and patterns described in this paper are not novel. Bracco et al [28] made 36 collections throughout the New World (Brazil, Peru, Venezuela, Guatemala, US), three from Africa (Guinea, Senegal, Uganda), and three from Asia (Singapore, Cambodia, Tahiti). They also detected two clades and concluded that “three percent of nucleotide divergence between these two clades is suggestive of a gene pool division that may support the hypothesis of occurrence of two subspecies of Ae. aegypti in the Americas.” Later the two clades were actually labeled as East and West African albeit based on only three haplotypes from long established laboratory strains [40]. Most recently a combined analysis of ND4 and CO1 also associated one clade (lineage 2) with West Africa [41].

We have recently discovered multiple chromosome inversions in Ae. aegypti s.l. [54] and M. Sharakova obtained direct visual evidence with Fluorescent In Situ Hybridization (FISH) for inversions on each arm of the third chromosome (unpublished). An obvious question arises as to how these inversions correspond to mosquitoes from the two clades and to the global versus African Ae. aegypti microsatellite clades [22].