A genotypically distinct, melanic variant of Anopheles arabiensis in Sudan is associated with arid environments

Background Anopheles arabiensis, an important malaria vector in Sudan and other countries in sub-Saharan Africa, exhibits considerable ecological and behavioural plasticity allowing it to survive in the harsh conditions of arid regions. It has been shown that adult populations of An. arabiensis in the semi-desert habitat of western Khartoum State survive through the long dry season in a state of partial aestivation, characterized by limited feeding activity and a degree of arrested ovarian development. Anopheles arabiensis in these sites occurs in two phenotypic forms. One is large and heavily melanized, the other has the typical characteristics of An. arabiensis as found elsewhere in Africa. The extent of genetic variation in these forms was examined in widely separated locations in Sudan, including Kassala, Gedaref and the Northern States between 1998 and 1999 and 2004 and 2006. Methods Each mosquito specimen was identified using standard morphological keys and a species-specific PCR test. Sequence variation in a 660 bp fragment of the mtDNA ND5 coding region was examined and the extent of genetic divergence between the forms was estimated from FST values using DNASP version 4.9. TCS 1.13 software was used to determine the genealogical relationships and to reflect clustering among mtDNA haplotypes. Results The melanic and normal forms were found in sympatry in Kassala, Gedaref and Khartoum states, with the melanic form commonest in the hottest and most arid areas. Both forms were encountered in the periods of study: 1998–1999, and 2004–2006. Only ten specimens of An. arabiensis were collected from the Northern State in February 2006, all of which were of the normal form. Based on the ND5 analysis, there was a marked subdivision between the normal and melanic forms (FST = 0.59). Furthermore, the melanic form showed more genetic variability, as measured by haplotype diversity (0.95) compared with the normal form (0.57), suggesting larger effective population. Conclusions This is the first demonstration of correspondent phenotypic and genetic structuring in An. arabiensis. The high level of genetic differentiation shown by the mtDNA ND5 locus suggests that the two forms may represent separate species. It is hypothesized that the melanic form is better adapted to hot and arid environments. Electronic supplementary material The online version of this article (doi:10.1186/1475-2875-13-492) contains supplementary material, which is available to authorized users.


Background
The main groups of malaria vectors in Africa are Anopheles gambiae, Anopheles funestus, Anopheles nili and Anopheles moucheti. Each of these comprise a complex or group of genetically distinct species that are similar in morphology, but vary in traits that affect their role in transmission of malaria [1][2][3]. The most important of these groups is the An. gambiae complex, which comprises eight closely related species that are distributed through sub-Saharan Africa and its outer islands [4][5][6][7]. Within this complex, An. gambiae, Anopheles coluzzii and Anopheles arabiensis are the most efficient vectors of human malaria in Africa [5,8,9].
In contrast, strong genetic differentiation in mtDNA (ND5) was detected between allopatric populations of An. arabiensis from the island of Reunion and the African continent, which was attributed to the low effective population size (Ne) on the island [32]. Lee et al. [36] reported a fixed nucleotide on the X chromosome between populations in East-southern Africa and those in Central Africa.
Anopheles arabiensis is the primary malaria vector throughout much of Sudan [37]. Other Anopheles species, such as An. funestus, An. nili, Anopheles pharoensis, Anopheles rufipes and Anopheles dthali are also present in the country, but they play a negligible role in malaria transmission. Malaria transmitted by An. arabiensis, continues to be a major health problem in Sudan [38,39]. It has long been postulated that populations of An. arabiensis undergo aestivation during the dry season [40,41]. It has been suggested that adults of a local population of An. arabiensis, in semi-desert habitat in Khartoum state (Sudan) are highly adapted to survive through the harsh long dry season, in a state of partial aestivation "with limited feeding activities and a degree of arrested ovarian development". Similar findings were recently reported for An. gambiae and An. coluzzii (formerly S and M molecular forms) in Mali [42][43][44][45][46]. Interestingly, these studies showed that whereas aestivation is a dry season survival strategy used by the M form of An. gambiae, populations of An. arabiensis and the S form of An. gambiae from the same area are more likely to rely on migration from distant locations [43].
It is hypothesized that the persistence of An. arabiensis populations in the arid areas in Khartoum State may be reflected in the genetic structure of these populations. To test this hypothesis, a molecular study was conducted on adults of An. arabiensis collected from irrigated sites along the White Nile and from an arid region, West of Khartoum, where previous studies showed the presence of aestivating adults of this species [40]. In these regions adults of An. arabiensis occur in two phenotypic forms, distinguishable by their size and extent of melanization.

Methods and study sites
Mosquito adult samples were collected from irrigated and non-irrigated habitats in Khartoum State (Central Sudan), Gedaref State (eastern Sudan), Kassala State (eastern Sudan) and Northern State. These locations are characterized by marked variations in rain precipitation and malaria endemicity ( Figure 1 and Table 1). Whereas Gedaref and Kassala States are in the zone of hyperendemic malaria, Khartoum and Northern states are considered to be hypo-, meso-endemic or zones free from malaria, respectively [39].

Khartoum state
Located in central Sudan, at the confluence of the White Nile and the Blue Nile, (15°30' -15°45' N and 32°15' -32°45' E), Khartoum State is characterized by a long dry and hot season between October -June followed by a short rainy season (July -September) with an average annual precipitation of 240 mm. Khartoum State has poor dry desert scrub vegetation, except along the riverbanks and in irrigated agricultural schemes. Sites used for collection of mosquitoes in Khartoum State were dry non- These villages lie along eastern bank of the El Rahad River, which is a seasonal watercourse that flows during and after the rainy season, June -December. Thereafter it fragments into small water pools during the dry hot season. Gedaref State is characterized with rich vegetation, especially during the rainy season. The main trees and bushes found in the study area are Balanites aegyptiaca, Acacia seyal, Acacia mellifera, Acacia nilotica, Acacia senegal, Combretum spp., Azadirachta indica and Ziziphus spina-christi. In the most of the places, the surface of the ground covered by Sorghum spp., Schoenfeldia ssp., Cynodon spp., Aristida spp., Cinchrus spp., and Brachiara spp. grass, which flourishes during the rainy and post-rainy season and dies out during the dry season.

Kassala state
Kassala State lies about 611 km East of Khartoum, at the edge of the semi-desert region of eastern Sudan, Gedaref State from the South and close to the border with Eritrea. The climate of the area is characterized by a hot dry season, which extends from March-June followed by a short rainy season (June -October) with an average annual rainfall of 400 mm. The vegetation of Kassala is similar to that found in Gadaref State, but with a lower

Northern state
Located in the northwest corner of Sudan bordering Egypt and Libya, the Northern State is an extremely dry desert, intercepted by a narrow stretch of seasonally flooded and irrigated areas around the Nile. Rainfall rarely exceeds an average of 1 mm per annum, in most years and the vegetation ranges from true desert to semi-desert scrub. Daily maximum temperatures are generally high reaching 47°C beween May and August, with desert temperatures cool at night, and a minimum temperature of c.25-29°C.

Morphological and molecular identification of mosquitoes
Mosquitoes were identified morphologically using standard taxonomic keys [47,48].
Since the only An. gambiae s.l. species found in the study areas is An. arabiensis, these taxonomic keys provided a sufficient tool for species identification. In order to confirm the identification of the An. gambiae complex to species level, a molecular diagnostic test was performed using the rDNA species-diagnostic PCR protocol [49] protocol. DNA was extracted from individual mosquitoes using a previously described method [50].

Amplification and sequencing of a mitochondrial DNA fragment
To study molecular variation of An. arabiensis populations, a 655 bp fragment of the mitochondrial nicotinamide adenine dinucleotide dehydrogenase gene subunit 5 (NADH-ND5) coding region sequences of mitochondrial DNA was PCR amplified as previously described [20]. The PCR mix comprised 50 μl reaction volume containing 1 μL of 1:200 DNA dilution, 50 pmol primers, DMP3A = 5'-AGG ATG AGA TGG CTT AGG TT-3'; 19CL = 5'-CTT CCA CCA ATT ACT GCT ATA ACA G-3'. The PCR product was purified using QIAquick PCR purification Kit (QIAGEN). The purified PCR product and cycle sequencing was performed with the ABI PRISM Dye Terminator Cycle Sequencing Kit (Applied Biosystems). DNA sequences were assembled, and analysis of nucleotide sequences was performed using an ABI PRISM 377 (Applied Biosystems) automated sequencer following the manufacturer's protocols. All sequences were deposited in NCBI GenBank (accession numbers KJ950294-KJ950360).

Mitochondrial ND5 DNA analysis
Prior to analysis, the 655 bp fragment of mtDNA (ND5) was confirmed by blasting against the corresponding An. gambiae sequence ( [51]; GeneBank accession No. L20934). Subsequently, ND5 sequences from individual mosquitoes were aligned using the Clustal W programme and subjected to genetic variation analysis. The frequency of each haplotype, haplotype diversity (Hd), average number of pairwise nucleotide differences (π) and average number of nucleotides segregating per site (S), the population mutation rate (2NM) based on number of segregating sites (θ) were computed using the program DnaSP version 4.9 [52]. Sequences of ND5 were tested for neutral evolution, using Tajima's D test [53] and Fu & Li's F tests [54]. The extent of nucleotide differentiation between An. arabiensis populations was calculated by estimating F ST values.
Gene genealogy network TCS 1.13 [55] was used to determine the genealogical relationships and inspect clustering among mtND5 haplotypes. This method is based on statistical parsimony network [56], and was chosen because it accepts the existence of ancestral haplotypes, which are assumed to be the most frequent haplotypes, according to coalescence theory [57].

Results
Although all specimens used in the study were morphologically and molecularly typed as An. arabiensis, two phenotypic forms of An. arabiensis were recognizable. One of these forms was larger, markedly darker and more metallic in colour than the other, which appeared as typical An. arabiensis. These forms are here labelled as the melanic (M) and normal (N) forms, respectively; all specimens were labelled accordingly.
Adult specimens of An. arabiensis were easily assigned to one of the two forms and no intermediate phenotypes were observed. The difference in body size was noticeable and measured. The body size of the adult melanic form was clearly larger than the normal form (mean body size and wing length =3.25 ± 0.22 mm and 2.68 ± 0.078 for the melanic form and 2.71 ± 0.19 mm and 2.37 ± 0.09 mm for the normal form, respectively). Both melanization and body size appeared to be inheritable characters. In colonization experiments, both forms produced corresponding adult offspring. The spatial distribution of normal and melanic forms in the different sites in Khartoum State is shown in Table 2. It is clear that the melanic form occurred more commonly in the hottest and most arid areas than in the irrigated areas (Tables 2 and 3, and Additional file 1).

Haplotype diversity and genetic variation in the mtND5 of Anopheles arabiensis
Genetic variation in the populations of An. arabiensis was studied by examining variation in a 655 bp (position 6896-7550) fragment of the mtND5 gene of An. gambiae [51], for a total of 232 An. arabiensis samples spanning the four study areas. Reference sequences have been deposited in GeneBank (accession numbers KJ950294-KJ950360). All polymorphic sites were silent codon sites and the direct sequencing revealed no characteristics of heteroplasmy. No insertion and deletion differences were found within any mtDNA sequences, confirming that the sequenced segment of ND5 gene represented mtDNA rather than pseudo genes or nuclear-transposed copies.
The results of Tajima's D test [51] and Fu & Li's F tests [54] on ND5 gene sequences of all An. arabiensis individuals examined in this study are shown in Table 4. Significant departure from the neutral theory was noticeable in each population of the melanic and normal forms, except in Kassala area. The deviation was higher in the normal form, indicating higher selection pressure in the melanic form.
A total of 67 mtDNA haplotypes were found among 232 An. arabiensis individuals, 35 of which represented    (Table 4). It was clear that, in each location, the haplotype diversity within melanic populations of An. arabiensis was higher than within normal populations and almost equal to the overall haplotype diversity of the two forms. This result is consistent with a smaller effective population size of the normal form. In North State, where the analysis was restricted to a small number of the normal form (10 specimens), the haplotype and nucleotide diversity of A. arabiensis population were 0.56 (P ≤ 0.001) and 0.00085 (P ≤ 0.001), respectively.
The data provide clear evidence of corresponding morphological and genetic structuring in populations of An. arabiensis in sympatry. The average level of mtDNA ND5 sequence divergence within Sudanese An. arabiensis populations is 0.57% (±0.0002) ( Table 4). This strong population subdivision within An. arabiensis populations was also supported by the F ST value of 0.59 (P ≤ 0.0001) between normal and melanic forms. This provides strong evidence for limited gene exchange (Nm = 0.36) between melanic and normal forms of An. arabiensis, even though the forms are sympatric (Table 5).

Mitochondrial ND5 genealogy estimation
In the Templeton network (Figure 2), the sequence data of mtND5 for An. arabiensis populations segregated into two major groups in addition to the minor varieties. Those two major groups of haplotypes correspond closely to the phenotypic variation detected within An. arabiensis populations, i.e. melanic and normal forms.

Discussion
The findings obtained in this study provide first evidence of correspondent morphological and genetic population structuring in An. arabiensis that exist in sympatry. These results contradict the current notion that An. arabiensis populations in sub-Saharan Africa are panmictic [19,20,24,[26][27][28][29][30][31]33,34,58]. Using relevant keys [47,48] and species-diagnostic PCR method [49], all mosquitoes used in the study were morphologically and molecularly confirmed as An. arabiensis. However, two clearly distinguishable phenotypic forms were found: a typical normal colour-size form and a larger heavily melanized form. Furthermore, our observations lead us to hypothesize that the melanic form is more adapted to arid hot environment, as deduced from relative higher abundance and larger effective population size. This adaptation may have been conferred by the increased level of melanization, which is known to provide protection against desiccation [59][60][61][62][63]. From a molecular viewpoint, the normal form is identical to An. arabiensis type form based on colour and mtND5 sequence. The mtND5 gene of the melanic form is different from An. arabiensis type form by up to four nucleotide substitutions (one apparently fixed and three other variable nucleotide substitutions), although these forms are sympatric in different collection sites. These nucleotide substitutions, which were found in melanic form, are similar to those found in An. arabiensis populations in Senegal [20].
The average level of mtDNA ND5 sequence divergence within Sudanese An. arabiensis populations (0.57%) is higher than the average level of mtND5 sequence divergence within An. gambiae (0.38%) and An. arabiensis (0.46%) populations across Africa [20]. Moreover, the average level of mtDNA ND5 divergence between species (An. gambiae and An. arabiensis) was only 0.46% per nucleotide.
The F ST value between sympatric populations of melanic and normal forms of An. arabiensis is significantly higher than previous F ST pairwise studies of the same  [20,32,36,64]. Moreover, intraspecific estimates of genetic differentiation (F ST ) within An. arabiensis populations were low (0.098) comparable to the interspecific F ST estimates between An. arabiensis and An. gambiae, which were recorded as 0.07 and 0.12 for sympatric and allopatric populations, respectively [32]. The failure to find the melanic form in Northern State may be a consequence of the small sample size, or ecological separation due to the highly restrictive habitats of An. arabiensis in this area. Elsewhere in this study, the two forms were found in sympatry suggesting that there is a strong level of reproductive isolation between them. In this part of Sudan, where no rain may fall for several years, there is a sharp contrast between extreme desert conditions and irrigated areas along the Nile. The only suitable breeding and resting sites for An. arabiensis exist in agricultural schemes along the Nile or the ponds created by the seasonal flood in agricultural schemes. On the other hand away from the Nile there is no possibility of temporary mosquito breeding and resting sites due to lack of vegetation, extreme aridity and sandy soil topography of the area and in addition lack of wellpumped tanks. The few specimens of An. arabiensis collected in this study were captured during a short visit focusing on the irrigated area, where permanent breeding sites are present. Future studies should survey the area for possible presence of the habitat of the melanic form.
In dry areas of Khartoum and Kassala there is no surface water. The villagers obtain their water from a number of scattered water tanks which pumps water from capped wells. There are number of small scattered farms of sorghum around each water tank. Anopheles arabiensis populations breed in temporary breeding sites available from continuous leaking of water from these water tanks. The establishment of these well-pump tanks -which is the only available source of water for the villager's uses -may serve as temporary breeding and resting sites for An. arabiensis in dry zone areas throughout the year. In Gedaref State, the study sites lie along eastern bank of El Rahad River, which is a seasonal watercourse that flows during and after the rainy season and then fragment into small water pools in the dry hot season that create temporary breeding sites.
The results of this study indicate that the melanic An. arabiensis are likely to be the mosquitoes said to have adapted to survive as adults through nine months of severe drought and heat [41]. This suggestion is based on the exclusive presence of this form during the hottest and driest parts of the year, its relatively larger size and its heavy melanization. Larger individuals should have smaller surface/volume ratio and, therefore, have higher resistance for desiccation. Larger individuals should also have higher accumulation of glycogen and lipids and, therefore, are more likely to increase their metabolic body water during dry conditions [59]. Furthermore, a number of recent studies demonstrated that melanization plays a major role in insect desiccation resistance by decreasing the permeability of the cuticle to water [59][60][61][62].
In Drosophila melanogaster, body melanization is a quantitative trait and shows significant levels of both withinand between-population variation [63]. Geographical populations of D. melanogaster from Africa, India and Australia exhibit clinal variation in melanization, which suggest adaptations to local climatic conditions [61]. Increased melanization has been associated with higher fitness under thermal as well as aridity stresses in D. melanogaster, i.e. a darker cuticle may improve thermoregulation as well as reduce cuticlar water loss [59,61,63]. The extreme dry zone areas appear to form barriers to gene flow between and among permanent wet irrigated and temporary breeding sites of dry areas An. arabiensis populations.
This study provides the first evidence of a high level of phenotypic and genetic sub-structuring of An. arabiensis in Sudan. This finding may have important implications for the ecology of An. arabiensis and the epidemiology and control of malaria in Sudan and other dry lands in Africa. In future studies, it will be important to examine other molecular markers and conduct mating studies to determine the degree of speciation between the two forms. Furthermore it would be interesting to compare the vector competence of the two forms and examine their role in malaria transmission. An understanding of the mechanisms and genes related to physiological adaptation in An. arabiensis may help explain how these mosquitoes survive through the dry season, and this in turn may help improve the control of malaria transmission in Sudan.

Conclusions
This study provides the first evidence of a high level of phenotypic and genetic sub-structuring of An. arabiensis populations in Sudan. Whether this genetic differentiation is an indication of recent speciation process or the presence of two species is unclear. Judging from the higher level of genetic variability in the melanic mosquitoes, it may be inferred that they experience less selection pressure and, therefore are better adapted to dry hot conditions of the collection sites, a phenomenon that has recently been recognized in other insect species. The marked difference between the two forms may have significant consequences for malaria transmission and its control in the region.

Additional file
Additional file 1: Polymorphic positions of mitochondrial DNA NAHD-dehydrogenase subunit 5 (ND5) gene in two forms of An. arabiensis collected from four collection sites (KS = Kassala,