Mosquitoes (Diptera: Culicidae) are a monophyletic group of true flies [1
], recognizable by their elongate adult mouthparts through which the females of most species feed on vertebrate blood. Mosquitoes occur throughout temperate and tropical regions, and well beyond the Arctic Circle, but are most diverse in tropical forest environments [5
]. A bewildering amount of morphological diversity parallels their spectacular radiation into virtually every conceivable collection of water, ranging from a few droplets trapped by plant parts to large bodies of fresh and brackish surface water, making mosquitoes "as ubiquitous as water" [6
]. The relationship between human health and those species that are medically important (<200 of 3,524 currently recognized; http://mosquito-taxonomic-inventory.info/
) has driven most mosquito research. Within this small subset of disease vector species, morphological similarities between close relatives (e.g
., cryptic or sibling species complexes) continue to pose practical and academic challenges to disease control, conventional taxonomy, and phylogenetic inference. Ironically, it is not the morphological similarity but rather the morphological diversity of mosquitoes that has confounded efforts to delimit many supraspecific groups and reconstruct their evolutionary history.
The Culicidae had an ancient origin, probably in the Jurassic [7
], consistent with the fossil record of their sister group Chaoboridae [9
]. Unfortunately, the sparse mosquito fossil record sheds no light on evolutionary relationships in the family. Traditional classification of Culicidae based on the phenetic framework of Edwards [7
] resulted in arbitrary groupings reflecting intuitive interpretation of morphological similarities. More modern classifications have incorporated important revisions of select genera and tribes based on explicit methodology, but in the absence of a comprehensive application of quantitative methods across the family, the result still does not entirely reflect evolutionary history [reviewed in [2
]]. Current classification divides the family into two subfamilies (Anophelinae and Culicinae), 11 tribes, and a minimum of 44 genera [2
] (Table ).
Species of Culicidae included in the phylogenetic analysis, with reference to their classification and distribution [after 2].
Generic-level relationships across all Culicidae have rarely been studied. The first attempt [14
] was based on comparative bionomics and morphology using intuitive methods typical of that time. Surprisingly, further attempts using modern cladistic methods were not made for nearly 50 years. The most comprehensive of these phylogenetic re-analyses employed 73 morphological characters to examine the relationships of the 38 genera then recognized [1
]. In general, almost none of the hypotheses raised by the 1951 phylogeny were supported in the 1998 reconstruction, with few exceptions including the monophyletic and basal position of the subfamily Anophelinae, and the monophyly of the tribes Sabethini and Culicini. However, most characters were homoplastic - some extensively - and many relationships were inadequately resolved. Although the Harbach and Kitching [1
] study challenged traditional generic groupings and reinforced the need for reappraisal, it also suggested that robust recovery of generic-level relationships of Culicidae may be difficult with morphological characters alone.
Only four higher-level phylogenies of Culicidae based on gene sequences have been published, each of which were very taxon-limited in scope. All were able to show that, in agreement with the morphological phylogenies, Anopheles
was sister to other sampled genera [4
]. The study by Miller et al. [4
], based on four mosquito species, was inconclusive and the more comprehensive study of Shepard et al. [15
], based on 18S rDNA sequences of 39 species representing nine genera, was unable to resolve deeper relationships. Of the four studies, only Besansky and Fahey [16
] employed a single-copy nuclear protein coding gene to assess relationships. Their study, based on the white
gene, included 13 species representing nine genera. When third codon positions were excluded, Anophelinae was recovered as a basal lineage and Sabethini, Culicini and Aedini were recovered as monophyletic, suggesting the potential of protein-coding sequences for reconstructing generic-level relationships within Culicidae.
The importance of sampling multiple genes when attempting to reconstruct species phylogenies is well recognized [18
]. Mitochondrial DNA and nuclear ribosomal DNA are convenient targets, due to conserved primer binding sequences and ease of amplification based on their typically high copy number. However, both can be problematic for resolving deep phylogenetic relationships. Mitochondrial DNA may exhibit a high mutation rate, skewed base composition, and even symbiont-induced biases [19
]. Beyond base compositional bias, ribosomal DNA also can be exceedingly difficult to align [21
]. Given the increasing availability of completely sequenced mosquito genomes, protein-coding nuclear genes represent a viable alternative as well as a rich and largely untapped resource.
In the study reported here, we explored the phylogenetic utility of six nuclear protein-coding genes: arginine kinase
, CAD, catalase
, and white
. As noted above, only white
was used previously in mosquitoes [16
]. All except catalase
have been used in other insect groups: CAD in bees, empidoid flies, and lacewings, among others [22
]; arginine kinase
in hymenopterans [25
in beetles [27
]; and hunchback
in Hawaiian drosophilids [28
]. We sequenced these six genes from 26 mosquito species representing 25 genera, and two chaoborid outgroup species. In addition, 80 morphological characters were scored from these mosquito and outgroup species. Our goals were twofold: (1) to estimate a generic-level phylogeny of Culicidae based on molecular and morphological evidence, and (2) to use fossils and sequence data to infer divergence times for major culicid lineages.