DNA barcodes distinguish species of tropical Lepidoptera

Abstract

Although central to much biological research, the identification of species is often difficult. The use of DNA barcodes, short DNA sequences from a standardized region of the genome, has recently been proposed as a tool to facilitate species identification and discovery. However, the effectiveness of DNA barcoding for identifying specimens in species-rich tropical biotas is unknown. Here we show that cytochrome c oxidase I DNA barcodes effectively discriminate among species in three Lepidoptera families from Area de Conservación Guanacaste in northwestern Costa Rica. We found that 97.9% of the 521 species recognized by prior taxonomic work possess distinctive cytochrome c oxidase I barcodes and that the few instances of interspecific sequence overlap involve very similar species. We also found two or more barcode clusters within each of 13 supposedly single species. Covariation between these clusters and morphological and/or ecological traits indicates overlooked species complexes. If these results are general, DNA barcoding will significantly aid species identification and discovery in tropical settings.

Fig. 1.

Lepidoptera families used in this study and their barcode sequence statistics. (a) Species used in this study versus the total known for ACG, Costa Rica, and the world (http://janzen.sas.upenn.edu and I. Kitching, personal communication). (b) Comparison of COI barcode variation and measures of conspecific and congeneric genetic distances in the three families. Distances are calculated by using the Kimura two-parameter model of base substitution (23).

Fig. 2.

Number of barcode sequences for each species. The total number of individuals analyzed per family is shown in parentheses.

Fig. 3.

A tree representation of COI barcodes for 4,260 individuals of three Lepidoptera families showing the clear separation of branches leading to individuals of Hesperiidae, Sphingidae, and Saturniidae (red, blue, and green, respectively). The original tree was made by using the NJ algorithm (13) based on Kimura two-parameter distances (23). To make this tree we modified the NJ tree by collapsing all of the branches with zero length, which creates the star-shape clusters in the tree (see Figs. 5–7 for detailed NJ trees for each family).

Fig. 4.

Patterns of COI divergence for the 315 species of Lepidoptera from ACG that were represented by three or more individuals. Minimum between-species divergence (Min-BSD) is plotted against maximum within-species divergence (Max-WSD) for each morphologically defined species. Points above the diagonal represent cases where species identification is straightforward. Species with overlapping COI barcodes (and therefore very low Min-BSD values) are shown as red dots. Species with two or more distinct barcode clusters showing covariation in biological traits are shown as green dots (see text for details). In two species of Hesperiidae, indicated by an arrow, Min-BSD values are slightly (≈10%) smaller than Max-WSD values; but individuals of these species are not confused in the NJ analysis because their clusters do not overlap (13) (see Fig. 5). Distances are calculated by using the Kimura two-parameter model of base substitution (23).