DNA barcodes of closely related (but morphologically and ecologically distinct) species of skipper butterflies (Hesperiidae) can differ by only one to three nucleotides

Unlike most species of Lepidoptera whose DNA barcodes have been examined, closely related taxa in each of three pairs of hesperiids (Polyctor cleta and P. polyctor, Cobalus virbius and C. fidicula, Neoxeniades luda and N. pluviasilva Burns, new species) seem indistinguishable by their barcodes; but that is when some of the cytochrome c oxidase I (COI) sequences are short and sample sizes are small. These skipper butterflies are unquestionably distinct species, as evidenced by genitalic and facies differences and by ecologic segregation, i.e., one species of each pair in dry forest, the other in adjacent rain forest in Area de Conservación Guanacaste in northwestern Costa Rica. This national park is the source of the specimens used in this study, all of which were reared. Larval foodplants are of no or problematic value in distinguishing these species. Large samples of individuals whose barcodes are acceptably long reveal slight interspecific differentiation (involving just one to three nucleotides) in all three pairs of skippers. Clearly, the chronic practice of various taxonomists of setting arbitrary levels of differentiation for delimiting species is unrealistic. Additional key words: Area de Conservación Guanacaste, Costa Rica, dry forest, rain forest, foodplants, genitalia, Neoxeniades pluviasilva Burns, n. sp. A DNA barcode is the base pair (bp) sequence of a short (~650 bp), standard segment of the genome (Hebert et al. 2003). In animals, this is part of the mitochondrial gene cytochrome c oxidase I (COI). Because the COI gene generally mutates at evolutionarily rapid rates, comparison of barcodes in a sample of individuals best reveals differentiation at low taxonomic levels. Hence barcodes can be extremely useful in distinguishing and identifying species. Coupling this concept with the idea of always comparing the same short length of COI across a wide diversity of taxonomic groups—and doing so with demonstrable success—is what led to the catchy name “DNA barcodes” (Hebert et al. 2003). Even though COI had been used effectively in various evolutionary and taxonomic studies at and around the species level well before this epithet appeared, in the few years since its introduction, barcodes have been used for their specific purpose with notable results and with rapidly increasing frequency. The rearing of myriad wild-caught caterpillars in Area de Conservación Guanacaste (ACG) in northwestern Costa Rica is now approaching its thirtieth year (for information about both site and rearing process, see Miller et al. 2006, Janzen & Hallwachs 2006, Burns & Janzen 2001). DNA barcodes (of a total of 4,260 reared adults) have been able to distinguish among almost 98% of 521 previously known species of the lepidopteran families Hesperiidae (skipper butterflies), Sphingidae (sphinx moths), and Saturniidae (wild silk moths) VOLUME 61, NUMBER 3 139 (Hajibabaei et al. 2006, Janzen et al. 2005). Rare cases where barcodes failed, which always involved closely related congeners, are worth examining in more detail. In this paper we treat three such pairs of congeneric skipper species (noted in Hajibabaei et al. 2006:table 1). We map the ecologic separation of the species in each pair in and very near ACG. We document the species status of each member of a pair (and describe one as new) on morphologic grounds. We discuss the various degrees to which larval diets, although specialized, are unreliable for species discrimination. And we show, by scrutinizing sequence length and composition, that DNA barcodes separate the species after all. Despite the immense value of DNA barcodes and the fact that they have often indicated overlooked species, it is important to consider characters besides the barcodes themselves—a point made repeatedly in the revelation of 10 cryptic species hiding under the one name Astraptes fulgerator (Walch) in ACG (Hebert et al. 2004). THE SPECIES PAIRS IN QUESTION Ecologic separation (Figs. 1-3). Each pair comprises a dry-forest species and a rainforest species. Parapatry of this kind is a recurrent distribution pattern among closely related lepidopteran species in ACG. In each pair of the following list, the dry-forest species comes first: Polyctor cleta Evans and P. polyctor (Prittwitz), Cobalus virbius (Cramer) and C. fidicula (Hewitson), Neoxeniades luda (Hewitson) and N. pluviasilva Burns (a new species described below). Polyctor is a pyrgine genus, and Cobalus and Neoxeniades are hesperiine genera. From our distribution data, parapatry in both pairs of hesperiine species appears to be complete (Figs. 2, 3) whereas that in the pyrgine pair does not (Fig. 1). Out of 211 reared individuals (wild-caught as caterpillars) of the rainforest species P. polyctor, four came from dry forest. The genitalia of these apparent strays have been KOH-dissected and thoroughly studied to be sure of FIG. 1. Spatial distribution of Polyctor cleta and P. polyctor in and near ACG. 140140 JOURNAL OF THE LEPIDOPTERISTS’ SOCIETY their specific determination. Because both species of this essentially parapatric pair of Polyctor eat the same three species of foodplants (Table 3), and because one of these plants occurs in both rain and dry forest, a female wandering from rain forest can find an attractive foodplant in dry forest and oviposit on it. The flight of these skippers is far stronger than necessary to travel the distance involved. Of the four P. polyctor caterpillars found in dry forest, three were eating the species of foodplant most often eaten by this skipper in rain forest (and because two of those were found on the very same plant, they are probably offspring of a single female); the fourth caterpillar was eating an exceedingly common, but strictly dry-forest, species that is by far the preferred foodplant of P. cleta. A small number of P. cleta caterpillars found in disturbed ecotone between dry and rain forest, and less than 2 km from the latter, were eating the main foodplant of P. polyctor. KOH-dissection and examination of the genitalia of the four adults reared from this group gave no hint of hybridization. Morphologic differences (Figs. 4-41; Tables 1, 2). In all three pairs, the brown ground color of the adult averages paler in the dry-forest species than it does in its rainforest counterpart (Figs. 4–27). This is especially evident when comparing long series of more or less recently reared (therefore unfaded) specimens. In both sexes of Polyctor, a spot spanning the distal end of the forewing cell is hyaline in P. cleta but opaque in P. polyctor. A male secondary sex character in these species of Polyctor comprises a tuft of long hairlike scales arising near the base of the dorsal hindwing costa, as well as an elongate patch of pale specialized scales embraced by the swollen beginning of vein 7 and a similarly swollen, closely adjacent length of vein 6; in both veins, swelling extends out to the end of the cell; and the hairlike scales are long enough to overlie the special patch. These presumably pheromone-disseminating hairs are mostly to entirely dark in P. cleta but pale (often orangish) in P. polyctor (cf. Figs. 4 and 6). FIG. 2. Spatial distribution of Cobalus virbius and C. fidicula in and near ACG. VOLUME 61, NUMBER 3 141 Though clearly variations on a theme, the male genitalia of these two Polyctor species differ in striking ways. Despite substantial individual variation, almost every genitalic part differs interspecifically to at least some extent; but it is the highly asymmetric valvae that differ most (see Table 1 and cf. Figs. 28–33). Even the less elaborate female genitalia are notably distinct in the two species (Table 2). Both species of Cobalus, which are predominantly brown, have a conspicuous white patch both dorsally and ventrally on a distal area of the hindwing. In ACG specimens, this patch is restricted to males of C. fidicula but expressed by both sexes of C. virbius (Figs. 8–11, 20–23), except for two females in which it is barely perceptible. Both species express it more fully ventrally than dorsally. In C. fidicula the patch stops before the outer margin so as to leave a narrow strip of dark brown ground color, ventrally the patch extends from mid space 1c to vein 6, and the white of the patch looks creamy on the ventral surface. In C. virbius the patch reaches the outer margin, ventrally extends from the tornus to vein 6, and looks pure white on both wing surfaces. Lateral orange scaling—broad on the outer side of the palpus and narrow behind the eye—is bright in C. fidicula but just dully suggested, and only on the palpus, in C. virbius. Cobalus fidicula is a little larger than C. virbius, and its forewing hyaline spots are likewise larger. The male genitalia (which are symmetric) differ in two obvious respects. The ventral distal division of the valva is longer and dorsally dentate in C. fidicula (cf. Figs. 35 and 37). The very broad uncus in dorsal view shows a pair of prominent lateral swellings in C. virbius (cf. Figs. 34 and 36). Neoxeniades pluviasilva Burns, new species (Figs. 3, 14, 15, 26, 27, 40, 41, 45, 46; Table 3) Etymology. The species name, a noun in apposition, comes from the Latin pluvia for rain and silva for forest. Diagnosis. This is a rainforest species whereas FIG. 3. Spatial distribution of Neoxeniades luda and N. pluviasilva in and near ACG. 142142 JOURNAL OF THE LEPIDOPTERISTS’ SOCIETY N. luda is a species of the dry forest (Fig. 3). At a glance, N. pluviasilva is darker than N. luda and does not express a large, pale, outer marginal area on the ventral side of the hindwing nearly as well (cf. Figs. 26, 27 with 24, 25). In females of N. pluviasilva, the double hyaline cell spot of the forewing extensively overlaps the spot in space 2 whereas in N. luda females, this forewing cell spot overlaps the spot in space 2 little or not at all (cf. Figs. 15, 27 with 13, 25). Description. Member, with N. luda, of mainly South American N. scipio species complex—t