EUROPEAN JOURNAL OF ENTOMOLOGY EUROPEAN JOURNAL ENTOMOLOGY Discovery of a remarkable new species of Lymanopoda (Lepidoptera: Nymphalidae: Satyrinae) and considerations of its phylogenetic position: An integrative taxonomic approach

. A new species of Lymanopoda Westwood, a cloud forest Neotropical genus of Satyrinae, is described from the páramo grasslands on an isolated, peripheral massif in the Colombian Central Cordillera of the Andes: L. ﬂ ammigera Pyrcz, Prieto & Boyer, sp. n. The genus Lymanopoda is species-rich (approx. 65 species) and its alpha taxonomy is relatively well researched. Relationships within the genus using molecular data have also been explored. The new species is outstanding for its golden yellow colour in males, not found in any other neotropical Satyrinae. Cladograms were constructed based on COI sequences of 47 species of Lymanopoda (~ 70% of the known species) including 17 from Colombia. The new species segregates in the “ tolima ” clade, which comprises four other high altitude Colombian species, as well as two from Ecuador. However, it is the comparative analysis of male genitalia, in particular the superuncus and valvae, which identi ﬁ ed its closest relatives, thus con ﬁ rming that genital characters can help re ﬁ ne molecular phylogenies. In addition to identifying species using mitochondrial DNA (mtDNA bar-codes), nucleotide sites with unique ﬁ xed states used to identify nine species of Lymanopoda from Colombia are also presented.

ing (CCDB), Ontario, Canada, using standard high throughput protocols (Ivanova et al., 2006;deWaard et al., 2008). PCR amplifi cation with a single pair of primers consistently recovered a 658 bp region near the 5' end of COI that included the standard 648 bp barcode region for the animal kingdom (Hebert et al., 2004). Complete specimen data including images, voucher deposition, GenBank accession numbers, GPS coordinates, sequences and trace fi les are accessible in the Barcode of Life Data System (BOLD) ( Table 3).
Sequence divergences for the barcode region were quantifi ed using the Kimura 2 Parameter model, employing the analytical tools in BOLD (BOLD alignment, pairwise deletion). This was done to determine whether there is a barcode gap (a break in the distribution among genetic distances of specimens belonging to the same species and those of specimens from different species), that would allow the identifi cation of the specimens examined. Genetic distances between species are reported as minimum pairwise distances, while intraspecifi c variation is reported as mean and maximum pairwise distances.
Several quantitative species delimitation algorithms for molecular data have been developed over the past decade, including approaches dedicated to DNA barcodes such as Automatic Barcode Gap Discovery (ABGD) and Refi ned Single Linkage (RESL) Analysis algorithm (Puillandre et al., 2012;Ratnasingham & Hebert, 2013). Each specimen with a sequence longer than 500bp automatically gains a BIN (Barcode Index Number) assignment on BOLD that is based on the RESL algorithm (Ratnasingham & Hebert, 2013). BINs may be merged when genetically intermediate specimens are added, or split when new records reveal a clear sequence divergence structure. Distance-based neighbour-joining (NJ) was used to reconstruct DNA barcode gene trees. Despite certain limitations, NJ has repeatedly been shown to perform well for species identifi cation (Huelsenbeck & Hillis, 1993;Kumar & Gadagkar, 2000;Mihaescu et al., 2009;Mutanen et al., 2016).

Phylogenetic relationships
A reconstruction of the phylogenetic relationships of species of Lymanopoda was done using the Maximun Likelihood (ML) method. Two species of Satyrinae were used as an outgroup: Corades chelonis and Lasiophila zapatoza. The analysis was done using the Phylogeny.fr platform (Dereeper et al., 2008(Dereeper et al., , 2010 and sequences were aligned using MUSCLE (v3.8.31) and confi gured for highest accuracy (MUSCLE with default settings).
The phylogenetic tree was constructed using the ML method implemented in the PhyML program (v3.1/3.0 aLRT). The HKY85 substitution model was selected assuming an estimated proportion of invariant sites (of 0.601) and four gamma-distributed rate categories to account for the percentage heterogeneity across sites. The gamma shape parameter was estimated directly from the data (gamma = 1.147). The reliability of internal branches was assessed using the aLRT test (SH-Like). The graphical representation and editing of the phylogenetic tree were done using TreeDyn (v198.3). Lymanopoda have conspicuous white, blue, and reddish ground colours, which may be marked with white or green patches. The underlying evolutionary rationale for this is still unknown but there is growing evidence that some kind of mimicry is involved (Pyrcz, in prep.). Yet, even in Lymanopoda the discovery of a species with shiny golden yellow males was extremely surprising as this kind of colouration is not only unique for the genus but also among all worldwide Satyrinae. Here we investigate its affi nities within the genus Lymanopoda and address some questions about the adaptive role of its colour pattern.

Morphological studies
Most of the material used in this study was obtained during fi eld-work by C. Prieto and P. Boyer in Colombia. Specimens used for morphological studies were examined in the Nature Education Centre (formerly Zoological Museum) of the Jagiellonian University in Kraków (CEP-MZUJ). Types and additional specimens were examined in major public museums including Instituto de Ciencias Naturales de la Universidad Nacional, Bogotá, Colombia (ICN), the Natural History Museum, London, UK (NHMUK), Museo de Agronomía de la Universidad Central, Maracay, Venezuela (MIZA), Staatliches Museum für Tierkunde, Dresden, Germany (MTD) and Zoologische Museum, Humboldt Universität, Berlin, Germany (ZMHB), as well as in the collections of Pierre Boyer (PB) and Carlos Prieto (RCCP).
The terminal parts of the abdomens (including the genitalia) were removed from the specimens and soaked in 10% KOH solution for 5-10 min. Subsequently, abdomens were preliminarily cleaned using soft tissue in water in order to expose genital parts. Water was removed from dissected genitalia using 90% and 95% solutions of ethanol. A Nikon digital camera DS-Fi1 and an Olympus SZX9 stereomicroscope were used for taking pictures of the dissections, which were then processed in Adobe PhotoShop 7.0 CE and Corel PHOTO-PAINT X3 programs to enhance focus and improve quality. The dissected genitalia were kept in glycerol in vials pinned under the corresponding specimens. Genital terminology largely follows Klots (1956). Adults were photographed using a Minolta E-500 digital camera. Colour plates were composed using Adobe PhotoShop version 8. The following abbreviations are used in the text: FW -forewing; HW -hindwing; D -dorsum; V -venter; HDP -hindwing dorsal median patch.

Material and sampling area
Partial nucleotide sequences of mtDNA cytochrome c oxidase subunit I gene (COI) of individuals from several populations occurring in the Andes in Colombia that were previously identifi ed morphologically, were analyzed. Tissue samples were extracted from identifi ed pinned specimens collected in the past 10 years, as it is less likely that sequence data can be obtained from old material. Altogether 79 specimens, representing 47 species, yielded a DNA sequence of over 400 base pairs (bp) in length. Specimens with shorter sequences were excluded from the analyses.

Molecular delimitation of species and barcodes
For the DNA analyses, 79 individuals of 47 species of Lymanopoda were included as well as 2 individuals of two different genera as an outgroup, Corades chelonis Hewitson and Lasiophila zapatoza (Westwood). One or two legs were removed from each dried specimen and stored in individual tubes. DNA extraction, amplifi cation and sequencing of the barcode region of the COI gene were carried out at the Canadian Centre for DNA Barcod- This species has the size, wing shape, and wing pattern similar to L. huilana Wreymer, 1911 andL. tolima Weymer, 1890 (depicted in Fig. 5), but males differ from both these species and from any other congener by the golden yellow colour of their upper and undersides. The females are whitish and thus nearly inseparable from the most closely related species, which are, however, not sympatric.

Description
Male. (Figs 1.1-1.4) Head: Eyes chestnut covered with long, black hairs; labial palps two and a half the length of head, covered with yellow and black hairs, dorsally also brown scales; frons with a tuft of brown hair; antennae reaching half length of the costa, chestnut with white scales at the base of each fl agellomere, club composed of 10 segments, strongly fl attened and dilated, brown, dorsally slightly lighter with a median groove. Thorax: Dorsally black, mostly naked, with some long but sparse silver hairs, tegulae covered with long, golden brown hairs; ventrally black but covered with long and dense yellow and white hairs; femora of second and third pair of legs black, with fi rst pair and tibiae and tarsi yellow, densely covered with scales. Wings: FW (length: 20-21 mm) triangular with a pointed apex, straight outer margin and shallow tornus; HW oval with a rounded apex and straight outer margin from vein M2 to tornus where bent nearly at a right angle, anal margin straight. FWD yellow of variable shade, between pale yellow (in older individuals) and golden yellow from basal to postmedian area, except for a greyish basal suffusion and an elongated patch in subapical area; distally dark brown with sharp basal notches along the discal cell and vein Cu2A, a dark brown ocellus in space Cu1A-Cu2A. HWD varying between pale yellow and golden yellow with a greyish basal and medial suffusion and with a series of minute, sub marginal black dots (and in some specimens more or less developed marginal dark patches between tornus and apex). FWV colour pattern similar to that on the upper side, but the yellow basal area invariably lighter, and all the dark brown elements are dull and barely visible except for the darker patch in the postdiscal area. HWV light orange almost lacking a pattern except for a lighter, elongated patch in discal cell and a darker brown area immediately behind discal cell; sub marginal tiny black spots as on the upper side. Abdomen: Black dorsally and laterally (covered with dense, velvet black hairs and scales), ventrally with sandy yellow scales and hairs. Genitalia (Figs 2.1-2.4, 3.6): Tegumen strongly sclerotized with a slightly bulged dorsal surface; superuncus prominent, reaching half length of the uncus, bifurcated; uncus stout with a sharp tip pointing downwards; gnathos reduced, blunt; subscaphium small and weakly sclerotized;  appendix angular, stout but short with a sharp tip; valva elongated, wide in basal half, narrower in the middle, ends with a wide serrated apex and a prominent processus pointing upwards; saccus short and fl attened dorso-ventrally; aede agus simple, tubular, the valva + saccus very slightly arched, with a smooth surface.
Female. (Figs 1.5, 1.6) Sexual dichroism prominent; yellow is replaced by white pigmentation, however the dark brown-blackish elements of the colour pattern are nearly identical, except that they are slightly larger on the FWD, entering more deeply into the discal cell. The HWD sub marginal black dots are also larger. Otherwise, the wing shape of the female differs slightly in being less elongated, especially the hindwings (FW length: 21 mm). Female genitalia not examined.
Molecular characterization. No intraspecifi c haplotype diversity was found in the available sequences (n = 5). The lowest overall mean distance to another member of the genus is 5.3% to L. tolima from Nevado del Ruiz. BIN number: BOLD: ADD7260. Diagnostic fi xed states and their position in the COI barcode sequence are depicted in Table 2.    Etymology. The specifi c epithet "fl ammigera" is the nominative feminine singular of "fl ammiger" from the latin "fl amma" (= fl ame) and -iger (gero) (= to carry, to bear), in reference to the intense orange-yellow colour of the males of this butterfl y.

Bionomics.
Males patrol at 1-2 m above the ground in the cloud forest -páramo ecotone. Males patrol around midday along the sunny edges of paths. The immature stages and larval food plants are unknown but are presumed to be Chusquea bamboo, as is the case with other species of Lymanopoda, common in the collecting area.
Distribution. This species is known only from the type locality, Páramo de Las Domínguez (Pan de Azucar in some maps), an isolated massif situated west of the main Colombian Central Cordillera range of the Andes. It occurs in the páramo grassland at 3300-3600 m, just above timberline. Adults were collected in January and July.

Species delimitation based on barcode analysis
A NJ tree was generated for 47 species and 79 individuals of Lymanopoda. When discrepancies between the DNA-based and standard taxonomy were found, the specimen was examined to confi rm that its morphological identifi cation was correct, and the alignment and trace fi les were carefully re-examined. It was found that 47 morphospecies were assigned to 44 BINs (Fig. 4), therefore showing a 94% of congruence between morphospecies and BINs. The morphospecies L. hyagnis, L. umbratilis and L. shefteli were placed in the NJ tree under the same BIN code due to the low genetic divergence of 0.39% between L. hyagnis and L. umbratilis, and 1.9% between L. shefteli and L. hyagnis. Similarly, the genetic divergence between the the morphologically very divergent L. caeruleata and L. caucana is 0.77% (Table 1). However, in all the cases, the identifi cation of an unknown specimen by matching its sequence to those in the reference library led to correct results. Nucleotide sites with unique fi xed states that were used to identify nine species (those represented by at least three specimens in our dataset) of Lymanopoda from Colombia are compared in Table 2.
The species of Lymanopoda examined have a mean intraspecifi c genetic distance of 0.05% (n = 78 comparisons of barcodes > 600 bp). Maximum intraspecifi c divergence was 0.77%. The mean interspecifi c genetic distance was 9.60% (n = 2848 comparisons of barcodes > 600 bp). Maximum interspecifi c divergence was 13.37% and minimum interspecifi c distance was 0.39%.

Lymanopoda phylogeny
A phylogenetic tree was constructed using the ML method for 47 species of Lymanopoda, including 17 from Colombia and 30 others whose COI sequences were available in GenBank (Fig. 5), out of ~ 65 known, which makes up 70% of all known species. The tree presents four main clades, one of which is called here for convenience "obsoleta" with 14 species including the two species, L. fl orenciaensis Salazar, Henao &Vargas, 2004 andL. maletera Adams &Bernard, 1979, not sequenced before, the "ionius" clade with 17 species, the "caucana" clade with fi ve species and "tolima" clade with eight species. The latter is subdivided into two clades, one of which includes two species not sequenced before, L. nevada Krüger, 1924 andL. paramera Adams &Bernard, 1979, whereas the other contains six species including L. fl ammigera sp. n. and two other species not included in the generic phylogeny produced previously (Casner & Pyrcz, 2010), L. tolima and L. casneri Pyrcz & Clavijo, 2016, the latter, however, sequenced by Marín et al. (2017). The resolution of this clade is low and presents a polytomy, therefore the position of the new species relative to other fi ve species is not established.

Colour patterns
The new species is remarkable fi rst of all because of its unusual golden-yellow colour of males, unique not only among other congeners but also within the entire speciesrich subtribe Pronophilina (over 650 species), and arguably among all neotropical and even worldwide Satyrinae. The evolutionary basis of this outstanding colouration is unknown but the hypothesis that this colouring is somehow related to mimicry, seems unlikely. This is because the Sulphur Colias dimera Doubleday, 1847, which is the potential model, although generally very common in the Colombian páramos and probably obnoxious, has not been detected in the region where L. fl ammigera sp. n. occurs. Other related Colombian species, such as L. huilana, L. tolima, L. zebra Rodríguez, 2007, L. casneri andL. melia Weymer, 1911, are predominantly white or black and white, which is certainly associated with thermoregulation and the limited solar radiation at high altitudes, and the higher absorption of UV. It could eventually also prove to be the case for L. fl ammigera although the optical qualities of its wing pigments and scales have not been investigated so far. It is worth pointing out that there are several similarly pigmented species of skippers (Hesperiidae) in the high tropical and temperate Andes within the genera Zalomes Bell, Wahydra Steinhauser, Hylephila Billberg, and one yet undescribed species of Racta Evans.
It is however puzzling why such unusual colour patterns evolved in just one isolated area whereas throughout the northern and central Andes most páramo species of Lymanopoda are predominantly white. On the other hand, it is true that the genus Lymanopoda is particularly plastic phenotypically and a number of species occurring in cloud forests or the forest-páramo ecotone have colour patterns that are unusual for the subfamily Satyrinae, for example, the blue L. hazelana Brown, 1943, L. samius Westwood, 1851and L. cinna Westwood, 1889, green patched L. marianna Staudinger, 1897or red L. inaudita Pyrcz, 2010. Some of these colour patterns are almost certainly due to mimetic relationships, an issue currently being investigated (Pyrcz, in prep.).

Barcoding
This study provides an initial assessment of the usefulness of DNA barcoding in Lymanopoda. The NJ tree analysis yielded high percentage of correct identifi cations in the genus Lymanopoda. In the tree, 94% of the morphospecies used in this study formed distinct clades and were assigned a Barcode Index Number (BIN) matching perfectly the morphology based identifi cations. In 6% of the cases, more than one morphospecies shares a BIN number with other species. These cases include fi ve species in this study: L. hyagnis, L. umbratilis, L. shefteli cluster together and have the same BIN number; and L. caeruleata and L. caucana also have the same BIN number. The former three species belong to a complex group of morphologically similar taxa occurring allopatrically in parallel valleys in the Madre de Dios upper basin in southern Peru and northern Bolivia, whose relationships are still not fully understood, and their separate specifi c status is yet to be confi rmed by more thorough taxonomic studies involving their spatial, geographic and altitudinal distribution patterns. L. caeruleata and L. caucana are allopatric species, morphologically easily separable by their predominantly blue (L. caeruleata) and brown (L. caucana) wing colour patterns and genital characters, so their separate specifi c status is strongly supported. Our results confi rm that DNA barcoding is a highly effi cient method for identifying species in the subfamily Satyrinae, as pointed out in another recent study on high Andean butterfl ies (Marín et al., 2017).

Phylogeny
The cladogram based on the COI marker produced for 47 species has to be considered as complementary relative to previous studies as it takes into consideration only one marker, compared to 40 species and 5 molecular markers (Casner & Pyrcz, 2010). We, however, chose to use only the COI marker because one of the key issues of this study was to investigate the robustness of barcoding relative to morphological traits in evaluating relationships within the genus Lymanopoda, in particular, between hypothetically closely related taxa. It is interesting, from this perspective, to point out that, regarding the subdivision of the genus into main monophyletic groups and, in particular, the basal position of the "caucana" clade comprising 5 species, the results are highly congruent with previous molecular (Casner  and morphological phylogenetic hypotheses (Pyrcz, 2001). The position of L. prusia, Heimlich, 1973, as a sister to the remaining species of Lymanopoda is, however, not confi rmed.
The "tolima clade", with 6 species in Casner & Pyrcz's paper, is here restricted to 4 species, two of which were not previously examined, L. huilana and L. fl ammigera sp. n. This well supported clade includes all the high altitude páramo species, examined so far, distributed from northcentral Colombia (Belmira) to Ecuador. Also, all of these species share a number of morphological synapomorphies, which support its monophyly. In this respect, the Peruvian species, L. inde Pyrcz, 2004 andL. eubagioides Butler, 1873, excluded from this clade, stand apart, and their position within this clade suggested originally by Casner & Pyrcz (op. cit.) should be reconsidered. Importantly, two white páramo species, L. nevada and L. paramera, found in isolated ranges in northern Colombia, were included in the molecular analysis for the fi rst time. Although they superfi cially resemble the species in the "tolima" clade by being predominantly white, they were placed in a separate clade, even if they still occur in the larger unit comprising the "tolima" clade and not in the other two large clades, "excisa" and "obsoleta".
By combining molecular and morphological data it is possible to determine the closest relatives of L. fl ammigera sp. n. within the "tolima" clade. COI based analysis is inconclusive in this respect in showing a polytomy. Comparisons of male genitalia show, however, that L. fl ammigera, L. casneri and L. tolima share a unique synapomorphy, a bifurcate, dorso-ventrally fl attened, prominent rounded superuncus. In L. huilana and L. hazelana the superuncus is considerably smaller and not bifurcated even if two lateral lobes are noticeable. Other characters are less evident, although the valvae of L. casneri and L. tolima are more similar, being short with a single prominent apical tooth, whereas the valvae of L. fl ammigera sp. n. are narrower in the middle and much longer, looking in this respect more like those of L. huilana. In L. melia, the sister species of L. tolima according to Casner & Pyrcz's (op. cit.) phylogeny, the superuncus is short and single. These data have important phylogeographical implications. L. tolima diverged in the fi rst place from L. huilana even though there is a continuity of páramo habitats between the areas in Quindío, Tolima and Valle del Cauca in the Central Cordillera with those in Cauca and Nariño further south where L. huilana occurs. On the other hand, there are currently no appropriate páramo habitats over 200 km between Quindio and the Páramo de Belmira in Antioquia where L. casneri is found. Apparently some more complex underlying paleoecological processes have resulted in the shaping of present day distributions of páramo Lymanopoda species in this part of Colombia.

Final considerations
This study highlights two important facts. It is confi rmed that genitalia, in particular those of males, are extremely valuable not only in alpha-taxonomy but also phylogenetically. Here a comparative analysis refi nes some data obtained using molecular tools. Of course, not in all taxa are male genitalia as informative, which depends mostly on the number of modifi cations leading to the evolution of noticeable phenotypical traits even in closely related groups of taxa. In the genus Lymanopoda such traits are appreciable. Secondly, our study confi rms the usefulness of the COI marker in species defi nition as well as in phylogenetic considerations, a role that has been questioned. Here, COI support data on 17 species of Colombian Lymanopoda helped refi ne the phylogeny of the genus, and is congruent in most aspects with the previously proposed arrangement based on fi ve markers. In other words, COI does work at least in the genus Lymano poda, even if in some other taxa of Lepidoptera this may not necessarily be the case. This study expanded our knowledge on the evolution of the genus Lymanopoda by adding seven more species to its phylogeny. Data for several key species are however still missing, in Colombia in particular for L. mirabilis (Staudinger), a high páramo species with unusual extremely elongated wings and atypical genitalia from the southern part of the Cordillera Oriental, and L. melendeza Adams from the Sierra del Cocuy that has some resemblance in both genitalia and colour patterns to the Venezuelan L. marianna Staudinger, known so far only from the holotype.

ACKNOWLEDGEMENTS.
For help and support in many ways we acknowledge A. Hausmann (Bavarian State Collection of Zoology, Munich, Germany) and C. Uribe Ortega and L.C. Gutierrez (Universidad del Atlántico, Barranquilla, Colombia). This research was supported by funds from the Georg Forster Research Fellowship Program of the Alexander-von Humboldt Foundation (Bonn), the Federal Ministry for Education and Research (Germany), the Vice-Rectorate for Research of the Universidad del Atlántico, Colombia under decision number 3247 of 12th June 2015, and by an internal grant of the Institute of Zoology and Biomedical Research of the Jagiellonian University, KZDS006320.