Description of three new species of Automeris Hübner, 1819 from Colombia and Brazil (Lepidoptera, Saturniidae, Hemileucinae)

Abstract The Saturniidae is one of the most emblematic families of moths, comprising nearly 3000 species distributed globally. In this study, DNA barcode analysis and comparative morphology were combined to describe three new species within the genus Automeris, which is the most diverse genus in the family. Automeris llaneros Decaëns, Rougerie & Bonilla, sp. nov., Automeris mineros Decaëns, Rougerie & Bonilla, sp. nov., and Automeris belemensis Decaëns, Rougerie & Bénéluz, sp. nov. are described from the Colombian Orinoco watershed, the Colombian Eastern Cordillera, and the area of endemism of Belém in the Brazilian Amazonia, respectively. They all belong to the Automeris bilinea (Walker, 1855) species subgroup, which comprises a number of species that are sometimes difficult to distinguish from each other using morphology alone. Here, the description of these three new species is based on significant differences from their closest relatives, either in terms of wing patterns, genitalia, DNA barcodes or a combination of these features.


Introduction
The Saturniidae family, popularised as the wild silkmoths, is one of the most emblematic families of moths, because of the giant size, colourful patterns, or tailed hindwings of some of its species. In the latest published checklist (Kitching et al. 2018), as many as 3454 valid species were recognised in eight subfamilies and 180 genera. The diversity of the family is highest in the neotropics, where it is represented by six subfamilies and nearly 2400 species that can be found in a broad range of habitats, including the southmost areas of South America, from sea level to elevations exceeding 4000m. The diversity in habitus of these moths is also extreme: size ranges from small moths a few centimetres in wingspan (e.g., in the genus Hylesia Hübner, 1820) to very large ones approaching a wingspan of 20 cm (e.g., in Arsenura Duncan, 1841); wing patterns can be cryptic, mimicking leaves (e.g., in Copaxa Walker, 1855), can harbour large eyespots (e.g., in Automeris Hübner, 1919), or can be aposematic in colour (e.g., in Citheronia Hübner, 1919); wings can be rounded (e.g., Dirphia Hübner, 1919) or elongated (e.g., in Syssphinx Hübner, 1819 andPtiloscola Michener, 1949), or some taxa have spectacular tailed hindwings (e.g., Copiopteryx Duncan, 1841).
Within subfamily Hemileucinae, the genus Automeris comprises species whose size ranges from small to very large; their main feature is the presence of a large eyespot on the dorsal surface of the hindwings. It is the most diverse genus within the family. In his monograph of the subfamily Hemileucinae, Lemaire (2002) listed 135 species in the genus, which he further organised into nine species groups based on the habitus and the structure of genitalia. Recently, the use of molecular approaches such as DNA barcoding, in addition to morphology, led to a significant increase in the pace of discovery and description of new species. Thus, in the past ten years only, as many as 155 new taxa were described, raising the total number of species in the genus to 313 (Kitching et al. 2018). In this paper, we use a combination of morphological features and molecular data (DNA barcodes) to propose the description of three new species from Colombia and Brazil within the group of Automeris bilinea (Walker, 1855), which was defined by Lemaire (2002) as a subgroup within the larger species group of Automeris illustris (Walker, 1855).

Specimen collecting
Specimens were collected in the following three localities: from July to August 1999 in the savannah landscapes of the Eastern Plains of Colombia (Meta department, TD and DB leg.); in December 2002 in the Boyacá department, Colombia, in an area of humid Andean forest (1500 m in elevation) with moderate level of forest fragmentation (G. Lecourt and DB leg.); and from April to July 2008 in the state of Pará, Brazil, in an area of Amazonian forest with moderate to high levels of forest fragmentation (TD leg.). Moths were attracted by a Mercury Vapour (MV) bulb powered by a small portable generator. A white sheet of 2 m height × 3 m width was used as a reflector. Collecting took place throughout each entire night, i.e., from 18:30 h to 06:30 h, in order to increase the probability of detecting species with different flight behaviours (Lamarre et al. 2015). Moths coming to the sheet were injected with ammonia, stored, and dried in labelled paper envelopes and brought to the lab to be mounted for morphological examination.

DNA barcoding and molecular analyses
DNA was extracted from dry legs removed from dry collection specimens of the suspected new species. We sampled two specimens of Automeris llaneros sp. nov., five specimens of Automeris mineros sp. nov. and 17 specimens of Automeris belemensis sp. nov., and we also included sequences of closely related species obtained from the Barcode of Life Data Systems (BOLD; Ratnasingham and Hebert 2007) and that had been generated as part of the DNA barcoding campaign for saturniid moths, coordinated by RR. Tissue samples were processed at the Canadian Centre for DNA Barcoding (CCDB). DNA was extracted using a routine silica-based 96-well extraction automation protocol (Ivanova et al. 2006). The part of COI used as a 'DNA barcode' (Hebert et al. 2003) was amplified with the primer set LepF1/LepR1 (Hebert et al. 2004), targeting a 658 bp fragment. The DNA extracts that did not amplify for the full-length DNA barcode were re-amplified with the internal primer pairs LepF1/MLepR1 and MLepF1/LepR1, targeting DNA fragments of 307 bp and 407 bp (Hajibabaei et al. 2006), respectively. All PCR amplifications were performed according to the standard PCR reaction protocol used in CCDB (Hajibabaei et al. 2005); PCR products were checked on a 2% E-gel 96 Agarose (Invitrogen, Burlington, ON, Canada). Unpurified PCR fragments were sequenced in both directions using the same primers as for the PCR reaction. The sequencing reactions followed CCDB protocols (http://ccdb.ca/resources/; Hajibabaei et al. 2005). All sequences were aligned and inspected for frame-shifts and stop codons for removal of editing errors and possible pseudogenes.
All records, including specimen and sequence data, and GenBank accession numbers, are given in Appendix 1, and are publicly accessible in the Barcode of Life Data system (BOLD) within dataset DS-AUTONSP (https://doi.org/10.5883/DS-AU-TONSP). An unrooted neighbour joining (NJ) tree was computed on BOLD V4 using p-distances and the BOLD aligner option to compare the sequences obtained from the specimens of the three new species and those of closely related taxa. BOLD was used to calculate uncorrected p-distances between newly described species and their closest relatives. We also used the barcode identification numbers (BINs), i.e., clusters of barcode sequences automatically generated in BOLD which have a high concordance with species, as an additional source of information for species discrimination (Ratnasingham and Hebert 2013).

Distribution maps
We present maps of the current distribution for the three newly described species and their closest species within the A. bilinea subgroup (seven species, Fig. 3). Records of the other species within the A. bilinea subgroup were collected from BOLD and were georeferenced and carefully curated for locality and species identity excluding any possible error (a complete list of specimens and associated data is given in Appendix 1). Paratypes (19 ♂♂ and 6 ♀♀). Colombia • 15 ♂♂ and 5 ♀♀, all same data as holotype; all specimens collected at MV light except one pair ab ovo, reared on Pyracantha regersiana in Rouen (France) by TD, and 3 ♂♂ and 2 ♀♀ ab ovo, reared on Quercus sp. in Bogotá (Colombia) by L.D. Ramirez and DB. Deposited as follow: 2 ♂♂ and 2 ♀ (Fig. 1B, F; allotype; BOLD SampleID: BC-Dec0547) in IAvH, 5 ♂♂ in MNHN (BOLD SampleID: BC-Dec0548, BC-Dec0549, BC-Dec0550), 6 ♂♂ and 1 ♀ in CTD, 1 ♂ in CFB, 4 ♂♂ and 2 ♀♀ in CCGM, 1 ♂ and 1 ♀ in CDB.
Diagnosis. Automeris mineros sp. nov. is similar to the reddish forms of A. midea, a species with a large and essentially Amazonian distribution (Fig. 3). However, the vivid coloration, which is occasional in the later, is consistent among all the specimens of A. mineros sp. nov. that have been examined. It is also possible to separate the two species by additional fine characters of the habitus. In A. mineros sp. nov., the antemedian area of the forewings appears lighter than the median area due to the presence of a dense dusting of yellow scales, while this zone is generally concolourous with or darker than the rest of the wing in A. midea (Fig. 4). The apex of the forewing tends to be less acute in A. mineros sp. nov. The width of the yellow periocellar ring of the hindwing is also wider in A. mineros than in most of the examined specimens of A. midea. Finally, the veins on the ventral side of the forewings are clearly highlighted in orange in A. mineros sp. nov., and are more sharply contrasting with the surrounding ground colour than in A. midea. Distinction based on male genitalia is less conclusive, although we can however note the truncated instead of triangular shape of the saccus, as well as the median plate of the gnathos which is less massive in A. mineros sp. nov. and clearly marked by a median projection. Interestingly, DNA barcodes clearly separate A. mineros sp. nov. from all closely related species in a distinct cluster of sequences (BIN BOLD:ABY4503; see fig. 4). The nearest neighbour is A. belizonensis (1.8% minimum p-distance), from which it can be distinguished by the more rounded shape of the forewings and the more vivid coloration of the hindwings.
Description. ♂ (Fig. 1A, E). Wingspan: 77-84 mm. Head: dark brown, labial palpi and antennae orange brown. Thorax: dorsally dark brown with red orange piliform scales on the ventral side; legs dark brown. Abdomen: dark orange brown with dark brown piliform scales on the dorsal side; eighth abdominal segment lacking any remarkable sclerotised structure. Forewings: length 40-42 mm, slightly elongated, rounded apex, straight outer margin; dorsal ground colour orange brown, suffused with yellow scales in the ante-and postmedian areas and, to a lesser extent, by pink scales in the median area; antemedial line faint, almost indistinct, only visible as yellow scales bordering its distal edge; postmedial line barely preapical (1-3 mm), slightly convex from apex to vein CuA2, then bent toward the anal margin, yellow in colour, lined distally by a line of black scales; discocellular mark rectangular, darker than the surrounding wing surface, with a dark brown spot in its centre, and three to four small spots of the same colour at its corners. Ventral side with a large dull orange area extending on the main basomedian area, with veins marked in distinctive orange scales; apical area dark brown, extending along outer margin and toward tornus. Postmedial line  well marked, dark reddish brown; marginal band diffuse, suffused with yellow scales and disappearing toward apex. Discocellular mark large and black, with a white discal spot in its centre and surrounded by a thin diffuse ring of yellow scales. Hindwings: basomedian area vivid orange-red with a 7-12 mm × 6-8 mm eyespot in its centre formed by, from its centre: a small well marked white pupil surrounded by a dark brown iris, a first large black periocellar ring, a second large yellow ring of the same width, and finally a barely visible line of black scales enclosing the eyespot. Postmedial line black and lunular, distally bordered by a thin line of yellow scales, and proximally by another thin line of yellow and black scales; postmedian area formed by a thin vivid orange-red band and bordered by a large orange-brown marginal band covered with yellow scales. Ventral side light brown, suffused with yellow scales, particularly on the marginal band; venation distally marked with orange scales. Postmedial line dark brown, becoming faint toward the anal margin; discal cell with a small white spot.
Wingspan ♀ (Fig. 1B, F): 94 mm. Head, thorax, and abdomen of the same colour as in the male. Forewings: length 47 mm, elongated, rounded apex, almost straight border; dorsal ground colour dark purplish brown, suffused with light grey scales in the median area, except for a large oblique band lacking these scales approximately half way between the dark rectangular discal mark and the apex; postmedian area suffused with yellow scales. Both ante-and postmedial lines yellow, the latter straight and slightly preapical (3 mm). Ventral side light orange brown, the venation marked by orange scales, marginal band suffused with orange scales; postmedial line dark brown; large black discocellular mark, with a large white discal spot in its centre. Hindwings: Basomedian area brownish red; eyespot with the same structure as in the male, slightly duller in colour; black and lunular postmedial line, bordered by a line of yellow scales on both edges; postmedian area brownish red; marginal band covered with yellow scales. Ventral side light brown, suffused with yellow to pink scales; venation distally marked by orange scales; discal point small and white.
Genitalia ♂ ( Fig. 2A): typical of the A. bilinea subgroup as described in Lemaire (2002). Uncus well developed, slightly extending beyond the distal end of the valves, with a broad bifid dorsal protuberance. Dorsal lobes of valves weakly developed and sharp. Median plate of gnathos strongly sclerotised with its posterior margin concave and with a small median projection. Saccus short and anteriorly truncated. Phallus straight, with a small lateral spine on its base; its posterior, tapering end slightly bent upward, with the weakly developed vesica expanding ventrally.
Genitalia ♀: not examined. Immature stages. Eggs were obtained from a wild collected female. Larvae hatched 22 days after and readily fed on Pyracantha regersiana (Rosaceae) in France (rearing #17 by TD) and on Quercus sp. (Fagaceae) in Colombia (rearing by L. D. Ramirez and DB). Native foodplants remain unknown. Rearing was successful in plastic boxes, feeding larvae with fresh branches changed every 2-4 days. Larvae completed six instars within two months on P. regersiana and pupated in a brown cocoon.
Eggs are white with a black micropyle, laterally flattened, 2 mm diameter × 0.8 mm height, laid in dense cluster of several dozens. First larval instar: head black. Body 4 mm upon hatching, 6 mm maximal length; pale yellow with black scoli and spines. Second instar: Head black. Body 7-8 mm maximal length; brownish yellow dorsally, dull yellow ventrally; scoli and spines dark brown. Third instar: Head black. Body: 14 mm maximal length; brownish yellow dorsally, green yellow ventrally; scoli and spines dark brown. Fourth instar: Head black. Body 19-20 mm maximal length; dark brown dorsally with fine light green stripes, light green ventrally; scoli and spines dark brown to black. Fifth instar: Head green. Body: 35-40 mm maximal length; light green colour with pink dorsal ornamentation, a lateral ivory strip ventrally and dorsally bordered with a thin black line; scoli and spines light green. Sixth instar: Same colour and ornamentation as previous instar; 35-40 mm maximal length. Pupa and cocoon: Last instar larvae spin a thin and supple cocoon of beige silk. Pupa 24-37 mm long, dark brown. Reared adults emerged from the cocoon early in the morning one to two months after pupation.
Distribution. Automeris mineros sp. nov. is known form the type locality only, in the Oriental Cordillera of Colombia near Muzo (Fig. 3), a region from which a number of new taxa of Saturniidae were described recently (Decaëns and Rougerie 2008).
Etymology. This species is named in reference to emerald mining, which represents an emblematic economic activity in the region surrounding the type locality. Diagnosis. Phenotypically, Automeris belemensis sp. nov. is closely related to A. cinctistriga and A. godartii from which it is difficult to separate based on wing patterns, particularly if we consider the phenotypic variability that characterises these species (see Fig.  3 for distribution map of these species). However, the examination of a large number of specimens reveals subtle characters that make it possible to differentiate them. The distinction from A. cinctistriga is possible because of the general shape of the forewings in A. belemensis sp. nov. usually less elongated with a less prominent apex, and by the ornamentation of the eyespot of the hindwings, whose yellow ring is thinner, and highlighted externally by the presence of a line of black scales which is usually lacking in A. cinctistriga. The distinction from A. godartii is more difficult, but the general colouration of the wings is, however, different in the 22 specimens examined of the two species. In A. belemensis sp. nov., the ground colour of the forewings is a lighter orange-brown than the grey-brown ground colour that is characteristic of A. godartii. The dusting of silver scales generally present in A. godartii, in particular in the preapical triangle of the forewings, is weakly marked or completely absent in A. belemensis sp. nov. The ante-and postmedial lines are also lighter in A. belemensis, standing out in a more contrasted way compared to the surrounding wing colour. Finally, the line of black scales surrounding the eyespot of the hindwings, and sometimes even the yellow periocellar ring, which are continuous in A. godartii, are frequently interrupted towards the subcostal area in A. belemensis sp. nov.

Automeris belemensis
It is likely that the male of A. godartii figured in Lemaire (2002: plate 40, fig. 4), originating from the state of Pará in Brazil, actually belongs to A. belemensis sp. nov.
With a long posteriorly produced uncus, male genitalia are similar to those of A. godartii, but also to those of A. lemensis, which is known only from the Gran Sabana region in southern Venezuela (Fig. 3). Interestingly, the DNA barcodes of the 11 sequenced specimens of A. belemensis sp. nov. form a distinct cluster in the NJ tree (BIN BOLD:AAA5242) with A. lemensis as the nearest neighbour with 2% minimum uncorrected p-distance (Fig. 4). However, both species can easily be distinguished by the ground coloration of their wings, which is much duller in the former, especially in the periocellar area of the hindwings, and an otherwise much rounder hindwing eyespot.
Description. ♂ (Fig. 1C, G). Wingspan: 70-71 mm. Head: dark brown, labial palpi and antennae orange brown. Thorax: dorsally dark brown and ventrally orange brown; legs dark brown. Abdomen: orange brown. Forewings: Length 35-37 mm, slightly elongated, slightly prominent apex, straight outer border; dorsal ground colour orange brown; ante-and postmedial lines thin and yellow, the latter slightly convex, becoming barely visible upon reaching the costal margin 3-4 mm from the apex; discocellular mark rectangular, darker than the surrounding wing surface, with four to six dark brown spots at its angles. Ventral side yellow brown, slightly darker on the outer margin, with the postmedial line underlined by dark brown scales, and a large black discocellular spot, marked in its centre by a small white discal dot. Hindwings: basomedian area dull orange with a large eyespot in its centre formed by, from its centre: a small black and white pupil, a large grey brown iris surrounded by a broad black periocellar ring, followed by a thin yellow ring, and finally by an external line of black scales. Postmedial line lunular, formed by a thin yellow line bordered proximally and distally by a wider black line; postmedian area dull orange; marginal band orange brown. Ventral side uniformly yellow brown, with a straight, weakly marked postmedial line, a faint zigzagging premarginal line, and with a small white discal spot.
Female unknown. Genitalia ♂ (Fig. 2B): similar general structure as in other species of the bilinea subgroup. Uncus well developed, long and bent downwards, apically barely bifid. Dorsal lobes of valves well developed and broadly triangular. Median plate of gnathos strongly sclerotised with its posterior margin convex, a developed median projection, and long lateral appendages. Saccus well developed, triangular and acute anteriorly. Phallus straight, with a small lateral spine on its base; its posterior end slightly bent upward, with a weakly developed vesica.
Distribution. Automeris belemensis sp. nov. is known from the lower Amazonian watershed in the Brazilian states of Pará and Maranhão, Brazil (Fig. 3). It has not been found despite of extensive collecting efforts in neighbouring region of French Guiana, and is also unknown from other areas of Brazilian Amazonia. This suggests it could be restricted to the Belém area of endemism as defined in Cardoso da Silva et al. (2005), which was recently highlighted as a hotspot for the diversity of so far undescribed moth species (Lamarre et al. 2016).
Etymology. Automeris belemensis sp. nov. is named as a reference to the area of endemism of Belém where this species has been found and to which it is likely endemic.
Diagnosis. Automeris llaneros sp. nov. is phenotypically very similar to A. cinctistriga, from which it is quite difficult to distinguish based on wing patterns alone. However, the two known specimens of A. llaneros sp. nov. have less elongated forewings with less acute apices than most of the examined specimens of A. cinctistriga. The background colour of the forewing is also duller in A. llaneros sp. nov., less orange, and the ante-and postmedian lines are finer, beige instead of yellow, and contrasting much less markedly with the general colour of the wings. Finally, the distance between the ante-and postmedian lines at the point where they join the anal edge of the forewings seems greater in A. llaneros sp. nov. (1 cm in the two known specimens) than in A. cinctistriga (4-7 mm). The DNA barcodes of A. llaneros sp. nov. are assigned to a different BIN than those of A. cinctistriga (see discussion), and the two species are very clearly separated in the DNA barcode tree, bringing additional support to their treatment as two distinct species.
DNA barcodes place A. llaneros sp. nov. near A. belizonensis, A. mineros sp. nov., and A. fieldi on the NJ tree (Fig. 4). This proximity seems to be confirmed by the examination of the male genitalia, whose bifid protuberance of the uncus is strongly developed, exceeding the tip of the valves, as in A. mineros sp. nov. (described here), A. belizonensis, and A. fieldi (Lemaire 1971;Brechlin and Meister 2014). In comparison, A. cinctistriga generally presents a less developed and less deeply indented uncus (Lemaire 1971). The new species is also easily distinguished from these close relatives based on its wing shape and patterns. For instance, A. belizonensis has more elongated forewings with more pointed apex, and an overall more vivid and orange colouring. Automeris mineros sp. nov. also has a very different coloration, notably due to the contrasting orange-red periocellar area of the hindwings. Finally, A. fieldi, a species occurring from the Pacific slopes of the Andes to Costa Rica (Fig. 3) and probably north to Honduras (Bénéluz, pers. comm.), stands out again by the slightly more elongated shape of the forewings, but also by the presence of a continuous ring of black scales external to the eyespot of the hindwings, which is lacking in A. llaneros sp. nov.
Description. ♂ (Fig. 1D, H). Wingspan: 72-74 mm. Head: Dark brown, labial palpi and antennae brown. Thorax: dorsally dark brown and ventrally light brown; legs light brown. Abdomen: dorsally orange brown, ventrally light brown. Forewings: Length 32-35 mm, slightly elongated, slightly prominent apex, straight to slightly convex outer border; dorsal ground colour dull orange brown; antemedial line thin and beige, doubled with a brown line proximally; postmedial line slightly convex, barely visible as it reaches the costal margin 2-3 mm from the apex, thin and beige, distally bordered by dark brown scales; basal and median areas concolorous; discocellular mark of the median area rectangular, darker than the surrounding wing surface, with three clearly visible dark brown spots at corners and one faint central spot. Ventral side yellow brown, with darker postmedial and premarginal lines, and a large black discocellular mark, marked in its centre by a white discal spot; venation marked with orange scales. Hindwings: Basomedian area dull orange suffused by dark brown scales, with a large eyespot in its centre formed by, from its centre: an almost completely white small pupil, suffused with few black scales, a large grey brown iris, surrounded by a large black ring and then a thinner yellow ring (in paratype specimen we observed a few black scales external to this outer yellow ring of the eyespot); postmedial line black, lunular, bordered proximally and distally by thin lines of yellow scales; postmedian area dull orange; marginal band large and grey. Ventral side uniformly orange brown, with a weak oblique postmedial line interrupted before it reaches the costal margin, a vestigial premarginal line forming darker U-shaped marks between veins, and a small white discal point.
Female unknown. Genitalia ♂ (Fig. 2C): very similar to those of A. mineros sp. nov. Uncus elongated, large and strongly bifid apically, largely extending beyond the valves. Valves relatively short, rounded, with a broad, rounded dorsal lobe; arms short and strongly curved. Median plate of the gnathos highly sclerotised, as wide as the saccus and subrectangular. Saccus well developed, triangular and acute anteriorly.
Distribution. Automeris llaneros sp. nov. is only known from the region of Carimagua and Orocué, in the Colombian part of the Orinoco watershed, the so called "Llanos Orientales" of Colombian Eastern Plains (Fig. 3). The region has been poorly investigated for saturniid diversity but is known to host a few endemics whose exact distributions need to be clarified (see for example Decaëns et al. 2005).
Etymology. This species is named in reference to the Llanos region, which refer to the large area of savannahs that cover most of the Colombian and Venezuelan Orinoco watershed.

Discussion
The description of three new species of Automeris within the highly cryptic A. bilinea species-subgroup is a new illustration of the value of DNA barcoding, when combined with morphological diagnostic characters and distribution data, in disclosing hidden diversity (Hebert et al. 2004). Our study illustrates two different methodological approaches, which can allow the identification and support the description of new species in such taxonomically difficult groups.
In the case of A. mineros sp. nov., the species had appeared clearly distinct from other known species by the study of its external habitus, showing in particular a unique coloration of the hindwing periocellar area. This singularity is however shared with some specimens representing extreme variants of another species belonging to this subgroup (A. midea). This led us to further explore the question, first by a comparison of male genitalia, used conventionally but whose discriminating characters between closely related species in this group sometimes remains uncertain or even equivocal (Lemaire 1971). Finally, we used DNA barcodes as an independent dataset to confirm the separation of A. mineros sp. nov. from its relatives.
For A. llaneros sp. nov. and A. belemensis sp. nov., the approach was different. The use of DNA barcodes first revealed the existence of cryptic species, i.e., species presenting different barcodes but that were not discriminable from their habitus, or differing from each other only by subtle characters. The subsequent analysis of the genetic distances revealed affinities with species which were not necessarily the most similar in their wing pattern, but such affinities were corroborated in both cases by the structure of the male genitalia. Thus, it is the combination of the genitalia, the habitus, and the DNA barcodes which concomitantly characterise these species within the group.
In the case of A. llaneros sp. nov., the problem of the real identity of A. cinctistriga, a species with which it is impossible to rule out confusion, also deserves to be considered carefully. Automeris cinctistriga was described from a male collected in Colombia, and Lemaire (1971) already stated that the exact identity of this species could be problematic, as the abdomen of the lectotype had been destroyed. It is also most likely that the type-locality of the lectotype, Bogotá (2600m asl), is wrong, as no similar species seems to fly at such high elevation Andean forests of the Colombian Eastern cordillera (pers. obs.). Within the material collected in the eastern plains of Colombia, we found several specimens corresponding to the description of A. cinctistriga based on their wing patterns, and DNA barcoding of these specimens revealed two distinct clusters of barcodes corresponding to two distinct BINs in BOLD. These two BINs were clearly separated in the barcode tree (Fig. 4), making A. cinctistriga in its former definition paraphyletic, and clearly suggesting that they correspond to two different species (Mutanen et al. 2016). In the absence of genitalia and available DNA sequences from the old lectotype of A. cinctistriga, it was formally impossible to define which of the two species corresponded to each of these BINs. We therefore adopted the more conservative position. We considered that A. cinctistriga corresponded to the BIN BOLD:AAA8445, which in BOLD, is defined by a large distribution around the Amazonian watershed (including Bolivia, Brazil, Colombia, Ecuador, French Guiana, Peru), perfectly fitting the distribution of A. cinctistriga as described in the literature (Lemaire 1971(Lemaire , 2002. The other BIN (BOLD:ABZ3239) was represented by the two Colombian specimens used herein as type material for the description of A. llaneros sp. nov. This position is also consistent with wing ornamentation, since the lectotype represented by Lemaire (1971) broadly presents the characteristics that we have attributed to A. cinctistriga.
Overall, the discovery of new species, including some cryptic ones, in a highly diverse genus such as Automeris does not represent a surprising finding. The cryptic diversity of Automeris has already been highlighted by recent taxonomic studies, in which traditionally recognised species with large geographical distributions have been divided into several new species based primarily on differences in DNA barcodes (Brechlin and Meister 2014). This underlines the extent of the taxonomic deficit that characterises the family Saturniidae, which nevertheless is among the best studied within the "Heterocera".
The fact that species traditionally recognised as having a wide distribution prove to be in fact complexes of cryptic species with more restricted distributions also raises new and interesting questions concerning the specificity of the different biogeographical areas which constitute the neotropics. The three regions from which the species described in our study originate are a perfect example. The saturniid fauna is considered to be made up of a mixture of endemics and widely distributed species, with variable proportions depending on the regions. For instance, the Eastern Cordillera of Colombia is considered to be a hotspot of diversity, harbouring a significant proportion of endemics (Lemaire 2002;Decaëns et al. 2007;Decaëns and Rougerie 2008), whereas faunas of the Orinoco and Amazon lowlands are generally considered to be dominated by species with wide distributions (Lemaire 2002). In all cases, however, it is likely that due to this cryptic diversity, a significant proportion of species diversity has been underestimated by the predominant use of morphology in previous estimations. This is well illustrated by our case study, where we describe three new species with probably restricted distributions based largely on the information provided by DNA barcoding. As described in other groups of organisms (Guarnizo et al. 2015), we can therefore expect that the generalisation of the use of DNA barcoding will continue to fragment the expansive distributions of widespread species, thus modifying our current perception of diversity distributed among biogeographical areas.
Lavelle (IRD-CIAT, Cali, Colombia), Marlucia Bonifacio Martins (MPEG, Belém, Brazil), and staff from the Carimagua Research Station (CIAT-CORPOICA, Colombia) for their support during field collecting in Brazilian Amazonia and Colombian Eastern Plains, and Gilbert Lecourt for kindly providing the specimens collected in the region of Muzo. We finally thank Paul Hebert for facilitating DNA barcode sequencing through the international Barcode of Life project.