Identity of the ailanthus webworm moth (Lepidoptera, Yponomeutidae), a complex of two species: evidence from DNA barcoding, morphology and ecology

During extensive ongoing campaigns to inventory moths of North America and Area de Conservacion Guanacaste (ACG), northwestern Costa Rica, we discovered that morphologically similar yponomeutid moths were assigned two diff erent names, Atteva ergatica Walsingham in Costa Rica and A. punctella (Stoll) in North America, but had identical DNA barcodes. Combining DNA barcoding, morphology and food plant records also revealed a complex of two sympatric species that are diagnosable by their DNA barcodes and their facies in Costa Rica. However, neither of the names could be correctly applied to either species, as A. ergatica is a junior synonym and A. punctella a junior homonym. By linking our specimens to type material through morphology and DNA barcoding, we determined that the ACG dry forest species, distributed from Costa Rica to southern Quebec and Ontario, should be called A. aurea, whereas the similar and marginally sympatric ACG rain forest species found in Central America should be called A. pustulella. Neotypes are designated for Phalaena Tinea punctella Stoll, 1781 and Deiopeia aurea Fitch, 1857. Atteva fl oridana has identical barcodes to A. aurea and provisionally maintained as a synonym.


Introduction
Th e ailanthus webworm moth is a conspicuous member of eastern North American micromoth assemblages and is commonly recorded in checklists and inventories as Atteva punctella (Stoll, 1781) (e.g.Heppner and Duckworth 1983;Ding et al. 2006; http://en.wikipedia.org/wiki/Ailanthus_webworm).Th e colloquial name refers to its association with the tree-of-heaven (Ailanthus altissima, Simaroubaceae), an ornamental introduced to Europe and North America from Asia.First planted in Philadelphia in 1784, the plant is now widely distributed across the United States and considered a serious invasive (Ding et al. 2006).Th e ailanthus webworm moth is native to the New World and its native hosts are trees in the genus Simarouba (Simaroubaceae).It is believed that once the expanding range of the tree-of-heaven reached southern Texas, this moth, presumably already present on native hosts, added this new host and expanded north on it (Becker 2009).
Th e name Phalaena (Tinea) punctella was recognized as a junior homonym almost immediately after its description but has been retained through several major works (Heppner and Duckworth 1983;Covell 1984;Heppner 1984).Th e two objective replacement names proposed were Tinea punctella Fabricius, 1787, and Crameria subtilis Hübner, 1822.Th e oldest valid name to replace Phalaena punctella is Tinea pustulella but this remained overlooked until recently (Heppner 2003).Over time seven more nominal taxa were synonymized under Atteva pustulella: aurea Fitch, 1857 (Deiopeia), compta Clemens, 1861 (Poeciloptera), compta fl oridana Neumoegen, 1891 (Oeta), A. edithella Busck, 1908, A. exquisita Busck, 1912, A. ergatica Walsingham, 1914, and A. microsticta Walsingham, 1914.Interestingly there were early suspicions that A. aurea and A. pustulella might represent diff erent species, the former distributed in the United States, the latter in South America, but at the time there was insuffi cient material to support this view (Walsingham 1897).A recent taxonomic review of New World Atteva (Becker 2009) introduced several nomenclatural changes and recognized three separate species within the long-standing concept of A. pustulella: A. pustulella, A. aurea, and A. fl oridana.Th e separation of aurea from pustulella introduced by Becker was based on data presented here.
Atteva is the sole genus constituting the subfamily Attevinae within the Yponomeutidae.Th e group has a pantropical distribution but at least one species (A.aurea) now has a range that extends into the temperate zone.No consistent hypotheses regarding the relationships, placement, and ranking of Attevinae have been published but the prevalent view is that they likely form a monophyletic group within the Yponomeutidae (Kyrki 1990;Landry and Landry 1998;Dugdale et al. 1998;Regier et al. 2009).
Th e authors are currently involved in extensive ongoing campaigns to inventory all Lepidoptera species in North America (Hebert et al. 2009;www.lepbarcoding.org) and in Area de Conservacion Guanacaste (ACG), Costa Rica (Janzen et al. 2005;Janzen et al. 2009;Burns et al. 2007;Burns et al. 2008;Burns et al. 2009; http://janzen.sas.upenn.edu/).It was during the course of these inventories that the problem concerning the identity of A. pustulella fi rst came to light.Incorporating DNA barcoding into inventories has been very eff ective at uncovering cryptic species (Hebert et al. 2004;Janzen et al. 2005;Janzen et al. 2009;Burns et al. 2009;Smith et al. 2007;Smith et al. 2008;Floyd et al. 2009;Hausman et al. 2009) and revealing, then subsequently correcting, taxonomic issues (e.g. the identity of Taygetis andromeda in Janzen et al. 2009).Th is is the case presented here with the ailanthus webworm moth.When identical barcode sequences were recovered for A. pustulella in Ontario and A. ergatica in ACG, the initial response was to change all records of A. ergatica to A. pustulella.Atteva ergatica was originally described from Central America and was long thought to be a synonym of A. pustulella in North America.However, further sampling revealed two adjacent barcode clusters representing two distinct species within ACG and separable by their DNA barcodes, wing patterns, host plant and ecosystem.Th rough examination of type material and literature investigation, we discovered like Becker (2009), that the name applied to the North American ailanthus webworm was inappropriate, but also that what had been thought to be intraspecifi c variation in facies across the ACG dry forest and rain forest ecosystems actually represented two species.
Th e goal of the present study was to determine the correct names for each Atteva species for inclusion into our inventories.However, we also view this study as an example of successful integrative taxonomic eff orts.Our conclusions were achieved through examination of Atteva type specimens, where available and involved a combined genitalia dissection and DNA extraction procedure (Knolke et al. 2005) and the sequencing of mini-barcodes (Hausman et al. 2009;Rougerie et al. personal communication).Where type specimens have been lost, original illustrations were examined.We provide DNA barcode, morphological and ecological diagnostics for the two closely related species in ACG, formerly united under A. pustulella and recently separated again (Becker 2009).

Methods
As part of ongoing inventories in ACG (Janzen et al. 2005;Janzen et al. 2009) and Ontario (Hebert et al. 2009), specimens assigned to the genus Atteva were submitted to the Canadian Centre for DNA barcoding at the Biodiversity Institute of Ontario at the University of Guelph for molecular analysis.Full-length (658 bp) DNA barcodes (Hebert et al. 2003;Floyd et al. 2009) were recorded for these specimens with standard protocols (www.dnabarcoding.ca).
Reared wild-caught specimens (see Janzen et al. 2009 for rearing methods) from the ACG caterpillar inventory are indicated by voucher codes of the form YY-SRNP-XXXXX, where the suffi x contains 1-5 digits.Light-caught ACG BioLep adult specimens have the same voucher code structure, but have 6 digits in the suffi x.
After the discovery that specimens identifi ed as A. ergatica in ACG and A. pustulella in North America had identical DNA barcodes, an eff ort was undertaken to examine and sequence relevant type material from national collections as well as more freshly collected specimens (see specimen records in Appendix I: Dataset 1).Following the protocol of Knolke et al. (2005) DNA was extracted from old type specimens during genitalia dissections for those which still had the abdomen intact, otherwise DNA was extracted from a leg.From these old DNA extracts, barcodes were assembled from six PCR amplicons (Hausman et al. 2009;Rougerie et al. in preparation).Wing-patterns and genitalia preparations were qualitatively examined and compared.
All sequences together with collateral information on the specimens were deposited in BOLD (www.barcodinglife.com)(Ratnasingham and Hebert 2008) in project ATTEV (Atteva of the New World).Sequences were also deposited in GenBank under accession numbers GU013569, GU692470-GU692541, HM034026-HM034136.A maximum parsimony tree was constructed in PAUP (Swoff ord 1998; using PAUPRat, http://users.iab.uaf.edu/~derek_sikes/software2.htm) to demonstrate the presence of diagnostic nucleotides for each species.

Results and discussion
Identical barcode sequences were recovered from so-called pustulella in Ontario and A. ergatica in ACG.Further sampling in ACG revealed two adjacent barcode clusters, one comprised of specimens entirely from the ACG rain forest and feeding on new shoot tips of Simarouba amara (an exclusively rain forest species) and the other of specimens entirely from ACG dry forest and feeding on new shoot tips of Simarouba glauca (an exclusively dry forest species).Once this was realized, it became obvious that these two species could be separated by their forewing colour pattern as well as by their barcodes.Further directed sampling then located the dry forest species feeding on both species of Simarouba in the several-km-wide dry forest-rain forest intergrade, in some cases on the same plant of S. amara side-by-side with the rain forest species.Yet more sampling further into the rain forest ecosystem (to the east of the dry forest) revealed that the dry forest species has now moved at least 20 km into the ACG rain forest ecosystem of S. amara, in open pastures and fi eld edges, and under more xeric conditions than would have been the case were the site still covered with original rain forest.In this circumstance, the rain forest species is at present much more abundant than is the dry forest species.
Our conclusion that the ACG specimens comprise two species is based on concordance (Avise and Ball 1990) between morphological, molecular and ecological characters that are diagnostic (Cracraft 1983) for each population, providing evidence for a hypothesis of genetic isolation.Th ere was no divergence in the genitalia among the species examined.Th is lack of a genital diff erence is consistent with previous morphological studies of this group (Becker 2009).Based on congruent wing patterns of the A. punctella type image (Figure 2A) and the wings of our rain forest cluster, we conclude that the correct name for the rain forest species is A. pustulella, which also fi ts with its rain forest distribution in the Neotropics (Becker 2009).Based on congruent wing patterns and identical DNA barcodes with the type specimens of A. ergatica and A. edithella (Figure 1), both synonymized under A. aurea (Becker 2009), we conclude that the ACG dry forest species, which extends north to southeastern Canada, should be called A. aurea (Figure 2).Th e two correct names along with synonyms and diagnostic characteristics are listed below, along with two other Atteva species included in our inventories.
Interestingly, A. aurea is more proximate in barcode to A. hysginiella than A. punctella (Table 1; Figure 1).Atteva hysginiella is geographically disjunct as a Galapagos Islands endemic (Figure 2), strikingly diff erent in coloration, and uses a diff erent host plant.Although intraspecifi c distances (Table 1) may seem high within these species compared to values reported in other studies (e.g.Janzen et al. 2009), we observed no relationship between genetic distance and geographic distance or morphological diff erences (Figure 1).Short sequence lengths obtained from older specimens (see specimen records in additional fi le 1: Datasheet 1) can make phenetic distances seem larger than if a full barcode sequence (658bp) were available for comparison.Th is is a problem seen when using percentage phenetic distances but it is clear from the maximum parsimony analysis (Figure 1) and the consensus barcodes shown in Table 1, that each species possesses a cohesive cluster of haplotypes with few polymorphisms, and diagnostic characters are present which separate all species.

Atteva pustulella (Fabricius)
Phalaena Tinea punctella Stoll, 1781: 164 Forewings.Th e pattern ranging from Uruguay and Argentina northwards to Costa Rica, consists of thick black outlinings around the white spots and the orange spots reduced, with the antemedial and medial orange fasciae divided into two large blotches each (Figure 3B).No other morphological features, either in colouration, other external characters or genitalia, distinguish this species from A. aurea.
Habitat and food plants.Atteva pustulella caterpillars (400+) have been found only in the ACG rain forest ecosystem and feeding only on new shoots of Simarouba amara.Where the ACG rain forest intergrades with ACG dry forest, to date all caterpillars from S. amara have been those of A. pustulella (while caterpillars of A. aurea have been found on both S. amara and S. glauca in the intergrade zone) (Figure 2).
Distribution.Th is species ranges from Costa Rica, where it meets aurea, southwards to Uruguay and Argentina.It is also present in the Antilles.Becker (2009)

Number of sequences Number of unique haplotypes
Mean intraspecifi c distance Atteva zebra 18 15 0.39 AACTTTATATTTTATTTTTGGAATTTGAAGAGG{A,G}ATAGTCGGTACCTCATTAA-GATTTTTAATTCGAGCAGAATTAGGTAATCCTGGATTTTTAATTGGAGATGATCAAATT-TATAATACTATTGTAACAGCTCATGCTTTCATTATAATTTTTTTTAT{A,G}GTTATACCTAT-CATAATTGGAGGATTTGGTAATTGATTAGTTCCTTTAATACTAGGAGCTCCTGATATAG-CATTCCCACGAATAAATAATATAAGTTTTTGATTACTACCCCCATCTCTTACTCTTTTA-ATTTCTAGAAGAATCGTTGAAAATGGAGCAGGAACTGGATGAACAGTTTATCCACCTT-TATCTTCTAATATTGCTCATGGAGGAAGTTCTGTTGATTTAGCTATTTTTTCTTTA-CATTTAGCCGGAATTTCATCTATTTTAGGAGCTATTAATTTTATTACTACAATTAT-TAATATACGAAGTAATGGTATAAATTTTGATCAAATACCTTTATTTGTTTGAGCT-GTAGGTATTACAGCTTTACTTTTATTGTT{A,G}TCATTACCTGTTTTAGC{A,G} GGAGCTATTACTATACTTTTAACTGA{C,T}CG{A,G}AATCTTAATACTTCATTTTTT-GATCCTGCAGGTGGAGGAGACCCAATCCTTTATCAACATTTATTT Neotype designations.Enquiries at several major institutions (AMNH, ANSP, BMNH, NMNH, NYSM, USNM, ZMUC) failed to locate the type specimens of A. aurea and A. punctella (=pustulella).Similar negative results were indicated by Becker (2009) in his recent taxonomic review of Atteva.Th e primary types of the oldest available names appear to be lost.Th e evidence brought here strongly supports the occurrence of two distinct species long subsumed under the name punctella.Th e differentiating characters are in forewing coloration, DNA barcodes, larval host plants, and habitats, as described here under each respective species and in the 'Results and Discussion' section.However, wing coloration is the only character that can be used to interpret the old descriptions.Th e proper application of the two names was enabled retrospectively after diff erences in life history and barcodes were discovered and superimposed on the diff erences in forewing pattern.
Th e original fi gure of Atteva punctella appears on Plate 372 in the fourth volume of Cramer's "De Uitlandsche Kapellen Asia, Africa en America" published in Amsterdam in 1780-1782 (copy in the Canadian Agriculture Library, Neatby Building, Ottawa).Following Cramer's death in 1780, the work was completed by his associate Caspar Stoll, who authored the section containing the description of punctella (Stoll 1781;ICZN 1958).Th e original colour illustration of punctella shows distinctly the diagnostic thick black lining and reduced orange markings in the forewing characteristic of the rainforest species.Its provenance from Surinam, presumably in the coastal area near or around Paramaribo, in what we deem to have been a primary rainforest area leads us to attribute that name to the rainforest species encountered in ACG.However, the fi gure in Cramer (Figure 3A) is a very small painting on which watercolours were manually applied on individual copies of the book, so it is not possible to know how accurately Th e distance values shown are % Kimura-2-parameter distances as calculated in BOLD.Th e DNA barcodes for each species is directly below the name and statistics for each species.Th e bold-face nucleotides highlight those bases that are unique to a species, and the bracketed bases indicate the location and nature of haplotype variants.the fi gure represented the specimen on which it was based.It was described from an unspecifi ed number of specimens.
Although we are able to distinguish the two species on forewing pattern, there is enough variation in that character, as well as in larval host and habitat for A. aurea to leave open the possibility of misapplication of that name if based on these features alone.On the other hand, the barcodes are unequivocal.Th erefore we deem it warranted to designate neotypes for both punctella Stoll and aurea Fitch in the interest of clarifying the taxonomic status of those nominal species and stabilizing nomenclature.Th e neotype was selected based on availability of a full barcode without ambiguous bases in addition to matching the forewing pattern of the original illustration of punctella and being from a rainforest location.While it would have been desirable to select a specimen from a locality "as nearly as practicable from the original type locality" (ICZN 1999, Art. 75.3.6),none from Surinam was available.
Forewings.Th e forewings are predominantly orange, with four bands of yellow spots outlined in black.Compared with pustulella the orange is more extended and the black outlines surrounding the white spots are thinner (Figure 5A).Th ere is a marked amount of variation in the number and size of pale spots but in spite of this variation aurea is clearly separated from pustulella (Figures 4-6).See Remarks below.
Habitat and food plants.Caterpillars are commonly encountered in ACG dry forest feeding on Simarouba glauca (100+ records) and in the ecotone between dry forest and rain forest, found feeding on both S. glauca and S. amara.Th e aurea population has extended at least 20 kilometers into ACG rain forest in anthropogenic artifi cially more xeric conditions, feeding on S. amara as well.It is found to the north of Costa Rica in dry forest to semi-desert ecosystems, and north to southeastern Canada, feeding on Ailanthus altissima.
Distribution.In North America, distributed wherever its non-native host occurs but vagrant adults are regularly found at its northern limit in eastern Ontario and southwestern Quebec beyond the host range.
Remarks.Th ere is signifi cant variation in forewing coloration as attested by several synonyms.More pronounced variants with reduced forewing black linings (edithella, exquisita) were described from semi-arid regions of northern Mexico and from Texas.Among barcoded specimens is an aberrant one (CNCLEP00027027 from Maryland) in which the forewing black reticulation failed to develop except asymmetrically in one small area of the right forewing (Figure 5D).Th at specimen was collected together with many others which displayed the normal coloration (Figure 5B-C).Th ere is evidence that colour features including the relative extent of black reticulation and intensity of orange and pale yellow may be aff ected by temperature, with the black tending toward reduction at higher temperature (Powell et al. 1973).
Specimens attributable to the fl oridana have barcodes unequivocally belonging to the aurea array of specimens (Figure 6C-D).Attempts to barcode the old type specimen of fl oridana failed.We maintain fl oridana as a synonym of aurea, consistent with the traditional treatment of other authors (Heppner and Duckworth 1983;Heppner 1984Heppner , 2003;;Covell 1984), but contrary to Becker (2009).Its status remains debatable, and it is possible for two species to have the same barcode.Barcode sharing is known to occur, albeit infrequently, among some close relatives in Lepidoptera (Hebert et al. 2009).Th e only evidence for considering fl oridana a separate species is the allegedly distinct forewing colour pattern with reduced spots (Figure 6), which is restricted to southern Florida and seems constant in sympatry with A. aurea (Becker 2009), whereas A. aurea varies considerably in forewing pattern over its wide geographic range.However, some specimens are nearly intermediate in pattern (Figure 6C) and would be questionably attributed to either aurea or form fl oridana from wing pattern alone.Larvae of the fl oridana form feed on Simarouba glauca and so overlap in host plant with aurea.It was reported that their larvae are distinct (Walsingham 1914: 329;Becker 2009), but this contention seems to be based on incorrectly interpreting a description of the fl oridana larva (Dyar 1897) in which the two species or forms were not actually compared.Th e alleged diff erence in larvae remains unverifi ed.Specimens seemingly of the form fl oridana examined by us were all collected early in April and May and it would be interesting to see if the form is seasonally related.Specimens of A. aurea were collected at various dates from May to August in the North American part of its range.
Although the type specimen of A. aurea, from Savannah, Georgia, appears to be lost, the description matches that of the ailanthus webworm moth and this is the oldest name applicable to North American populations of this Atteva.However, the original description is insuffi cient to determine the thickness of the forewing black lining and relative size of the orange and white spots, and there is no illustration of its type.Our attribution of that name is based on the original type locality which falls within the geographical range of the dry forest species and is far outside the range of the rain forest species.Th e neotype was selected based on availability of a full barcode without ambiguous bases in addition to matching the original description of the forewing pattern of A. aurea.While it would have been desirable to select a specimen from a locality "as nearly as practicable from the original type locality" (ICZN 1999, Art. 75.3.6),none from Georgia was available.Th e selected neotype is from northern Florida in a region with habitats similar to what occurs around Savannah, Georgia. .Th e specimen is spread and missing the left fore-and mid-legs which were removed for DNA extraction, but otherwise is in good condition (Figure 5A).Barcode sequence under GenBank accession number HM034047.
Habitat and food plants.Landry and Landry (1998) gave a detailed description of adults, immatures and life-history, with the caterpillar feeding on Castela galapageia (Simaroubaceae).
Distribution.Restricted to the Galapagos Islands (Figure 2).Th e type locality recorded as being in Panama has its origin in a labeling error (Becker 2009).
Forewings.Th e zebra pattern makes this species readily distinct from the other species presented here (Figure 3D).
Habitat and food plants.A. zebra is the common webworm of shoot tips of Simarouba amara saplings and adult trees in ACG rain forest (n = 123) It is more abundant than A. pustulella, but may be found on the same individual tree with A. pustulella and an occasional A. aurea in anthropogenic rain forest habitats.It has never been found on S. glauca or in ACG dry forest.
Distribution.Known only from Costa Rica and Panama.

Concluding remarks
Th is case study demonstrates the value of combining morphological, ecological and DNA barcode information when working with similar species.Atteva is an example where seemingly confusing morphological and ecological patterns, can be defi nitively partitioned in the light of discrete data such as DNA sequences.Th e integration and synthesis of inventories, each one necessarily regionally focused, is facilitated by DNA barcodes, an effi ciently communicated online character system.Th is was demonstrated by the fact that taxonomic problems surrounding the ailanthus webworm moth persisted in the ACG for 25 years and surfaced only recently.From the starting point of DNA barcode analyses it has been relatively straightforward to reach a taxonomic conclusion by joining taxonomic knowledge in the form of the name-bearing types with ecological and morphological information.Th e purported diffi cultly in obtaining barcodes from type material has been viewed as an obstacle to the melding of DNA barcoding information with other taxonomic information.
Recent studies (Hausmann et al. 2009), including this one, show that this is not necessarily the case.

Figure 1 .
Figure 1.One of 199 most parsimonious trees found by the PAUPRat analysis of Atteva barcodes sequences showing the presence of diagnostic nucleotide diff erences between the species.Th e scale bar shows the number of changes.Notable specimens are highlighted in red.

Figure 2 .
Figure 2. Map showing the distribution of Atteva specimens examined as part of this study.Notable specimens are highlighted in red.