Phylogenetic position of the ‘extinct’ Fijian coconut moth, Levuana iridescens (Lepidoptera: Zygaenidae)

Levuana iridescens Bethune-Baker, 1906, a day-flying moth purported to be endemic to the Fijian Island of Viti Levu and a former pest of its coconut palm trees, was last observed in 1956 and has been officially declared extinct by IUCN since 1996. The controversial classical biological control method that resulted in the (presumed) demise of this moth has given this species an iconic status in biological control studies. We investigated the sister-group relationships and phylogenetic placement of this moth using NGS-obtained ancient DNA sequences from museum specimens of L. iridescens collected in the 1920s, combined with 31 morphological characters used in earlier studies and 2 new characters. Our findings show that Levuana is most closely related to the Australian genus Myrtartona. The significance of these findings is discussed.


Introduction
The biological control program that resulted in the apparent extinction of the Fijian coconut moth, Levuana iridescens (Lepidoptera: Zygaenidae: Procridinae: Artonini) in the 1930s is often regarded as an example of a successful classical biological control program, or one of extinction of a native insect following the introduction of an exotic control agent [1][2][3][4][5][6][7][8][9][10][11][12][13]. In summary, the purposeful introduction of the Malayan parasitoid Bessa remota (Diptera: Tachinidae) isolated from the closely related zygaenid moth Palmartona catoxantha by entomologist Robert Tothill in 1925 resulted in a rapid decline and presumed extinction of the populations of L. iridescens in Viti Levu [5,11,14]. The moth was known only from Fiji, however, its endemicity continues to be debated [11].
In the Pacific Region, no zygaenids are known from Micronesia and Polynesia. Melanesia, on the other hand, has a rich zygaenid (Procridinae and Chalcosiinae) fauna, especially in the mountains of New Guinea. Two species are known from New Ireland: Homophylotis aenea Jordan, 1925 (Procridinae: Artonini) and Heteropan lycaenoides Walker, 1864 (Chalcosiinae: Heteropanini); one from the 'Bismarck Archipelago': Homophylotis doloides (Pagenstecher, 1900) (Procridinae: Artonini); and one from Woodlark Island: Heteropan cyaneus Jordan, PLOS [14] mention that they reared two 'Artona' species from larvae found on the Solomon Islands; one from Zingiberaceae, the another from banana (Musa sp., Musaceae). So far we have not been able to locate the material related to this rearing and therefore cannot verify these species. Banana (Musa sp.) is the larval hostplant of Pseudoamuria melaleuca (Jordan, 1908) in New Guinea [15]. Zingiberaceae are known as larval hostplants especially for species of the genus Amuria Staudinger, 1887. Despite its iconic importance in biological control studies, the sister-group relationships and phylogenetic placement of L. iridescens has never been properly studied. Hering (1922) [16], Tothill et al (1930) [15] and Bryk (1936) [17] all placed it in Chalcosiinae [18]. Kalshoven (1981) [19] considered L. iridescens to be conspecific with Palmartona catoxantha, a widespread Indo-Malayan species (and the source host for the parasitoid, B. remota). This synonymy was subsequently questioned [7]. Levuana has been placed in Artonini (Zygaenidae: Procridinae) since the description of this tribe by Tarmann in 1994 [20]. In his revision of the Australian Artonini, Tarmann [15] also presented a phylogeny for all Australian genera based on 31 morphological characters, but he did not include the genus Levuana.
Although the moth has been officially classified as 'extinct' by the World Conservation Monitoring Centre since 1996 [39], doubts have been expressed [7,10,11,15,40]. The last known specimen was collected in 1941 with further possible observational records in 1953 and 1956 [11,41]. It has been speculated that it might still exist in low numbers on Viti Levu [10]. If the moth is rediscovered, it is likely that measures will have to be considered by the Fijian government (or international conservation organizations) towards its protection. Conservation of imperiled species, such as L. iridescens, requires correct diagnosis of their taxonomic status for effective implementation of management actions. Analysis of phylogenetic affiliations is particularly crucial in management plans that involve translocation, re-introduction, population augmentation, or captive propagation [42]. In this study we investigate the phylogenetic position of Levuana within the Australian Artonini using morphological characters as well as DNA barcode sequence data obtained from old museum specimens for the first time.

Morphological data analysis
We coded thirty-one morphological characters used in a previous phylogenetic analysis of Australian Artonini [15] for Levuana, and added two new characters (32)(33) to new data matrix (S1 File). Character 19 was coded '0' for Adscita in Tarmann 2004 [15]; we re-coded this character as '2' instead, since A. statices was the model. Only few Adscita species have welldeveloped praebursa; the spermatophore is in the praebursa in most Artonini species, but it is usually in the corpus bursae in Adscita.
New dissections for both sexes of Levuana were made from USNM specimens (Fig 1), and morphological characters were coded by GMT using these specimens and dissections, as well as detailed descriptions and illustrations by Tothill et al in 1930 [14]. Adscita Retzius, 1783 (= Procris [Fabricius in Illiger], 1807), the type genus for the sister-tribe Procridini [21], was selected as outgroup. This new data matrix containing 14 taxa was analysed using PAUP � 4.0a164 [43] and a consensus tree was obtained. Branch and bound analysis yielded 43 trees of length 63, Cl = 0. 6825 (Cl excluding uninformatives = 0.6491), RI = 0.7143. The Majority Rule consensus tree is illustrated (Fig 2A).

Molecular data analysis
In recent years, DNA barcode data for most species of Australian Procridinae have become available through Barcode of Life Campaign (v4.boldsystems.org). We selected available sequences representing the genera of Australian Artonini sensu Tarmann 2004 [15] for phylogenetic analysis (Table 1). Prior to this study, four relevant genera (Amuria, Thyrassia, Palmartona and Levuana) had no representation in the DNA barcode library. In aiming to complete the generic coverage, we successfully obtained sequences for two of these genera: Four specimens of P. catoxantha from Malaysia (USNM) and eight specimens of L. iridescens from Fiji (CAS: 5, UCR: 1, USNM: 2) were subjected to next-generation (NGS) sequencing. A single leg was removed from each specimen and sent to Biodiversity Institute of Ontario in Guelph, Canada for DNA extraction, amplification, and sequencing. All work was performed in a dedicated clean lab using UV sterilized equipment to minimize the risk of contamination. DNA extraction and purification was performed following the methods outlined by Prosser et al [44] and the 658 bp COI barcode region was amplified for NGS. Briefly, primers normally used for amplifying six overlapping amplicons in separate reactions were multiplexed in two PCR reactions per sample, and the resulting amplicons from each sample were pooled and sequenced on an Ion Torrent PGM (Thermo Fisher) using a 318 v2 chip and OT2 400bp chemistry (Thermo Fisher). Approximately 4.2M sequence reads were generated, which were demultiplexed into an average of 353K (±143K) reads per specimen. For each specimen, reads were processed to remove primer sequences (Cutadapt v1.8.1) and then filtered based on a minimum quality score of QV20 and a minimum length of 100bp (Sickle v1.33). The reads were then dereplicated (Fastx Toolkit v0.0.14) and assembled into a full-length barcode contig by aligning to a reference barcode sequence [45]. Final sequences were deposited in GenBank (accessions MN555769-MN555779) and all NGS reads were deposited in the Sequence Read Archive. All four P. catoxantha and only five L. iridescens specimens yielded any usable DNA sequences. All of the barcode records are also publicly available in the BOLD dataset "DS-LE-VUANA", accessed at dx.doi.org/10.5883/DS-LEVUANA. Initial NJ trees were generated using analytical modules implemented in BOLD (v4.boldsystems.org). Maximum Parsimony and Maximum Likelihood analyses were conducted in PAUP � 4.0a164 [43], and MrBayes 3.2.7 [46] was used for Bayesian analysis. Result of the Maximum Likelihood analysis is presented ( Fig 2B).

Combined data analysis
The partitioned nexus of combined data (658 bp DNA barcodes and 33 morphological characters) was initially analysed using Mesquite 3.6 [47] and repeated in PAUP � 4.0a164 [43], which yielded similar results. Add & rearrange heuristic searches were implemented for tree inference, with 'Tree-value using Character Matrix' and Nearest-Neighbor-Interchange (NNI) tree rearranger criteria selected. Maxtrees was set to 1000. The analysis yielded 100 trees, a 50% majority rule consensus of which is shown (Fig 3). Since no DNA sequences were available for Amuria and Thyrassia, the position of these two taxa in the phylogeny was supported only by morphological characters. Evolution of morphological characters were subsequently investigated using 'trace character history' function in Mesquite.

Re-description
Forewing length in male 5.5-6.2 mm, in female 6.2-7.1 mm. Head of typical artonoid form (see Tarmann   Thorax dark greenish blue with metallic sheen and purple tinge, legs bluish green basally, ochreous distally; dorsal abdominal segments 1 and 8 in male and 1 and 7 in female metallic greenish blue, segments 2-7 (male) and 2-6 (female) ochreous with blue tinge near middle line, ventral abdominal segments ochreous with blue tinge, especially posteriorly. Wing venation in forewing with (R2+R3) stalked together, sometimes also together with M1, medial stem absent; hindwing with a short transverse vein between Sc and anterior margin of cell that can become vestigial.
The frenulum in Levuana female has only one bristle. This is a very important character that is not shared with any of the Australian Zygaenidae. Pollanisus, Onceropyga, Hestiochora, Turneriprocris, Myrtartona, and Australartona females have 3 bristles, Homophylotis has 2 or 3 (variable), Pseudoamuria and Palmartona have 2, and only the SE Asian genus Thyrassia has one bristle in both male and female. However, Thyrassia has many reductions and a Ctenuchid-like habitus that is very different from Levuana. Another important and so far unrecognized character is the distribution of scales on the wings in Levuana. The forewing upperside and the anterior part of the hindwing upperside until the medial stem and vein RR are densely covered with elongate triangular scales with a slightly bilobed distal end and mixed with smaller, almost needle-shaped scales that are partly arranged in a cross direction and situated below the larger scales, the posterior part of the hindwing is very translucent due to a single layer of less densely arranged upright standing needle-shaped scales. Only at the anal angle, two layers of needle-shaped scales can be found and there the hindwing is less translucent. All Australian genera and also Palmartona and Thyrassia are more densely scaled on the hindwing and are only in the central part of the hindwing sometimes more translucent.

Male genitalia
Uncus slender, with triangular base, short (shorter than in Pollanisus and Onceropyga), about half the length of dorsal length of valva, tapering towards and pointed at apex (as in Turneriprocris and not distally down-curved like in Myrtartona); tegumen long and slender, vinculum narrow, rounded, without ventral process; valva with fan-shaped translucent central part and a narrow sclerotisation dorsally and ventrally, without process, sacculus well-separated like in the Australian genera Pollanisus and Onceropyga (see [15]: 108, Figs 136, 136, 274), pulvinus stalked on a broad base, of similar form and in the same position as in Pollanisus. Phallus tubelike, approximately 5-times longer than broad, weakly sclerotized, the everted vesica of same shape as in Pollanisus, Onceropyga and Myrtartona, with one slender sclerotisation, somehow like the cornutus in Pollanisus and Onceropyga but shorter, without sharply pointed distal end, very similar to the structure in the vesica of Myrtartona mariannae (Tarmann, 2004) (see [15]: Fig 372) and a second sclerotisation that is forming an oval patch.

Molecular analysis
Our results support the status of L. iridescens as a distinct species with unique morphology and DNA barcode sequence that is very different from P. catoxantha (distance 10.1%). Analysis of morphological data alone revealed the position of Levuana in a cluster with Myrtartona and basal to six other genera (Fig 2A). This grouping was primarily supported by the absence of the medial stem in the venation of the forewing (character 7). The position of Levuana in this cluster (minus Myrtartona) was further supported by the monocotyledon larval hostplants (character 31).
Overall, DNA barcode distances were above 10% among the selected Procridinae (10.119%), with an average of 13.6% (Table 2). Various phylogenetic analyses of barcode sequences yielded varying topologies, as expected with phylogenies above species level that use saturated DNA barcode data alone. Support was generally low for deeper nodes, with few nodes consistently present (e.g. Australartona + Homophylotis and Onceropyga + Pollanisus) albeit with weak or no support. The position of Levuana was generally unstable across analyses of DNA barcode data alone. In Maximum Likelihood analyses, Levuana appeared in a clade with Turneiprocris, Hestiochora and Onceropyga, but with no support (Fig 2B). Analysis of combined data also placed Levuana next to Myrtartona, and sister to a clade containing Australartona, Artona + Amuria, and Palmartona + Homophylotis + Pseudamuria, with moderate support (Fig 3).

Discussion
Insects and their natural enemies, parasitoids in particular, have generally coevolved in their natural range and therefore understanding their phylogenetic affinities is crucial in biological control risk assessment. Insights into phylogenetic relationships also helps in more accurately predicting non-target hosts and assessing other risks of biological control introductions. For reasons of biosafety, host-specific parasitoids are often selected in preference over generalists, and so it is expected that closely related hosts are more likely to be at risk from attack of a parasitoid than more distantly related species. In Fiji, the introduction of Bessa remota to control the L. iridescens also affected Heteropan dolens (Druce 1888) (Zygaenidae: Chalcosiinae), another moth that suffered population decline, presumably due to parasitism by B. remota [48]. H. dolens was considered extinct until its rediscovery in 1963 [11].
The hypothesis that Levuana was not endemic to Fiji but probably introduced from another area of the Pacific remains unresolved, as is the native occurrence of its host, the pacific coconut Cocos nucifera, on Fijian coasts [49]. Lack of known records of Levuana prior to its first outbreak in 1877, the strictly-limited presence on Viti Levu (while most of the endemic species of Fiji are on several islands), and its later expansion into the surrounding islets support the idea that the moth was introduced to Fiji [11]. However, we find lack of records prior to its first discovery on Viti Levu in late 19th century irrelevant. Permanent winds, either from the sea to inland or vice versa, dominate most tropical islands including Fijian coasts, and only twice a day for about 30 minutes, when the winds change direction, the conditions are calm (GMT, personal observation). As far as we know, no attempts-either historical or recenthave been made to find L. iridescens in this window. These fragile little insects are most certainly only active around that time, and as in many other Artonini, it would be very difficult to encounter them only by active searching. This situation is similar in Palmartona catoxantha: Although this species is more common and widespread than L. iridescens, hardly anyone has observed or collected this species outside of outbreak years. Only one old specimen is known from Australia, near Brisbane (BMNH). Whether taken accidentally or a mislabeled specimen from elsewhere, this specimen was the reason for including Palmartona in the Australian revision of Artonini [15]. Absence of parasitoids attacking L. iridescens on Viti Levu, and thus the frequent and persistent outbreaks that regularly killed its host plants, may also suggest that L. iridescens was a late arrival to Viti Levu [11]. However, the notion that L. iridescens suddenly became a pest and had no natural enemies is not very well supported. The moth may have existed in small native colonies in its natural habitat along the coast and among the strong coastal winds, but its populations exploded only after farmers began to plant coconut trees inland and in large monocultures. Coastal winds are weak or absent on the inland, which would have created the ideal conditions for L. iridescens to develop en masse. Generalist parasitoids may not have been present on Fiji, explaining why none were associated with L. iridescens. Specialist parasitoids associated with other Lepidoptera, e.g. H. dolens, may occur on Fiji, although comprehensive studies are lacking. When the moth became a pest of coconut cultivations, no species-specific local parasitoids specialized on them, and thus the introduction of the non-native parasitoid (Bessa remota) as a biological control agent in the 1920s quickly and effectively ended the destruction of coconut palms by L. iridescens in Fiji.
The position of Levuana near Myrtartona and other Australian genera indicates that Levuana has likely split off from the Australian species-groups at a time before Pollanisus and the Onceropyga / Hestiochora group of genera had developed their striking synapomorphies (female inner genitalia without Petersen's glands but with the abdominal hair brush that covers the eggs, with presumably poisonous scales to protect the eggs). However, the genitalia characters of Levuana lack all the reductions of Homophylotis and lack also important characters from Myrtartona and the Artona-finger of Australartona. Alberti [50] was the first to give a differential diagnosis for the genitalia of this species in comparison with other South-East Asian and Australian Zygaenidae. He commented that the genitalia of Levuana are so similar to those of Pollanisus that he would not see a problem to treat these two taxa as two subgenera of Pollanisus. However, he did not change the generic status of Levuana because he did not have enough material for comparison from Australia and the adjacent islands to the North. The genitalia of Levuana seem to be intermediate between Myrtartona and Pollanisus, which indicates that Levuana may be of Australian (and not of Indo-Malayan) origin. Morphologically, the Indo-Malayan genera Artona, Amuria, Palmartona and Pseudoamuria are not very closely related to Levuana. It may well be that Levuana is an old relict that has been on Fiji for a long time, and is derived directly from a primitive Australian ancestor who also gave rise to Pollanisus and Onceropyga on one branch and Myrtartona and Turneriprocris on the other (see also discussion in [15]: 55). This also suggests that the hypothesis that L. iridescens is an introduced species to Fiji and originates from elsewhere is most unlikely.
Over the recent decades, the tone of discussion around the fate of Levuana seems to have shifted from "a classical example of successful biological control" [1] and "best documented case of extinction" [5,8] to "extinction of a native insect following the introduction of an exotic control agent" [12] and to 'possibly not extinct at all' [7,10,11,15,40]. Indeed, various factors indicate that the paucity of recent records since the conclusion of Levuana Campaign in 1930s has more to do with lack of proper surveys than anything else [11]. The moth may still survive in Fiji in low numbers, perhaps in small inaccessible offshore islands neighbouring Viti Levu where coconut palms are not maintained for commercial use (e.g. Nukulau and Makuluva) [14]. Without the aid of artificial pheromone attractants, which are currently unknown for Artonini, the best method would be to use malaise traps on tallest palms at canopy level [11]. The best time to look for adults of Levuana will probably be August to December, at times of day when the winds are calm. However, even if pupal cases or cocoon clusters are found, chemical analysis of silk amino-acid composition [11], as well as DNA barcodes-published for the first time through this study-can confirm their identity.
Supporting information S1 File. List of morphological characters used in phylogenetic analysis. For a detailed description and illustration of characters, see Tarmann 2004 [15]. (DOCX)