When taxonomy and biological control researchers unite: Species delimitation of Eadya parasitoids (Braconidae) and consequences for classical biological control of invasive paropsine pests of Eucalyptus

The invasive eucalyptus tortoise beetle, Paropsis charybdis, defoliates plantations of Eucalyptus nitens in New Zealand. Recent efforts to identify host specific biological control agents (parasitoids) from Tasmania, Australia, have focused on the larval parasitoid wasp, Eadya paropsidis (Braconidae), first described in 1978. In Tasmania, Eadya has been reared from Paropsisterna agricola (genus abbreviated Pst.), a smaller paropsine that feeds as a larva on juvenile rather than adult foliage of Eucalyptus nitens. To determine which of the many paropsine beetle hosts native to Tasmania are utilized by E. paropsidis, and to rule out the presence of cryptic species, a molecular phylogenetic approach was combined with host data from rearing experiments from multiple locations across six years. Sampling included 188 wasps and 94 beetles for molecular data alone. Two mitochondrial genes (COI and Cytb) and one nuclear gene (28S) were analyzed to assess the species limits in the parasitoid wasps. The mitochondrial genes were congruent in delimiting four separate phylogenetic species, all supported by morphological examinations of Eadya specimens collected throughout Tasmania. Eadya paropsidis was true to the type description, and was almost exclusively associated with P. tasmanica. A new cryptic species similar to E. paropsidis, Eadya sp. 3, was readily reared from Pst. agricola and P. charybdis from all sites and all years. Eadya sp. 3 represents the best candidate for biological control of P. charybdis and was determined as the species undergoing host range testing in New Zealand for its potential as a biological control agent. Another new species, Eadya sp. 1, was morphologically distinctive and attacked multiple hosts. The most common host was Pst. variicollis, but was also reared from Pst. nobilitata and Pst. selmani. Eadya sp. 1 may have potential for control against Pst. variicollis, a new incursion in New Zealand, and possibly Pst. selmani in Ireland. Our molecular data suggests that Pst. variicollis is in need of taxonomic revision and the geographic source of the beetle in New Zealand may not be Tasmania. Eadya sp. 2 was rarely collected and attacked P. aegrota elliotti and P. charybdis. Most species of Eadya present in Tasmania are not host specific to one beetle species alone, but demonstrate some host plasticity across the genera Paropsisterna and Paropsis. This study is an excellent example of collaborative phylogenetic and biological control research prior to the release of prospective biological control agents, and has important implications for the Eucalyptus industry worldwide.

Introduction Classical biological control of insects, involving the importation of a specialist parasitoid or predatory organism (agents) from the area of origin to control a pest (target), has proven to be an effective alternative to insecticide use for the control of numerous invasive species [1][2][3]. Classical biological control is now only considered when environmental safety concerns can be empirically evaluated. Success of classical biological control programs depends on ensuring: (1) an agent is sufficiently host specific to the pest to avoid significant non-target impacts; (2) a phenological match between the target and agent to facilitate efficient control and also prevent non-target impacts; and (3) the agent can survive and reproduce in the novel environment [4,2,[5][6][7].
Cryptic species present another challenge for successful biological control [8][9][10][11]. Cryptic species complexes are typically comprised of a set of related species that are morphologically indistinguishable or difficult to diagnose based on morphology alone. As the majority of taxonomic works have been based on only morphological characters, cryptic species complexes typically include a set of undescribed species [12]. Molecular taxonomy has revealed numerous cryptic species complexes, particularly within insects [13][14][15][16][17][18]. If a prospective biological control agent is a part of a cryptic species complex, multiple species may end up being released causing non-target effects, or if the wrong species is released, control of the target pest may not occur [11,7,19]. Thus, it is critical to test for the presence of cryptic species by sampling specimens from a wide range of localities from the agent's origin.
Eucalyptus plantations are an important wood and pulp fiber resource for the forestry industry in many countries in the world; however, the survival and expansion of these plantations in New Zealand and elsewhere are under threat due to the presence of insect defoliators, including several species of chrysomelid beetles across several continents [20][21][22]. Of invasive pests in New Zealand, the most serious is the Eucalyptus Tortoise Beetle, Paropsis charybdis Stål 1860 (Coleoptera: Chrysomelidae: Chrysomelinae: Paropsini) [23,24,21]. P. charybdis has two generations per year in New Zealand, with the first generation adults emerging in August and September after overwintering within the leaf litter and under bark (spring generation) [23,24] (Fig 1). There are four larval instars feeding on both expanding and adult leaves. Pupation of the spring generation occurs from November to December, and after approximately one month the second generation of beetles emerge (summer generation) [23,24]. The summer generation appears the most damaging. Beetles are present until autumn, and thus a third generation has been proposed but not yet proven [25]. Although native to Australia, P. charybdis is now found virtually throughout New Zealand following its initial invasion to a localized region in 1916 [24]. The extensive damage caused by P. charybdis is likely the result of high female fecundity, wide host range, and a lack of natural enemies attacking the spring generation [26,27].
Since the 1930s, classical biological control of P. charybdis in New Zealand has been attempted repeatedly through the importation and release of larval parasitoids (Tachinidae), egg parasitoids (Pteromalidae), and ladybird beetle predators (Coccinellidae) from Australia [28]. The majority of these agents failed to establish in New Zealand upon release [28]. The Although there are two generations of paropsine beetles in Tasmania, Eadya are univoltine, attacking only the first (spring) generation (dark gray). Eadya are larval parasitoids, typically attacking 2 nd instar larvae and emerging from the prepupal stage. Despite a second generation of beetle hosts (light gray) available, Eadya undergo a ten month obligate pupal diapause period. most successful biological control agent thus far was the pteromalid egg parasitoid, Enoggera nassaui (Girault, 1926) [28,21]. This parasitoid was easily reared, had high rates of parasitism within the laboratory, and became established in a number of release locations that initially led to a substantial decline of P. charybdis populations. However, parasitism of the first generation of P. charybdis by E. nassaui was consistently low [26]. This was likely due to a phenological mismatch between parasitoid and host as the egg parasitoid was active too late in the spring (approximately 30 days following the appearance of P. charybdis eggs) [29].To rectify the phenological mismatch, wasps from a cooler region of Australia were introduced [26,27]. Unfortunately, this agent has since been affected by an invasive obligate hyperparasitoid, Baeoanusia albifuncile Girault, (Hymenoptera: Encyrtidae), which reduced the success of the biological control program [30].
To date, biological control agents alone have not been able to control P. charybdis consistently. Insecticides are the only alternative for control [31,32]. Alpha-cypermethrin, a broadspectrum, synthetic pyrethroid can be used to control P. charybdis via aerial spraying. However, alpha-cypermethrin negatively impacts non-target fauna and thus, the Forest Stewardship Council (FSC), under which numerous Eucalyptus plantations in New Zealand are managed, has restricted the use of these chemicals [33,32]. Research is now aimed at introducing a successful biological agent that is effective in the cooler climates of New Zealand and active during the first generation of P. charybdis [27,34].
A new potential candidate is the solitary larval parasitoid, Eadya paropsidis Huddleston and Short 1978 (Hymenoptera: Braconidae: Euphorinae) [35,36]. This wasp is univoltine and attacks the first generation of paropsine beetles feeding on Eucalyptus in Australia (Fig 1). Eadya paropsidis was described along with E. falcata, as the only two known species in the newly erected, Australian endemic genus [37]. The two species are widely separated geographically, with E. paropsidis known from the Australian Capital Territory, New South Wales, Victoria and Tasmania, and E. falcata known from Western Australia [37][38][39]. The biology of E. falcata is unknown, but E. paropsidis has been reared in the field from Paropsis atomaria Olivier 1807 (synonym P. reticulata) on mainland Australia [37,39] and from Paropsisterna bimaculata (Olivier 1807) [38], Paropsisterna agricola (Chapuis 1987) [35], and P. charybdis in Tasmania (Allen, unpublished data). Although Eadya has been moved to Helconinae based on its placement in a one gene dataset [40], its biology and morphology and biology are consistent with its original placement in Euphorinae [41][42][43], including: attacking exposed chrysomelid beetles; forewing vein 2cu-a absent; forewing vein 3RS curved creating a small marginal cell; and metasomal tergum 1 petiolate [43]. In addition to these characters, species of Eadya can be identified by the presence of an inter-antennal carina and a closed second submarginal cell [43]. Rearing Eadya from field collections from a number of locations in Tasmania revealed two color morphs of the silk used to spin the wasp cocoon (Allen, unpublished data), suggesting the possibility of cryptic species of Eadya. However, due to a ten month obligate pupal diapause when much laboratory mortality happens, this species is frequently difficult to rear to an adult for morphological identification [36]. Hence using molecular phylogenetic approaches combined with host data from field collected paropsine beetle larvae, we set out to determine if E. paropsidis in Tasmania is: (1) one species or a group of cryptic species; (2) host-specific to P. charybdis and closely related Paropsini; and if so, (3) potentially suitable as an agent for biological control of P. charybdis in New Zealand. Wasps were collected from numerous localities across Tasmania over multiple years and reared to determine accurate associations with their paropsine beetle hosts. We utilized three molecular markers and morphology and present one of the most comprehensive datasets to investigate possible cryptic species and host specificity of a prospective classical biological control agent.

Taxon sampling
Eadya wasps and larval beetle hosts were collected from multiple field locations ranging from near sea level to sub-alpine (1000 m) in Tasmania, Australia across six years (2011-2016) from November to January (Fig 2). Specimens were collected by hand, sweep net, or malaise trap in the field. Wasps were reared to adulthood (n = 28), collected on the wing (n = 63) or dissected as larvae or pupae (n = 97) from collected paropsine beetle larvae (   Additional Eadya specimens were obtained through sentinel larval trials. These trials involved placing laboratory-reared, parasitoid free, 2 nd instar paropsine larvae on E. nitens branches in the field to assess levels of parasitism by species of Eadya. On each E. nitens tree, sentinel larvae were placed on foliage of a branch of approximately 1 cm diameter that was tied down firmly to a stake in the ground to prevent contact with other branches, and hence loss of sentinel larvae to neighboring branches. The stake and the branch leading to the main stem were smothered in Tanglefoot™ (The Scotts Company, Ohio, USA) to reduce predation and larval wandering. Branch foliage was then clipped back to approximately 0.33 m 2 . All insects and spiders that were located on that foliage were carefully removed. When confident that the foliage was free of arthropods, laboratory-reared beetle larvae were released onto each branch. Larvae were left for 72 hours before those remaining were carefully removed from each branch, into separate plastic aerated containers, one for each replicate, and returned to the laboratory for rearing to pupation or wasp emergence within a Contherm TM chamber set at 20 ± 1˚C and 16:8 L:D cycle. Emerged parasitoids were preserved in ethanol for molecular analysis.

CO1 Cytb 28S COI Cytb
Voucher numbers are referenced in Genbank Accession Numbers. Stage: A = Adult, L = Larva. Sex is only known for adult wasps. For host, P. = Paropsis, Pst. = Paropsisterna; Pst. variicollis Ã refers to the Pst. variicollis complex whose taxonomic status across southern Australia is unresolved and hence was undeterminable to an exact phylospecies. If the wasp was reared from a beetle that was extracted for DNA, the beetle voucher number is listed. Successful amplification of genes is listed with an "x" under the appropriate gene. For species clades and phylospecies, refer to the phylogenetic analyses (see results). https://doi.org/10.1371/journal.pone.0201276.t001 Taxonomy and biocontrol unite for control of invasive paropsine pests of Eucalyptus Three different paropsine beetles were reared in the laboratory for the sentinel trials: P. charybdis, Pst. agricola, and Pst. selmani. Paropsis charybdis larvae were obtained from colonies initiated each season from adults collected from Hobart, Tasmania off Eucalyptus ovata and E. viminalis. Pairs were maintained in cages at the University of Tasmania with E. viminalis branches at room temperature. Larvae of Pst. agricola and Pst. selmani were obtained as eggs laid on juvenile foliage of E. nitens from Moina (41˚32'27"S 146˚04'38"E), Northern Tasmania and maintained in the laboratory on cut juvenile leaves of E. nitens. A preliminary sentinel trial was conducted in an E. nitens plantation in Moina in December 2011 to establish appropriate methodology. For each replicate (tree), 25 larvae were placed per branch, with 6 replicates of Pst. agricola and Pst. selmani, and 4 replicates of P. charybdis. The sentinel trials were repeated between the 5 th and 18 th of December 2012 using just P. charybdis (n = 767) and Pst. agricola (n = 394) with higher numbers of larvae per tree (either 50 or 100) at the following sites: Ellendale (42˚38'07.24"S 146˚45'04.24"E) (4 replicates per species), Moina (3 replicates per species), Runnymede (42˚38'08.9"S 147˚33'57.9"E) (3 replicates of P. charybdis), Mount Nelson (45˚55' 42"S 147˚18'25"E) (4 replicates of P. charybdis), and The Lea (45˚56'43"S 1471 8'50"E) (2 replicates of P. charybdis), with the latter two sites being native vegetation rather than plantation sites.
Wherever possible, since paropsine beetles typically lay eggs in batches, reared Eadya specimens for molecular determination were taken from differing host larval groupings to maximize the chance that each Eadya were from different mothers. Beetle hosts included eight species from two different genera (Paropsis (abbreviated P.) and Paropsisterna (abbreviated as Pst.): Pst. agricola, Pst. bimaculata, Pst. nobilitata (Erichson 1842), Pst. selmani (only recovered from sentinel trials), Pst. variicollis (Chapuis 1877), P. aegrota elliotti Selman, 1983, P. charybdis, and P. tasmanica (Tables 1 and 2). The taxonomic status of Pst. variicollis is not clear, particularly with respect to two other names in use, Pst. obovata (Chapuis, 1877) and Pst. cloelia (Stål, 1860) (Chris Reid, Australian Museum, personal communication). This binomial could be valid or it may be a synonym of Pst. cloelia, and thus we refer to this taxon as Pst. variicollis Ã for the remainder of the paper to prevent further confusion. Further, an urgent revision is needed due to the recent invasion of New Zealand of Pst. variicollis Ã . Adult beetle voucher specimens were also sampled to have an accurately identified reference library to compare with DNA extracted from putatively identified beetle larvae (Table 2). This is important for field collected hosts, as larval paropsine beetle identifications can be challenging. Finally, several specimens of P. charybdis, Pst. variicollis Ã and one specimen of Pst. beata (Newman 1842) collected from New Zealand were also sampled ( Table 2). All wasp and beetle voucher specimens are maintained at the University of Central Florida Collection of Arthropods or the Australian National Insect Collection (Tables 1 and 2 and S1 Table).

Genetic sampling
A total of 188 wasps and 94 beetles were extracted for DNA and molecular analysis. Three gene regions were amplified, including two mitochondrial genes (Cytochrome oxidase I [COI] and Cytochrome b [Cytb]) and one nuclear gene (28S rDNA regions D1-D3 [28S]). COI has long been the standard for species delimitation in insects [45,16,46,18,47] including Braconidae [48,49,17,19]. Cytb is generally more conserved but can help provide an independent test to prevent overestimations of species [50]. Additionally, 28S has several variable regions (i.e., regions of ambiguous alignment) [51,52] that could potentially provide useful characters for species delimitation and thus was selected for amplification.
DNA was extracted and genes amplified from wasps (Table 1) and a subset of their beetle hosts ( Table 2). Genomic DNA extraction of the wasps and beetle hosts was done using the  DNEasy Mini Kit (Qiagen). The metasoma was separated from the adult and dissected pupal wasps to increase DNA concentration and ensure a voucher specimen was available postextraction for morphological examination. Larval wasps that had emerged from their host (prior to pupation) were ground with a sterilized pestle prior to extraction. Similarly, the associated beetle larvae from which the wasp emerged was also pulverized prior to extraction to ensure adequate DNA recovery as most beetle hosts were in poor condition after parasitization. DNA was also extracted from beetles collected as adults and vouchers retained and only COI was amplified as a tool to provide barcode confirmations on larval identifications. All PCR reactions were performed using 0.

Phylogenetic analyses
Gene alignments were completed for COI and Cytb by hand using the reading frame as a guide with Bioedit v.7.1.3 [54]. There were no indels present in either gene and thus the alignments were unambiguous. A modified [52] secondary structure model [51] was used to align 28S.

Regions of ambiguous alignment (RAAs), and regions of expansion and contraction (RECs)
were not excluded from the data set as we assumed most informative characters for this gene would be contained within these regions that can be hyper variable across genera. For each gene, the best fitting model of DNA sequence evolution for nucleotide analyses were determined using jModelTest v.0.1.1 [55] under the Bayesian Information Criterion (BIC). The model with the lowest calculated BIC score was considered the best-fitting model for each gene. Depending on the gene, either one or two species from other braconid subfamilies were used as outgroups Afrocampsis sp. (Acampsohelconinae) and Eumacrocentrus americanus (Cresson 1873) (Helconinae) to ensure the ingroup was monophyletic. For individual and concatenated data sets, Bayesian inference with two independent runs each with four chains and default priors was run in MrBayes v.3.2.6 [56]. An independent molecular model was applied to each partition in the concatenated data set (partitioned by gene) and different parameters of the model were unlinked to allow each partition to have its own set of estimations for parameters. The rate parameter was set to vary across different All analyses were performed for 5,000,000 generations sampling every 1000 th generation and the results from the two independent runs were summarized in a majority rule consensus tree after discarding the initial 25% of the trees for burn-in. Stationarity and appropriate mixing of the two independent runs were determined when the average standard deviation of split frequencies approached 0.01, the Potential Scale Reduction Factor (PSRF) for each parameter of the model was close to 1, and the overlay plots of both runs showed the number of generations versus the log probability of the data were heteroscedastic. Average intraclade and interclade genetic distances were calculated using Kimura's two-parameter model [57] using MEGA v.7.0.14 [58].

Morphological examination
We examined all adult specimens of Eadya collected for the purposes of this study (Table 1, adults), plus additional specimens collected in malaise traps, the paratype of E. paropsidis, the holotype and paratype of E. falcata, and some museum specimens (see additional material examined in S1 Table). We initially sorted specimens into morphotypes and observed four distinct morphospecies. One perfectly matched the description and paratype of E. paropsidis. Based on examination of all material, Eadya sp. 1 and sp. 2 were morphologically distinct with observable differences in several morphological characters from E. paropsidis or E. falcata. For example, Eadya sp. 1 and 2 do not possess a transverse carinae on the propodeum as in E. paropsidis, and have more impressed notauli than E. falcata. Further, Eadya sp. 1 lack median tubercles on the clypeus. However, the fourth morphospecies was very similar to E. paropsidis but was distinctly smaller in size. Parasitoids can vary in size due to the size of their host and nutritional factors during larval development. However, in their original description of Eadya, Huddleston and Short (37) noted variation across E. paropsidis, particularly a series of eight specimens that were smaller in size and had a less concave occiput. Although they chose not to describe these variants as a new species, we chose to separate the smaller specimens (as Eadya sp. 3) to test whether or not it was indeed a distinct species. Although we discuss the molecular results in context with the morphological examinations of morphospecies, descriptions of the new species are fully described in a separate paper as Eadya annleckiae Ridenbaugh 2018 (sp. 1); Eadya spitzer Ridenbaugh 2018 (sp. 2), and Eadya daenerys Ridenbaugh 2018 (sp. 3) [43].

Parasitism rates of paropsine beetles
A total of 2924 field collected paropsine beetle larvae across 10 beetle species, comprising over 135 independent collections (groups of larvae from same egg batch) were reared over six years (Table 3). Four beetle species had substantially higher (>18%) parasitism rates: P. tasmanica, P. variicollis Ã , P. charybdis and Pst. agricola, whereas no Eadya were reared from three beetle species. For the sentinel trials, the number of larvae recovered (n = 616) and the percent parasitized by Eadya or unidentified Tachinidae is presented in Table 4 for the preliminary trial at Moina in 2011 and the more substantial trials at five locations in 2012. The lack of parasitism of Pst. agricola by Eadya at Moina in 2011 was unexpected, but the timing of the trial was very late in the flight season for the species of Eadya that parasitizes this host. In the 2012 trials, species of Eadya parasitized beetle larvae at four of the five sites (Table 4). At two of the plantation sites in 2012, there were high levels of parasitism by tachinid flies.

Species delimitation in Eadya
For COI, a total of 672 characters and 177 taxa were included in the analysis, including one outgroup. Four distinct clades were recovered and well supported (pp = 1.0) (Figs 3A and S2) and assigned as putative species based on a phylogenetic species concept (monophyly with high support) [59,60] and a distinct barcoding gap (greater interclade distance than intraclade distance) [61]. Of the delineated morphospecies, Eadya sp. 1 corresponded to Clade A, Eadya sp. 2 to Clade B, E. paropsidis to Clade C and Eadya sp. 3 to Clade D (Fig 3A). Clade D had some interclade structure, labeled Clade D1 and D2 (S2 Fig). We consider both of these clades to correspond to Eadya sp. 3 because: (1) there are no amino acid differences between sequences of the members of these two clades; (2) the genetic distance between the two clades was only 1.1%; and (3) the two clades were not well supported. Thus, our four delineated morphospecies correspond perfectly to the four phylogenetic species delineated with COI. The average interclade genetic distances between all putative phylospecies of Eadya (Clades A-D) ranged from 8.7% to 31.2% (Table 5A), well above typical DNA barcoding thresholds (~2-3%) for species delimitation [62,14], including in Braconidae [49,19]. There was also very low average intraclade variation, with most clades exhibiting less than 1% genetic distance across all taxa, even though specimens within clades were sampled from different hosts, localities, and across different years (Table 1 and Table 5A). Since paropsine beetles lay eggs in batches, number of independent collections refers to the number of differing larval groupings collected to maximize the chance of finding Eadya parasitism. Ã Indicates a minimum as some independent collections were pooled.
https://doi.org/10.1371/journal.pone.0201276.t003 For Cytb, a total of 429 characters and 173 taxa were included in the analysis, including outgroups. The same four major clades corresponding to the four hypothesized morphospecies were recovered (Fig 3B and S3 Fig) and were well supported (pp ! 0.97). Unfortunately, only one taxon from Clade B was amplified for this gene (BJS553 -dissected larval specimen), but this taxon (Eadya sp. 2) was still recovered as sister to E. paropsidis. For 28S, a total of 893 characters and 162 taxa were included in the analysis and only two well supported clades (pp = 1) were recovered (Fig 3C and S4 Fig). Clade A (Eadya sp. 1) was congruent across all genes but all other taxa from COI and Cytb were recovered in a single clade. As this large Clade contains Eadya sp. 2, E. paropsidis, and Eadya sp. 3 (Clades B, C, and D, respectively), 28S appears to be too conserved for species delimitation in this group. There were limited substitutions across identified morphospecies within the large clade, even within hypervariable regions (RECs, RAAs, and RSCs) that may vary across closely related species [52]. A concatenated dataset was Taxonomy and biocontrol unite for control of invasive paropsine pests of Eucalyptus also analyzed with all three genes. The same major clades recovered across COI and Cytb were also recovered here albeit some with less support, but again supporting four distinct species of Eadya, including E. paropsidis (Fig 4). All taxa are clearly identifiable by morphology, although E. paropsidis and Eadya sp. 3 are very similar morphologically, with E. paropsidis having a more concave occiput, an emarginate occipital carina, and being larger in size.

Beetle species identification
COI was amplified from beetle remains regardless if the wasp was reared or dissected from the host. Due to degradation of host material, DNA extraction was successful for only 48 parasitized beetles: 35 Pst. agricola, one Pst. bimaculata, one P. aegrota elliotti, seven P. charybdis, and six specimens identified as Pst. variicollis Ã ( Table 2). All putatively identified larval material was recovered in well supported monophyletic clades with the correct adult reference voucher (Fig 5), indicating that larval identifications were accurate. All Pst. variicollis Ã samples were recovered in a strongly supported monophyletic clade. However, there was clade structure with respect to location, such that all Pst. variicollis Ã (from Tasmania) were recovered in a well-supported subclade, indicating distinct differences between samples from New Zealand and Tasmania. There was an average 1.6% genetic distance between the Tasmanian and New Zealand Pst. variicollis Ã (1.6%, Table 5B). Thus, these samples are either from the same species and the different clades are representative of population level differences between New Zealand and Tasmania samples, or they may represent different, but very closely related species. Regardless, the results highlight the need for a revision of Pst. variicollis Ã , which is particularly important as this species was discovered in New Zealand in 2016, representing another potential serious pest to the forest industry [63]. Samples of P. charybdis from New Zealand were recovered with samples from Tasmania, Australia in a well-supported clade, demonstrating no distinct population level differences between beetles from the two different countries. The average genetic distance among all P. charybdis was low at 0.3% (Table 5B).

Eadya paropsidis is a complex of species
Based on morphological examination, molecular data from three genes, and host-association data, Eadya paropsidis is not a single species, but rather a complex of species. Two of these species are cryptic, with limited morphological characters separating them: E. paropsidis and Eadya sp. 3. Interestingly, these two taxa were suspected to be different species in the original description of Eadya by Huddleston and Short (37). They state, "there is a series of eight specimens in ANIC [Australian National Insect Collection] which agree well with E. paropsidis except that the occiput is less concave, the propodeum is less abruptly divided and the insect smaller. More material is needed to decide if these specimens are succinctly distinct to be described as a new species or merely variants of E. paropsidis (p. 319)." Our morphological examinations along with molecular analyses confirm that the smaller variant is indeed a new species (Eadya sp. 3), and corresponds to Clade D in COI, Cytb, and the concatenated analysis (Figs 3A, 3B and 5). Although COI and Cytb were congruent, the D2-D3 region of 28S was a poor marker for species delimitation, as only two species were delimited from the 28S phylogeny ( Fig 3C). Substitutions in the 28S regions of ambiguity, where high rates of nucleotide variation are typically found [51] were minimal, ranging from no variation to a few single nucleotide polymorphisms. Genetic distances between species of Eadya for COI were high, ranging from 8.7 to 31.2%. In particular, Eadya sp. 1 had numerous genetic (over 30%) and morphological differences when compared to other species. Pupal cocoon color varied within species with Eadya sp. 1 (mostly white), Eadya sp. 3 (mostly brown) and Eadya paropsidis (mostly white) and was not therefore a reliable aid to species identification. Descriptions of all new species and a key to all species of Eadya can be found in Ridenbaugh et al. [43].

Eadya host plasticity
A list of all known host records for all four Eadya species is listed in Table 6. All species of Eadya can utilize multiple hosts, although some wasps have stronger associations with specific taxa. Two host records were only from sentinels, E. paropsidis from P. charybdis and Eadya sp. 1 from Pst. selmani (Table 6). Eadya sp. 2 was rarely found and not reared successfully to adulthood in the laboratory. All species of Eadya were reared from the target pest, P. charybdis. Eadya paropsidis was largely specific to P. tasmanica in Tasmania. However, from the original description [37], E. paropsidis was reared from P. atomaria in mainland Australia, and several subsequent studies list additional hosts for this species [38,39]. Considering our findings and by examining morphology of specimens, all Pst. agricola host records are actually Eadya sp. 3 and not E. paropsidis. This may also be the case for those records for Pst. bimaculata, though there are no specimens from these earlier records to confirm this. Eadya sp. 1 and 3 were recovered from multiple hosts. However, Eadya sp. 1 was most commonly associated with Pst. variicollis Ã or Pst. selmani and never from Pst. agricola or Pst. Posterior probabilities for major clades are listed near the relevant nodes. Trees connect at the arrows. Clades and corresponding putative species are labeled. Taxon names include voucher numbers, stage of wasp, beetle host name from which the wasps were reared, locality collected, and year of collection, as listed in Table 1. Hosts listed as Pst. variicollis Ã indicate this host is part of a complex of unresolved taxonomic status across southern Australia. The boxed inset has major clades collapsed based on a phylogenetic species concept for ease of viewing relevant clades. Scale bars refer to number of substitutions for tree branches.
https://doi.org/10.1371/journal.pone.0201276.g004 bimaculata, while Eadya sp. 3 was never reared from Pst. variicollis Ã or Pst. selmani, demonstrating strong species level differences in host usage despite some cross over in host taxa. Eadya sp. 3 almost exclusively used either Pst. agricola or P. charybdis, although Pst. bimaculata and Pst. nobilitata were rare hosts (Fig 5). Although the majority of Eadya sp. 3 collected were from Pst. agricola hosts, this reflects the relative abundance or availability of this host in our chosen field sites. Pst. agricola is far more abundant in the plantation locations sampled, whereas P. charybdis is rare and hard to collect at any location. Reasons for the relative rarity of P. charybdis in Tasmania are unknown, but we cannot rule out that this species suffers under high natural enemy loadings in Tasmania; P. charybdis is known to be a host to three species of phoretic mite in Tasmania [64] in addition to egg and larval parasitoids and ladybird predators. Practical difficulties in sampling P. charybdis also arise due to both adult and larval feeding preferences for flush adult foliage high in the crown (rather than waxy juvenile leaves) of Eucalypts in the subgenus Symphyomyrtus. First instar larvae of P. charybdis tend to scatter and feed singly on outermost branches often high in the crown; whereas, Pst. agricola feed gregariously on the waxy juvenile foliage within easier reach for sampling. Thus, our sampling may have been influenced by the biology of the beetles.
Host-plasticity is likely beneficial for reproductive success of the wasp. The ability to utilize multiple hosts increases the likelihood of successful parasitism due to the greater availability of resources across habitats [65]. This in turn decreases energy expenditure associated with host seeking. Although beneficial to the wasp, host-plasticity does have some implications for the suitability of these species as classical biological control agents.

Implications for biological control of P. charybdis, Pst. variicollis Ã , and other invasive paropsines
Although species of Eadya display host plasticity, they appear to be restricted to Paropsine beetles (Chrysomelinae) in two closely related and recently revised [66] genera. These beetles are similar across several biological features, including an overlap of spatial range, similar larval  Tasmania (NZL added if collected in New Zealand), and year of collection, as listed in Table 2. Clades are labeled with the identified beetle species based on the placement of the adult reference voucher specimens. Scale bar refers to number of substitutions for tree branches.
https://doi.org/10.1371/journal.pone.0201276.g005 phenology (temporal overlap), and related host plants (externally feeding on foliage of Eucalyptus species [67][68][69]44]. Additionally, relationships within paropsine beetles are closely linked to host plant usage on eucalypts [70]. Thus, species of Eadya are restricted to parasitizing a set of very closely related beetles, both phylogenetically and biologically, despite the ability to successfully parasitize multiple species. There are no native Eucalyptus in New Zealand and all paropsine beetles are invasive pests in that country [34]. Another new paropsine incursion was discovered in 2016, The eucalyptus variegated beetle (Pst. variicollis Ã ), which has further increased interest in species of Eadya as potential biological control agents [63]. Although there are no records of Eadya on beetles in any other genera, host specificity testing has not yet been completed. However, if Eadya is found to be host specific to beetles within these two genera, as expected from our results, then P. charybdis makes an excellent candidate for classical biological control. Withers, Allen [34] already selected a list of candidate species to test Eadya for potential non-target effects based on rigorous biological and phylogenetic criteria of native beetles and beneficial weed biological control beetles present in New Zealand.
Based on our data, all recent research [34][35][36]71] within this system has been conducted on Eadya sp. 3, as opposed to E. paropsidis. This is a promising as Eadya sp. 3 had the most records of parasitism from P. charybdis, relative to other species. Although there were more records from Pst. agricola, our sampling biases may have influenced part of this result. As Pst. agricola is not in New Zealand, there would be no additional resources for Eadya sp. 3 to utilize if released in that country, which should promote a successful biological control program. Thus, Eadya sp. 3 is the best candidate for importation for control of P. charybdis. This species was commonly collected across most localities, particularly on the wing, demonstrating a wide geographic range for this species.
Eadya sp. 1 could be a suitable candidate for classical biological control of the newly invaded Pst. variicollis Ã in New Zealand. As Eadya sp. 1 can attack both P. charybdis and Pst. variicollis Ã , which could provide an added benefit in the control of both pest species. However, negative impacts due to host competition on P. charybdis would need to be investigated if both Eadya sp. 1 and 3 were to be released. Results from this study indicate a careful population/species level study of the Pst. variicollis complex is necessary to determine the limits of this species. Eadya sp. 1 was also recorded from Pst. selmani, a Tasmanian paropsine that invaded Ireland in 2007 [72] and is a significant pest of Eucalyptus (plantations and cut foliage trade). Now that the Eadya species complex has been delimited, the next stage for any of these biological control programs will need to be thorough host specificity testing of the most appropriate Eadya species. In the case of P. charybdis biological control, as was the focus of this study, research will investigate Eadya sp. 3 against less closely related non-target beetles present in New Zealand [34].

Conclusions
For a successful biological control program the biological agent must be correctly identified, particularly in the context of potentially cryptic species complexes. This is essential to ensure an adequate assessment of the biological agent of choice as the host range, biological features, behavior, and potential for control may vary between species within these complexes [e.g. 7,34]. Prior to this study, it was assumed E. paropsidis was a single species due to limited taxonomic study on Eadya. However, based on our molecular and morphologic data, we now know E. paropsidis is not just one species, but a complex of species attacking Eucalyptus-feeding paropsine beetles in Tasmania. This research has important implications for the forest industry as species of Eucalyptus have been imported to numerous countries around the world for their pulp and fiber, and ornamental and oil producing properties. Eadya sp. 3 (formally called Eadya daenerys Ridenbaugh 2018) [43] is the most suitable candidate for release in New Zealand to control the eucalyptus tortoise beetle, P. charybdis. Eadya sp. 1 (formally called Eadya annleckiae Ridenbaugh 2018) [43] should be examined in future research for potential to control Pst. variicollis Ã in New Zealand and Pst. selmani in Ireland. However, a comprehensive molecular and morphological review of the taxonomic status of the Pst. variicollis Ã complex is needed. This study represents one of the most comprehensive biological control studies to delimit cryptic species and resolve host relationships through mass rearing, and analysis of morphological and molecular data in relation to hosts of a potential parasitoid biological control agent. It also represents a very successful case of biological control researchers collaborating with taxonomists early in the research pipeline, which is the best way to prevent unintended effects of natural enemy introductions to control pests. Finally, this study also provides the necessary data to create a model system to test theories on biological control and multitrophic community dynamics in invasion biology with respect to paropsine pests, their host Eucalyptus plants, and the suite of primary parasitoids that may regulate their populations. Posterior probabilities are listed near the relevant nodes for major clades. Clades and corresponding putative species are labeled. Taxon names include voucher numbers, stage of wasp, beetle host name from which the wasps were reared, locality collected, and year of collection, as listed in Table 1. Scale bar refers to number of substitutions for tree branches. (PDF) S4 Fig. Bayesian analysis of 28S without clades collapsed. Posterior probabilities are listed near the relevant nodes for major clades. Clades and corresponding putative species are labeled. Taxon names include voucher numbers, stage of wasp, beetle host name from which the wasps were reared, locality collected, and year of collection, as listed in Table 1. Scale bar refers to number of substitutions for tree branches. (PDF) S1 Table. All material examined for this study, including specimens for morphology and type material. (PDF) S2 Table. Primer sequences used in this study and references for sequences and cycling conditions. (PDF)